Skip to main content

Abstracts from the 2017 Society of General Internal Medicine Annual Meeting

“A PCMH MIND AND A PCMH HEART”: PATIENT, FACULTY, AND LEARNER PERSPECTIVES ON THE DEVELOPMENT OF AN INTERPROFESSIONAL TEAM-BASED TRAINING PRACTICE Daniel J. Coletti 4; Pratiksha Yalakkishettar5; Johanna Martinez4; Lauren Block4; Nancy A. LaVine3; Joseph Conigliaro2; Alice Fornari1. 1Hofstra NSLIJ SOM, Hempstead, NY; 2North Shore LIJ Health System, New Hyde Park, NY; 3Northwell Health, New Hyde Park, NY; 4Northwell Health, Great Neck, NY; 5Hofstra University, Hempstead, NY. (Control ID #2704946)

BACKGROUND: The benefits of interprofessional education (IPE) and clinical training have been documented for learners, yet less is known about the perceptions of the faculty members who facilitate the educational experiences, or the perspectives of patients who receive this interprofessional (IP) care. Aligning the priorities of these three stakeholders would inform a truly patient-centered medical home and also develop an IP workforce skilled and comfortable working in a PCMH delivery model. The objective of this report was to use focus groups to compare stakeholder attitudes about IP education and training.

METHODS: We conducted five groups with 42 participants (31 F, 11 M): A “learner” group, a faculty group and three patient groups. The learner and faculty group represented the disciplines of medicine (students and residents), psychology, pharmacy, and physician’s assistants. One of the patient groups was conducted in Spanish. This project was a formative research activity of IMPACcT (Improving Patient Access, Care, and cost through Training), a HRSA-funded program to expand the primary care workforce through IP education, training, and mentorship. Three raters analyzed group transcripts until common themes across the three groups emerged.

RESULTS: Themes present in all stakeholder groups were labeled 1) team engagement, 2) the role of technology in care delivery, 3) insurance/cost of care, 4) involving patients in the learning process, 5) time constraints, 6) scope of practice, and 7) autonomy, interdependence, and decision-making. Both similarities and distinct perspectives emerged across stakeholders when discussing these issues and most attitudes were positive. Learners, faculty, and patients each emphasized the importance of defining roles within a team and communicating roles to patients. Learners were excited about IP work and anticipated high quality interactions with other professions. Patients noted that participating in a teaching clinic was “more than about me” and described benefits receiving care from supervised trainees. Faculty perspectives, however, were more ambivalent. They noted that IP “is not a new thing” and questioned their ability to integrate PCMH mandates with an authentic mission to provide patient-centered care: “there’s a difference between…checking off all the boxes for PCMH and…really incorporating the spirit of it…like having a PCMH mind versus a PCMH heart.”

CONCLUSIONS: This is the first reported data comparing perceptions about IP education and care across these three stakeholder groups. Commonalities observed across the perspectives of patients, faculty, and learners suggests the need to attend to stakeholder priorities (e.g. regarding scope of practice and role definition) and bridging gaps between teaching PCMH principles and actually providing patient-centered, high quality care. Focus group material has been incorporated into project protocols for interprofessional huddling, communicating with patients, and in the content of our didactic curriculum.

“CONVERSATIONAL ADVICE”: A MIXED-METHODS ANALYSIS OF MEDICAL RESIDENTS’ EXPERIENCES CO-MANAGING PRIMARY CARE PATIENTS WITH BEHAVIORAL HEALTH PROVIDERS Patrick Hemming 1; Rachel Levine3; Joseph J. Gallo2. 1Duke University School of Medicine, Durham, NC; 2Johns Hopkins Bloomberg School of Public Health, Baltimore, MD; 3Johns Hopkins University, Baltimore, MD. (Control ID #2704644)

BACKGROUND: Integrated Behavioral Health Clinicians (IBHC’s) are increasingly common in residency primary care clinics. When IBHCs and residents co-manage patients, residents may learn new approaches to counseling and medical management. This survey sought to better understand the impact that patient co-management with IBHCs has on residents’ learning about behavioral health management.

METHODS: Residents were surveyed from 2 Internal Medicine (IM) and 3 Family Medicine (FM) residency training programs with behavioral health integration in residents’ continuity clinics. To assess the degree of face-to-face interaction during their most recent co-managed case, residents were asked whether or not the co-management included (1) a shared visit with the IBHC and/or (2) meeting face-to-face to discuss the patient’s care. Respondents were asked about other features of the visit, including: (1) whether or not they received feedback from the IBHC on their management, and (2) to rate the episode’s impact on the patient’s care. Associations between the degree of face-to-face interaction and these two outcomes were assessed using multiple logistic regression and linear regression. Residents were asked open-ended questions regarding what they had learned from co-managing with an IBHC. Qualitative responses were coded thematically by two researchers using grounded theory. Associations were examined between face-to-face interactions and the frequency of each major learning theme category.

RESULTS: 113 residents of 117 respondents had experience co-managing a patient and described their most recent experience co-managing a patient with an IBHC (overall response rate 72%, 117/163). Residents were significantly more likely to receive feedback on their patient management if they had a shared visit (Adjusted OR 3.0, 95% CI 1.2–7.6). Residents gave high ratings (mean rating 8.0/10) to the patient impact of their co-management with no association to whether or not they had a shared visit. Fourteen learning sub-themes were reported from the following major themes: BH skills awareness, interpersonal communication skills awareness, and newly-adopted attitudes toward BH. Residents who reported receiving feedback were more likely than those who did not receive feedback to report themes of interpersonal communication skills awareness (yes 26.6% vs. no 9.4%)

CONCLUSIONS: In residency clinic settings with BHI, residents have the opportunity for regular face-to-face co-management with IBHCs. Face-to-face co-management can facilitate increased feedback on skills and management and positively influence residents’ attitudes and perceived skills.

Selected Quotes from Each Major Theme Organized by Degree of Co-management and Receipt of Feedback (Yes or No)

  Theme 1 BH Skills Awareness Theme 2 Interpersonal Communication Skills Awareness  
No shared appointment or face-to-face meeting “I learned about the severity of my patient’s Post-traumatic Stress Disorder, which he had only mentioned to me briefly, and I got more information about his coping resources.” N No residents in this group reported this major theme “With BHI, patients have additional support and tend to be more compliant.” N
Met face-to-face, no shared appointment “I learned about ethical issues, physician obligations, and reporting requirements to law-enforcement.” Y “I learned about community resources available for IV drug addicts.” N “I learned to have a stronger communication with my patient’s family.” N “I learned different ways of addressing medication problems with the patient and how to approach her about her chronic pain.” Y “I developed more understanding of the impact of social stressors on health” N “The alternate perspective on disease often provides the key to treatment adherence” N
Shared appointment, no outside face-to-face meeting “I learned about quick interventions for my patient in a single visit.” Y “I learned different ways to presents medical information in ways that patients understand.” Y “It is helpful to have a neutral party and to coordinate the message that the patient receives.” Y
Shared appointment and face-to-face meeting “I learned how to address some of the most common issues--opioid dependence and depression.” Y “I learned various weight loss support groups and strategies for patients” Y “I learned to ask what other concerns patient has- if there is a part that is missing from what the patient is telling me- that can help for therapeutic relationship.” Y “The BHC gave conversational advice in steering the interview.” Y “I can delegate this aspect of care to the IBHC to free up time for me to deal with this patient’s complex medical conditions” N “I learned that behavioral health issues are complex and require special management considerations.” Y


BACKGROUND: Medical school and residency training programs rely on skilled clinician educators to provide high quality educational experiences. While faculty development programs are common, as of 2012, formal degree-granting programs in medical education existed at only ten institutions in the United Stated. To date, there have been limited outcomes reported for participants of degree-granting programs. Beginning in 2002, the University of Pittsburgh’s Institute of Clinical Research Education created both masters and certificate level degree-granting programs in medical education, which now have more than 10 years of graduates. Courses include those focused on teaching skills, medical education research skills, and faculty development. We sought to evaluate the program by surveying its participants with regards to their attitudes, self-reported skills, and career outcomes.

METHODS: All graduates of the program between 2004 and 2014 received an email invitation to complete an anonymous electronic survey regarding their satisfaction with the program as well as their perception of whether the master’s program adequately prepared them in domains pertinent to medical educators. Participants were also asked to upload their current CV from which data about educational leadership positions, curriculum development and national dissemination of education-related work was abstracted.

RESULTS: Out of 60 graduates, 47 completed the survey (78%) and out of those 45 uploaded their CV for analysis (75%). More than 90% of respondents agreed that due to completion of the program they were competent in applying principles of learning theory, clinical teaching, small group teaching, ability to give lectures, providing feedback to learners, curriculum development and evaluation, as well as conducting and evaluating educational research. 94% of respondents believed that they were a more effective educator than peers who did not complete a degree. CV abstraction revealed that 98% of respondents hold academic positions. Respondents represent more than 15 different medical specialties, though the majority (40%) are general internists. Of graduates surveyed, 76% held educational leadership positions. 93% published in peer reviewed journals, 67% published on an educationally related topic, 87% participated in curriculum development, and 67% engaged in mentorship. 13 respondents won teaching awards at their respective institutions.

CONCLUSIONS: Because degree-granting programs in medical education require great resource investment, the outcomes of such programs are relevant for institutional support and sustainability. Graduates of the degree granting programs at the University of Pittsburgh reported, because of their training, competence in several key domains crucial to success as a clinician educator. Abstraction of CV’s noted almost all hold academic positions in their representative specialties as well as document their prolific nature in several domains essential to academic success as a clinician-educator.

“GETTING LOST” IN HOSPITAL IS A SOURCE OF STRESS AMONG PHYSICIANS AND NON-PHYSICIAN HEALTHCARE PROVIDERS Grigorios G. Anagnostopoulos3; Michel Deriaz3; Jean-Michel T. Gaspoz1; Dimitri Konstantas3; Idris Guessous 1, 2. 1Geneva University Hospitals, Geneva 14, Switzerland; 2University of Lausanne, Lasuanne, Switzerland; 3University of Geneva, Geneva, Switzerland. (Control ID #2704923)

BACKGROUND: Navigating around large hospitals has been shown to be a stressful and time-consuming experience for all users of the hospital, including staff members. In addition, navigation difficulties in a hospital highlight suboptimal organisation. When staff encounters navigation difficulties this can lead to cost and efficiency issues and potentially put patient safety at risk. Despite the provision of an array of in-hospital navigational aids, ‘getting lost’ continues to be an everyday problem in these large complex environments. Within the framework of developing a navigation mobile app for the largest university hospital of Switzerland (Geneva University Hospitals), we aimed to identify the navigational needs of, problems encountered by, and consequences to primary care staff.

METHODS: A questionnaire was developed by a primary care physician and an indoor navigation specialist (both from Geneva University) based on factors identified through a review of the literature. The questionnaire was constructed to reflect all professional activities encountered in the primary care division and was sent in 2016 to both physicians and non-physician healthcare providers. The questionnaire includes the identification of current problems in way finding inside the hospital (stationary and ambulatory) and their impacts on the staff’s work.

RESULTS: Out of 169 eligible collaborators, 111 (65.7%) answered the questionnaire, 61.3% were completed by physicians, 56.7% of collaborators had >5 years of work at Geneva University Hospitals, 70.5 were female, mean age was 39.5 years (SD 10.3). The majority (52.3%) of the participants answered that they had faced difficulties in finding their destination in the hospital (36.9% occasionally; 14.4% often, 0.9% all the time). Only 9.9% answered “never” and 37.8% “rarely”. On average, participants estimated that they spent 11.7 min per week on searching for their destination or answering questions of others trying to reach their destinations. About 70% of the participants reported that the difficulty of finding one’s way in the hospital could be a source of stress for staff members. The participants expressed a very positive view over the prospect of the creation of an application for mobile phones that would guide staff to their destination in the hospital.

CONCLUSIONS: These results show that even for staff members of a large hospital, finding their destination is difficult, consume time, and might be a source of stress. Improving way finding could contribute to pursuing organizational change, improve efficiency, and decrease stress.

“IT WAS A LOT OF DIFFERENT THINGS BUT BASICALLY WE WERE BEING EVICTED.” PRECURSORS OF HOMELESSNESS AMONG ADULTS AGED 50 AND OLDER: FINDINGS FROM THE HOPE HOME STUDY Irene Yen; Pamela Olsen; Angela Allen; John Weeks; Kelly R. Knight; Margot Kushel. University of California, San Francisco, San Francisco, CA. (Control ID #2705865)

BACKGROUND: The median age of single homeless adults is approximately 50; little is known about older adults’ pathways to homelessness. Among older homeless people, 44% first experienced homelessness after age 50. We conducted in-depth interviews with older homeless adults about life course experiences and precipitants of homelessness, and examined whether these differed by age at first homeless.

METHODS: We recruited 24 participants from the HOPE HOME cohort, a population-based study of homeless adults 50 and older; we sampled purposively so that 1/2 experienced homelessness before age 50. Our interview focused on childhood circumstances, family and social relationships, criminal and victimization history, social services, criminal justice, education and employment, substance use, and precipitants of most recent homelessness. We developed a codebook using open coding. Two researchers coded transcripts independently, and then met to discuss and reach consensus. The research team identified key themes using thematic analysis.

RESULTS: Participants were 52–64 years old, 71% were African American. We identified five key themes: 1) adverse childhood experiences (ACE) had negative consequences throughout the life course; 2) substance use disorders derailed efforts at educational attainment, partnership, employment, and housing; 3) inadequate social ties with partners and family members limited housing options; 4) legal entanglements threatened employment, housing, and entitlement access; 5) institutional and structural racism impeded participants’ access to services and criminalized behaviors. Those with early onset homelessness reported a higher severity of ACE and earlier onset of mental health and substance use disorders which interfered with educational and occupational attainment and familial attachments. They were unable to identify discrete precipitants of homelessness. Those with late onset homelessness reported lower severity of ACE and behavioral health conditions. They reported discrete homelessness precipitants, including: catastrophic illness, marital dissolution, job loss, and eviction for non-financial reasons. Men with late onset homelessness reported feelings of shame and loss of masculinity, hampering their ability to seek assistance. African-American participants reported multiple experiences of racism which heightened their vulnerability to homelessness.

CONCLUSIONS: Older homeless adults with early and late onset homelessness shared risk factors for homelessness that differed in severity and timing. Preventing homelessness among those with early onset homelessness requires early interventions for those with ACE and early onset behavioral health conditions. Efforts to prevent late-onset homelessness should focus on interpersonal (e.g. marital dissolution) and structural (e.g. eviction) high risk periods. Homelessness prevention efforts should incorporate structural responses to racism.

“LET ME FINISH THIS NOTE AND THEN WE CAN TALK ABOUT THAT”: TRANSITIONS BETWEEN MULTITASKING AND SILENT ELECTRONIC HEALTH RECORD USE IN SAFETY NET VISITS Neda Ratanawongsa 1, 2; George Matta1, 2; Courtney R. Lyles2; Kaylin Yu3; Jennifer Barton5, 6; Christopher Koenig4; Ed Yelin1, 1; Dean Schillinger1, 2. 1UCSF, San Francisco, CA; 2University of California San Francisco, San Francisco, CA; 3Cornell University, Ithaca, NY; 4San Francisco State University, San Francisco, CA; 5Oregon Health & Science University, Portland, OR; 6VA Portland Health Care System, Portland, OR. (Control ID #2698401)

BACKGROUND: Electronic health record (EHR) use affects communication, but the impact may depend on EHR use styles. We explored how clinicians transition between multitasking and silent EHR use during safety net visits and the benefits & risks of these dynamic EHR styles.

METHODS: We conducted an observational study ≥2 months after certified EHR implementation in 5 academic public hospital clinics. We video-recorded encounters between English- and Spanish-speaking patients with chronic conditions and their primary and/or specialty care clinicians. Using Atlas.ti, we recorded durations of: multitasking EHR use (concurrent with patient talk or clinician talk); silent EHR use (clinician AND patient silent for ≥5 s); non-EHR tasks; and focused clinician-patient talk. We calculated median proportions of the total visit spent in these segments. Two investigators independently analyzed 3 videos and generated codes (editing organizing style), negotiating discrepancies to create a coding template. One investigator independently applied the template to the remaining videos. By consensus, we combined codes into themes and selected representative quotes.

RESULTS: We recorded 35 visits between 25 patients & 25 clinicians. Patients averaged 57 years in age; 44% were women; 48% were Hispanic/Latino; 28% reported < high school graduation; and 20% had limited health literacy (LHL). Among clinicians, 72% were women; 48% were primary care providers; 88% physicians; and they averaged 16 years since earning degrees. The median visit length was 25.5 min; 17% visits were in Spanish. The proportion of visit time with clinician EHR use was: 0% in 3 visits, ≤ 25% in 9 visits, >25–50% in 17 visits, and >50% of visit time in 6 visits. Multitasking EHR use occurred in 91% of visits, and silent EHR use in 60% of visits. A median of 28% of the visit was spent in multitasking EHR use (range 0–73%); 3% in silent EHR use (range 0–19%); 3.4% in non-EHR tasks (range 0–59%); and 31% in focused clinician-patient talk (range 2–81%). Table 1 shows the three core themes: 1) varying clarity in signaling the transition to silent EHR use; 2) breaking the silence; and 3) unaddressed patient concerns.

CONCLUSIONS: Most safety net encounters involved both multitasking EHR use and transitions in & out of silent EHR use. Clinicians may be unaware these transitions are occurring. Some patients use the silence to recall or reinsert their concerns; this silence could be particularly beneficial to LHL patients or introverts who need space to process and voice their thoughts. However, both EHR use styles pose risks to missing patient concerns. Future research should explore how different EHR-use styles affect clinician cognitive workload - including outcomes related to safety and clinical decision-making - as well as patient engagement and satisfaction.

Themes Related to EHR Use Styles

Varying clarity in signaling the transition to silent EHR use Some clinicians overtly voiced a need to pause to use the EHR, Most clinicians gave non-verbal signals (e.g., turning body or gaze) without verbal warnings.
Some clinicians drifted into silence without signaling awareness they were halting conversation with patients.
• “I’m going to look into the computer for a second, okay?”
• “Give me a minute, I want to review in the computer what we’ve done before.”
• “Let me just…oops…hold on one second.”
• After period of multitasking history- taking: “So…umm…we…” & drifts into 17 sec silence while typing.
• “There aren’t specific treatments…but they’re going to…uh…uh…uh…”
Breaking the silence Clinicians & patients may break silent EHR use, leading to multitasking. Clinicians may emerge from EHR tasks to re-engage patients. Patients often interjected “chit-chat,” but also voiced biomedical or psychosocial concerns. • After 43 sec, clinician explains: “I’m going to send this to your pharmacy now, ok? This is the new dose of the furosemide.”
• After 25 sec, pt asks “So are you ready for Christmas?”
• After 32 sec, pt asks: “So they didn’t talk to you either about my pap machine?”
• After 12 sec, pt lifts his foot “My [primary] doctor was worried about this…”. Clinician replies “Oh yeah…when did this happen?” and examines it.
Unaddressed patient concerns Both multitasking and silent EHR use offer risks of unaddressed patient concerns. Multitasking EHR Use:
• Pt describes friend who had a stroke: “He called me to tell me what happened.” Clinician: “So let’s just review, you’re taking simvastatin which is a cholesterol medicine…”
• “Yes and sometimes I have tremendous pain.” Clinician: “Mm-hm..and you’re still taking…?”
Silent EHR Use: Pt explains problems with getting medications on time from pharmacy. Clinician: “That’s weird” & resumes silent EHR use.

“WHY DON’T YOU LOOK THAT UP?” VERSUS “COLLABORATIVE SELF-DIRECTED LEARNING:” THE IMPACT OF RESIDENCY TRAINING CONTEXT ON SELF-DIRECTED LEARNING Adam P. Sawatsky; John T. Ratelle; Sara Bonnes; Jason Egginton; Thomas J. Beckman. Mayo Clinic, Rochester, MN. (Control ID #2698863)

BACKGROUND: Self-directed learning (SDL) is an important component of the Accreditation Council for Graduate Medical Education practice-based learning and improvement competency. Previous research has elucidated the significance of personal characteristics and learning process within SDL. However, there is a need to understand the impact of learning context on SDL within residency training. Therefore, we explored resident physicians’ perceptions of the learning context to characterize the roles of faculty and training programs in supporting resident SDL.

METHODS: We drew upon existing principles of SDL and used constructivist grounded theory to explore the effects of context on SDL during residency training. We conducted 7 focus groups (FG) with 46 internal medicine residents. FG guides were developed from literature and expert review. A trained facilitator moderated the FGs, which were transcribed verbatim. Transcripts were explored using open coding and analytic memos to guide subsequent FGs. Constant comparison and axial codes revealed themes related to SDL context. Themes were organized within a theoretical model of SDL to understand the impact of residency training context. The findings were confirmed with member checks.

RESULTS: Residents identified the roles of Individual faculty members and residency programs in supporting SDL throughout the learning process, including triggering SDL, identifying learning objectives, formulating learning objectives, utilizing resources, applying knowledge, and assessing learning. Three archetypes of faculty support for SDL emerged: 1) role modelling personal SDL; 2) challenging learners by asking, “Why don’t you look that up?”; and 3) engaging learners in “collaborative SDL.” These archetypes incorporate slightly different approaches to supporting SDL, and were seen as complementary means for engaging learners. Residents discussed the importance of residency programs creating and supporting a culture of SDL by providing time and resources for learning, structuring teaching opportunities that deliver a framework for future learning, incorporating opportunities for self-assessment, offering social learning opportunities, and teaching SDL skills.

CONCLUSIONS: SDL is viewed by residents as an internal orientation to learning that integrates process and personal characteristics, like motivation. This study highlights the effect of contextual factors, including the role of individual faculty members and the residency program, on the process of SDL during residency training. Residents elaborated archetypes of faculty involvement in SDL, and provided perspectives on achieving these strategies. The current findings can be used to assist residency programs with faculty development and creating a culture of SDL.

CROSS-DISCIPLINARY ROLE AGREEMENT IS STILL LACKING IN THE VA PCMH Karleen Giannitrapani 6; Linda Kim3; Alexis K. Huynh5; Susan E. Stockdale2; Alison Hamilton4; Lisa V. Rubenstein1. 1GLA VA, North Hills, CA; 2Greater Los Angeles VA Healthcare System, Sepulveda, CA; 3VA-Greater Los Angeles, Los Angeles, CA; 4Veterans Administration, Los Angeles, CA; 5Veterans Affairs, Sepulveda, CA; 6Veterans Health Administration, Menlo Park, CA. (Control ID #2707577)

BACKGROUND: With increasing demand for primary care services and a deficient supply of qualified providers to meet this need, new interdisciplinary team based models of primary care delivery, including the Patient Centered Medical Home (PCMH), have emerged. Under PCMH, the roles of supporting teams members expand in include top-of-license tasks allowing some tasks that historically fell to only physicians to be shared or redistributed. We explore the role expansion of clinical associates (CAs) (licensed practical nurses, health technologists and medical assistants). Specifically, in this study we query team members of a newly implemented patient centered medical home (PCMH) to identify facilitators and barriers of role self-efficacy, a belief of possessing the capacity to execute their new team based roles effectively.

METHODS: We employ longitudinal qualitative analysis involving two waves of 105 semi-structured interviews with interdisciplinary providers based in Veterans Health Administration (VA) PCMH teams. Primary data were collected in 2011–2012 (wave 1) and 2014 (wave 2). We assess team member experiences approximately 1 year after the implementation to identify facilitators and barriers to role self-efficacy; we compare these with facilitators and barriers identified 4 years post implementation.

RESULTS: In wave one, three themes that functioned as facilitators/barriers to the self-efficacy necessary for successful role expansion of CAs were identified: 1) role training 2) time and resources for roles and 3) cross-disciplinary role agreement. By wave two, most training complaints were resolved. Time and resources to complete expanded tasks as well as cross-disciplinary role agreement persisted as challenges. Specifically, insufficient coordination between medicine and nursing or administrative leadership about staff roles was unresolved.

CONCLUSIONS: The goal of freeing up physician time by sharing tasks effectively across a team, will likely not be realized without team member role self-efficacy, specifically time and resources to completed expanded role tasks and cross-disciplinary agreement about roles. When implementing interdisciplinary teams in primary care, existing disciplinary paradigms and leadership pathways are impacted and in turn impact team member role-self efficacy and ultimately team functioning. Investing resources in engaging managers and in interdisciplinary leadership approaches may be an important strategy for supporting PCMH team member role self-efficacy.

CROSS-SECTIONAL STUDY OF PHYSICIAN BURNOUT AND ORGANIZATIONAL STRESSORS IN A LARGE ACADEMIC HEALTH SYSTEM Kristine Olson 3; Seppo Rinne4; Mark Linzer2; Christine Sinsky1; Sandip Mukherjee3; Michael Bennick3; Ronald Vender3; Harlan M. Krumholz3; Theodore Long3. 1American Medical Association, Chicago, IL; 2Hennepin County Medical Center, Minneapolis, MN; 3Yale University School of Medicine, New Haven, CT; 4Center for Healthcare Organization and Implementation Research, Bedford, MA. (Control ID #2703601)

BACKGROUND: Physician burnout is common nationwide and tied to adverse outcomes for physicians, patients, and healthcare systems. The aim of this study is determine the prevalence of burnout among academic faculty, hospital employed, and community private practice physicians, and to identify remediable workplace stressors associated with physician burnout.

METHODS: This was a cross-sectional survey via email of the estimated 4118 clinicians affiliated with a large non-profit academic tertiary medical center. The sample represented three cohorts: academic faculty, hospital employed, and community private practice physicians. Burnout was measured by the Maslach Burnout Inventory, calculated as defined by a score ≥ 27 on emotional exhaustion or a score of ≥ 10 on depersonalization, or both. The perceived presence of seven workplace stressors was measured by the Mini-Z survey, and odds ratios for burnout were determined by separate multivariable logistic regression models, adjusting each for age, gender, full-time equivalent (FTE), and group (academic, employed, private practice).

RESULTS: Of the estimated 4118 clinicians invited by email, 1252 received the survey having clicked the link, and 557 responded (45% completion rate). The prevalence of burnout was 58.6%. Physicians who perceived poor control over workload (53.9%) were more likely to experience burnout (80.4% vs 33.1%, OR 7.6, 95% CI 4.7–12.2, p < 0.001). Physicians who perceived team effectiveness as unsatisfactory (16.8%) were more likely to experience burnout (84.3% vs. 53.4%, OR 7.2, 95% CI 3.1–17.1, p < 0.001). Physicians who perceived work atmosphere as hectic-chaotic (51.7%) were more likely to experience burnout (74.8% vs. 41.4%, OR 4.3, 95% CI 2.7–6.7, p < 0.001). Physicians who felt their values were not in alignment with department leadership (42%) were more likely to experience burnout (72.7% vs 48.0%, OR 3.2, 95% CI 2.0–5.1, p < 0.001). Physicians who felt their time for documentation was unsatisfactory (64.6%) experienced more burnout (68.4% vs. 40.7%, OR 4.2, 95% CI 2.6–6.9, p < 0.001), as did the 53.4% of physicians who felt the amount of time on the EMR at home was unsatisfactory (65.7% vs. 50.6%, OR 2.1, 95% CI 1.4–3.3, p = 0.036). Academic faculty experienced more burnout than those in private practice (62.2% vs 43.7%, p = 0.006). Burnout was most prevalent among trainees, early career physicians, and women.

CONCLUSIONS: In this cross sectional study, the prevalence of physician burnout was 58.6%, and highly associated with lack of control over workload, team effectiveness, a more chaotic work atmosphere, unsatisfactory time for documentation, a lack of value-alignment with departmental leadership, and the presence of EMR stress. Private practitioners experienced less burnout than academic faculty or hospital employed physicians. Academic faculty were most affected. Organizational leadership commissioned this work to guide interventions and begin serial monitoring.

GENDER DISPARITIES IN SMALL GROUP VERBAL PARTICIPATION AMONG 1ST YEAR MEDICAL STUDENTS Mary Clare Bohnett; Sylvia Peterson-Perry; Molly R. Rabinowitz; Mariah Peterson; Shadi Dowlatshahi. OHSU, Portland, OR. (Control ID #2698509)

BACKGROUND: It has been shown that gender disparities exist for female clinicians today and that female trainees are disadvantaged by the so-called “hidden curriculum”. However, research is lacking on specific targets at which to aim interventions. Class participation has been correlated with academic success, but it is unknown whether participation disparities exist during preclinical medical education, or whether increasing participation may be a locus for combating gender disparities. This descriptive pilot study explores gender disparities at the level of preclinical medical education, using verbal participation as a proxy metric. It aims to (1) quantify any gender disparities in verbal participation; (2) examine whether factors such as facilitator gender or inclusive facilitation impact verbal participation; and (3) encourage medical education initiatives that address gender gaps before the hidden curriculum widens them later in training.

METHODS: Observational data was collected over a 4-week period during a first year medical school course. Ten small work group sessions were observed and unique verbal participation events were counted by gender. Facilitator gender, gender composition of group, and inclusionary comments by facilitator were also recorded.

RESULTS: Descriptive statistics analysis revealed a disparity in average verbal participation between males and females of 2.5:1, when normalized by group gender composition. At least one instance of inclusive facilitation per session increased female participation, while a lack of inclusive facilitation doubled the disparity.

CONCLUSIONS: Verbal participation disparities exist, but are somewhat mitigated by inclusive facilitation. More research is needed on implications and interventions to promote gender equity.

figure a

Average verbal participation events, with and without inclusive comments by facilitator (per student per session), normalized for group gender composition

GEOGRAPHIC VARIATION IN PRESCRIPTION OPIOID USE DURING THE POSTPARTUM PERIOD Nora V. Becker 2; Brandon Maughan1, 2. 1Emergency Physicians Integrated Care, Salt Lake City, UT; 2University of Pennsylvania, Philadelphia, PA. (Control ID #2709687)

BACKGROUND: Morbidity from prescription opioid use varies substantially by geography. There is a lack of published literature on opioid use in the immediate postpartum period.

METHODS: We extracted claims data from the Optum Clinformatics Data Mart, a database of approximately 14 million commercially-insured individuals. The sample included women with claims for vaginal or Cesarean delivery during 2001–2013. Extracted data included age, race, inpatient diagnosis and procedural codes, and type of filled prescriptions. Primary outcome was proportion of women who filled an opioid prescription within four days of discharge. Outcomes were stratified by mode of delivery (uncomplicated vaginal, complicated vaginal, uncomplicated Cesarean, complicated Cesarean) and compared across US census divisions using two-tailed tests of proportions.

RESULTS: Among women with uncomplicated vaginal delivery (n = 905,584), 28.3% filled opioid prescriptions. Opioid use differed by 459% between the lowest-use division (Middle Atlantic, 10.7%) and highest-use division (East South Central, 49.2%) (p < 0.0001). Complicated vaginal deliveries (n = 80,510) had higher opioid use (42.1%) and similarly large variation among divisions (19.2% vs. 61.9%, p < 0.001). Opioid use after uncomplicated Cesarean (n = 357,813) and complicated Cesarean (n = 85,388) was similar (76.2% vs. 77.0%). The Middle Atlantic division had lower opioid use rates for Caesarean deliveries (61.0 and 60.1%, respectively) compared to other divisions (rates of 76.2–79.6 and 76.8–80.2%, respectively).

CONCLUSIONS: There is significant geographic variation in prescription opioid use following vaginal delivery. This variation in prescription opioid use may identify opportunities to improve prescription safety and reduce opioid-related harms among women in the postpartum period.

IMPLEMENTATION OF THE RURAL VA MULTICENTER MEDICATION RECONCILIATION QUALITY IMPROVEMENT STUDY (R-VA-MARQUIS) Caroline Presley 2, 4; Kathleene Wooldridge2; Susan Byerly2; Amy R. Aylor3; Christianne Roumie2, 4; Robert S. Dittus4, 2; Amanda S. Mixon1. 1VA Tennessee Valley Healthcare System and Vanderbilt University, Nashville, TN; 2Vanderbilt University Medical Center, Nashville, TN; 3Veterans Engineering Resource Center (VERC), Indianapolis, IN; 4VA Tennessee Valley Healthcare System, Nashville, TN. (Control ID #2704394)

BACKGROUND: Unintentional medication discrepancies at care transitions can contribute to adverse drug events. High quality medication reconciliation can decrease unintentional medication discrepancies but is difficult to implement.

METHODS: R-VA-MARQUIS was a feasibility study to improve inpatient medication reconciliation practices using a mentored-implementation design conducted from 2014–2016 in three VA hospitals caring for rural Veterans. An evidence-based toolkit of best practices in medication reconciliation was adapted to the VA setting. Distance mentors evaluated baseline medication reconciliation practices and guided local improvement teams as they implemented toolkit interventions through monthly site phone calls and yearly site visits. At each site, the number of unintentional medication discrepancies per Veteran was collected in a sample of control and intervention patients to monitor response to the interventions. These data were obtained by comparing a Best Possible Medication History (BPMH) taken by trained pharmacists with the medication lists in provider notes and orders upon admission and discharge. Data analysis was performed using traditional QI methods; unintentional medication discrepancies per Veteran were plotted on run charts and XmR charts for each site. Student’s t-test was used to compare control and intervention patients.

RESULTS: Sites 2 and 3 successfully implemented several toolkit components. Site 2 utilized pharmacy students to take BPMH on admission, implemented risk stratification, and trained providers on how to take BPMH. Site 3 implemented standardized discharge process and documentation, trained providers on how to take BPMH, and, most importantly, hired an inpatient clinical pharmacist. Facilitators of implementation at Sites 2 and 3 included committed multidisciplinary improvement teams and support from senior leadership. Site 1 was unable to implement any toolkit interventions and faced barriers of high turnover of staff and lack of multidisciplinary engagement on the improvement team. Data was collected on a total of 797 patients across the sites. Patients were 68.7 years old on average; 94.5% were male. Patients were on a mean of 10.1 medications at admission; 45.6% had been admitted at least one time in the previous year. At Site 2, unintentional medication discrepancies per Veteran were not improved in intervention patients compared to control (4.45 vs. 3.5, p = 0.012). At Site 3, unintentional medication discrepancies per Veteran were reduced in intervention patients compared to control and this reduction was sustained over the course of the study (1.79 vs 4.77, p < 0.001).

CONCLUSIONS: The mixed results by site highlight the complexity of implementing evidence-based practices for medication reconciliation. This study adds to understanding of how to implement best practices to improve medication reconciliation in smaller hospitals caring for rural Veterans.

UTILIZATION OF PRIMARY CARE AND PREVENTIVE HEALTH SERVICES AMONGST INDIVIDUALS IN SAME-SEX AND OPPOSITE-SEX PARTNERSHIPS; A CROSS-SECTIONAL OBSERVATIONAL ANALYSIS OF MEDICAL EXPENDITURES PANEL SURVEY (MEPS), 2003–2001 Ani Abrahamyan 1; Igor I. Bussel1; John R. Blosnich2, 1; Janel Hanmer1. 1University of Pittsburgh Medical Center, Pittsburgh, PA; 2VA Pittsburgh Healthcare System, Pittsburgh, PA. (Control ID #2706132)

BACKGROUND: The health of sexual minorities is national priority in the US. Prior publications reported health disparities amongst sexual minorities, including health risk behaviors, health service access/utilization, and differences in the prevalence of medical conditions. However, little evidence exists about differences in utilization of primary care and preventative health services. The objective is to describe the differences in utilization of primary care and preventive health services amongst individuals in same-sex (SS) and opposite-sex (OS) partnerships and to elucidate perceptions of the quality of care received.

METHODS: An observational cross-sectional analysis of the Medical Expenditures Panel Survey (2003–2011) was conducted on matched cohorts of 494 individuals in SS and OS partnerships. Measures included identification of a PCP, basic preventive care utilization, and perceptions of quality. Descriptive analysis was performed to determine frequencies for categorical variables and means for continuous variables.

RESULTS: SS partnered men had a higher rate of compliance with colon cancer screenings (i.e., FOBT or colonoscopy) than OS partnered men (50.3% vs 30.2%, P = 0.001). The rate of flu shots within the last 2 years was greater amongst SS partnered men compared to their OS partnered cohorts (35.5% vs 53.1%, P = 0.001). There were no differences among men or among women in the perception of quality of care, which was assessed by responses to the perceived timeliness and necessity of care, perceptions about providers listening, explaining, and showing respect, as well as the ease of seeing a specialist.

CONCLUSIONS: In sexual minorities who identify themselves as SS partnered, there appears to be minimal difference in utilization patterns of primary care and preventive health services with the exception of greater utilization of colon cancer screening and flu shot compliance by SS partnered men. These differences may be driven, in part, by HIV status and an overlap in screening modalities used to diagnose problems associated with receptive anal intercourse. The lack of differences in perception of quality of care is in conflict with earlier studies that showed greater dissatisfaction with care amongst SS partnered people. This could be explained by a changing political climate with increasing equality and improved quality of life for sexual minorities or it may simply be due to the use of a different sample of same-sex partnered people. These findings are based on a subset population that is in a partnership and identifies themselves as such, thus it may not necessarily be representative of sexual minorities as a whole.

10 year ASCVD RISK AND INCIDENT ASCVD IN THE SYSTOLIC BLOOD PRESSURE INTERVENTION TRIAL (SPRINT) Timothy B. Plante; Stephen P. Juraschek. Johns Hopkins University, Baltimore, MD. (Control ID #2706812)

BACKGROUND: Atherosclerotic cardiovascular disease (ASCVD) is the top cause of disability and death in the US. Guidelines recommend integrating 10y ASCVD risk into primary prevention counseling. Observational studies have shown fewer ASCVD events with lower SBP levels but it’s unclear if intensive SBP control below 140 mm Hg will further reduce risk. SPRINT randomized adults to standard or intensive BP control. A composite ASCVD outcome has not been reported. Whether intensive BP control reduced risk of incident ASCVD for SPRINT participants is unknown.

METHODS: SPRINT randomized 9,361 adults without diabetes or stroke to standard (135–139 mm Hg) or intensive (<120 mm Hg) SBP control. After excluding baseline ASCVD or ages outside of the risk estimator validated range, we calculated 10y ASCVD risk. HR compared intensive vs. standard BP control and ASCVD events, ASCVD or death, and death alone. Visualization of HR by level of 10y ASCVD risk for ASCVD events used a restricted cubic spline model with 95% CI relative to the median 10y ASCVD risk using Harrell’s method. Kernel density plots depicted distribution of 10y ASCVD risk.

RESULTS: After excluding ASCVD at baseline or age >79y, 6,911 participants remained. Mean age was 65y, 37% were female, 35% were Black, median 10y ASCVD risk was 16%. There was a trend towards fewer ASCVD events in the intensive BP control arm (n = 105 v. 136; HR 0.78; 95% CI 0.61–1.01). ASCVD events or death was reduced (0.80; 0.65–0.99) but not death alone (0.82; 0.61–1.10). The spline model showed increasing hazard of ASCVD events with increasing baseline 10y ASCVD risk. Intensive but not standard BP control showed increased hazard of ASCVD events at a 10y ASCVD risk level >30%.

CONCLUSIONS: Among adults at increased risk for ASCVD, intensive BP control is associated with a non-significant trend towards reduction in incident ASCVD. Modeling of treatment effects across levels of baseline risk demonstrates a positive association between intensive BP control and incident ASCVD events among those with the highest baseline risk, though few were in this risk range. Further investigation to delineate levels of risk that may benefit from this intervention are warranted.

figure b

5-STAR NURSING HOME (NH) REPORT CARD ASSOCIATED WITH PREVENTABLE HOSPITALIZATIONS FROM POST-ACUTE CARE Kira L. Ryskina 1; Tamara Konetzka3; Rachel M. Werner2. 1University of Pennsylvania, Philadelphia, PA; 2University of Pennsylvania and Philadelphia VA, Philadelphia, PA; 3University of Chicago, Chicago, IL. (Control ID #2697507)

BACKGROUND: In an effort to better inform consumers, Medicare began publicly reporting NH quality using a simple-to-understand 5-star rating. This resulted in increased attention to NH ratings from consumers and NHs themselves. While NH star ratings increased substantially since they were first published in December 2008, whether this represents improvement in quality or, instead, “teaching to the test” is unknown. Our objective was to test whether improvements in star ratings were associated with improvements in patient outcomes.

METHODS: We used Medicare hospital claims and NH clinical assessment data for 2,143,217 fee-for-service Medicare beneficiaries receiving post-acute care in 15,318 NHs Jan 2007 - Jun 2010. We used these data to recreate NHs’ star ratings over the study period and to observe hospitalizations and risk-adjustment variables. To account for the differences in patient exposure to NH care, our primary outcome was number of potentially preventable hospitalizations per 100 patients per month (pppm), weighted by the number of days in the NH. We compared the correlation between star ratings and hospitalizations before vs after Dec 2008 by estimating hospitalizations as a function of NH star ratings, an indicator for whether the hospitalization occurred before or after ratings were released, and the interaction between the two. All analyses included risk adjustment from the NH clinical data and NH fixed effects to account for heterogeneity across NHs. Standard errors were adjusted for clustering within NHs.

RESULTS: The association between the number of preventable hospitalizations and a NH’s star rating weakened after the ratings became public: the number of hospitalizations was 0.58 pppm higher in 5-star (best-rated) NHs (95% CI 0.06 to 1.11, p = 0.029) and 0.51 pppm lower in 1-star (worst-rated) NHs (95% CI −0.86 to −0.16, p = 0.004) after star ratings release vs. before (Figure). When stratifying stays by baseline characteristics, these differences were concentrated in NHs with low baseline quality and staffing levels.

CONCLUSIONS: Recent improvements in NH star ratings were not accompanied by improvements in preventable hospitalizations for post-acute care patients. This suggests improvements in star ratings were driven by potentially superficial changes that impact ratings, but not the underlying quality of care.

figure c

A COMMUNICATION INTERVENTION AIMED AT MEDICINE DOCTORS AND NURSES IMPROVES PATIENT SATISFACTION SCORES Jill Allenbaugh 2; Jennifer Corbelli2; Laurie Rack2; Carla Spagnoletti1. 1University of Pittsburgh, Pittsburgh, PA; 2University of Pittsburgh Medical Center, Pittsburgh, PA. (Control ID #2703658)

BACKGROUND: Patient satisfaction continues to play an ever-expanding role in healthcare, from reimbursement to hospital rating. However, effective strategies to improve Hospital Consumer Assessment of Healthcare Providers and Systems (HCAHPS) remain scarce. At the University of Pittsburgh Medical Center, a multidisciplinary patient experience committee aims to improve patient experiences through targeted interventions. The committee identified communication-specific HCAHPS scores, specifically for items “Doctors (nurses) explained things in a way you could understand,” as an area for intervention. We hypothesized a lack of adequate training in beside communication and identification of poor health literacy as reasons for suboptimal communication scores. The aims of our study were to (1) develop a multidisciplinary curriculum to teach physicians and nurses to best deliver complex medical information and (2) evaluate the effectiveness of the curriculum by examining HCAPHS scores for doctor and nurse communication.

METHODS: A clear health communication curriculum was developed for 112 internal medicine (IM) residents and 120 nurses from the general medicine wards. The curriculum was disseminated through 60–90 min workshops that were facilitated by IM clinician educators and a nurse educator. The content included didactic teaching on health literacy, small group discussion and simulated videos of optimal communication skills during bedside rounds and the discharge process. Data was collected from 422 HCAHPS surveys from patients discharged from these wards over a 6-month period. We compared that percentage of “top-box” scores (corresponding to a rating of 9 or 10 on a 10-point scale) on communication items between 3 months pre and 3 months post curriculum.

RESULTS: A total of 76 residents (participation rate 68%) and 80 nurses (participation rate 67%) who work in 5 wards across 3 hospitals attended. Percentage of “top-box” scores improved for all doctor and nurse communication-specific HCAHPS items after the intervention (See Table).

CONCLUSIONS: Our data shows that a multidisciplinary clear communication curriculum with a focus on health literacy can significantly improve doctor and nurse communication specific HCAHPS scores at a large academic hospital. As communication skills are essential to providing quality patient care, our curriculum has value for any medical specialty or healthcare system seeking ways to improve the patient experience.

  Pre (% top box) Post (% top box) P value
Doctors explained things in a way you could understand 63 70 .15
Doctors listened carefully to you 66 77 .02
Doctor treated you with courtesy and respect 80 84 .25
Overall communication with doctor 70 77 .10
Nurses explained things in a way you could understand 59 73 .003
Nurses listened carefully to you 60 72 .018
Nurses treated you with courtesy and respect 74 80 .199
Overall communication with nurses 65 75 .025


BACKGROUND: Scheduled telephone follow-up visits are one strategy for improving access to specialty care practices, primarily because telephone follow-up visits can be completed in less time (4–6 telephone visits per hour vs. 2–3 office visits per hour, for instance) with lower overhead costs. However, there is limited data on the relative advantage to patients from such a program. To inform ongoing improvement of the program, and to share the experience with payers, who do not currently reimburse scheduled telephone visits, we conducted a survey to evaluate patient-reported outcomes associated with telephone and office follow-up visits.

METHODS: Scheduled telephone follow-up visits were introduced in January 2015 as a substitute for some scheduled follow-up office visits in specialty care practices at UCSF. Physicians determine the appropriate patients based upon clinical condition and patient acceptability. Patients are not charged for the service and the UCSF physician receives a modest internal RVU credit. English-language surveys were delivered to all patients (>18 y.o.) who completed a follow-up visit (either telephone or office) at UCSF in Endocrinology, Hepatology, or Multiple Sclerosis clinics between March and May 2016 (n = 2741). Patients with email addresses on file were sent the survey electronically (Qualtrics LLC), and those without were contacted by phone.

RESULTS: 96% of patients had email addresses, of which 16% (n = 426) responded. An additional 8 patients without email addresses completed telephone surveys. The final study sample included 13 and 17% of patients with telephone visits and office visits, respectively. Respondents and non-respondents were similar with regard to clinic, sex, and mean age. Main results are show in Table below. Among telephone visit respondents, 87% reported that it replaced the need for an office visit. When telephone visit respondents were asked about their out-of-pocket costs and total time spent for previous office visits, they had similar responses as office visit respondents. 87% of telephone visit respondents and 85% of office visit respondents agreed that insurance companies should pay for telephone visits that substitute for an office visit, and more than half were willing to pay a co-payment of approximately $20 (median).

CONCLUSIONS: This study fills an important gap in understanding the patient’s perception of telephone follow-up care, and represents a critical first step in mobilizing health plans to pay for telephone visits.

  Phone Visit N = 110 Office Visit N = 324 P-value
Age, mean 55 years 56 years 0.525
Female sex 69% 60% 0.10
Very satisfied or satisfied with telephone visit 98% N/A  
Out-of-pocket costs, median (IQR) $0 (0,0) $50 (20,100) <0.001
Total Time Spent, median (IQR) 10 min (5, 50)* 240 min (150, 420)  
Urgent/emergency care visit 5% 4% 0.92

*imputed based on 10 min scheduled telephone visit time, and request of patients to be available for phone visit 30 min before and after scheduled visit time.

A COMPARISON OF MEDICATION-BASED VERSUS MEDICAL CLAIMS-BASED RISK ADJUSTMENT TO PREDICT 1-YEAR MORTALITY AMONG VETERANS DUALLY-ENROLLED IN VA AND MEDICARE PART D Thomas R. Radomski 3, 4; Xinhua Zhao4; Joseph T. Hanlon3, 4; Joshua M. Thorpe2, 4; Carolyn T. Thorpe2, 4; Florentina Sileanu4; John P. Cashy4; Jennifer A. Hale4; Maria K. Mor4; Leslie R. Hausmann4, 3; Julie M. Donohue6; K J. Suda1; Kevin T. Stroupe1; Chester Good4; Michael J. Fine4, 3; Walid F. Gellad5, 3. 1Hines VA Hospital, Hines, IL; 2University of Pittsburgh, Pittsburgh, PA; 3University of Pittsburgh School of Medicine, Pittsburgh, PA; 4VA Pittsburgh Healthcare System, Pittsburgh, PA; 5VA Pittsburgh/University of Pittsburgh, Pittsburgh, PA; 6University of Pittsburgh Graduate School of Public Health, Pittsburgh, PA. (Control ID #2700094)

BACKGROUND: There is systematic undercoding of medical comorbidities within administrative claims in the Veterans Health Administration (VA) as compared to Medicare. This undercoding may lead to bias when applying claims-based risk adjustment indices in studies of Veterans who dually use both VA and Medicare services, or when comparing medical outcomes between VA and Medicare. Medication-based risk adjustment models may be an unbiased method of risk adjustment in these circumstances or when medical claims are not uniformly available. Our objective was to adapt a medication-based risk index for use with combined VA and Medicare data and compare its prognostic accuracy to commonly used claims-based risk adjustment methods in predicting 1-year mortality.

METHODS: Our cohort was all individuals enrolled in both VA and Medicare Part D who filled at least one opioid prescription from either system in 2012. We adapted an existing VA-based medication risk index (Rx Risk-V) to also incorporate Medicare pharmacy and durable medical equipment claims. VA and Medicare medications/supplies were classified into 1 of 45 disease categories, applying previously derived category weights. Using the C-statistic (C), we compared the prognostic accuracy of the adapted Rx Risk-V + demographics (age, gender, race, Medicaid eligibility, and disability status) in predicting 1-year mortality to models that included demographics with and without alternative risk adjustment methods (prescription count, Charlson index, or Elixhauser index). We also compared the adapted Rx Risk-V model to models that included demographics, adapted Rx Risk-V, and Elixhauser or Charlson. We conducted a sensitivity analysis by restricting our cohort to dual enrollees who had a clinical encounter and/or received a medication within both VA and Medicare (i.e. dual users), as we expected the adapted Rx Risk-V model may perform better among this cohort.

RESULTS: The 271,343 Veterans in the overall cohort had a mean age of 70.5 years; 96.1% were male, 81.7% were non-Hispanic white, and 63.4% were dual users. Overall, 9.4% died within 1 year. The prognostic accuracy of the adapted Rx Risk-V (C = 0.76) was significantly greater than models using demographics alone (C = 0.72) or a prescription count (C = 0.74), but significantly lower than models using Charlson (C = 0.79) or Elixhauser (C = 0.79) (P < 0.001 for all comparisons). The model combining demographics, adapted Rx Risk-V, and Charlson provided significantly greater prognostic accuracy than all other models (C = 0.80, P < 0.001). We found similar patterns of relative model performance in analyses limited to dual users.

CONCLUSIONS: The adapted Rx Risk-V index, when used in combination with common claims-based indices, enhances prognostic accuracy in regard to 1-year mortality. Using this model in place of claims-based indices marginally lowers prognostic accuracy, but remains a viable method of risk adjustment when medical claims are not available or the use of claims-based indices may lead to bias.

A COMPARISON OF QUALITY END-OF-LIFE CARE IN PATIENTS WITH HEMATOLOGIC AND SOLID MALIGNANCIES: IDENTIFYING DEFICIENCIES IN PATIENT CARE Ramy Sedhom 1; Daniel Sedhom1; David Barile2. 1Rutgers Robert Wood Johnson Medical School, New Brunswick, NJ; 2University Medical Center of Princeton at Plainsboro, Plainsboro, NJ. (Control ID #2670839)

BACKGROUND: Limited data is available on the differences in end-of life care for patients with hematologic and solid malignancies. We retrospectively analyzed and compared the quality of end-of-life care between patients with hematologic malignancies and those with solid tumors.

METHODS: All adult patients who died of advanced cancer between 1/1/2015 and 5/1/2017 at our institution were included. We collected baseline demographics and indicators reflective of quality end-of-life care including: hospitalizations, emergency room visits, admissions to the intensive care unit, and chemotherapy use during the final 30 days of life. Data analysis included descriptive statistics, Chi-square tests, and multivariate logistic regression.

RESULTS: 57/316 (18%) of patients who died had hematologic malignancies. In the final 30 days of life, patients with hematologic cancers had more visits the emergency room (56%vs. 41%, P = 0.03), admissions to the hospital (84% vs. 45%, P < 0.001), >2 admissions (26% vs. 11%, P < 0.001), extended lengths of stay (48% vs. 6%, P < 0.001), transfers to the intensive care unit (42% vs. 8%, P < 0.001) and death (35% vs. 4%, P < 0.001). They also had more frequent chemotherapy use (48% vs. 12%, P < 0.001) compared to patients with solid tumors. In addition, patients with hematologic malignancies were less frequently referred to palliative care (5% vs. 21%, P = 0.02) or hospice (4% vs. 20%, P = 0.02). Chart review, and multivariate analysis, revealed more aggressive care among patients with hematologic malignancies.

CONCLUSIONS: Patients with hematologic malignancies receive different treatment at end-of-life and are less likely to receive palliative services when compared to patients with solid malignancies. Future research is needed to identify differences in overall patient care in order to provide higher quality care.

A CRITICAL APPRAISAL OF GUIDELINES FOR ELECTRONIC COMMUNICATION: DO CURRENT RECOMMENDATIONS NEED MODERNIZATION Joy L. Lee 2, 1; Nir Menachemi3; Michael Weiner1, 2. 1Indiana University, Indianapolis, IN; 2Regenstrief Institute, Indianapolis, IN; 3Richard M. Fairbanks School of Public Health at IUPUI, Indianapolis, IN. (Control ID #2704852)

BACKGROUND: Patients are increasingly interested in electronic access to providers (e.g., via emailing, secure messaging, or texting), and federal policies are encouraging secure messaging. Nevertheless, little is known about how providers should use electronic tools to communicate with patients. We sought to identify published evidence that underpins guidelines for electronic communication, and to evaluate evidence-based recommendations for electronic-communication between patients and members of the healthcare team. We identified a dearth of published work in both areas and instead critically appraise existing published guidelines and suggest an agenda for future work in this area.

METHODS: We performed a narrative review of provider-targeted guidelines on electronic communication, searching Ovid MEDLINE, EMBASE, and PubMed databases using relevant search terms (e.g. ‘patient’ and ‘email,’ ‘webmail,’ or ‘messaging’ and ‘guideline,’ or ‘standard’). We limited the search to articles published in English. We also manually searched the citations of relevant articles. We identified the themes and suggested practices of each guideline, and evaluated whether promoted practices reflect current practice.

RESULTS: We identified 11 guidelines on electronic communication between providers and their patients. Although many articles referenced emerging evaluations of electronic communications, no recommended practices were underpinned by evidence of the effectiveness of the practices. Within the guidelines, privacy and data security were major themes. Most guidelines recommend a discussion with patients regarding expectations for communication. They focused little on suggested practices regarding the content of electronic communications, communication delivery, and how best to achieve communication functions such as providing information and emotional support. Many guidelines are out of step with current practice regarding both how email is used and how patients and clinicians interact. For example, they include suggestions that providers should direct patients to use the auto-reply e-mail feature to acknowledge reading messages and to also direct patients to use specific subject headings in electronic communications.

CONCLUSIONS: Although current guidelines for electronic communication may be rooted in a solid conceptual basis, they have three major flaws: a) the guidelines are not founded on evidence, b) they offer little guidance on how best to use electronic tools to communicate effectively, and c) many guidelines are out of line with current practice. Just as clinicians and clinical trainees are being taught face-to-face communication with patients based on an evidence-rich foundation, the same rigorous work needs to be applied to electronic communication. Researchers need to evaluate and identify effective practices systematically, create a framework to evaluate quality of communication, and test the relationship between electronic communication delivery and quality of care.


BACKGROUND: Although substantial research demonstrates the effectiveness of integrated care models, there are very few studies that compare clinical outcomes in depression for a co-location model versus a collaborative care model (Krahn, Bartels, Coakley, et al., 2006; Unützer, Katon, Callahan, et al. 2002) As health care reform efforts support integrated model sustainability and scalability, understanding whether there are differential outcomes achieved can help practices allocate resources appropriately. An academic medical center with large ambulatory primary care practices (N = 19), serving primarily Medicare and Medicaid recipients with significant racial and ethnic diversity, began offering integrated care primarily through co-location (usually a licensed clinical social worker and a part time psychiatrist) in the Fall of 2014. In February 2015, 7 sites were chosen to begin implementation of the integrated collaborative care model (CCM) as part of a Health Care Innovations Award from the Center for Medicare and Medicaid Innovations. The CCM program aimed to improve care quality through the addition of a care manager to the behavioral health team, allowing enhanced “between visit” care and case reviews in a multidisciplinary team, facilitated by a measurement-informed care patient registry.

METHODS: A cross sectional natural experimental design compared depression symptom severity outcomes for patients attending Montefiore primary care sites employing co-located care (N = 12) and sites utilizing CCM (N = 7). Depression symptom severity was measured with the Patient Health Questionnaire 9 (PHQ-9). Patients were enrolled in the study if they scored 10 or above on the PHQ-9, indicating moderate to severe depression. Eligible patients receiving both intervention types had access to short-term, psychotherapy, concrete social services, and medication management. At 10 to 16 weeks (mean = 12 weeks) following enrollment, patients were re-administered the PHQ-9 by a trained and blinded independent assessor over the phone.

RESULTS: A total of 240 participants were enrolled (N = 122 at co-location sites; N = 118 at CCM sites). Significant within group reductions in depressive symptoms were observed in the co-location sites (difference, 2.23, p < 0.0003) and the CCM practices (difference, 5.04, p < .0001). Between group differences indicated patients in CCM sites demonstrated significantly greater reduction in depressive symptoms compared to patients at the co-location sites (difference, −2.81; p = .0005).

CONCLUSIONS: The CCM intervention appears to result in a significantly greater reduction in depressive symptoms compared to the co-location model across a range of Montefiore primary care clinics serving low income, diverse patients. Replication will be necessary in larger samples to further support these findings.

A CROSS SECTIONAL STUDY REVEALS AN ASSOCIATION BETWEEN ELECTRONIC CIGARETTE USE AND MYOCARDIAL INFARCTION Taher M. Tayeb; Talal Alzahrani; Nardos Temesgen; Ivan Pena. The George Washington University, Arlington, VA. (Control ID #2689119)

BACKGROUND: Electronic cigarettes (E-cigarettes) have grown in popularity as an alternative to traditional cigarettes especially among consumers who want to reduce the risk of morbidity and mortality associated with smoking. Nonetheless, a recent study showed that both E-cigarettes and traditional cigarettes cause an increase in oxidative stress and endothelial dysfunction, however this effect is less pronounced with E-cigarettes. Currently, there is a limited study that shows the impact of E-cigarette in the cardiovascular system. Therefore, data from the 2014 National Health Interview Survey (NHIS) was used to evaluate the effect of E-cigarettes on the cardiovascular system, specifically the effect on myocardial infarction(MI).

METHODS: Analysis of the 2014 National Health Interview Survey (NHIS) database was performed to examine the effect of E-cigarettes on MI. Initially, subjects were assigned to one of two groups: those with a history of MI and those without a history of MI. The t-test and chi-square test were subsequently applied to compare the different demographics and health characteristics between these two groups. A logistic regression model was then used to measure the association between E-cigarettes and history of MI. Data was adjusted for multiple risk factors for MI including age, gender, race, body mass index, income, the status of smoking cigarettes, and history of hypertension, diabetes, and hypercholesterolemia.

RESULTS: A total of 35,156 subjects were included in the final logistic model. [Summarize a number of baseline variables with respect to health and demographic characteristics] Analysis showed that increasing age (OR, 1.04; p <0.001), history of hypertension (OR, 2.72; p < 0.001), high cholesterol (OR, 2.19; p < 0.001), and diabetes (OR, 1.68; p < 0.001) are associated with an increased odds of myocardial infarction. With respect to smoking, increased frequency of smoking was associated with increasingly higher odds of MI when compared to patients who had never smoked: every day smokers (OR, 2.75 p < 0.001), some day smokers (OR, 2.39; p < 0.001), and former smokers (OR 1.80; p < 0.001.) In contrast, females (OR, 0.49; p < 0.001), Hispanics (OR, 0.62; p < 0.001), and people with higher incomes (OR, 0.93 [95% CI, 0.90–0.96]; p < 0.001) have lower odds of heart attack. With respect to Electronic cigarette use and MI, analysis revealed an odds ratio of 1.42 with p = 0.017.

CONCLUSIONS: Our findings indicate that Electronic cigarette use, when adjusted for other risk factors, is associated with a 42% increased odds of myocardial infarction. This increase in odds is consistent regardless of traditional cigarette smoking history. More studies are needed to further assess this risk.

A CROSS-SECTIONAL STUDY OF SELF-REPORTED COLORECTAL CANCER SCREENING STATUS IN US ADULTS WITH AND WITHOUT DISABILITIES Bliss Temple 2; Lisa I. Iezzoni1. 1Harvard Medical School, Boston, MA; 2University of California San Francisco, San Francisco, CA. (Control ID #2703635)

BACKGROUND: Colorectal cancer screening is a life-saving aspect of primary care recommended for most adults 50 to 75 years old. People with disabilities face difficulties accessing health care services and disparities in rates of cervical and breast cancer screenings. This study used a large, nationally representative database to examine rates of self-reported colorectal cancer screening to determine whether there are disparities for people with disabilities compared to those without disabilities.

METHODS: This was a secondary analysis of cross-sectional data from adults 50–75 years old in the 2013 National Health Interview Survey, which uses a multistage area probability design to sample the civilian non-institutionalized US population. The outcome was self-reported current colorectal cancer screening status. The main predictor was self-reported limitation in functional abilities (i.e. disability) in at least one of 12 domains. Severe disability and type of disability (physical, sensory, cognitive/intellectual, chronic disease, mental health) were secondary predictors. Covariates included age, sex, race, Hispanic ethnicity, and seeing a primary care provider (PCP) in the previous year. Logistic regression analyses were used to model the odds of current screening status. The model was evaluated for interactions, collinearity, goodness of fit, and misspecification.

RESULTS: Of the 13,332 subjects in the analysis, 51.3% reported any disability and 28.2% severe disability. Subjects with any disability reported a 59.3% current colorectal cancer screening rate; those with severe disabilities, 58.8%. Both groups had higher screening rates than those without disabilities (52.1%). Those with disabilities did not have significantly different adjusted odds (OR 1.05; 95% CI 0.98,1.13) of current screening compared with those without disabilities. Statistically significant covariates were age, Asian race, Hispanic ethnicity, and seeing a PCP within the past year. The odds ratio for those with versus without severe disability was not statistically significant (OR 0.95; 95% CI 0.87,1.03). Odds ratios for health condition subgroups compared to those without that condition were statistically significant only for the groups with physical health conditions (OR 1.15; 95% CI 1.06,1.24) and chronic diseases (OR 0.89, 95% CI 0.80,0.99).

CONCLUSIONS: Overall rates of colorectal cancer screening for people with and without disabilities were not significantly different, despite barriers to accessing care and disparities in other cancer screening rates. This may reflect an increased focus on health by people with disabilities compared to the nondisabled and/or fewer barriers to colorectal cancer screening than for other cancers. Policy initiatives aimed at decreasing health disparities for people with disabilities may be more productively focused on areas other than colorectal cancer screening.

A DIFFERENCE-IN-DIFFERENCE ANALYSIS OF TRENDS IN UTILIZATION FOLLOWING A PRIMARY CARE REDESIGN INTERVENTION Kevin Nguyen 1; Alyna T. Chien2, 4; David J. Meyers1, 3; Sara Singer1, 2; Meredith Rosenthal1. 1Harvard T.H. Chan School of Public Health, Boston, MA; 2Harvard Medical School, Boston, MA; 3Brown University School of Public Health, Providence, RI; 4Boston Children’s Hospital, Boston, MA. (Control ID #2699123)

BACKGROUND: Many academic medical centers (AMCs) are seeking novel approaches to deliver higher quality care and lower costs for patients cared for in primary care practices. Beginning in mid-2012, 20 practices across 6 AMCs affiliated with Harvard Medical School launched a learning collaborative to establish team-based care at both hospital- and community-based practices. We describe trends and changes in utilization 3 years before and 1 year following this intervention.

METHODS: The learning collaborative followed a sequenced change strategy, and relied on in-person learning sessions, monthly calls, and leadership training. We conducted difference-in-difference analyses using generalized linear models to compare changes in utilization between intervention practices and controls before and after the intervention using the Massachusetts All-Payer Claims Database (APCD) for 2009 to 2013. The APCD aggregates medical claims each year for private payers and Medicaid in the Massachusetts. The 20 intervention practices were identified by linking National Provider Identifier number to the Massachusetts Health Quality Partnership Physician database, which trackers physician hierarchy within practices and systems. We used propensity weights to identify controls that were of comparable size and affiliated with AMCs. Patients were attributed to intervention or control practices annually. Utilization measures included outpatient visits, emergency room (ER) visits, inpatient hospitalizations, ambulatory care sensitive inpatient (ACSIP) hospitalizations, and ambulatory care sensitive ER (ACSER) visits. We adjusted the utilization outcomes based on the number of months a patient was enrolled. We estimated percent change in intervention practices’ utilization relative to changes in controls’ utilization.

RESULTS: One year following the beginning of the intervention, we found a statistically significant reduction in outpatient use in intervention sites compared to controls by 0.72 visits per patient per year, or approximately 7.3% (p < 0.01). From baseline, ER use decreased in intervention sites compared to controls by 0.06 visits per patient per year (8.3%, p < 0.01) and inpatient hospitalizations in intervention practices decreased by 0.1 admissions per patient per year (19.2%, p < 0.01). ACSIP admissions decreased by 0.02 per patient per year in intervention sites compared to controls (28.6%, p < 0.01) and ACS ER decreased by 0.01 visits per patient per year (9.1%, p < 0.01).

CONCLUSIONS: We observed significantly larger reductions in utilization in intervention practices when compared to control practices. Our findings suggest that involvement in the learning collaborative was associated with significant reductions in healthcare utilization.

A DIGITAL LANGUAGE DIVIDE? THE RELATIONSHIP BETWEEN INTERNET MEDICATION REFILLS AND MEDICATION ADHERENCE AMONG LIMITED ENGLISH PROFICIENT PATIENTS Alejandra Casillas 2; Leo Morales3; Jonathan Grotts2; Chi-Hong Tseng1; Gerardo Moreno1. 1UCLA, Los Angeles, CA; 2University of California Los Angeles, Los Angeles, CA; 3University of Washington, Seattle, WA. (Control ID #2703287)

BACKGROUND: Use of an Internet portal to refill medicines positively affects medication adherence among English-speakers. However, no studies examine if Internet refills also affect medication adherence for patients with Limited English Proficiency (LEP). Our objectives were to 1) examine the relationship between use of an Internet medication refill system and medication adherence among linguistically diverse patients with chronic conditions and 2) compare this relationship between LEP and English Proficient (EP) patients.

METHODS: We analyzed 2013–2014 data from 509 surveyed adults with hyperlipidemia, hypertension, and/or diabetes mellitus in the Group Health Cooperative. Data were merged from a multilingual phone survey, plan enrollment, claims data, and electronic medical records. The primary outcome, medication adherence rate (%), was calculated for patients with at least two pharmacy fills over a 12-month period by the “Continuous Measure of Medication Gaps” (CMG) method. For Internet refill system use, patients were asked, “Have you used the health systems Internet site to refill any medications in the last 12 months?”. LEP status was captured by patient self-identification with a non-English primary language, and/or a claims record of interpreter use. We used multivariate linear regression models to examine the adjusted effects of use of the Internet refill system on medication adherence, and compared this association between LEP and EP patients. We adjusted for age, gender, education, insurance, chronic conditions, and number of medications.

RESULTS: 384 patients (75%) had a calculable CMG (to produce adherence rate): 134 EP and 250 LEP, included in the models. Patients who had calculable CMG’s were older, prescribed more medications, and more likely to have Medicare or Medicaid coverage, versus patients without a CMG. In unadjusted analyses, LEP patients had lower use of the Internet refill system (p < .001) and lower adherence versus the EP group (p < .001). In multivariate analyses, LEP status (b = −0.022, p = .047) was negatively associated with adherence. Use of the internet refill system (b = 0.030, p = .002), and Medicare or Medicaid insurance (b = 0.030, p = .007), were positively and independently associated with adherence. In stratified models, use of remote Internet refills was positively associated with adherence, even when examining EP (b = 0.029, p = .003) and LEP patients (b = 0.027, p = .049) separately.

CONCLUSIONS: We found that the use of remote medication refills through an Internet portal was independently and significantly associated with higher medication adherence in both EP and LEP patients. These findings suggest that LEP patients could be under-utilizing a beneficial Internet tool. Should our health-care system fail to ensure that LEP patients can digitally manage their health care (such as refilling medications) through an Internet portal, we face the risk of widening the existing health care gaps via this digital divide.


BACKGROUND: The rate of uninsured adults in the United States has declined by nearly 40% since the Affordable Care Act in 2013, nevertheless, more than 20 million adults in the US remain uninsured. Safety net programs are necessary to help address the needs of those who remain uninsured, but to what degree? We have examined the enrollment of people in a safety net breast and cervical cancer screening program (BCCCP), serving uninsured and low-income women in an area of Metro Detroit, Wayne County, MI, to better understand the magnitude and determinants of this need.

METHODS: We examined patterns of enrollment by age, ethnicity, and income, in our community-based breast and cervical cancer control program for the period from 2013 through 2015. We used logistic regression to estimate adjusted odds ratios for enrollment according to race/ethnicity.

RESULTS: Please see the attached graph depicting the number of breast and cervical cancer screenings of non-Hispanic and Hispanic women enrolled in our safety net program from January 2013 through October 2014. - The graph shows a decline in the number of non-Hispanic women enrolled in our program from October 2013 through July 2014. - The graph also demonstrates that Hispanic enrollment in the program remained steady during this time, which correlates with the implementation of the ACA in Michigan (April 2014). - Our adjusted Odds Ratio for being Hispanic in 2015 compared to 2013 was 4.44.

CONCLUSIONS: The ACA has reduced, but not eliminated the need for specific safety net programs in the United States. While enrollment in the BCCCP declined by 70% pre-post ACA, we have served over 15,000 women from 2013 through 2015 who, in the absence of this safety net program, would likely have remained unscreened. In particular, we note that the BCCCP has taken on increased importance for Hispanic women in our community, who remain disproportionately uninsured, here and elsewhere.

figure d

A NARRATIVE REVIEW OF PHYSICIAN PERSPECTIVES REGARDING THE SOCIAL AND ENVIRONMENTAL DETERMINANTS OF OBESITY Feenalie Patel 2; Ashley K. Haddad1; Debra A. Werner1; Monica E. Peek1. 1University of Chicago, Chicago, IL; 2University of Tennessee Health Sciences Center, Cordova, TN. (Control ID #2703874)

BACKGROUND: Obesity is a multifactorial disease with influences from biology, behavior, social norms, physical environment, and the larger health system. However, current guidelines for managing obesity discuss only biologic and behavioral therapeutic options, without suggesting ways to address the social and/or environmental determinants of obesity. In this review, we evaluate the ways that physicians conceptualize and address the social and environmental determinants of obesity to better understand current practice.

METHODS: We searched PubMed, Ovid, and PsycINFO databases and conducted hand searches of relevant bibliographies over the past 30 years. Search terms included variations on the following: obesity, attitude, perception, belief, knowledge, physician, adult, recommendation, treatment, cause, and etiology. Our final review included all articles that elicited physician perspectives on the causes or management strategies of obesity. We categorized causes and management strategies into the following domains, defined by the ecologic model of health: individual (i.e. biological, psychological, behavioral), social, environmental, and macro-level.

RESULTS: We found 1144 total articles, 24 of which met inclusion criteria. Eleven articles (46%) assessed physician perspectives on obesity causes and 18 (75%) assessed physician perspectives on obesity management. Of the 11 articles that discussed obesity cause, all (100%) evaluated physician perspectives on at least one non-individual domain, but no paper evaluated physician perspectives across all domains. Of the 18 articles that discussed obesity management, 12 (66%) evaluated physician perspectives on at least one non-individual domain, and again, none evaluated physician perspectives across all domains. When asked to rank the causes of obesity across multiple domains, physicians ranked individual causes as more important than social or environmental causes in all except one study. Physicians also used individual management strategies more frequently (and believed individual management strategies were more successful) than social or environmental management strategies in all studies.

CONCLUSIONS: No articles in our review allowed physicians to evaluate the causes or management strategies of obesity across all domains that might influence weight. However, within this limitation of available literature, our review suggests that clinicians rank individual characteristics as more important in the development and management of obesity than social or environmental factors. While clinicians may simply under-value the impact of social and environmental obesity determinants, this finding may also represent a lack of perceived ability or responsibility to address non-individual health determinants in clinical practice. As we move into the era of population health management, more research is needed to understand the most appropriate ways for physicians to address the social and environmental determinants of chronic diseases such as obesity.

A NEEDS ASSESSMENT FOR EMPANELMENT IN A RESIDENT CONTINUITY CLINIC Jennifer Bracey; Alanna Stone; Sara Turbow. Emory University School of Medicine, Atlanta, GA. (Control ID #2704973)

BACKGROUND: Empanelment is the act of assigning patients to a primary care provider (PCP). Improving empanelment could create greater satisfaction in clinic by increasing continuity and enabling closer relationships with patients, both of which could decrease resident clinic burnout and create interest in GIM careers. In this analysis, we sought to define the baseline accuracy of current PCP assignments for our resident continuity clinics, and we investigated if residents were engaging in empanelment by assigning themselves as PCP for patients without accurate PCP assignments.

METHODS: This analysis was performed in three IM resident continuity clinics at a large urban hospital. We examined charts for all patients scheduled in clinic on two dates in December 2016. We recorded if the patient had a PCP assigned prior to the visit and, if they did, whether the listed PCP was the resident the patient was scheduled to see that day, a different current resident, a graduated resident, or another type of provider (ex. advanced practice provider, subspecialist, or community provider). We reviewed the same charts after clinic to determine if patients without a PCP or with an incorrect PCP had been reassigned to the resident they saw that day. Patients who did not show to clinic or who were added to the schedule after our initial pre-clinic chart review were excluded from the analysis. SAS Studio was used for data analysis.

RESULTS: We reviewed 496 charts. Of those, 45.6% (n = 226) had the resident they were scheduled to see that day listed as their PCP. 13.1% (n = 65) had another current IM resident listed, 13.9% (n = 69) had a graduated resident listed, 10.5% (n = 52) had another type of provider listed, and 16.9% (n = 84) were unassigned. Of the 205 patients eligible to be assigned a new PCP at their visit, 44.4% (n = 91) came to the appointment. Five patients were assigned to PCPs other than the resident they saw that day. Of the remaining 86 patients, 45.4% (n = 39) had their PCP assignment changed to the resident they saw that day.

CONCLUSIONS: We found that residents assign themselves to be the PCP at a low rate. There was no statistically significant difference in the percent of patients reassigned to a current resident between patients in each of the “inaccurate” PCP categories (p = 0.7314). This also does not appear to differ by PGY level. This analysis was limited by high no-show rates: 42.3 overall and 67.9% for patients without an assigned PCP. Overall, this project shows a need for increased focus on empanelment in our resident continuity clinic. Our next steps will focus on resident-level interventions to: 1) educate them on empanelment and why this process is important and 2) strategize ways for them to improve empanelment. By improving empanelment, we may see greater resident satisfaction with clinic, less clinic burnout and an increased interest in GIM careers.

A NEW FRAMEWORK FOR STAKEHOLDER ENGAGEMENT IN EARLY STAGE TRANSLATIONAL SCIENCE Amy LeClair 1; Thomas Concannon3; Virginia Kotzias3; Allison M. Cole2, 2; Simona Kwon4; Alexandra Lightfoot5. 1Tufts Medical Center, Boston, MA; 2University of Washington, Seattle, WA; 3RAND Corporation, Boston, MA; 4NYU School of Medicine, New York, NY; 5University of North Carolina at Chapel Hill, Chapel Hill, NC. (Control ID #2705519)

BACKGROUND: Stakeholder and community engagement (SCE) is a national priority for the National Center for Advancing Translational Science (NCATS). There is an established framework for stakeholder engagement for the latter stages (T2-4) of translational science, but no such framework currently exists for early stage (T1) clinical researchers. Four Clinical Translational Science Award (CTSA) hubs launched a collaboration to develop a new framework for engaging communities and stakeholders in T1 research.

METHODS: We led structured individual and group discussions with T1 investigators to learn about: (1) the health decisions they seek to inform with research evidence, (2) the actors who make those decisions, and (3) the individuals and organizations that are affected by those decisions. Focus groups lasted, on average, one hour. Interviews lasted approximately 30 min. Both were recorded. Audio recordings were transcribed and de-identified, and transcripts were coded using Dedoose™. We used a deductive-inductive procedure to develop the framework for stakeholder engagement in T1 research. A deductive codebook was development from the focus group & interview guides; emergent themes were added and the codebook was revised after preliminary inductive analysis. Two coders analyzed all transcripts using a constant comparison approach. Analysis included input from research participants to assure the framework could be used by T1 researchers.

RESULTS: Eighteen individuals connected to four CTSA hubs participated in the discussions. Participants came from the fields ranging from basic chemistry and drug development to infectious disease and pediatrics and represented both methodological and topical experts. Preliminary analysis produced several findings. First, basic scientists can identify stakeholders, contrary to concerns that stakeholder engagement is not applicable at this stage. Second, stakeholders in early stage translational research (T1), as identified by the research participants, do not fit into the same framework as those further down the translational spectrum (T2-T4). A unique framework may therefore assist early stage researchers in the process of stakeholder engagement. Finally, participants indicated they would like more guidance on who, how, and when to engage them in their research.

CONCLUSIONS: By showing T1 researchers how to identify and involve their stakeholders in (1) defining research questions, (2) carrying out research activities, and (3) disseminating research evidence, this work has the potential to improve the use of basic science evidence in latter stages of translation from bench to bedside.

A NEW MODEL FOR EXPANDING AAMC MEDICAL EDUCATION RESEARCH CERTIFICATION Joshua T. Hanson; Jeffrey Jackson. University of Texas Health Science Center, San Antonio, San Antonio, TX. (Control ID #2702129)

BACKGROUND: It is difficult to provide large-scale faculty development in medical education scholarship. This type of development frequently requires formal training and, in response, the Association of American Medical Colleges (AAMC) has created the Medical Education Research Certificate (MERC) program. MERC courses are offered at national meetings; however, this can be costly and time consuming to the individual faculty member. In an effort to provide this to faculty at the University of Texas School of Medicine (UTSOM), we engaged with AAMC to host the MERC program locally. Our goal was to create a sustainable faculty development program in medical education scholarship by increasing opportunity while reducing resource utilization.

METHODS: MERC faculty traveled to the UTSOM campus where all the sessions were held. All workshop topics were offered in morning and afternoon sessions to accommodate clinical and administrative duties of the attendees. Attendees were recruited through the Division of General & Hospital Medicine, Office of Undergraduate Medical Education, and the Office of Graduate Medical Education. To measure impact of the program, ten-item pre and post surveys were administered and were constructed to measure increases in skills and knowledge of educational research. Individual paired comparisons were made between pre- and post- survey results by paired samples t-test and grouped according to certification completion. Survey data were compiled and analyzed using Microsoft Excel and Stata.

RESULTS: The total number of registrants was 53 and the average number of workshops attended was 5. Twenty-nine registrants completed certification, for which the criterion is completion of six sessions. The total cost of the week of sessions was $16,179, which calculates to $305 per registrant or $558 per certificate. Complete pre- and post-survey data were available for 38 registrants (71.7%). All items had mean increases self-reported knowledge and skills. Among those that completed certification, comparisons demonstrated statistically significant improved knowledge and skills in 10/10 domains. Among those that did not complete certification, 4/10 domains had statistically significant improvements.

CONCLUSIONS: Engaging with the AAMC to host the MERC program locally resulted in significant cost savings considering attendance and certification typically requires national meeting registration, travel costs, and is a multi-year endeavor. Certification was associated with increased knowledge and skills in all of the domains. Even partial attendance resulted in knowledge increases. This early success has led to the planning of a second annual MERC intensive program with recruitment of attendees from more departments. Future plans include measurement of sustained self-reported knowledge and measurement of increased scholarship after attendance and certification.

A NOVEL APPROACH TO BLINDING IN A RANDOMIZED CLINICAL TRIAL (RCT) OF A COMPLEX BEHAVIORAL INTERVENTION: THE PREVENTING AMPUTATION BY TAILORED RISK-BASED INTERVENTION TO OPTIMIZE THERAPY (PATRIOT) STUDY Xiao Xiao 1; Kimberly L. Stone1; Sarah Sullivan1; Yvonne Ye1; Madeline Russell1; Tova Bergsten1; Trina Wijangco1; Samantha M. Hill1; Sundar Natarajan2. 1Department of Veteran’s Affairs New York Harbor Health System, New York, NY; 2VA New York Harbor Healthcare System, New York City, NY. (Control ID #2705763)

BACKGROUND: In order to generate valid results in RCTs, it is important to collect unbiased data. Behavioral intervention trials are more difficult to blind than pharmacologic trials. In our ongoing behavioral trial, PATRIOT, which aims to prevent foot complications in diabetes through improved self-care, we are intervening both face-to-face as well as remotely. Consequently blinding is more difficult. Here we illustrate the different blinding processes used.

METHODS: In the PATRIOT trial, following randomization during the baseline visit, the intervention group receives computer-based education and demonstration on how to use a special foot thermometer. Following that, intervention participants receive a comprehensive intervention that includes regular telephone counseling and tailored mailings. The control group receives health prevention strategies not related to foot care. For this complex intervention, we needed to develop new strategies to maintain blinding at the participant, research assistant (RA), counselor, outcome adjudicator and data analyst levels. We created a “Blinding Tracker” to identify unblinded and blinded staff so that participants pass from unblinded to blinded staff so that only blinded staff collect data. The integrity of the study is maintained by careful monitoring of blinding with any break in blinding being contained.

RESULTS: To date, we have enrolled 221 participants. While participants know the arm to which they are randomized, we needed to make sure that the staff collecting data are blinded. Participants are educated throughout the whole study process by the telephone counselors about the importance of blinding using non-scientific descriptions. Though the counselors know their participant’s treatment assignment, they do not know their foot photography results and other outcomes. The RA’s, who conduct study visits and collect data, are blinded to treatment assignment. To improve efficiency and preserve blinding, we have different RAs for different phases of a participant’s progress. Initially, RAs are initially blinded, but by the end of their involvement with a particular subject, when data collection is done, they become unblinded so they can show the educational videos and demonstrate foot thermometer use. However, RA’s can also become unblinded prematurely. To date, out of 96 six-month visits, there have been 26 such instances. RAs are subsequently transitioned off a particular subject to be replaced by a blinded RA. We will also control for RA blinding in the analysis. The adjudicators who read the foot photographs to determine outcomes and the statisticians are blinded to treatment assignment.

CONCLUSIONS: Novel techniques have been used to achieve and maintain blinding, but it is resource intensive. While conducting complex trials, vigilance and responsiveness are needed. Finally, blinding information should be incorporated in the analysis in order to get the most valid results.


BACKGROUND: Palliative Care (PC) specialists assist patients with serious illness, their families, and their clinicians with medical decision making, symptom management, advance care planning, and end of life care. As the population ages, the need for PC services will increase, and will exceed the supply of PC clinicians. Primary care physicians will need to feel comfortable in navigating many primary PC skills.

METHODS: We describe a 2 week PC rotation for third year internal medicine (IM) residents at a large academic medical center. Between 2012 and 2016, 32 IM residents completed a PC elective rotation. The curricular components included didactic sessions and bedside teaching on a busy inpatient PC consult service and inpatient unit. All 32 residents completed a pre and post survey during the elective rotation in which they ranked their comfort level with primary PC topics using the RedCap survey tool (1 = Need further basic instruction, 2 = Competent to perform with close supervision, 3 = Competent to perform with minimal supervision, 4 = Competent to perform independently). They also ranked their overall comfort level in working with seriously ill and dying patients (1 = very uncomfortable to 10 = very comfortable).Surveyed topics included conducting a family conference, breaking bad news, discussing DNR orders, home hospice, shift in treatment to comfort care, treatment withdrawal and advance directives, performing a basic pain assessment, using oral and parenteral opioids, adjuvant analgesics, assessing and managing difficult symptoms, and assessing decision-making capacity.

RESULTS: Prior to beginning the rotation, the IM resident average comfort level was 2.5 (range 2 to 3.5) for the above topics, indicating that their comfort was between “competent to perform with close supervision” and “competent to perform independently.” After completion of the rotation, the average comfort level was 3.3 (range 2.4 to 3.5), indicating they felt that they were competent to perform these tasks with minimal supervision. IM residents also indicated an increase in their comfort level in working with seriously ill and dying patients, from 6.7 prior to 8.6 after the rotation.

CONCLUSIONS: We have shown that the self-reported comfort level of IM residents improves after a 2 week elective rotation with primary PC topics. At the end of the rotation, participants felt that they are able to perform these tasks with minimal supervision, which is reassuring as they complete residency. The limitations of this study are that it is an elective rotation, so we are unable to speak to the comfort level of IM residents who did not complete this rotation. This study raises questions as to the comfort level for PC topics of IM residents that have not participated in a PC rotation, and whether this improvement in resident comfort level translates to improved patient outcomes.

A PERSONAL TOUCH TO DECREASE NO-SHOWS Jay L. Mathur 2; Noor Khan1; Mehrshid Kiazand2. 1UPMC Mercy, Moon Township, PA; 2UPMC Mercy, Pittsburgh, PA. (Control ID #2699304)

BACKGROUND: Outpatient non-attendance, or no-shows, produces major consequences for both patients and providers. A recent review of 42 studies found that telephone reminders, mostly automated, were the most effective system to reduce no-shows (9.4% reduction). Our Internal Medicine Clinic continues to have high no-show rates (27.1% annually) despite automated reminders. We sought to understand the impact personal phone reminders may have on no-show rates vs automated reminders. We were particularly interested in evaluating whether PGY1 no show rates were impacted at the beginning of their training.

METHODS: Our teaching outpatient center is divided into two separate sections, Firms A and Z, each with its own resident panel, patient sets, and dedicated staff. All patients receive automated phone reminders 48 hours prior to their office visit. Only Firm A patients also received a personal phone reminder by a staff member during the first 4 months of the academic year. We analyzed the effect of this additional personal phone reminder versus automated reminders in improving no-show rates.

RESULTS: 4,121 scheduled clinic visits from 6/29/15 - 10/31/15 were analyzed. 1,067 (25.9%) no-shows occurred over this period. The mean average scheduled visits per resident was 72.4 for A firm and 75.0 for Z firm (p = .7188). The cancellation rate per resident was similar in both firms (A = 22.0%, Z = 20.8% p = .3951). The average rate of no-shows per resident was significantly less for firm A (A =23.3%, Z = 28.7%, p < .0007). PGY1’s did have a lower no-show rate in firm A although the numbers were not statistically significant (A = 23.5%, Z = 29.2%, p = .0639)

CONCLUSIONS: Personalized reminders may have resulted in significantly less no-shows in our study. We postulate that despite the cheaper and convenient alternative of automated telephone reminders, personalized reminders can reduce no-shows and lead to more efficient care and better outpatient training. Personalized reminders can also increase initial R1 clinic visits and subsequent continuity experience. Lack of statistical significance for R1 data from our review maybe from small sample size (firm A n = 10, Z n = 10).

A POSITIVE DEVIANCE APPROACH FOR LARGE-SCALE DISSEMINATION OF AN EVIDENCE-BASED HYPERTENSION PROGRAM. Shari Bolen 2; Thomas Love3, 2; Randall D. Cebul1. 1Case Western Reserve University, Chagrin Falls, OH; 2MetroHealth/Case Western Reserve University, Cleveland, OH; 3Case Western Reserve University, Cleveland, OH. (Control ID #2706516)

BACKGROUND: Improvements in blood pressure (BP) control across diverse populations is challenging. Although evidence-based interventions for BP improvement have been successfully disseminated within large integrated health systems serving insured patients, less is known regarding successful large-scale dissemination approaches of evidence-based interventions for hypertension management across multiple health systems serving diverse patients. Our objective was to compare BP control over time within diverse practices that participate in a large regional health improvement collaborative (called Better Health Partnership, BHP) which used a positive deviance approach to identify and disseminate a best practice for hypertension management across multiple health systems.

METHODS: We compared cross-sectional results for the 116,042 patients seen in July 2015 - June 2016 to the 97,847 seen in July 2011 - June 2012 at all 33 practices who reported to BHP throughout that period. Using clinics’ electronic health records (EHR), patients are included in BHP’s biannual reports if they have a diagnosis of hypertension, are 18+ years of age, not pregnant, and have at least two primary care visits in the past 2 years and one in the last year at a participating clinic. BHP used the reported clinic-level EHR data to identify a best practice for hypertension management between 2009–2011 based on a protocol in one health system that showed dramatic improvement and high achievement of BP control. BHP disseminated this best practice using twice yearly public reports and region-wide learning collaboratives; practice coaching for sites with lower levels of good BP control; and EHR-catalyzed patient-centered care.

RESULTS: In 2015–16, patients in the 33 practices were middle-aged (mean 62 years) and diverse (52% Female, 62% White, 32% Black and 3% Hispanic). Across seven Northeast Ohio counties, 28% lived in low-income neighborhoods (median income below $33,000). Patients in these practices showed similar characteristics across reporting periods except for insurance, which went from 5% Medicaid and 9% Uninsured in 2011–12 to 13.2% Medicaid and 3.2% Uninsured in 2015–16. Across all patients in the 33 practices, rates of good BP control (<140/90 mmHg) rose from 67 to 72% (a difference of 5.2 percentage points) from 2011–12 to 2015–16, with increases of at least 4.8 percentage points in all insurance, race/ethnicity, sex, income and educational subgroups. Most improved were patients of Hispanic ethnicity (from 62 to 71%) and the uninsured (from 59 to 66%).

CONCLUSIONS: A positive deviance approach substantially accelerated the dissemination and implementation of a best practice protocol for hypertension management across multiple health care systems with diverse patients in a regional health improvement collaborative. A specific focus on low income and minority populations may have led to the greater improvements seen in uninsured and Hispanic subgroups.

A PRECISION MEDICINE APPROACH FOR PATIENT-REPORTED MENTAL HEALTH SCORES Anthony T. Fojo 2; Katherine L. Musliner1, 3; Peter P. Zandi1; Scott L. Zeger1. 1Johns Hopkins Bloomberg School of Public Health, Baltimore, MD; 2Johns Hopkins School of Medicine, Baltimore, MD; 3University of Aarhus, Aarhus, Denmark. (Control ID #2706653)

BACKGROUND: In recent years, precision medicine approaches have been developed for a variety of diseases. Early successes have rested on genomic data and laboratory markers. Precision medicine techniques are underutilized in the field of mental health, partly because clinical management depends on patient symptoms and not genetic targets or biomarkers. The increasing use of the electronic medical record (EMR) and patient reported outcome (PRO) data offers an opportunity to develop algorithms to predict patient outcomes on an individual level. Our objective was to develop a precision medicine framework for mental health symptom scores, which we applied to data from a clinical trial of risperidone for schizophrenia.

METHODS: We constructed a novel, multilevel prediction model that learns from serial symptom scores and predicts future symptom trajectories. We fit our model to the trial data - comprising scores on the Positive and Negative Syndrome Scale (PANSS) at weeks 0, 1, 2, 4, 6, and 8 - using Markov-Chain Monte-Carlo. We used 10-fold cross-validation to evaluate predictive accuracy, and analyzed how many observations were necessary to accurately predict outcomes at study completion.

RESULTS: Predictions were most accurate for the subsequent clinic visit, with diminishing accuracy farther into the future. Predictions of outcomes at study completion performed better after at two observations; additional observations yielded only marginal improvements. We present a sample of predictions for an individual based off of the first two clinic visits in the figure. Predictions of symptom scores at study completion based on the first two clinic visits deviated from the observed scores by a mean of 17.6 (interquartile range 6.8 to 23.0) on a scale from 30 to 210, and uncertainty intervals contained the true observation 98% of the time.

CONCLUSIONS: We demonstrate how a rigorous statistical model can leverage repeated measurements to predict disease trajectory on an individual level. Our methodology could be extended to other diseases, such as depression, anxiety, or substance abuse, where PROs are measured serially to guide treatment. Our methods could also be extended to time frames beyond eight weeks, or used to predict outcomes such as suicidal behavior or hospitalization. The systematic use of repeated PROs offers the promise of precision medicine in the field of mental health.

figure e

A QUALITATIVE ANALYSIS OF SYMPTOMATOLOGY IN OLDER HOMELESS ADULTS: RESULTS FROM THE HOPE HOME STUDY Adam Bazari 1; Maria Y. Patanwala2; Colette Auerswald1; Margot Kushel3, 4. 1UC Berkeley-UCSF Joint Medical Program, Berkeley, CA; 2University of California, San Francisco, San Francisco, CA; 3UCSF/ZSFG, San Francisco, CA; 4UCSF, San Francisco, CA. (Control ID #2706528)

BACKGROUND: The homeless population is aging. Older homeless adults experience the premature development of age-related conditions, but less is known about their symptom burden or experience. Somatic symptoms (physical symptoms) increase with age. Social (loneliness) and existential (guilt, regret, hopelessness) symptoms are distinct from but can overlap with mental health symptoms such as anxiety and sadness. We used qualitative methods to characterize the experience of somatic, social, existential and social symptoms in older homeless adults.

METHODS: We conducted semi-structured interviews with participants from the HOPE HOME cohort, a longitudinal study of homeless adults age 50 and older. Inclusion criteria included: current homelessness and high somatic symptom burden, or at least one somatic symptom with high psychological, social, or existential symptom burden. Interviews covered participants’ views on symptom etiology, impact on daily activities, personal strengths and management strategies. We employed Grounded Theory to analyze our data, including open coding, selective coding, and writing theoretical memos.

RESULTS: We interviewed 25 participants (60% men, 80% African American, ages 52–71). We found three emergent themes: 1) Existential and social symptoms cause as much distress as somatic symptoms. “My back pain actually is pretty real because I’m sleeping on cement but the thing that really gets me is the future. Sometimes there is a hopeless feeling that comes on.” “Right now [my feelings are] combined because the more mental, the more physical I go through…the only thing that is really destroying me now is I want to see my mother.” 2) There is a clear relationship between the experience of homelessness and symptoms. Participants describe how the experience of homelessness causes existential suffering that can manifest as somatic symptoms. “When I start thinking that I’m not gonna get off of this situation, my body starts to hurt, my stomach gets nauseated. It’s burning like it’s on fire.” 3) Participants manage symptoms with strategies from within and outside of biomedicine, including formal healthcare, religion, use of alternative medicines, and reliance on social supports. “[When I’m feeling angry], I just start praising God until He comes and allow his spirit to wrap his arm around me; put me in a nice sleep. When I wake up, I feel a lot better.”

CONCLUSIONS: The traumatic experience of homelessness causes and exacerbates physical and psychological distress. Among older homeless adults, existential symptoms are intertwined with and as distressing as somatic symptoms. To provide optimal care for this population, health care providers should screen for and address a broad range of symptoms, including somatic, psychological, existential and social.

A QUALITATIVE STUDY OF HOSPITALISTS’ PERCEPTIONS OF PATIENT SATISFACTION METRICS ON PAIN MANAGEMENT Susan L. Calcaterra 1, 2; Anne Drabkin1, 2; Ingrid A. Binswanger2, 4; Joseph W. Frank2, 3; Jennifer Reich5; Stephen Koester5. 1Denver Health, Denver, CO; 2University of Colorado, Aurora, CO; 3Denver Veterans Affairs Eastern Colorado, Denver, CO; 4Kaiser Permanente Colorado, Denver, CO; 5University of Colorado, Denver, CO. (Control ID #2707048)

BACKGROUND: Patient satisfaction metrics for pain management impacts federal incentive payments to hospitals. Hospitalists are key players in the delivery of high quality healthcare and are often financially compensated by their institutions for high satisfaction scores. Hospital-based initiatives to promote pain management may unintentionally contribute to excessive opioid prescribing. We aimed to understand hospitalists’ perceptions of the impact of satisfaction metrics on pain management in their clinical practice.

METHODS: We conducted 25 open-ended, in-depth interviews with hospitalist physicians practicing in one of five hospitals (two university, one safety-net, one Veterans Affairs, and one private) located in Colorado or South Carolina. Four team members, two hospitalist physicians and two qualitative researchers, systematically analyzed transcribed interviews to identify emergent themes using a team-based, mixed inductive and deductive approach.

RESULTS: Hospitalists felt institutional pressure to earn high satisfaction scores for pain management, which they believed led to increased opioid prescribing. One hospitalist stated: “I’m well aware that I’m being watched for my patient satisfaction scores. So at least consciously it does not affect my position on opioids, which is not to say unconsciously it doesn’t affect my decision.” Hospitalists expressed concern that tying physician compensation to satisfaction scores commoditized pain to the patients’ detriment. One physician recalled: “In my community practice, we were incentivized to keep people happy. It was expected that we keep up those scores. There is always pressure and you certainly don’t want your patients to not be satisfied, but I think when you are given individual incentives based on that, it does sometimes change your procedures.” Hospitalists believed that satisfaction scores would likely improve with more time spent at the patient’s bedside discussing pain management alternatives and options, but felt they lacked the time to do so. One hospitalist explained: “I think the patient’s satisfaction is mostly related to their relationship with the doctor rather than the pain. If you actually take the time to talk to the patient and they understand and they agree with you, then of course, the satisfaction scores are not going to be low.” Finally, hospitalist perceived that hospital administrators interpreted high satisfaction scores as representative of quality healthcare delivery. One physician stated: “I think patient satisfaction is one of the things in medicine where it is so far removed from what is actually happening clinically at the bedside. I don’t trust administrators who worry on a daily basis about patient satisfaction.”

CONCLUSIONS: Patients should be educated about reasonable expectations for pain management in the hospital. Institutionally supported methods to improve physician-patient communication and engagement may promote both patient-centered, safe pain management and patient satisfaction.

A RANDOMIZED COMPARATIVE EFFECTIVENESS TRIAL OF A TRANSITIONAL CARE CLINIC: 180-DAY EFFECTS David T. Liss 1; Christine Schaeffer-Pettigrew1; Emily Finch1; Andrew J. Cooper1; Avani Sheth2; Ashanti D. Tejuosho1; Caroline Teter1; Ronald T. Ackermann1. 1Northwestern University, Chicago, IL; 2Cook County Health and Hospital System, Chicago, IL. (Control ID #2706621)

BACKGROUND: Care transition interventions following hospital discharge can reduce subsequent rehospitalization. Early experimental evidence is mixed, particularly for vulnerable populations. This randomized trial examined a transitional care clinic for high-risk patients with no trusted medical home.

METHODS: We conducted a pragmatic comparative effectiveness trial of the Northwestern Follow-up Clinic (NFC), a transitional care clinic where a multidisciplinary care team provided coaching and assistance navigating the post-discharge transition through clinic visits and phone outreach, medication management support, social service and behavioral health coordination and transitioning to a permanent medical home. Between September 2015 and February 2016, patients were randomized to either NFC referral or referral to a federally qualified health center (FQHC); we used a 3:1 allocation ratio to meet the NFC site’s capacity constraints. Following an emergency department (ED) visit, observation stay or inpatient admission, adults were eligible for inclusion if they had no medical home; were referred to the NFC by a hospital provider, or; expressed that they did not wish to return to their medical home or their medical home was insufficient to manage their needs. We examined differences in hospital care utilization during the 30 days after initial discharge (primary outcomes) and over 90-day and 180-day follow-up. Negative binomial regressions estimated differences in rates of: A) ED visits; B) observation stays; C) inpatient admissions, and; D) total hospital visits (i.e., A-C combined). Regressions adjusted for age, sex, insurance, race/ethnicity, driving distance, prior NFC use and type of initial hospital visit.

RESULTS: A total of 654 eligible patients were randomized and included in analyses (490 NFC arm, 164 FQHC arm). The sample was balanced on all sociodemographic characteristics; most patients were uninsured (34.6%) or insured through Medicaid (49.7%). At 30 days, intent-to-treat analysis demonstrated non-significant differences in adjusted rates of ED visits (incidence rate ratio [IRR] 0.82, p = 0.46), observation stays (IRR 0.61, p = 0.19) and inpatient admissions (IRR 0.67, p = 0.16), but there were fewer total hospital visits in the NFC arm (IRR 0.65, p = 0.04). After 90 days, NFC patients had fewer adjusted inpatient admissions (IRR 0.60, p = 0.03) and total hospital visits (IRR 0.60, p = 0.01). Over 180 days, NFC patients had fewer adjusted ED visits (IRR 0.62, p = 0.04), observation stays (IRR 0.52, p = 0.02), inpatient admissions (IRR 0.51, p = 0.003) and total hospital visits (IRR 0.49, p < 0.001).

CONCLUSIONS: Over 180-day follow-up, rates of all utilization outcomes under study were significantly lower in the NFC arm. These results demonstrate the potential of transitional care clinics to improve transitions and reduce hospital care use in high-need populations.

A RANDOMIZED TRIAL OF DIFFERENT LEVELS OF INPATIENT MEDICAL ATTENDING SUPERVISION OF TRAINEES Kathleen M. Finn 2; Christiana Iyasere2; Yuchiao Chang1; Joshua Metlay3. 1Harvard Medical School, Boston, MA; 2MGH, Boston, MA; 3Massachusetts General Hospital, Boston, MA. (Control ID #2703602)

BACKGROUND: While the relationship between resident work hours and patient safety has been extensively studied, there is limited data on the impact of different levels of inpatient attending supervision of internal medicine residents on patient safety or resident education.

METHODS: We conducted a 9-month clustered randomized trial on an academic inpatient medicine teaching service where 22 experienced faculty were asked to provide two different types of resident supervision: “routine oversight” (joining morning rounds to review new admissions only) and “responsive oversight” (joining morning rounds to review both new admissions and established patients). Each faculty served as their own control and the order of routine vs. responsive oversight was randomized for different 2-week blocks on service. The primary outcome was the rate of potential medical errors using an established algorithm based on medical record review. Secondary outcomes included number of orders written, consultations and radiology studies. Resident education was evaluated with a time motion study to assess resident participation on rounds and surveys to assess both resident and attending perceived educational value of the two models.

RESULTS: A total of 1503 patients (5,691 patient-days) were included in the analysis. While the medical-error rate decreased by 9% from the routine oversight to the responsive oversight rounding models, this was not statistically significant (109.0 vs. 93.7 per 1000 patient days, P = 0.26). There was no difference in mean orders written in the morning (7 am–12 pm) or afternoon (12 pm–5 pm) in either routine vs. responsive models (63.8 vs. 81.3 total orders per day - morning, P = 0.12), in consultation notes (11.3 vs. 13.2 total consults notes per day, P = 0.31) or radiology studies (5.7 vs. 6.2 studies done per day, P = 0.89). A time motion analysis of 161 work rounds, found no difference in mean length of total time spent discussing established patients in the two models (202 +/− 46 min-routine vs. 202 +/− 57 min responsive, P = 0.99). The time a supervising resident spoke on work rounds did not change (58 +/− 20 min vs. 57 +/− 20 min, P = 0.58) but the time interns spoke did decrease when the attending joined rounds (64 +/− 19 min vs. 55 +/23 min, P = 0.008). In surveys both residents and interns reported that when an attending joined rounds they were less efficient, had decreased autonomy and had more fear of being judged. Attendings believed that the team knew their plan of care better though residents did not perceive this.

CONCLUSIONS: In an academic inpatient medical service, increased attending supervision did not significantly reduce medical errors or affect number of orders written, consultations or radiology studies. Residents perceived less autonomy and efficiency, though a time motion analysis found no difference in length of rounds or level of supervising resident participation.

A RANDOMIZED TRIAL OF INPATIENT LABORATORY TEST PRICE TRANSPARENCY IN THE ELECTRONIC HEALTH RECORD: THE PRICE TRIAL Mina Sedrak 3; Jennifer S. Myers4, 1; Dylan Small1; Irving Nachamkin4; Justin B. Ziemba5; Dana Murray4; Gregory Kurtzman1, 4; Jingsan Zhu1; Wenli Wang1; Deborah Mincarelli4; Daniel Danoski4; Brian Wells4; Jeffrey S. Berns4; Patrick Brennan4; Bill Hanson4; C J. Dine1, 4; Mitesh Patel1, 2. 1University of Pennsylvania, Philadelphia, PA; 2Crescenz VA Medical Center, Philadelphia, PA; 3City of Hope Comprehensive Cancer Center, Duarte, CA; 4University of Pennsylvania Health System, Philadelphia, PA; 5Johns Hopkins School of Medicine, Baltimore, MD. (Control ID #2700021)

BACKGROUND: While many health systems are considering increasing price transparency at the time of order entry, the evidence of its longer-term impact on clinician ordering behavior is unclear. Few randomized trials exist and among those conducted, they are single site, of limited duration (<6 months) and outcomes are not adjusted for patient characteristics or comorbidities.

METHODS: The Pragmatic, Randomized Introduction of Cost Data in the Electronic health record (PRICE) trial was a randomized, controlled trial conducted at three hospitals to evaluate the effect of displaying Medicare allowable fees for inpatient laboratory tests on clinician ordering behavior comparing changes during a one-year pre- and one-year post-intervention period. Test groups were randomly assigned to display fee data (intervention, n = 30) or not (control, n = 30) in the electronic health record. The main outcome measure was change in number of tests ordered per patient-day. Secondary outcome measures were change in number of tests performed per patient-day and associated fees per patient-day for tests ordered and performed. Multivariate adjusted models were fit to outcome measures adjusting for time trends and patient demographics, insurance, disposition, and comorbidity severity.

RESULTS: The sample included 98529 patients comprising 142921 admissions. Pre-intervention trends of order rates among the intervention and control groups were similar. In adjusted analyses of the intervention group compared to the control group over time, there were no significant changes in overall test ordering behavior (0.05 tests ordered per patient-day; 95% CI, −0.002, 0.09; P = 0.06) or associated fees (0.24 US dollars per patient-day; 95% CI, −0.42, 0.91; P = 0.47). Subset analyses found small but significant differences in tests ordered per patient-day based on patient intensive care unit (ICU) stay [(Patients with ICU stay: −0.16, 95% CI: −0.31, −0.01, P = 0.04)(Patient without ICU stay: 0.13, 95% CI, 0.08, 0.17; P <0.001)] and the magnitude of associated fees [(top quartile of tests based on fee value: −0.01, 95% CI, −0.02, −0.01; P = 0.04)(bottom quartile: 0.03, 95% CI, 0.002, 0.06; P = 0.04)]. Adjusted analyses of tests that were performed found a small but significant overall increase in the intervention group relative to the control group over time (0.08 tests performed per patient day, 95% CI, 0.03, 0.12; P < 0.001).

CONCLUSIONS: Displaying Medicare allowable fees for inpatient laboratory tests in the electronic health record did not lead to any meaningful or consistent changes in overall clinician ordering behavior. To our knowledge, this is one of the largest and longest evaluations of a randomized introduction of price transparency in a health care setting.

A RANDOMIZED TRIAL OF USING LOSS AVERSION AND ENHANCING GROUP-BASED INCENTIVES IN A PHYSICIAN PAY-FOR-PERFORMANCE PROGRAM Amol S. Navathe 1; Andrea B. Troxel2; Amanda Hodlofski1; Kristen Caldarella1; Amelia Bond1; Qian Huang1; Shireen E. Matloubieh1; Lee Sacks3; Pankaj Patel3; Kevin G. Volpp1; Ezekiel J. Emanuel1. 1University of Pennsylvania, Philadelphia, PA; 2New York University, New York, NY; 3Advocate Physician Partners, Chicago, IL. (Control ID #2706589)

BACKGROUND: Pay-for-performance (P4P) is the most pervasive value-based payment scheme despite little evidence of positive effects on patient outcomes. Yet, Medicare’s upcoming Merit-based Incentive Payment System (MIPS) will expand traditional P4P even further. In this study, we examine whether two principles from behavioral economics, loss aversion and social pressure, can improve the effectiveness of physician financial incentives.

METHODS: We conducted a cluster randomized 3-arm clinical trial among 43 primary care and specialist practices at Advocate Physician Partners (APP), Chicago, IL. The active phase of the trial was January 1, 2016 to December 31, 2016, with follow-up through March 31, 2017. The control arm (Arm 1) was compromised of 21 physicians enrolled in the existing physician incentive payment design with quarterly performance feedback and a one-time bonus check (to be distributed April 1, 2017). Arm 2 tested the use of loss aversion among 26 physicians, in which incentive accounts were pre-funded with the total incentive dollars available over the 2016 program year. Arm 3 evaluated the performance of 24 physicians in an enhanced social pressure design that increased the group portion of the bonus to 50 percent. We performed an intention-to-treat analysis of treatment assignment on quality measure performance, including wellness (e.g. body mass index (BMI) screening), chronic care (e.g., appropriate beta blocker prescription), and mental health (e.g. depression screening) measures. Pairwise comparisons were evaluated with the Tukey range test. NOTE: Because the trial is still in follow-up period we present blinded results below - at the SGIM conference we will present final unblinded results.

RESULTS: There were 16,375 attributed patients with conditions related to APP P4P quality measures in the 2016 program year. Significant differences in performance were observed between the trial arms for 7 measures across 3 categories, though this tended to be true only in the greater than 65 age population. For example, for the ‘beta blockers for chronic heart failure’ measure in the chronic care category, we observed scores of 73% in Arm Y, 67% in Arm X, and 68% in Arm Z, differences of 5 and 6% (P = 0.033), however there was no significance for patients less than 65 years (P = 0.75). Similarly, for blood pressure measurement among cardiovascular disease patients over 65, we observed a score of 90% in Arm Y, 59% in Arm X, and 49% in Arm Z, differences of 31 and 41 percentage points (P = 0.028), while there was no significance for patients less than 65 years (P = 0.66). We observed similar trends for other measures such as ‘depression screening and follow-up’ and ‘BMI assessment.’

CONCLUSIONS: The trial results suggest that the intervention in Arm Y may be a useful technique to make physician P4P incentives more effective. The performance improvements for patients over 65 may in part be explained by applicability of the quality measures.

A SAFE SPACE FOR SOLIDARITY, CONVERSATION, AND FINDING MEANING IN MEDICINE: REFLECTIVE WRITING WORKSHOPS LED BY NEAR-PEERS DURING THIRD-YEAR CLERKSHIPS Lorenzo R. Sewanan 2; Kayleigh Herrick-Reynolds2; Priscilla Wang1; Andi Shahu1; Daniel Zheng2; John Encandela2; Anna Reisman1. 1Yale School of Medicine, New Haven, CT; 2Yale University, New Haven, CT. (Control ID #2704999)

BACKGROUND: Third-year medical students (MS3s) routinely experience mentally and emotionally challenging situations. They often lack the opportunity to debrief about these experiences due to time constraints, clinical hierarchy, and peer isolation during clerkships. Medical schools are increasingly incorporating reflective writing into curricula. A growing body of literature demonstrates the positive impact of reflective writing for medical students, including increased empathy and improved learning. These sessions are usually conducted by faculty facilitators, whose presence may prevent truly open discussion. We proposed that reflective writing workshops led by near-peers built into the third-year curriculum would provide a more effective model for honest reflection.

METHODS: Students and facilitators were medical students at Yale School of Medicine. The study occurred over a one-year period as MS3s rotated through four core clerkships. Facilitators were trained upperclassmen volunteers. MS3s were split into groups of 10–15 and required to attend a workshop at the end of each rotation. Rules of engagement were reviewed (confidentiality, title IX, commitment to active, judgment-free participation and listening). Workshops followed a structured but flexible series of questions progressing from warm-up questions to more challenging prompts. Evaluations were voluntary and anonymous, including (1) a paper survey of open-ended questions at the end of the session and (2) a longer, online end-of-year survey. We used a randomized iterative consensus process within a phenomenological framework to analyze the qualitative data. Major themes arising from this formal categorization process were used, along with representative quotes.

RESULTS: 266 session questionnaires were collected across 28 sessions, and 82/102 possible participants filled out the end-of-year survey. Most respondents (62%) indicated that they would attend the workshops even if not required. Our qualitative analysis revealed that the workshop provided a safe space for reflection and sharing. Many students were surprised to learn how willing peers were to share experiences. Students valued having protected time and found the prompt-based format beneficial. Common themes included challenges of patient care, medical hierarchy, interprofessional tension, and burnout. Students found that the format of writing and sharing provided solidarity within the class. Many stated that the workshops helped them process experiences not discussed before and encouraged them to focus more on emotions in clinical care. Most indicated that they would rather share negative experiences in a group led by near-peers than by faculty. Many described a desire to write and engage in active reflection in the future as a result of workshop participation.

CONCLUSIONS: The near-peer reflective writing workshops provided protected space to process difficult clerkship experiences, increased solidarity and support within the class, and encouraged MS3s to actively reflect.

A SIMULATED NIGHT ON CALL (NOC): ASSESSING THE ENTRUSTMENT OF NEAR GRADUATING MEDICAL STUDENTS FROM MULTIPLE PERSPECTIVES. Adina Kalet 1, 1; Tavinder Ark2; Kinga L. Eliasz1, 1; Mike Nick1, 1; Grace Ng1, 1; Demian Szyld3, 1; Sondra Zabar1, 1; Martin V. Pusic1, 1; Thomas S. Riles1, 1. 1New York University School of Medicine, New York, NY; 2University of British Columbia, School of Population Health, Vancouver, BC, Canada; 3Harvard Medical School, Boston, MA. (Control ID #2705395)

BACKGROUND: The AAMC has identified 13 Entrustable Professional Activities (EPAs) that all entering residents should be expected to perform on day 1 of residency without direct supervision regardless of specialty choice. We developed an immersive, Night on call (NOC) simulation to understand the measure of entrustment of all 13 Core EPAs from the perspective of patients, nurses, attendings, and peers.

METHODS: NOC is a 4-hour simulation, during which a medical student rotates through a series of authentic clinical coverage scenarios including: 4 standardized patient (SP) cases with varying degrees of complexity, each of which require first answering a call from a standardized nurse, (SN), then evaluating a SP with the SN in the room, making immediate management decisions and writing a coverage note; a phone call to an attending (Attn, an experienced clinician) to orally present (OP), and discuss the case, formulation of a clinical question and finding a best answer using digital library resources (EBM), a test of ability to recognize a pre-entrustable peer, and a handoff of 4 cases to a peer (HOff, portrayed by an senior medical student). Competency assessments were based on validated tools where available. Each rater provided an entrustment judgment. This included 9 raters providing a total of 16 entrustment judgments: 4 SPs and 3 SNs (1 rating competency and 1 rating communication each), 1 Attn based on OP, 1 peer rating based on the HOff (1 item each). Raters were trained in both case portrayal and rating reliability. This study is IRB approved. After exploring the relationships among competency measures and entrustment judgements, to test the hypothesis that NOC measures trustworthiness of our near graduates, we conducted a one-factor (entrustment) confirmatory factor analysis (CFA) with the 16-entrustment items allowing the ratings from the same raters and between raters on the same case to correlate. The CFA was conducted with a means and variance adjusted weighted-least squares estimation (WLSMV) to take the ordinal distributions of the entrustment items into account.

RESULTS: 73 medical students (39 women; Age 26.5 (+2.6) years) completed NOC. The one-factor CFA model fit the data (χ2 = 155.27, df = 112, p < .001, CFI = 0.97, TLI = 0.97, RMSEA = 0.07, p > 0.05). All but 2 of the 16 factor loadings were greater than 0.3, (Attn factor loading = 0.23 and the SP ratings from the first clinical case of NOC sequence (0.21)).

CONCLUSIONS: A single-factor model with 16 measures fit the entrustment framework within an ecologically valid simulated workplace suggesting that an individual student’s clinical trustworthiness is measurable across discrete work activities. This work provides an assessment framework for the educational handoff from medical school to residency to ensure quality of care and patient safety.

A SYSTEMATIC REVIEW OF THE EXPERIENCES OF HEALTH CARE STAFF ON USING ELECTRONIC SYSTEMS FOR THE FOLLOW-UP OF PATIENT TEST RESULTS. Abdulaziz A. Mohammed 1; Alaa Bagalagel2; Ahmad Noor2; Andrew K. Husband1; Simon P. Forrest1; Sarah P. Slight1. 1Durham University, Teesside, United Kingdom; 2King Abdulaziz University, Jeddah, Saudi Arabia. (Control ID #2706937)

BACKGROUND: Health Information Technology has the potential to improve the transfer of patient test results; [1] however, abnormal test results are still missed even with the use of electronic systems. The aim of this review was to explore health care staff experiences of using electronic systems for the follow-up of patient test results.

METHODS: The review followed the PRISMA guidelines and was registered on PROSPERO (CRD42016042944). Four databases were searched from Jan 2005-July 2016: Embase, Medline, CINAHL and PsycINFO. Primary articles were included if they focused on (1) any type of Electronic Health Records (EHRs) used for the follow-up of test results, (2) healthcare staff opinions and views of system(s), (3) any type of test result(s), (4) all disease states in both adult and paediatric populations, and (5) any health care setting. Commentaries, editorials, letters and any studies not available in English language were excluded. Three reviewers independently screened the titles, abstracts, and full texts of articles for inclusion, and also assessed the quality of those included using the Critical Appraisal Skills Programme (CASP) tool.

RESULTS: Our search returned 1,178 publications, 79 of which were duplicates. One thousand and eighty three were eliminated at the title (600), abstract (478) and full text stages (5). Fifteen articles and one abstract met our inclusion criteria. Six main themes were identified: (1) systems’ design, (2) workload, (3) systems’ infrastructure, (4) communication, (5) training, and (6) feedback. Users highlighted a number of benefits of using electronic systems (e.g. ability to access test results quickly). However, some reported having a lack of knowledge about system features and receiving insufficient training. [2] A proper training could highlight existing features specific to alert managements, which might be helpful in the follow-up of patient test results. On the other hand, lack of knowledge and training could increase follow-up time using electronic systems and contribute to a heavy workload. [2]

CONCLUSIONS: Users need to receive sufficient training on the use of electronic systems. More research is needed to explore how electronic systems could be further improved. This review only focused on healthcare staff experiences and did not cover patients’ experiences. References: Singh H, Arora H, Vij M, Rao R, Khan M, Petersen L. Communication outcomes of critical imaging results in a computerized notification system. J Am Med Inform Assoc. 2007;14:459–466. doi: 10.1197/jamia.M2280. Hysong SJ, Sawhney MK, Wilson L, et al. Understanding the management of electronic test result notifications in the outpatient setting. BMC Medical Informatics and Decision Making. 2011;11:22. doi:10.1186/1472-6947-11-

A SYSTEMATIC REVIEW ON THE IMPACT OF SURGICAL INCIDENTS ON MEDICAL AND NON-MEDICAL OPERATING STAFF. Naresh Serou 1, 2; Lauren M. Sahota1; Andrew K. Husband1; Simon P. Forrest1; Krishna Moorthy2, 4; Sarah P. Slight1, 3. 1Durham University, Teesside, United Kingdom; 2Imperial College Healthcare NHS trust, London, United Kingdom; 3Newcastle upon Tyne Hospitals NHS Foundation Trust, Newcastle upon Tyne, United Kingdom; 4Imperial College London, London, United Kingdom. (Control ID #2700381)

BACKGROUND: The occurrence of surgical incidents can have a devastating effect, not only on the patient and their families, but also on the health professionals involved.

METHODS: The aim of this review was to explore the effect of surgical incidents on operating staff and what different coping strategies they used. This review followed the PRISMA-P reporting guidelines and was registered with the PROSPERO database (number: 42016042175). Studies were eligible for inclusion if they were primary research or reviews that focused on the effect of surgical errors on operating theatre staff in any health care setting. We were interested in articles that included data about the professional and personal impact of a surgical incident on staff, and the different coping mechanisms used by staff. We conducted the search in June 2016: MEDLINE in Process (Ovid) Jan 1950-Present, EMBASE (Ovid) Jan 1974-Present, CINALH 1982-Present, and PsycINFO 1967-Present. A customised data extraction form was used to capture pertinent information from included studies and the Critical Appraisal Skills Programme (CASP) tool to appraise their quality.

RESULTS: A total of 3,918 articles were identified, with 667 duplicate articles removed, and 3,251 excluded at the title (540), abstract (2,386) and full text (304) stages. Twenty-one articles (19 full text articles, 2 review articles) were included in the final review, eight of which focused on the impact of a surgical incidents on surgeons and anaesthetists. Only two articles involved theatre nurses and operating theatre team. Five key themes emerged: 1) the emotional impact on health professionals, 2) organisation culture and support, 3) individual coping strategies, 4) learning from surgical complications, 5) recommended changes to practice. Most articles reported that health professionals used different coping strategies in the aftermath of a surgical incident. Some surgeons discussed the event with their peers (more likely amongst senior surgeons), while others chose to reflect on the incident privately. Irrespective of different coping mechanisms used, informal open discussion with peers was viewed as helpful to regain self-confidence and positive thinking. Studies recommended establishing a mechanism by which deaths, serious incidents and never events were reviewed at an organisational level.

CONCLUSIONS: Health professionals can suffer emotional distress and use different coping strategies in the aftermath of a surgical incident. This review excluded studies that focused solely on the effect of malpractice claims on health professionals; these studies may have provided further insights on the emotional effects of incidents. Future research is needed to explore whether the impact of surgical complications differs amongst the wider operating theatre team.

ABNORMAL LFTS: FINDING THE NEEDLE IN THE…STACK OF NEEDLES? Andrew Schreiner 2; William P. Moran2; Jingwen Zhang1; Justin Marsden2; Patrick D. Mauldin2. 1MUSC, Charleston, SC; 2Medical University of South Carolina, Charleston, SC. (Control ID #2707614)

BACKGROUND: Primary care physicians commonly encounter liver function tests in patients. The prevalence of abnormal liver function tests (LFTs) ranges from 9 to 21.7% in the general population, and the cause often remains unknown. Abnormal LFTs may indicate a wide range of pathologic processes, including cirrhosis, a leading cause of liver transplantation and death in patients with liver disease. The goal of this research is to better understand the burden of abnormal liver function tests in the primary care setting.

METHODS: We performed a retrospective analysis of patients with abnormal liver function testing in the primary care setting. We searched the electronic health record of an academic, internal medicine, patient-centered medical home (PCMH) at the Medical University of South Carolina (MUSC) over an 8 year period with patients seen by the clinic at least once, and with at least 1 abnormal liver function test element. Abnormal liver function tests were stratified by the degree to which they were abnormal and the pattern of abnormality.

RESULTS: Since 2008, the clinic provided care to 30,891 unique patients, of whom 26,907 underwent liver function testing, and 11,945 possessed at least one abnormal result. Total bilirubin values 3 times the ULN occurred in 722 patients, AST 3 times the ULN occurred in 2,177 patients, ALT 3 times the ULN in 1,399 patients, and alkaline phosphatase 3 times the ULN occurred in 362 patients. The proportion of these patients retested with similar results is shown in Table 1 (attached). Cholestatic patterns occurred in 2.9% of the test results, mixed patterns in 17.6% of the results, and hepatocellular injury patterns in 79.6% of the results.

CONCLUSIONS: Patients with abnormal liver function tests present to primary care physicians with remarkable frequency. Though most tests results fall into patterns of mild abnormality (<2 times the ULN) with a hepatocellular injury pattern, a wide array of abnormalities and patterns do appear.

figure f

ACCESS TO A SCALE AND SELF-WEIGHING HABITS AMONG PUBLIC HOUSING RESIDENTS Carolyn Bramante 1; Jeanne M. Clark2; Kimberly Gudzune2. 1Johns Hopkins, Baltimore, MD; 2Johns Hopkins University, Baltimore, MD. (Control ID #2702362)

BACKGROUND: Regular self-weighing is an effective weight management tool. This activity is contingent on an easily accessible scale; however, few studies examine how many people actually have regular access to a scale. Financial constraints might preclude the purchase of a scale, particularly among low-income populations that are disproportionately affected by obesity. Our objective was to determine the proportion of low-income public housing residents who have access to a scale and their self-weighing habits, as well as determine individual characteristics associated with scale access.

METHODS: We conducted a cross-sectional survey of randomly selected households in public housing developments in Baltimore, MD. We asked participants if they had “regular access to a scale where they can weigh themselves,” and to describe their self-weighing habits (“I weigh myself daily” with responses dichotomized as ‘never or hardly ever/no scale access’ versus ‘some/about half/much of the time/always’). We then used t-tests or Chi2 tests, as appropriate, to examine the association of scale access with the following characteristics: age, gender, marital status, education, employment, food insecurity, smoking status, physical activity, diet, BMI, and health status.

RESULTS: Overall, 266 head of households participated (48% response rate). Mean age was 45 years, 86% were women, 95% were black, and 54% were obese. Only 32% reported having access to a scale; however, 78% of people who had access reported engaging in at least some self-weighing. Table 1 compares characteristics between those with and without access to a scale. Residents who had access to a scale were significantly older (p = 0.03) and significantly less likely to be disabled/unemployed (p = 0.01) or food insecure (p < 0.01). We found no other statistically significant associations with any other factors.

CONCLUSIONS: While only one-third of public housing residents have access to a scale, those who do have access self-weigh with some regularity. Economic status appears to be a factor influencing scale access in this population, as we found a significant inverse association with markers of low economic status. Therefore, addressing economic barriers to scale ownership may be a reasonable target for future weight management interventions in low-income populations.

figure g

ACCESS TO MEDICAL CARE FOR AMERICANS WITH EXCHANGE-BASED INSURANCE: EARLY EVIDENCE FROM THE NATIONAL HEALTH INTERVIEW SURVEY Ilana B. Richman 3; Douglas K. Owens1; Jay Bhattacharya4; Steven Asch2. 1VA Palo Alto/Stanford University, Stanford, CA; 2VA/Stanford, Menlo Park, CA; 3Yale University School of Medicine, New Haven, CT; 4Stanford University School of Medicine, Stanford, CA. (Control ID #2706424)

BACKGROUND: In 2016, an approximate 12.7 million Americans purchased health insurance through exchanges established by the Affordable Care Act. Plans sold on the exchanges, though, have been criticized for offering excessively narrow networks, which may limit access to care. The goal of this study was to evaluate access to care for those with exchange-based coverage compared to those with employer-sponsored insurance.

METHODS: We used data from the National Health Interview Survey, an annual, nationally representative survey, during 2014 and 2015. We included adults ages 18–64 who had either an employer-sponsored plan or an individual plan purchased on an exchange. In evaluating access to care, we used logistic regression to adjust for age, sex, race/ethnicity, poverty status, employment status, health status, history of uninsurance, and survey year. Results are expressed as predicted probabilities.

RESULTS: Our analysis included 15,795 adults, of whom 2,023 had purchased insurance on an exchange. Exchange participants were older (43.6 vs 41.3 years, p < 0.001) and were less likely to be white (74% vs 84%, p < 0.001) than those with employer sponsored coverage. Those with exchange-based plans were less likely to be working (67% vs 85%, p < 0.001), were more likely to have a household income below 400% of the federal poverty level (74% vs 40%, p < 0.001), and were more likely to be in poor or fair health (10% vs 5.6%, p < 0.001). Those who purchased insurance on the exchanges were less likely to identify a usual source of care (79% vs 84%, p < 0.001) and were more likely to have had trouble finding a primary care physician (6.6% vs 3.2%, p < 0.001) than those with employer-sponsored coverage. Those with exchange-based coverage were also more likely to have waited for appointments (7.2% vs 5.1%, p = 0.007) and were more likely to have had trouble affording specialty care (6.8% vs 3.9%, p < 0.001). In spite of these differences, those with exchange-based coverage were not less likely to receive treatment for chronic conditions including hypertension (82% vs 83%, p = 0.70) and hypercholesterolemia (72% vs 76%, p = 0.32). Lastly, despite barriers to care, those with exchange-based coverage were more likely to report that their coverage was an improvement from the year prior (31% vs 11% p < 0.001), though more also reported that their coverage was worse (18% vs 11%, p < 0.001).

CONCLUSIONS: Although exchange-insured patients faced some barriers to care compared to those with employer-sponsored insurance, differences were small and compared favorably to previously published rates among the uninsured. Those with exchange-based coverage also received care for chronic conditions at rates similar to those with employer-sponsored insurance. Many with exchange-based insurance viewed their coverage as an improvement, likely reflecting a prior history of uninsurance or underinsurance. Tracking access to care will be even more crucial as the federal policy environment changes.

ACCESSIBILITY OF OUTPATIENT BEHAVIORAL HEALTHCARE IN ONE METROPOLITAN AREA Hannah Spellman2; Daniel J. Coletti1; Leslie Rosenberg1; Lauren Block 1, 2. 1Northwell Health, Lake Success, NY; 2Hofstra Northwell School of Medicine, Hempstead, NY. (Control ID #2706395)

BACKGROUND: The USPSTF recommends screening for depression in primary care, provided that adequate systems are in place to ensure timely and “appropriate” follow-up. It is unclear whether adequate access exists for patients with non-emergent behavioral health needs. The objective of this project was to assess availability of outpatient behavioral healthcare within two densely populated counties in the New York metropolitan area.

METHODS: The cross-sectional survey included all community-based, hospital-affiliated, and university-affiliated outpatient behavioral health centers located in Nassau and Queens Counties in NY. Facilities were compiled from the Northwell General Internal Medicine Directory of Behavioral Health Services and service directories from the local county Department of Health. Calls were made by a single caller with a predetermined script weekdays from June through July 2016, between 10 am and 4 pm. Data collected were summarized and analyzed with descriptive statistics.

RESULTS: The inclusion criteria were met by 40 facilities. Contact was made with 93% of facilities within three calls, but only 50% of facilities on the first call. Over half of the facilities (55%) had intake appointments available within the week; however, only 20% had treatment appointments within the week, and only 3% offered an appointment with a prescribing provider within a week. Within one month, 53% of facilities had psychotherapy treatment appointments and 32% also had medication evaluation appointments. Two-thirds (65%) of facilities had waits of more than a month to see a prescribing provider. The availability of additional treatment services varied widely, with 68% offering a provider who spoke a language other than English. 86% of facilities accepted an assortment of public and private insurance plans, and 76% of facilities had an income-based sliding payment scale for uninsured patients.

CONCLUSIONS: The availability of intake appointments masks longer wait-times for medication management. Shorter wait times to intake may help patients with urgent needs find adequate treatment, and longer wait times to medication management may be suitable for patients with mild behavioral health problems. However, behavioral healthcare may be inadequate for patients with moderate diagnoses who will not be treated urgently, but who should not wait the length of time to an appointment. More research is needed to assess the true adequacy of behavioral healthcare for adults in our region and nationwide. Integrating additional behavioral healthcare services into primary care settings might improve access.

ACCULTURATION IS ASSOCIATED WITH DIETARY PATTERNS IN SOUTH ASIANS IN AMERICA Meghana D. Gadgil 2; Namratha R. Kandula1; Alka M. Kanaya2. 1Northwestern University, Chicago, IL; 2University of California, San Francisco, San Francisco, CA. (Control ID #2699896)

BACKGROUND: Cultural beliefs and practices may affect dietary patterns in South Asians. We aimed to determine whether the strength of cultural beliefs and practices affects adherence to a particular dietary pattern in South Asians in the United States.

METHODS: We conducted a cross-sectional analysis including South Asians aged 40–84 years without known cardiovascular disease who were enrolled in a community-based cohort called Mediators of Atherosclerosis in South Asians Living in America (MASALA). A validated food frequency questionnaire, a multi-dimensional measure of traditional cultural beliefs and assessment of cultural practices were collected at baseline. We used principal component analysis with varimax rotation to determine dietary patterns, and multivariable logistic regression models for associations with cultural beliefs and practices.

RESULTS: Of the 892 participants included in our analysis, 47% were women. We identified 3 major dietary patterns, which we termed “Animal Protein,” “Fried snacks, Sweets, High-fat dairy” and “Fruits, Vegetables, Nuts, Legumes.” A weaker traditional cultural beliefs score was associated with higher odds of adherence to the Animal Protein dietary pattern [OR(95%CI) 1.66 (1.40, 1.97)] and lower odds of adherence to the Fried snacks, Sweets, High-fat dairy pattern [0.76 (0.64, 0.90)]. Participants who do not fast [3.34 (1.27, 8.83)], eat South Asian food in restaurants [5.56 (1.09, 28.29)], shop at South Asian grocery stores {11.91(2.26, 62.71)], eat outside of the home [0.26 (0.16, 0.42)] or who have fewer South Asian friends [5.50 {1.13, 26.63)] had higher odds of consuming the “Animal Protein” dietary pattern (all p < 0.05). Those who do not fast [0.25 (0.10, 0.63)] or shop at South Asian grocery stores [0.15 (0.04 .53)] were less likely to consume the “Fried Snack, Sweets and High-fat dairy” dietary pattern (all p < 0.05). There was no significant association of strength of cultural beliefs or practices with the Fruits, Vegetables, Nuts and Legumes pattern.

CONCLUSIONS: South Asians in the United States with stronger traditional cultural beliefs and practices were more likely to consume the Fried Snacks, Sweets and High-fat dairy dietary pattern, and less likely to consume the Animal Protein pattern. Prevention programs should consider how to modify these dietary patterns as part of comprehensive risk reduction in South Asians.

ACUTE AMBULATORY ASSESSMENT TO AVOID ADMISSION (5ALPHA): A QUALITY IMPROVEMENT STUDY Tara O’Brien 2, 1; Robert Wu2, 3; Ian Stanaitis1; Geetha Mukerji2, 1; Minnie Rai1; Sam Sabbah2, 3. 1Women’s College Hospital, Toronto, ON, Canada; 2University of Toronto, Toronto, ON, Canada; 3University Health Network, Toronto, ON, Canada. (Control ID #2700001)

BACKGROUND: In Canada, changing patient demographics caused by aging and increased patient multimorbidity has contributed to an increased number of emergency department (ED) visits and hospitalizations. The University Health Network (UHN) in Toronto, Ontario has experienced a 6% yearly increase in ED visits and general internal medicine (GIM) hospitalizations. The goal of this quality improvement (QI) initiative is to prevent GIM referrals and inpatient admissions at UHN by providing rapid follow up to patients in a short stay medical unit staffed by a GIM physician, the Acute Ambulatory Care Unit (AACU) at Women’s College Hospital (WCH).

METHODS: This was a pre-post design study, using the Model for Improvement QI framework. For UHN ED patients who required an urgent GIM consultation but were stable for discharge, we provided a follow up in the AACU within 24–48 hours of discharge. Consent was obtained from patients and data was collected prospectively. Process measures were collected including patient demographics, need for inpatient admission, and impact of AACU as perceived by the ED physician. The primary outcome was the monthly rates of GIM consults at the Toronto General Hospital (TGH) and hospitalizations at UHN. Qualitative data was collected using a patient experience survey. The investigators met monthly to review referrals and patient outcomes and PDSA cycles were used to improve the model of care delivery over time.

RESULTS: A total of 788 patients were seen in the AACU over a one-year period. The mean age of patients was 56, 52% females and 46% had 3 or more chronic medical conditions. The three most common reasons for referral were anemia (7%), hypertension (6%) and abdominal pain (6%). The rate of TGH GIM consults remained stable at a mean of 570 consults per month despite a 3.5% increase in ED visits. Similarly the number of inpatient admissions at UHN remained unchanged with a mean of 975 per month. While in the AACU, 36% of patients had medical imaging, 27% had subspecialty consultation and 24% had non-invasive cardiac testing. Additional follow up in the AACU was required in 58% of patients and 2.5% required transfer back to UHN for admission. The goal of the referral as perceived by the ED physician was to avoid admission, avoid GIM consult and provide rapid follow up in 27 37 and 63% of cases respectively. Results from the patient experience survey demonstrated that 77% of patients were extremely satisfied with their care in the AACU and 92% would recommend this experience to other patients.

CONCLUSIONS: This QI initiative successfully provided safe rapid follow up to GIM patients discharged from the ED in the AACU, resulting in perceived avoidance of GIM consult or admission and high patient satisfaction. Our cross-institutional model of care likely contributed to the flattening of inpatient admissions at UHN which had been steadily increasing. This model of care is a potential scalable solution to address the problem of hospital overcrowding.

ACUTE MYOCARDIAL INFARCTION READMISSION RISK PREDICTION MODELS: A SYSTEMATIC REVIEW OF MODEL PERFORMANCE Lauren N. Smith; Anil N. Makam; Douglas Darden; Ethan Halm; Oanh K. Nguyen. UT Southwestern Medical Center, Dallas, TX. (Control ID #2700088)

BACKGROUND: Hospitals are subject to federal financial penalties for excessive 30-day hospital readmissions for acute myocardial infarction (AMI). Prospectively identifying AMI patients at high risk for readmission could help prevent 30-day readmissions by enabling targeted interventions. However, the effectiveness of AMI readmission risk prediction models is unknown.

METHODS: We systematically searched Ovid MEDLINE, Embase, the Cochrane Library, and CINAHL databases from inception through March 2016 for studies of risk prediction models for 30-day hospital readmission among adults with AMI. Two independent reviewers abstracted data and assessed the risk of bias.

RESULTS: We identified 4532 unique titles; after full-text review, we identified 8 studies describing 9 unique risk prediction models across both academic and community hospitals. Seven studies assessed models predicting all-cause 30-day readmissions; of these, we included only the 5 studies that assessed model performance in a validation cohort in our final analysis (Table). Three studies only included patients ≥65 years old. Observed readmission rates ranged from 13.0–19.7%. Four models used administrative data, 1 used electronic health record data, and 1 used ‘hospital data’ not otherwise specified. Models included between 7–37 predictors; comorbidities, demographic characteristics, and utilization metrics were the most frequently included predictors. Most models (n = 4), including the Centers for Medicare and Medicaid Services (CMS) AMI administrative model had modest discrimination (C-statistic range 0.62-0.66). A modified version of the CMS AMI model enriched with idiosyncratic combinations of SES variables had the best discrimination (C-statistic 0.76 for both); but this was assessed among Medicare fee-for-service beneficiaries in a single state, potentially limiting generalizability. Predicted readmission rates ranged from 8% among the lowest risk individuals to 35.7% among the highest risk individuals, though these data were unavailable for 3 of 5 studies.

CONCLUSIONS: We found a limited number of validated AMI readmission risk prediction models with modest predictive ability. Future modeling strategies should assess the potential impact of including more clinically detailed data on improving model performance.

figure h

ADAPTING A HEALTH SYSTEM INTERVENTION FROM KAISER PERMANENTE TO IMPROVE HYPERTENSION MANAGEMENT AND CONTROL IN A LARGE NETWORK OF SAFETY NET CLINICS Valy Fontil 1, 2; Reena Gupta1; Kirsten Bibbins-Domingo1, 2. 1University of California San Francisco, San Francisco, CA; 2Zuckerberg San Francisco General Hospital, San Francisco, CA. (Control ID #2703857)

BACKGROUND: Nearly half of Americans have uncontrolled blood pressure (BP). Health system interventions that include evidence-based hypertension (HTN) treatment algorithms have successfully improved BP control in high-functioning integrated health systems such as Kaiser Permanente (KP), but it is unclear whether these interventions can work in safety-net clinics that disproportionately care for populations at highest risk for HTN. National efforts to improve population control and to reduce disparities in HTN will require adaptation of proven interventions to safety-net settings.

METHODS: We used the precede-proceed framework to adapt KP’s treatment algorithm within a multi-component health system intervention for HTM management, implement it in a network of 12 safety-net clinics in the San Francisco Health Network, and evaluate its impact on BP control. We evaluated the program’s effectiveness in improving the BP control rate (proportion of patients whose BP is at the target goal) and use of recommended medications such as fixed-dose combination drugs (FDC) at 24 months post-intervention. We used linear regression analyses to assess for post-intervention trends in BP control and use of FDCs. We assessed BP control rates by race and examined trends in ACE-inhibitor monotherapy in African Americans, which is associated with lower efficacy and delays in achieving BP control.

RESULTS: At 24 months post-intervention, there was a significant trend in increasing rates of BP control (65 to 72%, P < 0.01). Improved BP control occurred in blacks (59 to 67%, P < 0.01) Whites (65 to 73%, P < 0.01), and Latinos (63 to 73%, P < 0.01). Improved BP control among Asians did not reach statistical significance (71 to 75%). Use of fixed-dose combination drugs increased from 8 to 15% (P < 0.01). ACE-inhibitor monotherapy among blacks trended downward but remained high (33 to 14% P < 0.01).

CONCLUSIONS: Adaptation of a HTN treatment intensification algorithm from Kaiser Permanente led to improved BP control at a large network of safety-net clinics. Our findings can inform best practices for health system interventions to improve BP control at safety-net clinics which must play a pivotal role in achieving improved BP control and reducing HTN disparities.

figure i

ADDICTION TREATMENT ORGANIZATIONS’ ADOPTION OF ELECTRONIC HEALTH RECORDS: A NATIONAL SURVEY AT THE ONSET OF HEALTH REFORM Peter D. Friedmann 1, 2; Donna Wilson1, 2; Christina Andrews4; Harold Pollack3; Thomas D’Aunno5; Keith Humphreys6. 1UMass Medical School - Baystate, Springfield, MA; 2Baystate Health, Springfield, MA; 3University of Chicago, Chicago, IL; 4University of South Carolina, Columbia, SC; 5New York University, New York, NY; 6Stanford University, Palo Alto, CA. (Control ID #2705630)

BACKGROUND: Electronic Health Record (EHRs) are essential technology for health care organizations to participate in health reform innovations. Addiction treatment organizations were excluded from federal initiatives to promote adoption of EHRs; their use of EHRs is uncertain.

METHODS: This analysis uses data from a 2014 survey of a national random sample of addiction treatment organizations. Directors and supervisors responded to detailed questionnaires about program structure and processes, including adoption of EHRs.

RESULTS: Of 692 responding programs, directors of 390 (weighted 55%) reported having an EHR and another 160 (weighted 24%) planned to adopt one within 2 years. Only 157 (weighted 22%) participated in Health Information Exchanges. In multinomial logistic models, correlates of having an EHR were state funding for EHR adoption (OR 3.5, 95% CI 1.2–10.6); private not-for-profit (OR 3.6, 95% CI 1.4–9.2) or public (OR 4.6, CI 1.1–19.5) compared to for-profit ownership; having accreditations (e.g. JCAHO, CARF) (OR 2.96, CI 1.4–6.5); being a new program (OR 8.3, CI 1.4–50.9); having an average caseload greater than 40 clients (versus 10; OR 10.2, CI 2.4–43.5); and reporting more perceived barriers to an EHR (OR 0.81, CI 0.72–0.92). Planning for EHR adoption within 2 years was correlated with increasing perceived competition (OR 2.5, 95% CI 1.1–5.6). State implementation of the Affordable Care Act did not influence EHR adoption.

CONCLUSIONS: Only 55% of addiction treatment programs had an EHR at the advent of health reform. Another 24% planned to adopt with 2 years. State funding, non-profit or public ownership, larger caseload, and fewer perceived barriers appeared to be related to EHR adoption. Fewer than one-quarter participate in Health Information Exchanges. These technological limitations may constrain addiction treatment organizations ability to participate in innovations and modernization resulting from health reform.

ADDRESSING COMPLEX PATIENTS’ PSYCHOSOCIAL PRIORITIES DURING TIME-LIMITED PRIMARY CARE VISITS Eilann C. Santo 1; Michelle T. Vo2; Richard W. Grant2. 1Kaiser Permanente Northern California, San Francisco, CA; 2Kaiser Permanente Northern California, Oakland, CA. (Control ID #2706484)

BACKGROUND: Providing patient-centered primary care during time-limited visits is challenging. Complex patients in primary care often present with both medical and psychosocial concerns. Given the important impact of psychosocial issues on medical management, addressing patients’ psychosocial concerns is essential to providing effective, patient-centered clinical care. As part of an ongoing multi-site, randomized trial examining patient priorities, we had the unique opportunity to examine how patient self-identified psychosocial vs medical visit priorities were addressed by their primary care physicians.

METHODS: Patients identified their top 1 or 2 visit priorities in the waiting room prior to a primary care visit as part of the Aligning Patient and Provider Priorities clinical trial (ClinicalTrials.Gov NCT02707146). We categorized patient visit priorities as either psychosocial (e.g. depression, anxiety, substance use, stress, or personal safety) or medically related. We performed a structured chart review of visit progress notes, after-visit summaries, and follow up secure email messages to determine whether the provider addressed the patient’s visit priority. Evidence that the priority was addressed was based on documentation in progress note, change to problem list, prescription of diagnostic or therapeutic interventions, referral to specialist, or communication about the problem through after visit summary or secure messages. We used Fisher’s exact tests to compare provider action for psychosocial vs. medically related patient priorities.

RESULTS: Patients in our cohort (n = 103) had a mean age of 62 (SD 12) years; 69.9% were female, 43.0% were African American, and 53.4% had a history of a mental health disorder. Nearly one quarter of patients (18.6%) reported a psychosocial concern as one of their top two priorities for their primary care visit. Overall, patients listed 137 unique priorities (1.3/patient). Psychosocial priorities were less likely to be addressed during the visit compared to medically related priorities (62.5% vs 91.2% p = 0.001). Similarly, psychosocial priorities were less likely to receive clinical action or follow up (50.0% vs 86.7%, p < 0.0002) or post-visit information from a primary care doctor than medically related health priorities (12.5% vs 33.6%, p = 0.05).

CONCLUSIONS: Our findings suggest that patient psychosocial priorities are less likely to be addressed by primary care physicians than medically related priorities. To provide truly patient-centered primary care for patients with complex disease, physicians should be sure to acknowledge patients’ psychosocial concerns and provide the same guidance and follow up as medically related problems.

ADDRESSING THE SOCIAL NEEDS OF HYPERTENSIVE PATIENTS: THE ROLE OF PATIENT-PHYSICIAN COMMUNICATION AS A PREDICTOR OF MEDICATION ADHERENCE Antoinette Schoenthaler 1; George Knafl2; Kevin Fiscella3; Gbenga Ogedegbe1. 1NYU School of Medicine, New York, NY; 2UNC Chapel Hill, Chapel Hill, NC; 3University of Rochester School of Medicine & Dentistry and Wilmot Cancer Center, Rochester, NY. (Control ID #2701644)

BACKGROUND: Poor medication adherence is a significant public health problem in patients with hypertension. The patient-physician relationship offers an ideal opportunity to address patient non-adherence as physicians’ communication skills contribute to as much as 50% of the quality of care patients’ receive. Despite this evidence, there is no empirical data on how the informational and relational aspects of patient-physician communication affect patient’s actual medication-taking behaviors. The aim of this study was to evaluate the impact of patient-physician communication on medication adherence among a sample of primary care physicians and their Black and White hypertensive patients.

METHODS: Cohort study of 92 hypertensive patients (mean age: 60 years) and 27 primary care physicians (mean age: 36 years) in two safety-net primary care practices in New York City. Patient-physician encounters were audiotaped at baseline; medication adherence data were gathered continuously over the 3-month study with an electronic monitoring device. Audiotape analyses of patient-physician communication were coded using the Medical Interaction Process System. Six categories of communication were computed: 1) patient centeredness; 2) patient assertiveness; 3) psychosocial focus; 4) information exchange; 5) physician disclosure-promoting; and 6) physician verbal dominance. Frequencies of content codes were also calculated for the proportion of the discussion specific to: hypertension; antihypertensive medications; patient-physician social conversation; and discussions about patient’s social circumstances (i.e., patient’s living situation, relationship with spouse/partner). Adaptive statistical modeling methods based on likelihood cross validation were used to analyze the adherence data.

RESULTS: The majority of patients were Black, 58% women, and most were seeing the same physician for at least 1 year. Approximately half of physicians were White (56%), 67% women, and have been in practice for 6 years. Fifty-eight percent of patients exhibited poor adherence to prescribed antihypertensive medications. Two categories of patient-physician communication predicted poor medication adherence in the multivariate adaptive logistic regression model: fewer discussions about patients’ social circumstances (OR: 6.03, 95CI:2.15-17) and about their antihypertensive medications (OR: 5.64, 95CI:1.49-21.3).

CONCLUSIONS: The odds of poor medication adherence are nearly six times greater when patient-physician interactions do not address patients’ social circumstances or their medication regimen. These findings support the importance of adherence counseling and attending to the social determinants of health in routine care of low-income patients in ambulatory practices.

ADOLESCENTS AND ADULTS WITH DOWN SYNDROME RARELY RECEIVE AGE-, GENDER-, AND SYNDROME-SPECIFIC PREVENTIVE HEALTHCARE Kristin M. Jensen 1, 2; Elizabeth J. Campagna2; Elizabeth Juarez-Colunga2, 2; Desmond K. Runyan2; Allan V. Prochazka1. 1University of Colorado School of Medicine, Aurora, CO; 2University of Colorado, Aurora, CO. (Control ID #2705315)

BACKGROUND: Down syndrome (DS) has known preventive healthcare recommendations that require lifelong supervision. Given the increasing life expectancy of persons with DS into their 50s, they now face age- and gender-specific health conditions in addition to their DS-specific comorbidities. However, their medical care is increasingly provided by PCPs less familiar with the special health risks of DS. In this retrospective study, we evaluate adherence to DS- and age/gender-specific preventive recommendations among adolescents and adults with DS.

METHODS: Using Medicaid claims data (2006–2010) in CA, CO, MI, and PA, we defined our cohort as DS patients ≥12yo who were enrolled in Medicaid for ≥45/60mo without concurrent Medicare. Providers were considered PCPs if they billed ≥10 well exams in a year. Patients without a PCP were excluded (n = 450) for a total of 3501 patients. PCPs with ≥80% of their well-examinations billed as well-child or well-adult examinations were categorized as child- or adult-focused, respectively. The remaining PCPs were classified as mixed-focus. Age-appropriate PCPs are defined as having a focus consistent with a patient’s age: 12–17yo = child-focused, ≥26 = adult-focused; 18–25yo are in transition and can be appropriately seen by any PCP. Levels of adherence to DS- and age/gender-specific preventive screening activities were classified as good (≥80%), moderate (50– <80%), and poor (<50%). Differences in adherence were evaluated using Pearson’s chi-squared tests.

RESULTS: Our cohort was 52% male, with 21% adolescent (12-17yo), 32% transition age (18-25yo), and 47% adult (≥26 yo). 52% of the cohort (40% of adults) received primary care from child-focused PCPs, with the remaining split equally between adult- and mixed-focus PCPs. Less than half of our cohort had ≥1 well examination during the study (44% of those with an age-appropriate PCP vs. 37% of those without an age-appropriate PCP, p < 0.001). Influenza vaccination rates were similarly low (age-appropriate PCP 33%; age-inappropriate PCP 38%, p = 0.003). Most preventive healthcare recommendations had poor adherence (<50%) regardless of whether they were DS- or age/gender-specific: well examinations; vaccinations; breast, cervical, and colon cancer screenings; sleep apnea; and hearing. Lipids, vision, prostate cancer, and thyroid screenings met moderate adherence (50– < 80%). Echocardiograms to screen for acquired valve disease were highly utilized in patients with preexisting congenital heart disease (89%) but less so in those without congenital heart disease (22%, p < 0.001)

CONCLUSIONS: Adherence to age/gender- and DS-specific preventive healthcare recommendations was low during this 5 yr study in adolescents and adults with DS, regardless of patient age or PCP focus. Given the medical complexity of this otherwise vulnerable population, this represents a significant opportunity to improve primary care for persons with DS to decrease morbidity and improve overall health.

ADVANCED CARE PLANNING AT AN ACADEMIC PRIMARY CARE CLINIC Carlene P. Partow; Omar Mesina; Leslie Sheu; Brent Kobashi. University of California, San Francisco, San Francisco, CA. (Control ID #2692158)

BACKGROUND: Advanced care planning (ACP) is becoming increasingly important in a primary care setting due to the growing aging population. Despite this, recent studies show that conversations around ACP between patients and their providers are low, as low as 30% even among terminally ill cancer patients.[1] Providers and clinic leaders are turning to completion of Advanced Directive (AD) and, in some cases, Physician Orders for Life-Sustaining Treatment (POLST) forms as objectives measures of physician ACP conversations. Little is known about ACP practices in academic primary care settings where patients with complex medical issues often receive care.

METHODS: We sought to understand current local practice of AD and POLST form completion at our urban academic primary care clinic at the University of California, San Francisco, and explored physician barriers to having these conversations. We extracted patient data from our clinic’s electronic health record to determine AD and POLST completion rates. Concurrently, we administered an anonymous 13-question electronic survey to physicians, residents, medical students and nurse practitioners. The survey was designed to: 1) assess provider comfort conducting ACP conversations, 2) identify the greatest barriers to ACP, and 3) gain feedback on interventions to increase rates of AD and POLST completion. In addition, we collected qualitative data through provider focus groups.

RESULTS: As of October 7, 2016, 2,578 out of 23,216 (11%) of patients have ADs completed and only 618 (3%) have POLSTs completed. Among patients 60 years of age and older, 19% had an AD and 6% had a POLST. Fifty-nine of the 77 (77%) providers completed the survey. Thirty-four of the 59 providers (58%) did not believe AD/POLST completion rates were satisfactory and 41% were unsure. Fifty-seven (97%) providers identified time as the greatest barrier to ACP. Additional barriers included logistical uncertainty with AD form completion (29%) and concern with patient discomfort (24%). Provider focus groups confirmed that time was the most important barrier and provider discomfort was a greater factor for residents.

CONCLUSIONS: Completion of AD and POLST forms are crucial for successful ACP, but completion rates in our academic primary care setting remains low. Our provider surveys and focus groups suggest there are opportunities for improving completion rates and establishing best practices for introducing ACP to patients. As next steps, we plan to provide both patients over the age of 60 and their providers with AD and POLST forms during appointment check-in to encourage patients and providers to discuss ACP during their visit. Additionally, flyers will be placed in exam rooms to remind both patients and physicians to complete AD and POLSTs. We hope this intervention will serve as a model for improving the rate of ACP in an outpatient setting. [1] Temel, J.S., Greer, J.A., Admane, S. et al. J GEN INTERN MED (2010) 25: 150. doi:10.1007/s11606-009-1161-z

ADVERSE CHILDHOOD EXPERIENCES IN ADDICTION MEDICINE PATIENTS RECEIVING BUPRENORPHINE THERAPY Kristin Rodriguez; Scott Davis; Zach Ferguson; Krishna Suthar; Martina Jelley. University of Oklahoma School of Community Medicine, Tulsa, OK. (Control ID #2706820)

BACKGROUND: It is well established that adults with a significant history of adverse childhood experiences (ACEs) are more likely than their non ACE-affected counterparts to experience poor health outcomes. In a prior study conducted at the University of Oklahoma School of Community Medicine (OUSCM) in 2012, a team of ACE researchers gathered data on 354 patients in Internal Medicine, Family Medicine, and Community Health that indicated the prevalence of ACEs was higher (0–1 ACE, 35.5%; 2–3 ACEs, 27.3%; 4 or more ACEs, 37.2%) than the cohort examined in the original CDC ACE study (0–1 ACE, 62.1%; 2–3 ACEs, 25.4%; 4 or more ACEs, 12.5%). Data from that original ACE study also indicate that individuals with exposure to 4 or more ACEs have a 7- to 10-fold increased risk for drug abuse and addiction compared to their non ACE-affected peers. Data from the OUSCM study indicated a mean ACE score associated with any substance abuse in our patient population of 3.62. Having experienced traumatic events has been acknowledged as a significant risk factor for substance abuse. To our knowledge, there is no available literature examining the prevalence of ACEs in patients treated for substance use disorder with buprenorphine.

METHODS: We conducted a survey of current adult patients in the OU addiction medicine practice to determine the prevalence of ACEs in patients treated for substance use disorder with buprenorphine. Patients were approached to complete the self-administered survey during a regular clinic visit with their addiction medicine physician. The survey included demographics, ten ACE screening questions, and information about the patient’s substance use history.

RESULTS: Forty-eight patients completed the survey. Analysis indicated that 20.8% of these patients had experienced 0–1 ACE, 20.8% 2–3 ACEs, and 58.3% had 4 or more ACEs. The prevalence of 4 or more ACEs in this sample of patients is much higher than than our primary care population (37.2%) or CDC (12.5%) samples. The mean ACE score was 4.52. Patients in this sample reported high levels of several of the ACE categories: emotional abuse (52.1%), emotional neglect (47.9%), parental divorce (71.1%), household alcohol or drug abuse (78.7%), household mental illness (47.9%), and family member incarceration (29.2%).

CONCLUSIONS: We found that the prevalence of ACEs in a sample of patients treated for substance use disorder with buprenorphine is considerably higher than in the general population and in our primary care patients. Screening for ACEs in this population and investigating treatment approaches that include ACE awareness deserves further study, as this may improve success rates of medication assisted addiction treatment.

AGE, GENDER AND RACIAL DIFFERENCES IN THE ASSOCIATION BETWEEN TIME-VARYING DEPRESSIVE SYMPTOMS AND MORTALITY Kelsey Bryant 2; Deanna Jannat-Khah5; Yulia Khodneva6; Jessica R. Singer4; Monika M. Safford3; Nathalie Moise1. 1Columbia University Medical Center, New York, NY; 2Columbia University Medical Center, New York Prebyterian, New York, NY; 3Weill Cornell Medical College, New York, NY; 4New York Presbyterian Hospital - Columbia University Medical Center, New York, NY; 5Weill Cornell Medical School, New York, NY; 6University of Alabama, Birmingham, AL. (Control ID #2706480)

BACKGROUND: Depressive symptoms relapse and remit, and the timing of the relationship between depressive symptoms and mortality remains poorly elucidated. We recently demonstrated that time-varying depressive symptoms may be associated with all-cause and noncardiovascular (CVD) disease mortality, suggesting a proximal association. Here, we aim to identify whether age, gender, and race modify the association between time-varying depressive symptoms and mortality.

METHODS: Reasons for Geographic and Racial Differences in Stroke (REGARDS) is a prospective cohort study of black and white individuals (≥45 years) without a history of CVD or active cancer recruited between 2003 and 2007. The associations between time-varying depressive symptoms (CES-D ≥4 vs. <4) and all-cause and nonCVD mortality were measured using Cox proportional hazard regression analyses adjusting for demographic (age, gender, region, income, health insurance, education), medical (blood pressure, cholesterol, antihypertensives, statins, aspirin, antidepressants, cardiovascular disease, body mass index, diabetes, renal disease, pulmonary disease and cognitive impairment), behavioral (smoking, alcohol use, physical inactivity, medication adherence), and physiologic (C-reactive protein, functional status, and stress) risk factors. P-values for interaction terms were reported. All results were stratified by self-reported health status to isolate the effect of depressive symptoms on mortality.

RESULTS: Of the 29491 participants included in the analysis, 3253 (11%) had elevated depressive symptoms at baseline. The mean age was 65 (9.4) years; 55.1% were female; 41.1% black, 46.4% reported excellent/very good health. Depressive symptoms were measured at baseline and on average 5 and 7 years later. In those with excellent/very good self-reported health, time varying depressive symptoms were significantly related to: all cause mortality, particularly in those 45–65 years (vs. ≥65 years) (aHR = 1.56 [1.01-2.42] vs. aHR = 1.48 [1.21–1.81], p = 0.018) and nonCVD mortality, particularly in males (vs. females) (aHR = 1.72 [1.28–2.31] vs. aHR = 1.44 [1.04–2.00], p = 0.006) and blacks (vs. whites) (aHR = 1.80 [1.28–2.54] vs. aHR = 1.47 [1.10–1.96], p = 0.025).

CONCLUSIONS: In this large cohort of non-institutionalized adults with excellent/very good self-reported health status, we demonstrate that depressive symptoms appear to be associated with an increased proximal risk of mortality, particularly in those 45–65 years old, blacks, and males. In light of recent increases in mortality rates amongst middle-aged males, our study lends further support to calls for improved diagnosis and treatment in these high-risk groups who often fail to come into contact with healthcare providers.

ALCOHOL POLICY CHANGES AND 22-YEAR TRENDS IN INDIVIDUAL ALCOHOL CONSUMPTION IN A SWISS ADULT POPULATION: A 1993–2014 CROSS-SECTIONAL POPULATION-BASED STUDY Shireen Dumont 1; Pedro Marquez-Vidal4; Thierry Favrod-Coune1; Jean-Marc Theler1; Jean-Michel T. Gaspoz2; Barbara Broers1; Idris Guessous1, 3. 1Geneva University Hospitals, Geneva 14, Switzerland; 2University Hospitals, Geneva 14, Switzerland; 3University of Lausanne, Lausanne, Switzerland; 4University Hospital of Lausanne, Lausanne, Switzerland. (Control ID #2704839)

BACKGROUND: Evidence on the impact of legislative changes on individual alcohol consumption is limited. We assessed trends in individual alcohol consumption of a Swiss adult population following the public policy changes that took place between 1993 and 2014, while considering individual characteristics and secular trends.

METHODS: We used data from the “Bus Santé” study, an annual health survey conducted in random samples of the adult population in the State of Geneva, Switzerland. Individual alcohol intake was assessed using a validated food frequency questionnaire. Individual characteristics including education were self-reported. Seven policy changes (six about alcohol and one about tobacco) that occurred between 1993 and 2014 defined 6 different periods. We predicted alcohol intake using quantile regression with multivariate analysis for each period adjusting for participants’ characteristics and tested significance periods. Sensitivity analysis was performed including drinkers only, the 10th percentile of highest drinkers and smoker’s status.

RESULTS: Data from 18,963 participants collected between 1993 and 2014 (aged 18–75 years) were included. Between 1993 and 2014, participants’ individual alcohol intake decreased from 7.1 to 5.4 g/day (24% reduction, p < 0.001). Men decreased their alcohol intake by 34% compared to 22% for women (p < 0.001). The decrease in alcohol intake remained significant when considering drinkers only (28% decrease, p < 0.001) and the 10th percentile highest drinkers (24% decrease, p < 0.001). Consumption of all alcoholic beverages decreased between 1993 and 2014 except for the moderate consumption of beer, which increased. After adjustment for participants’ characteristics and secular trends, no independent association between alcohol legislative changes and individual alcohol intake was found.

CONCLUSIONS: Between 1993 and 2014, alcohol consumption decreased in the Swiss adult population independently of policy changes.

figure j

Study period divided in 6 periods according to policy changes between 1993–2014

ALGORITHMS FOR MANAGING CONCERNING BEHAVIORS IN PATIENTS PRESCRIBED OPIOIDS FOR CHRONIC PAIN: A DELPHI STUDY Jessica S. Merlin 4; Sarah Young8; Joanna L. Starrels1; Soraya Azari5; E. J. Edelman7; Jamie Pomeranz9; Payel J. Roy2; William Becker6; Jane M. Liebschutz3. 1Albert Einstein College of Medicine & Montefiore Medical Center, Bronx, NY; 2Boston University Medical Center, Boston, MA; 3General Internal Medicine, Boston Medical Center, Boston University School of Medicine., Boston, MA; 4University of Alabama at Birmingham, Birmingham, AL; 5University of California, San Francisco, San Francisco, CA; 6VA Connecticut Healthcare System, West Haven, CT; 7Yale University School of Medicine, New Haven, CT; 8SUNY Binghamton, Binghamton, NY; 9University of Florida, Gainesville, FL. (Control ID #2700506)

BACKGROUND: Opioid prescribing and addiction have increased dramatically in the last decade. Current guidelines recommend careful monitoring of patients prescribed long-term opioid therapy (LTOT) for chronic pain, and will likely lead to increased identification of concerning behaviors. Minimal research informs strategies to manage these behaviors, which pose an important management challenge and potential source of burnout for primary care providers. Our objective was to establish consensus treatment approaches for common and challenging concerning behaviors that arise among patients on LTOT.

METHODS: We used a Delphi approach, which allows for generation of consensus by a panel of experts. Participants were experts in chronic pain and opioid prescribing recruited from professional societies and other expert groups. The Delphi process consisted of four online rounds: brainstorming to identify common and challenging behaviors, identification of management strategies for each behavior, and two rounds to establish consensus and explore disagreement/uncertainty.

RESULTS: The most frequently-cited common and challenging behaviors were missing appointments, taking opioids for symptoms other than pain, using more opioid medication than prescribed, asking for an increase in opioid dose, aggressive behavior, and alcohol and other substance use. Results are synthesized in 7 management algorithms (see Figure for example). Across algorithms, participants agreed that patient education and information gathering were important approaches. Participants also agreed that stopping opioids is not an important intiial approach to these behaviors.

CONCLUSIONS: These algorithms provide much-needed guidance on how to manage several important concerning behaviors among patients on LTOT. Future research is needed to investigate the impact of the algorithms on patient outcomes. The long-term goal of this program of research is to provide primary care providers with a decision-support algorithm to manage concerning behaviors in patients on LTOT.

figure k


Dina H. Griauzde; Laurence F. McMahon; Linda M. Balogh; Floyd J. Brinley; John Crump; Mark Ealovega; Audrey Fan; Yeong Kwok; Kristen Krieger; Thomas O’Connor; Elisa Ostafin; Heidi Reichert; Jennifer Meddings. University of Michigan, Ann Arbor, MI. (Control ID #2698154)

BACKGROUND: Ambulatory care-sensitive conditions (ACSCs) are a group of acute and chronic conditions for which early and effective management in the primary care setting may prevent an Emergency Department (ED) encounter. Among insurers, ambulatory care-sensitive ED encounters are used as a quality metric to guide payment. Accordingly, the Patient Centered Medical Home (PCHM) model aims to reduce ambulatory care-sensitive ED encounters through strategies such as extended clinic hours, yet it is unknown whether patients attempt to access their medical home prior to ED presentation. Further, certain symptoms may not be appropriate for management in the primary care setting (e.g. chest pain), even if they are ultimately attributed to an ambulatory-sensitive condition (e.g. gastroesophageal reflux). This study aims to (1) characterize patterns of ED utilization for ACSCs among patients with established primary care within a General Medicine medical home and (2) to describe the appropriateness of the care location (e.g. ED vs. primary care clinic).

METHODS: We conducted a retrospective chart review using our institution’s Electronic Health Record of ambulatory care-sensitive ED encounters that occurred between January 1, 2014 and December 31, 2014 among patients of a General Medicine medical home. Ten General Medicine physicians reviewed a random sample (n = 263) of these encounters and abstracted from the medical record the day and time of ED presentation and the source of ED referral (e.g. patient self-referral vs. physician referral). Physicians also assessed the appropriateness of the care location (e.g. Emergency Department vs. primary care). Inter-rater reliability was assessed using the kappa statistic.

RESULTS: Compared to all other days of the week, the fewest number of ED visits occurred on weekend days, and nearly half of patients (47 percent) with ACSCs presented to the ED after business hours (8 am to 3:59 pm). Most patients (83 percent) were self-referred to the ED. Among the 119 cases considered appropriate for management by General Medicine, almost all (95 percent) of these patients were self-referred to the ED. The ED was considered the most appropriate care location for over half (54 percent) of the reviewed encounters.

CONCLUSIONS: Patients diagnosed with ambulatory care-sensitive conditions often present to the ED without contacting their medical home. Frequently, the ED is indeed the most appropriate location to manage the patient’s presenting condition.


BACKGROUND: In the setting of high and rising prevalence of type 2 diabetes among US adults, guideliens for clinical interventions and quality targets abound. Yet, how to organize medical care to efficienty achieve these best practices is not clear. Visit frequency yielding the greatest value for both individual patients and society remains especially uncertain. We sought to determin whether poor glucose control is associated with more frequent ambulatory visits among adult patients with type 2 diabetes who are medically homed at High Value Healthcare Collaborative (HVHC) member organizations.

METHODS: We conducted a retrospective cohort study of patients 18 years and older with type 2 diabetes who received at least 12 months of ambulatory care at organizations that were members of the High Value Healthcare Collaborative between February 2013 and February 2015. HVHC is a voluntary association of 14 non-profit US health care organizations with a common goal of improving quality while reducing costs. Our primary exposure is the patient HbA1c value at the start of a qualifying 12 month observation period. Our main outcome measure is the mean number of ambulatory visits with primary care or endocriniology clinicians per person in the 12 months following index HbA1c, and our secondary outcome measure is the proportion of those visits provided by an endocrinology specialist. We adjusted for age, sex, race, ethnicity, and poverty status.

RESULTS: We examined visit frequency for 103,220 patients from nine system-specific cohorts ranging in size from 693 to 30,973 patients. Race, ethnicity, sex, poverty status, and index HbA1c value varied substantially across sites. In unadjusted analyses, the overall mean ambulatory visit rate was 4.6 visits per year and ranged from 2.0 to 5.4 visits per year across member organizations. The proportion of annual visits with an endocrinology specialist was 3.7% overall and ranged from 0.3 to 11.4%. The association between index HbA1c value and annual visit rate was very modest ranging from 4.5 for patients with an index HbA1c less than 7.0 to 4.8 for those with an index HbA1c value of 9 or greater.

CONCLUSIONS: Mean ambulatory visit frequency varied across HVHC organizatons but only correlated meaningfully with index HbA1c value at 2 sites. In contrast, the proportion of visits delivered by endocrinology specialists was low across sites but did increase with higher index HbA1c values across all sites studies

AN ANALYSIS OF STUDENTS’ REFLECTIVE WRITING PROVIDES A WINDOW INTO PROFESSIONAL IDENTITY FORMATION, RESILIENCE, AND BURNOUT PROTECTION IN THE DENVER HEALTH LONGITUDINAL INTEGRATED CLERKSHIP (DH-LIC) Michelle Cleeves 2; Mim Ari3; jennifer gong4; Jennifer Adams1. 1Denver Health, Denver, CO; 2Denver Health and Hospital Authority, Denver, CO; 3Cook County, Chicago, IL; 4University of Colorado School of Medicine, Aurora, CO. (Control ID #2706503)

BACKGROUND: Burnout, prevalent among practicing physicians, is also found in U.S. medical students. Undergraduate medical educators are therefore ideally placed to address burnout during the clinical years when professional identity begins to take shape. Students in longitudinal integrated clerkships (LICs) complete core clerkships longitudinally working with core faculty preceptors. LIC students finish 3rd year with more positive attitudes about their clerkships, stronger patient-centeredness, and more confidence in their patient care abilities compared to traditional 3rd year students. In the DH-LIC, students learn exclusively in Denver’s only safety-net hospital system, providing care for the underserved for the duration of their 3rd year. The DH-LIC curriculum incorporates reflective exercises designed to help students process their experiences in this setting. Through qualitative analysis, these reflections provide insight into students’ professional identity formation, resiliency, and strategies protecting against burnout.

METHODS: DH-LIC students completed reflective writing exercises followed by facilitated small group discussions throughout 3rd year. Topic included: social determinants of health, boundary setting with patients, and closure with patients. The authors analyzed 2 years of essays (n = 45 essays, 15 students). Through an iterative process we identified themes, then independently coded to find agreement among coders on consistent themes within and across essays. We then totaled the percentage of students reflecting on each theme.

RESULTS: Reflections were firmly rooted in understanding the system of disadvantage constraining patient care (100% of students recognized the impact of patient disadvantage & 87.7% healthcare system failure on patients’ health) and a commitment to treat patients as people (60% demonstrated sympathy & empathy for patients and 73% humility when learning from patients). Reflections also revealed behaviors & attitudes that demonstrated efficacy & self-monitoring. Common themes included: advocating for & engaging with patients (86.67%), understanding provider vulnerability & learning to set professional boundaries (73%), valuing continuity with patients (60%), & feeling valued by patients & preceptors (60%). More than 60% of students identified these behaviors & attitudes as key to their role and identity as physicians. Students identified the LIC structure (100%) and positive & negative role models (73%) as contributing to these experiences.

CONCLUSIONS: Through the course of the LIC year, students demonstrate a developing sense of professional identity consonant with that needed to provide compassionate care to the underserved. Analysis of the reflective essays acted as a window into the themes gleaned from the students’ experience. Student reflections demonstrated self-efficacy, identification of and commitment to key values, and resilience in the face of challenges.

AN E-LEARNING MODULE ON CHRONIC LOW BACK PAIN IN OLDER ADULTS: EFFECT ON MEDICAL RESIDENT ATTITUDES, CONFIDENCE, KNOWLEDGE, AND PRACTICE PATTERNS Zachary G. Jacobs 2; Michael Elnicki1; Subashan Perera2; Debra K. Weiner2. 1University of Pittsburgh, Pittsburgh, PA; 2University of Pittsburgh Medical Center, Pittsburgh, PA. (Control ID #2672639)

BACKGROUND: Chronic low back pain (CLBP) negatively impacts the lives of millions of Americans each year, posing an enormous financial burden; this is in large part due to an inadequacy of chronic pain education. The goal of this study is to investigate the feasibility of using an online module to teach medical residents about CLBP in older adults, and to determine its impact on their attitudes, confidence, knowledge, and ability to evaluate and manage CLBP in the clinic.

METHODS: All categorical internal medicine residents at the University of Pittsburgh Medical Center from 2015–2016 were assigned to intervention (N = 73) or control groups (N = 70) based on clinic schedule. The intervention group was instructed to complete an online, self-guided module previously developed by a panel of experts. The control group was exposed instead to the Yale “Office-Based Medicine Curriculum” on CLBP. Knowledge, attitudes and confidence were assessed pre- and post-intervention. Knowledge was assessed with 1) the KnowPain-12 survey (a validated, 12-item questionnaire), and 2) ten multiple choice CLBP-specific questions written by pain specialists. A reviewer masked to group assignment conducted a retrospective review of resident clinic encounters, rating physical exams and diagnoses as either beginner or advanced based on terms used in documentation.

RESULTS: Survey results from pre- (N = 44) and post-intervention (N = 42) showed no improvement on the 10-item multiple choice test or the KnowPain-12 survey in either group (60% average on both metrics). There were tendencies for greater improvements in the intervention group compared to controls in confidence in managing fibromyalgia (2.4 to 2.9 vs 2.5 to 2.5; p = 0.06) and leg length discrepancy (1.8 to 2.5 vs 1.5 to 1.9; p = 0.05). Those exposed to the module were also more likely to use more advanced diagnosis codes (15% vs 5%) and physical exam documentation (62% vs 45%) compared to controls.

CONCLUSIONS: This study demonstrates the use of an online module is a feasible method for teaching medical residents about CLBP. One of the most startling findings is the paucity of knowledge amongst participants. These data highlight the importance of developing effective methods for educating clinicians about chronic pain. While the module did not lead to greater overall improvements on knowledge tests, it did lead to improved resident confidence, and greater sophistication in evaluating patients with CLBP.

figure l

AN ELECTRONIC HEALTH RECORD-ENABLED (EHR) UNIVERSAL MEDICATION SCHEDULE TO PROMOTE ADHERENCE: A PRAGMATIC TRIAL Guisselle del Salto 1; Deesha Patel1; Laura M. Curtis1; Ruth Parker3; Fred Rachman4; Sarah S. Rittner4; Elizabeth Adetoro4; Andrew Hamilton4; Michael S. Taitel2; Jenny Jiang2; Amisha Wallia1; Michael S. Wolf1. 1Northwestern University, Chicago, IL; 2Walgreens Co, Deerfield, IL; 3Emory University, Atlanta, GA; 4Alliance of Chicago Community Health Services, Chicago, IL. (Control ID #2705084)

BACKGROUND: Many patients misunderstand prescription medication (Rx) instructions, and this has been repeatedly linked to medication errors, adverse drug events and poor adherence. Patients with limited literacy and English proficiency, and those with more complicated drug regimens, are especially at risk of unintentional misuse. We collaborated with federally qualified health centers and a national pharmacy chain to impart a health literacy ‘best practice’ - the Universal Medication Schedule (UMS) - via EHR and mobile technologies to better convey dosing instructions and help organize multi-drug regimens among type 2 diabetic patients.

METHODS: We randomized 59 providers from 11 practices to usual care or one of two intervention arms: 1. UMS Arm: Leveraging the GE Centricity EHR platform, patients received 1) UMS Rx instructions that guided patients to take all of their medication at 4 standard time intervals (morning, noon, evening, bedtime); 2) a single-page, plain language medication information sheet in English and Spanish that was designed following health literacy best practices was generated with an After Visit Summary (AVS) once ordered; and 3) A complete list of patients’ current medications mapped to UMS intervals was automatically generated with the AVS to help patients consolidate their entire regimen to the 4 standard daily intervals. 2. UMS + SMS: Patients also were exposed to the UMS tools at the point of care, but also received SMS text reminders about when to take medicines based on UMS intervals. 452 English and Spanish-speaking type 2 diabetic patients who were prescribed 3+ medications for chronic conditions and could receive text messages were assessed by phone at baseline, 3 months, and 6 months. Patient outcome measures include medication knowledge, demonstrated proper use, medication adherence (measured by self-report and pill count) and clinical outcome measures.

RESULTS: In multivariable analyses, no significant differences in any of the outcomes were found by study arm. By six months, however, patients receiving the UMS + SMS intervention were significantly more knowledgeable of their medications’ indication (Adjusted Odds Ratio (AOR) 1.43, 95% Confidence Interval (CI) 1.06-1.93, p = 0.02) and adherent to their regimen according to self-report of missed doses (AOR 2.33, 95% CI 1.02-5.31, p = 0.045). No differences in rates of proper use, adherence via pill count, or clinical outcomes count were noted. In addition, no interactions between study arm and literacy level or language were found to be statistically significant.

CONCLUSIONS: Our trial tested embedding the UMS and tools within an EHR platform at primary care practices. While we did find an effect on self-reported adherence and even knowledge, this health literacy intervention did not improve adherence via the more objective metric of pill count, nor clinical outcomes. Furthermore, these benefits only occurred with the UMS + SMS intervention, and not UMS arm alone.

AN EMOTIONAL WELL-BEING CAMPAIGN: MESSAGING FOR THE PUBLIC Jessica Kaltman 1; Sophie Feller1; Lello Tesema1; Enrico G. Castillo1, 2; Kenneth B. Wells1. 1UCLA, Los Angeles, CA; 2Los Angeles County, Los Angeles, CA. (Control ID #2706351)

BACKGROUND: As part of the Affordable Care Act, the National Prevention Council comprising multiple federal agencies and led by the Office of the Surgeon General (OSG) was created. In 2011, the council released the National Prevention Strategy which identified seven prevention priority areas to target in order to improve the health of the nation. Mental and emotional well-being (EWB) is one of the seven priority areas setting the foundation to develop a national EWB campaign. An important aspect of a successful campaign is to understand the terminology currently used to discuss this issue amongst the American public. We conducted a rapid review of the popular media to understand how the concept of EWB is portrayed to the public.

METHODS: A rapid review of popular media sources was conducted between June 8, 2016 to July 5, 2016 to assess the terminology used by the media to discuss the subject of EWB. Google alerts, which prospectively pulls articles from media outlets on topics specified by search terms, were created for 18 terms related to EWB. These terms were chosen from the academic tradition (e.g., thriving, flourishing, coping, positive psychology, grit and resilience), interviews with key community and academic stakeholders and commonly used terminology (e.g., well-being, wellness, emotional fitness, and emotional well-being). Additionally, the websites for New York Times, Washington Post, The Atlantic, and New Yorker magazine were searched separately, as Google Alerts excludes these outlets. A team of three people reviewed over 150 media articles that used the terminology chosen. Articles were evaluated for use of the terminology.

RESULTS: Our findings demonstrate that happiness, resilience and stress pervade the public media. Thriving, and occasionally grit, are terms applied to topics not related to well-being and are more often used to describe crops, athletes, sports teams and economies. Importantly, EWB did not appear to be a widely-referenced term in the popular media. Happiness was used in ways that most closely related to EWB research and interventions.

CONCLUSIONS: A review of the popular media lends insight into how the media process and the public receive information related to EWB. These preliminary findings may carry implications on the challenges in developing unified public messaging on EWB. The preference of the media to use terms other than EWB should be taken into consideration when messaging about an EWB initiative.


BACKGROUND: Random blood glucose (RBG)-based case findings strategies perform better than national screening guidelines to detect undiagnosed diabetes in community-based studies. However, the performance of electronic medical record (EMR)-based RBG predictive models using real-world clinical data is unknown.

METHODS: We conducted a retrospective cohort study using EMR data from a large, integrated, safety-net health system and developed a model to predict diabetes (A1C ≥ 6.5 or fasting blood glucose (FBG) ≥ 126 mg/dL) and dysglycemia (A1C ≥ 5.7 or FBG ≥ 100 mg/dL). Non-pregnant primary care patients, age 18–64, with an index visit and a resulted diabetes screening test (A1C or FBG) between June 1, 2011 and December 31, 2014 and ≥1 outpatient RBG value 12 months prior to the screening test were eligible. We excluded patients with diagnosed diabetes or prediabetes on or 18 months before the index visit using ICD9 codes and lab results. Patient demographics, comorbidities, and diabetes risk factors were extracted from the EMR. We built RBG-only and a RBG+risk factor (RF) models to predict diabetes and dysglycemia using logistic regression with backwards selection to retain age, race, BMI, and hypertension in the final RBG+RF model. The most recent RBG prior to the date of the screening test was the primary predictor. In a subgroup analysis of patients with >1 RBG value (N = 8,060), we also modeled mean RBG as a predictor. We report model c-statistics as a measure of discrimination.

RESULTS: A total of 22,058 patients met inclusion criteria. Mean age was 49 years and mean BMI was 31. Overall, 60% were female, 84% were non-Caucasian, 59% had hypertension, and 54% had a family history of diabetes. All patients satisfied American Diabetes Association Screening criteria and 62% met the 2015 USPSTF Diabetes screening criteria. C-statistics in the RBG-only models were 0.88 (undiagnosed diabetes) and 0.73 (undiagnosed dysglycemia). In the RBG+RF model, c-statistics to detect undiagnosed diabetes (0.91) and dysglycemia (0.78) were significantly higher than the RBG-only models (p < 0.01). Both the RBG-only and RBG+RF models performed significantly better than the 2015 USPSTF Screening to detect undiagnosed diabetes (0.57) and dysglycemia (0.61) (p < 0.001). In the subgroup with >1 RBG, c-statistics for the RBG +RF models using mean RBG performed slightly better than those using the most recent glucose to detect undiagnosed diabetes (0.91 vs. 0.89) and dysglycemia (0.79 vs 0.77) (P < 0.01).

CONCLUSIONS: Predictive models using only a single, most recent RBG to detect diabetes and dysglycemia are superior to national screening guidelines, and adding diabetes risk factors commonly available in the EMR further improves performance. Further validation using prospectively collected screening tests is needed to address potential biases in retrospective EMR data.

AN EVIDENCE MAP OF THE WOMEN VETERANS’ HEALTH RESEARCH LITERATURE (2008–2015) Elisheva Danan 1; Kristine E. Ensrud2; Erin E. Krebs1; Eva Koeller1; Tina L. Velasquez1; Roderick MacDonald1; Nancy Greer1; Timothy Wilt1. 1Minneapolis VA Health Care System, Minneapolis, MN; 2Univeristy of Minnesota/VA Health Care System, Minneapolis, MN. (Control ID #2705633)

BACKGROUND: Women comprise the most rapidly growing population of Veterans seeking care at the Department of Veterans Affairs. Clinical, research and policy initiatives to serve female Veterans exist, yet the corresponding literature has not been systematically reviewed since 2008. In 2015, VA Women’s Health Services and the VA Women’s Health Research Network jointly requested an updated literature review to assess the scope of research related to female Veterans, determine if research was aligning with priority areas and guide planning.

METHODS: As part of the VA Evidence-based Synthesis Program we conducted a systematic review and created an evidence map of research related to female Veterans’ health published from January 2008 to December 2015. We searched the MEDLINE, CINAHL and VA HSR&D databases to find literature related to female Veterans’ health. From included studies, we extracted study characteristics including healthcare topic, design, participant number, proportion women, research setting, funding source, and more. We organized and presented results via visual representation and text summary within and across healthcare topics. We identified patterns, strengths and gaps.

RESULTS: We identified 2,276 abstracts that were independently assessed for relevance. Eligible articles (k = 437) were sorted into 39 healthcare topics and described according to 14 additional study characteristics. Nearly half the articles were related to mental health (207/437 studies, 47%), particularly post-traumatic stress disorder, military sexual trauma, and substance abuse (these 3 topics comprised 61% of the mental health articles). Few studies addressed common chronic diseases often seen by general internists such as diabetes and hypertension (3 studies) or depression and anxiety (4 studies). Nearly all (396/437, 91%) articles described observational studies. Less than 2% (8/437) presented primary findings of randomized trials. We found growth in the number of articles published per year. In the first half, 2008–2011, 134 studies were published, whereas more than double that (303 articles) were published from 2012–2015. The number of studies within smaller VA priority topic areas increased over time, including post-deployment health, reproductive health, access and utilization, and healthcare organization and delivery. We excluded many identified studies that included female Veterans from our evidence map because they did not stratify results by gender, thereby limiting their applicability to female Veterans.

CONCLUSIONS: Female Veterans’ health and healthcare literature grew substantially from 2008–2015. Observational studies in mental health make up the majority of recent research, though additional research agenda priority areas demonstrate progress. Methodological and health condition gaps remain. This evidence map can inform clinical, research, and policy initiatives.

AN EXPENSIVE WHITE COAT Salam Hawa 1; Raynata Ramkhelawan1; Ravneet Randhawa1; Niket Sonpal2. 1American University of Antigua, Newark, NJ; 2Touro College of Osteopathic Medicine, New York, NY.

(Control ID #2706288)

BACKGROUND: A dream is a mere goal sometimes cut short by reality; this is one realization made by many medical students striving towards their career. As the years pass and the financial burden grows, the reason of dreaming of a medical degree, whether an innate passion, parental demand, or simply a change in career path, is no longer the only significant factor in making career decisions. The purpose of this survey is to collect data to investigate the level of impact of how accumulated debt imposes on the choice of specialty by medical students and residents. Studies have shown a trending decline in primary care with a simultaneous inclination towards specialty medicine.

METHODS: The survey consisted of 26 questions administered via Google forms and made accessible for 20 days. The questions were designed to acquire the demographics of each student pertaining to the accumulated loans, level of financial understanding, and furthermore how accumulated debt affects their medical specialty choice.

RESULTS: At the end of the 20 day period, there were 118 responses collected. Of total respondents, 56.8% were in the 1st or 2nd year of medical school, 40.7% were in there 3rd or 4th year, and 2.5% were in their 1st to 3rd year of residency. Of 118 respondents, only 117 responded regarding total accumulated debt; and among those responses, 14 will be excluded due to their lack of a specified answer. Among the remaining respondents, 30.09% have an accumulated debt of less than $60,000, 34.95% have between $60,001 to $140,000, 22.33% have between $140,001 to $220,000, and 12.6% have above $220,001. When asked if loan amounts were sufficient to cover their obligations, there were 116 responses. Of those responses, 28 were excluded due to lack of a specified answer. Of the remaining responses, 60.22% stated their loans were sufficient and 39.77% stated they were not able to meet their financial obligations with their loans. When asked if the amount of accumulated debt provoked a change in the choice of specialty, 117 responses were collected. Among those, 17 responses will be excluded due to their lack of specified answer. Among the remaining responses, 52% stated their choice did not change and 48% stated it did.

CONCLUSIONS: Upon review, most students fall within the debt range of $60,001 to $140,000 and the majority of students stated loans sufficiently covered their needs and furthermore indicated debt was not a factor in choice of field. However, there is a substantial 39.77% of students who counteract this notion and a greater 48% who indicated their interest in pursuing a high paying specialty serves as a means to pay their accumulated debt. Further comparisons in the changing rates of medical school costs versus compensation rates for physicians will be essential in defining the root of the problem.

AN IN-DEPTH CASE STUDY OF ADAPTING PATIENT EXPERIENCE DATA COLLECTION FOR LOWER LITERACY PATIENT POPULATIONS USING TABLETS IN CLINIC Courtney Lyles1; Lina Tieu 1; Alicia Hobbs2; Erin E. Curtis2; Urmimala Sarkar1. 1University of California, San Francisco, San Francisco, CA; 2San Francisco Department of Public Health, San Francisco, CA. (Control ID #2705455)

BACKGROUND: Healthcare policy supports the inclusion of patient experience in healthcare quality measurement and reporting. However, response rates to the gold-standard, paper-based Consumer Assessment of Healthcare Providers and Systems (CAHPS) surveys are low and differ significantly by racial/ethnic and language subgroups. There is little discussion about whether CAHPS captures the true domains of patient healthcare experience, how to improve reporting for underserved patient populations, or whether data collection via mobile technology might result in higher response rates from more representative populations.

METHODS: Our study sought to 1) design and evaluate tablet-based CAHPS administration, and 2) conduct formative qualitative work to create shorter, lower literacy patient experience items (informed by CAHPS) and explore broader concepts of patient experience among vulnerable patients. We partnered with a start-up company Shift Health to create a tablet interface that was visually attractive and simple to use and created multi-lingual, low-literacy questionnaires. Next, we used 25 think-aloud interviews with patients to validate a new literacy-appropriate tablet questionnaire compared to the standard paper version, eliciting their perspectives about using a tablet-based survey to report their care experiences.

RESULTS: Of the 25 participants, 52% were male, 80% were non-White, and 68% had limited health literacy; the mean age was 53. Almost a quarter (24%) and over a third (36%) had never used a smartphone or tablet, respectively. Patients answered the majority of questions similarly on the paper vs. tablet versions, but strongly preferred the tablet, even among those without significant previous experience using mobile technology: “This [the tablet] is quite easy to do, because everything is easy to read. You just literally have to point your finger at your answer.” Patients emphasized the importance of reporting feedback to their clinics, particularly concerning wait times, quality of provider communication, and ability to access care, and valued the option to complete the survey at the point of care: “It’s better in the clinic because you’re in the same environment.” The final literacy-appropriate questionnaire made concrete improvements to the CAHPS items, including 1) reducing the total number of items from 31 to 17 while preserving core domains, 2) reducing the reading level required to understand survey questions from 7th to 5th grade (Flesch-Kincaid test), and 3) adding open-ended questions to capture patient-directed perspectives of care.

CONCLUSIONS: As we move forward with federal policy supporting patient experience data collection, this study provides clear next steps to ensure underrepresented and vulnerable patient perspectives are engaged and represented in this process. If designed with patient input, tablet-based surveys may be a feasible and effective method for collecting patient experience data at the point of care.

AN INTERDISCIPLINARY STRATEGY FOR IMPROVING HAND HYGIENE ON AN INPATIENT MEDICINE UNIT Katherine A. Hochman 2; Nicole Adler1; Lisa Gumbrecht3; Brian Bosworth1. 1NYULMC, New York, NY; 2New York University School of Medicine, New York, NY; 3NYU Langone Medical Center, New York, NY. (Control ID #2701187)

BACKGROUND: The CDC reports that 5% of hospitalized patients develop hospital acquired infections, which are responsible for 100,000 deaths annually. Poor hand hygiene compliance on the Medicine service placed patients at higher risk for infection and was the impetus for our Clean Hands Save Lives Initiative.

METHODS: The Clean Hands Save Lives initiative was a triple-prong systems-based strategy that required the leadership of the medical director and nurse manager and the engagement of the entire floor. First, unit leadership dedicated a portion of the the morning unit-based safety huddle to identify daily handwashing champions. Each day a new group of champions was identified, including 2 nurses, one one floor patient unit technician, and two physicians. Champions were responsible for reinforcing correct hand hygiene procedures in real time, promoting a culture of “if you see something, say something.”Each week, 35 different health care providers were hand-hygiene champions, hardwiring best practice. Second, proper hand hygiene procedures were reinforced at the safety huddle several times a week and Purell dispensers were installed outside every patient room.Third, an email was sent to each team member regarding the hand hygiene initiative at the start of each rotation. Real-time feedback on hand hygiene technique was provided by unit leadership.

RESULTS: At the start of the initiative in quarter 1 of 2015, hand hygiene compliance for the 17 East Medical Unit was at 64%. By quarter 1 of 2016, hand hygiene compliance was at 93 and has remained above 90% for the past 4 quarters (Figure 1).

CONCLUSIONS: Successful implementation of our Clean Hands Save Lives Initiative on a hospitalist led medicine unit was due in large part to making this a daily focus of all members of the team, leading to unit culture change. The interdisciplinary approach to the problem, daily reinforcement of the initiative, regular education of unit staff and ease of practicing proper hand hygiene all were contributing factors to its success and sustainability. The initiative is now practiced in all units on the Medicine service.

figure m

AN INTERVENTION TO REDUCE BLACK-WHITE CANCER TREATMENT DISPARITIES: THE CASE OF EARLY STAGE NON-SMALL CELL LUNG CANCER Samuel Cykert 1; Lloyd Edwards2; Paul Walker3; Rohan Arya4; Peggye Dilworth-Anderson2. 1University of North Carolina, Chapel Hill, NC; 2UNC Gillings School of Global Public Health, Chapel Hill, NC; 3ECU Brody School of Medicine, Greenville, NC; 4University of South Carolina School of Medicine, Columbia, SC. (Control ID #2702897)

BACKGROUND: African-Americans (AA) with lung cancer experience a higher annual death rate compared to Whites (W) with AA men particularly affected (79 vs 66 deaths per 100,000 population). Despite this risk, Bach and others have shown that treatment rates for AA patients with early stage disease lag behind rates for W. In this report, we describe results from a multisite interventional trial designed to optimize lung surgery and overall treatment rates for all patients and reduce treatment disparities.

METHODS: Baseline rates for curative surgery and radiation were established using 3-year chart review for all patients with biopsy proven lung cancer at 3 institutions. We then prospectively recruited AA and W patients with newly diagnosed stage I or II non-small cell lung cancer. Patients were identified at initial visit either through biopsy proven disease or a Bayesian probability algorithm. Informed consent was obtained. Intervention components for enrolled patients included: (1) entry into a real time electronic registry that provided warnings for missed appointments or unfulfilled, pre-designated milestones in care, (2) quarterly feedback of treatment rates by race and co-morbid illness to cancer care personnel, and (3) provision of a specially trained navigator to enhance communication with affected patients and between patients and the care team. Descriptive statistics, bivariate analyses, and logistic regressions were performed with a primary outcome of surgical treatment for lung cancer cure and a secondary outcome of surgery or stereotactic radiation for cure.

RESULTS: Baseline surgical rates from the retrospective analysis of 714 early stage, non-small cell patients were 69% for W and 66% for AA patients. Using logistic regression controlling for comorbidities, COPD, age, and other demographic data, we determined the odds ratio for surgery for AA compared to W lung cancer patients as 0.64 (95% CI 0.43, 0.96). When surgery was combined with stereotactic radiation for cure, the treatment rates improved to 80% for W and 76% B. The same regression analysis applied to combined treatment rates yielded an OR of 0.61 (95% CI 0.40, 0.96) for AA vs W. 244 patients were recruited prospectively including 89 AA (36%) for the intervention. The mean age for this group was 65.7 years; 54% were women. Findings in the intervention group show an overall surgical rate of 74% (74.8% W, 71.4%AA, p = 0.6). Combined treatment rates increased to 91.9% for W and 94.1% AA patients (p = 0.5). Logistic regression was performed comparing the intervention group to the baseline group. Results showed that overall treatment improved for both W and AA, the surgical and overall treatment disparity resolved, while age, COPD, and clinical stage remained significant predictors of treatment.

CONCLUSIONS: Results from an interventional study designed to optimize lung cancer treatment and narrow Black-White treatment disparities appear promising and could be applicable in reducing other cancer disparities.


BACKGROUND: Current guidelines suggest the use of the more specific Wells score could safely reduce the number of unnecessary scans. There is a lack of research to support whether these guidelines apply to the Black population.

METHODS: A retrospective descriptive study to determine the diagnostic utility of VQ scan was conducted among patients who were seen during January 2012 to January 2016. The study population included patients who underwent VQ scan for evaluation of pulmonary embolism. A total of 180 charts were reviewed and 49 were excluded due to poor quality data. A review of the initial history and physical progress notes, as well as discharge summaries was performed. Clinical presentation and Wells probability of pulmonary embolism were compared with the results of the scan.

RESULTS: We collected data on 131 patients who underwent VQ scan for evaluation of pulmonary embolism. The median age of the study population was 63.02 ± 16.12 years. Majority of the study population 121 (92.4%) was black. Sixty four (48.9%) VQ scans were done for a low clinical probability for pulmonary embolism as defined by the well’s clinical score. The most common clinical presentations were shortness of breath (SOB) 74 (58%), Leg pain 39 (29.8%), chest pain 36 (27.4%), and syncope 4 (3.1%). Sixty two (96.9%) patients with low clinical probability had low probability VQ scan(p = 0.030). Among patients who did CT angio and VQ scan, low probability scan was noted in patients with no pulmonary embolism in CT in 25(96.2%) (p = 0.006).

CONCLUSIONS: Using the accepted guidelines in which a high pretest probability leads to further imaging and a low probability leads to a D-dimer blood test, use of the more specific Wells score could safely reduce the number of unnecessary scans. [13]The combination of a negative D-dimer with a low or moderate clinical probability can safely exclude pulmonary embolism in many patients.[14]This was reflected in our study, which showed that in patients with normal D-dimer, the result of VQ scan was low probability in 100%. However, no statistically significant correlation was observed between elevated d-dimer and results of VQ scan. In view of this finding, we suggest even in patients with elevated d-dimer other possible diagnosis should be explored. The study showed strong correlation between low clinical probability and low probability VQ scans and its utility to safely rule out PE in a predominantly black population. Studies conducted in other populations have detected similar findings.

Correlation between the clinical pretest probability with *VQ scan probability. Sixty two (96.9%) patients with low clinical probability had low probability VQ scan.(p = 0.030)

  Low Probability VQ scan n(%) Intermediate Probability VQ scan n(%) High Probability VQ scan n(%) Total n(%)
Low clinical probability 62(96.9%) 2(3.1%) 0(0.0%) 64(100.0%)a
Moderate clinical probability 46(93.9%) 0(0.0%) 3(6.1%) 49(100.0%)
High clinical probability 10(76.9%) 1(7.7%) 2(15.4%) 13(100.0%)

*VQ - Ventilation perfusion scans. p- Value: a 0.030

ANXIETY AND DEPRESSION IN UNIVERSITY STUDENTS EXPERIENCING INTIMATE PARTNER VIOLENCE: PRELIMINARY FINDINGS OF A CROSS-SECTIONAL STUDY Danny Lee 1; Arif Pendi2; Alfonzo Valdez2; Jose L. Aguilar3; Kate Basia Wolitzky-Taylor4; Joshua Lee4; Kasim Pendi3; David Safani2. 1Virginia Commonwealth University School of Medicine, Yorba Linda, CA; 2University of California Irvine, Orange, CA; 3University of California Riverside, Riverside, CA; 4University of California Los Angeles, Los Angeles, CA. (Control ID #2685507)

BACKGROUND: Intimate partner violence (IPV) is an under-reported but commonly occurring phenomenon affecting people of all ages. Among university students, IPV and its consequences are not thoroughly studied. Given that anxiety and depression may follow IPV, it is imperative to research IPV and its potential effects in university students, a group already characterized by high rates of anxiety and depression. Furthermore, recent reports have suggested that the mental health sequelae of IPV may cause greater long-term harm than physical injuries. This may indicate a need for internists to screen for anxiety or depression in this sub population. Thus, this study had the following objectives: (1) report the prevalence of IPV in university students and (2) investigate the association between IPV and both anxiety and depression.

METHODS: A cross-sectional study design was employed; an anonymous survey was sent to all professors at a large public university in the United States with a request to forward the link to undergraduate and/or graduate students. The instrument consisted of a socio-demographic questionnaire, HITS Domestic Violence Screening Tool, Generalized Anxiety Scale 7 (GAS-7), and Patient Health Questionnaire-9 (PHQ-9). Full-time students over the age of 18 were included; part-time students or those below the age of 18 were excluded. Participants that screened positive for IPV according to HITS were compared via t-test to those that screened negative in terms of their continuous scores on the GAS-7 (for anxiety) and PHQ-9 (for depression). All standardized scales (HITS, GAS-7, and PHQ-9) have displayed adequate validity and reliability according to the literature. Analyses were conducted with IBM® SPSS® Version 22.

RESULTS: Completed responses (n = 396; 24% female; 86% undergraduates) were scored on the HITS Domestic Violence Screening Tool. Approximately 5% of student respondents (18 of 396 participants) screened positive for IPV. This group exhibited greater anxiety (10.83 ± 5.711 versus 6.23 ± 5.355; p = 0.003) and greater depression (12.06 ± 6.384 versus 7.34 ± 6.266; p = 0.007) compared to respondents that screened negative for intimate partner violence.

CONCLUSIONS: A small but significant minority of respondents reported IPV. This group was associated with significantly greater anxiety and depression, suggesting that experience of intimate partner abuse is correlated with increased generalized anxiety and depression severity. As a result, this sub-group of students may benefit from augmented screening efforts for anxiety and depression by general internists. Given that university students have already been associated with a large burden of anxiety and depression, those that have experienced intimate partner violence may subsequently experience greater disorder severity.

ARE ALL PATIENTS IN FEDERALLY QUALIFIED HEALTH CENTERS EQUALLY READY TO USE HEALTH INFORMATION TECHNOLOGY? Nazia Naz S. Khan; Karen Kelly-Blake; Gurpreet Chahal; Saman Kandola; Zhehui Luo; Ade B. Olomu. Michigan State University, East Lansing, MI. (Control ID #2705525)

BACKGROUND: Health information technology (HIT) has emerged as an innovative way to engage, empower patients and improve access to health care. However, patients navigating health care in low resource communities may not benefit from this potential even when information and communication technologies are readily available. Our objectives were: 1) to assess Federally Qualified Health Center (FQHC) patient readiness and interest in adopting telecommunication, i.e. cell phone, text message and email to communicate with healthcare providers, and 2) to determine whether there are differences in the readiness and usage of technology between FQHC clinics.

METHODS: This analysis is part of the larger Office-Guidelines Applied to Practice (Office-GAP) study designed to improve secondary prevention of heart disease for DM and CVD patients in FQHCs in Michigan. Office-GAP is a quasi-experimental design with 2 FQHC clinics. 499 patients were enrolled in Office-GAP program. 304 patients completed the Technology Usage Readiness Survey, which evaluated their readiness, interest, and barriers to health information technology for communication with providers. There were 153 patients in Clinic 1 and 151 in Clinic 2. Data were analyzed using descriptive statistics and logistic regression models. We adjusted for education, race, insurance, smoking and depression between the two clinics.

RESULTS: Mean age was similar (52 years +/− 11) between Clinic 1 and 2. In Clinic 1, 37% had no high school degree compared to 52% in Clinic 2. Whites comprised 37% of patients in Clinic 1 and 24% in Clinic 2. There was no difference in the percentage of Black patients between the two clinics. We compared Clinic 1 and Clinic 2 patient responses in adjusted analyses: have a cell phone (OR = 1.3, 95% CI = 0.7–2.7); ability to text controlling for having a cellphone (OR = 0.9, 95% CI = 0.4–1.8); like doctors to text controlling for having a cellphone (OR = 1.8, 95% CI = 1.0–3.2); access to internet/email (OR = 1.6, 95% CI = 0.9–2.9); like doctors to email (OR = 1.5, 95% CI = 0.8–2.9). Out of 151 patients in Clinic 2, 55 patients were immigrants: 38% had cell phone; 16% had the ability to text; 13% like doctors to text, 22% have access to internet/email, and 9% like doctors to email them about their conditions.

CONCLUSIONS: We found a statistically significant difference in patient readiness to engage/communicate with their providers using text messages between the two FQHCs. Immigrant patients are less ready to use mobile health (m-Health). With the push for m-Health, assessment of individual clinic patient populations is necessary before telecommunication interventions are implemented. With differences in patient population who have cellphone, ability to text and email it is evident that many patients in FQHCs are not ready to adopt m-Health.

ARE LONGITUDINAL INTEGRATED CLERKSHIPS (LICS) AT A TIPPING POINT? A NORTH AMERICAN SURVEY OF INTERNAL MEDICINE CLERKSHIP DIRECTORS Lindsay A. Mazotti 4, 6; Jennifer Adams1; Bruce Peyser2; Katherine C. Chretien5; Briar Duffy7; David Hirsh3. 1Denver Health, Denver, CO; 2Duke, Durham, NC; 3Harvard Medical School, Cambridge, MA; 4Kaiser Permanente Oakland, Oakland, CA; 5Washington DC VAMC, Washington, DC; 6University of California San Francisco, San Francisco, CA; 7Univeristy of Minnesota, Minneapolis, MN. (Control ID #2688452)

BACKGROUND: LICs are an example of educational innovation, fueled by a desire to improve educational continuity and address the “hidden curriculum.” The number of North American (NA) institutions adopting LICs is increasing, but the extent of the change remains unknown. Clerkship Directors (CDs) in Internal Medicine (IM) can provide information on LIC program growth, as every school has an IM clerkship. CDs serve as leaders within institutions and are likely aware of or directly involved in LICs. This study aims to quantify the number of current and planned LICs at NA schools and to characterize the intended purpose of starting LICs.

METHODS: In 2015, the Clerkship Directors in Internal Medicine (CDIM) survey included a section on LICs. This section queried the number of current and planned LICs, their duration and capacity, and the purpose of starting a new LIC.

RESULTS: The response rate was 76% (94/123) of CDIM schools, representing 67% (94/146) of NA medical schools. Thirty six percent (n = 34/94) of responding schools have ≥1 LIC. Thirty schools have year-long LICs. The majority (22/34) of LICs are <5 years old. Nineteen institutions are planning a new LIC or increasing the number of students participating in LICs, nine of which are schools without existing LICs. CDs believe LICs are being implemented to foster continuity of care, support students’ patient-centeredness, advance interprofessional education, and address workforce shortages. Four schools are converting their entire class to an LIC experience.

CONCLUSIONS: The number and scope of LICs are increasing across NA, especially in the last 5 years. The increase in LICs may be due to early adopters having demonstrated the benefit of LICs, the compatibility with sought-after goals of medical education, the value of tests of change, and the ability to observe the change first before adopting. Applying the theory of diffusion innovation1 suggests that a critical momentum may now exist such that the LIC innovation may now be self-propagating. 1 Rogers EM. Diffusion of Innovations.1995.

figure n

ARE PATIENTS TRANSFERRED TO HOSPITALS THAT CAN APPROPRIATELY TREAT THEM? Stephanie Mueller 1; Jie Zheng3; Endel J. Orav1; Jeffrey L. Schnipper2. 1Brigham and Women, Boston, MA; 2Brigham and Women’s Hospital, Boston, MA; 3TH Chan Harvard School of Public Health, Boston, MA. (Control ID #2703663)

BACKGROUND: Patients are often transferred between hospitals to provide access to specialty care, although receiving hopsitals are not always chosen based solely on patient need. We evaluated the frequency with which patients are transferred to hospitals with availability of required specialty services.

METHODS: We performed a retrospective analysis using 2013 CMS Master Beneficiary and Inpatient claims files merged with 2013 American Hospital Association data. Eligible beneficiaries included those continuously enrolled in Medicare A/B with an acute care hospitalization claim, excluding Medicare managed care and ESRD beneficiaries. We defined transferred patients as those with corresponding “transfer in/out” claims, or either claim with corresponding hospital admission/discharge within 24 hours. We identified top primary diagnoses at time of transfer using ICD-9 codes, selected those requiring specialty services via expert opinion, and used McNemar’s test to compare the availability of each specialty service between receiving and transferring hospitals, stratified by diagnosis.

RESULTS: Of the 101,507 transferred patients, 46,030 (45.3%) had a primary diagnosis requiring a specialty service. For each diagnosis, availability of each corresponding specialty service was more often present at the receiving compared to the transferring hospital (Table), although depending on diagnosis, in 35–85% of transfers, services were just as available at the transferring hospital, and in 7–28% of transfers, the receiving hospitals did not have availability of any corresponding specialty service.

CONCLUSIONS: In this national evaluation of transferred patients, we found patients requiring specialty services are transferred to hospitals more likely to have appropriate services compared to transferring hospitals, implying appropriate transfer. However, exceptions to this rule were frequent, suggesting other possible reasons for transfer in these instances, warranting further investigation into hospital transfer practices.

figure o

ASKING IS NEVER BAD. I WOULD VENTURE ON THAT! : PATIENT PERCEPTIONS OF PAIN SCREENING IN VA PRIMARY CARE Karleen Giannitrapani 4; Marie C. Haverfield3; Roger T. Day3; Jesse Holliday3; Steven Dobscha1; Robert D. Kerns2; Karl Lorenz3. 1Portland VAMC, Portland, OR; 2VA Connecticut, West Haven, CT; 3VA Palo Alto Healthcare System, Menlo Park, CA; 4Veterans Health Administration, Menlo Park, CA. (Control ID #2705833)

BACKGROUND: It is now more than 15 years since the ‘Pain as the 5th Vital Sign’ routine pain screening initiative was implemented in Department of Veterans Affairs (VA) primary care settings. However, the patient perspective on and experience of routine pain screening has not been characterized. The objective of this study is to understand patients’ perspectives on and preferences about pain screening in primary care.

METHODS: We conducted semi-structured qualitative phone interviews with 35 Veterans from five outpatient clinics associated with three large VA medical centers. These were recorded and professionally transcribed; all personally identifying patient information was redacted. Using content analysis methods, we coded transcripts for perceptions of and experiences with routine pain screening in VA primary care. We compared codes across transcripts to identify key themes.

RESULTS: We identified five key themes: 1) Patients affirm it is important to screen for pain because some patients, particularly older ones, may not feel comfortable bringing it up: “if your focus is on managing pain then the question should always come from the provider… to ask a veteran 65-plus to bring up pain… (they) may not be accustomed to talking about their pain because in prior generations that was a sign of weakness.” 2) Patients believe it is important to ask about function in relation to pain. One patient suggested: “ ask has your pain ever inhibited you from completing a task this last week?” if you leave it…open like that… if you say, “a task,” you allow the veteran to determine what level that task is.” 3) Patients want to be screened for pain over being screened for something more broad like quality of life: “I’d rather have the nurse ask me about pain because the nurse can’t do anything more for my quality of life”. 4) Patients recommend asking about pain over a look back period and not just pain now: “when you ask me right then and there I might not be feeling any pain. But the thing that drove me to the clinic might’ve been a 5 or a 6.” 5) In order for it to be an ‘appropriate’ use of clinic visit time patients want to know providers are going to use the information collected in pain screening: “I’m a process guy…the question is what do you do with that information about my pain.”

CONCLUSIONS: Patients endorse the perceived value of pain screening, especially when they perceive pain to be actionable, and because some patients may not report it. Patients highlighted the perceived value of function and chronicity in assessing pain. These perspectives can inform improving the identification and management of pain in primary care.

ASSESSING MOTIVATION AND BARRIERS TO WEIGHT LOSS Christina Fahey; Abigail Lawson; Garrett Oberst; Dylan Woolum; David Rudy; Stephanie A. Rose. University of Kentucky, Lexington, KY. (Control ID #2705367)

BACKGROUND: Obesity and its comorbidities continue to increase in prevalence across the United States. While patients and physicians are increasingly recognizing the importance of weight loss, patients continue to express barriers to losing weight. Understanding these barriers, both medical and personal, can improve care by tailoring weight loss programs to each individual. The aim of this study was to assess barriers to weight loss in patients enrolled in a primary-care based weight loss program, through studying motivations for weight loss and self-efficacy in avoiding situations that trigger poor dietary choices.

METHODS: Patients with a BMI ≥30 kg/m2 or a BMI ≥25 with obesity related comorbidities were invited to enroll in the Physician Weight Management Clinic in a university-based Internal Medicine Clinic. Patients were asked to complete a survey that included validated questions from the Treatment Self-Regulation Questionnaire (TSRQ), which measures motivation (autonomous [AM], versus controlled [CM]) for weight loss, and validated questions from the Weight Efficacy Lifestyle Questionnaire (WEL-Q), which measures self-efficacy in the ability to resist making poor dietary choices in certain situations. All weight loss clinic patients were asked to complete the survey, but only patients who gave consent to be part of the study were analyzed.

RESULTS: Of 114 consented patients, 27 completed the survey. Mean patient AM for weight loss was 6.2, while mean CM was 3.0 (ranges = 1 [low motivation] to 7 [high motivation]). Patients felt most confident in resisting poor dietary choices in situations involving Positive Activities (mean confidence = 6.4), followed by situations involving Physical Discomfort (6.3), situations involving Social Pressures (6.0), situations involving Availability (5.3), and situations involving Negative Emotions (4.6) (range 0 [low confidence] to 9 [high confidence]).

CONCLUSIONS: Of the surveyed patients, we found that internal confidence for losing weight was greater than external motivators. We also found that situations in which patients had the least confidence in resisting poor dietary choices were those associated with negative emotions. Patients had the most confidence in resisting poor choices in situations associated with positive activities. Limitations include generalizability, as we surveyed patients already in a weight loss program. Future goals include comparing individual barriers to weight loss outcomes, administration of a follow-up survey every 3 months to re-assess self-efficacy and motivations, and comparing motivation and confidence between patients enrolled in and not enrolled in a weight loss program. Assessing individual barriers to weight loss is critical in order to tailor weight loss programs appropriately. This research shows the areas patients struggle with the most and thus provides physicians with the best areas in which to focus treatment.

ASSESSING OUTCOMES FOR MEDICAL STUDENTS INVOLVED IN THE EMORY PIPELINE PROGRAM Caroline B. Maness 4; Kevin Luk4, 2; Chidiogo Anyigbo4, 2; Jennifer O. Spicer1; Jordan Rose3; Yolanda Hood4; Robert Lee4. 1Emory University School of Medicine, Decatur, GA; 2Rollins School of Public Health, Atlanta, GA; 3Emory Center for Science Education, Atlanta, GA; 4Emory University School of Medicine, Atlanta, GA. (Control ID #2698482)

BACKGROUND: The Emory University Pipeline Program is a multi-tiered mentorship and educational program comprised of students from five Atlanta Public Schools, Emory College, Rollins School of Public Health, and Emory University School of Medicine that provides a three year health science enrichment curriculum for underprivileged high school students. High School students in the program receive longitudinal mentorship from Emory undergraduates in the areas of college preparation and career interest development. They also receive in-depth exposure to the health science fields through enrichment curricula taught by graduate students. While Pipeline has conducted annual evaluations of high school and undergraduate members’ satisfaction with the program, career skills development, and health science knowledge, Pipeline has not been evaluated in terms of its effect on graduate members’ careers. Specifically, our study sought to assess the impact of Pipeline on medical student alumni’s teaching ability, communications skills, career preparation, and mentorship involvement.

METHODS: We created a 28 question online survey that was sent via email to the 40 alumni who participated in Pipeline while medical students at Emory University School of Medicine as of 2014. The survey assessed what effect participation in Pipeline had in the following domains: teaching ability and comfort, communication skills, career preparation and fulfillment, and mentorship. A portion of survey responses were scored on a 1 (Strongly Disagree) to 5 (Strongly Agree) Likert scale and the second survey portion inquired about continued participation in extracurricular teaching and mentoring. Likert scale responses of 4 or 5 were used to indicate a beneficial impact of Pipeline, and scores of 3 or below were interpreted to indicate that Pipeline had no impact or a negative impact.

RESULTS: Out of 40 alumni who received the survey, 24 began the form and 21 completed all sections. Pipeline alumni reported benefit in their teaching ability in small group settings (100%, n = 23) and lecture settings (91.3%, n = 23), and 78.3% (n = 23) of respondents reported engaging in teaching activities outside of those required on service. 82.6% (n = 23) of alumni reported participating in Pipeline improved their ability to communicate with patients, and 91.3% (n = 23) indicated that Pipeline strengthened their ability to communicate with undergraduates and medical students. Many of the respondents (79.2%, n = 24) agreed that participating in Pipeline helped clarify the types of settings in which they would like to practice. Furthermore, 95.5% (n = 22) of Pipeline alumni endorsed continued participation in mentorship relationships beyond Pipeline.

CONCLUSIONS: The Emory Pipeline Program enhanced medical students’ teaching and communication abilities. Pipeline alumni also continued to participate in high rates with teaching and mentorship activities. Participation in Pipeline programs should be encouraged to facilitate the development of clinician educators.

ASSESSING THE EFFECTIVENESS OF A TRAINING INTERVENTION ON COMMUNITY HEALTH CENTER STAFF’S ABILITY TO IMPLEMENT DIABETES GROUP VISITS Ivana I. Barouhas 2; Sarah P. Hermans4; Erin M. Staab2; Amanda Benitez5; Amanda Campbell1; Cynthia T. Schaefer3; Michael T. Quinn2; Arshiya A. Baig2. 1Midwest Clinicians’ Network, East Lansing, MI; 2University of Chicago, Chicago, IL; 3University of Evansville, Evansville, IN; 4American University, Washington, DC; 5Enlace Chicago, Chicago, IL. (Control ID #2703682)

BACKGROUND: Group visits, which supplement one-on-one medical care with patient education and social support, have been shown to improve outcomes among adults with diabetes. We designed and evaluated the effectiveness of a training intervention to prepare community health center (CHC) staff to implement and sustain diabetes group visits. We also identified facilitators and barriers to the adoption of this care model, our aim being to help guide best practices in its widespread implementation.

METHODS: This 18-month pilot study included monthly webinars and two in-person learning sessions designed to guide the execution of six monthly group visits by twenty-six CHC staff at seven sites within the MidWest Clinicians’ Network. We conducted surveys assessing staff preparedness to implement group visits before Learning Session 1 (LS1), after LS1, and after Learning Session 2 (LS2). Most questions were scored on a scale from 1 = strongly disagree to 5 = strongly agree. Questions asking the degree to which factors were barriers to group visits were scored from 1 = major barrier to 4 = not a barrier. We used paired t-tests to assess for change between pre- and post-LS1 values and between post-LS1 and post-LS2 values. P < 0.05 was considered significant. We performed a qualitative analysis of audio-recorded team check-ins and telephone interviews to identify challenges and successes of the group visit programs.

RESULTS: The surveys showed increased staff awareness of: group visit barriers from 3.74 ± 0.76 (mean ± SD) pre-LS1 to 4.30 ± 0.47 post-LS1 (P = 0.004), and to 4.83 ± 0.38 post-LS2 (P = 0.0005); benefits of the group visit model from 3.96 ± 0.71 pre-LS1 to 4.48 ± 0.51 post-LS1 (P = 0.0001); and key factors for group visit success from 3.00 ± 0.92 pre-LS1 to 4.26 ± 0.45 post-LS1 (P = 0.0000). From post-LS1 to post-LS2, the degree to which transportation, concerns regarding individual medical attention, and concerns regarding patient privacy impeded group visits improved from: 2.04 ± 0.76 to 2.50 ± 0.79 (P = 0.0144), from 3.00 ± 0.68 to 3.67 ± 0.49 (P = 0.0036), and from 3.00 ± 0.83 to 3.56 ± 0.51 (P = 0.0014), respectively. General preparedness to conduct group visits increased from 3.78 ± 0.85 pre-LS1 to 4.17 ± 0.39 post-LS1 (P = 0.01). Group visit facilitators included: thorough preparation and knowledge about the care model; a motivated team; experienced diabetes educators; supportive CHC leadership; and adequate space for the visits. Common barriers included patient barriers (recruitment, transportation, motivation), and staff barriers (time limitations).

CONCLUSIONS: While CHC staff noted several barriers to the implementation of diabetes group visits, our training intervention increased staff confidence, preparedness, and awareness of key group visit elements. Facilitators to sustaining group visits included preparation-, team-, and facility-related factors. Future research is needed to assess our training program in a larger sample of CHC sites and assess the impact of diabetes group visits on patient outcomes.

ASSESSING UTILIZATION OF PRIMARY CARE PRACTICES BY PATIENTS WITH SELF-REPORTED EXCESSIVE ALCOHOL USE Dennis Keselman 1; Jeanne Morley2, 1; Laura Harrison3; Linda DeMasi3; Vladislav Fomin1; Megan O’Grady4; Nancy Kwon3, 1; Jonathan Morgenstern2, 1; Joseph Conigliaro2, 1; Sandeep Kapoor3, 1. 1Hofstra Northwell School of Medicine, Hempstead, NY; 2Northwell Health, Great Neck, NY; 3Northwell Health, New Hyde Park, NY; 4The National Center on Addiction and Substance Abuse, New York, NY. (Control ID #2705317)

BACKGROUND: There is widespread consensus that screening patients for alcohol misuse and providing brief interventions in primary care settings could lead to improvements in clinical outcomes through early identification, reduction in alcohol misuse, and reduction of alcohol related comorbidities. This would explain the healthcare cost savings seen with SBIRT interventions. However, there are no large-scale data on the primary care utilization patterns of patients who potentially misuse alcohol. The Screening, Brief Intervention, and Referral to Treatment (SBIRT) program implemented at a Level-III Patient-Centered Medical Home incorporates universal screening of patients for potential alcohol misuse, and data from the alcohol screens, together with demographic data, can be used to better understand the patterns of primary care utilization in specific populations.

METHODS: Patients were screened by frontline staff for potential alcohol misuse during all appointments. Patients who screened positive for potential alcohol misuse (n = 998), along with a cohort of patients of similar demographics who screened negative (n = 968) between 7/1/15 and 6/30/16 were included in a chart review. The chart review obtained data from AllScript Electronic Health Record about each patient’s SBIRT alcohol “pre-screen” score using the AUDIT-C standardized tool, demographic information, insurance type, appointment type, number of visits to the PCMH, and the total scheduled appointment time for all their visits to the PCMH. The data was inputted into a HIPPA-certified database system (REDcap) and then analyzed using SPSS and SAS Studio.

RESULTS: Patients that pre-screened positive for potential alcohol misuse used primary care less (mean = 87.79 min and 2.50 visits) than the control group of patients who pre-screened negative (mean = 110.33 min and 3.44 visits) over the 1-year period (p < 0.0001). Hispanic/Latino, African-America/Black, middle aged (35–64 years), uninsured, and Medicaid patients have the largest disparities in primary care utilization between the pre-screen positive and negative groups (p < 0.01).

CONCLUSIONS: This study shows that patients who drink at higher-risk levels utilize primary care services less, as opposed to more. Hispanic/Latino, African-American/Black, middle-aged, uninsured, and Medicaid patients have the largest disparities in primary utilization between the pre-screen positive and negative groups. This underutilization highlights the importance of identifying patients in need of services at every potential touch point to maximize opportunity for intervention, as well as enhancing outreach and patient engagement. Understanding these primary care utilization patterns will better illustrate future directions for improving care for different patient populations, including those at risk for alcohol misuse.

ASSESSMENT OF A CURRICULUM ON INTIMATE PARTNER VIOLENCE (IPV) FOR INTERNAL MEDICINE RESIDENTS Emily R. Insetta 1, 2; Colleen Christmas2, 1. 1Johns Hopkins Bayview Medical Center, Baltimore, MD; 2Johns Hopkins University, Baltimore, MD. (Control ID #2701490)

BACKGROUND: Intimate partner violence (IPV) is a widespread problem in the US, affecting 1 in 4 women and 1 in 7 men. IPV is associated with detrimental physical, mental, and behavioral health consequences. However, there are no standardized IPV requirements in medical education. Studies show that physicians in training and practice have limited IPV knowledge and desire more education. Existing curricula for medical students and practicing physicians have improved IPV knowledge and screening. There are little data for IPV education in residency, particularly for internal medicine residents who are primary care providers during training and beyond. This study aimed to evaluate a curriculum designed to improve residents’ knowledge, attitudes, and practices in caring for patients who are IPV victims.

METHODS: 15 first-year internal medicine residents at Johns Hopkins-Bayview participated in two, 1-hour classes during July and August of 2016. The first class included a speaker with IPV experience, a case discussion, and a didactic presentation. The second part reviewed evidence for IPV interventions and focused on patient-doctor communication using role plays. Data were collected via voluntary, pre- and post-curriculum surveys with questions adapted from the validated tool, PREMIS (Physician Readiness to Manage Intimate Partner Violence Survey). Changes in IPV knowledge, attitudes, and practices were compared using 2-tailed t-tests. Additional responses evaluating the curriculum in the post-test were described.

RESULTS: 15 first year residents participated in the baseline survey and 12 participated after the educational intervention. IPV knowledge was high at baseline and did not improve except for one question about health consequences of IPV (p = 0.032). Assessment of attitudes showed an existing recognition of IPV and its health impact, which did not change. The curriculum produced significant changes in self-efficacy including improved confidence in detecting IPV (p = 0.022), documenting IPV (p = 0.000035), and referring to resources (p < 0.00001). Participants reported increased comfort with managing their emotions regarding IPV (p = 0.0080) and discussing IPV with female (<0.00001) and male (p = 0.021) patients. Self-reported frequency of IPV screening for females remained the same (p = 0.45) but improved for male patients (p = 0.033). All participants agreed or strongly agreed that after the curriculum they would be more skillful in discussing IPV, more likely to screen for IPV, and better equipped to refer to IPV resources.

CONCLUSIONS: A curriculum for residents improved their confidence and comfort in addressing IPV. After training, all participants felt they were more prepared to screen for IPV, discuss IPV, and refer to resources. Incorporating IPV curricula into residency may promote greater detection of IPV and better care for victims.

ASSESSMENT OF ADHERENCE TO DEPRESSION MANAGEMENT GUIDELINES USING UNANNOUNCED STANDARDIZED PATIENTS: ARE RESIDENT PHYSICIANS EFFECTIVELY MANAGING DEPRESSION IN PRIMARY CARE? Sondra Zabar 1; Kathleen Hanley1, 3; Amanda Watsula-Morley1; Lisa Altshuler1; Heather Dumorne1; Andrew B. Wallach2, 3; Barbara Porter2, 3; Adina Kalet1; Colleen Gillespie1. 1NYU School of Medicine, New York, NY; 2Bellevue Hospital, New York, NY; 3Gouverneur Healthcare Services, New York, NY. (Control ID #2705368)

BACKGROUND: All physicians need to be skilled at diagnosing, treating, and managing depression. We designed an unannounced standard patient (USP) case to assess residents’ clinical skills in addressing depression and explored how those skills are associated with residents’ general clinical skills in order to design targeted curriculum on depression.

METHODS: The USP was a 26 y.o. male presenting as a new patient to a clinic complaining of fatigue and problems sleeping. Goals of the case were to diagnose a common presentation of depression and make a treatment/follow-up plan. The USP was trained to have a positive PHQ 2 & PHQ 9, family history of depression, and be willing to engage in medication and/or therapy if offered. A post-visit checklist was used by the SPs to assess communication, patient education, and assessment skills using behaviorally anchored items rated as not done, partly done, or well done. A systematic chart review was conducted to examine treatment, quality of documentation, and referrals. Case fidelity was checked by audiotape and confirmed by PHQ 9 score in the EHR. Evidence-based treatment was defined as prescribing an SSRI and/or providing a psychiatric referral; if neither of those, scheduling follow-up for within 2 weeks.

RESULTS: 122 residents saw the USP case from 2009–2015. Mean visit length = 45 min, SD 25 (14 to 183 min). The patient was screened for depression with a PHQ 2 in 93% of visits; 82% also had a PHQ 9. Overall, 77 residents (63%) provided appropriate treatment: 8% prescribed an SRRI, 23% provided a referral, 19% did both, 7% prescribed a sleep aid and <2 week follow-up, and 43% provided a combination of these treatments. 45 residents (37%) did not provide appropriate treatment: 27 (60%) prescribed a sleep aid and follow-up >2 weeks and 18 (40%) provided no treatment/referral and follow-up >2 weeks. There were no differences in exploration of medical history or substance use, but 83% of residents who treated appropriately had a PHQ 9 compared to 62% of residents who did not treat appropriately. 71% also included depression on the problem list compared to 13% of residents who did not treat appropriately. Residents who treated appropriately had significantly better clinical skills assessed by the USP including: overall communication (71% vs. 54%, p = 0.00), information gathering (72% vs. 55%, p = 0.01), relationship development (75% vs. 60%, p = 0.03), patient education (55% vs. 21%, p = 0.00), and patient activation skills (33% vs. 13%, p = 0.01).

CONCLUSIONS: Although almost all residents obtained the relevant information, only about 50% of residents diagnosed depression. PHQ 9 appears to be associated with providing more effective treatment, supporting the importance of health system screening protocols. Residents’ communication and depression-specific patient education and activation skills seem to be related to how they identify and manage depression, suggesting that interventions to build these skills may lead to higher quality care.

ASSESSMENT OF SELF-REPORTED MEDICATION ADHERENCE TO INHALED COTICOSTEROIDS AMONG ADULTS WITH ASTHMA AND CHRONIC OBSTRUCTIVE PULMONARY DISEASE. Neelima C. Tangirala 2; Li Chen2; Michael S. Wolf4; Rachel O’Conor4; Juan P. Wisnivesky3; Alex Federman1. 1Icahn School of Medicine at Mount Sinai, New York, NY; 2Mount Sinai Health System, New York, NY; 3Mount Sinai School of Medicine, New York, NY; 4Northwestern University, Chicago, IL. (Control ID #2707002)

BACKGROUND: Regular use of inhaled corticosteroids (ICS) is the standard of care for patients with persistent asthma and chronic obstructive pulmonary disease (COPD). In research, adherence to ICS is often measured using the 10-item Medication Adherence Report Scale (MARS), a self-reported medication adherence assessment designed to minimize social desirability bias. However, there are limited data on the validity of this measure.

METHODS: Data for this analysis were obtained from the baseline interviews from two cohort studies that examined the association of health literacy with self-management behaviors among adults ages 65 and older with asthma (n = 452) and adults ages 40 and older with COPD (n = 393). Patients for both studies were recruited from primary care practices in New York City and Chicago. In both studies, research assistants administered the MARS in English or Spanish as part of the baseline assessment. ICS adherence was also objectively measured over a 4-week period, within 3 months of the baseline interview, using electronic monitoring devices that record every actuation of the ICS inhaler, both metered dose and dry powdered devices. Adequate adherence by MARS assessment was defined as a score ≥4.5 based on prior literature, and by electronic monitoring as actuation of ≥80% of doses prescribed. We assessed the criterion validity using correlations between self-reported adherence and electronic adherence. Receiver Operating Characteristic (ROC) curve analysis of the continuous value for MARS scores against the dichotomized outcome of adherence as measured electronically was also performed.

RESULTS: In the study of asthma patients, 78% were prescribed ICS medications and MARS and electronic adherence data were available for 75 and 53% of all patients respectively. In the study of COPD patients, 30% were prescribed ICS medications and MARS and electronic adherence data were available for 87 and 24% of all patients respectively. There were no significant differences between patients missing MARS or electronic data and those with complete data for either study. Among patients with asthma, the continuous values for adherence measured by self-report and electronically were weakly correlated (r = 0.33, p < 0.001); similarly, the agreement between the discrete measures was weak (kappa 0.30, p < 0.001). Findings were similar for COPD patients: r = 0.29, p = 0.003; kappa 0.27, p = 0.005. Area under curve (AUC) values generated from ROC analysis was 0.68 and 0.69, for asthma and COPD patients, respectively.

CONCLUSIONS: Commonly used measure for adherence performed weakly compared to electronic monitoring in separate populations of patients with asthma and COPD. Investigators measuring self-reported medication adherence among patients with these pulmonary diseases should consider using alternative instruments or using objective measures exclusively

ASSESSMENT OF THE SAFETY AND TOLERABILITY OF THE ANTIBIOTIC RIFAXIMIN FOR IRRITABLE BOWEL SYNDROME (IBS): A POOLED ANALYSIS OF THREE PHASE 3 STUDIES Christopher Chang 4; Mark Pimentel1; Zeev Heimanson3; Brian E. Lacy2. 1Cedars-Sinai Medical Center, Los Angeles, CA; 2Dartmouth-Hitchcock Medical Center, Lebanon, NH; 3Salix Pharmaceuticals, Calabasas, CA; 4University of New Mexico School of Medicine, Albuquerque, NM. (Control ID #2699361)

BACKGROUND: IBS prevalence in North America is approximately 12%, and studies suggest females are almost twice as likely to have IBS as males. Substantial differences in the gut microbiota (eg, composition and diversity) have been observed in patients with IBS vs healthy individuals. The nonsystemic antibiotic rifaximin 550 mg, taken 3 times daily for 14 days, is indicated for diarrhea-predominant irritable bowel syndrome (IBS-D) in adults. Given that IBS symptoms may be episodic or chronic, patients may require repeated courses of therapy. To gain a better understanding of the rifaximin safety profile, a pooled analysis evaluated safety and tolerability of rifaximin in IBS.

METHODS: This was a post hoc pooled analysis of two phase 3 studies (Trials 1 and 2; rifaximin 550 mg or placebo 3 times daily for 2 weeks) and double-blind retreatment phase of a phase 3 study (Trial 3; 2 courses of rifaximin 550 mg or placebo 3 times daily for 2 weeks, with courses separated by 10 weeks). Safety was assessed through 10 weeks posttreatment in Trials 1 and 2, and through 4 weeks posttreatment after second treatment course in Trial 3. Analysis included patients with ≥1 dose of study medication and ≥1 postbaseline safety assessment.

RESULTS: The analysis included 952 rifaximin-treated and 942 placebo-treated patients. Most AEs were mild to moderate in intensity. The most common AEs (≥2% of patients in rifaximin group) were upper respiratory tract infection, headache, and nausea (Table).. Increased alanine aminotransferase levels were reported in 1.6% (rifaximin) and 1.2% (placebo) of patients. Constipation was reported in 0.7% (rifaximin) and 1.6% (placebo) of patients. Only 1 serious AE in rifaximin group (alcohol withdrawal syndrome) was considered related to study drug. One patient developed a serious AE of Clostridium difficile posttreatment. This patient had a history of C difficile and received a 10-day course of cefdinir for a urinary tract infection; the AE was not considered to be drug-related but likely due to cefdinir exposure.

CONCLUSIONS: This pooled safety analysis of 952 rifaximin-treated patients supports the safety and tolerability profile of rifaximin in an adult population with IBS.

Table. AE Summary

Parameter, n (%) Rifaximin (n = 952) Placebo (n = 942)
Any AE 480 (50.4) 477 (50.6)
Severe AEs 39 (4.1) 59 (6.3)
Drug-related AEs 81 (8.5) 109 (11.6)
Serious AEs 14 (1.5) 19 (2.0)
AEs leading to study discontinuation 9 (0.9) 9 (1.0)
Upper respiratory tract infection 47 (4.9) 47 (5.0)
Headache 42 (4.4) 51 (5.4)
Nausea 39 (4.1) 31 (3.3)
Diarrhea 34 (3.6) 25 (2.7)
Abdominal pain 32 (3.4) 37 (3.9)
Nasopharyngitis 29 (3.0) 43 (4.6)
Sinusitis 24 (2.5) 23 (2.4)
Urinary tract infection 23 (2.4) 26 (2.8)
Bronchitis 22 (2.3) 22 (2.3)

ASSOCIATION BETWEEN GENDER MINORITY STATUS AND SELF-REPORTED MENTAL AND PHYSICAL HEALTH IN THE U.S. Carl G. Streed 3; Ellen P. McCarthy2; Jennifer Haas1. 1BWH, Boston, MA; 2Beth Israel Deaconess Medical Center, Harvard Medical School, Boston, MA; 3Brigham & Women’s Hospital, Boston, MA. (Control ID #2703097)

BACKGROUND: Gender minorities, including transgender and gender non-conforming individuals, experience disparities in mental and physical health. Prior investigations are limited in size and geographic scope due to the lack of national survey data on gender identity.

METHODS: We pooled 2014 and 2015 Behavioral Risk Factor Surveillance System (BRFSS) data and analyzed measures of participant-reported mental health and physical function among gender minority adults (transgender 1,173, representing an estimated 727,983 adults; gender non-conforming 270, representing 184,038 adults) compared to cisgender peers (314,450 representing 162,400,000). We used bivariable and multivariable analyses to identify differences by gender minority status. We performed logistic regression to assess the effects of age, education, race, employment status, and income. We performed analyses using SAS 9.4 and utilized survey procedures to account for the complex sampling design and were weighted to reflect national estimates.

RESULTS: Compared to cisgender adults, gender minority adult respondents are younger (19.2% vs 24.4% under 30 years-old), less likely to be white (69.9% vs 56.7% white, non-Hispanic), less likely to be employed (57.2% vs 50.8% employed) and have lower annual household incomes (15% vs 22.1% less than $20,000; all p <0.001). After adjusting, gender minority participants are at increased odds compared to cisgender peers of self-reported fair or poor health and self-reported serious difficulty concentrating, remembering, or making decisions. In sub-group analyses, transgender respondents are at increased odds compared to cisgender peers of self-reported serious difficulty concentrating, remembering, or making decisions. Gender non-conforming respondents are at increased odds compared to cisgender peers of self-reported fair or poor health, self-reported serious difficulty concentrating, remembering, or making decisions, and self-reported limitation in any way because of physical, mental, or emotional problems (all p < 0.001).

CONCLUSIONS: Significant self-reported mental and physical health disparities exist for transgender and gender non-conforming adults in the United States; these disparities persist even when adjusting for known protective factors, such as employment or income. These disparities in self-reported mental and physical health beg the need for additional research into potentially unmeasured stressors transgender and gender non-conforming people experience. In addressing mental and physical health, clinicians should be sensitive to the need to be aware of the unique needs of gender minorities.

Adjusted Odds Ratio of Gender Minority Status and Self-Reported Mental and Physical Health Outcomes

  Gender Minorities OR [95% CI] Sub-Group Analyses
Transgender (MTF or FTM) cOR [95% CI] Gender Non-Conforming OR [95% CI]
Fair/Poor Health 1.25 [1.10, 1.42]* 1.13 [0.98, 1.31] 1.89 [1.43, 2.49]*
Difficulty concentrating, remembering, or making decisions 1.51 [1.30, 1.75]* 1.48 [1.19, 1.66]* 1.97 [1.44, 2.69]*
Limitation in any way because of physical, mental, or emotional problems 1.13 [1.00, 1.28] 1.04 [0.91, 1.19] 1.56 [1.22, 2.07]*

OR adjusted Odds Ratio; CI Confidence Interval; MTF male-to-female; FTM female-to-male

* p < 0.001

ASSOCIATION BETWEEN PURPOSE IN LIFE AND GLUCOSE CONTROL AMONG OLDER ADULTS Dina H. Griauzde 2, 3; Michele Heisler3; HwaJung Choi2; Claire Ankuda2; Tyler N. Winkelman4, 3; Jeff Kullgren1. 1Ann Arbor VA Healthcare System and University of Michigan, Ann Arbor, MI; 2University of Michigan, Ann Arbor, MI; 3Ann Arbor VA/University of Michigan, Ann Arbor, MI; 4University of Michigan, St. Paul, MN. (Control ID #2707330)

BACKGROUND: Greater purpose in life is associated with lower rates of certain chronic conditions such as cardiovascular disease and stroke. Whether purpose in life can protect against development of prediabetes or type 2 diabetes is unknown. The objectives of our study were to examine associations between purpose in life and subsequent (1) change in hemoglobin A1c (HbA1c) and (2) development of clinically meaningful categories of abnormal glucose metabolism (i.e., prediabetes or type 2 diabetes) among US adults. We hypothesized that greater levels of purpose in life would be associated with smaller increases in HbA1c and lower rates of developing abnormal glucose metabolism compared to lower levels of purpose in life.

METHODS: We conducted a national study of 3,907 adults age 50 and older who did not have type 2 diabetes or prediabetes and were participants in the Health and Retirement Study. Baseline purpose in life was measured using a 7-item, validated adaptation of Ryff and Keyes’ Scales of Psychological Well-Being and grouped into tertiles (high, medium, low). We used multivariable linear regression to examine the association between baseline purpose in life and change in HbA1c from baseline to 4-year follow-up. Multivariable logistic regression was used to examine the association between baseline purpose and incident prediabetes or type 2 diabetes over the same period.

RESULTS: After adjusting for sociodemographic factors, body mass index, physical activity, and physical and mental health factors, change in HbA1c was 0.07 percentage points lower among participants with high purpose than those with low purpose (95% CI −0.12 to −0.02; p = 0.011). Participants with high purpose had lower odds of developing prediabetes or type 2 diabetes than those with low purpose (adjusted odds ratio 0.78; 95% CI 0.62 to 0.98; p = 0.037).

CONCLUSIONS: Among older adults, greater purpose in life is associated with a lower incidence of prediabetes or type 2 diabetes. Strategies to promote greater purpose in life should be tested as part of type 2 diabetes prevention efforts.

ASSOCIATION OF CITIZENSHIP STATUS WITH KIDNEY TRANSPLANTATION IN PATIENTS ON MEDICAID Jenny I. Shen 1, 2; Daniel Hercz2; Lilly Barba1, 2; Holly Wilhalme2; Erik L. Lum2; Edmund Huang2; Leslie K. Salas1; Sitaram Vangala2; Keith Norris2. 1LaBiomed at Harbor-UCLA Medical Center, Torrance, CA; 2Medicine, University of California, Los Angeles, Los Angeles, CA. (Control ID #2692159)

BACKGROUND: Although undocumented immigrants can receive emergency dialysis in the US regardless of their ability to pay, most states do not provide them with subsidized care for renal transplantation, a more cost-effective form of renal replacement therapy that provides better outcomes for patients. The objective of the study was to determine whether undocumented immigrants have similar outcomes to U.S. citizens after receiving kidney transplants covered by Medicaid.

METHODS: In this observational cohort study, we identified from the United States Renal Data System and followed for 5 years all adult patients on Medicaid who received their first kidney transplant from 1990–2011 and categorized them by citizenship status. We applied Cox proportional hazards frailty models with transplant center as a random effect to estimate the hazard ratios (HR) and 95% confidence intervals (CI) for all-cause graft loss.

RESULTS: Of 10,495 patients, 8660 (82%) were U.S. citizens, 1489 (14%) were permanent residents, and 346 (3%) were undocumented immigrants. Undocumented immigrants were younger, healthier, on dialysis longer, and more likely to have had a living donor. 71% were transplanted in California, and 61% were transplanted after 2005. We identified 42 graft failures in undocumented immigrants (3.7/100 person-years) and 2,445 in U.S. citizens (7.9/100 person-years). Undocumented immigrants had a lower unadjusted risk of graft loss compared to US citizens (HR 0.48, 95% CI: 0.35–0.65). Results were attenuated but still significant when adjusted for demographics, comorbidities, dialysis, and transplant-related factors (HR 0.67, 95% CI: 0.46–0.94).

CONCLUSIONS: Only 3% of patients with Medicaid transplanted in the US were undocumented immigrants. The transplants occurred mostly in California, have increased in recent years, and our study suggests that they do no worse, and perhaps even better, than US citizens. Policymakers should consider expanding coverage for kidney transplantation in undocumented immigrants is it leads to high quality outcomes while lowering the overall costs of renal replacement therapy in these patients.

ASSOCIATION OF HYPERTENSION AMONG SELF-REPORTED BINGE DRINKERS IN A PATIENT-CENTERED MEDICAL HOME WITH A SCREENING, BRIEF INTERVENTION, AND REFERRAL TO TREATMENT PROGRAM Vladislav Fomin 1; Jeanne Morley2, 1; Laura Harrison3; Linda DeMasi3; Dennis Keselman1; Megan O’Grady4; Nancy Kwon3, 1; Jonathan Morgenstern2, 1; Joseph Conigliaro2, 1; Sandeep Kapoor3, 1. 1Hofstra Northwell School of Medicine, Hempstead, NY; 2Northwell Health, Great Neck, NY; 3Northwell Health, New Hyde Park, NY; 4The National Center on Addiction and Substance Abuse, New York, NY. (Control ID #2705251)

BACKGROUND: 70 million adults in the United states have hypertension, of which only half is properly controlled. Binge drinking has been found to be associated with a modest increase in blood pressure and in certain populations, with an increased prevalence of hypertension. There is not an established link between self-reported binge drinking and prevalence of hypertension in a diverse urban population. This study sought to compare self-reported drinking patterns with blood pressure and hypertension status at the time of visit.

METHODS: The Screening, Brief Intervention, and Referral to Treatment (SBIRT) protocol has been implemented within the workflow of a Level-III Patient-Centered Medical Home, in which every patient is given a standardized three question pre-screen (AUDIT-C) to evaluate their drinking patterns. We used data from this pre-screen in connection with a linked AllScripts Electronic Health Record to collect demographic and patient visit data. 998 patients who screened positive for potential alcohol misuse over the past year (7/1/15-6/30/16) were controlled matched based on age, gender, race and ethnicity (n = 968). AllScripts Electronic Health Records were linked to the patients’ responses to the pre-screen questionnaire and were used to collect patient data at time of visit. Hypertensive status was qualified as either having a systolic blood pressure greater than 140, having a previous diagnosis of hypertension, or being currently treated for hypertension. Data was analyzed using SPSS software. The primary statistical tool used was a logistic regression with covariates.

RESULTS: When adjusted for age, BMI, and tobacco usage, there was no association found between the rate of hypertension and binge drinking compared with those who screened negative for potential alcohol misuse (p >0.05). Similarly, there was no association found between hypertension and those who screened positive for potential alcohol misuse when compared to those who screened negative (p > 0.05). When compared to non-binge drinkers, binge drinkers had an average of 3.7 units (mmHg) higher systolic blood pressure with no significant difference in diastolic blood pressure (p > 0.05). When compared to those who screened negative for potential of alcohol misuse those who screened positive had an average of 2.5 units (mmHg) higher systolic blood pressure and 1.4 units higher diastolic blood pressure (p < 0.05).

CONCLUSIONS: This study illustrates that self-reported binge drinking and alcohol consumption over healthy NIAAA guidelines did not correlate with a diagnoses of hypertension in our population of patients. The modest trend towards increased blood pressure highlights the importance of understanding a patient’s alcohol consumption when addressing, counseling and managing elevations in blood pressure. Irrespective of diagnosis of HTN, alcohol misuse is associated with higher BP’s which may have long term implications if treated early.

ASSOCIATION OF PLANT-BASED PROTEIN WITH SLEEP QUALITY AND DURATION IN WOMEN Allison Crawford 1; Brooke Aggarwal1; Heidi M. Greenberger1; Ming Liao2; Marie-Pierre St-Onge1. 1Columbia University, New York, NY; 2Columbia University Medical Center, NEW YORK, NY. (Control ID #2705746)

BACKGROUND: Experimental evidence has demonstrated that diet composition affects sleep; however, examinations in a non-experimental context are few. The purpose of this study was to assess the relation between sleep duration and quality and plant-based dietary protein intakes.

METHODS: Women, age 20–75 (n = 106), were recruited from an urban academic health center. Study population characteristics were as follows: mean age 40+/−18 years, 61% white, 83% employed, 93% college degree, 24% current/former smokers, 61% chronic disease, and 44% BMI <25 kg/m2. Participants taking sleep medications (n = 20) and/or participants with missing data (n = 7) were excluded from the analyses; analytic population n = 82. Sleep quality and duration were assessed using the Pittsburgh Sleep Quality Index (PSQI) and the Insomnia Severity Index (ISI). Dietary history was obtained using the Block Brief Food Frequency Questionnaire (FFQ). Food items were coded as plant or animal based. Total protein was calculated for each food using the USDA Food Composition Database. The sum of daily protein from plant-based foods was calculated as a percentage of total daily caloric intake. The proportion of energy consumed from protein included only food items coded as plant-based in this analysis. The associations between the proportion of plant-based protein energy intake and sleep variables were assessed using logistic and linear regression models. Logistic regression models included sample medians for plant-protein intake (≥4% vs. <4) and ISI score (<7 vs ≥7); recommended cut points were used for sleep duration (<7 vs. ≥7 h) and sleep quality (PSQI score <5 vs ≥5). Multivariate adjustment included age, race, education, income, and smoking status.

RESULTS: Sample mean sleep duration was 6.70+/−1.35 hours, sleep quality scores 5.50+/−3.74, and mean plant-protein daily consumption 4.07%+/−1.51%. In bivariate linear models, percentage of plant-based protein intake was positively associated with sleep duration (ß = 0.22, p = 0.04). Higher percentage of plant-based protein intake was also associated with better sleep quality (ß = −0.571, p = 0.02), and lower insomnia (ß = −0.87, p = 0.046). In bivariate logistic models, consuming ≥4% of energy from plant protein was associated with lower odds of short sleep duration (OR 0.27, CI 0.10–0.74, p = 0.01) and there was a trend for higher sleep quality (OR 0.41, p = 0.07). In multivariate adjusted logistic models, the association of plant-based protein intake with sleep duration was statistically significant (p =0.01) and marginally significant for sleep quality (p > 0.05/< = 0.11) and in the expected direction for both.

CONCLUSIONS: Plant-based protein consumption in women was associated with sleep duration and possibly with better sleep quality. Completion of cohort recruitment will increase statistical power to test these associations. Limitations include possible confounding for unmeasured covariates.

ASSOCIATION OF STRUCTURED ASYNCHRONOUS VIRTUAL VISITS FOR HYPERTENSION FOLLOW-UP IN PRIMARY CARE WITH BLOOD PRESSURE CONTROL AND USE OF CLINICAL SERVICES David M. Levine 2; Ronald F. Dixon3; Jeffrey A. Linder1. 1Brigham and Women’s Hospital, Boston, MA; 2Brigham and Women’s Hospital and Harvard Medical School, Boston, MA; 3Massachusetts General Hospital and Harvard Medical School,, Boston, MA. (Control ID #2704278)

BACKGROUND: Optimal management of hypertension requires frequent monitoring and follow-up. Novel, pragmatic interventions are needed to engage patients, maintain blood pressure control, and enhance access to busy primary care practices. “Virtual visits” are structured asynchronous online interactions between patients and their primary care clinicians to extend medical care beyond the initial office visit. Virtual visits have high patient and clinician satisfaction, but clinical outcomes and subsequent utilization of primary and other health care are unknown.

METHODS: We studied patients presenting to primary care practices with a principle diagnosis of hypertension at a large health care network in the Northeast United States. The exposure group was every patient who had engaged in a hypertension virtual visit. After a visit where hypertension was discussed, clinicians could ask their patients to follow-up online. A patient entered blood pressure, adherence, and other items. Once submitted, the clinician responded with treatment decisions. The usual care group was drawn from all patients who presented to a primary care office for hypertension in the same network, but where virtual visits were not yet available. For both groups, the “pre-visit” and “post-visit” time periods were 180 days before and after the index visit. We used logistic regression to create a propensity score of receiving a virtual visit. We propensity-score matched on age, gender, language, partner status, veteran status, smoking, race/ethnicity, health insurance, total chronic conditions, mean pre-visit systolic blood pressure (SBP), total pre-visit antihypertensive medications, and pre-visit primary care visits, specialist visits, emergency visits, and inpatient admissions. We developed multivariable linear and negative binomial regression models to examine the difference-in-differences in mean SBP change, primary care visits, specialist visits, emergency department visits, and inpatient admissions from the pre to the post period between virtual visit and usual care patients.

RESULTS: Of the 1,051 virtual visit patients and 35,050 usual care patients, we propensity-score matched 878 patients from each group. Both groups were 61 years old, 43% female, 85% White, and had about 5 chronic conditions. Virtual visit patients completed a mean of 2.5 (95% CI, 2.3 to 2.7) virtual visit encounters. Compared to usual care, virtual visit patients had an adjusted 0.9 (95% CI, 0.4 to 1.5) fewer primary care office visits and 0.4 (95% CI, 0.03 to 0.7) fewer specialist visits. There was no significant adjusted difference in SBP control (0.6 mmHg [95% CI, −2.0 to 3.1]), emergency department visits (0.0 more visits [95% CI, 0.0 to 0.01]), or inpatient admissions (0.0 more visits [95% CI, 0.0 to 0.1]).

CONCLUSIONS: In a propensity-matched cohort of primary care patients with hypertension, virtual visit participation was associated with equivalent blood pressure control and reduced in-office primary and specialty care utilization.

ASSOCIATION OF TRIGLYCERIDES TO HDL RATIO WITH CARDIOMETABOLIC DISEASE OUTCOMES May Yang; Joseph Rigdon; Sandra A. Tsai. Stanford University School of Medicine, Stanford, CA. (Control ID #2701588)

BACKGROUND: In the US annually, cardiovascular disease (CVD) is the leading cause of death for men and women and roughly 1.4 million people receive a new diagnosis of diabetes. Due to the importance of early intervention, accurately assessing patients’ risk for cardiometabolic disease, such as CVD and diabetes, may be clinically beneficial. An elevated plasma triglyceride (TG)/high-density lipoprotein cholesterol (HDL-C) ratio reflects insulin resistance and has been associated with a higher cardiometabolic risk in European and Asian populations. Whether TG/HDL-C predicts CVD is less understood. A database of ethnically diverse patients in an academic center was used to study the association between TG/HDL-C and cardiometabolic disease outcomes.

METHODS: The Stanford Translational Research Integrated Database Environment (STRIDE) identified patients who were over the age of 50 with available data from January 2008 through December 2015. Demographic information and laboratory results were extracted from the database. Data on cardiovascular risk factors and CVD and diabetes diagnoses were collected through chart review. Using a TG/HDL-C ratio measured in 2008, high-risk patients were identified based on previously studied cutoffs (>2.5 for females, >3.5 for males). Logistic regression modeling with adjustments for relevant confounders was used to estimate associations between risk category and diagnosis of CVD or diabetes in follow-up.

RESULTS: One thousand patient charts were reviewed, 29.5% of which were categorized as high-risk by TG/HDL-C ratio. Of the 762 subjects who did not have a diagnosis of diabetes in 2008, 26.1% were categorized as high-risk by TG/HDL-C. There were 14.6% of high-risk TG/HDL-C subjects and 7.1% of low-risk TG/HDL-C subjects diagnosed with diabetes in follow-up. After adjusting for cofounders, high-risk TG/HDL-C ratio was associated with a diabetes diagnosis in follow-up (OR = 1.61, 95% CI: 0.93–2.78) with borderline significance. Asian ancestry significantly increased the risk of diabetes diagnosis in follow-up (OR = 4.82, 95% CI: 2.4–9.7). Of the 772 patients free of CVD diagnoses in 2008, 28.6% were categorized as high-risk by TG/HDL-C. There were 15.4% of high-risk TG/HDL-C subjects and 7.4% of low-risk TG/HDL-C subjects diagnosed with CVD in follow-up. After adjusting for confounders, high-risk TG/HDL-C ratio significantly increased the risk of CVD diagnosis in follow-up (OR = 1.79, 95% CI: 1.03–3.1).

CONCLUSIONS: TG/HDL-C ratio can be used in conjunction with other cardiometabolic risk factors such as body mass index (BMI) to determine risk of diabetes and CVD. The logistic regression findings further suggest that alternative TG/HDL-C criteria may be necessary to identify Asian patients most at risk for diabetes.


Laura Hawks 6; Maria R. Khan2; Kathleen A. McGinnis3; Joy D. Scheidell2; Christian Grov1; Amy C. Justice5; Emily A. Wang4. 1CUNY School of Public Health, New York, NY; 2New York University School of Medicine, New York, NY; 3VA Pittsburgh Healthcare System, Pittsburgh, PA; 4Yale School of Medicine, New Haven, CT; 5Yale University, West Haven, CT; 6Yale University, New Haven, CT. (Control ID #2708548)

BACKGROUND: While new HIV infections have decreased in recent years, new infections are not decreasing among men who have sex with men (MSM), especially MSM of color. Studies have shown that, in general, people at risk for and living with HIV have increased interaction with the criminal justice system than the general population, but little is known about the relationship between criminal justice history and HIV risk among MSM. This study seeks to measure associations between recent and prior history of incarceration and HIV risk behavior and viral load in HIV-positive MSM.

METHODS: The Veterans Aging Cohort Study Survey Substudy (VACS Survey) is an observational cohort of veterans living with HIV that began in 2002 that was designed to evaluate the role of alcohol use with clinical outcomes. We analyzed VACS Survey data from the 2011–2012 follow-up, and include men who reported having at least one male sex partner in the year prior to the survey and who provided data regarding incarceration history (N = 487). The independent variable of interest is self-reported history of incarceration (recent, ever, or never), and the dependent variables are detectable viral load and HIV risk behaviors: injection drug use (IDU), 2+ sexual partner in past 12 months, sex without a condom in the past 12 months, sex without a condom due to alcohol or drug use, and sex with partners who had other sexual partners. Covariates included age, race, education, relationship status, income, and homelessness.

RESULTS: Participants had a mean age of 52 years (sd = 9.5) and included 292 (60%) African-American, 131 (27%) white, and 64 (13%) of other race/ethnicity. The prevalence of incarceration history was 40%. In both unadjusted and adjusted models for age, education, marital status, income, and homelessness, recent and prior incarceration versus no incarceration were strongly associated with having a detectable HIV viral load (recent adjusted odds ratio [AOR] 3.25 95% CI: 1.35–7.83; prior AOR: 3.30 95% CI: 1.84–5.91). Those with a history of incarceration also had significantly higher odds of past 12 month IDU (recent AOR 17.11, 95% CI: 2.78–105.48; prior AOR 6.10, 95% CI 1.15–32.44). Recent and prior incarceration were strongly associated with having two or more sex partners (recent AOR 2.74, 95% CI: 1.25–6.02; prior AOR 1.71, 95% CI 1.11–2.64) and engaging in sex without a condom due to alcohol (recent AOR 14.34, 95% CI: 3.17–64.74; prior AOR 8.30 95% CI: 2.25–30.63) or drug use (recent AOR 5.74 95% CI 1.83–18.03; prior 2.90 95% CI: 1.13–7.42 respectively) and with partners who had other partners (recent AOR 3.31, 95% CI: 1.42–7.72; prior AOR 2.62, 95% CI 1.47–4.68).

CONCLUSIONS: Among HIV positive MSM, incarceration is linked with multiple proximate determinants of HIV transmission including detectable viral load, IDU and high risk sexual activity. This study highlights the importance of targeting this population for intervention following release from incarceration as a means of reducing new HIV infections.


Elizabeth S. Goldsmith 1, 2; Amy Gravely1; Agnes C. Jensen1; Indulis R. Rutks1; Erin E. Krebs1. 1Minneapolis VA Health Care System, Minneapolis, MN; 2University of Minnesota, Minneapolis, MN. (Control ID #2701126)

BACKGROUND: Chronic musculoskeletal pain affects over 60% of veterans in primary care clinic and is the most common cause of physical disability in the US. Despite this, measuring pain-related physical function remains a challenge. Though self-report measures are the gold standard for pain outcome assessment, their relationship to objective physical performance is unclear. Our objective was to examine associations of self-reported pain-related functional impairment and pain severity with objective physical performance among veterans with chronic musculoskeletal pain.

METHODS: The Strategies for Prescribing Analgesics Comparative Effectiveness (SPACE) trial enrolled 264 veterans (primary care patients of the Minneapolis VA Health Care System) aged 21–80 with chronic (≥6 months) musculoskeletal back or hip/knee arthritis pain for randomization to opioid or non-opioid medication therapy. At baseline, participants completed two self-report measures of pain-related functional impairment—Brief Pain Inventory (BPI) Interference scale and Roland Disability Questionnaire (RDQ)—and a measure of pain severity, the BPI Severity scale. Participants completed study-staff-assessed physical performance measures including 6 meter gait speed, chair stand test, and the Fullerton Advanced Balance Scale (FAB). We compared baseline self-reported and objective measures via general linear regression models adjusted for participant age and sex, considering BPI Interference as the primary measure.

RESULTS: 259 patients had complete pain-related measures at baseline. On average, each one point higher on the 0–10 BPI-Interference scale was associated with 0.2 s slower 6 meter walk (p < 0.0001), 0.5 fewer chair stands in 30 s (p < 0.0001), and 1.1 points lower on the Fullerton Advanced Balance Scale (p < 0.0001). Similarly, higher self-reported pain severity (BPI-Severity) and pain-related impairment on the RDQ were also significantly associated with worse objective physical performance by all measures (gait speed, chair stands, and FAB).

CONCLUSIONS: Measures of self-reported pain-related functional impairment and pain severity were strongly associated with objective physical performance in this population of veterans with chronic musculoskeletal pain. These findings support the validity of patient-reported measures of pain-related function. Future analyses will examine whether self-reported pain measures and physical performance respond similarly to pain treatment over the 12-month SPACE trial. Further research may identify subcategories of patients and pain conditions in which self-reported pain measures relate differently to observed physical performance, and may clarify roles of additional pain dimensions such as pain sensitivity.

AT-THE-BEDSIDE WALKING INTERDISCIPLINARY ROUNDS - STREAMLINED COMMUNICATION BUT NOT YET THE ANSWER FOR PATIENT SATISFACTION Katherine A. Hochman 3; Nicole Adler2; Ramon Jacobs1; Brian Bosworth1; Anne Meara1; Regina Presa1; Thomas Sedgwick1; Patricia Lanzelloti1; Lisa Gumbrecht1. 1NYU Langone Medical Center, New York, NY; 2NYULMC, New York, NY; 3New York University School of Medicine, New York, NY. (Control ID #2701239)

BACKGROUND: Prior To March 2016 interdisciplinary rounds were held in the back of the nurses’ stations on two inpatient medial units. Patients satisfaction scores around the discharge process and communication were consistently low. Medical director and nurse manager dyad leadership teams redesigned interdisciplinary rounds to improve communication between the patient and the health care team.

METHODS: With the At-The-Bedside Walking Interdisciplinary Rounds initiative, every patient is visited by the entire interdisciplinary team each afternoon. The interdisciplinary team consists of the Hospitalist, the unit nurse manager, the bedside nurse, the care manager, the social worker and the medicine resident. Each visit takes 3–5 min and is led by the resident, who starts by introducing every member of the team. Importantly, the patient is surrounded by every person on the team (a design to be literally and figuratively patient centric). Rounds are structured around four simple questions designed to effectively communicate the diagnosis, the milestones for discharge and the discharge disposition and date. Updated information is written on the patient’s white board, located at the foot of the bed. Patients and caregivers have an opportunity to ask clarifying questions. Moreover, the patient can experience first-hand the collaboration that takes place amongst the team members with a streamlined and unified message. The team will use a video language access network for interpreter services for those patients who feel more comfortable speaking in their native language. For those patients who prefer not to discuss discharge planning in large groups, members of the team will return individually.

RESULTS: When comparing pre (Q1CY2016, N = 81) and post (Q2CY2016, N = 80) intervention top box HCAHPS patient satisfaction scores, the results were mixed. Care transitions improved slightly from 46 to 48 and communication with doctors increased from 74 to 75%. Communication with nurses decreased from 78 to 75%. Discharge information, however, improved from 78 to 84%.

CONCLUSIONS: While these early results are disappointing, we believe that patient centered care starts with streamlined communication at the bedside with the interdisciplinary team. We will be tweaking how to better contextualize these rounds for patients in the future.

ATTITUDES TOWARD PEOPLE LIVING WITH HIV AMONG MEN WHO HAVE SEX WITH MEN USING AND NOT USING PREP: IMPLICATIONS FOR PREP DISSEMINATION Evan Rausch 1; Patel V. Viraj2. 1Montefiore Medical Center, New York, NY; 2Montefiore Medical Center/Albert Einstein College of Medicine, New York, NY. (Control ID #2706094)

BACKGROUND: HIV Pre-exposure Prophylaxis (PrEP) with antiretroviral medications is highly effective in preventing HIV acquisition, but social barriers to access have hampered adoption among certain groups at highest risk, including young Black and Latino men who have sex with men (YBLMSM). Little is known about the relationship between PrEP adoption and attitudes toward people living with HIV (PLWHA). We conducted a qualitative study interviewing YBLMSM on their views on PrEP and HIV to identify differences between PrEP adopters and non-adopters and inform PrEP uptake interventions.

METHODS: We conducted individual semi-structured interviews with YBLMSM recruited through Facebook from July 2015 to November 2015. Patients met inclusion criteria if they were between 18 and 29 years of age, had engaged in condomless anal sex with a man in the past year, were HIV negative by self-report, were living or working in the Bronx, New York, and were fluent in English or Spanish. We continued interviews until reaching thematic saturation and used a modified grounded theory approach for analysis. Two investigators independently coded all interviews using an iteratively developed codebook.

RESULTS: Participants (n = 19) all identified as gay and had a primary care provider. About half (n = 10) used PrEP at the time of interview (adopters). Three themes related to attitudes toward PLWHA emerged that differed between adopters and non-adopters: relationships with people with HIV, HIV status disclosure, and severity of HIV. Most participants knew a PLWH, but adopters were more likely to have a close friend with HIV or multiple friends with HIV. Although both adopters and non-adopters had experienced sexual encounters where an PLWH withheld their status, those who adopted PrEP viewed the experience in a more positive light and framed it in the context of challenges around HIV stigma and disclosure. Non-adopters were more likely to view HIV as a severe diagnosis, while adopters were more likely to view HIV as a chronic illness.

CONCLUSIONS: Although PrEP is used to prevent HIV acquisition, our findings suggest that PrEP adoption in YBLMSM is associated with more accepting views of PLWH. Reduction of HIV stigma is a potential focus of interventions promoting PrEP adoption, offering a way to unify future HIV treatment and HIV prevention strategies around a common goal of reducing HIV disparities.

ATTRITION AMONG WOMEN VETERAN COMMUNITY CARE USERS NEW TO VA Susan M. Frayne 1; Ciaran S. Phibbs1; Elizabeth M. Yano2; Donna L. Washington2; Fay Saechao1; Eric Berg1; Sarah Friedman1; Andrea Finlay1; Katherine J. Hoggatt2; Alison Hamilton2. 1VA Palo Alto, Palo Alto, CA; 2VA Greater Los Angeles HSR&D Center, Sepulveda, CA. (Control ID #2702970)

BACKGROUND: The Veterans Health Administration (VA) purchases care in the community for many patients, and disproportionately for women. Little is known about the patient experience of community care, especially for new patients who are forming their first impressions of VA. For women newly joining VA, we characterized facility-level variability in use of community care, then examined whether use of community care predicted attrition from VA.

METHODS: Among women Veterans with 1+ VA primary care (PC) clinic visit in fiscal year 2011 who were new to VA (no VA or fee basis use in the prior 8 years), those with at least one outpatient or inpatient non-VA (fee) medical care visit in the year following T0 (date of first outpatient visit) were “community care users.” Attrition was defined as no VA care in person-specific years 2–3 following T0.

RESULTS: Among the 18,395 new women VA primary care patients nationally, 5,555 (30.2%) were community care users in their first year of VA care. Across facilities, 1.7 to 69.9% (median 35%) were community care users. The attrition rate of was 7.8% for community care users vs 19.3% for others; this pattern was similar for patients at VA Medical Centers (VAMCs) and Community-Based Outpatient Clinics (CBOCs). In a logistic regression, odds of attrition were lower for community care users than for others (adjusted OR 0.43, 95% CI 0.38-0.48), controlling for age, race/ethnicity, marital status, urban/rural residence, service-connected disability, receipt of care from a designated women’s health provider, and receipt of care in a VAMC versus CBOC.

CONCLUSIONS: Nearly one in three new women PC patients received community care at least once within a year of joining VA, although community referral rates varied markedly by facility. Those who received part of their care through VA’s community care program had less than half the odds of attriting from VA in the subsequent 2 years compared to women who did not. While further inquiry is needed regarding how assignment to a facility with low versus high reliance on community care impacts the patient experience, these findings offer some reassurance that sending new women patients out to the community for care not available at the local VA does not adversely affect attrition, a patient-centered construct.

AWARENESS OF BULLYING IN INTERNAL MEDICINE RESIDENCIES: RESULTS OF A NATIONAL SURVEY OF INTERNAL MEDICINE PROGRAM DIRECTORS Manasa Ayyala 1; Saima Chaudhry3; Donna Windish5; Denise Dupras2; Shalini Reddy4; Scott Wright1. 1Johns Hopkins University School of Medicine, Baltimore, MD; 2Mayo Clinic, Rochester, MN; 3Memorial Healthcare System, Hollywood, FL; 4University of Chicago Pritzker School of Medicine, Chicago, IL; 5Yale University, Cheshire, CT. (Control ID #2705010)

BACKGROUND: Bullying in medical education has been described as a significant and ubiquitous problem in studies of residents and medical students. American and international studies cite upwards of 50% of trainees reporting experiencing bullying, most commonly in its verbal form. Being bullied creates psychological pressure which can lead to mental health consequences including depression, anxiety, and post-traumatic stress disorder. The authors conducted this study to describe perceptions of internal medicine (IM) program directors about the bullying of internal medicine residents.

METHODS: The 2015 Association of Program Directors in Internal Medicine (APDIM) Annual survey was sent to 368 program directors (PDs) with APDIM membership, representing 92.9% of the IM residency programs. To understand internal medicine PDs’ perspectives and awareness about bullying in their programs, several questions about bullying were included in the annual survey. Bivariate analyses were performed on PD characteristics and program characteristics with regard to their answer to the following specific question, “To the best of your knowledge, were one or more of your trainees bullied during the last academic year?”

RESULTS: A total of 227/368 (61.6%) of PDs responded to the survey. Less than one third of respondents (71/227, 31%), reported being aware of bullying in their residency program during the previous year. There were no significant differences between those reporting bullying in their programs and those who did not when gender, tenure as PD, geographic location, or specialty of PD were considered in the analyses (all p > 0.05). Those who acknowledged bullying in their program were more likely to agree that bullying was a problem in graduate medical education (p < 0.0001), and that it had a significant negative impact on the learning environment (p < 0.0001).

CONCLUSIONS: Most IM PDs believe that bullying does not occur in their training programs. Because bullying is thought to negatively affect the learning environment and threaten the well-being of trainees, program directors may wish to more proactively assess its actual prevalence.

AZITHROMYCIN IS ASSOCIATED WITH IMPROVED SURVIVAL IN PATIENTS WITH HEALTHCARE-ASSOCIATED PNEUMONIA Eric Mortensen 1; Russell Attridge2; Antonio Anuzeto3. 1University of Connecticut Health Center, Farmington, CT; 2University of the Incarnate Word, San Antonio, TX; 3South Texas Veterans Health Care System, San Antonio, TX. (Control ID #2705318)

BACKGROUND: Many studies of patients hospitalized with healthcare-associated pneumonia (HCAP) have demonstrated worse survival for those patients who received HCAP concordant therapy as compared to community-acquired pneumonia (CAP) concordant therapy. We hypothesize that this is because azithromycin is a part of CAP concordant therapy, but not HCAP concordant therapy. Therefore the purpose of this study was to examine the association of azithromycin use with 30-day mortality for patients hospitalized with HCAP after adjusting for potential confounders.

METHODS: We conducted a retrospective national study using Department of Veterans Affairs administrative data of patients hospitalized at any Veterans Administration acute care hospital. We included patients >65 years hospitalized with pneumonia in fiscal years 2002–2012 with at least 1 HCAP risk factor. HCAP risk factors include prior hospital admission <90 days, nursing home residency, prior outpatient intravenous antibiotics <90 days, and hemodialysis. Our outcome was 30-day mortality. We used a generalized linear mixed effect model (“multilevel regression”) to control for potential confounders, including sociodemographics, comorbid conditions, prior outpatient health care utilization, and severity of illness, as well as for the admitting hospital.

RESULTS: We identified 26,113 patients who meet the inclusion criteria and 7,864 (29.8%) received azithromycin as part of their inpatient antibiotic therapy. After adjusting for potential confounders, we found that 30-day mortality was significantly lower in those who received azithromycin (odds ratio 0.60, 95% confidence interval 0.55-0.64) as part of their inpatient antibiotic therapy.

CONCLUSIONS: In patients hospitalized with pneumonia with HCAP risk factors, we identified a significant association between azithromycin use and 30-day mortality. Our study suggests that for patients with HCAP, azithromycin use may be associated with improved outcomes. Randomized clinical trials are needed to identify what are the best empiric antibiotics regimens for patients hospitalized with healthcare-associated pneumonia.

BARRIERS AND FACILITATORS AFFECTING THE IMPLEMENTATION OF SUBSTANCE USE SCREENING IN PRIMARY CARE CLINICS: A QUALITATIVE STUDY OF PATIENTS, PROVIDERS, AND STAFF Jennifer McNeely 1, 1; Pritika Kumar1; Traci Rieckmann3; Erica Sedlander1; Sarah Farkas1; Joseph Kannry2; Aida C. Vega2; Eva Waite2; Lauren Peccoralo2; Richard N. Rosenthal2; Dennis McCarty3; John Rotrosen1. 1NYU School of Medicine, New York, NY; 2ICAHN School of Medicine at Mount Sinai, New York, NY; 3Oregon Health and Science University, Portland, OR. (Control ID #2707175)

BACKGROUND: Alcohol and drug use is a leading cause of morbidity and mortality that frequently goes unidentified in medical settings. As part of a multi-phase study to implement the NIDA Common Data Elements for collecting substance use screening information in electronic health records (EHRs), we interviewed key clinical stakeholders with a goal of identifying barriers and facilitators affecting the implementation of substance use screening in primary care clinics.

METHODS: Focus groups and individual qualitative interviews were conducted with 67 stakeholders, including primary care patients, medical providers (faculty and resident physicians, nurses), and medical assistants, in two urban academic health systems. Themes were identified, discussed, and revised through an iterative process, and mapped to the Knowledge to Action (KTA) framework (Graham, 2006), which guides the selection and implementation of new clinical practices.

RESULTS: Factors affecting implementation based on KTA elements were identified from participant narratives. Identifying the problem: Participants unanimously agreed that having knowledge of a patient’s substance use is important because of its impacts on health and medical care, that substance use is not properly identified in medical settings, and that universal screening is the best approach. Adapting knowledge: The majority of patients and providers stated that the primary care provider should play a key role in substance use screening and interventions. There was discrepancy of opinion regarding the optimal approach to delivering screening. Some felt that patients should self-administer questionnaires, while others thought that patients would be more comfortable having face-to-face discussions with their primary care provider - though not with other members of the care team. Many providers reported that being able to take effective action once unhealthy substance use is identified is crucial. Assessing barriers: Patients expressed concerns about confidentiality, ‘denial’, and providers’ lack of empathy. Barriers identified by providers included individual-level factors such as lack of knowledge and training, and systems-level factors including lack of time, resources, and space, disjointed communication between members of the medical team, and difficulty accessing addiction treatment.

CONCLUSIONS: Based on these findings, we designed and are testing an implementation strategy utilizing universal screening, patient self-administered questionnaires, and EHR-integrated clinical decision support to assist providers in conducting brief motivational counseling and linking patients to behavioral health services, to address unhealthy substance use in primary care clinics.

BARRIERS AND MOTIVATORS TO DIABETES CONTROL IN A NAVIGATOR-PHOTOVOICE INTERVENTION: A QUALITATIVE ANALYSIS Jenny Jia; Lisa M. Quintiliani; Ve Truong; Cheryl Jean; Karen E. Lasser. Boston Medical Center, Boston, MA. (Control ID #2690649)

BACKGROUND: Diabetes mellitus is a major public health issue that disproportionately affects minorities and individuals with low socioeconomic status and has major implications on patients’ daily lives. We developed and pilot tested a community-based diabetes self-management intervention using photovoice and group education sessions to explore diabetes disease control.

METHODS: Patients with poorly controlled diabetes living in Boston census tracts that we had previously identified as having large numbers of patients with diabetes and who receive primary care at Boston Medical Center, the largest safety-net hospital in New England, were invited to participate in 6 group sessions held at a local library. Sessions included photovoice, a methodology whereby patients use photography and write accompanying narratives to facilitate deeper thinking about controlling their diabetes. We designed three photovoice missions, each with different prompts for participants to take photos in their day-to-day lives related to diabetes. After each mission, a patient navigator facilitated discussions to elucidate patients’ experiences. The navigator also invited guest speakers to provide educational presentations (e.g., dentistry) based on requests from participants. Sessions were audio recorded and transcribed. We coded transcripts for themes of barriers and motivators in diabetes control.

RESULTS: The majority of participants (n = 16) were 67% female (67%), and African American or Black (89%). Less than half had greater than a high school education. Participation in the group sessions was 68.8% amongst all participants averaged over all sessions. The most common barriers mentioned by participants were 1.) issues with self-control over diabetes self-management (“I have a love affair with food.”), 2.) lack of dietary knowledge (“Does it matter if it’s white, wheat, or whole wheat [bread]?”), and 3.) frustration and fatigue towards self-management of diabetes (“I’ve gotten to the point where I’m rebellious. I don’t take insulin and I won’t take my medication.”). Subanalysis of statements related to frustration and fatigue showed that common root issues include the chronic nature of diabetes and no demonstration of improvement in participants’ disease control despite individual efforts. Motivators included interpersonal elements, such as the intervention’s group sessions (“I enjoy coming down here because you don’t feel so alone because a lot of times you don’t mention diabetes but it’s always in your head.”), clinicians, family members, and caregivers.

CONCLUSIONS: The combination of perceived lack of self-control and frustration and fatigue, often related to the chronicity of diabetes and poor outcomes in disease control, is concerning for the development of burnout in the diabetic patient population. Interventions to build resilience in diabetics should be considered. Participants were generally receptive to the group sessions, which may be enhanced by hosting sessions at a community-based location.

BARRIERS AND STRATEGIES FOR TAPERING LONG-TERM OPIOID MEDICATIONS: A QUALITATIVE STUDY OF PRIMARY CARE PROVIDER EXPERIENCES Laura C. Kennedy 2; Ingrid A. Binswanger3, 1; Shane Mueller1, 3; Cari Levy5, 4; Daniel Matlock4, 2; Susan L. Calcaterra4, 6; Steve Koester3; Joseph W. Frank4, 5. 1Kaiser Permanente Colorado, Denver, CO; 2University of Colorado, Aurora, CO; 3University of Colorado Denver, Denver, CO; 4University of Colorado School of Medicine, Aurora, CO; 5VA Eastern Colorado Health Care System, Denver, CO; 6Denver Health Medical Center, Denver, CO. (Control ID #2701272)

BACKGROUND: Since 1999, prescribing of opioid medications quadrupled, as has the rate of drug overdose deaths. Growing observational evidence describes the dose-dependent risks of opioids, but evidence of long-term benefits is lacking. Healthcare providers increasingly seek to taper or discontinue chronic opioid use, but little evidence exists on how to approach this challenging process. We sought to explore primary care providers’ (PCPs) experiences discussing and implementing tapers of long-term opioid therapy with patients in their clinical practice.

METHODS: We conducted 6 semi-structured, in-person focus groups with PCPs. A multidisciplinary team developed the focus group interview guide, with areas of focus on discussing risks and benefits of opioid medications with patients, approaches to opioid tapering, and barriers and facilitators to tapering opioids. Focus groups were recorded, transcribed, and coded in ATLAS.ti. We performed qualitative analysis in a mixed inductive-deductive manner and identified themes through an iterative, multidisciplinary, team-based approach.

RESULTS: PCPs (N = 40; 70% non-Hispanic white, 85% physicians) at a large academic medical center, an urban safety-net hospital, and a VA medical center participated. Emergent themes were identified in 2 domains: 1) barriers to tapering, and 2) strategies to facilitate tapering. Barriers to tapering opioids include providers’ pessimism, limited time and lack of patient-provider trust. One provider noted, “The patient has to be receptive to the idea. If you dont have that as a baseline, I dont think theres anything you can say where they will be willing to do it.” Describing the impact of time constraints, one provider stated, “This is never the only conversation Im having with a patient in that room and patients never come to discuss that, so to sneak that into a visit where they might have three or four other agenda itemscan be quite difficult.” Regarding trust, one provider perceived that patients “fear that if they tell me their pain is better I might decrease their medicine.” Facilitators emerged as both provider- and system-level strategies. Providers sought to connect with patients by demonstrating empathy (“I understand youre in pain. I get it”), using simple language (“This puts you at higher risk of death”) and planning ahead (“Were going to talk about coming down so its almost like a quit date”). A clinic-wide upper limit on opioid dosing supported their efforts to taper opioids. One provider noted, “The triggering point of the discussion was that new change in policy establishing a new ceiling dose of opiates. So, its out of my hands.”

CONCLUSIONS: Discussing and implementing opioid tapering presented significant challenges for time-strapped providers. A feasible primary care model to support tapering of long-term opioid therapy may involve communication training for providers, dedicated time to discuss tapering opioids, and clinic-wide policies establishing upper-limits on opioid dosing.

BARRIERS TO IMPLEMENTATION OF A TRANSITIONAL CARE INTERVENTION: A QUALITATIVE ANALYSIS Jeffrey L. Schnipper 1, 3; Hilary Heyison1; Cherlie Magny-Normilus1; Elyse R. Park2; Nyryan V. Nolido1; Ryan Thompson2, 3; Gwen Crevensten2, 3. 1Brigham and Women’s Hospital, Boston, MA; 2Massachusetts General Hospital, Boston, MA; 3Harvard Medical School, Boston, MA. (Control ID #2706649)

BACKGROUND: Transitions from hospitals to the ambulatory setting are high risk periods for patients. Many interventions have been tried, with varying degrees of success, and often the problem has been with implementation rather than theoretical efficacy of the intervention. The objective of this analysis was to better understand barriers to implementation in the context of a real-world quality improvement study.

METHODS: We developed, implemented, refined, and evaluated a multi-faceted, multi-disciplinary transitions intervention across two hospitals and 18 primary care practices within a Pioneer Accountable Care Organization. We then conducted focus groups of 2–8 clinicians each that were part of or affected by the intervention: inpatient and outpatient physicians, nurses, pharmacists, and care coordinators. We used a semi-structured qualitative interview guide to standardize data collection. Preliminary themes, including barriers to implementation, were identified a priori using the Systems Engineering Initiative for Patient Safety model of work-system design, then iteratively refined. Each focus group session, approximately 60 min each, was audio-recorded, transcribed, and then uploaded into NVivo 11 (QSR, Melbourne, 2012). Two coders independently coded all transcripts then met to resolve discrepancies.

RESULTS: A total of 8 focus group interviews were conducted, involving 21 participants. Several barriers to implementation of the intervention were identified. Regarding those related to the inherent nature of the intervention, barriers included complexity of certain tasks, such as post-discharge phone calls, relative to the personnel carrying them out, as well as the number of personnel involved. Logistical barriers included different modes of communication between inpatient and outpatient nurses, having enough advanced warning of a pending discharge to perform medication counseling, and lack of staffing to conduct the intervention on weekends. Handoffs between roles and locations was a barrier, with inadequate information systems to keep everyone on the same page and communicate important issues to be followed up. Several providers linked the lack of sufficient staffing for this intervention to the institution’s lack of commitment to improving transitions of care. The barriers most often cited by providers were lack of communication and lack of time availability/”bandwidth,” followed by lack of institutional commitment, difficult patient population, competing priorities, variable staffing, and logistics.

CONCLUSIONS: This qualitative analysis identified several barriers to the success of transitional care interventions. While some barriers were specific to the intervention being studied, many provide generalizable lessons learned to improve the success of future transitional care interventions.

BEYOND HEALTH INSURANCE: REMAINING DISPARITIES IN U.S. HEALTH CARE IN THE POST-ACA ERA Benjamin D. Sommers 1, 2; Caitlin McMurtry1; Robert Blendon1; John Benson1; Justin Sayde1. 1Harvard School of Public Health, Brookline, MA; 2Brigham & Women’s Hospital, Boston, MA. (Control ID #2704169)

BACKGROUND: The Affordable Care Act (ACA) has reduced the U.S.’s uninsured rate to an historic low, though it faces an uncertain future after the 2016 election. However, coverage is only one of many factors that may contribute to racial and income-based disparities in health care access, affordability, and quality. There has been little post-ACA analysis of remaining disparities in health care and how much of a role health insurance coverage still plays in these gaps.

METHODS: Using a novel national telephone survey of over 8,000 Americans conducted in late 2015, we examined disparities between low-income and high-income adults, and between racial/ethnic minorities and whites. We examined three main outcomes: self-reported quality of care, cost-related delays in care, and emergency department (ED) use due to lack of available appointments. We conducted a series of cross-sectional regression analyses, starting with models that only took into account income or race, and then sequentially adjusted for health insurance, state of residence, demographics, and health status, in order to assess the extent to which disparities were mediated by insurance differences. Then we used multivariate regression to assess respondents’ views of whether quality and affordability had improved over the past 2 years and whether the ACA was helping them.

RESULTS: Quality of care ratings were significantly worse among lower-income adults than higher-income adults. Only 10–25% of this gap was explained by health insurance coverage. Cost-related delays in care and ED use due to lack of available appointments were nearly twice as common in the lowest-income group, and less than 40% of these disparities was explained by insurance. There were significant racial/ethnic gaps: reported quality of care was worse among blacks and Latinos than whites, with 16–70% explained by insurance. In contrast to these disparities, lower-income and minority groups were generally more likely than whites or higher-income adults to say that the ACA was helping them and that the quality and/or affordability of care had improved in recent years.

CONCLUSIONS: Our post-health reform survey shows ongoing stark income and racial disparities in the health care experiences of Americans. We find evidence that the ACA has helped narrow some of these gaps, suggesting that a potential repeal of the law poses significant risk particularly to low-income groups and racial/ethnic minorities. But even if the ACA survives in some form under the new administration, the law’s coverage expansion should not be considered the primary solution to racial and socioeconomic disparities in health care, since health insurance coverage only explains a moderate portion of ongoing disparities in affordability, quality, and access. Additional policy attention will be needed to address these serious problems in the post-ACA era.

BP REPAIRED: EFFECT OF REPEAT MEASUREMENT OF INITIALLY ELEVATED BLOOD PRESSURE ON BLOOD PRESSURE CONTROL Douglas Einstadter 2; Shari Bolen3; Randall D. Cebul1. 1Case Western Reserve University, Chagrin Falls, OH; 2MetroHealth Medical Center, Cleveland, OH; 3MetroHealth/Case Western Reserve University, Cleveland, OH. (Control ID #2704668)

BACKGROUND: Hypertension (HTN) affects nearly 70 million people in the US. However, only 53.0% of HTN patients seen in primary care have their BP controlled to <140/90 mm Hg according to national surveys. BP measurement error is recognized as a major cause of poor BP control. Although many aspects of measurement error have been described, the effect of a second BP measurement on overall clinic-measured BP control in real-world practices has not been reported. We evaluated the effect of a second BP measurement on the rate of elevated BP among more than 48,000 patients with diagnosed HTN and followed in primary care at an urban safety-net health system.

METHODS: In mid-2015, we introduced an electronic health record (EHR) prompt to remind staff to repeat a BP measurement for HTN patients whose initial BP was ≥140/90 mm Hg. To evaluate the effect of repeating the BP, we obtained all recorded BP values for HTN patients who were seen in a primary care clinic between January and December, 2016 at MetroHealth, an urban safety-net system in Cleveland, Ohio. We compared the rate of BP control based on the first recorded BP with that of the final recorded BP at each visit. We also determined the change in systolic BP (final-initial) overall and by age, sex, race, insurance and initial systolic BP subgroups.

RESULTS: From January through December, 2016 there were 130,687 primary care office visits for 48,007 patients with HTN. The mean age was 61 years, 59% were female, 43% black, and 30% Medicaid or uninsured. The initial BP was ≥140/90 at 50,684 (39%) visits and an initially high BP was re-measured at 42,464 (84%) visits. The median change (final-initial) in systolic BP was −10 mm Hg (IQR −2 to −18). Among those with a repeat BP, 37% of final BPs were <140/90. Thus, repeating the BP increased the overall HTN control rate by 12% (i.e., 39% × 84% × 37%) from 61 to 73%. Control of BP on repeat measurement was not associated with age, sex, race or insurance type. However, BP change was positively associated with the initial BP value; the higher the initial systolic BP, the greater the change in final systolic BP. While the absolute BP change was greatest for patients with the highest initial systolic BP, those closest to the control threshold were most likely to have normalization of their BP on repeat measurement.

CONCLUSIONS: Repeat measurement of an initially elevated BP resulted in a meaningful improvement in overall BP control, comparable to that seen with addition of an antihypertensive medication. The absolute BP change was greater for patients with a higher initial systolic BP, but patients with an initial BP value closer to the goal were more likely to normalize on repeat. Implementing routine repeat measurement for an initially elevated BP may contribute to improved decision-making around hypertension management and should be included as a component of programs to improve BP control.

BREAST CANCER DIAGNOSIS AND TREATMENT BEFORE AND AFTER HIGH-DEDUCTIBLE INSURANCE ENROLLMENT James F. Wharam 1; Fang Zhang2; Christine Lu2; Anita Wagner2; Larissa Nekhlyudov4; Craig C. Earle3; Stephen B. Soumerai2; Dennis Ross-Degnan2. 1Harvard Medical School and Harvard Pilgrim Healthcare, Boston, MA; 2Harvard Medical School and HPHCI, Boston, MA; 3Ontario Institute for Cancer Research, Toronto, ON, Canada; 4Harvard Medical School, Boston, MA. (Control ID #2702914)

BACKGROUND: Primary care physicians play a key role in ensuring appropriate cancer screening, initial workups for breast cancer, and timely transitions through subsequent oncologic care. High-deductible health plans (HDHP) have recently become the predominant commercial health insurance product in the US and the centerpiece of Affordable Care Act replacement proposals. However, HDHP effects on patients undergoing evaluation and treatment for cancer are unknown.

METHODS: We studied Optum health insurance claims and enrollment data derived from a large national health insurer. We used a controlled segmented survival design to examine employer-mandated HDHP transitions, minimizing selection bias. We included approximately 273,000 intervention group women age 25–64 without evidence of breast cancer prior to inclusion. These women were continuously enrolled for 1 year in a low-deductible (≤$500) plan followed by up to 4 years in a HDHP (≥$1000) after an employer-mandated switch. Women were included on a rolling basis and followed between 2004–2012. The comparison group comprised contemporaneous, 1:1 propensity score-matched women whose employers offered only low-deductible plans. Primary outcome measures included times to first diagnostic breast imaging (diagnostic mammogram, breast ultrasound, or breast MRI), breast biopsy, breast cancer diagnosis, and surgical resection (lumpectomy or mastectomy) and were analyzed using segmented survival models adjusted for age group, race/ethnicity, education level, poverty level, US region, index month, and duration of enrollment prior to baseline.

RESULTS: At follow-up, HDHP members experienced delays in receipt of diagnostic imaging (adjusted hazard ratio, aHR: 0.97 [0.94,0.99]), biopsy (aHR: 0.94 [0.89,0.99]), breast cancer diagnosis (aHR: 0.82 [0.74,0.90], Figure), and surgical resection (aHR: 0.90 [0.84,0.98], Figure) compared with controls (p-values comparing follow-up aHRs with baseline aHRs: 0.013, 0.026, 0.002, and 0.004, respectively).

CONCLUSIONS: HDHP enrollment was associated with delays in breast cancer diagnostic testing, diagnosis, and surgical resection. Such delays might lead to adverse long-term breast cancer outcomes among HDHP members. Primary care physicians, oncologists, and population health managers should consider closely monitoring women at risk for breast cancer who are enrolled in HDHPs.

figure p

Time to first (c) breast cancer diagnosis and (d) breast surgical resection 12 months before and 48 months after a mandated HDHP switch (red line), compared to contemporaneous control group women who remained in low-deductible plans (green line).


BACKGROUND: Since 2005, the U.S. Preventive Services Task Force (USPSTF) has recommended genetic counseling for women at increased risk of breast cancer or ovarian cancer due to BRCA mutations, however, only 18% of primary care physicians use a formal breast cancer risk calculator to identify women at high risk. As a result, we sought to evaluate the feasibility and impact of two methods of assessing the risk of having a genetic mutation increasing breast cancer risk (i.e., in person and via the patient portal) within a primary care setting in a federally qualified community health center (FQHC).

METHODS: We identified women ages 18 and older with established care with Internal Medicine, without a breast cancer diagnosis. To assess possible need for genetic counseling and testing, we administered the Breast Cancer Genetics Referral Tool (B-RST). This test takes approximately 5 min to complete and relies primarily on patient-reported family history. Results are reported as “positive” or “negative.” A positive result connotes increased risk of a genetic mutation and the possible need for detailed genetic counseling. For in-person assessment, we approached all adult women in the waiting area at the FQHC prior to a scheduled appointment and introduced the study. A research assistant administered the B-RST survey in English or Spanish to consenting women. We are currently administering the B-RST through a secure, online patient portal to remaining eligible women who are enrolled in the portal. These women receive an online, secure message with information about breast cancer risk and the B-RST survey. All patients were informed of their results and were given information to assist with interpretation of results. For those patients with positive results, study staff also notified their primary care providers, who were provided detailed information on how to refer patients for genetic counseling, if desired.

RESULTS: For in-person recruitment, we identified 177 eligible women, of whom 168 consented to participate for a 95% recruitment rate. The mean age of participants was 46 years; 144 (95%) were Latina, and 103 (61%) preferred Spanish. Of women surveyed in-person, 38 (23%) reported a family history of breast or ovarian cancer and 4 (7%) had positive scores on the B-RST. For on-line assessments, we sent messages to 10 women so far, of whom 3 responded, for a response rate of 30%. Mean age was 44 years, and 7 (70% were Latina). One person had a family history of breast or ovarian cancer; none had positive B-RST scores. We are awaiting final data on assessment results from both methods, as well as data regarding referrals for genetic counseling and testing.

CONCLUSIONS: We identified previously unrecognized risk of BRCA gene mutations in a substantial number of women, highlighting the importance of developing systematic methods of risk assessment in this population.

BURNOUT AMONG PRIMARY CARE PROVIDERS AND STAFF IN SMALL TO MEDIUM SIZED PRIMARY CARE PRACTICES: EARLY FINDINGS FROM EVIDENCENOW Samuel T. Edwards 2, 3; Miguel Marino2; Bijal A. Balasubramanian1; Steele H. Valenzuela2; Rachel Springer2; Leif Solberg4; Alex Preston1; Deborah Cohen2. 1University of Texas School of Public Health, Dallas, TX; 2Oregon Health & Science University, Portland, OR; 3VA Portland Health Care System, Portland, OR; 4HealthPartners Institute, Minneapolis, MN. (Control ID #2705392)

BACKGROUND: Clinician burnout is associated with poorer health care quality but it is unknown what practice characteristics are associated with clinician and staff burnout in small to medium sized primary care practices.

METHODS: Study Design and Sample: We performed a cross-sectional survey of a large nationwide sample of small to medium sized primary care practices participating in the Agency for Healthcare Research and Quality’s EvidenceNOW initiative. The seven cooperatives of EvidenceNOW work with 1685 primary care practices in seven regions of the United States to improve the delivery of heart health care. Surveys were fielded at the beginning of EvidenceNOW from 9/22/2015-10/03/2016. Measures: A practice survey was administered to practice leadership that collected characteristics such as practice size, ownership, staffing, patient demographics, provider panel size, urban/rural location, and registry use. A practice member survey administered to practice clinicians and staff included questions on job type, years employed, hours worked per week, and burnout. Burnout was assessed using a single item 5-level measure as used in the Physician Worklife Study. A score of 3 or higher indicated burnout. Analysis: We conducted a person-level, generalized estimating equation logistic regression to assess the relationship between practice/person characteristics and burnout. Analyses accounted for clustering of clinicians/staff by practice using robust sandwich standard errors.

RESULTS: Of 1685 enrolled practices, 1468 returned surveys (87% response rate). We analyzed a subset of 904 practices in which >1 practice member survey and a practice survey was returned. Among these practices, 5,953 practice members returned surveys, a mean of 6.6 respondents per practice. Burnout was present in 16.9% of respondents overall, but was higher among physicians (22.3%) than other job types (nurse practitioner = 18.9%, clinical staff = 18.3%, non-clinical staff = 16.8%). Correlates of burnout included physician respondent (MD vs. non clinician, OR 1.38, 95% CI 1.07–1.79), non-solo practice (2–5 clinician practice vs. solo practice, OR 1.92, 95% CI 1.36-2.71), more years in practice (4–8 years in practice vs. < 4 years OR 1.63, 95% CI 1.38-1.93) and rural location (urban vs. rural OR 0.65, 95% CI 0.46–0.91). Clinician panel size, patients seen per week per clinician, staff/clinician ratio and practice ownership were not associated with burnout.

CONCLUSIONS: Burnout is prevalent in small to medium sized primary care practices, but at a lower frequency than seen in prior studies and physicians have only slightly higher levels of burnout than other staff members. Members of solo practices have lower odds of burnout, while clinician panel size, patients seen per week and staffing levels are not associated with burnout, suggesting practice autonomy is a more important determinant of burnout than workload. Policy efforts to bolster primary care should focus on promoting practice characteristics associated with less burnout.

BURNOUT AND COPING IN INTERNAL MEDICINE RESIDENTS Brielle Spataro 1; Sarah A. Tilstra4; Doris Rubio2; Melissa McNeil3. 1UPMC Presbyterian Hospital, Pittsburgh, PA; 2University of Pittsburgh, Pittsburgh, PA; 3University of Pittsburgh Medical Center, Pittsburgh, PA; 4University of Pittsburgh School of Medicine/Medical Center, Pittsburgh, PA. (Control ID #2705365)

BACKGROUND: Physician burnout is a syndrome characterized by emotional exhaustion, depersonalization, and a low sense of personal accomplishment. It is thought to be a consequence of chronic work-related stress. Coping is how one deals with stress and may affect burnout. The objective of our study was to determine the correlations between assessed burnout and coping mechanisms.

METHODS: All internal medicine residents at the University of Pittsburgh are assessed annually in June for burnout using the Maslach Burnout Inventory-General Survey (MBI-GS) beginning prior to the start of the internship year. Residents are considered to have at least one manifestation of professional buronout if they have a high level of either emotional exhaustion or cynicism. In addition, all residents are asked to complete a Brief COPE survey which measures 14 different coping mechanisms. We analyzed all surveys completed from 2010–2015. Spearman correlation was used to determine the relationship between emotional exhaustion and cynicism and coping at each time point. Residents were stratified into post graduate year (PGY).

RESULTS: We analyzed 1,144 surveys from 587 residents. There were 397 pre intern and 309 PGY-1, 227 PGY-2, and 211 PGY-3 assessments respectively. 577 (50%) of the surveys were from women. When evaluating the relationship between burnout and maladaptive coping mechanisms at each PGY level, we found small to moderate correlations between emotional exhaustion with behavioral disengagement (r = 0.18–0.37), self-blame (r = 0.27–0.45), self-distraction (0.18–0.32,), and venting (0.15–0.47). There were small to moderate correlations with cynicism and behavioral disengagement (r = 0.19–0.40), self-blame (r = 0.24–0.35), self-distraction (r = 0.14–0.34) and venting (r = 0.12–0.38). Denial (r = 0.13–0.20) and substance use (r = 0.15–0.22) had small correlations with emotional exhaustion. There were also small correlations with denial (r = 0.10–0.26), humor (r = 0.13–0.21) and substance use (r = 0.10–0.29). with cynicism. When evaluating the relationship between adaptive coping and burnout, we found a small correlation with the use of acceptance with emotional exhaustion (r = 0.10–0.24) and cynicism (r = 0.11–0.15). No other adaptive coping mechanisms consistently correlated at each PGY with emotional exhaustion or cynicism.

CONCLUSIONS: The use of maladaptive coping mechanisms correlated with professional manifestations of burnout. The strongest correlations were seen with behavioral disengagement, self-blame, self-distraction and venting. This suggests that the assessment of coping mechanisms could identify residents at risk for burnout and could serve as an early warning sign for high risk residents.

CAN A WEB-BASED TRACKING & FEEDBACK TOOL IMPROVE BREAST CANCER TREATMENT AT SAFETY-NET HOSPITALS? Nina A. Bickell 3; Maria Castaldi2; Ajay Shah5; Alan Sickles6; Theophilus Lewis7; Shalini Arora8; Kevin Clarke12; Peter Pappas9; Margaret Kemeny10; Anitha Srinivasan11; Rebeca Franco3; Kezhen Fei1; Michael Parides1; Ann S. McAlearney4. 1Icahn School of Medicine at Mount Sinai, New York, NY; 2Jacobi Medical Center, Manhasset, NY; 3Mount Sinai School of Medicine, New York, NY; 4The Ohio State University, Columbus, OH; 5Bronx-Lebanon, Bronx, NY; 6NYU Lutheran Medical Center, Brooklyn, NY; 7Kings County Hospital Center, Brooklyn, NY; 8Elmhurst Hospital Center, Elmhurst, NY; 9The Brooklyn Hospital Center, Brooklyn, NY; 10Queens Hospital Center, Jamaica, NY; 11Metropolitan Hospital Center, New York, NY; 12Newark Beth Israel Medical Center, Newark, NJ. (Control ID #2688645)

BACKGROUND: System failures, a cause of underuse in which doctors order care, patients don’t refuse but care doesn’t ensue, tend to happen more at safety net hospitals (SNH). Tracking & follow-up approaches that close referral loops in SNHs may reduce underuse. In a randomized trial, we tested a Tracking & Feedback (T&F) tool to reduce underuse of adjuvant breast cancer treatment.

METHODS: We recruited 5 community & 5 municipal SNHs that serve a large proportion of minorities in the NYC metropolitan area. We implemented rapid case ascertainment, a T&F tool and trained point persons at each site to determine if women with newly operated stage 1–3 breast cancer, connected with the oncologist since such connections are associated with getting treated. The tool created a daily “to do” reminder for point persons to ascertain if patients were seen by the oncologist. Point persons then relayed this information to surgeons to follow through as they deemed necessary. Underuse includes: no RT after lumpectomy or mastectomy with >4 positive nodes; no chemo for HR- and no hormonal therapy for HR+ tumors >1 cm; no trastuzumab for Her2+ tumors. We interviewed key informants about tool usefulness. We conducted intention to treat and pre-post analyses to assess tool and implementation effectiveness, respectively.

RESULTS: Pre-intervention, despite randomizing hospitals, intervention (INT) hospitals had fewer whites (4% vs 14%; p = 0.0005), poorer follow-up approaches (0.68 vs 0.80; p = 0.07), less Medicaid & uninsured patients (36% vs 62%; p < .0001) and more underuse (28% vs 15%; p = 0.002) compared to control (CNTL) hospitals; comorbidities and stage were similar. The RCT found no difference in underuse rates (9% at INT & 11% at CNTL hospitals; p = 0.8). Because randomization did not result in equivalent distributions, we modeled pre- (N = 403) and post (N = 191) populations controlling for time period and clustering & found that hospitals with better follow-up (OR = 0.82; 95% CI: 0.71–0.96) had less underuse. In settings with poor follow-up & tracking approaches, key informants found the tool useful.

CONCLUSIONS: While the RCT findings were negative, they suggest a T&F tool may help reduce underuse in SNHs with poor follow-up capabilities.

CAN WE LINK STANDARDIZED ASSESSMENT OF RESIDENTS’ CLINICAL SKILLS WITH PATIENT OUTCOME DATA? Adina Kalet 2; Colleen C. Gillespie2; Lisa Altshuler2; Heather Dumorne2; Kathleen Hanley2; Andrew B. Wallach1; Barbara Porter2; Sondra Zabar2. 1Bellevue Hospital, New York, NY; 2New York University School of Medicine, New York, NY. (Control ID #2702671)

BACKGROUND: At Bellevue Hospital Center (BHC), we have a robust Unannounced Standardized Patient (USP) program, where trained actors portraying real patients in the clinical setting, incognito, assess the residents’ skills following their visit. We sought to determine the relationship between USP ratings of residents’ skills and clinical outcomes among the residents’ continuity patient panels to define educationally sensitive patient outcomes.

METHODS: We assembled a retrospective cohort of PGY 2 internal medicine residents with at least 2 USP visits between 7/1/14-6/30/15 and ambulatory care patient panels at BHC. The two outcome variables were the percentage of hypertensive patients in the residents’ panel with blood pressure (BP) <140/90, and the average of the most recent glycosylated hemoglobin (HbA1C) result among the residents’ patients with diabetes. The predictor variables included mean USP ratings of residents’ clinical skills and mean faculty rating of the residents’ clinic notes (scored for quality on a 0 to 3 scale). USPs used a behaviorally anchored checklist (not done, partly done, well done) for the following domains: communication, case specific assessment, patient education, physical examination, professionalism, management plan, patient satisfaction, and patient activation measure. We tested the correlations between USP scores with BP and HbA1C control, and then developed multivariate, linear regression models of USP scores on BP and HbA1C scores, respectively, each controlling for Avg. Chronic health score (ACHS, derived by scoring different clinical conditions by acuity and used to determine if the panel is getting sicker over time) and total number of patients in the panel (TNPP) because these variables were correlated with both the outcome and predictor variables.

RESULTS: 29 PGY 2 residents had a mean of 2.5 (SD 1.0) USP visits during the study period. Residents’ patient panels size varied (median 124, range 62–171) and mean patient age was 48 years (SD 1.4). Patient Activation scores were correlated with Average Chronic Health Score (r = .482, p = .008) and Panel Average last A1c (r = −.311, p = .10). Patient activation scores explained 16% variance in the mean panel last HgA1c, (adjusted R2 .137, p = .08). Case specific Assessment & Patient Education skills across USP cases explained 21.5% of the variance and the Average Chart Note Score explained 14.4% of the variance in % of Hypertension controlled (adjusted R2.378, p < .009).

CONCLUSIONS: This exploratory study suggests that learnable resident clinical skills are associated with quality of care indicators for HTN and DM control. In particular, being able to activate patients, assess and educate them and write high quality notes are pathways to quality care. Next steps are to confirm these findings in a larger dataset. Doing so will help align medical education with patient safety and care quality and provide guidance for educational and clinical research aimed at improving the health of populations served.

CARDIOVASCULAR GENETIC RISK TESTING FOR TARGETING STATIN THERAPY IN THE PRIMARY PREVENTION OF CARDIOVASCULAR DISEASE: A COST-EFFECTIVENESS ANALYSIS Jamie A. Jarmul 2, 3; Mark J. Pletcher4; Stephanie Earnshaw5; Morris Weinberger3; Daniel Jonas2; Christy Avery3; Kristen Hassmiller Lich3; Stephanie Wheeler3; Michael Pignone1. 1Dell Medical School, UT-Austin, Austin, TX; 2UNC-Chapel Hill, Durham, NC; 3Gillings School of Global Public Health, Chapel Hill, NC; 4University of California-San Francisco, San Francisco, CA; 5Research Triangle Institute, Research Triangle Park, NC. (Control ID #2705618)

BACKGROUND: It is unclear whether additional risk stratification using novel risk factors improves clinical decision-making about statin use for primary prevention in patients with intermediate atherosclerotic cardiovascular disease (ASCVD) risk beyond traditional risk factors. Cardiovascular genetic risk testing is one option for risk stratification. The objective of this study was to estimate the cost-effectiveness of testing for a 27-SNP cardiovascular genetic risk score (cGRS) to target statin therapy in the primary prevention of ASCVD.

METHODS: The UNC-RTI Coronary Heart Disease Prevention Model is a state-transition Markov model that can be used to compare incidence of ASCVD, mortality, quality of life, and costs with and without a prevention intervention, for specific clinical scenarios. In the model, a specific clinical scenario is defined by age, sex, and ASCVD risk factors, including systolic blood pressure, total cholesterol, HDL cholesterol, smoking status and anti-hypertensive medication use in non-diabetic, ASCVD-free individuals. We updated this model to evaluate the cost-effectiveness of testing for a 27-SNP cGRS. We tested a set of clinical scenarios that included 45-year-old, 55-year-old, and 65-year-old men and women, each with an intermediate 10-year ASCVD risk of 7.5%. Our primary outcome measure was cost per quality-adjusted life-year (QALY) gained. We performed one-way and two-way deterministic sensitivity analyses for key parameters, including statin disutility, statin medication costs and cGRS testing costs, as well as a global probabilistic sensitivity analysis.

RESULTS: Under base case assumptions for statin disutility (disutility = 0.001) and cost ($4/month), the preferred strategy is to treat all patients at 7.5% risk with statins without cGRS testing. Results were similar for patients at lower or higher ASCVD risk (2.5–10%).In deterministic sensitivity analyses, we found that for certain clinical scenarios, such as a 65-year-old man with a 10-year predicted ASCVD risk of 7.5%, cGRS testing can be cost-effective under a very limited set of assumptions; for example, when the cost of obtaining a cGRS test is $100, statin cost of $15/month and statin disutility is 0.013, the preferred strategy (using a willingness-to-pay of $50,000/QALY) is to obtain a cGRS test and treat if cGRS is intermediate or high. However, even when a cGRS testing strategy was preferred, the probabilistic sensitivity analysis showed that the probability of cost-effectiveness at a willingness-to-pay threshold of $50,000/QALY was less than 50%.

CONCLUSIONS: Our analyses demonstrate that testing for a 27-SNP cGRS is not a cost-effective approach for targeting statin therapy in the primary prevention of ASCVD.

CARDIOVASCULAR HEALTH EFFECTS OF 100% FRUIT JUICE VERSUS WHOLE FRUIT IN POSTMENOPAUSAL WOMEN: RESULTS FROM THE WOMEN’S HEALTH INITIATIVE Brandon Auerbach 1, 1; Alyson Littman1; Lesley F. Tinker2; Joseph Larson2; James Krieger1, 3; Bessie Young1, 1; Marian Neuhouser2. 1University of Washington, Seattle, WA; 2Fred Hutchinson Cancer Research Institute, Seattle, WA; 3Healthy Food America, Seattle, WA. (Control ID #2702245)

BACKGROUND: One hundred percent fruit juice is rich in nutrients like potassium and polyphenols, but it is also high in naturally occurring sugars and thus may be associated with adverse cardiometabolic health effects. We investigated whether 100% fruit juice and whole fruit were independently related to incident hypertension or incident type 2 diabetes.

METHODS: We included postmenopausal women 50–79 years of age enrolled in the Women’s Health Initiative. The risk of incident hypertension was analyzed in 80,539 participants and risk of incident diabetes in 114,219 participants. One hundred percent fruit juice and whole fruit intake were assessed by baseline food frequency questionnaire. Standardized questionnaires assessed medical history and other characteristics at baseline and every 6–12 months during follow-up. Cox regression, adjusted for demographic, socioeconomic, behavioral, and dietary variables (including total energy intake) was used to estimate hazard ratios (HR) for the associations between 100% fruit juice and whole fruit consumption and incident hypertension and diabetes during a mean of 7.8 years of follow-up.

RESULTS: In multivariable analyses comparing the highest versus lowest quintiles of consumption, there was no association between 100% fruit juice consumption (highest quintile mean 9.5 ± 3.7 ounces/day versus lowest quintile mean 0.5 ± 0.4 ounces/day) and incident hypertension (HR 1.00, 95% CI 0.97–1.03) or incident diabetes (HR 0.98, 95% CI 0.92–1.04). There was also no association between whole fruit consumption (highest quintile mean 2.5 ± 0.6 servings/day versus lowest quintile mean 0.3 ± 0.1 servings/day) and incident hypertension (HR 1.02, 95% CI 0.98–1.05) or incident diabetes (HR 1.03, 95% CI 0.96–1.10).

CONCLUSIONS: Consumption of 100% fruit juice up to 9.5 ounces/day or whole fruit up to 2.5 servings/day was not associated with risk of incident hypertension or diabetes among postmenopausal US women.

CAREGIVER BURNOUT IN A SCRIBE MODEL AND PHYSICIAN FINANCIAL PRESSURES IN OUTPATIENT PRIMARY CARE PRACTICES Anita D. Misra-Hebert 1; Jacqueline Fox2; Lei Kou1; Sarah Schramm1; Michael B. Rothberg1. 1Cleveland Clinic, Cleveland, OH; 2Cleveland clinic, Cleveland, OH. (Control ID #2705757)

BACKGROUND: Primary care redesign initiatives in outpatient practice include evolving roles for front-line staff, including physicians, medical assistants, nurses and administrative staff. Changing roles and added responsibilities may contribute to caregiver burnout and may also affect physicians’ perception of financial pressures. We assessed burnout levels of caregivers practicing in a scribe model vs. usual care and physician perception of financial pressures in outpatient primary care practices in a large, integrated health system.

METHODS: We created a survey including the Maslach Burnout Inventory (MBI) as well as a 1-item question from the RAND-American Medical Association Survey addressing perceived source of financial pressures for physicians. The survey was distributed electronically using REDCap to 813 outpatient primary care employees including physicians, medical assistants, nurses, and administrative staff at diverse practice sites in an integrated health system.

RESULTS: Ten respondents who reported not working in primary care were excluded. Of the remaining 803 participants, 285 surveys were completed (response rate 35%), including 76 physicians, 79 medical assistants(MAs) and 47 nurses; 65% of respondents worked in Internal Medicine and 35% in Family Medicine at 29 practice sites. Twenty-one percent of respondents worked at the health system for < 2 years and 33% for > 10 years. Nine (12%) physicians worked with a scribe and 26 (33%) MAs and 4 (33%) licensed practical nurses (LPNs) performed scribing duties. On the MBI, 38% of physicians scored high for emotional exhaustion, 24% for depersonalization, while 58% scored high on personal accomplishment. Among MAs and LPNs, 27% scored high for emotional exhaustion, 9% for depersonalization, and 45% on personal accomplishment. No significant differences in MBI scores were noted between physicians or MAs and LPNs who worked in a scribe model compared to those who did not. When asked which factors were important in determining compensation, physicians rated the following as “very important:” Factors reflecting your own productivity (45%), Results of satisfaction surveys completed by your patients (23%), Specific measures of quality of care such as rates of preventive care services for your patients (28%), Results of practice profiling i.e. comparing your pattern of using medical resources with that of other physicians (16%), and The overall financial performance of the practice (31%).

CONCLUSIONS: A significant number of physician and non-physician staff in outpatient primary care report burnout even while reporting high personal accomplishment. Interestingly, burnout scores were unrelated to participating in a scribe model vs. usual care. Physicians perceive financial pressures related to their own productivity, quality measures, and patient satisfaction as very important to their compensation. Efforts to reduce stressors on caregivers in outpatient primary care teams are needed.

CEFTRIAXONE VERSUS AMPICILLIN/SULBACTAM ON THE TREATMENT OF ASPIRATION PNEUMONIA -A PROPENSITY SCORE ANALYSIS FROM A DATA OF JAPANESE MULTICENTER REGISTRY Shinya Hasegawa 1; Atsushi Shiraishi2; Makito Yaegashi1. 1Kameda Medical Center, Kamogawa, Japan; 2Kameda Medical Center, Chiba, Japan. (Control ID #2696173)

BACKGROUND: Incidence of aspiration pneumonia is increasing in Japan in association with increase of the number of elderly people. In Japan, ampicillin/sulbactam (ABPC/SBT) and ceftriaxone (CTRX) is preferred for the treatment of aspiration pneumonia with the aim of covering the full spectrum of the oral anaerobes and for the treatment of community acquired pneumonia (CAP), respectively. However, comparison of ABPC/SBT with CTRX on the treatment of aspiration pneumonia has not been investigated.

METHODS: This study is an analysis of prospectively registered data from 4 Japanese hospitals for patients with community-onset pneumonia from September 2011 to January 2013 (the Adult Pneumonia Study Group-Japan: APSG-J). Aspiration pneumonia is defined as the one including at least one of the aspiration related factors which was set preliminarily. A total of 637 consecutive patients with aspiration pneumonia treated by CTRX or ABPC/SBT were enrolled. Propensity score was estimated from the 29 pretreatment variables including age, sex, comorbidities, use of oral steroids, aspiration related factors, vital signs, laboratory data, and findings of a chest x-ray. Propensity score matching successfully got 164 patients into CTRX and ABPC/SBT groups. The primary endpoint defined as in-hospital mortality was compared between the two groups.

RESULTS: In the propensity-matched cohort, the primary endpoint was observed in 22 patients (6.7%) over a median follow-up period of 25.9 days. In-hospital mortality rates were not significantly different with CTRX compared with ABPC/SBT (4.3% [95% CI: 1.7–8.6] versus 9.1% [95% CI: 5.2–14.7], P = 0.12).

CONCLUSIONS: This study found that mortality benefit of patients treated with CTRX as a first line treatment for aspiration pneumonia was not superior to that with ABPC/SBT.

CENTRAL ROLE OF RELATIONSHIPS IN PROMOTING CAREERS IN GLOBAL HEALTH Brent C. Williams; Jason Bell; Katherine Hughey; Patricia Mullan. University of Michigan, Ann Arbor, MI. (Control ID #2700138)

BACKGROUND: Medical school curricula in global health most often center around providing coursework and structured, often scholarly, field experiences. Few studies have examined the role of non-curricular aspects of students’ experience in facilitating careers in global health. To guide and refine the Global Health and Disparities (GHD) Path of Excellence at the University of Michigan Medical School, now in its sixth year, we examined the relative value of curricular and non-curricular aspects of the GHD Path to medical students.

METHODS: The GHD Path includes: a) four-year mentoring relationship with a GHD Advisor; b) completion of a scholarly field project; c) small group activities in the second year that include group exercises and interactions with GHD faculty and senior students; and d) a team-based Mini Field Project in the second year focusing on leadership skills. In the spring of 2016 we administered a survey to the 41 graduating UM Medical students who participated in GHD throughout medical school. For each component of the GHD Path, students were asked to rate the extent to which {component of GHD} “provided VALUE to you” and “provided a positive IMPACT on your professional development.” Response categories were “Strongly Disagree (SD)”, “Disagree (D)”, “Neutral (N)”, “Agree (A)” and “Strongly Agree (SA)”.

RESULTS: Twenty-seven (67%) of the 41 students completed surveys. Other than the capstone project, all components rated as high value or impact by >80 % of students concerned relationship-building (Table).

CONCLUSIONS: Building personal and professional relationships is as important as field experience, and more important than coursework, to developing a career in global health among medical students. Possible mechanisms include role modeling, increased motivation through sense of community, and resources available through social networks. Providing these results are confirmed in future studies, programs designed to promote careers in global health should create, nurture, and measure opportunities for students to develop personal and professional relationships related to their career paths.

Percent of students who Agree/Strongly Agree that “{GHD activity} provided ‘VALUE to you’ and a ‘Positive IMPACT on your Professional Development’“

Your GHD Advisor 89 78
Interaction with other students 89 89
Capstone project 82 81
Other GHD Faculty 78 81
Meet the professor dinners 74 52
M2 mini-field project 67 67
Noon seminars 67 48
Small group seminars 63 52

CENTRALIZED VACCINE REMINDER/RECALL TO IMPROVE ADULT VACCINATION RATES AT AN URBAN SAFETY NET HEALTH SYSTEM Laura P. Hurley 1, 3; Brenda Beaty4, 3; Steven Lockhart3; Dennis Gurfinkel3; Kristin Breslin1; L Miriam Dickinson2, 3; Anne Libby2, 3; Allison Kempe3, 2. 1Denver Health, Denver, CO; 2University of Colorado, Aurora, CO; 3University of Colorado, Anschutz Medical Campus, Aurora, CO; 4Unviersity of Colorado, Aurora, CO. (Control ID #2702047)

BACKGROUND: Adult vaccination rates are well below Healthy People 2020 goals. Reminder/recall (R/R) is a proven method to increase immunizations, but is underutilized. Centralized R/R using a public health collaborative reduces the burden of an individual practice conducting reminder/recall, and, in prior pediatric trials, has been shown to be effective at increasing immunization rates. However, centralized R/R has not been tested in adult populations. Our objectives were to assess effectiveness and implementation costs of centralized vaccine reminder/recall using the Colorado Immunization Information System (CIIS) vs. usual care for adult vaccine delivery.

METHODS: From September 2015 to April 2016, we randomized 25,039 healthy adults, 16,897 high-risk adults, and 5332 adults ≥ 65 to centralized R/R (R/R) or usual care (TAU) within an urban safety net healthcare system in Denver, CO. In the centralized R/R arm, healthy adults who needed an influenza and/or Tdap, and high-risk adults and seniors who needed an influenza, and/or Tdap, and/or pneumococcal vaccine (PPSV23 or PCV13) were sent up to 3 reminder/recalls by autodialed telephone or mail. Documentation of receipt of any of needed vaccines in CIIS within six months was the primary outcome. We assessed effectiveness of the intervention using multivariable modeling controlling for gender, age, race, ethnicity, insurance type and history of vaccine refusal. Implementation costs were assessed with activity-based accounting for time spent on R/R tasks and supplies in the R/R arm.

RESULTS: In the healthy 19–64 year-old population, 18.6% in the R/R group vs. 17.9% in the TAU group received any needed targeted vaccine (AOR 1.06, 95%CI 0.98-1.14). In the high-risk 19–64 population, 28.1% in the R/R group vs. 26.7% in the R/R group received any needed targeted vaccine (AOR 1.07, 95%CI 1.001-1.15). In the senior population, 33.2% for the R/R group vs. 29.8% for the TAU group received any needed targeted vaccine (AOR 1.15, 95%CI 1.02-1.30). Receipt of influenza vaccine for the 19–64 high-risk and senior populations was the source of the difference in receipt of any needed vaccine in R/R and TAU arms. 26.6% of 19–64 high risk adults in the R/R group received influenza vaccine compared to 25% in the TAU group (p = 0.02) and 32% of seniors in the R/R group received influenza vaccine compared to 28.6% in the TAU group (p = 0.007). Total R/R resource costs were $14,868.16 for the trial, and the cost per person receiving at least one vaccine was $3.29.

CONCLUSIONS: Centralized R/R was effective at increasing influenza vaccination rates in vulnerable adult populations over a short period of time, without burdening the practices and at a low cost.

CHALLENGES WITHIN INPATIENT ROUNDS: A MULTI-INSTITUTIONAL QUALITATIVE STUDY Raphael Rabinowitz 3; Oliver Hulland3; Jeanne M. Farnan2; Lisa Kearns4; Michele Long5; Bradley Monash1; Priti Bhansali6; H. Barrett Fromme2. 1University of California, San Francisco, Mill Valley, CA; 2University of Chicago, Chicago, IL; 3The University of Chicago Pritzker School of Medicine, Chicago, IL; 4The Ohio State University, Columbus, OH; 5University of California San Francisco, San Francisco, CA; 6Children’s National Medical Center, Washington, DC. (Control ID #2701124)

BACKGROUND: Attending rounds is an important activity for both patient care and education at teaching hospitals. There is a relative lack of studies addressing stakeholders’ perceptions of challenges and barriers to the rounding experience. This study characterizes perceptions of medical students, pediatrics and internal medicine housestaff and faculty about the challenges faced during inpatient rounds.

METHODS: The authors conducted focus groups with a purposive sample of medical students and internal medicine and pediatrics interns, residents, and faculty at 4 teaching hospitals to determine their perceptions of the purpose of rounds. Participants volunteered unsolicited comments about challenges faced during rounds without directly being asked this question. The constant comparative method was used to identify themes and codes in a secondary analysis of these comments.

RESULTS: The study identified 4 themes: problems with how rounds are run, challenges to teaching, external problems, and problems specific to family-centered rounds (FCR). Problems with how rounds are run related to challenges that emerged from improper or unsatisfactory rounding facilitation by team leaders, and included excessive variability in rounds, poorly defined roles and expectations, and problems with presentations. Challenges to teaching included tension between education and service, decreased physical exam instruction, and decreased role modeling. External problems included systemic issues that impacted rounds such as time pressures and problems with technology. Problems with FCR were exclusively mentioned during pediatrics focus groups, mostly by residents, and encompassed difficulties introduced by the incorporation of patients and families on rounds, including decreased learning, loss of complexity in discussion, and unhelpful for patients and families.

CONCLUSIONS: Challenges described by study participants varied by training level and specialty. Conflicting frustrations with the rounding experience by its participants indicates widespread misunderstanding of the priorities of other stakeholders in the rounding experience. The acknowledgement of problems related to how rounds are run suggests that faculty development initiatives and resident as teacher (RAT) curricular innovations aimed at standardizing a set of best practices may have a significant impact on educational outcomes for new participants on rounds. More research is needed to identify these practices and the best methods to promote them. A tension between teaching objectives and external problems imposed by a compressed and fragmented workday due to duty hours implementation underscores a need for creative innovations in rounding efficiency to maximize their bedside educational value. Finally, resident-specific concerns about FCR indicates the need to better educate residents about the value of patient interactions and family-centered rounds.

CHANGES IN HEALTH AND WELL-BEING AMONG FIRST-YEAR RESIDENTS Christopher Wee 1; Jacob A. Petrosky2; Lauren Mientkiewicz3; Krishna K. Patel4; Charles Kwon2; Allan Siperstein5; Xiaobo Liu2; Michael Rothberg2. 1Cleveland Clinic Foundation, Cleveland, OH; 2Cleveland Clinic, Cleveland, OH; 3Akron Children’s Hospital, Akron, OH; 4Mid America Heart Institute, Kansas City, MO; 5Cleveland, Cleveland, OH. (Control ID #2703204)

BACKGROUND: Training conditions for physicians may affect their ability to serve as healthy role models to patients. There is limited data, however, on how trainees’ health and wellness habits are affected by GME training. Our objective was to measure differences in self-reported health and health behaviors at the beginning and end of the PGY-1 year.

METHODS: We invited incoming PGY-1 trainees in all specialties at the Cleveland Clinic during orientation to participate in an anonymous online survey, through a mass e-mail invitation. The 31-question instrument included previously validated questions regarding self-reported health habits, well-being, and nutrition practices, including average hours slept, average days exercising per week, and servings of fruits and vegetables consumed. Self-reported health status was assessed on a scale of 1 (excellent) to 5 (poor). At the end of the year, a follow-up survey was sent to those who had responded initially. Baseline and follow-up measures were compared using paired T-test or Wilcoxon signed rank sum test as appropriate.

RESULTS: Of 170 interns who were sent invitations, 59 (35%) started the initial survey and 34 (58%) of these responded to the follow-up survey. Only the 34 individuals who completed both surveys were included in the longitudinal analyses. Average age was 27, 17 were in internal medicine, 5 in a preliminary year, 4 in pediatrics, and the remainder in other specialties. Twelve were married or in long-term relationships, and 50% lived alone. Only 3 of the respondents had children. Over the course of the year, self-reported weight did not change significantly (157.2 lbs initially v 154.8 lbs at follow-up), number of hours slept decreased (7.3 v 6.3 hrs p < 0.001), as did days exercised per week (3.3 v 2.4 days p = 0.005), but the length of workouts did not change. There was no change in alcohol consumption. Breakfast and lunch habits changed significantly, with more people skipping breakfast (17.7% v 41.2%) and eating lunch out at follow-up (38.2% v 70.4%). Dinner habits and consumption of fruits and vegetables did not change. Respondents noted feeling less happy and more worn out at follow-up. When asked to rate their general health, at both the initial and follow-up survey, the median answer was 2 (very good). However, when asked to compare their health at the time of the survey to 1 year prior, during the initial survey, the average respondent answered “about the same,” while during the follow-up survey, they answered, “somewhat worse than 1 year ago” (p < 0.001).

CONCLUSIONS: While training to become health professionals, interns’ own health habits deteriorated. Further studies should explore the relationship between health habits and burnout, and whether adherence to constructive habits could mitigate the effects of trainee burnout.

CHANGES IN SELF-REPORTED GENERAL HEALTH, PHYSICAL HEALTH, AND MENTAL HEALTH FOLLOWING THE AFFORDABLE CARE ACT’S MEDICAID EXPANSION Tyler N. Winkelman 1, 2; Virginia W. Chang3, 3. 1University of Michigan, Ann Arbor, MI; 2VA Ann Arbor Healthcare System, Ann Arbor, MI; 3New York University, New York, NY. (Control ID #2699823)

BACKGROUND: The adoption of Medicaid expansion in some states and not others provided a unique natural experiment to study the effects of Medicaid. Research stemming from this natural experiment suggests that Medicaid expansion increased health insurance coverage, improved access to care, and reduced cost-related barriers to prescription drugs among low income individuals. Findings with respect to health outcomes, however, have been more mixed. Therefore, we analyzed recently released national data from the Behavioral Risk Factor Surveillance System (BRFSS) to assess the relationship between Medicaid expansion and self-reported health measures among low-income individuals.

METHODS: We used 2011–2015 BRFSS data, which provided 3 years of data prior to implementation of Medicaid expansion and 2 years of follow-up data in the majority of expansion states. Our study sample consisted of all individuals age 18–64 with household incomes below $15,000, targeting individuals who would have qualified for Medicaid coverage in expansion states. As in prior work, we excluded five states that had previously expanded Medicaid. Our outcomes were self-reported general health, poor physical health days, poor mental health days, and disability following Medicaid expansion. We used a difference-in-differences approach to estimate the effect of Medicaid expansion on our outcomes of interest. Our key independent variable was equal to 1 for individuals living in states where expansion was in effect during the month of their interview. Estimates were obtained with multivariable linear probability models and adjusted for age, race/ethnicity, sex, education, marital status, and children, as well as state-level and quarter-year fixed effects. We used BRFSS sampling weights and estimated robust standard errors clustered at the state level to account for serial autocorrelation.

RESULTS: In adjusted analyses of the influence of Medicaid expansion, we found that expansion was associated with a significant reduction in fair/poor self-rated health (2.5 percentage points [95% CI, −3.5 to −1.5]). While expansion was not associated with a statistically significant change in the number of poor physical health days (−0.20 days [95% CI, −0.68 to 0.28]), it was associated with a significant reduction in the number of poor mental health days (−0.52 days [95% CI, −0.99 to −0.04]). Change in disability prevalence did not vary between expansion and non-expansion states (P = 0.73). Adjusted linear time trends prior to expansion (2011 to 2013) for all outcomes were similar in expansion and non-expansion states (P > .05 for all comparisons).

CONCLUSIONS: To our knowledge this is the first national study to report positives changes in self-reported general health following the ACA’s Medicaid expansion provision, driven by changes in mental health. Whether these trends continue to improve will likely depend on whether policymakers choose to improve or repeal the ACA in the coming months.

CHANGES TO OUTPATIENT HYPERTENSION AND DIABETES MEDICATIONS IN OLDER ADULTS FOLLOWING UNRELATED HOSPITALIZATIONS Timothy Anderson 1; Siqi Gan1, 2; Kathy Fung1, 2; Ying Shi1, 2; Michael A. Steinman1, 2. 1University of California, San Francisco, San Francisco, CA; 2San Francisco VA Medical Center, San Francisco, CA. (Control ID #2705026)

BACKGROUND: Transient elevations of blood pressure and serum glucose are common in hospitalized older adults. Although these perturbations usually have little long-term significance, they may lead inpatient clinicians to make changes to outpatient regimens, even when patients are hospitalized for conditions that do not typically require aggressive blood pressure or glucose management. Unnecessary changes made to stable outpatient regimens during a hospital stay can easily become over-treatment once patients return home, exposing them to increased risks of serious adverse drug events including hypoglycemia, syncope and falls. We thus sought to evaluate how often patients’ diabetes and hypertension regimens are changed following hospitalization for unrelated conditions.

METHODS: We used national VA and Medicare data to assemble a retrospective cohort of veterans age 65 years and older who received regular care in VA outpatient settings, had hypertension and/or diabetes and were hospitalized in a VA medical center in 2011 with pneumonia, urinary tract infection or venous thromboembolism. Using VA pharmacy dispensing records, we identified medications in use prior to hospitalization and medications prescribed in the peri-discharge period. We compared admission medications to medications in use at discharge to identify changes in regimens. Changes were classified as medication additions, discontinuations and dose changes.

RESULTS: 9,795 veterans were included in our cohort of whom 7,635 had hypertension and 4,079 had diabetes. The majority of patients were male (98%) and the median age was 78 (IQR 70–84). Patients were prescribed a median of 7 medications on admission (IQR 5–10). Patients with hypertension were prescribed a median of 2 anti-hypertensives (IQR 1–3) and patients with diabetes a median of 1 hypoglycemic medications (IQR 0–2). 4,836 (63%) patients with hypertension experienced a change to their anti-hypertensive medications, of which 39% experienced at least one anti-hypertensive discontinuation and 38% at least one anti-hypertensive addition. 1,589 (39%) patients with diabetes experienced a change to their diabetes medications, of which 23% experienced at least one hypoglycemic discontinuation and 20% at least one hypoglycemic addition. Overall, medication changes were common, 94% of patients experienced at least one medication change following hospitalization and the median number of medication changes was 5 (IQR 3–7). The majority of patients experienced at least one medication addition (81%) and at least one medication discontinuation (72%).

CONCLUSIONS: Over half of patients in a national cohort of older veterans hospitalized for common medical conditions experienced changes to their outpatient hypertension and diabetes medications at discharge. These high rates of changes have the potential to impact adherence and expose patients to serious adverse drug events upon discharge.


Caleb Murphy; Jill Bowman Peterson; Alisa Duran. University of Minnesota, Minneapolis, MN. (Control ID #2703276)

BACKGROUND: Daily lab testing despite patients’ clinical and lab stability is identified as an area of inappropriate health care spending in the Society of Hospital Medicine’s Choosing Wisely list.1 Approximately 25% of daily labs are inappropriate, leading to overtreatment and increased spending, yet guidelines for identifying such labs do not exist.2 The objective of this study was to create and apply the Minnesota Lab Appropriateness Criteria (MLAB) to properly identify inappropriate labs compared to the national average of 25%.

METHODS: MLAB were created by two internists at the University of Minnesota experienced in hospital medicine and high value care, reviewed by the Alliance for Academic Internal Medicine High Value Care Workgroup, and revised. 50 medicine admissions (hospitalization 2–10 days, non-ICU, non-cirrhotic patients) were randomly selected. Using MLAB, two reviewers independently rated appropriateness of basic metabolic panels (BMP) and complete blood counts (CBC) from each hospitalization using both a dichotomous scale (DS; appropriate/inappropriate) and a three-point Likert scale (LS; 1 = inappropriate, 2 = equivocal, 3 = appropriate).

RESULTS: 461 daily labs (253 BMPs, 208 CBCs) from the 50 admissions were reviewed. Using MLAB, 24.1% (95%CI 18.8–29.4%) of BMPs and 25.0% (95%CI 19.1–30.9%) of CBCs were rated inappropriate on the DS. On the LS, 20.2% (95%CI 15.2–25.1%) of BMPs were inappropriate and 7.1% (95% CI 4.0–10.3%) were equivocal, while 16.8% (95%CI 11.7–21.9%) of CBCs were inappropriate and 12.0% (95%CI 7.6–16.4%) were equivocal. When comparing raters on the DS, κ was 0.68 (95%CI 0.58–0.78) for BMPs and 0.77 (95%CI 0.68–0.87) for CBCs. Weighted κ on the LS was 0.58 (95%CI 0.49–0.67) for BMPs and 0.62 (95%CI 0.52–0.72) for CBCs.

CONCLUSIONS: Using MLAB, raters identified 24.1% of BMPs and 25.0% of CBCs as inappropriate, consistent with previously reported figures. This suggests MLAB correctly identifies inappropriate daily lab ordering. When assessed on the LS, inappropriate lab rates dropped as more labs were identified as equivocal. Yet appropriate test rates also dropped; thus, equivocal tests were not just comprised of tests previously identified as inappropriate on the DS. Interrater reliability between raters showed moderate to substantial agreement using both the DS and LS; with more agreement on the DS. MLAB offer an accurate and reliable method of assessing BMP and CBC appropriateness, with potentially important applications in high value care initiatives and medical education. These criteria, when applied prospectively, could potentially translate to cost savings across health systems and help reduce unnecessary daily lab testing in the hospital. References: 1. Bulger J, Nickel W, Messler J, et al. Choosing Wisely in Hospital Medicine: Five opportunities for improved healthcare value. J Hosp Med. 2013. 8:486–92. 2. Zhi M, Ding EL, Theisen-Toupal J, et al. The landscape of inappropriate laboratory testing: a 15-year meta-analysis. PLoS ONE. 2013. 8:e78962.

CLINICAL DECISION SUPPORT (CDS) TOOLS FOR ACE INHIBITOR THERAPY IN HEART FAILURE: HELPFUL OR HASSLE? Anne Press 2; Jonathan Austrian1; Saul Blecker2. 1NYU, New York, NY; 2NYU School of Medicine, New York, NY. (Control ID #2706210)

BACKGROUND: Electronic health record (EHR)-based clinical decision support tools (CDS) incorporate individualized data to produce patient-specific recommendations at the point-of-care. However, these tools are often limited in their effectiveness, which may be due to poor consideration of usability. The purpose of this study was to evaluate the utilization of a CDS intervention to increase prescription of Angiotensin Converting Enzyme inhibitor (ACEi) or Angiotensin Receptor Blocker (ARB) for patients with heart failure.

METHODS: We performed a retrospective study of hospitalized patients with heart failure from the time of CDS implementation, 7/10/13, through 11/30/15. The CDS that we investigated offers providers an opportunity to prescribe an ACEi or ARB or report a contraindication to therapy for patients with documented heart failure. All patients with an EF ≤ 40% who were not on an ACEi or ARB at time of discharge were included in the study. We identified the number of patients for whom the CDS triggered; of those, we categorized provider response as: dismissed, ordered an ACEi/ARB, or contraindication reported. We then performed manual chart review to identify the CDS reported contraindication and structured chart abstraction with standard guidelines to identify gold standard contraindications. We compared each CDS contraindication to gold standard contraindications to determine their accuracy.

RESULTS: Out of the 618 subjects who had an EF ≤ 40% but no ACEi or ARB at the time or discharge, 435/618 (70%) had a triggered CDS. Of these 435 subjects for who a CDS was triggered, 180 (41%) were dismissed, 225 (52%) had a contraindication response and 30 (7%) had a prescription for an ACEi/ARB therapy. Overall the accuracy of the documented CDS was 42% (Table 1).

CONCLUSIONS: The CDS that we reviewed was poorly utilized and contraindications documented in the tool poorly correlated with patient clinical status reflected elsewhere in the EHR. These findings identify this CDS as a possible impedance to user workflow. One way to improve CDS tools at the point of care is through thorough usability testing and consideration of physician workflow prior to implementation.

Table 1: Clinical Decision Support (CDS) recorded contraindication to therapy and accuracy of reported contraindication versus gold standard of chart abstraction. Contraindications are listed in order of display in the CDS tool

Contraindication: CDS Accuracy
Allergies 4 50%
Aortic Stenosis 4 100%
Patient refusal 9 Not Assessed
Angioedema 0
Hyperkalemia 9 78%
Hypotension 43 34%
Renal Artery Stenosis 2 0%
Renal impairment 135 44%

*there were 19 cases with contraindication reported as “other”

CLINICAL DECISION SUPPORT DECREASES VARIABILITY IN CTPA YIELD Safiya Richardson; Lauren McCullagh; Sundas Khan; Vinodh Mechery; Guang Qui; Salvatore Pardo; Thomas McGinn. Hofstra Northwell School of Medicine, New York City, NY. (Control ID #2705685)

BACKGROUND: Effective clinical decision support (CDS) will be instrumental in bringing the best available evidence to the point of care. Current estimates indicate that only 20–50% of care provided is evidenced based. CDS has demonstrated an ability to improve management and standardize care. Practice variability is a particular challenge in Emergency Rooms (ERs) where overcrowding and medical urgency lead to an overdependence on intuition. Here we focus on the impact of a CDS tool designed to standardize provider estimation of pre-test probability of pulmonary embolism (PE) before computed tomography pulmonary angiogram (CTPA). We measure the change in variability of the percent of tests positive for PE (CTPA yield) after CDS implementation.

METHODS: A CDS tool was built in Allscripts Sunrise Electronic Medical Record at a large academic tertiary care hospital. Tool design was based on previously published work examining ER workflow and usability testing of triggers. Orders for CTPA, V/Q scan or d-dimer are routed to the PE Wells Calculator which presents Wells’ Criteria, calculates risk of PE and offers an order set based on risk. We performed a before and after analysis. To avoid potential seasonal variation, due to the start of new trainees in July, we determined the pre-intervention control period to be April to June, 2014 and the post-intervention period to be April to June, 2016. The primary outcome event was CTPA yield, a validated measure of CTPA ordering accuracy. Only providers who ordered CTPAs during both the pre- and post-intervention period were included in the analysis. Variability was assessed using the coefficients of variation.

RESULTS: A total of 45,105 patients were seen in the ER during the analysis periods, resulting in 910 CTPA orders completed to evaluate for PE. During the pre-intervention period monthly CTPA yield varied from 4 to 9%, demonstrating wide variability in accuracy of provider assessment of pre-test probability for PE. During the post-intervention period the yield demonstrated less variability, varying from 10 to 12%. The coefficient of variation was 0.67 during the pre-intervention period vs 0.06 during the post-intervention period. These were significant results as there was non-overlap of the 95% CI for the coefficients of variation for pre- and post- intervention.

CONCLUSIONS: CDS designed to assist provider estimation of pre-test probability of PE decreased monthly variability in CTPA yield in the ER. Thoughtfully designed and extensively usability tested CDS can improve quality and standardize care.

CLINICAL MEDICAL STUDENT RESILIENCE AND EXPERIENCES WITH STRESSFUL CLINICAL EVENTS Jennifer Houpy; Wei Wei Lee; James N. Woodruff; Amber Pincavage. University of Chicago, Chicago, IL. (Control ID #2693005)

BACKGROUND: Although medical students face numerous stressors during their clinical years, little is known about their resilience. The objective of this study was to characterize medical student resilience and experiences with difficult clinical events.

METHODS: Anonymous electronic surveys were provided to all third year (MS3) and fourth year (MS4) medical students at the University of Chicago in the spring of 2016 to assess resilience (10 item Connor Davidson Resilience Scale (CD-RISC 10)), experiences with difficult clinical events, and burnout (non-proprietary single-item burnout measure).

RESULTS: 62 MS3s and 55 MS4s (response rate 66%) completed surveys. The mean CD-RISC 10 score was 28.2 ± 6.37 (range 10–40, possible range 0–40, 40 indicating most resilient) and lower than in a general population sample (32.1 ± 5.80, p < 0.001). Mean resilience was higher in males (30.47 ± 6.14 vs. 26.43 ± 6.02, p = 0.001), MS4s (29.68 ± 5.98 vs. 26.91 ± 6.47, p = 0.023), and those reporting no burnout symptoms (30.44 ± 5.44 vs. 25.00 ± 6.29, p < 0.001). There was no significant difference based on age, undergraduate major, or path to medical school. Over 80% of students had experienced the following clinical events: dealing with difficult patients, difficult family discussions, systems issues, poor team dynamics, chronic narcotic patients, and difficult encounters with other staff. 55% had experienced medical errors. Students found poor team dynamics most stressful. After difficult clinical events, 71% of students reflected on them often, 62% would prefer to discuss them with their team that same day, and only 5% of would prefer not to discuss them. Most students had talked to peers (91%) about difficult clinical events, while only 37% discussed them with their attending and 60% with the team resident. MS4s spoke with attendings (48% vs. 27%, p = 0.043) and residents (72% vs. 49%, p = 0.017) more than MS3s did. Resilience was higher in students who reported having skills to cope with difficult clinical events (29.47 ± 5.91 vs. 22.98 ± 5.93, p < 0.001) and who were comfortable discussing medical errors with peers (30.48 ± 5.99 vs. 25.14 ± 5.64, p < 0.001). Most students (64%) believed resilience training would be helpful and best delivered during third year (66%). Only 27% believed they had sufficient training. The top topics identified included difficult team interactions, finding meaning in daily work, and dealing with disappointment.

CONCLUSIONS: Resilience in clinical medical students was lower than in the general population. Students had some insight into their resilience. Although students wanted to discuss difficult clinical events with their team, most students discussed them with peers. Students endorsed a need for resilience training during the third year. More curricula promoting resilience are needed.

CLINICIAN BURNOUT IN SMALL NORTH CAROLINA PRIMARY CARE PRACTICES APPEARS LOWER THAN EXPECTED: AN INITIAL SNAP SHOT FROM EVIDENCE NOW. Samuel Cykert 2; Darren A. DeWalt1; Bryan Weiner3; Janet Freburger4. 1UNC School of Medicine, Chapel Hill, NC; 2University of North Carolina, Chapel Hill, NC; 3University of Washington, Seattle, WA; 4University of North Carolina Chapel Hill, Chapel Hill, NC. (Control ID #2706811)

BACKGROUND: Given the rapid implementation of electronic health records (EHR), new reporting programs, and the looming changes anticipated from the Merit-based Incentive Payment System and alternative payment models, healthcare leaders and policymakers fear that clinicians, especially in small practices, will experience severe burnout. Evidence Now is a national initiative funded by AHRQ designed to reduce cardiovascular risk among adult patients in small primary care practices and study the effect of practice facilitation as a means of accelerating this effect. We report initial results of a baseline survey of North Carolina (NC) clinicians who are participating in the project. We focus on a validated, single item measure of burnout. Prior studies using national samples documented 60% of the workforce experiencing burnout.

METHODS: The full practice member survey was sent to clinicians from 186 practice sites between February and November of 2016. The items include a validated single item burnout question, an adaptive reserve scale (ARS), and multiple questions concerning the use of evidence-based guidelines. The burnout item consisted of 5 possible responses: 1) I enjoy my work. I have no symptoms of burnout. 2) Occasionally I am under stress and I don’t always have as much energy as I once did but I don’t feel burned out. 3) I am definitely burning out and have one or more symptoms of burnout such as physical and emotional exhaustion. 4) The symptoms that I’m experiencing won’t go away. I think about frustrations a lot. 5) I feel completely burned out and often wonder if I can go on. I am at the point where I may need some changes or may need to seek some sort of help. We also obtained information on payer mix, personnel, and EHR satisfaction.

RESULTS: 158 practice member surveys were received from the practice sites. The mean burnout score was 1.87 (SD +/− 0.5). 18% reported no burnout while 60% agreed with response 2. 20% of clinicians admitted to burnout described in response 3 while 1% said the symptoms were persistent (response 4). No one agreed to the most severe response (5). A negative correlation between ARS and burnout was strong (r = −.57, p < 0.001). Linear regression was performed to identify factors associated with worse burnout including EHR satisfaction, FTE clinicians, FTE staff, provider visits per day, practice ownership, payer mix, PCMH status, regular data discussions, and ARS. The r-squared for the model was 0.38. PCMH recognition and FQHC status were associated with higher burnout while high adaptive reserve and more visits per provider were associated with less.

CONCLUSIONS: Burnout was lower than expected in a large cross section of clinicians in small NC primary care practices suggesting a resilient workforce in these practices. While the associations of higher burnout in FQHC’s and lower burnout with high adaptive reserve were expected, associations with PCMH recognition and higher provider volume were somewhat paradoxical and need further exploration.

CLOSING THE SPECIALTY REFERRAL LOOP: AN IMPERATIVE FOR A RESILIENT PRIMARY CARE NETWORK Malhar Patel; Colin O’Leary; Priscille Schettini; Kevin Shah. Duke University School of Medicine, DURHAM, NC. (Control ID #2702417)

BACKGROUND: A critical function of large primary care practices is connecting patients with specialists. Consistently closing the loop on all patient referrals is a patient safety imperative as identified by The Joint Commission. Our project sought to identify referral completion rates, gaps in referral documentation, and how wait times of referrals affect completion rates.

METHODS: From a large primary care network, we analyzed specialty referrals (N = 106,885) in FY2016 across 22 high volume subspecialties, excluding procedural and ancillary service referrals. For referrals with appointment scheduling data, we characterized appointment completion, cancellation rates, and wait times. We stratified these referrals into three categories of wait times (<30 days, 31–60 days, and >60 days) and analyzed each category’s completion rates.

RESULTS: For 93,584 referrals, a total of 106,885 appointment scheduling attempts were made in FY2016. Some unique referrals had multiple scheduling attempts. Of the 106,885 appointment scheduling attempts, 63,360 (59.3%) had known appointment dates, while 43,525 (40.7%) referrals had no recorded appointment date. Of the referrals without appointment dates, 40.3% were indicated as scheduled (although no date was provided) and 59.7% were not scheduled for reasons including patient self-scheduling or declining. We restricted our analysis further to the 63,360 appointments with documented appointment dates. Of these, 56.9% resulted in completed appointments, 29.3% in cancelled appointments, 7.3% in appointments scheduled after FY2016, and 6.5% in patient no shows. Of canceled referrals, 53.6% were rescheduled, while 46.4% were not. Almost a third of referrals (30.9%) had wait times longer than 30 days. Average wait times for each specialty ranged from 10.6 to 84.7 days. Referrals in wait time categories <30 days, 31–60 days, and >60 days had appointment completion rates of 65.8%, 43.3%, and 29.3%, p < .001, respectively. Sub-analyses of 3 specialties (endocrinology, nephrology, and urology) demonstrated that, while wait times varied by specialty, the rate of appointment completion consistently decreased as wait times increased.

CONCLUSIONS: Our analysis of specialty referrals in a large academic primary care practice demonstrates opportunities to close the loop on specialty referrals. Specifically, we found numerous referrals that were not scheduled, did not have appointment dates, or did not have documented completion. Our analysis showed a correlation between increasing wait time and decreasing referral completion rate, as well as wait time variability by specialty. These analyses are likely generalizable to primary care practices making a large number of referrals. This study underscores the need for methods to track patients and ensure the loop is closed between primary and specialty care. This would not only improve patient safety and care, but also advance the Quadruple Aim by reducing burnout in health care providers frustrated by incomplete care for their patients.

CLUSTER ROUNDS: A NEW TECHNIQUE IN BEDSIDE MEDICINE Nousha Hefzi; Emmanuel Akintoye; Courtney M. Moore; Sajith Matthews; Diane L. Levine. Wayne State University School of Medicine, Detroit, MI. (Control ID #2707041)

BACKGROUND: Proficient physical examination (PE) skills are critical tools in physical diagnosis. A decline in the ability to correctly perform the PE has led to adverse preventable medical errors including missed or delayed diagnosis and incorrect diagnosis.[1] To strengthen the PE skillset of junior medical students on the Internal Medicine clerkship, we utilized an innovative method to teach PE and diagnosis. “Cluster rounds” involves an attending physician gathering multiple patients with the same disease to demonstrate the variability of findings in patients with the same diagnosis. The goal of cluster rounds is to 1) improve PE skills, 2) improve diagnostic skills through the PE, 3) provide an opportunity for direct observation of bedside skills, and 4) increase the confidence of medical students when performing a history and PE on their own.

METHODS: Cluster rounds involve a 5 min history and a 10–15 min PE per patient for a given diagnosis. 11 sessions with 4 patients per a session were conducted. Each session focused on a single diagnosis (endocarditis, congestive heart failure, inflammatory bowel disease, breast cancer, osteomyelitis, arthritis/gout, lead toxicity, opioid toxicity, COPD, and cirrhosis). 8–9 students attended each session with 4 performing the actual PE. Pre- and post-session surveys using a Likert scale were developed to assess 70 student impressions. Impact of the cluster rounds on the students’ self-perception was assessed in two ways: first, by percentage increase in perception to at least the level of ‘strongly agree’ using McNemar’s test. Second, by the difference in the median Likert score before and after cluster round using Wilcoxon signed-rank test.

RESULTS: Prior to the cluster round, most participants felt to some degree that they were able to perform the PE on their own without assistance. However, only 2–14% of the participants strongly agreed to each of the questions. Compared to the pre-assessment, there was significant increase in the proportion of participants that moved to the level of ‘strongly’ or ‘very strongly’ for able to perform (14% vs 69%, p < 0.001), comfortability (12% vs 62%, p < 0.001), confidence (7% vs 60%, p < 0.001), able to teach (2% vs 52%, p < 0.001) and able to use PE finding to make diagnosis (14% vs 67%, p < 0.001). In addition, there was significant increase in the median Likert score across all endpoints: 3 vs 4 (p < 0.001), 3 vs 4 (p < 0.001), 3 vs 4 (p < 0.001), 2 vs 4 (p < 0.001), and 3 vs 4 (p < 0.001) respectively.

CONCLUSIONS: All participants agreed that cluster rounds complemented their IM clerkship curriculum and was valuable in enhancing their PE skillset. This preliminary data illustrates the potential for this innovative method to enhance students’ clinical practice skills in comfortably performing a PE and utilizing its findings to make a diagnosis, as well as teaching the exam to peers. With further studies, cluster rounds can potentially be a critical educational experience with regards to students’ comfort in PE.

COLLABORATE AS A MEASURE OF PATIENT-CENTERED CARE: A VALIDATION OF A POTENTIAL POINT-OF-CARE ASSESSMENT IN THE VA. Barbara G. Bokhour 1, 3; Laurel Radwin2; Glyn Elwyn4; Mark Meterko5. 1ENRM Veterans Affairs Medical Center, Bedford, MA; 2VA Boston Healthcare System, Boston, MA; 3Boston University School of Public Health, Boston, MA; 4Dartmouth Medical School, Hanover, NH; 5US Department of Veterans Affairs, Bedford, MA. (Control ID #2705971)

BACKGROUND: Many organizations have identified patient-centered care as a critical aspect of quality. Yet it is unclear how and when to best measure patients’ perceptions of the extent to which their care meets the criteria of patient-centeredness. The objective of this paper was to assess the validity of the CollaboRATE, a brief measure of patient-centered communication (PCC) with potential for point-of-care administration. The CollaboRATE asks respondents to evaluate their providers’ efforts to (1) help them understand their health issues, (2) listen to the things that matter most to them about their health issues, and (3) include what matters most in choosing what to do next in the patient’s care.

METHODS: We conducted a mail survey of outpatients (n = 1019) and inpatients (n = 767) at 8 Department of Veterans Affairs medical centers, half of which were designated as leaders in patient-centered care. Surveys included 1) the CollaboRATE; 2) the Communication Assessment Tool (CAT), an established, longer measure of patient-centered communication; 3) measures of patient satisfaction; 4) self-reported health and functional status (PROMIS −29); and 5) the Lorig measure of healthcare self-efficacy. We calculated Cronbach’s alpha to assess reliability, and Pearson correlations to assess concurrent validity with the CAT and construct validity with the various proximal and distal patient-reported outcomes.

RESULTS: Internal consistency reliability of the CollaboRATE was excellent in both the inpatient and outpatient samples (r = 0.96, 0.97, respectively). Concurrent validity of the CollaboRATE was demonstrated by statistically significant and strong correlation with the CAT (r. = 0.84, 0.85); p < .001). Construct validity was demonstrated in both samples by a strong and statistically significant correlation with the proximal outcome of overall care satisfaction in the outpatient and inpatient samples (r = 0.81 and 0.74, respectively; p < .001), and by weak to modest but conceptually consistent and statistically significant positive relationships with the more distal outcomes of patient-reported physical health (r = 0.12, 0.15, p < .001) and social functioning (r = 0.18, 0.21, p < .001) and healthcare self-efficacy (r = 0.23, 0.25, p < .001), and negative relationships with anxiety, depression, fatigue, and sleep quality and disturbance (r’s −0.10 to −0.27).

CONCLUSIONS: We found strong evidence of reliability and validity for the CollaboRATE in both inpatient and outpatient samples of Veterans. These findings support the use of the CollaboRATE as a brief measure of patient-centered communication. In the outpatient setting, CollaboRATE could be administered immediately following an outpatient visit using any of a variety of modes of administration such as, a Kiosk check-out stop, hand-out postcard-size survey with drop-off box, or using a survey application on a tablet or laptop computer. In the inpatient setting this could be conducted at bedside or immediately prior to discharge.

COLLABORATIVE CARE FOR DEPRESSION AMONG PATIENTS WITH LIMITED ENGLISH PROFICIENCY: A SYSTEMATIC REVIEW Maria E. Garcia 3; Lisa Ochoa-Frongia3; Nathalie Moise1; Adrian Aguilera4; Alicia Fernandez2. 1Columbia University Medical Center, New York, NY; 2UCSF, San Francisco, CA; 3University of California, San Francisco, San Francisco, CA; 4University of California, Berkeley, Berkeley, CA. (Control ID #2706421)

BACKGROUND: Individuals with limited English proficiency (LEP) face unique challenges in accessing and engaging in mental health care. While many studies report improvement in depressive symptoms among patients enrolled in collaborative care, few studies have evaluated the effectiveness of this model among LEP individuals, whose depression rates are high.

METHODS: To investigate the effectiveness of collaborative care in the treatment of depressive symptoms among individuals with LEP, a systematic review of English articles was performed using online PubMed, PsychINFO, CINAHL and EMBASE databases thru September 15, 2016. Studies were included if they were RCTs, cohort, or case–control studies evaluating collaborative care for depression in primary care among LEP patients. Included studies described the 3 key components of collaborative care: 1) care coordination and care management, involving multi-professional patient care with a social worker or therapist, primary care providers, and a consulting psychiatrist, 2) regular/proactive monitoring and treatment to target using validated clinical rating scales by a designated care manager, and 3) regular, systematic psychiatric caseload reviews and consultation for patients demonstrating no clinical improvement. Two reviewers independently examined titles and abstracts and extracted information on participants’ characteristics, type of intervention, features of collaborative care, preferred participant language, and outcome measures.

RESULTS: The search yielded 15 papers, representing 9 studies: 5 RCTs, 3 cohort studies and 1 case–control study. The studies had 4885 total participants and 2564 LEP participants (52%). The majority of LEP participants (2335, or 91%) spoke Spanish. The heterogeneity in study design and outcome definitions precluded a meta-analysis; only 2 studies had the same outcome definition. Follow-up times varied greatly, and ranged from 3 months to 2 years. Four of five RCTs reported more improvement in depressive symptoms among individuals in the intervention group compared to the usual care group. The last RCT had similar improvement between intervention and control groups but had high rates of psychiatry follow-up among both groups. Among non-RCTs, 2 studies reported that Spanish language preference was associated with depressive symptom improvement; the last 2 non-RCTs demonstrated improvement in depressive symptoms but had no usual care control groups.

CONCLUSIONS: All but one study reported improved depressive symptoms in LEP individuals treated with collaborative care. While limited by the small number of LEP participants (particularly Asian patients) and the heterogeneity of study designs and outcomes, this systematic review suggests that collaborative care may be an effective model for the treatment of depression among LEP populations. Future studies should more explicitly detail participants’ language preference and cultural adaptations made to the model to treat LEP individuals.

COLLABORATIVE CARE TO ALLEVIATE SYMPTOMS AND ADJUST TO ILLNESS (CASA): PRIMARY EFFICACY RESULTS FROM THE CASA RANDOMIZED CLINICAL TRIAL OF A PALLIATIVE SYMPTOM AND PSYCHOSOCIAL CARE INTERVENTION IN HEART FAILURE David Bekelman 5, 4; Larry Allen5; Brack Hattler4; Edward P. Havranek2; Diane Fairclough1; Connor F. McBryde3; Paula Meek5. 1Colorado School of Public Health, Auroa, CO; 2Denver Health Medical Center, Denver, CO; 3Denver VA Medical Center, Denver, CO; 4Eastern Colorado Health Care System, Denver, CO; 5University of Colorado AMC, Aurora, CO. (Control ID #2704174)

BACKGROUND: Palliative care provided by specialists shows promise in improving symptoms and quality of life in patients with chronic heart failure. However, there is limited high-quality data for what works in the outpatient setting. Furthermore, because there are relatively few palliative care specialists, scalable interventions that can be used in routine outpatient care are needed. The CASA trial determined whether a team-based intervention improved health status (i.e., symptoms, function, and quality of life) and other outcomes in outpatients with heart failure compared to usual care.

METHODS: Patients with heart failure and poor self-reported health status were recruited from a VA, an academic health system, and an urban safety net health system. Patients with dementia, metastatic cancer, major mental illness, or active substance abuse were excluded. The CASA intervention included a nurse who addressed persistent symptoms (e.g., shortness of breath, fatigue, pain) and a social worker who provided psychosocial care. Patients were also reviewed with a study primary care provider, cardiologist, and palliative care physician who wrote orders for medications and tests for patients’ primary providers to consider. The primary outcome was heart failure-specific health status at 6 months, measured using the Kansas City Cardiomyopathy Questionnaire (range, 0–100, higher is better; the study was designed to have 90% power to detect a clinically meaningful difference of 6). Secondary outcomes included depression (Patient Health Questionnaire-9), overall symptom distress (General Symptom Distress Scale), specific symptoms (PEG pain, PROMIS fatigue, shortness of breath), hospitalizations, and mortality. Data were analyzed using mixed models.

RESULTS: 314 patients were randomized (157 intervention, 157 control). Participants were generally male (77%), white (63%), with a mean age of 65.5 years, and 57% had reduced ejection fraction. At 6 months, mean KCCQ score improved 5.5 points in the intervention arm and 2.9 points in the control arm (difference, 2.7; 95% confidence interval −1.3, 6.6; p = 0.19). Among secondary outcomes, depressive symptoms and fatigue improved at 6 months with CASA (effect sizes of −0.29 and −0.30, respectively, p = 0.02 for both). There were no changes in overall symptom distress, pain, shortness of breath, or hospitalizations. Mortality at 12 months was similar (CASA, 10/157; usual care, 13/157; p = 0.52).

CONCLUSIONS: This randomized trial of the CASA intervention did not demonstrate a significant improvement in heart failure-specific health status. Secondary outcomes of depression and fatigue, which have been difficult symptoms to address in heart failure, did improve. Alternate or more intensive interventions should be evaluated to improve health status in the symptomatic heart failure population.

COLLECTING DATA ON PATIENTS’ SEXUAL ORIENTATION, GENDER IDENTITY, AND SEX ASSIGNED AT BIRTH IN A CLINICAL SETTING: PERSPECTIVES OF PROVIDERS, PATIENTS, AND REGISTRATION STAFF Karey Kenst 2; Robert Coulter3; Aswita Tan McGrory2; Lenny Lopez1. 1University of California San Francisco, San Francisco, CA; 2Massachusetts General Hospital, Boston, MA; 3University of Pittsburgh, Pittsburgh, PA. (Control ID #2707079)

BACKGROUND: The Institute of Medicine recommends that healthcare organizations collect information about patients’ sexual orientation, gender identity, and sex assigned at birth (SOGI). Healthcare organizations are beginning to collect these data. This study investigated providers’, registrars’, and patients’ perspectives to: (1) understand the issues and nuances regarding collecting SOGI data; and (2) elucidate preferred processes for asking about SOGI.

METHODS: We conducted a mixed methods study with participants of diverse sexual orientations and gender identities. Using a random sampling strategy, we conducted surveys with 104 patients and administered follow-up semi-structured interviews with 16 patients. We also conducted surveys with 13 primary care providers and semi-structured interviews with 15 registrars, 13 LGBTQ (lesbian, gay, bisexual, transgender, and queer) providers, and 18 LGBTQ community members. Surveys and interviews covered the following topics: opinions about collecting SOGI data in clinical settings; storing, accessing, and using SOGI data; cognitive testing of previously recommended SOGI questions; and processes for administering questions. Applied thematic analyses were conducted with a grounded theory approach.

RESULTS: LGBTQ community members and patients overwhelmingly wanted healthcare providers to know their SOGI. Patients and providers agreed that knowing SOGI information would help providers avoid making assumptions about LGBTQ patients’ identities and health care needs. All participants thought SOGI data could also be used to monitor for disparities. Barriers included: LGBTQ participants feared discrimination or mistreatment and providers were concerned with making patients uncomfortable. Registrars had the most hesitancy and discomfort asking SOGI, fearing they would upset patients and be unable to address patients’ distress. However, they thought that ongoing training would be required to appropriately support registrars. LGBTQ participants wanted additional response options that were inclusive of a wide range of identities (e.g., non-binary, asexual). Transgender participants wanted to be asked about current gender identity first followed by sex assigned at birth. A private setting for registration was recommended and patients and providers thought SOGI data should be stored in their medical records and be accessible to clinical providers. Participants believed ongoing education of patients, providers, and registrars about SOGI data collection and usage was necessary for successful implementation.

CONCLUSIONS: Participants thought that collecting SOGI was important and that staff training and policies were essential to assure appropriate implementation and equitable use of these data. There are many nuances to collecting SOGI data but collecting it allows healthcare organizations to identify disparities and implement targeted initiatives to improve care and achieve health equity for LGBTQ populations.

COMMUNICATION GAPS BETWEEN PCPS AND HOSPITALISTS AROUND TESTS PENDING AT THE TIME OF DISCHARGE Karen J. Blumenthal; Megan M. Meehan; Ryan Thompson. Massachusetts General Hospital, Boston, MA. (Control ID #2703692)

BACKGROUND: Effective communication and care coordination between primary care physicians (PCPs) and hospitalists are essential for safe care transitions between inpatient and outpatient settings. Inadequate care coordination at discharge can result in a lack of necessary follow-up, and in some cases may lead to patient harm. Previous work has suggested that as many as 41% of patients discharged from a medicine service are discharged with one or more test results pending. Standardizing communication between PCPs and hospitalists, particularly around follow-up of tests results pending at discharge, could result in fewer missed test results, less delay in action on these results, and improved patient outcomes. The objective of this study was to determine current communication practices between PCPs and hospitalists at our institution with regard to test results pending at discharge from the inpatient medicine service.

METHODS: We administered a cross-sectional web-based survey to all PCPs (n = 165) whose patients are admitted to a hospitalist/inpatient medical team, and all hospitalists (n = 96) working at a single, large academic medical center. The survey asked about frequency and perceived efficacy of communication between hospitalist and PCP with regard to pending test results. We summarized survey responses using frequencies to compare communication practices between PCPs and hospitalists.

RESULTS: Fifty-two percent of PCPs and 46% of hospitalists responded to the survey. Our survey found significant differences between hospitalists and PCP regarding knowledge about tests pending at discharge. Seventy-five percent of hospitalists reported “always” (34%) or “usually” (41%) making the PCP aware of tests that need follow-up after discharge, while only 26% of PCPs reported “always” (4%) or “usually” (22%) being made aware by hospitalists of pending tests that need follow-up after discharge (p < 0.001). Additionally, 77% of hospitalists reported that they expect that the PCP will follow-up the results of pending tests after discharge. In contrast, only 22% of PCPs report that it is “always” or “usually” clear who is responsible for following-up test results pending at discharge.

CONCLUSIONS: Our study suggests that there is a large gap in perceptions and expectations between PCPs and hospitalists around pending tests at discharge. Our study also revealed ambiguity around who is responsible for following-up pending test results after discharge. These results highlight the need for interventions to develop and standardize the communication of pending tests at the time of discharge as a step to improving patient safety and outcomes following discharge.

COMMUNICATION SKILLS AND VALUE-BASED MEDICINE: UNDERSTANDING RESIDENTS’ VARIATION IN CARE USING UNANNOUNCED STANDARDIZED PATIENT VISIT Kathleen Hanley 1, 3; Amanda Watsula-Morley1; Lisa Altshuler1; Heather Dumorne1; Adina Kalet1; Barbara Porter2, 3; Andrew B. Wallach2, 3; Colleen Gillespie1; Sondra Zabar1. 1NYU School of Medicine, New York, NY; 2Bellevue Hospital, New York, NY; 3Gouverneur Healthcare Services, New York, NY. (Control ID #2705415)

BACKGROUND: Training residents to effectively practice value-based care is challenging. We hypothesized that residents with better communication skills would order fewer unnecessary tests and prescribe more appropriate care. We used a USP case of a patient with uncontrolled asthma to examine the relationship between value-based care and communication skills.

METHODS: A 25 year-old female USP presented as a new patient to a medicine resident’s clinic, reporting asthma since childhood with worsening symptoms over the past few months. At the time of the visit, she was using her albuterol inhaler multiple times daily, without any additional asthma treatment, and was unsure whether she was using it properly. Data was collected using two forms of assessment: a post-visit USP checklist and a systematic review of the corresponding clinic note to examine treatment recommendations including referrals and quality of documentation. The USP checklist measured communication, patient education, and assessment skills. Each response option included descriptive behavioral anchors and was rated as not done, partly done, or well done. Domain scores were calculated as percent items rated well done.

RESULTS: 141 USP visits were made from 2009 to 2016 with a mean visit length = 88 min, SD = 28 min (range: 40 to 180 min). Almost all residents (92%) evaluated the patient’s asthma with a pulmonary examination. The most common treatment prescribed was albuterol and an inhaled steroid, with or without a spacer (79%). The majority of residents (53%) did not order any additional studies; 21% ordered one study, and 26% ordered two or more studies. Study orders fell into one of three categories: gold (appropriate/recommended: PFTs, flu shot, HIV), grey (pulmonary consult, HCG), or inappropriate (TSH, A1C). Across the 141 visits, 129 studies were ordered; 46% were gold, 5% were grey, and 49% were inappropriate. The most common study ordered was a PFT (31%). 87% of single study orders were gold, but 92% of multiple orders included at least one inappropriate study. Residents who did not order any studies had significantly higher patient education and counseling skills than residents who ordered one or more studies (54% vs 34%, p = 0.00) and were more likely to explain how to correctly use an inhaler than residents who ordered one or more studies (48% vs 27%, p = 0.01). These residents also had significantly higher management and treatment skills (61% vs 39%, p = 0.00) and overall communication skills (68% vs 55%, p = 0.01). There were no significant differences between groups in medications prescribed or in quality of documentation.

CONCLUSIONS: Effective communication skills may contribute to value-based care through appropriate patient education and ordering of fewer inappropriate studies. Rigorous curricula and assessment of resident’s patient education skills should be in place to help both patients and health care system achieve value-based care.

COMMUNITY HEALTH SPECIALIST TRAINING FOR LIBRARIANS: IMPROVING KNOWLEDGE AND CONFIDENCE IN ADDRESSING THE SOCIAL DETERMINANTS OF HEALTH Anna U. Morgan 2; Bernadette A. D’Alonzo2; Roxanne Dupuis2; Alexander Reisley2; Heather Klusaritz1; Carolyn C. Cannuscio1. 1Perelman School of Medicine, U. of Pennsylvania, Philadelphia, PA; 2University of Pennsylvania, Philadelphia, PA. (Control ID #2706512)

BACKGROUND: Hosting over 1.5 billion in-person visits annually, and often serving as a lifeline for the highly vulnerable, public libraries can be partners in population health. Public library staff, however, report feeling ill-equipped to address the health and social challenges facing many of their patrons. The aim of this study was to train public library staff in pertinent health-related topics and evaluate their comfort, confidence, and preparedness in assisting vulnerable patrons.

METHODS: Our interdisciplinary team of individuals with medical, social work, and public health backgrounds conducted a needs assessment of library staff and local residents in Philadelphia to determine the most pressing challenges facing the community. We then developed a 12-hour, case-based training curriculum for library staff. Topics addressed in the training--homelessness, mental illness and substance use, immigration, and trauma--were selected based on findings from the needs assessment. Each case was designed to help staff “recognize” high-risk patrons, “engage” them in conversation, and “refer” them to appropriate community-based services. Participants were surveyed before and after each session and asked to rate on a scale of 1–10 how “comfortable,” “confident,” and “prepared” they felt when presented with each case. Responses (before and after) were compared using a paired t-test. Participants were also interviewed 4 months post-training about their daily work, how they had used the training, and for additional feedback. Interviews were transcribed and analyzed for key themes using an iterative process.

RESULTS: 11 individuals participated in the training (81% female, age 26–65, 45% African-American, 45% white, 9% Asian). Staff included managers, adult and children’s librarians, library assistants, and security guard. Participants reported a significant improvement in comfort, confidence, and preparedness in almost all cases. Nine (81%) of the library staff completed semi-structured interviews. Overall, participants were positive, stating that the curriculum boosted their confidence about their current work. They also reported feeling more capable of assisting vulnerable patrons. However, several had not yet used many of the resources presented because they had not encountered patrons whom they identified as needing extra assistance.

CONCLUSIONS: A 12-hour case-based training curriculum to teach public library staff to “recognize, engage and refer” patrons to appropriate resources significantly increased the preparedness, comfort, and confidence of staff in assisting vulnerable patrons with health and social needs, thereby increasing the capacity for public libraries to serve as partners in improving population health. Future training programs should strive to provide resources in a more accessible format, continue to assist staff in defining their role as public health partners, and assess the impact such training programs have on vulnerable patrons.

COMMUNITY HEALTH WORKER SUPPORT VERSUS COLLABORATIVE GOAL-SETTING FOR DISADVANTAGED PATIENTS WITH MULTIPLE CHRONIC DISEASES: A RANDOMIZED CLINICAL TRIAL David Grande 1; Judith A. Long1, 3; Nandita Mitra2; Hairong Huo1; Robyn A. Smith1; Shreya Kangovi1. 1Perelman School of Medicine at the University of Pennsylvania, Philadelphia, PA; 2University of Pennsylvania, Philadelphia, PA; 3Corporal Michael J. Crescenz VAMC, Philadelphia, PA. (Control ID #2707904)

BACKGROUND: A growing number of Americans have multiple chronic conditions. Collaborative goal-setting between patients and providers can improve chronic disease control, but may be insufficient in patients with unaddressed socioeconomic needs. Community health workers (CHWs), trained laypeople who share socioeconomic background with patients, can effectively improve chronic disease outcomes. Unfortunately, most prior CHW interventions have been disease-specific. The study team created IMPaCT (Individualized Management for Patient-Centered Targets), a standardized CHW intervention that addresses ‘upstream’ socioeconomic and behavioral barriers across diseases. In a prior randomized clinical trial of hospitalized patients with a variety of diagnoses, this intervention improved post-hospital access to primary care, mental health and quality of care while decreasing recurrent hospital readmission. For this study we adapted the IMPaCT intervention for use in the outpatients with multiple chronic conditions. Our objective was to compare the effects of collaborative goal-setting versus goal-setting plus CHW support on chronic disease control.

METHODS: We performed a single-blind, randomized clinical trial in two urban academic adult internal medicine clinics. We enrolled 302 participants who were residents of high-poverty neighborhoods, uninsured or publicly insured, and diagnosed with ≥2 chronic diseases (diabetes, obesity, tobacco dependence, hypertension). All participants met with their primary care provider to collaboratively set a chronic disease management goal for one of their multiple conditions. Patients randomly assigned to CHWs also received six months of tailored support.

RESULTS: The mean age of the cohort was 56.3 years (SD 13.1), 94.7% were black and 96.3% had a history of a traumatic event. Participants were diagnosed with an average of 2.5 of the eligibility chronic conditions. We found differences in the six-month change in chronic disease control between CHW support vs goal-setting alone arms (△HbA1c −0.4 vs 0.0, △BMI −0.3 vs. -0.1, △CPD −5.5 vs. -1.3, △SBP −1.8 vs. -11.2, overall p = 0.08). Patients receiving CHW support also showed greater improvements in mental health (2.3 vs. -0.2, p = 0.008) and reported higher quality primary care that was comprehensive (49.2% vs. 39.7%, p = 0.01) and supportive of disease self-management (62.9% vs 38%, p = 0.0002). Thirty-two percent of patients in the goal-setting arm were hospitalized at one-year versus 23% in the CHW support arm (p = 0.11). There were no differences in patient activation or self-rated physical health.

CONCLUSIONS: CHW support led to modest improvements in diabetes, obesity and smoking, but not in hypertension. CHW support improved mental health and quality of primary care, and may have reduced hospitalizations. Standardized CHW interventions can improve key health outcomes across multiple diseases for high-risk patients.

COMPARING LIFE-YEARS LOST VS. NUMBER OF DEATHS IN THE US OVER THE PAST 20 YEARS Glen B. Taksler; Michael B. Rothberg. Cleveland Clinic, Cleveland, OH. (Control ID #2698725)

BACKGROUND: Leading causes-of-death are typically reported by number of deaths. Assessing life-years lost to each cause may provide better context.

METHODS: We assembled national data on life-years lost to each cause-of-death and analyzed changes over the past 20 years, using all death certificates filed with the National Vital Statistics System for 2015 and 1995. Specifically, we defined “life-years lost” as remaining life expectancy for each decedent’s age, sex and race. We summed life-years lost across all individuals in 2015 and 1995, and calculated the share of life-years lost to each cause-of-death. To better understand reasons for the change in life-years lost between 1995 and 2015, we redefined life-years lost as the product of 3 terms: the mortality rate, population size and life expectancy. We employed a first-order Taylor series expansion for each age (single-year increments), sex, race and cause-of-death subgroup, allowing us to approximate the change in life-years lost by a weighted average of the change in disease-specific mortality rates, the change in population size and the change in life expectancy.

RESULTS: Heart disease caused the most deaths in both years, but cancer was responsible for 23% more life-years lost in 2015. Life-years lost to heart disease declined 6% since 1995, led by a 42% decline for acute myocardial infarctions, whereas life-years lost to cancer increased 16%. Accidents caused twice the share of life-years lost as deaths in 2015 (10.6% vs. 5.4%), and rose in importance since 1995 because of a 4.5-fold increase in life-years lost to accidental poisonings, primarily overdoses. The entire gains of the past 20 years in preventing and treating HIV were offset by the increase in accidental deaths. Suicides and homicides disproportionately burdened younger individuals. Improvements in disease-specific mortality rates contributed to a 28% reduction in life-years lost since 1995, and saved 1.5 million more life-years for heart disease than cancer. However, total life-years lost increased by 17%, due to population growth (mostly for ages 50–64 y, 16.1% contribution) and longer life expectancy (mostly for white males, 5.0% contribution). Among elders aged ≥80 y, increases in disease-specific mortality rates for Alzheimer’s disease caused more growth in life-years lost than all other increases combined. Measuring life-years lost highlighted racial disparities in heart disease (+20.8% for black males vs. -4.6% for white males), homicides and perinatal conditions.

CONCLUSIONS: Life-years lost may provide better context than number of deaths for understanding mortality trends. Focusing on life-years provides a better estimate of the societal burden of disease (e.g., cancer has already surpassed heart disease as the leading cause of life-years lost) and highlights racial disparities. Based on current trends, future progress in secondary prevention and treatment of chronic heart conditions, cancer, addiction and neurodegenerative disease appear critically important.

COMPARING PATIENTS’ EXPERIENCES WITH ECONSULT AND REFERRAL FROM PRIMARY TO SPECIALTY CARE: RESULTS FROM A NATIONAL SURVEY Sara L. Ackerman 3; Scott Shipman4; Drayton Moody1; Meaghan Quinn4; Ariana Afshar2; Nathaniel Gleason1. 1UC San Francisco, Nathaniel Gleason, CA; 2UCSF, San Francisco, CA; 3University of California, San Francisco, San Francisco, CA; 4American Association of Medical Colleges, Washington, DC. (Control ID #2706848)

BACKGROUND: eConsults enable primary care providers (PCPs) to request advice from specialists via an electronic health record. Research has demonstrated high clinician satisfaction with eConsults as well as their potential to improve specialty care access. However, to date there has been no large-scale assessment of patients’ experience with eConsult. As part of a national dissemination of an eConsult model developed at the University of California, San Francisco, we sought to compare the experiences of eConsult patients with patients referred for an office appointment with a specialist.

METHODS: An email-initiated online survey was administered at 9 academic medical centers in 2016. We identified all referral and eConsult orders to participating specialties from primary care visits with adult patients. We filtered out patients with: a) orders from PCPs who did not submit at least one eConsult in the prior month, to control for provider influences on patient experience; and b) more than one order in the prior month, to reduce confusion about which referral or eConsult was the subject of the survey. A total of 28,160 patients (eConsult/ref) were invited to participate during the nine-month study.

RESULTS: 7,264 patients responded (14% eConsult; 86% referral), a 26% response rate. Respondents’ mean age was 57, 66% were women, and on average they reported more formal education and less ethnic diversity than the general population (83% identified as white). Referral patients were more likely than eConsult patients to identify the correct order type resulting from their primary care visit (72% vs. 40%; p < 0.001). Nearly all respondents (96% eConsult; 99% referral) agreed with their PCP’s order decision. Approximately 82% of both referral and eConsult patients were satisfied with the specialist’s recommendations. Among patients who had received information from a PCP about their recent eConsult (84%), 96% agreed that the PCP’s communication of the specialist’s recommendations was prompt; 97% agreed that the PCP clearly explained the specialist’s advice; and 79% agreed they had an opportunity to ask questions. 52% of communications from PCPs to patients about eConsult results took place via secure email; 33% were by phone or in-person. 32% of referral patients and 77% of eConsult patients would prefer an eConsult rather than an in-person appointment with a specialist for a similar problem in the future.

CONCLUSIONS: Across nearly all measures, eConsult patients were as satisfied as standard referral patients with their consultation experience. The high proportion of referral patients who expressed a preference for eConsult for a similar problem in the future also suggests acceptability among patients previously unfamiliar with the service. The lack of awareness of eConsult orders among many respondents, and delays in conveying the specialist’s advice, suggest room for improvement in patient involvement in eConsult decision making and communication.

COMPARING THE VALUE OF CARE DELIVERED BY U.S. NURSE PRACTITIONERS, PHYSICIAN ASSISTANTS, AND PHYSICIANS IN THE EMERGENCY DEPARTMENT John N. Mafi 1, 2; Peter Smulowitz3; Robert Brook2, 1; Bruce E. Landon4, 5. 1David Geffen School of Medicine at UCLA, Los Angeles, CA; 2RAND Corporation, Santa Monica, CA; 3Beth Israel Deaconess Medical Center, Harvard Medical School, Boston, MA; 4Harvard Medical School, Boston, MA; 5Beth Israel Deaconess Medical Center, Boston, CA. (Control ID #2706948)

BACKGROUND: Prior research has found that nurse practitioners and physician assistants (NPs, PAs) provide similar quality and efficiency of care to physicians for routine and low-complexity care in the ambulatory setting. Few studies, however, have compared the value of care between NPs/PAs and physicians managing potentially higher acuity cases in the emergency department (ED) setting.

METHODS: Using nationally representative data from the National Hospital Ambulatory Medical Care Survey on visits to ED providers, we compared utilization of health services among NPs/PAs and physicians from 2009–2013. Outcomes included overall use of hospital admissions, imaging studies, diagnostic tests, procedures (e.g., suturing), and medications given in the ED or upon discharge. We also evaluated use of 4 low value services recently highlighted by the Choosing Wisely Campaign for 5 common conditions: use of (1) CT/MRI for back pain, headache, or syncope, (2) plain x-rays for back pain, (3) antibiotics for upper respiratory infection or skin abscess after drainage, (4) opioids for back pain or headache, as well as hospital admissions for each of the 5 conditions. We excluded red flags for each condition (e.g., hypotension on presenting vital signs for skin abscess) and estimated weighted logistic regression models adjusting for age, sex, race/ethnicity, triage severity assigned by a triage nurse, whether the patient came from a nursing home, primary diagnosis category (e.g., cardiovascular, neurological), chronic comorbidities (e.g., diabetes, HIV), insurance status (e.g., Medicare), urban location, geographic region, and year.

RESULTS: We identified 100,535 physician visits and 20,387 NP/PA visits (39% of NPs/PAs saw patients alone vs. 61% of NPs/PAs saw patients alongside a physician), representing ~515 million ED visits during the study period. Compared with physicians, NP/PA patients were younger (mean age 42.7 vs. 46.5 years, p < 0.001) and had lower triage severity (6.7% vs. 12.2% immediate or emergent cases, p < 0.001). In adjusted analyses, NPs/PAs utilized similar amounts of hospital admissions (~10%, aOR 0.97 [0.84-1.11], imaging (~50%, aOR 0.96 [0.90–1.02]), diagnostic tests (~79%, aOR 0.92 [0.84–1.01]), and procedures (~49%, aOR 1.04 [0.95–1.15]), but used more medications (83.9% vs. 81.5%, aOR 1.18 [1.04–1.33]) compared with physicians. When focusing on use of low value services, utilization patterns were largely similar except for use of antibiotics: 59.7% for NPs/PAs vs. 53.7% for physicians, aOR 1.28 [1.03–1.58].

CONCLUSIONS: In this large and nationally representative analysis, NPs/PAs prescribed more medications and low value antibiotics than physicians after controlling for patient characteristics, although practice patterns for other services remained otherwise equivalent between the two groups. These findings suggest that aside from greater medication and antibiotic use, NPs/PAs seem to provide similar value and efficiency of care compared with physicians among the measures we studied.

COMPARING THE VISIT CHARACTERISTICS AMONG PHYSICIANS AND ADVANCED CARE PRACTITIONERS Aditi Patel; Parth Parikh; Xiaobo Liu; Michael B. Rothberg. Cleveland Clinic, Cleveland, OH. (Control ID #2707362)

BACKGROUND: To reduce physician burden and decrease costs, health systems are increasingly employing advanced care practitioners, including nurse practitioners (NP) and physician assistants (PA). Studies have shown that patient satisfaction and quality of care are similar between primary care physicians (PCP) and NPs. These comparisons generally focus on specific problems and have not assessed the complexity of the patient or the visit. Our objective was to understand patient and visit characteristics of same day visits among the three provider types.

METHODS: A retrospective analysis of adult patients with same day office visits to either a physician, PA or NP was performed. Of the 19,864 patients screened, 400 patients seen by each provider were randomly selected. Patient demographics (age, race, sex and insurance type) were extracted electronically. Two team members manually extracted the following visit-related variables: visit complexity (simple problems addressed by following algorithms or complex problems requiring diagnostic acumen), duration of complaint (acute vs. chronic), number of additional problems addressed, number of prescription and non-prescription medications prior to the visit and groups requiring extra attention (i.e. patients on insulin, opiates, warfarin and oral hypoglycemic). The ANOVA F-test or Kruskal-Wallis test was used to evaluate relationships between continuous measures and setting (described as Median, IQR). Pearson’s chi-square test or Fisher’s exact test was used to assess associations between categorical measures.

RESULTS: After excluding 9 repeat visits within the study period, our sample contained 1191 unique visits. Physicians were less likely to see patients compared to PAs and NPs who were female (60.6% vs 70.7% vs 66.9% p = 0.01), African American (3.8% vs 14.3% vs 14.1% p < 0.001) or who had commercial insurance (59.3 vs 72.2 vs 62.4 p < 0.001). There was no difference in frequency of acute problems (84.1% vs 88.7% vs 85.4% p = 0.69). Physicians were more likely to address additional problems compared to PA and NP’s (35.6% vs 25.8% vs 25.3% p = 0.001), and problems that require diagnostic acumen (44.9% vs 33.8% vs 30.3% p < 0.001), At the time of the visit, patients of NPs took more prescription medications (Median 4, IQR 2–6) than patients of physicians (3,1-5) or PAs (2,1–4) (p < 0.001).

CONCLUSIONS: The patients seen by physicians, PAs and NPs differ in a number of ways that suggest they are not interchangeable..Patients of NPs take more prescription medications but physicians see more complex medical complaints. Physicians also address additional problems more often.

COMPARISON OF MEDICAL STUDENT SELF ASSESSMENT AND FACULTY ASSESSMENT OF PERSONAL AND PROFESSIONAL DEVELOPMENT SKILLS Christopher Mattson; Jeanne M. Farnan; James N. Woodruff; Wei Wei Lee. University of Chicago, Chicago, IL. (Control ID #2706759)

BACKGROUND: In 2013, the Association of American Medical Colleges (AAMC) included “Personal and Professional Development (PPD)” as a competency domain to be taught in medical school. While most schools have elements of PPD programming in place, assessment of student progress towards competency attainment has not been well studied. In addition, little research has focused on assessing differences between student self assessment and faculty assessment of PPD skills.

METHODS: After conducting a literature review on core PPD competencies, we developed a 48-question survey consisting of Likert scale questions and open-ended questions assessing PPD goals that was given to MS2 students in February 2016. A parallel 8-item survey was developed for faculty to evaluate relevant PPD domains during the MS2 clinical skills (CS) course and Scholarship and Discovery (S&D) experience in the spring and summer of 2016. For our data analysis, we compared the student self-assessments to matched faculty assessments from their CS preceptors and S&D mentors. Using paired t-tests, we analyzed the data for agreement between student self-ratings and faculty ratings. We also analyzed the data for agreement between the two different faculty raters.

RESULTS: There were 28 students for whom there was a self-assessment, a CS assessment and a S&D assessment. Self-assessment scores were significantly lower than CS assessments for each of the 8 PPD competencies assessed. Self-assessments were significantly lower than S&D assessments for 7 of the 8 competencies. There was no statistically significant difference found for “Complying with rules and regulations”. There was also no statistically significant difference between CS assessments and S&D assessments for any of the 8 PPD competencies. Of the 28 S&D mentors included in the data set, 17 (60.7%) rated the student as being “Highest performance (top 10%)” in comparison to other medical students with whom the mentor had worked previously. Eight (28.6%) rated the student as being “Above average performance (11-25%)”, 1 (3.6%) rated the student as being “Average performance (26-50%)” and 2 (7.1%) said they had never before mentored a medical student on a scholarly project. Thirty-five out of fifty six (62.5%) total faculty assessments gave a rating of 5 out of 5 for all competencies.

CONCLUSIONS: Over half of the S&D mentors rated their students as being in the top 10% of students which suggests that grade inflation contributes to the discrepancy in student and faculty assessments. This is supported by the fact that the majority of CS and S&D faculty uniformly rated students 5/5 for all domains. The data is consistent with trends in the literature, which have shown grade inflation in other aspects of medical education. The data also makes it difficult to provide meaningful PPD feedback to students. Future work should focus on faculty training to improve PPD assessment skills and also on providing opportunities for students and faculty to review and discuss discordant scores.


BACKGROUND: While literature supports usage of computed tomography (CT) to assess pericardial thickness and presence of pericardial effusion (PE), how CT compares to echocardiography (echo) in determining PE size or findings of tamponade physiology (TP) is unknown. Our goal is to compare echo and CT in sizing a PE and determining TP.

METHODS: A retrospective chart review was performed using 122 patients who had an echo that showed a PE and who had a chest CT within three days of the echo. The presence of TP and qualitative sizes of PE on CT and echo were obtained and compared (Table 1).

RESULTS: Patients with TP on echo had a range of PE sizes on CT from trace to large, but 90% of those with TP on echo had a PE of at least moderate size on CT. All patients with large PE by CT had TP on echo. TP by CT did not predict with significance TP by echo [Odds ratio 2.09; 95% confidence interval (CI): 0.58 to 7.09, p = 0.24]. The correlation of sizing between echo and CT was variable, being strongest at the size boundaries [Spearman’s ρ = 0.784, p < 0.0001]. If a PE was trace or absent on CT, 84% of PEs were sized similarly on echo. If a PE was moderately large to large on CT, 86% of PEs were sized similarly on echo. If a PE was sized between small and moderate on CT, only 66% of PEs were sized similarly on echo.

CONCLUSIONS: Our data suggest that identifying TP by CT does not correlate with those findings on echo and that sizing between modalities does not correlate well for mid-range sized PEs.

figure q

Table 1: Frequencies of qualitative PE size by CT and echocardiogram


BACKGROUND: Service is intrinsic to patient care. However, recent ACGME resident survey questions about service and education imply a conflicting relationship between the two and seem to equate “service” with “scut”. This study uses clinical vignettes to investigate this question and aims to (1) investigate resident and faculty perceptions of service and education (2) ascertain what aspects of patient care positively and negatively affect perceptions, and (3) evaluate perceptions of how the program balances service and education.”

METHODS: This cross-sectional study design used clinical vignettes to investigate resident and faculty perceptions of service and education at a medium-sized university-affiliated community internal medicine program. We used an anonymous electronic survey instrument to collect qualitative and quantitative data responses. Participants for the study were recruited from the following groups: Internal Medicine Trainees (IMT) (81 residents, chief residents and fellows) and Teaching Faculty (TF) (50 core faculty and program leaders). Analyses included quantitative and qualitative assessments. Main themes were identified from vignette comments and multiple choice questions to showcase how residents and faculty conceptualize “service” and “education.” Likert scale data were collapsed into a 4-point scale (strongly disagree/disagree, neutral, agree/strongly disagree, not applicable).

RESULTS: Eighty participants completed surveys for an overall response rate of 61% (N = 45; 56% IMT; N = 35; 70% TF). Overall, 21% of respondents perceived that excessive service obligations negatively impacted the learning environment, with trainees being significantly more likely than faculty to report agreement with this statement (33% vs 6%, respectively; p = .001). Trainees were also significantly more likely than faculty to agree that excessive service obligations contributed to duty hour violations (21% vs 6%, respectively; p = .03) in their program. With some notable exceptions, faculty and trainees generally agreed on the relative service and educational values of vignette activities. Several activities ranked high in terms of both service and educational value, such as running a family meeting and seeing an urgent care clinic patient. Both quantitative and qualitative responses suggested that high “service” activities fell into two main categories: tasks that could easily be completed by non-physicians and patient care with little autonomy. For example, the top 4 reported choices in a list of activities that constituted “inappropriate or excessive service obligations” were: “calling insurance companies for prior authorizations (N = 56)”, “caring for pts when you have little autonomy (33)”, “completing forms for patients (30)”, and “making clinic appointments for patients at discharge (29)”.

CONCLUSIONS: Service and education are not mutually exclusive. Residents perceive maximal education and service value in situations involving both complexity and connection.

COMPLEX CARE ROUNDS: A COORDINATED INTERDISCIPLINARY STRATEGY FOR CARING FOR OUR SICKEST PATIENTS Katherine A. Hochman 2; Anne Meara2; Regina Presa2; Thomas Sedgwick2; Ramon Jacobs1. 1New York University School of Medicine, New York, NY; 2NYU Langone Medical Center, New York, NY. (Control ID #2700019)

BACKGROUND: Extended stay (>30 day stay) cases strain hospital resources, psychologically burden staff, increase average length of stay (ALOS) and ultimately are not patient-centric. In 2008, extended stay patients represented 15% of the average daily census on the Medicine service. The objective of our intervention was to devise a sustainable strategy to reduce the number of extended stay patients on the medicine service and overall ALOS.

METHODS: In 2008, we convened a proactive interdisciplinary team to engage in addressing barriers to care progression and discharge. The goal was to reduce the number of extended stay patients on the medicine service and overall ALOS. The team was comprised of a hospitalist physician and leadership from Care Management, Social Work, Ethics, Finance and post-acute care. Complex Care rounds occurred twice weekly. All patients with LOS of 5 days or more were presented in a standardized format, “The Four Questions”, to ensure a shared mental model. These questions are 1.) Why is the patient hospitalized? 2.) Why is the patient STILL hospitalized? 3.) What has to happen for this patient to be discharged? and 4.) When and where is this patient being discharged to safely? A special emphasis was placed on barriers to discharge (ranging from medical, social, financial, psychological), advanced care planning and contingency planning. The hospitalist physician served as a liaison to the medical staff regarding clarification of endpoints to admission.

RESULTS: Since the initiation of Complex Care Rounds, the percent of extended stay patients on the average daily census dropped from 15 to 1.5% (Figure 1). Our ALOS dropped from 6.81 days (2008) to 5.69 days (2016). It is important to note this decrease in ALOS occurred despite two important factors: 1.) our case mix index has increased dramatically from 1.47 to 1.81 during that same period and 2.) the initiation and vast expansion of our observation program siphoned off the short stay cases.

CONCLUSIONS: We have demonstrated a strategic and sustainable approach for reducing and managing patients with extended stays. Since we have achieved our initial goals, it is time to reinvent these rounds. Our new focus is mobilizing resources within the community based on disease entity and psychosocial needs.

figure r

COMPREHENSIVE OSCES AS OPPORTUNITIES FOR FACULTY TO MAKE ENTRUSTMENT JUDGMENTS: HOW ARE STANDARDIZED PATIENT ASSESSMENTS OF SKILLS PERFORMANCE ASSOCIATED WITH FACULTY ENTRUSTABILITY JUDGMENTS? Colleen C. Gillespie 1; Kathleen Hanley1, 2; Jasmine A. Ross1; Jennifer Adams1, 2; Sondra Zabar1, 2. 1NYU School of Medicine, New York, NY; 2Gouverneur Healthcare Services, New York, NY. (Control ID #2705883)

BACKGROUND: Entrustable Professional Activities (EPAs) and milestones are expert judgments made based on many formative assessments. Their validity is dependent on the number of assessments but attention is increasingly being paid to having a “fair” sample of observations equally distributed across residents and contexts. OSCEs provide such a consistent, fair sample of behavior assessed under controlled conditions but have mostly used been used to provide granular skills feedback. We explore how faculty judge the “entrustability” of residents based on observing OSCE cases and then how these entrustments relate to OSCE skills performance.

METHODS: In an 11-case OSCE for primary care residents (n = 25; PGY1-3), SPs rated skills in communication (information gathering, relationship development, education/counseling), assessment, patient education (case-specific), physical exam, professionalism, treatment plan, patient satisfaction and patient activation. Summary scores were calculated as % items rated well done (vs not or partly done; internal consistency > .72). Faculty observers then judged how much supervision the resident would need in actual practice to handle the case: 1-requires direct supervision, 2-requires indirect supervision, 3-ready for unsupervised practice, or 4-can supervise others. Mean entrustment rating across cases was correlated with clinical skills.

RESULTS: Mean entrustment =2.46 (SD .37), falling between requires indirect supervision and ready for unsupervised practice. On average, residents were judged to need direct supervision in .40 cases (SD .65), indirect supervision in 4.76 (SD 2.03), ready for unsupervised practice in 2.92 (SD 1.80), and able to supervise others in 1.07 cases (SD 1.15) with PGY1 residents needing direct and indirect supervision in more cases than PGY2 and 3 (p = .037). Associations between OSCE performance and faculty entrustment ranged from essentially zero (communication sub-domains of information gathering and education/counseling; case-specific patient education; patient satisfaction) to negative (communication sub-domain of relationship development r = −.25, p = .16; professionalism r = −.21, p = .22) to positive (case-specific assessment, r = .35, p = .07; physical exam r = .30, p = .13; treatment plan r = .40, p = .04; patient activation r = .51, p = .008). Associations between skills performance and entrustment ratings varied by case.

CONCLUSIONS: OSCEs provide a valuable opportunity for faculty to make entrustment judgments based on observing the same, complete encounter across many trainees. Entrustment judgments appear to be capturing elements of competence related to but different from SP assessments of performance, including especially “bottom line” aspects of practice such as assessment, physical exam, treatment plans and patient activation. Interestingly, we consider patient activation skills to be an “educationally sensitive patient outcome” i because both teachable and associated with patient outcomes and our results support the importance of this skill set.

CONDITION-RELATED SIDE EFFECTS AND BELIEFS ABOUT CANCER AND DIABETES Chloe Soukas 1; Kimberly A. Muellers1; Juan Wisnivesky2; Jenny J. Lin1. 1Icahn School of Medicine at Mount Sinai, New York, NY; 2Mount Sinai School of Medicine, New York, NY. (Control ID #2707558)

BACKGROUND: Urinary dysfunction is a side effect of both prostate cancer treatment and as some diabetes medications. In patients with both newly diagnosed and chronic conditions, it is important to consider how symptoms impact patient beliefs about illness. We undertook this study to assess the effect of cancer diagnosis on experience of urinary symptoms in patients with diabetes. We also evaluated how differential experience of side effects impacts patients’ beliefs about diabetes among diabetes patients with or without cancer.

METHODS: We recruited patients with diabetes and early-stage prostate cancer and enrolled age-, gender- and hemoglobin A1c-matched patients with no cancer history. All patients were prescribed an oral hypoglycemic agent, and all cancer patients began treatment in the past 6 months. Urinary symptoms were assessed using the EPIC short form and dichotomized by no symptom versus any symptom during the past four weeks. Patient beliefs were assessed by the Beliefs and Illness Perceptions Questionnaire. We used chi-squared tests to assess differences in cancer and non-cancer patients’ symptoms, reported beliefs about diabetes, and symptom experience and itemized beliefs stratified by cancer status.

RESULTS: We surveyed 31 prostate cancer and 89 diabetic non-cancer patients. We found no overall difference in cancer and non-cancer patients’ beliefs (all items p > 0.05). Cancer patients were more likely to report urine leakage, poor urinary control, poor overall urinary function, painful urination, and frequent urination in the past 4 weeks. There was no difference in experience of weak urine stream. Non-cancer patients had associations between urine leak and more emotional impact (55% vs. 25%, p = 0.045; p > 0.1 for cancer); between poor urinary function and less emotional impact (23% vs. 50%, p = 0.025; p > 0.05 for cancer); and weak urine stream and concern (100% vs. 71%, p = 0.032; p > 0.1 for cancer). Among cancer patients, there were associations between poor urinary control and concern (87% vs. 38%, p = 0.013; p = 1.0 for non-cancer); between painful urination and emotional impact (67% vs. 20%, p = 0.043; p = 1.0 for non-cancer); and between frequent urination and impact on daily life (35% vs. 0%, p = 0.033 for cancer; p > 0.05 for non-cancer). Both cancer (45% vs. 0%, p = 0.026) and non-cancer (45% vs. 21%, p = 0.012) patients reported more emotional impact with frequent urination.

CONCLUSIONS: Patients with diabetes and prostate cancer were more likely to experience urinary problems in the past 4 weeks than other diabetes patients. There was no difference between cancer and non-cancer patients’ beliefs about diabetes. Although urinary symptoms appeared to differentially affect cancer and non-cancer patients’ beliefs about diabetes, the results of this analysis reveal a complex relationship between cancer, symptom experience and disease belief. Further exploration of symptom attribution and beliefs about illnesses in patients with multiple conditions may be valuable to improving disease management.

CONFLICTING MAMMOGRAPHY SCREENING GUIDELINES: WOMEN’S PERCEPTIONS OF SCREENING EFFICACY Michael G. Knight 2, 3; Carmen E. Guerra1; Marilyn M. Schapira2. 1Perelman School of Medicine, University of Pennsylvania, Philadelphia, PA; 2University of Pennsylvania, Philadelphia, PA; 3Crescenz VA Medical Center, Philadelphia, PA. (Control ID #2707227)

BACKGROUND: Breast cancer continues to be the second leading cause of cancer death, and has the highest rate of new cancer cases, among US women. National initiatives have worked to combat these persistent rates through community health education, and increasing access to mammography. However, such initiatives now face the challenge of conflicting recommendations surrounding mammography. In this study, we sought to describe the awareness of conflicting mammography guidelines, and its association with perceptions about mammography among a group of women who had previously undergone screening mammography.

METHODS: A survey was designed with questions, adapted from the Theory of Planned Behavior, to assess the awareness of conflicting mammography guidelines and attitudes about mammography. A convenience sample of women over the age of 21, who had previously undergone mammography, was used. Participants were recruited from community events in the New York area. Awareness of conflicting mammography guidelines was defined by a response of “strongly disagree”, “disagree”, or “neither agree or disagree” for one or more statements that experts agreed on the age of initiation, screening interval, or age of discontinuation of routine mammography. Beliefs that having a mammogram would prolong life or prevent breast cancer were measured on a 5-point Likert scale from 1 (Strongly Disagree) to 5 (Strongly Agree). Descriptive statistics were used to define the study population, and generate response distributions for survey items. Bivariate analysis using chi-square or t-tests was used to examine the relationship between awareness of conflicting guidelines and attitudes about mammography.

RESULTS: There were 70 study participants; 87% were Black, 15% had a family history of breast cancer, and 2% had a personal history of breast cancer. Sixty-one percent (61%) were aware of conflicting mammography screening guidelines. When compared to participants who were not aware of conflicting guidelines, these women were less likely to agree with the following: Having a mammogram would “ me to live longer” mean (SD), 3.4 (1.2) vs. 4.1 (1.0) (p = 0.013) and “ me in the prevention of breast cancer” 3.3 (1.4) vs 4.1(1.1) (p = 0.012). Participants who were aware of the conflicting guidelines were also more likely to believe that private insurance companies made recommendations for mammography screening 25.6% vs 7.4% (p = 0.057). There was no difference in other perceived barriers, behavioral control, or intentions to have a mammogram.

CONCLUSIONS: Awareness of conflicting mammography guidelines in women who have undergone mammography is associated with perceptions of decreased efficacy of mammography with respect to breast cancer prevention and mortality reduction. These results warrant further study and consideration by healthcare providers and guideline generating organizations on the possible association that the ongoing controversy on mammography guidelines may have with patient perceptions about mammography.

CONSUMERS AND USE OF A LARGE NATION-WIDE PRIMARY CARE TELEMEDICINE SERVICE Kathryn A. Martinez; Mark Rood; Nikhyl Jhangiani; Adrienne Boissy; Michael B. Rothberg. Cleveland Clinic, Cleveland, OH. (Control ID #2702041)

BACKGROUND: Advancements in mobile technology have ushered in a new area of consumer-directed health applications. Interest is high, however, little is currently known about the provision of primary care via telemedicine. The objective of this study was to characterize consumers and their use of a large nation-wide telemedicine service.

METHODS: We analyzed all completed primary care encounters between January 2013 and August 2016 from the Online Care Group primary care telemedicine service, one of the largest telehealth companies in the U.S. Patient characteristics, including age, sex, and geographic region were provided by the user. Encounter characteristics, including time of day, wait time, visit length, connection type (mobile device, computer, or telephone), and whether patients used a coupon were recorded by the telemedicine system. Patient diagnosis and prescription receipt were recorded by the visit provider. At encounter conclusion, users rated their satisfaction with their visit provider and with the telemedicine system, on scales of 0 to 5 stars. They were also asked where they would have sought care otherwise (doctor’s office, emergency department, urgent care/retail clinic, or done nothing). We generated descriptive statistics regarding users and use of the telemedicine service.

RESULTS: There were 56,863 completed telemedicine encounters during the study period by 601 providers; 60% of patients were female, mean age was 35.2 years (Interquartile Range (IQR): 27–45 years) and 9% were <18 years. Most were from the South (34%), followed by Midwest (29%), West (23%), and Northeast (14%). The majority (62%) accessed the system via mobile device and a third (32%) by computer. Twenty-two percent used a coupon. Mean wait time was 5.1 min (IQR: 1.2-6.1 min) and mean visit length was 7.1 min (IQR: 3.4-8.8 min). Most accessed the service during working hours: 27% called between 7 am and noon and 34% called between noon and 5 pm. Only 6% of users called between midnight and 6 am. Eighty-five percent rated their visit provider 5 stars, and 78% rated the telemedicine service 5 stars. Thirty percent were diagnosed with upper respiratory infections, followed by urinary conditions (8%), dermatological conditions (5%), and mental health/lifestyle issues (4%). Overall, 57% of users received a prescription. Had they not used the telemedicine service, 44% reported they would have visited an urgent care/retail clinic, 28% would have sought care at a doctor’s office, 21% would have done nothing, and 6% would have gone to the emergency department.

CONCLUSIONS: Users were young and most accessed the system via mobile device. Both average wait time and visit length were short. Use of the telemedicine service appears to serve primarily as a substitute for urgent care/retail clinics, and to a lesser extent physician visits. However a significant number of users would have forgone care entirely were the service not available.

CONTRACEPTIVE COUNSELING QUALITY AND EFFECTIVE CONTRACEPTIVE USE AMONG WOMEN VETERANS: DATA FROM THE ECUUN STUDY Lisa S. Callegari 3, 6; E. Bimla Schwarz1; Xinhua Zhao5; Maria K. Mor4; Sonya Borrero2. 1University of California, Davis, Sacramento, CA; 2University of Pittsburgh and VA Pittsburgh, Pittsburgh, PA; 3VA Health Services Research & Development, Seattle, WA; 4VA Pittsburgh Healthcare System, Pittsburgh, PA; 5VA Puget Sound Health Care System, Pittsburgh, PA; 6University of Washington, Seattle, WA. (Control ID #2706328)

BACKGROUND: Nearly half of pregnancies in the US are unintended, and the majority of these occur in women not using or inconsistently using contraception. Understanding associations between contraceptive counseling quality and effective contraceptive method use could inform efforts reduce unintended pregnancy.

METHODS: We analyzed data from the “Examining Contraceptive Use and Unmet Need among Women Veterans” (ECUUN) study, a national telephone survey of women veterans ages 18–45 years who received primary care in VA in the past year. This analysis was limited to women at risk of unintended pregnancy (heterosexual intercourse within the past year; not pregnant or trying to conceive; no history of hysterectomy, sterilization, or infertility) who reported receiving contraceptive counseling in the past year. Counseling quality was measured using six Likert-scale questions assessing whether patients’ concerns were addressed, risks/benefits of options were discussed, and patients’ views were elicited regarding which method they thought was best for them. Composite scores of how many items to which participants responded “agree” or “strongly agree” were calculated (0–6) and categorized as high (score = 6), moderate (score 3–5), and low (score < 3) quality. Our outcome was contraceptive method used at last sex, categorized as highly effective (intrauterine device, implant), moderately effective (injection, pill, patch, ring), and less effective (barrier, withdrawal)/no method. Logistic regression assessed associations between counseling quality and use of effective methods (moderately or highly effective) and, among those using effective methods, use of highly effective methods, controlling for age and race.

RESULTS: Of the 365 women veterans in our sample, 42 (12%) reported low quality counseling, 144 (40%) moderate quality, and 179 (49%) high quality. In unadjusted analyses, women reporting low quality counseling were less likely to use highly effective methods compared to women reporting high or moderate quality counseling (2% vs 20 and 18%, respectively) and more likely to use less effective methods/no method (36% vs 21 and 19%, respectively), with similar proportions (60-63%) using moderately effective methods across counseling quality groups [overall p = 0.02]. In adjusted analyses, women reporting low quality counseling were less likely to use effective methods than women reporting high quality counseling (adjusted OR[aOR]:0.42, 95%CI 0.20, 0.89). Among the 286 women using effective methods, women reporting low quality counseling were less likely than women reporting high quality counseling to use a highly effective method (aOR:0.12, 95%CI 0.003, 0.81).

CONCLUSIONS: Quality of contraceptive counseling is associated with effective contraceptive method use among women Veterans. Facilitating higher quality contraceptive counseling may represent an opportunity to improve contraceptive outcomes and reduce unintended pregnancy.

CORONARY ARTERY DISEASE SEVERITY MODIFIES THE ASSOCIATION BETWEEN GLYCEMIC CONTROL AND MORTALITY AND INFORMS PERSONALIZED DIABETES MANAGEMENT TARGETS Sridharan Raghavan 3, 4; Wenhui G. Liu3; Michael Ho1; Margaret E. Plomondon3; David Magid3; David Saxon2, 3; Steven M. Bradley5, 3; Thomas M. Maddox1. 1Eastern Colorado Health Care System, Denver, CO; 2University of Colorado, Aurora, CO; 3VA Eastern Colorado Health Care System, Denver, CO; 4University of Colorado School of Medicine, Denver, CO; 5Minneapolis Heart Institute, Minneapolis, MN. (Control ID #2702512)

BACKGROUND: While professional society guidelines recommend a personalized approach to diabetes management that takes into account an individual’s comorbidities and diabetes-related complications, there is little evidence to guide individualization of diabetes treatment targets. We examined whether the severity of coronary artery disease (CAD) modifies the association between glycemic control and mortality and thus provides guidance in personalizing diabetes treatment.

METHODS: We studied 17,394 veterans with type 2 diabetes who underwent elective cardiac catheterization in the Veterans Affairs (VA) health care system between 2005 and 2013. The primary exposure was HbA1c level over 2 years of follow-up after index catheterization. The primary outcome was 2-year all-cause mortality, and cardiovascular versus non-cardiovascular mortality was a secondary outcome. CAD severity was categorized as no CAD, non-obstructive CAD, or obstructive CAD based on angiographic findings. Using multivariable Cox proportional hazards regression, adjusting for demographic and clinical covariates, we estimated the association between time-varying HbA1c and mortality, with an interaction term to determine effect modification by CAD severity.

RESULTS: Mean HbA1c at baseline was 7.5%; 61% had obstructive, 22% had non-obstructive, and 17% of participants had no CAD. Relative to a reference HbA1c of 6.5-7%, HbA1c <6% was associated with increased mortality risk (HR 1.55 [1.25, 1.92]), whereas HbA1c categories above 7% were not. We observed a significant interaction between HbA1c and CAD severity (interaction p = 0.0005). HbA1c <6% was associated with increased risk of mortality among individuals with obstructive and non-obstructive CAD, but not in those with no CAD (Figure). In secondary analysis, HbA1c <6% was associated specifically with increased risk of non-cardiovascular but not cardiovascular mortality.

CONCLUSIONS: CAD severity may inform diabetes treatment goals. In particular, the short-term risk of harm associated with lower glycemic control targets was limited to individuals with CAD on angiography. Measures of cardiovascular disease burden may help physicians optimize individualized glycemic control targets, especially with respect to minimizing harm.

figure s

CORRELATES OF OPIATE MISUSE OF PATIENTS ON CHRONIC OPIATE THERAPY FOR CHRONIC NON-CANCER PAIN IN AN ACADEMIC, SAFETY-NET PRIMARY CARE CLINIC Smita Y. Bakhai 1; Bright Thilagar1; Jessica Reynolds1; Kenneth Leonard2. 1SUNY at Buffalo, Williamsville, NY; 2SUNY at Buffalo, Buffalo, NY. (Control ID #2690370)

BACKGROUND: Misuse and deaths associated with prescriptions opiates has increased dramatically. Efforts to curb the misuse of prescription opiates have focused on recognizing inappropriate use by patients in treatment for chronic non-cancer pain (CNCP). The primary objective of this study was to evaluate the correlates of opiate misuse based on a failed urine drug test (UDT) among patients on chronic opiate therapy (COT) for CNCP in a primary care setting.

METHODS: Utilizing medical records, we conducted a cross-sectional study of 206 subjects, between April 2011-April 2012, who were prescribed COT for at least 3 months’ duration for CNCP. UDT results were used to classify subjects into three groups: 1) No Misuse, positive UDT for prescribed opiate and negative for illicit drugs and non-prescribed control substances; 2) Misuse, positive UDT for illicit drugs or non-prescribed control substances and positive or negative for prescribed opiates or overdose; 3) Potential diversion, negative for prescribed opiates and negative for illicit/un-prescribed controlled substances. Variables of interest were UDT results, patient demographics and medical, psychiatric and substance abuse history, missed medical appointments and non-adherence to non-opiate medications for other chronic diseases. Non-adherence was defined as physician documentation of medication non-adherence in medical records during clinic visit. We conducted an initial analysis comparing all groups using chi square analyses for categorical variables. Continuous variables were analyzed with ANOVA with Bonferroni corrected post hoc tests. We conducted two logistic regression analyses, one comparing groups using only those variables that were predictive of group status in the initial analyses.

RESULTS: The most commonly prescribed opiate was hydrocodone. The most commonly misused substances were marijuana, cocaine, benzodiazepines and non-prescribed opioids. Of the 206 records analyzed, 80 (38%) had a no misuse, 91 (44%) had misuse and 35 (17%) had potential diversion. In bivariate analyses, positive history of smoking (OR 3.90, 95% CI 1.69–9.03), substance use (OR 7.02, 95% CI 2.56–19.20), missed medical appointments (OR 2.85, 95% CI 1.44–5.63), and non-adherence to other medications correlated with opiate misuse group (OR 18.86, 95% CI 8.73–40.74). In logistic regression, only substance use history (OR 4.32, 95% CI 1.27–14.64) and non-adherence with non-opiate medications (OR 13.22, 95% CI 5.81–30.10) correlated with opiate misuse.

CONCLUSIONS: Non-adherence with non-opiate medications and missed appointments were significant correlates of opiate misuse based on failed UDT in a primary care setting. This study highlights the significance of reviewing medication and appointment adherence, in addition to substance abuse and smoking history.

COST OF RETRIEVING PRESCRIPTION DRUG MONITORING PROGRAM REPORTS: COMPARISON OF PHYSICIAN VERSUS DELEGATE MODELS Marcus Bachhuber 2; Marc Larochelle1; Jessica Merlin3; Sean M. Murphy4. 1Boston University School of Medicine and Boston Medical Center, Boston, MA; 2Montefiore Medical Center/Albert Einstein College of Medicine, Bronx, NY; 3University of Alabama Birmingham, Birmingham, AL; 4Washington State University, Pullman, WA. (Control ID #2705307)

BACKGROUND: Prescription drug monitoring programs (PDMPs) are state-level databases of dispensed controlled substances. Clinical guidelines recommend, and an increasing number of states mandate, periodically obtaining PDMP patient reports prior to writing controlled substance prescriptions. Although the task of retrieving reports is a clerical function, it often falls to the prescribing physician. While delegates such as administrative staff members have access to PDMPs in many states, no study has compared the costs of delegates assisting physicians to obtain PDMP reports to the costs of relying solely on physicians to obtain them.

METHODS: We conducted a cost analysis with a 1 year time horizon, from the perspective of physicians (if self-employed) or their employers. We obtained estimates of the frequency of controlled substance prescribing by primary care providers from the 2011–2013 National Ambulatory Medical Care Survey. We measured time to retrieve PDMP reports by calculating the mean of 64 query attempts by 13 physicians in 5 states (30s to log in and 40s to retrieve a PDMP report). We defined two PDMP usage cases based on the frequency of queries: comprehensive (before every Schedule II-IV prescription), selective (before new Schedule II- IV prescriptions and only every 6 months for continuing prescriptions), and minimal (limited to before new Schedule II or III prescriptions and annually for continuing prescriptions). For each PDMP usage case, we calculated the labor costs (monetary value of physician or delegate time) and media costs (paper and ink).

RESULTS: Nationally, primary care providers write a new Schedule II-IV controlled substance prescription at 5.0% of visits and write a continuing Schedule II-IV controlled substance prescription at 17.8% of visits. For comprehensive, selective, and minimal PDMP usage, primary care providers spend an annual mean of 12, 11, and 6 hours, respectively, retrieving PDMP reports. The mean annual cost of this activity is an estimated $1509, $1387, and $762, respectively. The mean annual cost of the delegate model is $1105, $1023, and $710, respectively. Compared to the physician model, the delegate model results in a cost savings of $404 (27%), $364 (26%), and $52 (7%).

CONCLUSIONS: Retrieving PDMP patient reports—a clerical function—represents an opportunity cost to the payer of physician services in that it may result in forgone patient care. Alternatively, physicians may perform this activity after work hours, which could contribute to burnout. A model where delegates assist physicians can only reduce costs by up to one quarter. Ultimately, automation and integration of PDMP data into electronic health records may further reduce physician labor costs. Physicians, health care systems, and states should collaborate to ensure that PDMPs are employed in a manner that maximizes their benefit to patients and society, while minimizing their associated costs.


BACKGROUND: The Advisory Committee on Immunization Practices recommends a single dose of the live attenuated herpes zoster vaccine (HZV) in persons aged ≥60 years. Because the efficacy decreases to zero after 10 years, many vaccinated patients will soon be subject to an increased risk of herpes zoster and post-herpetic neuralgia. A booster might be necessary to extend protection but its cost-effectiveness is unknown. We aimed to determine the cost-effectiveness of a HZV booster and its optimal timing in immunocompetent adults first vaccinated at aged ≥60 years.

METHODS: We updated a validated Markov model to follow patients life-time. From the societal perspective, we compared costs and quality-adjusted life years (QALYs) for no booster versus booster options. We examined a booster anytime between 1–20 years after the first dose, and conducted the analysis for persons having the first dose at different ages: 60, 70, and 80 years. The model started with the ‘Healthy’ state and patients could move between health states over time. Compared to no booster group, patients with a booster had an additional reduction in disease incidence and complications proportional to vaccine efficacy. Because patients entered the model already vaccinated, costs and side effects of the first dose were not included. We assumed the booster had the same efficacy and waning rate as the initial vaccination. Transition probabilities and other model inputs were based on published literature. We derived vaccine efficacy from the Shingles Prevention Study and its long-term data. All costs were expressed in 2015 US dollars ($). Costs and QALYs were discounted at 3%/year. Results were presented as incremental cost-effectiveness ratio (ICER) ($/QALY) and a threshold of $100,000/QALY was used to determined cost-effectiveness.

RESULTS: Cost-effectiveness of the booster varied by age and time since vaccination. The booster cost less than $100,000/QALY for all patients at least 5 years after the initial dose, but was most cost-effective at around 10 years. In one-way sensitivity analysis, none of the variables caused the ICER to exceed $100,000/QALY. In probabilistic sensitivity analysis, a booster at 10 years had >80% the probability of being cost-effective.

CONCLUSIONS: Under current assumptions, a booster dose of HZV could be cost-effective for vaccinated patients 10 years after initial vaccination. Future data on booster’s efficacy could change the conclusion.

figure t

CREATING CONTINUITY ON A 7 ON 7 OFF HOSPITALIST STAFFING MODEL Christopher M. O’Donnell; Marsha Stern; Traci Leong; Ethan Molitch-Hou; Bruce Mitchell. Emory University, Atlanta, GA. (Control ID #2702955)

BACKGROUND: Transitions of care are a potential pitfall for miscommunication and an area of increased risk for patients. This study implemented a novel staffing model in a 7 on 7 off shift-based staffing system that focused on continuity of care on an urban academic hospitalist service. The purpose of the model was to allow hospitalists more continuity with their patients and to measure how this is associated with mortality, length of stay (LOS), and readmissions.

METHODS: This was a retrospective cohort in which matching 6 months of patient data were compared pre and post implementation of a new staffing model in 2016. The model changed from a drip system of admissions with dedicated rounders and admitters to one where rounding hospitalists admitted their own patients for daytime hours on the first four days of their week. They did not admit on their last three days. Total hospitalists involved in care were assessed as a measure of handoffs. Outcome measures included mortality, length of stay, and 30-day readmission rate.

RESULTS: We identified 2,102 patients admitted to the hospital medicine service from January 1st 2015 to June 30th 2015 and 1,870 patients admitted during the same time in 2016. Patient characteristics were found to be similar between years. There was on average 2.76 hospitalists per patient encounter in 2015 versus 2.19 in 2016 [p < 0.001]. Average length of stay was significantly reduced by over a half day from 6.34 days in 2015 to 5.7 days in 2016 [p = 0.002]. Of patients admitted prior to intervention, 46.52% had just one rounder during the hospital course while 57% had one rounder in the intervention group [p < 0.001]. In 2016, 23% of patients had one provider from admission to discharge compared to none in 2015 [p < 0.001]. There was no difference in mortality between groups. There was a direct relationship between the number of hospitalists and death in both groups with an increase odds ratio of 1.37 with each additional hospitalist added to the patient’s care [p < 0.001]. There was no statistical difference in 30-day readmissions between years.

CONCLUSIONS: Improvement in hospitalist continuity significantly decreased patient LOS in this novel seven day staffing model. There was not a significant difference in mortality or 30-day readmission rate between groups, but there were increased odds of death with additional hospitalist handoffs. Implementing a new staffing model focusing on continuity is an inexpensive and potentially cost effective intervention for a hospitalist service.

figure u

CUSTOMIZING EMR REGISTRY REPORTING: USING DISEASE MANAGEMENT REGISTRY DATA TO IMPROVE COMMON ACO DIABETES METRICS John Voss; William Clay; Douglas Thaggard; Ira Helenius; Katharine Schlag. University of Virginia, Charlottesville, VA. (Control ID #2703575)

BACKGROUND: Electronic medical records (EMRs) offer promise to facilitate chronic illness care quality improvement (QI) but the data may be difficult to access in readily usable form for individual clinician QI efforts. To improve diabetes mellitus (DM) control in an internal medicine resident (IMR) clinic, we developed and disseminated customized outcomes reports using EMR disease management registry data for 61 2nd & 3rd year IMRs in an internal medicine residency clinic.

METHODS: Using a locally developed structured PDSA tool, IMRs self-designed HbA1c or HBP control or foot exam QI projects for implementation over 6 months (3rd year IMRs) or 12 months (2nd year IMRs). Each IMR received a monthly report for their own clinic DM patients displaying summary run charts of HbA1C control, HBP control, and foot exam completion as well as a list of individual patients and their respective clinical values. Residents were instructed to use customary available clinic resources to achieve their project goals. To create reports, DM registry data were downloaded biweekly and processed via an R script to generate reports displaying run charts of HbA1C control (mg/dL, proportion <8%, proportion >9%), HBP control (mm Hg systolic and diastolic, % < 140/90, % < 150/90), foot exam completion (% requiring) and lists of unique patient values and individual clinician versus clinic population means. After 12 months, DM outcomes were assessed using mixed model linear or logistic regression controlling for IMR, IMR project selection, and elapsed time.

RESULTS: IMRs selecting HbA1c control projects reduced their patient’s HbA1c values from a mean of 8.33 (sd = .43) to 8.07 (sd = .31) compared to 7.82 (sd = .38) to 7.84 (sd = .57) for all other residents (b coefficient for trend = 0.005 mg/dL per week, overall p < 0.001.) No initial meaningful difference existed in the mean proportion of patients needing a foot exam between foot exam (FE) IMRs (0.67) and non-foot exam (NFE) IMRs (0.68). Although this proportion decreased significantly over time for all residents, the proportion of patients needing a foot exam at the end of the project for FE IMRs (0.38) was substantially lower relative to NFE IMRs (0.49), and the difference in this change over time was marginally significant, p = 0 .08. IMRs electing to improve HBP control were unable to demonstrate improvement over time. A contributing factor may be a significantly higher proportion of patients with a BP > 150/90 at baseline (0.35) for the HBP control residents relative to non-HBP control residents (0.25), p = 0 .02.

CONCLUSIONS: Structured tools and custom EMR reports can be developed using free and readily available software to assist clinicians to improve chronic illness care metrics of interest in the era of value-based purchasing. These tools alone may be insufficient to improve more challenging outcomes like hypertension control without additional resources beyond individual clinician effort.

DAILY MARIJUANA USERS IDENTIFIED IN PRIMARY CARE AND EMERGENCY SBIRT SETTINGS: CHARACTERISTICS AND SCREENING RESULTS Sandeep Kapoor 1, 3; Jeanne Morley2, 3; Kristen Pappacena4; Cherine Akkari4; Camila Bernal4; Charles Neighbors4, 5; Mark Auerbach1, 3; Nancy Kwon1, 3; Jonathan Morgenstern2, 3; Joseph Conigliaro2, 3; Megan O’Grady4, 5. 1Northwell Health, New Hyde Park, NY; 2Northwell Health, Great Neck, NY; 3Hofstra Northwell School of Medicine, Hempstead, NY; 4The National Center on Addiction and Substance Abuse, New York, NY; 5Yale Univeristy, New Haven, CT. (Control ID #2705043)

BACKGROUND: Marijuana is the most widely used illicit substance in the US. Its rates of use as well as rates of daily users have steadily increased over the past several years. Within primary care settings, individuals who screen positive for drugs most commonly use marijuana. An estimated 25–50% of daily users will develop cannabis use disorder, making them a potentially important focus in healthcare settings.

METHODS: This study examined the characteristics and screening results of daily marijuana users identified in healthcare settings as part of an interdisciplinary Screening, Brief Intervention, and Referral to Treatment (SBIRT) program. Patients presenting to a Primary Care Practice (PCP) or an Emergency Department (ED) were screened for risky substance use with the AUDIT and DAST-10 and, if positive, were further assessed on psychosocial factors, substance use severity, and demographics (n = 1604).

RESULTS: Of the patients who screened positive and participated in the assessment, 44% used marijuana at least one day in the past 30 and 15% used marijuana 25 or more days. These daily users (n = 239) had relatively low DAST-10 scores (M = 3.02) and 42% screened into the moderate or higher risk category on the DAST-10 (3+). Using multivariate logistic regression analyses, we identified significant predictors of 1) being a daily user (vs. non-daily) and 2) screening as moderate/high risk (vs. low) on the DAST-10. Among the marijuana users (n = 739), daily use was significantly predicted by being male, Latino, and an ED patient. Among daily users, significant predictors of screening moderate/high on the DAST-10 were: younger age, being a user of other drugs, and more frequent alcohol use.

CONCLUSIONS: Findings suggest that while SBIRT programs in healthcare settings are likely to encounter daily marijuana users, the majority may screen at low risk for health or psychosocial problems. Patients who screen into higher risk categories are likely to be younger and using other substances. Therefore, interventions for daily marijuana users may need to be tailored such that interventions for lower risk patients focus on reducing marijuana use, while interventions for higher risk patients be adapted for multiple substances and younger users.

figure v

Predictors of daily marijuana use (using ≥ 25 days/month) among all marijuana users (n = 739)

DAY OF THE WEEK AND PHYSICIAN DECISION MAKING Michael L. Barnett 4, 1; Jeffrey A. Linder2; Hannah Neprash3; Anupam B. Jena3. 1Brigham and Women, Boston, MA; 2Brigham and Women’s Hospital, Boston, MA; 3Harvard Medical School, Boston, MA; 4Harvard T. H. Chan School of Public Health, Boston, MA. (Control ID #2705781)

BACKGROUND: Physicians make dozens of clinical decisions daily, which may create the potential for “decision fatigue,” whereby decision-making effort slumps over time. However, little evidence has explored this phenomenon across the week for multiple clinical decisions. We hypothesized that full-time, but not part-time, physicians would display increasing decision fatigue during the workweek for effort-intensive decisions such as avoiding narcotic prescribing and counseling on smoking cessation.

METHODS: We performed a cross-sectional analysis of the nationally representative National Ambulatory Medical Care Survey of office visits by adults from 2000–2012. The main analysis was restricted to “full-time” physicians who saw patients for ≥4 weekdays in the surveyed week. The key exposure was a visit occurring the beginning of the week (Monday/Tuesday) versus the end of the week (Thursday/Friday). We measured the rates of 3 clinical measures potentially vulnerable to decision fatigue: antibiotics for upper respiratory infections (URIs), narcotics for acute low back pain (LBP) and smoking cessation counseling for active smokers. We estimated the rates of each outcome in the beginning vs. the end of the week using logistic regression adjusting for patient and visit characteristics. We compared results for “full time” physicians with a sample of “part-time” physicians seeing patients for ≤3 weekdays in the week who should be less vulnerable to decision fatigue across the week. All estimation of results took account of the multistage probability design of the surveys.

RESULTS: From 2005–2012, the study sample contained 82,725 office visits eligible for at least one of the outcomes examined. Office visits in the beginning versus end of the week were similar across all patient characteristics measured (all p > 0.20). Comparing the end of the week to the beginning, after adjustment physicians were significantly more likely to prescribe narcotics for LBP (OR 1.18, 95% CI 1.05-1.32, p = 0.006). In contrast, physicians were less likely to perform smoking cessation counseling at the end of the week (OR 0.78, 95% CI 0.67-0.92, p = 0.002). Neither of these measures had statistically significant differences across the week among physicians working ≤3 days in the week (p > 0.17). There was no significant difference in antibiotic prescribing for URIs by day of the week for either group (p = 0.83).

CONCLUSIONS: We observe evidence consistent with decision fatigue among physicians over the workweek for 2 of 3 measures examined: increased rates of narcotic prescribing for LBP and decreased rates of time-intensive smoking cessation counseling. We observed these differences for full-time, but not part-time, physicians, arguing against patient selection in the end of the week driving these findings. These results suggest that strategies to improve the quality of care delivery should also address the potential contribution of decision fatigue for full-time clinicians.

DE-ESCALATION OF BROAD SPECTRUM ANTIBIOTICS FOLLOWING NEGATIVE CULTURES IN PNEUMONIA: RATES AND OUTCOMES Michael B. Rothberg 2; Abhishek Deshpande2; Peter B. Imrey2; Pei-Chun Yu2; Peter K. Lindenauer1. 1Baystate Medical Center, Springfield, MA; 2Cleveland Clinic, Cleveland, OH. (Control ID #2707495)

BACKGROUND: For patients at risk for multidrug resistant organisms, including those with healthcare associated pneumonia and other severe pneumonias, recommended empirical therapy includes one agent with activity against methicillin-resistant Staphylococcus aureus (MRSA) and another with activity against Pseudomonas. Following negative blood and respiratory cultures, guidelines are unclear as to whether antibiotics should be de-escalated. We assessed the rate of de-escalation and associated outcomes among a large cohort of pneumonia patients.

METHODS: This retrospective cohort study assessed adult patients admitted with pneumonia from 2010–2015 to 168 US hospitals that participate in the Premier data base, providing administrative and microbacteriological data. We included only patients who had blood or respiratory cultures and were begun on vancomycin together with an antipseudomonal drug other than a quinolone by hospital day 1. Patients with any positive culture or who died or were discharged before hospital day 4 were excluded. De-escalation was defined as stopping vancomycin and the antipseudomonal drug by the 4th hospital day after cultures were obtained, but continuing another antibiotic. Patients were matched on propensity for de-escalation and compared on subsequent transfer to intensive care (ICU), inpatient mortality, length of stay (LOS), cost and 30-day readmission. The propensity model included patient demographics, co-morbidities and treatments received on day 4 after cultures. Certain treatments (e.g. mechanical ventilation, vasopressors) served as proxies for severity of illness. We also compared patient outcomes across quartiles of de-escalation at the hospital level.

RESULTS: We identified 22,400 patients who met inclusion criteria. Following cultures, 4114 (18.4%) had both drugs stopped within 4 days. Hospital rates of de-escalation ranged from 0 to 67%. Median age was 72 years, 54% were male, and 39% were admitted to the ICU. Compared to patients without de-escalation, those de-escalated had similar demographics, but fewer co-morbidities and less severe disease at 1 and 4 days after culture. After propensity-matching, there were no significant differences in any of 83 measured variables. In the matched sample, patients who were de-escalated had 21% shorter LOS, 21% lower costs, and lower odds of subsequent transfer to ICU (OR 0.28; 95% CI 0.17 - 0.47)) and inpatient mortality (OR 0.71, 95% CI 0.60 - 0.83). There were no significant associations between hospital quartile of de-escalation and inpatient mortality, cost, LOS or readmission.

CONCLUSIONS: In a large US inpatient database, <20% of pneumonia patients had antibiotic coverage de-escalated following negative cultures, but there was wide variation by hospital. De-escalated patients and hospitals with high rates of de-escalation did not have worse outcomes. Following negative cultures, de-escalation appears safe in selected patients.

DECEDENTS OF OPIOID OVERDOSE: INJECTION-RELATED OVERDOSE DEATHS DIFFER FROM NON-INJECTION RELATED DEATHS. Emily E. Hurstak 1, 2; Christopher Rowe3; Emily Behar3; Caitlin Turner3; Rachel Cabugao3; Phillip Coffin3. 1UCSF, San Francisco, CA; 2San Francisco Free Clinic, San Francisco, CA; 3San Francisco Department of Public Health, San Francisco, CA. (Control ID #2703722)

BACKGROUND: Opioid overdose is the leading cause of unintentional injury death in the United States. Death certificates often exclude details on individual causative drugs and may underestimate heroin-related overdose deaths due to challenges in relying on toxicology alone. We propose a strategy to differentiate overdose deaths related to injection drug use (IDU) from non-injection related opioid overdoses in order to improve understanding of opioid overdose epidemiology and to guide prevention efforts.

METHODS: We reviewed all accidental drug overdose cases occurring in San Francisco County from 2006–2012 using information from the California Electronic Death Reporting System. We included cases if the death occurred in San Francisco County, was attributed to intoxication from heroin or opioid analgesics, and was not determined to be a suicide or homicide. We extracted descriptive data from decedent case narratives from the San Francisco Office of the Chief Medical Examiner (OCME), including death scene details and health history. We created a definition for injection-related overdose deaths based on review of toxicology, decedent history, and death scene details. We compared opioid analgesic overdose cases to cases involving injection drug use.

RESULTS: Between 2006–2012, we identified 255 deaths that involved IDU and 552 deaths attributed to opioids without evidence of IDU. Sixty-percent (59.6%) of IDU-related deaths could be designated as definite heroin-overdoses. Decedents of IDU-related overdose were younger (45.2 vs. 48.6 years, p < 0.001) and more likely to be male (76.9% vs. 67.2%, p = 0.005). Decedents of IDU-related overdose were more likely to be found in a single room occupancy hotel (38.4% vs. 25.3%, p < 0.0001) or a public space (14.5% vs. 8.2%, p = 0.005), compared to a private residence (40.0% vs. 56.9%, p < 0.001). IDU-related overdoses were more likely to be discovered by a passerby (8.6% vs. 4.5%, p = 0.02) or a program staff member (37.3 vs. 29.9, p = 0.04), rather than by a cohabitant (35.3% vs. 45.2%, p = 0.02). Fifty percent (49.8%) of IDU-related overdose decedents had a known prior history of IDU, compared to 10.3% of non-IDU overdose deaths. Benzodiazepines contributed to the cause of death in 8.2% of IDU-related overdoses compared to 24.3% of non-IDU deaths (p < 0.001).

CONCLUSIONS: We created a definition for IDU-related opioid overdose by incorporating historical details about the decedent and death scene. Using this definition, we identified differences in overdose cases that involved injection drug use compared to those that did not. Traditional opioid overdose prevention efforts target people with injection drug use. However, only a small proportion of decedents of opioid analgesic overdose have histories of injection drug use and our results suggest differences in the drug use patterns of injection-related and non-injection related overdoses. Increasing our understanding of opioid overdose will improve overdose education programs and outreach efforts to populations at risk.


BACKGROUND: Fresh Frozen Plasma (FFP) is a major source of coagulation factors, often indicated in bleeding due to coagulopathy. It is frequently transfused inappropriately, especially due to inadequate adherence to evidence-based transfusion guidelines. Like any other blood product, FFP transfusion is expensive and associated with serious transfusion reactions. Studies have shown that interventions that emphasize transfusion guidelines can significantly decrease inappropriate utilization of FFP. Aim of this project was to evaluate current FFP transfusion practices at our community hospital and improve utilization by reinforcing transfusion guidelines via the Computerized Physician Order Entry (CPOE) system in addition to training sessions.

METHODS: Stage I included review of FFP transfusion orders over a period of six months from April 2015 to September 2015. Evaluation included number of patients, units of FFP transfused and indications. Indications were deemed appropriate or inappropriate in accordance with institutional transfusion guidelines. Massive transfusion protocols and plasmapheresis were excluded. In October 2015, a new ordering system was incorporated in CPOE which required ordering physicians to choose from a menu of evidence-based indications for FFP transfusion. Training sessions were conducted for residents and attending physicians to reinforce transfusion guidelines. Stage II commenced after implementation of the new system, from November 2015 to May 2016. Records from the new ordering system were evaluated using the same parameters as Stage I.

RESULTS: In Stage I, total of 310 units of FFP were transfused to 101 patients. During Stage II, total of 251 units of FFP were transfused to 80 patients. A 19.04% decrease in number of units transfused and 20.8% decrease in number of recipients, after implementation of the new ordering system. Inappropriate transfusions that did not meet institutional guidelines decreased from 61 units to 25 units (from 19.6 to 9.9%). Most common indications for FFP transfusion were Gastrointestinal and Intracranial bleeding with coagulopathy. Most common inappropriate indications were bleeding without coagulopathy and elevated INR without evidence of bleeding.

CONCLUSIONS: Incorporation of transfusion indications in the electronic ordering system has the potential to decrease inappropriate Fresh Frozen Plasma transfusions by prompting ordering physicians to select evidence-based indications. This may in turn reduce cost and potentially reduce adverse transfusion reactions. This is a continuing project which will assess several other parameters in the upcoming months including number of transfusion reactions before and after implementation of the new system, reasons for inappropriate orders, and affect on pre and post-transfusion INR. Our goal is to continue education and awareness about evidence-based transfusions in order to improve quality of patient care.

DEPRESCRIBING IN POLYPHARMACY PATIENTS: A PILOT STUDY OF PRESCRIBING PRACTICES IN AN ACADEMIC PRIMARY CARE CLINIC. James L. Wofford 4; Carolyn F. Pedley2; Claudia L. Campos4; Feben Girma3; Melanie Martin1. 1Wake Forest, Greensboro, NC; 2Wake Forest, Winston-Salem, NC; 3Wake Forest Baptist Medical Center, Winston-Salem, NC; 4Wake Forest University, Winston-Salem, NC. (Control ID #2704790)

BACKGROUND: “Deprescribing”, the process of tapering, or discontinuing of drugs with the goal of minimizing polypharmacy and improving outcomes, has now become a national movement. Whether or not proponents of deprescribing can counter the forces that encourage polypharmacy (pharmaceutical marketing, attention to clinical inertia, consumer demand), a better understanding of prescribing practices in the faced of polypharmacy would be useful for ambulatory education. We explored the prescribing actions of clinicians in a single urban academic primary care clinic to determine the prevalence and types of medication changes and to compare prescribing patterns of senior clinicians with resident physicians

METHODS: We performed a one-week audit of all medication lists for return continuity visits at this community-based, academic primary care clinic, Each EMR medication list was examined for medication additions, changes, and/or deletions during that visit. Patients with polypharmacy were defined as having 10 or more medications documented on their medication list. We compared prescribing actions of senior clinicians with those of resident clinicians, and for patients with and without polypharmacy.

RESULTS: 186 patients who attended a return continuity visit during the designated audit week were seen by one of 34 clinicians (121 patients to 11 senior clinicians, 65 to 21 resident physicians) Patients had a mean age of 56.8 (+12.8) years, and 24% (46/187) were aged 65 or older. 18% (33/186) were seen by a clinician other than their designated PCP during the return clinic visit. The number of medications listed in the EMR averaged 10.0 + 5.4, but did not significantly differ by age group (mean 11.6 (+0.8) for >65 years versus 9.4 (+0.5) for <65). Fifty percent (93/186) of patients had 10 or more medications. Medication lists were only slightly longer for senior clinicians than for residents (10.3 (+0.5) versus 9.3 (+0.7)). Sixty percent (111/185) of return visits reflected at least one medication change to the list during the clinic visit. At least one medication was added (36.2% (67/185)), changed (34.5% (64/185)), or deprescribed (16.2% (30/185)), respectively, Resident clinicians more often added medications than did senior clinicians (45.3% (24/64) vs 32.3% (39/121), p = .08), but there was no significant difference between resident physicians and senior clinicians in medication dosing changes (39.1% (25/64) vs 34.7% (42/121)), or deprescribing of a medication (14.1% (9/64) versus 17.4% (21/121)). There was no signficant difference in deprescribing for polypharmacy patients (>9 medications) versus non-polypharmacy patients (12.9% (12/93) versus 19.6% (18/93) p = .21).

CONCLUSIONS: In this academic-affiliated primary care clinic, the rate of medication addition was more than twice the rate of deprescribing. Senior clinicians were no more likely than resident cliniciansto to deprescribe medications.The presence of polypharmacy was not associated with a higher rate of deprescribing.

DEPRESCRIBING: A SURVEY STUDY OF PATIENT ATTITUDES AND EXPERIENCES THAT PREDICT MEDICATION DISCONTINUATION Amy Linsky 1, 3; Steven R. Simon1; Kelly Stolzmann1; Mark Meterko2. 1VA Boston Healthcare System, Boston, MA; 2VHA Office of Analytics and Business Intelligence (OABI), Bedford, MA; 3Boston Medical Center, Boston, MA. (Control ID #2699348)

BACKGROUND: Polypharmacy is associated with adverse medication effects. One potential solution is deprescribing, which is the intentional, proactive, rational discontinuation of a medication that is no longer indicated or whose potential harms outweigh potential benefits. We sought to characterize patient characteristics, attitudes and healthcare experiences associated with prior medication discontinuation.

METHODS: We conducted a mail survey in August-October, 2015 using the Patient Perceptions of Discontinuation (PPoD) instrument. We randomly sampled 1600 Veterans receiving primary care at Veterans Affairs medical centers nationally with ≥ 5 concurrent prescribed medications, oversampling women. The primary outcome was response to: “Have you ever stopped taking a medicine (with or without your doctor’s knowledge)?” We restricted analyses to those who answered “No” or “Yes.” The primary predictors of interest were eight validated attitudinal scales, all on a 1–5 scale (Beliefs about Medications Questionnaire - Overuse, Trust-provider, CollaboRATE, Medication Concerns, Provider Knowledge, Interest in Stopping Medicines, Patient Involvement in Decision Making, and Unimportance of Medicines). Other predictors included demographics, health status, and health care experiences. Multivariable logistic modeling associated patient factors with prior medication discontinuation.

RESULTS: Respondents (n = 803; adjusted response rate, 52%), were predominantly male (85%); non-Hispanic white (68%), age ≥ 65 years (60%), with generally poor (16%) or fair (45%) health. Participant attitudes toward medications and their providers were generally favorable. Respondents disagreed that medicines were unimportant and overused, with mean scores of 2.39 and 2.91 on the Unimportance of Medicine and the BMQ-Overuse scales, respectively. Respondents held generally positive views of providers, with mean scores of 3.75 on Provider Knowledge and 3.56 on Trust-provider. Concurrently, patients were generally interested in stopping medicines (mean 3.42). Just over 1 in 3 patients (34%) reported having stopped a medicine in the past. In a multivariable logistic regression model (p < 0.001, pseudo-R2 0.31, c-statistic 0.82), factors associated with experiencing discontinuation included being told to stop a medicine or asking to stop a medicine, greater interest in deprescribing and in shared-decision making, and higher education. Factors associated with decreased discontinuation were more prescriptions, higher trust in provider, and seeing a VA clinical pharmacist.

CONCLUSIONS: This study suggests that more highly educated patients who have interest in deprescribing and in shared-decision making may be more receptive to discontinuation discussions. Future research evaluating how to incorporate this survey and its findings into clinical workflow through the design of clinical decision-support interventions may help promote safe, appropriate, and rational use of medications.

DEPRESSION AMONG HEALTHCARE WORKERS—CONTRIBUTION OF PERSONALITY, WORK CHARACTERISTICS, COPING, AND BURNOUT Wei Duan-Porter 2; Daniel Hatch3; Jane Pendergast1; Guy Potter1. 1Duke University School of Medicine, Durham, NC; 2Minneapolis VA Health Care System, Minneapolis, MN; 3Duke University Center for the Study of Aging and Human Development, Durham, NC. (Control ID #2705619)

BACKGROUND: Relative contributions of individual and work factors to the development of depression remain unclear. There is also debate about the relationship between depression and burnout. We sought to examine the association of individual and work characteristics with depressive symptoms over 12 months. We evaluated the contribution of demographics, personality, work characteristics, and coping to differences in depression between individuals. Then, we investigated how burnout was associated with differences in depression both between and within individuals.

METHODS: Participants were recruited from nurses and associated healthcare staff employed by a large academic health system in the Southeastern U.S. Baseline assessments included demographics, health information, work characteristics, coping, burnout, and depression. Participants then completed monthly surveys on burnout and depression. Selected participants had full baseline data, and a minimum of 3 completed follow-up assessments (n = 281, 89% of these had data for at least 9 of 12 months). We used random effects hierarchical linear modeling to examine differences in depression between and within individuals over time. We added categories of predictors in a step-wise fashion to successive models, beginning with only demographics, then personality, work characteristics, coping, and finally burnout.

RESULTS: Most participants were female (92%), white (82%), and nurses (90%). Demographic factors were not significantly associated with depression, but personality, work characteristics, and coping all had significant fixed effects. Personality accounted for 36% of variation between individuals, while work characteristics and coping explained an additional 5 and 8%, respectively. Burnout was significantly associated with depression, but the 2 dimensions—exhaustion and disengagement—had contrasting effects. Whereas baseline exhaustion contributed to differences in depression between individuals (coefficient 2.44, p <0.001), time-varying follow-up exhaustion scores were not associated with further changes in depression within individuals over time (coefficient 0.37, p = 0.09). Time-varying disengagement scores were associated with changes in depression within individuals (coefficient 0.52, p = 0.01), but baseline disengagement was not significantly associated with differences between individuals (coefficient −0.51, p = 0.32).

CONCLUSIONS: Personality, work characteristics, coping, and burnout were associated with depression among healthcare workers. Higher exhaustion was associated with more depressive symptoms when comparing individuals, and higher disengagement over time was associated with increased depressive symptoms within individuals longitudinally. Our findings suggest that interventions to promote wellbeing for healthcare workers should address both work and individual factors. Careful characterization of type and severity of existing burnout may impact the expected benefits and relevant timeframe for evaluating future interventions.

DEPRESSION AND BURNOUT AMONG MEDICAL STUDENTS DURING 4 yearS OF MEDICAL SCHOOL: A SINGLE CENTER STUDY Padmini Ranasinghe 1; Brandyn D. Lau2. 1Johns Hopkins School of Medicine, Baltimore, MD; 2The Johns Hopkins University School of Medicine, Baltimore, MD. (Control ID #2703533)

BACKGROUND: Previous studies reported that medical students experience symptoms depression and burnout at a higher rate compared to the general population. Medical students also have a higher rates of suicidal ideation. Burnout rate is up to 50% among medical students: 53% of medical students reported experiencing high emotional exhaustion and/or depersonalization. Recent systematic review, estimated that prevalence of depression or depressive symptoms among medical students was 27.2 and that of suicidal ideation was 11.1%. Despite the deleterious consequences of these symptoms, only a small minority (up to 15.7%) of medical students ultimately seek treatment.

METHODS: In this longitudinal single center survey study, health and wellness of medical students across 4 years of medical school were evaluated using a combination of validated and adapted assessment tools. Standardized surveys were administered annually between 2009 and 2014. This study was approved by the Institutional Review Board and consent was obtained at the time each first survey was administered. Students who matriculated in 2009, 2010, or 2011 were included in this report. Each participant completed 5 surveys: one at the beginning of each year of medical school and an exit survey post-graduation. PHQ 9 was used to screen for depression and burnout was assessed based on a single question of “Can you classify your current level of burnout”.

RESULTS: Three hundred sixty (360) students matriculated to the undergraduate medical education program between 2009 and 2011. 284 (78.9%) were continuously enrolled during pre-clinical and clinical years, and were eligible for inclusion. In this study 128 medical students (45.1%) completed the health and wellness survey each year. During the first year of medical school, 4.7% of respondents reported symptoms of at least minor depression. During second year 6.6% of respondents reported symptoms of at least minor depression, 24.9% in third year, and 32.7% in fourth year. The proportion of respondents who self-reported at least some symptoms of burnout was low during first year (4.2%), increased in second year (18.8%), and remained high in third year (17.7%) and very high in fourth year (31.3%).

CONCLUSIONS: Medical students’ psychological health including depression and burnout are high during medical school: symptoms of at least minor depression and self-reported burnout increased each year in the school. This suggest we as generalist involved in undergraduate medical education should be aware of these subtle changes which have life long implications. We also should be mindful and available for students to become effective and compassionate student centered teaches

DEPRESSIVE SYMPTOMS PREDICT SHORT TERM MORTALITY IN THOSE WHO REPORT EXCELLENT OR VERY GOOD STATE OF HEALTH Nathalie Moise 1; Yulia Khodneva3; Joshua S. Richman3; Karina Davison1; Ian M. Kronish1; Jonathan Shaeffer4; Monika M. Safford2. 1Columbia University Medical Center, New York, NY; 2Weill Cornell Medical College, New York, NY; 3University of Alabama, Birmingham, AL; 4University of Colorodo, Denver, CO. (Control ID #2704319)

BACKGROUND: The relationship between elevated depressive symptoms, which often relapse and remit, and mortality has been poorly elucidated given the use of single measurements of depressive symptoms, the unclear role of health status and incomplete covariate adjustment. We aimed to determine whether time varying depressive symptoms predict all-cause and cause-specific mortality in those with self-reported excellent or very good health.

METHODS: The Reasons for Geographic and Racial Differences in Stroke (REGARDS) is a national, population-based longitudinal study conducted from 2003–2007 in general continental U.S. communities. 29,491 black and white U.S. adults aged 45 years and older randomly sampled within race-sex-geographic strata with available baseline depressive symptoms. Elevated depressive symptoms were defined as CES-D-4 ≥ 4 measured at baseline and on average 5 and 7 years later. Cox proportional hazard regression models were constructed to assess cancer, non-cardiovascular (CVD), CVD and all-cause mortality. All analyses were stratified by baseline self-reported health status.

RESULTS: Overall, 11.0% had elevated depressive symptoms; the average age was 64.9 years, 55% were female, 41% black, 54% had poor, fair or good health, and 16% died in the follow up period. Time-varying depressive symptoms were significantly associated with nonCVD (aHR = 1.29, 95% CI 1.16–1.44) and all-cause (aHR = 1.24, 95%CI 1.14–1.39), but not cancer (aHR = 1.15, 95%CI 0.96–1.38) or CVD (aHR = 1.13, 95%CI 0.98–1.32) death adjusting for age, gender, race, education, income, insurance status, region, diabetes, blood pressure, body mass index, kidney disease, hyperlipidemia, CVD, cognitive impairment, lung disease, aspirin, antidepressant, statin or antihypertensive use, smoking, physical inactivity, alcohol, medication non-adherence, C-reactive protein, physical health, and perceived stress. Depressive symptoms were particularly related to all-cause (aHR = 1.48, 95%CI 1.27–1.78), CVD (aHR = 1.37, 95%CI 0.99–1.91), nonCVD (aHR = 1.54, 95%CI 1.24–1.92) and cancer (aHR = 1.36 95% 0.97–1.91) death in those who reported excellent or very good health. Baseline analyses yielded similar results.

CONCLUSIONS: Depressive symptoms confer both a short-term risk for all-cause mortality, CVD, non-CVD death and cancer death, particularly in those with excellent or very good health. These findings may have implications for timely treatment, regardless of health status.

DERIVATION OF A CLINICAL PREDICTION MODEL TO PREDICT UNCHANGED INPATIENT ECHOCARDIOGRAMS Craig G. Gunderson 1, 3; Elizabeth S. Gromisch2, 3; John J. Chang1, 3; Brian Malm2, 3. 1Yale Medical School, West Haven, CT; 2Yale University School of Medicine, New Haven, CT; 3VA Connecticut Healthcare System, West Haven, CT. (Control ID #2705132)

BACKGROUND: Transthoracic echocardiography (TTE) is one of the most frequently ordered tests in healthcare. Repeat TTE defined as TTE done within 1 year of a prior TTE represent 24–42% of all studies. In the present study we derive a clinical prediction model to predict unchanged repeat TTE with the goal of defining a subset of studies that are potentially unnecessary.

METHODS: Single-center retrospective cohort study of all hospitalized patients who had a repeat TTE between October 1, 2013 and September 30, 2014. Patients with repeat TTEs within the past year were grouped into those with major changes on repeat TTE and those with mild or no change. Logistic regression was used to define variables that were independent predictors of major changes and a score was constructed based on the regression coefficients.

RESULTS: During the study period 601 patients had at least one inpatient TTE. Of these 601 patients, 211 (35%) had a prior TTE within the past year. Of those 211 patients, 67 (32%) had no change, 66 (31%) had minor changes and 78 (37%) had major changes. On logistic regression, 5 variables were independent predictors of major new TTE changes, including history of intervening acute myocardial infarction, intervening cardiothoracic surgery, major new EKG changes, prior valve disease, and prior chronic kidney disease. Using the β-coefficient for each of these variables we defined a clinical prediction model which we named the CAVES score. The prevalence of major TTE change for the full cohort was 35%. For the group with a CAVES score of −1 that probability was only 5.6%, for the group with a score of 0, the probability was 17.7%, and for the group with a score ≥1 the probability was 55.3%.The bootstrap corrected c-statistic for the model was 0.78 (95% CI, 0.72–0.85) indicating good discrimination.

CONCLUSIONS: We report on the derivation of a 5 variable clinical prediction model called the CAVES score to predict major changes in repeat TTEs. Overall the CAVES score had good discrimination and calibration. If further validated it may be useful to predict repeat TTEs that are unlikely to have major changes.

figure w

DETERMINING THE APPROPRIATENESS OF MEDICATION ALERTS OVERRIDDEN AS INACCURATE WARNINGS Christine Rehr 1, 2; Adrian Wong1, 4; Diane L. Seger2; David W. Bates1, 3. 1Brigham and Women’s Hospital, Boston, MA; 2Partners Healthcare, Wellesley, MA; 3Harvard Medical School, Boston, MA; 4MCPHS University, Boston, MA. (Control ID #2704776)

BACKGROUND: Medication clinical decision support has shown progress in reducing adverse drug events and costs, and improving patient outcomes (1). However, research has also shown that many of these warnings are frequently overridden, and often inappropriately, likely because of alert fatigue and poor alert design (2). This study was performed to evaluate the alerts providers have overridden with the reason “inaccurate warning” in order to learn more about the alerts perceived to be erroneous.

METHODS: The data for this study were overrides occurring in the intensive care units (ICU) at a major academic medical center between June 15 and November 15, 2016. The overrides were analyzed by two independent reviewers using predetermined criteria for appropriateness. The three alert types analyzed were chosen based on clinical significance and frequency for ICU patients: drug-drug interactions (DDI), drug-allergy interactions (DAI), and dose alerts.

RESULTS: There were 26501 overridden alerts in the ICUs, 268 of which were documented with the reason “inaccurate warning” (1.0%). Of the 15 DDIs, all were inappropriate overrides, and all but one were for QT prolongation. Eight of the 48 DAIs (16.7%) were inappropriate overrides, as the patient had a true allergy and had not tolerated the medication before. Two of the 13 renal dose overrides (15.4%) were found to be inappropriate, whereas the 11 appropriate overrides were for correct renal replacement therapy doses or for patients without renal insufficiency. Two of 4 weight-based dose overrides (50%) were inappropriate, as the provider should have used the patient’s weight to calculate the correct dose. Five of 13 overdose overrides (38.5%) were inappropriate, whereas some of the appropriately overridden alerts had incorrect overdose thresholds (23.1%) or were alerts on correct electrolyte protocols (23.1%), meaning 46.2% of these overdose alerts should never be fired.

CONCLUSIONS: This study shows that DDIs are not appropriately overridden with the “inaccurate warning” reason, whereas some dosing and DAI alerts are. The appropriate DAI overrides indicate that there should be a better mechanism for showing only accurate alerts, such as removing documented allergies based on prior tolerance or incorrect ingredient match. The appropriate dose overrides indicate that the alerts need to better identify patient conditions, dosing mechanisms, and protocols, in addition to altering thresholds and cancelling certain electrolyte alerts. References 1. Bates DW, Teich JM, Lee J, et al. The impact of computerized physician order entry on medication error prevention. J Am Med Inform Assoc 1999;313–21. 2. Nanji KC, Slight SP, Seger DL, et al. Overrides of medication-related clinical decision support alerts in outpatients. J Am Med Inform Assoc 2014;21:487–91.

DETERMINING THE OPTIMAL SCREENING INTERVAL FOR TYPE 2 DIABETES MELLITUS USING A RISK PREDICTION MODEL Andrei Brateanu 1; Thomas Barwacz2; Lei Kou1; Sihe Wang1; Anita D. Misra-Hebert1; Bo Hu1; Abhishek Deshpande1; Nana Kobaivanova1; Michael B. Rothberg1. 1Cleveland Clinic, Cleveland, OH; 2University Hospital, Cleveland, OH. (Control ID #2703815)

BACKGROUND: Progression to diabetes mellitus (DM) is variable and the screening time interval not well defined. The American Diabetes Association and US Preventive Services Task Force suggest screening every 3 years, but evidence is limited. The objective of the study was to develop a model to predict the probability of developing DM and suggest a risk-based screening interval.

METHODS: We included non-diabetic adult patients screened for DM in the Cleveland Clinic Health System if they had at least two measurements of glycated hemoglobin (HbA1c), an initial one less than 6.5% in 2008, and another between January, 2009 and December, 2013. Cox proportional hazards models were created. The primary outcome was DM defined as HbA1C greater than 6.4%. The optimal rescreening interval was chosen based on the predicted probability of developing DM.

RESULTS: Of 5084 participants, 100 (4.4%) of the 2281 patients with normal HbA1c and 772 (27.5%) of the 2803 patients with prediabetes developed DM within 5 years. Factors associated with developing DM included HbA1c (HR per 0.1 units increase 1.20; 95%CI, 1.13–1.27), family history (HR 1.31; 95%CI, 1.13–1.51), smoking (HR 1.18; 95%CI, 1.03–1.35), triglycerides (HR 1.01; 95%CI, 1.00–1.03), alanine aminotransferase (HR 1.07; 95%CI, 1.03–1.11), body mass index (HR 1.06; 95%CI, 1.01–1.11), age (HR 0.95; 95%CI, 0.91–0.99) and high-density lipoproteins (HR 0.93; 95% CI, 0.90–0.95). Five percent of patients in the highest risk tertile developed DM within 8 months, while it took 35 months for 5% of the middle tertile to develop DM. Only 2.4% percent of the patients in the lowest tertile developed DM within 5 years (figure 1).

CONCLUSIONS: A risk prediction model employing commonly available data can be used to guide screening intervals. Based on equal intervals for equal risk, patients in the highest risk category could be rescreened after 8 months, while those in the intermediate and lowest risk categories could be rescreened after 3 and 5 years respectively.

figure x

DEVELOPING A PEER SUPPORT PROGRAM TO TARGET HIGH RISK PRIMARY CARE PATIENTS Karin M. Nelson 1, 2; Jennifer L. Williams1; Marie C. Lutton1; Julie Silverman1, 2; Kamala B Jain1, 2; Tiffanie Fennell1, 2; Matthew R. Augustine1, 2; Kristen E. Gray1, 2; Walter Kopf1, 2; Christopher Vanderwarker1, 2. 1VA Puget Sound Health Care System, Seattle, WA; 2University of Washington, Seattle, WA. (Control ID #2701160)

BACKGROUND: Cardiovascular disease (CVD) is the leading cause of mortality in the US and among Veterans, and CVD risk factors remain sub-optimally controlled. The objective of the current study is to describe a novel strategy using patient-level geographic information systems (GIS) data to identify areas of highest CVD risk and recruit Veterans with well-controlled hypertension living in these areas to become peer health coaches as part of the “Veteran peer Coaches Optimizing and Advancing Cardiac Health (Vet-Coach)” study.

METHODS: We spatially joined Veteran addresses of patients enrolled in VA Puget Sound primary care or women’s clinic from VHA administrative data to the 2010 Census Tract Boundary file to obtain the census tract code. Using these data, we generated spatial distributions of hypertension prevalence by census tract within a 5-mile and 10-mile radius of our clinic. We then selected 60 census tracts (from total of n = 398 in King County, WA) with the highest rates of hypertension: 38–44% of patients with a diagnosis of hypertension (n = 24 census tracts), 45–50% (n = 25 census tracts), and 51–65% (n = 11 census tracts). To identify potential peer health coaches, we provided each primary care clinic team (n = 41 MDs, n = 35 nurses) with a list of their patients from the target areas who had a diagnosis of hypertension (n = 964 patients) and at least one clinic visit during the prior year. During a clinic staff meeting, we asked primary care providers and team nurses to review their list and nominate Veterans they felt might be a good peer health coach, based on the following criteria: successful use of VA services; successful management of hypertension or other chronic conditions; and experience utilizing clinic and community resources. We used data from the 2006–2010 U.S. Census Bureau American Communities Survey to characterize census tract demographics.

RESULTS: Using this method, primary care clinic teams nominated n = 73 Veterans living in the 60 targeted census tracts. We then sent the nominated patients an introductory letter prior to conducting follow-up calls to assess interest in becoming a Vet-Coach, which provides reimbursement based on an hourly rate. We interviewed n = 12 Veterans for the position and selected 5 peer coaches. Compared to other census tracts in King County, the targeted census tracts had higher rates of poverty [% of residents under 200% of federal poverty, 32% vs. 22%, p < 0.001] and a higher percentage of non-white residents [49.4% vs. 33.8%, p < 0.001].

CONCLUSIONS: Using this area-based method, we successfully recruited patients from low-income neighborhoods with high rates of hypertension. Results of the Vet-Coach recruitment will be discussed along with lessons learned, with a specific focus on developing a peer support program targeted to high risk patients that is integrated into primary care clinic work flow.

DEVELOPING A PREDICTIVE MODEL FOR MEDICINE READMISSIONS Michael R. Trautwein 1; Stacy Schwartz2; Karen M. Freund3; Saul N. Weingart1; Geneve Allison1; Phanicharan Sistla1; Lori Lyn Price1; Jana C. Leary1; Alexander Pavoll1. 1Tufts Medical Center, Boston, MA; 2Tufts Medical center, Boston, MA; 3Tufts University School of Medicine, Boston, MA. (Control ID #2706110)

BACKGROUND: 30-day preventable hospital readmissions are a measure of resource utilization and an indicator of health care quality. We compared attributes of patients who were readmitted within 30 days to those who were not in order to develop a predictive model that could inform interventions to prevent unnecessary readmissions.

METHODS: We abstracted electronic medical and billing records for patients admitted to the medicine service at Tufts Medical Center from 9/5/13 - 8/31/16. We included potential predictors of readmission included demographic characteristics, primary language, discharge unit, hospital service, number of medications, zip code (to determine distance to Tufts and as a proxy for income), discharge disposition, physical therapy evaluation, any ICU stay, number of admissions, observations and emergency department encounters in the previous 6 months, primary insurance, warfarin use, alcohol and substance abuse, diabetes, opioid use, Charlson comorbidity score, and day of discharge (weekday vs. weekend). Variables associated with readmission in bivariate analyses (p < 0.05) were used to create a multivariate logistic regression model with backward elimination based on a derivation cohort of discharges from 1/1/14 - 11/30/15. We validated the model using a validation cohort that included discharges 1/1/16 - 8/31/16, using the C-statistic to evaluate the models. We used beta coefficients from the derivation model to develop a scoring system to predict risk of readmission. We compared our model with the LACE predictive model.

RESULTS: The development dataset consisted 7971 admissions; mean age of the patients (SD) was 61 (18), 55% were men, 70% were white, 10% black, 13% Asian and 6% Hispanic. There were 3214 admissions in the validation cohort. In bivariate analysis, factors associated with readmission were Charlson comorbidity score, ICU use, number of medications, admissions, observations and emergency department encounters in the previous 6 months, discharge disposition, opioid use, diabetes, insurance, discharge service, and physical therapy consultation. In the multivariate analysis, 6 variables associated with unplanned readmissions included Charlson comorbidity score, any ICU stay, number of medications, prior admission/observation encounters within 6 months, discharge disposition, and hospital service. Discharges to skilled nursing and long term care facilities had the greatest odds of readmission. The C-statistic for the predictive model from the development dataset was 0.67 and 0.66 for the validation dataset, both higher than c = 0.63 using the LACE model.

CONCLUSIONS: We developed a model with moderate ability to predict unplanned readmissions. In our setting, variables predicting readmission were related to medical complexity, and pointed to the importance of ensuring vulnerable patients’ safe transitions to skilled nursing and longterm care facilities.

DEVELOPMENT AND PRELIMINARY FINDINGS OF DIABETESSISTERSVOICES - AN ONLINE COMMUNITY TO ENGAGE WOMEN WITH DIABETES ABOUT RESEARCH AND HEALTHCARE PRIORITIES Peijin Han1, 5; Wanda Nicholson2; Anna Norton3; Richard Singerman4; Aditi Sundaresan2; Wendy L. Bennett 1. 1Johns Hopkins School of Medicine, Baltimore, MD; 2University of North Carolina, Chapel Hill, NC; 3DiabetesSisters, Bolingbrook, IL; 4TrustNetMD, San Diego, CA; 5Johns Hopkins Bloomberg School of Public Health, Baltimore, MD. (Control ID #2705252)

BACKGROUND: Women with or at high risk of diabetes are an important clinical group with unique health concerns across the life course- from preconception health and pregnancy to middle age and the menopause transition. Engaging women with diabetes is critical to efforts to inform a patient-centered research agenda to improve health outcomes. Our objective was create an online community for women with any type of diabetes (type 1, type 2, gestational or pre-diabetes) to engage about research priorities to address their health care needs and improve diabetes management, use in-person and web-based strategies for recruitment to the on-line site and measure trends in engagement and re-engagement of site participants.

METHODS: In phase 1, we partnered with DiabetesSisters, a national advocacy organization for women with diabetes, and a diverse stakeholder advisory board to develop the online community, termed DiabetesSistersVoices (aka DSVoices). In phase 2, we employed several recruitment strategies, including social media (e.g. email, Twitter, Facebook), web-based newsletters and printed media to engage women with diabetes. Eligibility included women living with or at risk for diabetes who were age 18 years and older and currently living in the U.S. Participants completed an on-line consent process and survey questions about socio-demographic characteristics, self-reported health status, technology use and social support. Once enrolled, trends in participants’ activities, including posting questions, sharing experiences of living with diabetes and searching for educational resources are monitored and tracked by an on-line moderator.

RESULTS: To date, 94 women have enrolled (83% type 1, 86% Caucasian, median age = 46), from 31 states in the U.S. Most learned about DSVoices from Facebook, Twitter, friends and relatives. More than 90% of participants regularly use social networking sites (mainly Facebook and Twitter) and 80% of them use every day. The most popular discussion topics are: “type 1 and lada (latent autoimmune diabetes in adults)”, “type 2”, “popular websites”, “healthy living” and “physical activity and exercise”. Priority topics for research and healthcare have focused on (1) medication management (2) eating disorders (3) meal planning (4) mental health (5) exercise.

CONCLUSIONS: This formative data describes the refinement and formal testing of an on-line community for women with diabetes to set research priorities. Additional engagement strategies to enroll minority or low-income women and those in rural areas are underway. Our efforts can contribute to the development of novel engagement methods and the identification of patient-centered priorities for research and care for women with or at risk for diabetes.

DEVELOPMENT AND VALIDATION OF AN ELECTRONIC HEALTH RECORD MODEL FOR PREDICTING 30-DAY READMISSIONS IN ACUTE MYOCARDIAL INFARCTION: THE AMI READMITS SCORE Oanh K. Nguyen 1, 1; Anil N. Makam1, 1; Christopher Clark2; Song Zhang1; Sandeep R. Das1; Ethan Halm1, 1. 1UT Southwestern Medical Center, Dallas, TX; 2Parkland Health & Hospital System, Dallas, TX. (Control ID #2694487)

BACKGROUND: Readmissions after hospitalization for acute myocardial infarction (AMI) are common, but the few available risk prediction models have poor predictive ability and are not readily usable in real-time. We sought to develop and validate an AMI readmission risk prediction model from electronic health record (EHR) data available on the first day of hospitalization, and to compare model performance to the Centers for Medicare and Medicaid Services (CMS) AMI model and a validated multi-condition EHR model.

METHODS: EHR data from AMI readmissions from 6 diverse hospitals in north Texas from 2009–2010 were used to derive a model predicting all-cause non-elective 30-day readmissions to any of 75 hospitals in the region, which was then validated using five-fold cross-validation.

RESULTS: Of 826 consecutive index AMI admissions, 13% were followed by a 30-day readmission. The AMI READMITS score included seven predictors, all ascertainable within the first 24 hours of hospitalization (Table 1A). The AMI READMITS score was strongly associated with 30-day readmission in our cross-validation cohort: ≤13 points = extremely low risk (bottom quintile, mean predicted risk 3%); 14–15 points = low risk (4th quintile, predicted risk 7%); 16–17 points = moderate risk (3rd quintile, predicted risk 11%); 18–19 points = high risk (2nd quintile, predicted risk 16%); and ≥20 points = extremely high risk (top quintile, predicted risk 35%). The READMITS score had good discrimination with comparable performance to the CMS model in our cohort; it had improved discrimination, reclassification, and calibration compared to a multi-condition EHR model (Table 1B).

CONCLUSIONS: The AMI READMITS score accurately stratifies patients hospitalized with AMI into groups at varying risk of 30-day readmission. Unlike claims-based models which require data not available until after discharge, READMITS is parsimonious, easy to implement, and leverages actionable real-time data available from the EHR within the first 24 hours of hospitalization to enable early prospective identification of high-risk AMI patients for targeted readmissions reduction interventions.

figure y


BACKGROUND: COPD is a leading cause of morbidity and mortality, and is underdiagnosed in primary care settings. Though benefits of COPD screening in asymptomatic adults with mild disease are unproven, COPD therapies reduce exacerbations and improve health status in symptomatic individuals with more severe COPD. Feasible screening approaches to identify undiagnosed cases of more severe COPD in primary care settings may reduce the clinical burden of COPD. We created a screening tool to identify patients with moderate-to-severe (mod-sev) COPD and assessed its performance in a diverse sample of adult primary care patients at a large safety-net hospital. The tool consisted of a tablet-based questionnaire linked to an electronic peak expiratory flow (PEF) meter.

METHODS: We used data from the Framingham Heart Study (FHS) 8th Offspring Exam to derive 2 multiple logistic regression models for predicting mod-sev COPD in men and women. Candidate predictor variables included demographics, smoking status, selected symptoms from the ATS-DLD-78 Respiratory Disease Survey, and PEF. We used backward selection to identify variables that independently predicted mod-sev COPD at p < 0.2. We then loaded questions to capture final predictors onto a tablet computer which also contained a brief video on how to use a PEF meter. The system was refined using input from user-consultants. Its performance was then validated in 188 clinic patients at Boston Medical Center (BMC) who had a clinical diagnosis of COPD or no known lung disease. Next, we prospectively assessed the system in a ‘testing’ cohort of primary care patients without known COPD. The system classified predicted risk of mod-sev COPD as low (0–4%), medium (5–19%) or high (≥20%). In each study cohort COPD status was established by spirometry, and we calculated the system’s ability to discriminate between mod-sev and no or mild COPD (c-statistic). In the testing cohort we determined system sensitivity and specificity at 2 cut points.

RESULTS: The final model for men and for women both had 8 variables: 3 demographic, PEF, smoking status, and 3 relating to respiratory history. The table describes the 3 study cohorts and shows the system’s c-statistic in each. In the ‘testing’ cohort, system sensitivity and specificity were 0.94 and 0.36 when a positive test was defined as a predicted risk of ≥ 5%, and 0.81 and 0.55 when set at ≥ 20%.

CONCLUSIONS: Our system to identify undiagnosed mod-sev COPD showed good discrimination, sensitivity and specificity in a diverse group of primary care patients, particularly at a positivity criterion of ≥ 20%. Screen-positive patients would then require spirometry for confirmation.

Derivation (FHS) N = 2,574 Validation (BMC) N = 190 Testing (BMC) N = 399
Age (yrs), mean [SD] 66 [8.8] 59 [7.2] 56 [9.4]
Female (%) 55 54 58
Smoking Status (%)
 Current 7 49 29
 Former 52 32 23
Race/Ethnicity (%)
 Black 0 52 53
 Non-Hispanic White 100 40 28
 Mod-Severe COPD (%) 17 36 26
Model C-Statistic
 Men 0.92 0.87 0.73
 Women 0.91 0.83 0.67

DEVELOPMENT OF AN INTERNAL MEDICINE “BOOT CAMP” FOR MEDICAL STUDENTS David C. Chu 1; Rachel S. Casas1; Laura D. Hallett3; Frank Schembri4; Ryan Chippendale2. 1Boston Medical Center, Boston, MA; 2boston university, Boston, MA; 3University of Massachusetts, Worcester, MA; 4Boston Medical Center, Bostn, MA. (Control ID #2689273)

BACKGROUND: Internal medicine “boot camps” are an emerging strategy to ease the challenging transition from fourth year of medical school to internship, but prioritized topics vary by institution. We aimed to determine the most important content for a boot camp curriculum though a needs assessment at the Boston University School of Medicine (BUSM).

METHODS: Fourth year medical students applying to internal medicine residency programs (N = 47) and first year internal medicine interns (N = 40) at BUSM were anonymously, electronically surveyed about managing 27 clinical scenarios, skills, and procedures. We chose these topics based upon existing medicine boot camp curricula and internal discussion. Responses were ranked on 4 point scales of importance or interest. Participants were also asked open-ended questions about their concerns for intern year and the most important topics for inclusion in a boot camp elective.

RESULTS: Survey response rates were 51% (N = 24) for medical students and 58% (N = 23) for interns. Of clinical scenarios, medical students prioritized chest pain (N = 21, 88%), shock (N = 21, 88%), and severe electrolyte disturbances (N = 18, 75%) as ‘very important’ for inclusion. Current interns ranked chest pain (N = 19, 83%), acute hypoxemia (N = 15, 65%) and altered mental status (N = 15, 65%) as ‘very important’ for inclusion. Both medical students and interns rated time management and ECG interpretation among the most important clinical skills. Medical students were ‘very interested’ in almost all procedures, with thoracentesis, paracentesis, and arterial blood gases ranking highest (N = 17 for each, 71%). Interns expressed lower interest in procedures than medical students, with only arterial blood gases (N = 12, 52%) receiving more than three ‘very important’ responses. In open-ended responses, medical students and interns were most concerned about managing acutely decompensating patients (N = 7 and N = 5 respectively) and time management (N = 6 and N = 3 respectively). Additionally, medical students stated concerns about procedural skills (N = 3).

CONCLUSIONS: Medical students and interns have similar priorities in acute clinical scenarios and skills for inclusion in an internal medicine boot camp. However, medical students expressed more interest and concern about learning procedures compared to interns. This needs assessment will guide development of a boot camp curriculum at BUSM, and can inform similar electives at other institutions.

DEVELOPMENT OF PREDICTIVE SCORING SYSTEM OF FALLS IN ADULT INPATIENTS USING DEGREE OF BED-RIDDEN AND COGNITIVE FUNCTION Masaki Tago 1; Naoko E. Furukawa1; Yoshimasa Oda2, 1; Shu-ichi Yamashita1. 1Saga University Hospital, Saga, Japan; 2Yuai - Kai Foundation & Oda Hospital, Kashima, Japan. (Control ID #2690741)

BACKGROUND: Falls among elderly can degrade their quality of life and worsen prognosis. The incidence rate of falls in inpatients is higher than that of community-dwelling persons aged ≥65 years. Various factors, such as previous falls, lower-extremity weakness, age, female gender, had been reported as risk factors of falls. Some predictive formulas of falls have been developed; however, they used complicated factors, which make them hard-to-use. In the present study, we developed and validated a new formula predictive model of risk of falls of about our adult inpatients using the scales of the degree of bed-ridden and decline of cognitive function in daily living before admission. Both scales are widely used in the Japan’s public certification of long-term care need.

METHODS: We retrospectively analyzed the adult patients admitted to Yuai-kai Foundation Oda Hospital, Japan, from April 2012 to January 2015. The data (including age, sex, activity of daily living, the degree of bed-ridden and decline of cognitive function in daily living before admission, past medical history, medications, presence or absence of past history of falls and the route of admission) were derived from the charts. Multivariable analysis was performed with the variables gathered on admission, which had significant difference between patients with or without falls during the hospital stay, in the univariate analysis. We developed a new predictive formula with the independent variables selected by the multivariate analysis. The model was validated.

RESULTS: There were 371 falls of all 8031 admitted adult patients during the period of present survey. The median age of all patients was 77 years (20–104), 49% were men, and the median hospital stays was 9 days (1–189). In the univariate analysis, the proportion of patients with the age older than 75, emergency admission, referral forms, the use of hypnotic medications, hospital stay longer than 9 days, visual impairment, undergoing surgical operation, undergoing rehabilitation, sequels of cerebral infarction disturbing everyday life, and history of previous falls occupied higher percentages in fallen patients. There were also significant differences in the degree of bedridden, cognitive decline, and the necessity of assistance in daily living between patients with or without falls. The new prognostic formula is shown bellow: y = (older than 75 years) × 6 + (male) × 4 + (emergency admission) × 5 + (previous fall) × 3 + (the degree of bedridden = J) × 12 + (the degree of bedridden = A) × 18 + (the degree of bedridden = B) × 20 + (the degree of bedridden = C) × 17 + (the degree of cognitive decline = 2) × 4 (0 ≤ y ≤ 42, be apply to each condition = 1). The test of goodness of fit by ROC curve showed AUC of 0.789 (95%CI: 0.770 - 0.808, p < 0.001).

CONCLUSIONS: We developed a new and accurate predictive scoring system of falls in adult inpatients using degree of bed-ridden and cognitive function, which are much more easily gathered than those used in the previous reported models.

DIABETES AMONG WOMEN VETERANS FOUR DECADES AFTER WAR: THE HEALTHVIEWS STUDY Eric M. Schmidt1; Eileen M. Stock2; Tracey Serpi2; Yasmin Cypel3; Kathryn M. Magruder4; Amy M. Kilbourne5; Suad El Burai Felix2; Avron Spiro6; Rachel Kimerling7; Beth Cohen8; Susan M. Frayne 1. 1VA Palo Alto Health Care System/Stanford, Menlo Park, CA; 2VA Perry Point, Perry Point, MD; 3VA Central Office, Washington, DC; 4VA Charleston, Charleston, SC; 5VA Ann Arbor, Ann Arbor, MI; 6VA Boston, Boston, MA; 7VA Palo Alto, Palo Alto, CA; 8VA San Francisco, San Francisco, CA. (Control ID #2702859)

BACKGROUND: Military service can affect health through injuries, exposures, trauma, and physical demands. Less is known about the effects of war zone service on later-life risk for chronic diseases such as diabetes, particularly in women. In a national, population-based sample of Vietnam-era women Veterans, we examined whether incidence of diabetes following wartime service was greater among those who served in Vietnam compared to those who served in a non-combat location.

METHODS: In HealthVIEWS, a large epidemiologic study of Vietnam-era (1965–1973) women Veterans, those who were alive and locatable in 2011 were invited to complete a survey and telephone interview to characterize their health status (response rate 67%); sampling was stratified based on wartime location of service: in Vietnam (VN), near Vietnam (NV), or in the U.S. (US). An extended Cox regression model tested differences, by wartime location, in time from the start of Vietnam-era wartime service to self-reported diabetes onset. A time-varying effect of Vietnam service was entered to meet model assumptions before adjusting for demographics and military service factors.

RESULTS: Of 4,503 women in the analytic cohort, 17.7% developed diabetes; for them, mean time to diabetes onset differed for VN vs NV vs US (32.9 vs 33.2 vs 31.2 years; p < 0.01). Following enlistment into military service, those who served in VN (vs in US) were at three-fold lower relative risk of diabetes, and this difference persisted after adjustment (HR = 0.33, p < 0.01 at T0). The time-varying effect of Vietnam service was also significant, indicating diabetes risk accumulated faster in the VN group (vs US); diabetes risk was essentially the same in the VN vs US groups after 30 years (HR = 0.99, p = ns). The NV (vs US) group did not have excess in diabetes risk. Older age, non-white race/ethnicity, service in the Army or Marine Corps, lower military rank, and enlisting at the start of heightened periods of combat were associated with increased risk for diabetes.

CONCLUSIONS: War zone service was associated with lower risk for diabetes early in wartime military service but faster subsequent accumulation of diabetes risk, leading the VN and US groups to have nearly identical diabetes risk by three decades after military service. This more rapid accumulation of risk after war zone service, despite superior initial health, could potentially result from factors such as weight gain or mental health conditions, which might manifest in unique ways among women. Attention to preventive medicine and healthy lifestyle during military service among women currently in the armed forces may help to avert or delay diabetes onset later in life.

DIABETIC KETOACIDOSIS RISK AFTER INITIATING AN SGLT2 (SODIUM-GLUCOSE COTRANSPORTER 2) INHIBITOR: A POPULATION-BASED COHORT STUDY Michael Fralick; Sebastian Schneeweiss; Elisabetta Patorno. Brigham and Women’s Hospital, Boston, MA. (Control ID #2697981)

BACKGROUND: Recent case reports to the FDA suggest sodium-glucose cotransporter 2 (SGLT2) inhibitors might increase a patient’s risk of diabetic ketoacidosis (DKA). The objective of this study was to assess and quantify this risk.

METHODS: We conducted a population-based new initiator cohort study using the Truven MarketScan database. Patients who newly initiated either an SGLT2 inhibitor or a dipeptidyl peptidase-4 (DPP4) inhibitor, a medication not associated with DKA, between April 1, 2013 and December 31st, 2014 were included. We excluded patients with type 1 diabetes, past DKA, or end-stage renal disease. The primary outcome was hospitalization for DKA within 180 days of initiating an SGLT2. We matched new initiators of SGLT2 inhibitors with new initiators of DPP4 inhibitors using 1:1 propensity score (PS) matching. The PS was calculated through a logistic regression model including covariates related to demographics, diabetes severity, comorbid conditions, and healthcare utilization. The primary time-to-event analysis used Cox proportional hazard model to examine the incidence rate and hazard ratio of DKA.

RESULTS: We identified 61,708 patients who newly initiated an SGLT2 inhibitor and 94,285 who newly initiated a DPP4 inhibitor. After PS-matching, our study included 45,789 new users of SGLT2 inhibitors and 45,789 new users of a DPP4 inhibitors. Differences in baseline demographics, comorbid conditions, diabetes severity, and healthcare utilization were well balanced amongst these two groups. Of the patients included, half were male, the average age was 60, 65% had hypertension, 9% had coronary artery disease, 57% were prescribed metformin, 19% were prescribed insulin, and most were being managed by a primary care physician. There were 58 DKA events (4.23 events per 1,000 person-years) among initiators of an SGLT2 inhibitor compared to 29 DKA events (2.01 per 1,000 person-years) among initiators of DPP4 inhibitors. The unmatched/unadjusted risk of DKA within 180 days of initiating an SGLT2 inhibitor was 1.9 times (HR 1.9, 95% CI 1.4 to 2.6) greater than the risk for those initiating a DPP4 inhibitor and the risk heightened after PS matching (HR 2.1, 95% CI 1.3 to 3.3). The risk was further heightened within 60 days of initiating an SGLT2 inhibitor (PS matched HR 3.0, 95% CI 1.6 to 5.5) and the increased risk of DKA was not specific to the elderly (age < 65, PS matched HR 2.3, 95% CI 1.4 to 3.6).

CONCLUSIONS: SGLT2 inhibitors were associated with an increased risk of DKA. Physicians should counsel their patients accordingly and consider ordering bloodwork when patients present with symptoms suggestive of DKA (e.g., nausea, vomiting, polyuria, etc.).

DIET AND EXERCISE DO WORK FOR WEIGHT LOSS: THE SUCCESS OF AN INTERDISCIPLINARY OBESITY CLINIC IN A PRIMARY CARE SETTING Abigail Lawson; Christina Fahey; Garrett Oberst; Dylan Woolum; David Rudy; Stephanie A. Rose. University of Kentucky, Lexington, KY. (Control ID #2700120)

BACKGROUND: Despite efforts, obesity continues to be a problem in the United States. Since 2011 Medicare has recognized obesity as a disease. This change has purported to remove the stigma behind obesity and has encouraged its treatment in the medical setting as opposed to commercial settings alone. The aim of our study is to assess relevant weight loss in an outpatient primary care-based weight loss program. We hypothesize that relevant weight loss, defined as loss of 5 or 10% of body weight or loss of ≥3 kg at 6 months, can be achieved in the primary care setting.

METHODS: Patients with a BMI ≥30 kg/m2 or a BMI ≥25 with comorbidities related to obesity were enrolled in a weight loss program located at a large university-based Internal Medicine clinic. Most weight loss was achieved using diet and behavioral motivation, with a few patients also trying weight loss medications. Visits were covered by insurance similar to a primary care physician visit. Patients were seen by an Obesity Medicine-certified physician each visit and sometimes by a dietitian. Visits lasted 1 hour the first visit and ½ hour on subsequent visits. Information obtained for patients included diet history, family weight history, dietary assessment, 24-hour food recall. Visits typically occurred monthly. Demographic information, weight, waist/hip/neck circumferences, lab values, medications, and medical history were obtained on each patient. Mean weight loss, number of patients losing ≥ 5 and ≥10% of body weight, and number of patients who lost ≥3 kg at 6 months were measured. Comparisons were analyzed using t-tests that compared differences among male and female patients for continuous variable characteristics. Only those patients who consented to the study were analyzed.

RESULTS: 115 patients consented to be in the study. Patients were a mean age of 47.5 years, 76% female, and a mean BMI of 41 kg/m2. Patients lost an average of 10lbs (SE = 1.6), with men (M) losing slightly more weight than women (F) (14.2 vs. 8.8lbs). 21.7% of patients (M = 5, F = 20) lost ≥5% of their initial body weight, 8.7% of patients (M = 2, F = 8) lost ≥10% of their initial body weight, and 56.5% of patients (M = 17, F = 48) lost at least 3 kg at 6 months of program participation. Patients lost an average of 2.9lbs after 1 month, 5.4lbs after 3 months, and 7.2lbs after 6 months. More weight lost was positively correlated with more clinic visits, with patients who attended ≥6 visits losing an average of 17.3lbs.

CONCLUSIONS: This study demonstrates that behavioral modification and dietary change can promote relevant weight loss, and that this success can be achieved in the outpatient primary care setting. Patients continued to lose weight the longer they participated in the program, unlike many programs where weight plateaued or was regained over time. Future goals include the study of medication changes, changes in disease states and labs, and methods to make this program easier for primary care physicians to deliver.

DIFFERENCES IN EMERGENCY DEPARTMENT CARE FOR PATIENTS WITH LIMITED ENGLISH PROFICIENCY lucy schulson 1; Victor Novak1; Peter Smulowitz1; Bruce E. Landon2. 1Beth Israel Deaconess Medical Center, Boston, MA; 2Harvard Medical School, Boston, MA. (Control ID #2706552)

BACKGROUND: The United States is becoming more culturally and linguistically diverse. Research has focused on disparities in care for such patients, however little is known about how limited English proficiency (LEP) impacts care in the emergency department (ED). LEP patients may be particularly vulnerable in the ED because of its faster paced and higher acuity nature, which might lead to differences in the care received when there are language barriers.

METHODS: We used administrative data from a single academic ED to study all patients 18 years or older who presented to the ED between September 1, 2013 and September 1, 2015. We excluded patients presenting with a psychiatric or substance related complaint, who died in the ED, who were not charged for their visit, who left without being seen, or who refused treatment. Patients with > 5 ED visits in the calendar year were also excluded from analysis. The primary exposure of interest was whether the patient had identified a language other than English as their preferred language during registration. Our primary outcomes of interest were the number of diagnostic studies (laboratory studies, imaging tests, and cardiac tests) performed during the patient’s initial visit, whether the patient was admitted to the hospital, and, for those discharged from the ED, the rate of 72-hour unscheduled return visits. We estimated multivariable logistic regression or Poisson or negative binomial modeling as appropriate controlling for age, sex, insurance type, triage acuity score, time of presentation, and discharge diagnosis category.

RESULTS: We studied ED visits for 54,417 EP and 5415 LEP individuals. LEP patients were older (50.3% 60 and older vs 33.2% for EP, p < .0001), more likely to be female (57.6% vs. 54.6%, p < .0001), less educated (18% vs. 0.9% with 8th grade or less, p < .0001), and more likely to have public insurance (29.6% vs. 11.7%, p < .0001). In unadjusted analysis, LEP patients were more likely to have a laboratory test (80.6% vs. 75.4%, p < .0001), plain x-rays (71.8% vs. 64.7% p < .0001), and CT scans (5.9% vs. 5.2% p = .0417), but less likely to undergo MRI testing (0.8% vs. 1.1% p = .03). LEP patients were also more likely to have an EKG (52.4% vs. 45.1% p < .0001) or a stress test (4.1% vs. 2.9% p < .0001). Finally, LEP patients were also more likely to be admitted, (38.8% vs. 34.2%, p < .0001) and to have an unscheduled 72-hour return visit (1.8% vs. 1.6% p < .0001). Preliminarily, in adjusted analyses the findings were consistent.

CONCLUSIONS: LEP patients receive more care in the emergency department than EP patients. They also are more likely to be admitted and to return to the ED within 72 hours of discharge. These disparities may be due to communication challenges, and suggest that interventions be targeted to improve the value of care they receive.

DIFFERENCES IN GOALS DURING RESIDENCY TRAINING BETWEEN THE UNITED STATES AND JAPAN: TIME TO ADDRESS GAPS BETWEEN COMPETENCIES AND TRAINEE SELF-IDENTIFIED GOALS Hirotaka Kato3; Alfred Burger4; Ken Emoto5; Reiko Sakama 2; Yuki Uehara2; Ankur Segon3; Jenny J. Lin1. 1Icahn School of Medicine at Mount Sinai, New York, NY; 2Juntendo University School of Medicine, Tokyo, Japan; 3Medical College of Wisconsin, Milwaukee, WI; 4Mount Sinai Beth Israel, Icahn School of Medicine at Mount Sinai, New York, NY; 5Aso Iizuka Hospital, Iizuka, Japan. (Control ID #2698472)

BACKGROUND: Medical educators worldwide aspire to transition from time based training to a competency based medical education model. The degree of alignment between competencies and residents’ self-identified training goals, as well as variations in different countries have not been studied. This study sought to examine (1) how residents’ goals match proposed universal competencies, (2) differences in residents’ self-identified training goals between the United States (US) and Japan, and (3) prevalence of fatigue among residents.

METHODS: An online survey targeting postgraduate trainees in both the US and Japan was conducted between January and June 2016. The authors recruited two internal medicine residency programs at an urban university program and an affiliated medical center in New York City as well as one university based and four non-university based programs in Japan. The survey questions included demographics, milestones across eight competency areas proposed as universal, and eleven Iowa Fatigue Scale (IFS) questions. A 4-point scale was used (not important, a little bit important, somewhat important, very important) and responses were dichotomized into 1 (very important) and 0 (others) to compute statistical significance with Chi-Square or Fisher’s exact test. Presence of fatigue and severe fatigue were defined as IFS >30 and IFS > 40, respectively. Chronbach alfa (0.79) confirmed internal consistency for IFS.

RESULTS: 165 out of 393 potential respondents (42%) completed the survey. 58% were US trainees. The proportions of “very important” in systems-based practice milestones were generally low in both countries (<70%). Significant differences were observed across all competency areas due to lower rating by postgraduate year (PGY) 1 respondents in Japan. Among the PGY 2 and 3 respondents, the statistically significant items (p ≤ .05) included (a) Qualified handoff (US 86% versus Japan 59%), (b) Supervision for team members (80% vs. 41%), (c) Symptom relief (88% vs. 50%), (d) Evidence-based practice (94% vs. 69%), (e) Education for team members (71% vs. 48%), (f) Continuous learning (94% vs. 48%), (g) Communication with health professionals (92% vs. 71%), (h) Difficult conversations (96% vs. 74%), (i) Administrative responsibilities (36% vs. 68%), (j) Compassion and respect (100% vs. 84%), and (k) Sensitivity to diversity (91% vs. 58%). The prevalence of fatigue (42% vs. 81%), and severe fatigue (4% vs. 19%) was significantly higher in Japan (p < .01).

CONCLUSIONS: There were significant differences in training goals between resident physicians in the US and Japan. US trainees were more likely to see competencies as their own goals during residency compared to their Japanese counterparts. GME training should ideally be designed so that resident physicians can align their goals with milestones and competencies.

DISCHARGING THE DATA: UNDERSTANDING THE STRUCTURE OF THE HOSPITAL DISCHARGE SUMMARY IN DELIVERING PATIENT CARE Nancy T. Skehan; Bruce Barton; Sheri Keitz. University of Massachusetts Medical School, Worcester, MA. (Control ID #2700517)

BACKGROUND: Major clinical errors, increased emergency visits, and hospital readmissions can all result from poor communication between providers as patients transition from inpatient to ambulatory settings. Lapses can include medication errors, pending test results, or lack of critical information at discharge. The hospital discharge summary has been identified as a vital tool for relaying information at discharge and its content has been well defined by numerous national organizations. However, there is no standard structure for the discharge summary. We conducted a survey of residents, hospitalists and primary care providers (PCPs) in Internal Medicine, Pediatrics and Family Medicine to determine if they identify and share a definition of an accepted structure for the discharge summary, as compared to a History and Physical (H&P). We also assessed prior discharge summary training and providers’ comfort in discharge communication. The goal was to conduct a needs assessment of existing knowledge and attitudes about discharge summaries, ultimately to develop systems and educational interventions.

METHODS: A single-site cross-sectional survey of the target population was conducted at a tertiary care academic center and its community affiliates from April-May 2015. Providers were asked to identify the expected location of key clinical components in the discharge summary, as well as components of an H&P. Study data were collected and managed using REDCap electronic data capture tools. The statistical analyses generated the frequencies and percentages for each location by the specified subgroup. All analyses were conducted using SAS Version 9.3 (SAS Institute, Cary, NC).

RESULTS: 415 participants were surveyed, with 220 respondents. Respondents identified the location of the H&P elements with 80-96% concordance, compared to 42–82% in discharge summaries. In assessing knowledge, 72% of respondents felt they had not received formal training. 50% reported observing an adverse patient outcome with 77% citing difficulty in locating information as a primary cause. A subgroup analysis of hospitalists and PCPs revealed that hospitalists expect the final diagnoses at the beginning, where PCPs expect it to be at the end (p < 0.001).

CONCLUSIONS: Compared to the H&P, physicians do not identify a common sequence for the key components in discharge summaries. Moreover, hospitalists who generate the discharge summary do not have the same definition of its structure compared to PCPs, who receive the document. This may lead to difficulty in locating information within the document, and may contribute to medical error. These data show a need for standardized structure for the discharge summary with an accompanying educational and systems interventions to decrease the likelihood of major clinical errors.

DISPARITIES IN QUALITY OF PRIMARY CARE BY RESIDENTS AND STAFF PHYSICIANS: IS THERE A CONFLICT BETWEEN TRAINING NEEDS AND EQUITY IN CARE? Utibe R. Essien 1, 2; Jonathan R. Abraham1; Alaka Ray1; Wei He1; Yuchiao Chang2; Daniel E. Singer1; Steven J. Atlas1. 1Massachusetts General Hospital, Boston, MA; 2Harvard Medical School, Boston, MA. (Control ID #2694122)

BACKGROUND: Ambulatory education for internal medicine (IM) residents is vital to training. As a majority of training is spent on the inpatient wards, concern remains about the quality of outpatient care residents provide. Prior studies compared patient outcomes of resident and staff primary care physicians (PCPs) but few examined quality of care following mandated increases in ambulatory time. We compared outcomes of patients seen by resident and staff PCPs controlling for patient characteristics and time in network.

METHODS: We compared cohorts of patients receiving care from IM residents and staff PCPs at 16 primary care clinics (PCCs) affiliated with Massachusetts General Hospital from 2005–2015. A validated algorithm was used to attribute patients to resident and staff PCPs. We assessed diabetes and CAD outcome measures, cancer screening rates and resource utilization in 2015. We report outcomes adjusted for age, gender, race, language, insurance, education, area-based median household income, clinic site, Charlson score and physician-defined complexity using generalized linear regression models. To further control for confounding between resident and staff patients including time in network, we compared outcomes from 1:1 propensity score matched subgroups.

RESULTS: Among all PCC patients, 10,538 (7.5%) were attributed to resident and 130,706 to staff PCPs. Compared to staff patients, resident patients were more likely to be male, African-American, receive Medicaid or be uninsured, live in low-income areas and have higher comorbidity scores (all p < 0.05). Resident patients were less likely to speak English and completed less education. Compared to staff patients, resident patients had lower chronic disease outcomes, cancer screening rates and higher rates of admission and ED visits (Table 1). The differences in all outcomes remained significant in propensity score-matched subgroups (data not shown).

CONCLUSIONS: In this single primary care network, major differences in sociodemographics and comorbidities exist among patients seen by resident and staff PCPs. After controlling for confounding using multivariable modeling and propensity matching, resident patients still had poorer chronic disease, cancer screening and resource utilization outcomes. Unmeasured confounders may partially account for the outcomes observed. Novel approaches to engage trainees in team-based practice and population health management may be needed to prevent disparities in quality of care in academic PCCs while meeting resident educational needs.

Quality of Care between Resident and Staff PCP Patients

Quality of Care and Resource Utilization Measures Resident PCP Patients
n = 10,538
Staff PCP Patients
n = 130,706
Adjusted Relative Risk
(95% CI)
Adjusted p-value
Coronary Artery Disease (CAD) - LDL Level <100 mg/dL 49.6% 61.8% 0.82
(0.74, 0.89)
Diabetes - HbA1c Level < 9% 74.2% 81.1% 0.95
(0.91, 0.98)
Breast Cancer Screening 70.4% 83.6% 0.89
(0.87, 0.92)
Cervical Cancer Screening 69.6% 82.8% 0.92
(0.89, 0.94)
Colorectal Cancer Screening 65.9% 77.8% 0.93
(0.90, 0.95)
Admissions (per 100 pts) 15.5 6.5 1.75*
(1.59, 1.93)
Emergency Department (ED) Visits (per 100 pts) 25.8 11.9 1.47*
(1.35, 1.60)

*Risk Ratio

DO DIABETES GROUP VISITS IMPROVE HEALTH OUTCOMES AMONG PATIENTS IN COMMUNITY HEALTH CENTERS? Arshiya A. Baig 1; Erin M. Staab1; Amanda Benitez3; Sarah P. Hermans1; Yue Gao1; Sandra Ham1; Amanda Campbell4; Cynthia T. Schaefer2; Michael T. Quinn1. 1University of Chicago, Chicago, IL; 2University of Evansville, Evansville, IN; 3Enlace Chicago, Chicago, IL; 4Midwest Clinicians’ Network, East Lansing, MI. (Control ID #2705087)

BACKGROUND: Diabetes group visits, shared appointments in which patients receive self-management education in a group setting and have a medical visit, are an innovative way to deliver diabetes care and have been shown to improve clinical outcomes. Community health centers (CHCs) play a vital role in providing care to patients with diabetes. No studies have systematically implemented diabetes group visits in a network of U.S. CHCs.

METHODS: We designed a group visit intervention in partnership with Midwest Clinicians’ Network. We trained staff from six CHCs in five states to conduct six monthly group visits with 8–10 adult patients with uncontrolled diabetes (glycosylated hemoglobin, A1C, ≥ 8%). Primary outcome was change in A1C from baseline to 6- and 12-month follow-up. Secondary outcomes were changes in diabetes self-care (diet, exercise, foot care, blood glucose testing), diabetes-related quality of life (dissatisfaction with diabetes control, diabetes-related social worry), and diabetes distress (emotional burden, physician distress, regimen distress, interpersonal distress). Analysis of secondary clinical outcomes, including LDL, blood pressure, and weight, is in progress, and we are currently collecting data on matched control patients at each site.

RESULTS: Fifty-one patients were enrolled across the six sites (mean age 55 ± 12 years, 67% female, 38% African American, 38% white, 22% Latino, 9% American Indian/Native American). Average baseline A1C was 10.2% ± 1.7%. Patients attended an average of 3.5 ± 1.9 group sessions. Compared to baseline, A1C was significantly lower at 6 months (8.8% ± 2.3%, p = 0.0002) and 12 months (9.0% ± 2.0%, p < 0.0001). Patients improved in two self-care areas from baseline to 6 months: days per week following a healthful eating plan (3.3 ± 2.4 vs. 4.8 ± 1.4, p = 0.023) and testing blood sugar (4.8 ± 2.4 vs. 6.1 ± 1.2, p = 0.001). Dissatisfaction with diabetes control decreased significantly (2.3 ± 0.8 vs. 1.8 ± 0.6 [1 = least to 5 = most problematic], p = 0.029) at 6 months. Participants experienced significantly less emotional burden (2.6 ± 1.4 vs. 2.0 ± 1.4 [1 = least to 6 = most distressed], p = 0.049), regimen distress (3.0 ± 1.5 vs. 2.0 ± 1.2, p = 0.014), and overall diabetes distress (2.3 ± 1.2 vs. 1.8 ± 1.1, p = 0.048). There were no significant changes in foot care, exercise, physician or interpersonal distress, overall diabetes-related quality of life, or diabetes-related social worry. The intervention was well-received by patients: 93% gained confidence in their ability to manage diabetes, 93% agreed the content was relevant to their lives, and 90% said attending group visits was not a burden.

CONCLUSIONS: Among a sample of adults with uncontrolled diabetes who receive care in CHCs, diabetes group visits improved glycemic control, diabetes self-care, diabetes-related quality of life, and diabetes distress. Future research should assess the cost of group visits and facilitators and barriers to group visit implementation in CHCs.

DO HISPANIC-SERVING HOSPITALS HAVE WORSE PATIENT OUTCOMES? CASE OF ACUTE MYOCARDIAL INFARCTION INPATIENT MORTALITY Lenny Lopez 5; Michael Paasche-Orlow3; Nancy R. Kressin4; Eun Ji Kim2; Meng-Yun Lin1; Jennifer E. Rosen6; Amresh D. Hanchate3. 1Boston Meidcal Center, Boston, MA; 2Boston University, Boston, MA; 3Boston University School of Medicine, Boston, MA; 4Dept of Veterans Affairs and Boston University, West Roxbury, MA; 5University of California San Francisco, San Francisco, CA; 6MedStar Washington Hospital Center, Washington DC, DC. (Control ID #2707004)

BACKGROUND: Hospital quality is associated with patient outcomes. Examination of hospital effects on disparities in patient outcomes has been limited to black vs. white comparisons. Our objective was to extend such analyses to include ethnicity and to examine the role of hospitals’ ethnic composition and other characteristics on patient outcomes. We estimated inpatient risk adjusted mortality for acute myocardial infarction (AMI) discharges by race/ethnicity and examined variations in mortality by patient and hospital characteristics at high proportion Hispanic-serving hospitals compared to other hospitals.

METHODS: We used patient level data from state inpatient discharge data (2010–11) from 14 states (AZ, CA, CO, FL, IL, MA, MD, NV, NJ, NY, OR, PA, TX, VA) that account for 85% of the US Hispanic population. We obtained hospital characteristics from the American Hospital Association and we applied the Agency for Healthcare Research and Quality’s Inpatient Quality Indicators protocol to AMI discharges. Using the proportion of Hispanic patients out of all discharges at each hospital, we stratified all hospitals into deciles. The top 3 deciles served a majority of Hispanic patients (79%), but smaller proportions of white (22%), black (34%) and Asian (53%) patients. We adjusted for patient risk using Elixhauser comorbidity indicators to compare inpatient mortality differences by race/ethnicity (Hispanics, non-Hispanic blacks, Asians and non-Hispanic whites). Using logistic regression models we estimated race/ethnic differences in risk-adjusted mortality and stratified these differences across Hispanic-serving hospital deciles.

RESULTS: There were 513,639 AMI discharges (10.8% Hispanics, 9.5% blacks, 2.9% Asians and 76.8% whites). Overall inpatient mortality rate was 6.1%. Inpatient mortality was higher among Hispanics (odds ratio (OR) = 1.09, p < 0.05) and Asians (OR = 1.09, p < 0.05), but no different among blacks (OR = 0.97), compared to whites, after adjusting for age, sex and comorbidities. Across hospital deciles by Hispanic-serving proportion, the share of Hispanic patients ranged from 0.1% in the lowest decile to 46.3% in the highest, and the latter were more likely to be safety-net, for-profit, government-owned or have a low annual average AMI volume. Risk-adjusted inpatient mortality exhibited a U-shape with respect to Hispanic-serving hospital deciles with the lowest mortality in the middle, decile 5 hospitals. Relative to decile 5 hospitals, mortality was higher in the top 3 deciles: decile 8 OR = 1.21; decile 9 OR = 1.14; decile 10 OR = 1.29 (all p-values < 0.05).

CONCLUSIONS: Hispanic patients have a higher rate of AMI inpatient mortality than white patients. Hospitals that disproportionately serve Hispanic AMI patients serve a lower portion of white and other minority patients and have higher inpatient mortality compared to other hospitals. Quality improvement at these hospitals could reduce Hispanic disparities in healthcare outcomes.

DO MEDICARE’S PAY-FOR-PERFORMANCE PENALTIES HAVE A BITE? THE CUMULATIVE 2013–2016 EXPERIENCE. Amresh D. Hanchate 4, 1; Meng-Yun Lin2; Danny McCormick3; Ge Bai6; Gerard F. Anderson5; Souvik Banerjee1; Michael Paasche-Orlow1. 1Boston University School of Medicine, Boston, MA; 2Boston Medical Center, Boston, MA; 3Harvard Medical School/Cambridge Health Alliance, Cambridge, MA; 4VA Boston Healthcare System, Boston, MA; 5Johns Hopkins University, Baltimore, MD; 6Johns Hopkins University, Washington, DC. (Control ID #2701207)

BACKGROUND: In 2013 the Centers for Medicare & Medicaid Services’ (CMS) introduced pay-for-performance programs - the Hospital Readmissions Reduction Program (HRRP) and Hospital Value Based Purchasing (HVBP) program - that instituted Medicare payment penalties (and bonuses) for hospitals based on patient outcome performance measures. However, due to the small size of annual penalties, the potency of the programs to influence quality improvement has been questioned. Little is known about how often hospitals are subject to repeated penalties, the cumulative penalty size and associated financial loss. Using data from 2013–2016 we assessed the frequency of repeated penalties, revenue loss relative to hospital financial performance indicators, and characteristics associated with higher financial loss.

METHODS: We obtained data on Medicare participating hospitals’ penalty/bonus payment rates for the HRRP and HVBP programs (2013–2016) from CMS’ Final Impact Rule files. We merged these data with hospital data on acute inpatient care revenues from Medicare patients and overall patients, and net income (operating profits), obtained from CMS’ Healthcare Cost Report Information System (HCRIS) files (2013–2014, the program years for which data is available). We used the 2013–2014 annual average of the HCRIS financial performance indicators, adjusted for inflation, as the baseline indicator of hospital financial status. We applied the cumulative penalty (and bonus) payment rates during 2013–2016 to the baseline financial measures to obtain the expected cumulative penalties ($) as a proportion of annual net income (income share) and of annual overall inpatient revenues (revenue share). Using linear regression models we examined the associations of both income and revenue share with several hospital characteristics: Medicare share of hospital inpatient volume (Medicare share), safety-net status, bed size and non-profit status.

RESULTS: Our study cohort included 2,192 hospitals. During 2013–2016, 1,695 hospitals (77%) experienced a cumulative penalty from both programs combined; 72 and 41% hospitals were penalized at least 3 years under HRRP and HVBP, respectively. Among penalized hospitals, the maximum cumulative penalty rate was 9.5% (median = 1.8%), and the maximum penalty amount was $18,869 per bed (median = $1,961). In over half the penalized hospitals, income share exceeded 3.2%, and revenue share exceeded 0.4%. 363 hospitals (16.6%) experienced penalties exceeding 2% of income share and 1% of revenue share. Income and revenue shares were higher in hospitals with higher Medicare share of total discharge volume; hospitals with >50% Medicare share experienced higher penalties amounting to 6.5% income share and 0.6% revenue share. Safety-net status, larger bed size and non-profit status were also associated with higher financial loss.

CONCLUSIONS: Repeated penalties from HRRP and HVBP programs are common, and for many hospitals the cumulative penalties represent a sizable share of net income and revenues.

DO PRESCRIBED OPIOIDS INCREASE RISK OF COMMUNITY-ACQUIRED PNEUMONIA REQUIRING HOSPITALIZATION? E. J. Edelman 1; Kirsha S. Gordon2; Janet P. Tate1; William Becker2, 1; Kendall Bryant3; J. R. Gaither1, 2; Cynthia Gibert4; Adam Gordon5; Brandon D. Marshall6; Maria Rodriguez-Barradas7; Jeffrey H. Samet8; Kristina Crothers9; Amy C. Justice1, 2; David A. Fiellin1. 1Yale University, New Haven, CT; 2VA Connecticut Healthcare System, West Haven, CT; 3National Insitutes of Health (NIH), Bethesda, MD; 4Veterans Affairs Medical Center, Washington, DC; 5University of Pittsburgh and VA Pittsburgh Healthcare System, Pittsburgh, PA; 6Brown University, Providence, RI; 7Michael E. DeBakey VAMC, Houston, TX; 8Boston University School of Medicine, Boston, MA; 9University of Washington, Seattle, WA. (Control ID #2707005)

BACKGROUND: As opioids have respiratory and immunosuppressive properties, we aimed to determine whether prescribed opioids impact risk of community-acquired pneumonia requiring hospitalization (CAP) among patients with (HIV+) and uninfected individuals.

METHODS: We conducted a longitudinal analysis of data from 2000 to 2012 among HIV+ and uninfected controls in the Veterans Aging Cohort Study (VACS). Patients were followed upon receipt of their first outpatient medication prescription dispensed from the Veterans Affairs Healthcare System (VA). Patients were excluded if during their first year of observation they 1) died; 2) had an opioid prescription >14 days supply to capture patients who had no or only minimal opioid receipt; or 3) had evidence of severe medical illness with a VACS index score (a validated measure of mortality risk) >100. Additionally, patients were excluded if: 1) they ever received chemotherapy or immunosuppressive medications; 2) had an initial opioid prescription with an average daily morphine equivalent dose exceeding 100 mg. Using outpatient VA pharmacy data, the exposure of interest, opioid receipt, was categorized as a time-varying 4-level variable based on receipt in the current and prior 30-day intervals as: 1) none, indicating no opioid receipt in the prior 60 day interval; 2) previous only, indicating opioid receipt only in the immediate prior 30 day interval; 3) current only, indicating opioid receipt in the current 30 day interval only, or 4) both, indicating opioid receipt in the prior and current 30 day interval. The outcome, CAP, was defined based on ICD-9 codes 480–487 and receipt of antimicrobials by the third day of hospitalization. Baseline covariates included demographics (age, gender, race/ethnicity), comorbidities (HIV, HCV), smoking status and VACS index score. We created time-varying Cox proportional hazards models to assess the association between opioid receipt and incident CAP.

RESULTS: Among our analytic sample (n = 56,630), patients were followed for a mean of 8.1 years. The mean age was 49 years, 98% were men, 39% white; 48% black; 9% Hispanic; 4% other and 34% were HIV+. Eleven percent of the sample had an incident CAP during the follow-up period. In adjusted analyses, compared to none, those with opioid receipt in the previous interval only [adjusted hazard ratio (aHR) [95% confidence interval] = 1.51 [1.31, 1.74]), current interval only (aHR [95% CI] = 4.35 [3.99, 4.73]), and both intervals (aHR [95% CI] = 1.82 [1.66, 2.01]) were more likely to have incident CAP.

CONCLUSIONS: Opioid receipt is independently associated with an increased risk of pneumonia, with greatest risk when initiating treatment. These data should inform discussions regarding benefit and harms associated with prescribed opioids. Among those for whom prescribed opioids are deemed necessary, efforts to address modifiable risk factors for pneumonia should be prioritized.

DO PRIMARY CARE PATIENTS PRESENTING WITH RECTAL BLEEDING RECEIVE RECOMMENDED COLONOSCOPIES? Sanja Percac-Lima 1, 2; Lydia E. Pace1, 3; Kevin H. Nguyen4; Charis N. Crofton5; Katharine A. Normandin5; Sara Singer1, 4; Meredith Rosenthal4; Alyna T. Chien1, 5. 1Harvard Medical School, Boston, MA; 2Massachusetts General Hospital, Boston, MA; 3Division of Women’s Health, Brigham and Women’s Hospital, Boston, MA; 4Department of Health Policy and Management, Harvard T. H. Chan School of Public Health, Boston, MA; 5Division of General Pediatrics, Department of Medicine, Boston Children’s Hospital, Boston, MA. (Control ID #2699120)

BACKGROUND: Missed and delayed cancer diagnoses are a major ongoing concern in primary care. Rectal bleeding can be the first symptom of colorectal cancer and clinical guidelines recommend colonoscopies for most patients with rectal bleeding. This study examined how often patients presenting to primary care with rectal bleeding receive recommended colonoscopies and how colonoscopy rates differ based on patient characteristics and utilization patterns.

METHODS: We conducted a cross-sectional study of 258 patients aged 40–80 who received primary care at 15 practices across 5 academic medical centers affiliated with Harvard Medical School between 2012 and 2016. In this period, practices were involved in learning collaboratives aimed at establishing team-based care and improving preventive cancer screening. We used ICD-9 codes indicative of rectal bleeding to randomly select 20 patients per practice across the 4 study years, oversampling for Medicaid-insured and non-Medicaid patients in a 1:1 manner. We followed CRICO (Controlled Risk Insurance Company) guidelines to determine if a colonoscopy was recommended. Three research assistants used a 2,620-item tool to abstract patient data across 4 electronic health record systems and dually abstracted 10% of charts (the Kappa = 0.95); primary care physicians adjudicated disagreements. We used multivariate logistic regression to examine the degree to which receiving recommended colonoscopies was associated with patient age, sex, non-White race/ethnicity, insurance type, number of chronic conditions, primary care no show rates, and number of primary care and subspecialist visits unrelated to rectal bleeding.

RESULTS: Of the 258 patients with rectal bleeding, 45% (115) were female, average age was 55 (SD 10) years, 60% (155) were non-White (e.g. Hispanic, Black, Other), and 49% (126) had Medicaid. CRICO guidelines indicated that 91% (234) of patients presenting with rectal bleeding should have received a colonoscopy; 55% (131 of 234) did within 1 year of presentation. The average time to colonoscopy was 76 (SD 81) days. Age, sex, insurance type, number of chronic conditions, primary care no show rates and number of subspecialist visits unrelated to rectal bleeding were not significantly associated with colonoscopy receipt. In multivariable analyses, non-White patients were more likely to receive a colonoscopy than White patients (OR = 2.05, CI [1.15,3.65] p = 0.02), and the number of primary care visits unrelated to rectal bleeding was negatively associated with the odds of receiving a colonoscopy (OR = 0.80, CI [0.66,0.97] p = 0.03).

CONCLUSIONS: Care for primary care patients presenting with rectal bleeding is sub-optimal. Prioritization of issues unrelated to rectal bleeding may explain some of these deficiencies. However, in contrast to previous studies in the same setting, non-White patients were more likely to receive recommended care.


Leah Zallman 2, 3; Robert Joseph2; Emily Benedetto2; Ellie Grossman1; Lisa Arsenault3; Assaad Sayah2. 1Cambridge Health Alliance, Somerville, MA; 2Cambridge Health Alliance, Cambridge, MA; 3Institute for Community Heatlh, Malden, MA. (Control ID #2704023)

BACKGROUND: Behavioral health integration (BHI) is an increasingly employed strategy for improving patient outcomes, reducing costs, and improving patient experience. The impact of these programs on primary care providers’ (PCPs’) experience caring patients with mental health and substance use disorders, in particular PCPs’ perceptions of behavioral health (BH)-primary care (PC) system functioning and knowledge, are poorly understood. We aimed to examine the impact of BHI on PCPs’ perceptions of behavioral health BH-PC system functioning and knowledge.

METHODS: We implemented BHI based on evidence-based models consisting of seven elements: (1) Screening for mental health and substance use disorders, (2) Training of PC teams, (3) Integration of BH providers into PC teams, (4) Roll-out of unlicensed mental health care managers and establishment of a BH registry, (5) Psychiatry consult service, (6) Site-based BHI meetings, and (7) Site self assessments. The intervention was rolled out in early integration sites during 2 years and late integration sites during the subsequent 2 years. In this observational pre-post study, we administered an anonymous online survey annually to PCPs; 381 PCPs at 11 primary care clinics participated. We examined changes in perceptions with chi-square tests and fisher’s exact tests. We also conducted multivariable logistic regression analyses controlling for provider and site level characteristics.

RESULTS: The proportion of PCPs with high BH-PC systems functioning scores quadrupled from 14 to 55% (p <0.0001) and high knowledge scores increased from 63 to 85% (p < 0.001). Larger increases were demonstrated in early integration sites during the first 2 years and in late integration sites during the latter 2 years of the survey. Adjusting for participant and site level characteristics did not change these outcomes.

CONCLUSIONS: BHI improves PCP perceptions of BH-PC system functioning and knowledge. BHI is a strategy for improving PCP experience caring for patients with mental health and substance use disorders.

DOES BEIJING’S DIAGNOSIS-RELATED GROUP PAYMENT REFORM PILOT IMPROVE QUALITY OF ACUTE MYOCARDIAL INFARCTION CARE? Adrienne N. Poon 1; Kit Yee Chan2; WeiYan Jian3. 1George Washington School of Medicine and Health Sciences, Clifton, NJ; 2University of Edinburgh, Edinburgh, United Kingdom; 3Peking University, Beijing, China. (Control ID #2706989)

BACKGROUND: In China, quality of acute myocardial infarction (AMI) management has previously been found to be poor with uneven guideline implementation. Previous studies have supported the use of DRG payment systems to help drive improvements in quality of care. In 2012, China’s first pilot diagnosis-related group (DRG) payment system was implemented in Beijing. This study aims to explore whether this DRG payment pilot has improved quality of AMI management.

METHODS: Discharge and claims data from the Beijing Employee Basic Health Insurance Scheme were obtained from 14 tertiary care hospitals from January 2010 to September 2012. In October 2011, 108 DRGs for specific conditions/procedures were piloted in 6 hospitals while fee-for-service (FFS) payment was continued in controls. Four AMI DRGs included: percutaneous coronary intervention (PCI) procedure with stent that is complicated or uncomplicated and the higher risk coronary artery bypass grafting (CABG) surgery with and without cardiac catheterization. Regression was performed with a differences-in-differences design to assess whether DRG pilot implementation improved quality of care measures.

RESULTS: AMI DRG cases at pilot hospitals included 1,221 (88.9%) complicated PCI with stent, 69 (5.0%) uncomplicated PCI with stent, and 84 (6.1%) CABG surgeries. Of eligible DRG cases at pilot hospitals post-reform, 397 (28.9%) were ultimately paid through FFS including 69.1% of CABG surgeries. DRG payment led to a significant 72.2% reduction in in-hospital mortality though a 7.1% reduction in prescription of optimal AMI medications at arrival. DRG-eligible cases ultimately paid through FFS had a 24.1% increase in expenditures per admission, out-of-pocket (OOP) payment, and 15.4% increase in length of stay compared to cases paid through DRG. For higher quality uncomplicated PCI with stent cases, there were significant reductions in expenditures (19.7%) and OOP payments (29.5%). For FFS medically managed and PCI only cases, there was an 8.9% reduction in OOP payments and a 5.4% increase in in-hospital mortality.

CONCLUSIONS: DRG payment for cardiac procedures stimulated significant reductions in in-hospital mortality while reducing costs for high quality uncomplicated PCI with stent cases showing potential for DRG payment to improve quality of AMI in Chinese hospitals. Outcomes in terms of in-hospital mortality, however, were worse likely due to access to expensive but potentially life-saving procedures only if payment was guaranteed. Furthermore, more expensive and medically risky cases were shifted towards FFS payment, which is a major gap. Given that the majority of PCI cases had procedural complications as well as poor-adherence to medication guidelines, there is an urgent overall need to improve overall quality of AMI care and create incentives to stimulate adherence to evidence-based practice.

DOES CARDIOVASCULAR DISEASE RISK IN POSTMENOPAUSAL SOUTH ASIAN WOMEN CATCH UP TO MEN Diana Thiara 2; Feng Lin3; Namratha R. Kandula1; Alka M. Kanaya3. 1Northwestern University, Chicago, IL; 2University of California at San Francisco, San Francisco, CA; 3University of California, San Francisco, San Francisco, CA. (Control ID #2703172)

BACKGROUND: South Asians (SA) bare a greater burden of cardiovascular disease (CVD) and have the highest rate of mortality from ischemic heart disease than other ethnic populations in the US. While studies have identified some modifiable risk factors associated with cardiovascular disease in SA populations, no study has examined whether there is sex difference in subclinical atherosclerosis among postmenopausal women and similarly aged men.

METHODS: We conducted a cross-sectional analysis including postmenopausal women and men age 50–84 years old without known cardiovascular disease from the Mediators of Atherosclerosis in South Asians Living in America (MASALA) study. Post-menopausal status was ascertained by self-report, and women with oophorectomy were excluded. High resolution ultrasonography was done to measure common carotid intima media thickness (CCA) and internal carotid intima thickness (ICA), and cardiac CT scan was performed to quantify coronary artery calcium (CAC). Regression models were used to examine the relationship between sex and ICA, CCA, and CAC, separately. Using multivariate analyses, we serially adjusted for traditional CVD risk factors (age, hypertension, diabetes, LDL, BMI, waist circumference, smoking, alcohol use, education, statin use, and family history of heart attack), behavioral factors (exercise and total caloric intake), psychosocial factors (depression, anxiety, burden and traditional cultural beliefs).

RESULTS: Of 576 SA (44% women), men had a mean age of 62 and women 59 years (p < 0.001). Men had significantly higher diabetes prevalence (34% vs 24%, p = 0.005), alcohol use (47% vs 16%, p < 0.001) and smoking (current or former: 31% vs 3%, p < 0.001), and were more likely to have higher levels of education than women (89% vs 82%, p = 0.02). Men also had higher rates of hypertension (55% vs 44%, p = 0.01) and larger waist circumferences (95.6 cm vs 90.0 cm, p < 0.001). Women had higher LDL than men (115 vs 104 mg/dl, p < 0.001) and higher depressive symptoms (CES-D score, 8.4 vs 7.2, p = 0.05). In unadjusted analyses, men had higher mean CCA (0.96 mm vs 0.88 mm, p < 0.001) and ICA (1.37 mm vs 1.21 mm, p < 0.001) and were more likely to have any prevalent CAC (70% vs. 33%) and higher extent of CAC compared to women. Multivariate adjustment for traditional CVD risk factors attenuated the association between sex and CCA and ICA completely, however the differences in CAC between sexes persisted. Men had a higher prevalence of any CAC (OR 4.67 (95% confidence interval, 2.63-8.28)) and extent of CAC (β 1.7, p < 0.001) in adjusted analyses.

CONCLUSIONS: While older postmenopausal women have similar prevalence of carotid intima media thickness compared to men, the prevalence and extent of CAC remains higher in men. These findings suggest that South Asian women, like women in other race/ethnic groups, may not benefit from aggressive statin therapy for primary prevention of coronary heart disease.

DOES DEPRESSION AFFECT GLYCEMIC CONTROL AND DELIVERY OF GUIDELINE-RECOMMENDED CARE IN PATIENTS WITH HEART FAILURE AND DIABETES Margaret Zupa; Kaleab Abebe; Yan Huang; Amy Anderson; Bea Herbeck Belnap; Bruce L. Rollman. University of Pittsburgh, Pittsburgh, PA. (Control ID #2703184)

BACKGROUND: Heart failure (HF) and type 2 diabetes mellitus (T2DM) are common, often comorbid chronic medical conditions which demand adherence with complex treatment regimens. While clinical practice guidelines can assist clinicians in providing evidence-based care, co-morbid depression may impair effective management of HF and T2DM and thereby increase afflicted individuals’ risk of morbidity and mortality. However, little is known about the impact of depression in patients with both conditions. We therefore explored the effect of depression on glycemic and blood pressure control and delivery of guideline-recommended HF and T2DM care among patients enrolled in the NHLBI-funded Hopeful Heart Trial presently examining the impact of treating depression in HF patients.

METHODS: We screened patients hospitalized with systolic HF (ejection fraction (EF) ≤ 45%) and NYHA class II-IV symptoms for depression with the Patient Health Questionnaire (PHQ-2), and telephoned 2 weeks after discharge to administer the PHQ-9. We classified patients as depressed if they screened positive on PHQ-2 and scored ≥ 10 on the follow-up PHQ-9, and recruited a randomly sampled cohort of non-depressed subjects who screened negative on PHQ-2 and scored <5 on follow-up PHQ-9 as a control group. We collected sociodemographic information from patient self-report and abstracted the hospital discharge summary for medical diagnoses, medication use, cardiac ejection fraction, blood pressure, and HbA1c. We used student’s t-test or chi-square test, when appropriate, to compare depressed and non-depressed HF patients with T2DM.

RESULTS: From 3/2014 to 11/2016, we enrolled 545 HF patients including 260 (48%) with T2DM. Depressed (N = 211) and non-depressed (N = 49) subjects were similar by mean age (65 years), gender (58% male), EF (28%), and marital status (45% married). Seventy-six percent of depressed patients were white compared with 55% of non-depressed; 23% of the depressed group was African American compared with 41% of non-depressed (p = 0.01 for racial distribution). Critically, baseline glycemic control as measured by mean HbA1c (8.2% vs 7.7% p = 0.16), systolic (126 vs 123 mmHg p = 0.25) and diastolic (71 vs 70 mmHg p = 0.53) blood pressure, and rates of HF and T2DM guideline-adherent prescription of ACE/ARB (63 and 69% p = 0.40) and statins (80 and 73% p = 0.35) were similar between depressed and non-depressed patients.

CONCLUSIONS: At baseline, comorbid depression was not related to glycemic or blood pressure control or delivery of guideline-adherent care to patients with HF and T2DM. Further research is needed to identify practices that improve adherence with guideline-based care for medically complex patients with HF and T2DM and elucidate the long-term impact of depression on patients with these conditions.

DOES NURSE CARE MANAGEMENT OF PATIENTS AT HIGH RISK FOR ALCOHOL USE DISORDERS IMPROVE DRINKING OUTCOMES? RESULTS OF THE CHOOSING HEALTHIER DRINKING OPTIONS IN PRIMARY CARE (CHOICE) TRIAL Katharine Bradley 1; Jennifer F. Bobb1; Evette Ludman1; Laura Chavez5; Andrew J. Saxon3; Joe O. Merrill6; Emily Williams4; Eric Hawkins3; Ryan Caldeiro1; Gwen Lapham1; Julie E. Richards1; Amy K. Lee1; Daniel Kivlahan2. 1Group Health Research Institute, Seattle, WA; 2VA Puget Sound, Seattle, WA; 3VA Puget Sound Health Care System, Seattle, WA; 4Health Services Research and Development (HSR&D) Seattle Center of Innovation for Veteran-Centered and Value-Driven Care, Seattle, WA; 5Ohio State University, Columbus, OH; 6University of Washington, Seattle, WA. (Control ID #2706223)

BACKGROUND: Importance: Alcohol use disorders (AUD) are common and under-recognized, and most patients never receive treatment. Experts recommend that AUDs be managed as chronic conditions in primary care, but the effectiveness of AUD care management unknown. Objective: To test whether a year-long care management intervention by registered nurses (RNs), compared to usual care, improved drinking outcomes at 12 months among patients at high risk for AUDs.

METHODS: Design: Randomized, controlled, encouragement trial. Setting: Three primary care clinics in one Veterans Affairs (VA) Health Care System Participants: VA outpatients who reported heavy drinking twice weekly or once weekly if previous AUD treatment, and consented to participate in a trial in which they might be offered additional alcohol-related care. Intervention: Proactive outreach and engagement by an RN who offered 12 months of repeated brief counseling using motivational interviewing (with or without biomarker monitoring), shared decision-making about treatment options, and support accessing preferred care including AUD medications.

RESULTS: Main Outcomes and Measures: Percent heavy drinking days (%HDD) on a 28-day timeline follow-back (TLFB), and percent of patients with “good drinking outcomes” (GDOs), assessed by blinded telephone interviewers, at 12 months. GDO is defined as drinking below recommended limits (TLFB) without alcohol-related symptoms (past 3 months) on the Short Inventory of Problems. Results: Among 304 participants (91% male, mean age 51.4 years), 73% met criteria for DSM-IV AUDs. Most (91%) intervention patients engaged in care with study RNs: 6% 1 visit, 33% 2–5, 20% 6–9, and 32% ≥ 10 visits. Intervention patients were more likely to receive alcohol-related care (42% vs 26%; p = 0.005), reflecting increased use of AUD medications (32% vs 8.4%; p < 0.0001), but no differences in specialty alcohol treatment or mutual help. There were no significant differences main outcomes at 12 months. Among usual care and intervention patients, respectively, %HDD were 35 and 39% (p 0.44), and percent with GDOs were 20 and 15% (p = 0.32), respectively. Findings at 3 months were similar.

CONCLUSIONS: Conclusions and Relevance: The CHOICE intervention for AUDs engaged patients in alcohol-related care and increased use of AUD medications, but did not improve drinking outcomes at 12 months.

DOES PHYSICIAN GENDER EFFECT ORDERING PRACTICE Nina Garza 2; Fawaz Georgie2; Alexander Horbal2; David E. Willens1; Alexis C. Haftka-George2. 1Henry Ford Health System, Detroit, MI; 2Henry Ford Hospital, Detroit, MI. (Control ID #2701850)

BACKGROUND: Breast cancer cases represent 14.6% of all new cancer diagnosis in the United States. However, medical societies disagree when, and how often, we should be screening. The United States Preventative Service Task Force recommends starting at 50 years of age and screening biennially, while others recommend starting at 40 years of age, or screening annually. This has created an environment where physicians must decide which guideline is best for their patients. We sought to find out what factors influence this decision, specifically if the gender of the ordering physician was associated with different screening practices in women aged 40–49.

METHODS: We examined every office visit for female patients age 40–49 with an internal medicine (IM), family medicine (FM) or gynecology (Gyn) provider in our health system between July 1, 2015 to May 30, 2016. Patients with a history of breast cancer or other malignant neoplasm were excluded. The association between physician gender and mammogram ordering rates was assessed via chi-squared testing. Other factors, such as comparison between specialties, were assessed via multivariable binary logistic regression.

RESULTS: In female patients aged 40–49, female physicians are more likely to order mammograms than male physicians overall. This disparity between genders was largest in internal medicine. Gynecology physicians order mammograms at a higher rate than internal medicine or family medicine physicians. Women aged 45–49 were more likely to receive a mammogram order than women aged 40–44. Also, black patients were less likely to receive a mammogram order compared to white patients.

CONCLUSIONS: Physician ordering practices do appear to vary by gender, however, this pattern is also influenced by specialty. The decision also seems to be effected by the age and/or race of the patient. The results of this study support the need for more research in factors contributing to preventive healthcare disparities.

DOES STIGMATIZING LANGUAGE IN PATIENT CHARTS NEGATIVELY IMPACT PHYSICIAN ATTITUDES AND CLINICAL DECISION MAKING? AN EXPERIMENTAL STUDY Anna P. Goddu 1; Katie O’Conor1; Sophie Lanzkron1; Mustapha Saheed1; Somnath Saha2; Monica E. Peek3; Carlton Haywood1; Mary Catherine Beach1. 1Johns Hopkins University, Baltimore, MD; 2Portland VA Medical Center, Portland, OR; 3The University of Chicago, Chicago, IL. (Control ID #2698556)

BACKGROUND: Clinician bias is a known contributor to health disparities. Research on implicit bias has primarily focused on individuals, rather than the transferability of such bias to others. Language used to describe patients may reflect physician bias, and also influence opinions of patients’ subsequent physicians. Few studies have evaluated written physician notes as a source of implicit bias. This study’s objective was to assess whether stigmatizing language written in a patient chart is associated with a subsequent clinician’s bias towards the patient and clinical management.

METHODS: Medical records from 82 inpatient and ED records were abstracted for stigmatizing language and used to create a single, hypothetical clinical case. Medical students and residents (internal medicine and emergency medicine) were then asked to complete an online survey which randomized them to 1 of 2 physician chart notes, one with the stigmatizing language and one without. The notes presented medically-identical information about the same hypothetical patient, a 28 y/o man with sickle cell disease in a vaso-occlusive crisis. For example, the stigmatizing note described that Mr. R was hanging out with friends outside McDonald’s and that Mr. R reported some “stressful situations” in his life. The neutral note stated that Mr. R spent yesterday afternoon with friends and reported recent stress. We assessed attitudes towards the hypothetical patient using the previously-validated Positive Attitudes towards Sickle Cell Patients Scale (range 5–35) and pain management decisions (residents only) using two multiple-choice questions (composite range 1–7 representing intensity of pain treatment). We used t-tests and Wilcoxon rank-sum tests to calculate differences in attitudes and treatment based on study arm assignment.

RESULTS: Respondents included 233 medical students and 180 residents (overall response rate 58%). About half (55%) identified as male. About half (54%) identified as white, 27% as Asian, 14% as Hispanic/Latino, and 10% as African-American. There were no differences between study arms in gender, race/ethnicity or year of training. Medical students and residents who reviewed the stigmatizing-language note had more negative attitudes towards the patient (20.6 stigmatizing vs. 25.6 neutral, p < 0.001). Further, residents exposed to the stigmatizing vs. neutral language note made less aggressive pain management decisions (5.56 stigmatizing vs. 6.22 neutral, p = 0.003).

CONCLUSIONS: Stigmatizing language written by one provider in a patient’s medical record can negatively impact subsequent providers’ attitudes towards the patient and influences clinical decision-making. Future work should explore which dimensions of stigmatizing language are most harmful. Attention to the language used in medical records may promote patient-centered care and reduce healthcare disparities for stigmatized populations.

DOES WISDOM PROTECT AGAINST DEPERSONALIZATON AMONG MEDICAL STUDENTS? John Schorling; Margaret L. Plews-Ogan; Rachel Kon; Tabor Flickinger; Justine E. Owens. University of Virginia, Charlottesville, VA. (Control ID #2701545)

BACKGROUND: Burnout among medical students is an important issue, especially the development of depersonalization which has been associated with loss of empathy and unprofessional behavior. Wisdom has also received recent attention. Branch suggests “wisdom is what we should be striving for in our development as clinicians”. Ardelt describes wisdom as: understanding deeper meanings, knowing limits of knowledge, tolerating ambiguity, engaging in reflection, having compassion and being other-centered. In cross-sectional analyses, we found that medical students high in wisdom had lower burnout scores. The purpose of this study was to evaluate whether students who score higher in wisdom are less likely to develop depersonalization over time.

METHODS: All students at one medical school were sent annual online surveys during the 2014–5 and 2015–6 years that included Ardelt’s 3-Dimensional Wisdom Scale (3D-WS) and the Maslach Burnout Inventory. Scores of 10 or greater for depersonalization (DP) were considered high. ANOVA was used to compare mean scores on the 3D-WS by high and low DP status.

RESULTS: The survey was sent twice to 473 students. 143 (30%) had paired data from the 3D-WS and MBI for both years. The mean age was 25. 52% were female. Students were classified as those who never had high DP (n = 76), had low DP in the first year of the study but became high in the second (n = 20), had high DP in the first year but became low in the second (n = 18), or had high DP both years (n-29). Students who ever met criteria for high DP had similar mean wisdom scores (3.6), which did not vary significantly by year but which were significantly lower than those who never had high DP (3.9, p < .001).

CONCLUSIONS: Students who never met criteria for high depersonalization had significantly higher wisdom scores than the scores of those who ever had high depersonalization. This suggests that having high wisdom as measured by the 3D-WS might be protective with regard to developing depersonalization in medical school.

figure z

Association between wisdom scores and high and low DP. The first set of columns had low DP both years, the second went from low to high, the third from high to low, and the fourth were high both years.

DON’T CLICK IT, REDUCE THE STICK IT: A QUALITY IMPROVEMENT INITIATIVE TO REDUCE DAILY INPATIENT LABORATORY DRAWS Clark A. Veet 1; Orighomisan Pessu2; Gregory M. Bump3. 1UPMC, Pittsburgh, PA; 2University of Pittsburgh Medical Center, Pittsburgh, PA; 3University of Pittsburgh School of Medicine, Pittsburgh, PA. (Control ID #2695261)

BACKGROUND: It is accepted that diagnostic testing is over-used in stable hospitalized medicine patients, including the use of daily labs such as complete blood counts and electrolyte monitoring. The Society for Hospital Medicine recommends eliminating this process in stable patients without new symptoms. This project focused on assessing patient understanding and preferences regarding daily labs. It also desired to obtain baseline daily laboratory data and devise a model to reduce unnecessary phlebotomy to improve patient satisfaction and reduce unnecessary testing.

METHODS: This initiative had two parts. Part one identified patient understanding and preference regarding daily labs via a four question survey. Part two quantified the baseline lab ordering habits before and after educational intervention which included presentations to housestaff and attendings. The rates of lab draws were compared using a t-test.

RESULTS: Part one identified patient preferences and attitudes regarding daily inpatient lab draws. 79 inpatients completed a four question survey focusing on expected and desired frequency of daily lab draws. 21% of respondants felt doctors ordered too many labs and only 45% of respondants expected daily labs. Part two measured the number of daily lab draws on an inpatient housestaff medicine service. A total of 2839 patient-days were analyzed on 7 medicine teams over the course of two months. At baseline, 75.7% of patients had CBCs, 78.9% had BMPs, and between 23 and 46% had Ca, Mg, and PO4. After intervention, 76.3% of patients had CBCs, 76.9% had BMPs, and between 23 and 49% had Ca, Mg, and PO4. There was no statistically significant difference between these groups.

CONCLUSIONS: This inpatient quality improvement initiative aimed to understand patient perspectives on daily labs and measure change associated with educational intervention on daily lab rates. Part one provided meaningful data regarding patient preferences and showed that patients do not necessarily expect “daily labs” as part of the hospital experience and do not equate it with optimal care. While other institutions have implemented similar projects with reductions in daily lab rates, we saw no significant difference after intervention due to many factors. On the individual level, junior housestaff are often responsible for order entry and may be uncomfortable without labs due to diagnostic uncertainty. Other factors include expectations from senior team members. Ineffective signout to covering team members may also create issue, as a patient without labs ordered may be perceived as an error. Interventions at other institutions have showed that a multi-disciplinary approach is needed to reduce daily labs. These include provider and nursing education as well as cultural shifts. Interventions to limit the number of days ordered on CPOE may be helpful, though it is met with pushback from generalists and specialists alike. At this institution, the project continues with additional interventions being implemented.

EARLY IMPACTS OF THE ACA ON OUT-OF-POCKET AND INSURANCE PREMIUM SPENDING Anna Goldman 2; Stephanie Woolhandler1; David Himmelstein1; David Bor2; Danny McCormick2. 1CUNY School of Public Health, New York, NY; 2Harvard Medical School/Cambridge Health Alliance, Cambridge, MA. (Control ID #2700865)

BACKGROUND: The Patient Protection and Affordable Care Act (ACA) has reduced the number of Americans who report being unable to afford care, but critics have cited exchange plans’ high premiums and deductibles. We examined changes in out-of-pocket (OOP) and household premium spending following implementation of the ACA’s coverage expansions on January 1, 2014.

METHODS: We analyzed data from the Medical Expenditure Panel Survey (MEPS) on non-elderly adults’ (age 18–64) OOP spending and premium contributions (for families that paid any private insurance premiums) prior to (2008–2013) and after the implementation of the ACA (2014), the most recent year of available data. We examined spending by persons in four groups based on family income: Poor (0–138% of the federal poverty level (FPL) - the eligibility threshold for Medicaid in most states); Lower income (139–250% of FPL - eligible for subsidized premiums and reduced cost-sharing on ACA exchanges); Middle income (251–400% of the FPL - eligible only for premium subsidies); and Higher income (>400% of FPL). We used linear regression models to estimate pre- to post-ACA changes in mean annual OOP spending and premium contributions, controlling for secular trends in the pre-ACA period. We used logistic regression to determine whether implementation of the ACA was associated with a reduction in the odds of health spending exceeding commonly used thresholds of affordability: Household OOP spending exceeding 10% of income (5% was also examined for the poor and near-poor); and family premium spending exceeding 9.5% of family income (the threshold defined by the ACA). Analyses employed weights to generate national estimates, and spending figures were adjusted to 2014 dollars using the consumer price index.

RESULTS: The 2008–2014 MEPS included 137,833 non-elderly adults. ACA implementation was associated with a significant $97.68 (SE, 38.75, p = 0.01) decrease in OOP spending for the lower income group, but not for others. We found no post-ACA change in the odds of household OOP spending exceeding >10% of family income for any income group; however, using the 5% threshold, the odds declined for lower income families (OR 0.81 [95% CI, 0.66–0.98]; p = 0.03). Mean premium contributions did not change post-ACA for any income group, although the odds of premium costs exceeding 9.5% of family income fell for the poor (OR 0.72 [95% CI 0.52–0.98]; p = 0.04). The rate of high-burden premium spending (>9.5% of family income) was 12 times higher in the poor income group compared to the wealthiest in the pre-ACA period (64.8% vs. 5.3%) and 11 times higher in the post-ACA period (58.7% vs. 5.2%).

CONCLUSIONS: In the first year after the law was implemented, the health spending burden on most families was unaffected by the ACA, and remained quite high. This has likely contributed to mixed political support for the law. Reforms that minimize OOP and premium spending among non-wealthy households might enjoy a more positive reception from the American public.

ECHOES OF BURNOUT IN INTERNAL MEDICINE RESIDENT NARRATIVE ESSAYS Emily Gordon. Rutgers New Jersey Medical School, Newark, NJ. (Control ID #2705234)

BACKGROUND: Burnout is a prevalent problem faced by general internists. It starts in medical school, continues through residency, and is a commonly cited cause for drop out among attending physicians. Narrative medicine is a practice that addresses the need to share experiences and challenges by way of telling stories. Introducing residents to narrative medicine could be a framework to identify and mitigate burnout, as well as cultivate resilience among trainees.

METHODS: From 2014–2015, all New Jersey Medical School Internal Medicine residents participated in a narrative medicine workshop. The workshop began with a sample reading of a narrative medicine piece by the course leader. The residents were then asked to submit a brief narrative essay on a memorable patient experience. Informed consent was requested from all participants to analyze their de-identified writing samples for common themes. Each sample was examined by the principal investigator for signs of burnout, as defined by the American Academy of Family Physicians Position Paper on Physician Burnout. We then calculated the frequency that such sentiments were expressed within and across writing samples.

RESULTS: Forty-three residents were asked to participate, and 39 consented to inclusion of their essays in this analysis. Of those, 13 (33%) contained statements concerning for burnout. Some examples were: “I realize from watching patients die that I have not enjoyed my life enough.” “I wake up crying. I’m exhausted and angry. A friend implores me to take the day off but I’m fearful of how I will be perceived at work.” “…hardened, impersonal, emotionless, well-oiled machines. This is the way a good doctor is supposed to be.”

CONCLUSIONS: Narrative medicine can be a powerful tool for identifying signs of burnout among Internal Medicine residents. In addition, sharing of patient stories in groups can help trainees to reflect the commonality of challenging patient experiences, which might mitigate feelings of burnout. As one resident stated in her essay: “These reflections remind us of how often we do form strong, meaningful connections with our patients.”

EFFECT OF A SOCIAL INCENTIVE-BASED GAMIFICATION INTERVENTION USING WEARABLE DEVICES AND SMARTPHONES ON PHYSICAL ACTIVITY: THE BE FIT RANDOMIZED CLINICAL TRIAL Mitesh Patel 1, 2; Emelia J. Benjamin3, 4; Kevin G. Volpp1, 2; Caroline Fox4; Dylan Small1; Joseph Massaro3, 4; Jane Lee4; Victoria Hilbert1; Maureen Valentino4; Devon Taylor1; Emily Manders3, 4; Karen Mutalik3, 4; Jingsan Zhu1; Wenli Wang1; Joanne Murabito3, 4. 1University of Pennsylvania, Philadelphia, PA; 2Crescenz VA Medical Center, Philadelphia, PA; 3Boston University, Boston, MA; 4Framingham Heart Study, Boston, MA. (Control ID #2700003)

BACKGROUND: Social networks have been demonstrated to influence individual health behaviors, but interventions that leverage social incentives within these networks to change health behaviors have not been well examined. The objective of this study was to test the effectiveness of a gamification intervention that used insights from behavioral economics to enhance social incentives to increase physical activity in the community.

METHODS: The Behavioral Economics Framingham Incentive Trial (BE FIT) was a randomized clinical trial that recruited 206 adults comprising 97 groups of two or three family members in the Framingham Heart Study and occurred between December 2015 and August 2016. Participants used a wearable device or smartphone application to establish a baseline step count and selected a step goal increase for a 12-week primary intervention period and a 12-week follow-up period. Participants in both the control and intervention arms received daily feedback on their performance for 24 weeks. During the first 12 weeks, participants in the intervention arm played a game (including points, levels, and lifelines) with their family members that was designed using insights from behavioral economics to enhance social incentives such as peer support, accountability, and collaboration. The primary outcome was mean proportion of participant-days the step goal was achieved during the primary intervention period. Secondary outcomes included mean proportion of participant-days the step goal was achieved during the follow-up period and mean daily steps during the intervention and follow-up periods.

RESULTS: During the 12-week intervention period, participants in the intervention arm achieved physical activity goals on a significantly greater proportion of participant-days (Mean, 0.53 vs. 0.32; adjusted difference, 0.27 [95% CI, 0.20–0.33]; P < .001) and had signficantly greater increase in daily steps compared to baseline (Mean, 1661 vs. 636; adjusted difference, 953 [95% CI, 505–1401]; P < .001). During the 12-week follow-up period, participants in the intervention arm achieved physical activity goals on a significantly greater proportion of participant-days (Mean: 0.44 vs. 0.33, adjusted difference: 0.12 [95% CI, 0.05–0.19]; P < .001) and had a significantly greater increase in daily steps compared to baseline (Mean, 1385 vs. 798; adjusted difference, 494 [95% CI, 170–818]; P < .01).

CONCLUSIONS: Among groups of family members in the community, a social incentive-based gamification intervention was effective at increasing physical activity during the 12-week intervention. After the intervention completed, differences were smaller but sustained through the 12-week follow-up period. To our knowledge, this is one of the first clinical trials to demonstrate how gamification and social incentives can be used to change health behaviors in the community.

EFFECT OF DIRECT-TO-CONSUMER ADVERTISING (DTCA) ON STATIN USE IN THE UNITED STATES Hsien-Yen Chang 1; Irene Murimi1; Matthew Daubresse1; Dima Qato2; Sherry Emery2; G. Caleb Alexander1. 1Johns Hopkins Bloomberg School of Public Health, Baltimore, MD; 2University of Illinois at Chicago, Chicago, IL. (Control ID #2705658)

BACKGROUND: The value of direct-to-consumer advertising (DTCA) of prescription drugs is widely debated, as is the effect of DTCA on prescription sales and health care utilization. We examined the association between DTCA intensity for statin medications and prescription sales and cholesterol-related health care utilization.

METHODS: We conducted an ecological study for 75 designated market areas (DMAs) from 2005 to 2009 in the United States using linked data regarding: (1) televised DTCA volume for rosuvastatin (Crestor™) and atorvastatin (Lipitor™) derived from Nielsen television ratings; (2) non-DTCA marketing and promotion derived from IMS Health Integrated Promotion Services; (3) retail, mail order and long-term care prescription drug sales derived from IMS Health Xponent; (4) prescription drug and ambulatory care health care utilization derived from Truven MarketScan; and (5) contextual factors such as health care density and socioeconomic status derived from the Area Resource File. We derived information for each month at each DMA and used multi-level negative binomial regression to account for nesting of individuals within DMAs.Main outcomes and measures were: (1) Volume of total, new and refilled prescription sales for the advertised statins; (2) number of statin prescriptions dispensed to commercially insured individuals; and (3) high cholesterol-related outpatient visits among the commercially insured.

RESULTS: The average intensity of rosuvastatin and atorvastatin ad exposures per household varied substantially across DMAs. After adjustment for socioeconomic, demographic and clinical characteristics, each 100-unit increase in advertisement viewership was associated with a 2.22% (95% confidence interval [CI] 0.30 to 4.19%) increase in rosuvastatin and atorvastatin sales. Similar patterns were observed between DTCA and statin dispensing among the commercially insured. DTCA was associated with increases in high cholesterol-related outpatient visits among adults 18–45 years of age (3.15% increase in visits per 100-unit increase in viewership, 95% CI: 0.98 to 5.37%) but not among those 46–65 years of age (0.51%, 95% CI: −1.49 to 2.55%).

CONCLUSIONS: DTCA for statins is associated with increases in statin utilization and hyperlipidemia-related outpatient visits, especially for young adults.

EFFECT OF HEALTH PLAN FINANCIAL INCENTIVE OFFERING ON EMPLOYEES WITH PREDIABETES Anita D. Misra-Hebert; Bo Hu; Phuc H. Le; Michael B. Rothberg. Cleveland Clinic, Cleveland, OH. (Control ID #2705493)

BACKGROUND: Prediabetes, defined as a glycosylated hemoglobin(HbA1c) level between 5.7–6.4% or a fasting glucose level between 100–125 mg/dl, may be improved or reversed with lifestyle interventions including diet, exercise, and weight loss or may progress to diabetes. Worksite wellness programs that offer financial incentives for participation in wellness activities or disease management may be effective in improving the health of employees with prediabetes at a population level. We studied the effect of offered employee health plan financial incentives on outcomes for employees with prediabetes.

METHODS: We conducted a retrospective cohort study using electronic medical record data. Exposures included fixed health plan financial incentives offered starting in 2009 through 2010, and then a premium discount starting in 2011 tied to program participation and achievement of goals. A total of 1115 employees identified with prediabetes from 2008–12 were matched per year, based upon when prediabetes was diagnosed, to 5705 non-employee patients with commercial insurance using propensity matching with baseline HbA1c, age, race(African American, White, Other), gender, and body mass index (<=25, 25.1–29.9, 30–35, and >35) included as covariates. We created longitudinal linear mixed models using the data for all employees and non-employees from the time that prediabetes was diagnosed to assess yearly changes in HbA1c, weight, and low density lipoprotein cholesterol (LDL).

RESULTS: Pooled analyses from 2010–12 beginning with the second year of the incentive program showed that the yearly change in HbA1c was greater for employees than for matched non-employees (−0.10% vs. −0.07%, respectively; p for difference in change = 0.002)). Pooled analyses for weight change and LDL cholesterol from 2010–12 showed that employees lost more weight than matched non-employees (−2.35 lbs vs. −0.02 lbs, respectively; p for difference in change = <0.001) and decreased LDL more than matched non-employees (−3.76 vs. −1.25, respectively; p for difference in change = 0.04).

CONCLUSIONS: A worksite wellness program offering employee health plan financial incentives for participation in wellness activities or disease management was associated with small improvements in HbA1c, weight, and LDL in employees with prediabetes at a population level. Longitudinal benefits may be greater than we observed over 3 years.

EFFECT OF NORMS ON LABORATORY AND IMAGING TESTING (ENLITEN): RANDOMIZED CONTROLLED TRIAL Kira L. Ryskina 1, 1; Constance J. Dine1, 1; Yevgeniy Gitelman1; Damien Leri1; Mitesh Patel2, 1; Gregory Kurtzman1; Lisa Y. Lin1; Andrew Epstein1, 1. 1University of Pennsylvania, Philadelphia, PA; 2University of Pennsylvania, New York, NY. (Control ID #2697527)

BACKGROUND: Normative feedback is an increasingly popular strategy to modify physician behavior that uses performance report cards. Little is known about the effect of normative feedback on the practice of cost-effective care by physicians-in-training (residents). Our objective was to test the effect of such feedback on the ordering of laboratory tests by internal medicine residents.

METHODS: This was a single-blinded randomized controlled trial ( #NCT02330289). Between January and June 2016, 198 resident-blocks on six general medicine teams at the Hospital of the University of Pennsylvania were cluster randomized with equal allocation to two arms: (1) those emailed a summary of their routine lab ordering (e.g. complete blood count, metabolic panel, liver function, and common coagulation tests) vs. the service average, link to a continuously updated personalized dashboard containing patient-level details, and snapshot of the dashboard; and (2) those who did not receive the intervention. The email, timed to arrive in the middle of each resident’s 2-week service to allow for baseline and follow-up periods, was followed by a reminder email. Our primary outcome was the count of routine lab orders placed by a resident per patient-day. Secondary outcomes included counts of imaging tests and non-routine lab orders. Outcomes were analyzed using an intent-to-treat approach. Weekend days were excluded. We modeled the count of orders per doctor-patient-day during the 2nd week of each resident’s service-block (post-intervention) as a function of trial arm and the resident’s order count in the 1st week (pre-intervention). The outcome was modeled using negative binomial models with adjustment for clustering within teams.

RESULTS: 114 unique residents participated (63 PGY1s and 51 higher-level residents), representing 833 resident-days. The mean count of labs per doctor-patient-day in the baseline period was 1.30 (SD 1.77) for the intervention group and 1.41 (SD 2.12) for controls (p = 0.27). We did not observe a statistically significant difference in routine lab ordering between residents who received normative feedback vs. those who did not; except for day 3 after the intervention, when orders were lower in the intervention arm by 0.44 labs per doctor-patient-day compared to controls (95% CI: 0.02, 0.85, p = 0.04). After stratifying the intervention cohort into residents who opened the summary email (74%) and those who did not (26%), we found that the change in ordering was concentrated among those residents who assessed the email. We did not observe a spillover effect to non-targeted orders.

CONCLUSIONS: A personalized dashboard reporting lab orders did not change targeted behavior among internal medicine residents, although there was a small decrease in orders among residents who accessed the information. Efforts to increase the physicians’ uptake of normative feedback during training may improve the effectiveness of normative feedback interventions to optimize physician behaviors.

EFFECTIVENESS OF A PEER NAVIGATION INTERVENTION TO MAINTAIN VIRAL SUPPRESSION AMONG HIV+ MEN AND TRANSGENDER WOMEN RELEASED FROM A LARGE MUNICIPAL JAIL: RESULTS OF A RANDOMIZED CONTROLLED TRIAL William E. Cunningham 1, 1; Terry Nakazono1; Mark Malek1, 2; Steve Shoptaw1, 1; Robert Weiss1; Susan Ettner1, 1; Nina Harawa1, 1. 1University of California, Los Angeles, Los Angeles, CA; 2Los Angeles County Sheriff’s Department, Los Angeles, CA. (Control ID #2702734)

BACKGROUND: There are 1.2 million people living with HIV (PLH) in the US with 56,000 new infections annually. HIV diagnosis, linkage to and retention in care, and adherence to antiretroviral therapy (ART) are steps in the HIV care continuum (HCC) that enable consistent viral suppression (VS), extend longevity and prevent further transmission. Criminal justice-involved populations (CJI) are important to engage in the HCC as about one in seven PLH pass through correctional facilities each year. While incarcerated, PLH receive ART and achieve VS more consistently than after release, yet no published interventions have shown sustained VS post-release. To address this gap, we conducted a randomized controlled trial of a peer navigation (PN) intervention.

METHODS: From December 2012 through June 2016 we enrolled and followed PLH being released from Los Angeles County (LAC) jail. Eligible participants were: 1. Age 18+ years; 2. Men or transgender women with documented HIV diagnoses; 3. English- or bilingual Spanish-speaking; and 4. LAC residents. At baseline, we interviewed, measured viral load (VL), and randomized PLH 1:1 to the 12-session PN intervention or usual care (UC) control. We trained lay peer navigators (PNs) to act as role models who could “walk PLH through” HCC steps. PNs assessed barriers, unmet basic needs and facilitators, and counseled on goal setting and problem solving during the first session held in jail. Post release, PNs continued counseling while they accompanied PLH to two HIV care visits, then facilitated communication with providers during the visits. The primary outcome was VS at 12 months. We used repeated measures logistic random intercept analysis to model VS outcomes over time.

RESULTS: Intervention (n = 180) and control (n = 176) arms were comparable; the sample was mostly black (42%) and Latino (31%), of low socioeconomic status, and uninsured (56%). Half (51%) had detectable VL at baseline. Study retention was high: 89% at 3 months, 80% at 6 months, and 70% at 12 months and did not differ by arm. Main findings are shown in Table 1: the adjusted probability of VS was 19% greater in the PN arm than the control arm (p < 0.01) at 12 months.

CONCLUSIONS: Although VS did not increase significantly from baseline to follow-up, the PN intervention was successful at preventing the declines in VS typically seen after release from incarceration. The intervention maintained VS over 12 months compared to the declines observed in the controls. While our PN intervention is likely to help PLH maintain VS after release from other large municipal jails, future research should examine ways to strengthen the intervention so that VS levels are increased above baseline levels.

Table 1. LINK LA: Adjusted probabilities (95%CI) of undetectable HIV viral load (VL)

Study Arm Baseline 3 Months 12 Months
PN Intervention (n = 180) 0.49 (0.39, 0.59) 0.53 (0.42, 0.63) 0.49 (0.37, 0.61)
UC Controls (n = 176) 0.52 (0.42, 0.62) 0.37 (0.27, 0.47)* 0.30 (0.21, 0.42)**

*p < =.05; **p < =0.01


BACKGROUND: Because studies have shown worsening of outcomes during the transition from pediatric to adult-oriented care, increasing emphasis has been placed on improving this transition for adolescents and young adults with chronic illness. Steps in the transition process for internists include ensuring a smooth transfer to adult-care and maintaining access to care for young adults after transfer. Multiple interventions to improve transition have been studied, and systematic reviews evaluating those interventions have been conducted, though these often focused on a particular disease or a particular set of outcomes. Summarizing the available systematic reviews will serve as one means to understand the transition intervention literature more broadly. Thus, we sought to assess the effectiveness of interventions designed to improve the transition from pediatric to adult-oriented care by performing a systematic review to identify and describe published systematic reviews on the topic.

METHODS: We searched PubMed, Cochrane, PsycInfo, and CINAHL to identify relevant systematic reviews. Two authors reviewed all unique abstracts resulting from our searches, and using pre-specified inclusion and exclusion criteria, determined which articles did not meet study criteria. The remaining abstracts underwent full text review by two authors, resulting in the final set of articles for inclusion in the review. We summarized the included articles qualitatively and evaluated their quality using AMSTAR (A MeaSurement Tool to Assess systematic Reviews) criteria.

RESULTS: The searches resulted in 347 unique abstracts. Of those, 9 systematic reviews were found to meet all inclusion criteria. The included systematic reviews found 48 unique studies assessing interventions to improve transitional care. Only 2 of the 48 studies were RCTs. Most studies used a pre-post design or a historical control group for comparison. Studies used a variety of outcomes, including patient satisfaction with the process, clinic attendance, and disease-specific parameters, like HbA1C. None of the reviews found studies addressing transition in the primary care setting, and only 6 addressed interventions in adult-oriented settings, while the rest were set in pediatric clinics (n = 12) or combined pediatric and adult settings (n = 30). The included systematic reviews had scores of 4–8 using AMSTAR criteria, suggesting fair to good quality.

CONCLUSIONS: We found that very few RCTs have been conducted to address the effectiveness of interventions to improve transition from pediatric to adult-oriented care. Importantly, we found no studies to address transition in primary care settings and most studies in this area were conducted in pediatric or combined settings. RCTs or other comparative studies are needed to determine best practices for transition. More studies are needed assessing transition interventions in adult-oriented care settings. Studies addressing transition in primary care settings are also needed.

EFFECTIVENESS OF OPIOID THERAPY VERSUS NON-OPIOID MEDICATION THERAPY FOR CHRONIC BACK AND OSTEOARTHRITIS PAIN OVER 12 MONTHS: A PRAGMATIC RANDOMIZED TRIAL Erin E. Krebs 3, 4; Siamak Noorbaloochi3, 4; Matthew J. Bair2, 1; Amy Gravely3; Agnes C. Jensen3; Kurt Kroenke1, 2. 1Indiana University, Indianapolis, IN; 2Roudebush VA, Indianapolis, IN; 3Minneapolis VA Health Care System, Minneapolis, MN; 4University of Minnesota, Minneapolis, MN. (Control ID #2693271)

BACKGROUND: Chronic back pain and osteoarthritis (OA) pain are among the most prevalent and disabling conditions in primary care. Although these conditions are often treated with opioids, no randomized trials have evaluated effectiveness of long-term opioid therapy. The objective of the Strategies for Prescribing Analgesics Comparative Effectiveness (SPACE) trial was to compare benefits and harms of opioid therapy versus non-opioid medication therapy over 12 months.

METHODS: SPACE is a 12-month single-masked pragmatic randomized trial. Eligible patients were seen in VA primary care for moderate-severe chronic back pain or hip/knee OA pain despite analgesic use. Participants were randomized to 1 of 2 active treatment arms—opioid therapy or non-opioid medication therapy—using a computer to conceal allocation. Multiple FDA-approved drugs were included in each arm. Within each arm, medications were adjusted to target pain and functional goals. Masked assessors collected outcomes at 3, 6, 9, and 12 months. Primary and secondary pain outcomes were pain-related function (Brief Pain Inventory Interference; BPI-I) and pain intensity (BPI Severity; BPI-S), respectively. The primary harm outcome was an adverse symptom count. Individual pain response was 30% improvement from 0 to 12 months. Analysis was intent-to-treat. Primary comparisons of means used mixed models controlling for baseline values. Response rate comparisons used chi-squares.

RESULTS: Of 240 randomized participants, 97.5% (n = 117 in each arm) completed 12-month outcomes. Mean age was 57.8 years (range 21–80), 87.9% were male, 85.5% were white, and primary pain location was back in 65 and knee/hip in 35%. At baseline, mean BPI-I and BPI-S scores were 5.4 (SD 1.8) and 5.4 (1.5) in the opioid arm and 5.6 (2.0) and 5.4 (1.2) in the non-opioid arm. Over 12 months, pain-related function improved in both arms; mean 12-month BPI-I scores did not differ (3.4 opioids vs. 3.2 non-opioids; p = 0.584). The proportion with pain-related function response was 59.0% for opioids and 60.1% for non-opioids (p = 0.789). Pain intensity improved more in the non-opioid arm, with 0.47 points (95% CI: 0.909, 0.035) greater reduction in mean BPI-S score over 12 months; mean 12-month BPI-S was 4.0 opioids vs. 3.5 non-opioids (p = 0.034). The proportion with pain intensity response was 41.0% for opioids and 53.9% for non-opioids (p = 0.049). The adverse symptom count did not differ between arms at 12 months (mean 6.9 opioids vs. 6.7 non-opioids, p = 0.682).

CONCLUSIONS: In this pragmatic trial designed to be maximally applicable to primary care treatment decision-making, long-term pain outcome data showed no significant advantage of opioid therapy compared with non-opioid medication therapy. In the context of prior studies that have documented higher rates of serious harms among patients receiving opioid therapy, our findings support the recent CDC recommendation that non-opioid therapies are preferred over opioids for chronic pain.

EFFECTS OF DISCONTINUING LONG-TERM OPIOID THERAPY IN PATIENTS WITH CHRONIC PAIN Jawad M. Husain 1, 2; Marc Larochelle1, 2; Julia Keosaian1, 3; Olivia Gamble1, 3; Ziming Xuan1, 3; Karen E. Lasser1, 2; Jane M. Liebschutz1, 2. 1Boston Medical Center, Boston, MA; 2Boston University School of Medicine, Boston, MA; 3Boston University School of Public Health, Boston, MA. (Control ID #2700245)

BACKGROUND: Opioid prescribing guidelines for chronic pain recommend discontinuing opioid therapy if the risk-benefit ratio becomes unfavorable. However, reasons for discontinuing opioids and effects of discontinuation are largely unknown. Providers express concern for losing patients to follow-up, resulting in untreated pain or addiction and opioid-seeking from alternate sources.

METHODS: The Transforming Opioid Prescribing in Primary Care (TOPCARE) study is a cluster-randomized controlled trial to improve opioid prescribing among patients on chronic opioid therapy for non-cancer pain. We analyzed data from an urban safety-net hospital primary care practice participating in the trial. Primary care providers (PCPs) randomized to the intervention were supported by a nurse care manger, electronic registry, academic detailing, and electronic tools; control PCPs received electronic tools only. We conducted a retrospective chart review of patients in both study arms whose PCPs discontinued opioids by the final 60 days of the 12 month intervention period, as noted in the electronic health record (EHR). Key outcomes from the EHR audit include: reason for discontinuation; ≥1 PCP visit and ≥1 pain-related emergency department (ED) visit within 6 months after discontinuation. We dichotomized discontinuation reasons as misuse/aberrant monitoring or other, which included pain resolved, inadequate analgesia, and adverse effects. We used chi-square tests to compare the frequency and reason for discontinuation between intervention and control patients. We used unadjusted logistic regression to determine if study arm or reason for discontinuation was associated with PCP follow-up or pain-related ED visits.

RESULTS: Intervention patients discontinued opioids more frequently (60/331 [18%]) compared with control patients (28/233 [12%]; p = 0.049). Misuse/aberrant monitoring was the most common reason for discontinuation, and was more frequent among intervention (41/60 [68%]) vs. control patients (13/28 [46%]; p = 0.049). Similar proportions of patients discontinued due to other reasons in both trial arms. In the 6 months following opioid discontinuation, 70% of the patients returned for ≥1 PCP visit and 17% had a pain-related ED visit. Intervention patients were less likely to have PCP follow-up after discontinuation (odds ratio [OR] 0.3; 95% CI 0.1–0.9), and more likely to have pain-related ED visits (OR 8.2; 95% CI 1.0–66.0) than control patients. Discontinuation due to misuse/aberrant monitoring was not associated with PCP follow-up (OR 0.8; 95% CI 0.3–2.0) or pain-related ED visits (OR 1.9; 95% CI 0.6–6.6).

CONCLUSIONS: The TOPCARE intervention led to higher rates of opioid discontinuation attributable to misuse or aberrant monitoring. However, intervention patients discontinued from opioids were less likely to return for follow-up and more likely to seek care in the ED for pain. Future research should evaluate strategies to manage pain and keep patients engaged in care after opioid discontinuation.

EFFECTS OF HEALTH LITERACY, PATIENT ACTIVATION, AND COGNITION ON HEALTH OUTCOMES OF OLDER ADULTS Marina Arvanitis; Rachel O’Conor; Julia Yoshino Benavente; Laura M. Curtis; Michael S. Wolf. Northwestern University, Chicago, IL. (Control ID #2707612)

BACKGROUND: Limited health literacy (HL) has been repeatedly linked to poor health outcomes, yet interventions to address HL-associated disparities have been variable or ineffective. To address this, some have proposed expanding HL to include factors such as patient activation. We aimed to assess the degree to which patient activation and cognition may explain associations between HL and common intermediate health outcomes in older adults.

METHODS: We analyzed baseline data from Health Literacy and Cognition in Older Adults (LitCog): a prospective study of 900 adults, ages 55–74 from participating health centers in Chicago. During structured interviews, participants completed assessments of HL (Test of Functional HL in Adults (TOFHLA), Newest Vital Sign (NVS), Rapid Estimate of Adult Literacy (REALM)), patient activation (Patient Activation Measure (PAM)), and cognition (fluid cognitive abilities). We assessed participants’ diabetes (DM) control by Hemoglobin A1c (HbA1c), blood pressure (BP) control by average systolic and diastolic pressures, and renal function by estimated glomerular filtration rate (eGFR) measurements from the medical record. We used bivariate statistics to describe participant characteristics, HL, activation, and cognition, and assess the relationships between these factors and health outcomes. We used a series of logistic regression models to examine the effects of patient activation and cognition, both individually and together, on the relationship between HL and each health outcome.

RESULTS: Nine hundred participants completed baseline interviews; their mean age was 63, 63% were female, and 71% were high school or college graduates. HL was limited in 32% by TOFHLA, 54% by NVS, and 28% by REALM. Most (84%) were highly activated per PAM, which did not vary by HL. Of participants with DM (n = 166), 35% had poor DM control (HbA1c >7%). Of all participants with available data, 6% had poor BP control (>140/>90), and 27% had renal dysfunction (eGFR <60). Limited HL by NVS and TOFHLA was significantly associated with all poor health outcomes. Using the TOFHLA, we found that the associations between limited HL, poor DM control, poor BP control, and renal dysfunction were largely explained by cognition, which attenuated the odds of poor DM control by 54% (OR 2.8 (95% CI 1.4,5.5) to 1.3 (0.3,3.3)), poor BP control by 24%, and poor renal function by 19%. The inclusion of patient activation did not affect the associations between HL and health outcomes, nor the attenuation of these associations by cognition. We completed the same analyses for HL by NVS and REALM, with similar findings.

CONCLUSIONS: Despite its growing links to health outcomes, patient activation does not appear to influence the associations between limited HL and poor health outcomes in older adults, which are largely explained by cognition. Interventions aimed at reducing HL-associated disparities in the elderly should focus on reducing the cognitive burdens placed on patients by the health care system.

EFFECTS OF THE AFFORDABLE CARE ACT’S MEDICAID EXPANSION ON ACCESS TO CARE BY RACE AND ETHNICITY John P. Shelley; Emily Levitan; Favel Mondesir; April Agne; Wilson Smith; Maria Pisu; Yufeng Li; Meredith Kilgore; Janet Bronstein; Andrea Cherrington. University of Alabama at Birmingham, Birmingham, AL. (Control ID #2697548)

BACKGROUND: Racial disparities in access to care have been well documented. The Affordable Care Act has the potential to mitigate these disparities with Medicaid expansion having the greatest prospect due to disproportionately high levels of poverty in minority populations. The expansion has been shown to increase access, but few have focused on disparities and none have focused on comparisons between changes by state expansion status. Our analysis tests the hypotheses that (1) access has improved for all groups regardless of race, ethnicity or expansion status and that (2) greater improvements occurred among minorities and in expansion states.

METHODS: We pooled cross-sectional data from the 2011–2015 Behavioral Risk Factor Surveillance System. Our sample includes adults ages 18–64 identifying as either white, black, or Hispanic of any race, further stratifying the Hispanic population by questionnaire language to capture heterogeneity. The expansion group (n = 668,064) included 24 states and the District of Columbia and the non-expansion group (n = 615,863) included 21 states; states expanding in 2014 or 2015 were excluded. Respondents interviewed in 2011–2013 and 2014–2015 were coded as pre- and post-implementation, respectively. Outcome variables were change in proportion of respondents having health insurance, a personal physician, not delaying care in the past year, and high-need respondents (reporting fair/poor health or a chronic condition) having a routine check-up in the past year. Difference-in-difference analyses were unadjusted to measure changes in observed disparities.

RESULTS: Access to care increased significantly for nearly all subgroups, regardless of race or expansion status, but were generally larger in minority groups. Expansion was associated with significantly greater increases in proportion of respondents insured in the white (+1.1%, p < 0.001), black (+0.9%, p < 0.001), and English-preferring Hispanic (+2.5%, p < 0.001) populations. Conversely, the change for insured Spanish-preferring Hispanics was significantly lower in expansion states (−2.1%, p = 0.047). Expansion was also associated with significantly greater increases in having a personal physician in the white (+0.7%, p = 0.049), black (+2.5%, p = 0.005), and English-preferring Hispanic populations (+3.3%, p = 0.009). Increases in respondents without cost-related care delays were only higher in expansion states for white respondents (+0.5%, p < 0.001). Increases in high need respondents with a check-up were also higher in expansion states but only significantly so for white respondents (+1.8%, p < 0.001). All other differences between states by status were not significant.

CONCLUSIONS: As a whole, the Affordable Care Act has improved access to care among all racial and ethnic groups with greater improvements in minority populations. Improvements in access associated with expansion were larger for some populations suggesting that expanding Medicaid can be an important step for improving access overall and moving towards equity.

EFFICACY AND SAFETY OF CANAGLIFLOZIN IN PATIENTS WITH TYPE 2 DIABETES BASED ON HISTORY OF CARDIOVASCULAR DISEASE OR CARDIOVASCULAR RISK FACTORS Michael J. Davies 1; Katherine Merton1; Ujjwala Vijapurkar2; Maurice Cuffee1; Jacqueline Yee2; Rong Qiu2. 1Janssen Scientific Affairs, LLC, Titusville, NJ; 2Janssen Research & Development, LLC, Raritan, NJ. (Control ID #2698864)

BACKGROUND: Treatment of patients with type 2 diabetes mellitus (T2DM) and a history of cardiovascular (CV) disease or CV risk factors may present clinical challenges due to the presence of comorbid conditions and use of concomitant medications. Canagliflozin (CANA), a sodium glucose co-transporter 2 (SGLT2) inhibitor, has been shown to improve glycemic control, body weight, and blood pressure (BP) with a favorable tolerability profile in a broad range of patients with T2DM. This post hoc analysis assessed the efficacy and safety of CANA in patients with T2DM based on CV disease history/risk factors.

METHODS: These analyses were based on pooled data from four 26-week, placebo (PBO)-controlled, Phase 3 studies of CANA 100 and 300 mg in patients with T2DM (N = 2313; mean A1C, 8.0%; body weight, 89 kg; systolic BP [SBP], 128 mmHg). Changes from baseline in A1C, body weight, and SBP at Week 26 were assessed in subgroups of patients based on history of CV disease (Y/N), history of hypertension (Y/N), baseline statin use (Y/N), and number of CV risk factors (0/1 vs ≥2). Safety was based on adverse event (AE) reports.

RESULTS: CANA 100 and 300 mg lowered A1C, body weight, and SBP versus PBO in patients across subgroups. At Week 26, A1C reductions with CANA 100 and 300 mg relative to PBO were generally similar in patients with history of CV disease (−0.95 and −1.07%) versus no CV disease (−0.71 and −0.90%), history of hypertension (−0.72 and −0.89%) versus no hypertension (−0.73 and −0.95%), baseline statin use (−0.77 and −0.99%) versus no statin use (−0.69 and −0.85%), and ≥2 CV risk factors (−0.74 and −1.02%) versus 0/1 CV risk factor (−0.72 and −0.87%). Similar body weight and SBP reductions were also seen with CANA versus PBO across subgroups. Incidence of AEs, AEs leading to discontinuation, and serious AEs was similar across subgroups.

CONCLUSIONS: CANA was efficacious and generally well tolerated in patients with T2DM regardless of CV disease history/risk factors.

EFFICACY OF MULTI-LINGUAL MOBILE APPLICATIONS TO PROMOTE PATIENT-PROVIDER COMMUNICATION AMONG ASIAN AMERICANS Judith M. Walsh 1; Janice Y. Tsoh1; L. E. Goldman2; mandana khalili2; Ginny gildengorin1; Tung T. Nguyen1. 1UCSF, San Francisco, CA; 2Zuckerberg San Francisco General Hospital, San Francisco, CA. (Control ID #2699407)

BACKGROUND: Asian Americans (AA) are the fastest growing racial group in the U.S. and are high users of technology; few studies have evaluated the impact of mobile technology on health outcomes.

METHODS: Using patient-centered methods, we developed 2 mobile applications (app) for the iPad in English, Cantonese, Mandarin, and Vietnamese. One app addressed hepatitis B and C (HEP), the other addressed healthy Nutrition, Physical Activity and weight (NPA). Primary care providers (PCP) and their patients were randomized to either HEP or NPA. Eligible patients were AA age 18 and older who spoke one of the 4 languages, had not had HEP B screening as per USPSTF recommendations, and who received primary care at UCSF or ZSFG. Patients used the app prior to a PCP visit. The app included questions leading to audio-visual messages delivered by doctors to address the patient’s response. A printout summarizing the patient’s concerns and recommendations was generated for the provider. Data were collected via survey prior to the use of the app and immediately after the visit. General linear models adjusted for provider clusters were used to compare the 2 arms on participants’ perceived helpfulness of the intervention, self-report of patient-provider discussions of health topics during the visit, and provider’s recommendations of HEP screening tests.

RESULTS: The sample included 362 AA (41% ZSFG and 59% UCSF) with a mean age of 58.4 years (range: 18–93), 63% females, 59% married, 33% with high school or less education, and 39% currently employed. The majority (81%) were born outside of the U.S. including 33% from China, 24% from the Philippines and 7% from Vietnam; 39% were limited English proficient. A majority (76%) had heard of HEP B; 53% had heard of HEP C. 220 participants (of 60 PCPs) were assigned to HEP and 142 (of 48 PCPs), to NPA. A majority (85%) completed the entire video and 55% reported giving the printout to their provider at the visit. Participants in both groups liked the app they used (81%), perceived it as “somewhat” or “very” helpful (80.5%), and thought it helped them to talk with their PCP about their health (85.5%). Compared to NPA, higher proportions of participants in the HEP group reported having had a discussion with their PCP about HEP B (73% vs. 21%, p < 0.001) or HEP C (66% vs. 17%, p < 0.001). Compared to HEP, NPA participants were more likely to report a PCP discussion on nutrition (56% vs. 34%, p < 0.001), physical activity (60% vs. 42%, p < 0.01), and weight (56% vs. 31%, p < 0.002). HEP participants were more likely than the NPA participants to report that their provider recommended testing for HEP B (57% vs 17%, p < 0.001) or HEP C (42% vs. 13%, p < 0.001).

CONCLUSIONS: Patient-centered multi-lingual mobile applications for health promotion were well received by AA in two healthcare settings and led to more patient-provider discussion. Further research is needed to assess the impact of the mobile applications on behavior and health outcomes.

EFFICACY VERSUS EFFECTIVENESS OF DIABETIC MEDICATIONS Anish Vani; Keith Goldfeld; Michael Cantor. NYU School of Medicine, New York, NY. (Control ID #2672008)

BACKGROUND: Diabetic medications are approved by the Food and Drug Administration (FDA) based on efficacy in controlled clinical settings. Real world effectiveness may often vary from the results of clinical trials. We attempted to quantify the “efficacy versus effectiveness” gap at our institution.

METHODS: We performed a retrospective analysis of all patients prescribed oral diabetic medications at New York University Langone Medical Center, a large urban tertiary care center, using data extracted from our electronic health record from October 2009 to September 2014. We limited the dataset to include newly diagnosed diabetic patients who were started on monotherapy or a combination pill. We excluded patients who were prescribed insulin or multiple diabetic agents. We used two-tailed Welch’s t-tests to compare the initial Hemoglobin A1c (HbA1c), final HbA1c, and total HbA1c reduction to clinical trial data used for FDA approval, with statistical significance defined as p < 0.05.

RESULTS: Among the 566 of the 2452 patients who met our inclusion criteria, the mean age was 62.9 +/− 12.2 years, the majority of whom were male (51.1%), white (61.1%), non-Hispanic (80.4%), and prescribed metformin monotherapy (70.3%). The mean treatment time for each drug class ranged from 32–39 weeks. The mean initial HbA1c for all groups was 8.95%, the mean final HbA1c was 7.44%, and the mean HbA1c reduction was 0.51%. For almost all drug classes, the initial HbA1c was lower than the initial HbA1c from efficacy trials. For most drug classes, effectiveness was about half of the expected value from efficacy trials. Most patients achieved their goal HbA1c levels [Table].

CONCLUSIONS: In our study population, real world effectiveness did not reflect measures of efficacy from controlled clinical settings. The results highlight the need for clinicians to realize that a medication may not perform as well as it does in a controlled setting and they may need to increase the intensity of treatment more quickly than expected. Also, replicating the results of clinical trials in an actual clinical setting is challenging, both from the perspective of data availability and clinical practice patterns.

figure aa

NYU versus FDA Clinical Trial HbA1c Levels

EHR-BASED MEDICATION SUPPORT AND NURSE-LED MEDICATION THERAPY MANAGEMENT: THE NORTHWESTERN AND ACCESS COMMUNITY HEALTH NETWORK MEDICATION EDUCATION STUDY (NAMES) RANDOMIZED CONTROLLED TRIAL Stephen Persell 2; Danielle Lazar3; Elisha Friesema4; Kunal N. Karmali2; Ji Young Lee1, 2; Alfred Rademaker1; Michael S. Wolf1, 2. 1Northwestern University, Chicago, IL; 2Northwestern University, Evanston, IL; 3Access Community Health Network, Chicago, IL; 4University of Minnesota, Minneapolis, MN. (Control ID #2701908)

BACKGROUND: Nurse-led medication management in primary care may improve medication use. Electronic health record (EHR) tools may offer low-cost approaches to support medication management.

METHODS: We performed a three-arm 12-clinic-level cluster-randomized trail in a Chicago area network of federally qualified community health centers that compared: usual care, EHR-based medication management tools alone (printed medication lists to prompt review at each visit and automated plain-language information after the visit), and EHR tools plus nurse-led medication management support (one-on-one counseling about medication regimens to clarify discrepancies and identify medication-taking concerns plus follow-up telephone calls after new prescriptions to perform medication review). We recruited patients from randomized clinics who were ≥18 years of age with ≥3 prescribed medications and suboptimal blood pressure (≥130/80 mm Hg if diabetes or ≥135/85 mm Hg diastolic if not). Outcomes at 1 year included systolic blood pressure (primary outcome, using a standard protocol), blood pressure control (<140/90 mm Hg), basic understanding of medication indications, reconciliation between patient-report and the medical record, and self-reported 4-day adherence. Analyses of study arm effects used generalized linear models accounting for subjects’ baseline outcome values (which sometimes differed by study arm) and clinic-level random effects.

RESULTS: 920 participants enrolled; 796 completed 1-year follow up. Of subjects completing follow up, mean age was 53 years, 69% were women, 87% were African American, 67% had high school education or less and 47% had limited health literacy. Subjects used a mean of 5.4 medications and 25% used 7 or more. At 1 year, systolic blood pressure was higher with EHR-tools only (+3.3 mm Hg [95% CI 0.2 to 6.5]) vs usual care and lowest in the EHR + nurse arm (−5.5 mm Hg [−3.1 to −7.8]) vs EHR-only, (−2.1 mm Hg [−5.4 to +1.1]) vs usual care. Compared to usual care, blood pressure control at 1 year was better in the EHR + nurse arm (odds ratio [OR] 1.21 [1.01-1.45]) but worse in the EHR-only arm (0.68 [0.58-0.78]). Understanding of indications was not changed with either intervention compared to usual care. Medication reconciliation substantially improved with both interventions, OR (95% CI) compared to usual care for antihypertensives, chronic disease medications, and all medications with EHR-tools only: 1.7 (1.2-2.4), 2.3 (1.3-4.2), and 4.6 (1.7-13), respectively and for EHR tools + nurse: 1.9 (1.4-2.7), 2.1 (1.2-3.5), 4.9 (1.6-15), but did not significantly differ between active intervention groups. Medication adherence was worse than usual care in the EHR-tools only arm for all chronic disease medications, OR 0.83 (0.71-0.98).

CONCLUSIONS: Both interventions improved medication reconciliation. Only nurse-led medication management plus EHR tools improved hypertension control. EHR tools only may have had unintended effects on blood pressure control.

ELECTRONIC LEARNING IN GRADUATE MEDICAL EDUCATION: A NATIONAL SURVEY OF RESIDENCY PROGRAM DIRECTORS Christopher M. Wittich 2; Anoop Agrawal1; David A. Cook2; Andrew J. Halvorsen2; Jayawant N. Mandrekar4; Saima Chaudhry3; Denise Dupras2; Amy Oxentenko2; Thomas J. Beckman2. 1Baylor College of Medicine, Houston, TX; 2Mayo Clinic College of Medicine, Rochester, MN; 3Memorial Healthcare System, Hollywood, FL; 4Mayo Clinic, Rochester, MN. (Control ID #2706097)

BACKGROUND: E-learning - the use of computers, mobile devices, and internet to learn - is a prevalent instructional approach. However, little is known about the role of e-learning in residency education. Our goal was to determine utilization of e-learning by United States (U.S.) internal medicine residency programs, program director (PD) perceptions of e-learning, and associations between residency program characteristics and e-learning utilization.

METHODS: This was a national survey performed in collaboration with the Association of Program Directors in Internal Medicine of 368 internal medicine residency programs (92.9% of the 396 U.S. programs) in 2016. Program directors were asked about utilization, resources for implementation, and perceptions of e-learning. To assess representativeness of the programs sampled, characteristics of survey responders were compared with survey non-responders for 5 publicly-available variables using the Fisher exact test or Welch t-test, as appropriate. Associations between PD characteristics and PD perceptions of e-learning scores were assessed using a multiple analysis of variance (ANOVA) model. Multiple logistic regression was used to generate odds ratios (ORs) and test associations between program characteristics and the regular use of synchronous and asynchronous e-learning. For the continuous predictors of ACGME-approved positions, ABIM 3-year rolling pass rate, percentage of positions filled by international medical graduates, number of hospital beds, and mean PD perception of e-learning score, the adequacy of bivariate models assuming linearity of the log odds was checked using Hosmer-Lemeshow goodness-of-fit tests.

RESULTS: A total of 214 (58.2%) PDs completed the e-learning survey and 85 (39.7%) reported that they utilize synchronous e-learning at least sometimes, somewhat often, or very often. More programs (153, 71.5%) reported that they utilize asynchronous e-learning at least sometimes, somewhat often, or very often. Most programs (N = 168, 79%) did not have a budget to integrate e-learning into their educational curricula. Mean (SD) scores for the PD perceptions of e-learning ranged from 3.01 (0.94) to 3.86 (0.72) on a 5-point scale; the overall mean (SD) score was 3.45 (0.54). There was a higher odds of synchronous e-learning in programs that had a budget for implementation (OR, 3.0 [95% CI, 1.04–8.7]; p = 0.04).

CONCLUSIONS: U.S. residency programs could be better resourced to integrate e-learning technologies. Asynchronous e-learning was used more than synchronous, which may be a result of busy resident schedules and duty hour restrictions. Programs use locally and externally-developed e-learning resources with roughly similar frequency. PD perceptions of e-learning, while above the scale median, are only modest. Future work should determine why PDs are hesitant to integrate e-learning despite evidence of its effectiveness.

ELECTRONIC MEDICAL RECORD DOCUMENTATION HAS POOR AGREEMENT WITH PATIENT SURVEYS FOR RECEIPT OF INPATIENT TOBACCO CESSATION CARE. Ethan Kuperman 2, 3; Jennifer Chapin4; David A. Katz1. 1University of Iowa, Iowa City, IA; 2University of Iowa Carver College of Medicine, Iowa City, IA; 3Iowa City VA Medical Center, Iowa City, IA; 4Iowa VA Medical Center, Iowa City, IA. (Control ID #2702062)

BACKGROUND: Joint Commission (TJC) created a set of ORYX quality measures to assess inpatient tobacco treatment, as care provided during and after admission may increase long term cessation. The objective of this study was to quantify the agreement between documented tobacco treatment in the electronic medical record (EMR) and patient-reported receipt of ORYX-concordant care in Veterans Administration (VA) inpatients.

METHODS: We retrospectively reviewed the EMR of daily smokers admitted to a general medical inpatient unit at 2 VA medical centers and were previously enrolled in a guideline implementation trial. Patients hospitalized for < 72 hours, significant cognitive impairment, or with terminal illness were excluded. Three measures were evaluated using TJC audit criteria: 1) TOB-1: Screened for tobacco use within 24 hours of admission. 2) TOB-2: Offered intensive cessation counseling AND pharmacotherapy during admission. 3) TOB-3: Offered cessation counseling AND pharmacotherapy at discharge. We compared the EMR audit data with the corresponding patient responses from a survey administered within 72 hours of discharge during the trial. Chart reviewers were blinded to survey data. Agreement between EMR audits and patient surveys was calculated using Cohen’s κ.

RESULTS: Results: The analysis sample included 138 patients (mean age = 59, 97% male). Overall 7-day point-prevalence abstinence at 6-month follow-up was 26%. Over 90% of patients met TOB-1 for both EMR audit and patient survey, but κ was <0.01. No patient met TOB-2 on EMR review and only 25% passed by patient survey. For TOB-3, 14.6% passed on EMR audit and 13.0% passed on patient survey (κ = 0.25).

CONCLUSIONS: Conclusions: Performance on TJC tobacco cessation measures from the EMR did not correlate with patient-reported tobacco treatment. Changes in documentation templates and increasing staff awareness may improve correlation between metric performance and care delivery.

ELEVATED HBA1C IN UNITED STATES VETERANS AND RISK OF INCIDENT DIABETES AND ALL-CAUSE MORTALITY Jordan Davis 5; Mengling Liu5, 1; Farrokh Alemi3; Scott Sherman1, 4; Sundar Natarajan1, 4; Ashley Jensen2; Sanja Avramovic3; Esther Levy5; Richard B. Hayes5, 1; Mark D. Schwartz1, 4. 1NYU School of Medicine, New York, NY; 2University of Calgary, Calgary, AB, Canada; 3George Mason University, Fairfax, VA; 4Department of Veterans Affairs New York Harbor, New York, NY; 5New York University, New York, NY. (Control ID #2704151)

BACKGROUND: United States Veterans are at excess risk for Type 2 diabetes and early mortality. Our objective is to determine the impact of prediabetes and related risk factors on the occurrence of diabetes and mortality in this at-risk population.

METHODS: At the primary care practices of the VA New York Harbor (VA NYHHS) during 2004–2014, we identified 15,173 diabetes-free Veterans, among those who received 2 or more hemoglobin A1c tests (HbA1c). Among these participants, we identified 14,361 veterans with HbA1c values below the diabetic range (i.e., <6.5% HbA1c) and characterized these individuals with respect to selected risk factors. We followed these individuals through 2014 for incident diabetes and all-cause mortality. Cox proportional hazard regression was used to relate HbA1c levels, age, sex, race/ethnicity, anthropometric measures, and comorbid cardiovascular conditions (ischemic heart disease, cerebral vascular accident, congestive heart failure and peripheral vascular disease) to incident diabetes and all-cause mortality (Hazard Ratio [HR] and 95% confidence intervals).

RESULTS: Among 8,145 Veterans with prediabetes (HbA1c 5.7–6.4%), 1,170 (14.4%) developed diabetes and 1,139 (14%) died during the course of follow-up. Compared to 5,292 normoglycemic Veterans (HbA1c: 5.0–5.6%), 4,207 prediabetics in the moderate risk group (HbA1c 5.7–5.9%) had a greater than 2-fold increased risk of incident diabetes (HR 2.46 [2.08–2.92]), and those 3,938 in the prediabetic high risk group (HbA1c 6.0–6.4%) had a greater than 5-fold risk (HR 5.70 [4.88–6.65]). Furthermore, all-cause mortality was increased in 924 participants with low glycemia (HbA1c <5.0%: HR 1.40 [1.17–1.68]) and among those 812 in the diabetes risk range (HbA1c ≥6.5%: HR 1.44 [1.22–1.71]) compared with the normoglycemic group. Excess all-cause mortality was not observed among the prediabetic group, compared to the normoglycemic group.

CONCLUSIONS: Among Veterans, prospective risk of transition to Type 2 diabetes ranged from 2.5-fold to 5.7-fold among prediabetics, depending on HbA1c level. Patients with HbA1c < 5.0 and those in the diabetic range had increased risks of all-cause mortality, while prediabetics showed no excess mortality. The higher risk population (HbA1c 6.0–6.4%) is an important group to target with diabetes prevention efforts.

ELUCIDATING MEDICAL TRAINEES’ BARRIERS TO INCIDENT REPORTING Jose R. Valery 2; Fernando Stancampiano1. 1Mayo Clinic, Jacksonville, FL; 2Mayo Clinic Florida, Ponte Vedra, FL. (Control ID #2703654)

BACKGROUND: Incident reporting is an important tool used by hospitals to improve patient care by systemically eliminating errors. Although medical trainees are at the forefront of healthcare delivery their incident reporting is very low. The Department of Health and Human Services released a report noting that 86% of events are unreported [1]. As low as 6.2% of medical trainees complete an incident report [2]. The ACGME has recognized the importance of including patient safety in medical education and launched the Clinical Learning Environment Review (CLER) Program in 2012 to stimulate such initiatives. Elucidating reasons for medical trainees’ reporting habits is essential in developing a successful educational intervention. In this study, we aim to refine existing knowledge on resident attitudes and behaviors pertaining to incident reporting. We hypothesize that non-technical aspects of incident reporting, such as concern about negative consequences or lack of faith in the effectiveness of the system, are significant in addition to technical barriers such as knowing how to report and having enough time to do so.

METHODS: Medical trainees’ attitudes and behaviors regarding incident reporting were evaluated through an anonymous and voluntary REDCap survey distributed via email to 207 medical and surgical trainees at the Mayo Clinic in Jacksonville, FL, in June of 2016. The trainees had between 11 and 83 months of clinical experience.

RESULTS: 24.6% (51/207) of trainees responded to the survey after 3 distributions. 17.6% (9/51) of respondents were interns and 68.9% (35/51) were male. 25.4% (13/51) had previously filed a report and 49% (25/51) reported not knowing how to file a report. Fear of personal and colleague repercussions from filing a report were significant with 35.2% (18/51) and 39.2% (20/51) respectively. Only 7.8% (4/51) thought filing a report would not contribute to patient safety. 37% (19/51) indicated that time was a significant barrier to reporting.

CONCLUSIONS: Our study suggests that technical aspects of incident reporting are significant and seem to be consistent with recently published data. However, the significance of non-technical aspects as barriers was found to be higher in our trainee population than previously reported [3]. Further research is needed to confirm the relative significance of barriers to incident reporting among medical trainees. When compared with published studies, our results suggest a significant heterogeneity in known barriers across institutions. As the generalizability of results may be limited, each institution may benefit from a similar survey to elucidate specific barriers prior to developing an educational intervention.

EMBEDDED DECISION SUPPORT DOES NOT LEAD TO RECOMMENDED HEPARIN DOSING Sarah W. Baron; William Southern. Montefiore, Bronx, NY. (Control ID #2704934)

BACKGROUND: Anticoagulation with heparin decreases recurrent thromboembolic events, but delays to therapeutic levels are associated with worse outcomes. In previous research using paper charting, a weight-based Heparin Nomogram improved time to therapeutic anticoagulation among hospitalized patients. However, it is unclear if the Heparin Nomogram remains effective in computerized patient order entry (CPOE) systems. We examined the effect of incorporating the Heparin Nomogram in a CPOE system on anticoagulation-related process measures among adult inpatients.

METHODS: We conducted a retrospective pre/post analysis of a validated decision-support tool, the Heparin Nomogram, by embedding it into the CPOE system of a large, academic, multi-hospital system in the Bronx, NY. Following implementation, all access to intravenous heparin orders displayed titration guidelines, weight and partial thromboplastin time (PTT) measurements. The primary outcome was time to therapeutic anticoagulation, defined as hours until PTT measurement between 46 and 70 s. Secondary outcomes were appropriateness of initial drip and initial bolus doses. All orders of intravenous heparin were examined for timing and dosing of drips and boluses, as well as PTT measurements. Time to therapeutic PTT was plotted using the Kaplan-Meier method. Differences between pre and post groups were tested using a log rank test. Percentages of each group receiving appropriate dosing were examined. Differences between groups were tested with chi-squared tests. Orders were analyzed three months prior to and three months following implementation.

RESULTS: A total of 1728 unique initial heparin orders were examined. Time to therapeutic PTT did not differ significantly between groups (Figure, p = 0.33). There were no significant differences in percent of patients receiving the appropriate initial drip dose (27% vs. 29%, p = 0.50), or bolus dose (7% vs. 8%, p = 0.50).

CONCLUSIONS: Embedding a Heparin Nomogram into the Computerized Patient Order Entry system did not significantly change time to therapeutic anticoagulation or initial heparin dosing. It remains unclear how best to use CPOE systems to support providers in dosing intravenous heparin for optimal anticoagulation among inpatients.

figure ab

EMERGENCY PHYSICIAN OPIOID PRESCRIBING PATTERNS AND RISK OF LONG-TERM USE Michael L. Barnett 2; Andrew Olenski1; Anupam B. Jena1. 1Harvard Medical School, Boston, MA; 2Harvard T.H. Chan School of Public Health, Boston, MA. (Control ID #2705830)

BACKGROUND: Rising rates of opioid prescribing and opioid-related overdose deaths have increasingly affected the elderly Medicare population, among whom hospitalization rates for opioid overdoses grew five-fold from 1993–2012. Growing opioid overuse may be partly driven by physician prescribing. However, the extent to which individual physicians vary in opioid prescribing and the implications of that variation for patients’ long-term opioid use and adverse outcomes are unknown.

METHODS: We performed a retrospective analysis of Medicare beneficiaries with an index emergency department (ED) visit during 2008–2011 without opioid prescriptions in the 6 months prior. To address selection bias, we relied on the fact that patients are unlikely to choose their ED physician once they have chosen a facility. Assigning patients to emergency physicians within a hospital, we categorized physicians as high- or low-intensity opioid prescribers based on their relative quartiles of prescribing rates within the same hospital. We compared rates of long-term opioid use, defined as 6 months of days supplied, in the 12 months subsequent to an ED visit among patients treated by high- vs. low-intensity prescribers, adjusting for patient characteristics. We also assessed rates of hospital encounters (ED visits or admissions) for opioid-related complications in older adults such as falls/fractures.

RESULTS: Our sample contained 215,678 and 161,951 patients treated by low- and high-intensity prescribers, respectively. Patient characteristics, including ED diagnoses, were similar across both groups. Opioid prescribing rates varied widely within hospital between high and low-intensity prescribers (24.1% vs. 7.3%). Overall, long-term opioid use at 12 months was significantly higher among patients treated by high vs. low-intensity prescribers, 1.51% vs. 1.16% (unadjusted OR 1.31, 95% CI 1.24–1.39). After adjustment, this difference changed minimally (adjusted OR 1.30, 95% CI 1.23–1.37). Rates of hospital encounters for falls/fracture were significantly higher in the 12 months subsequent to an index ED visit for patients treated by high- vs. low-intensity opioid prescribers (4.56% vs. 4.28%; adjusted OR 1.07, 95% CI 1.03–1.11).

CONCLUSIONS: We found over three-fold variation in emergency physician opioid prescribing rates within the same hospital associated with increased rates of long-term opioid use and hospitalizations for falls/fractures among patients treated by high-intensity opioid prescribers. We reduced selection bias by taking advantage of quasi-randomization of patients to ED physicians within the same facility, supported by the minimal change in our results with adjustment. These results suggest that an increased likelihood of receiving an opioid for even one encounter could drive significant future long-term opioid use and potentially increased adverse outcomes among the elderly.

END-OF-VISIT PRACTICES TO ENSURE OUTPATIENT SAFETY: RESIDENT PHYSICIANS’ PERFORMANCE IN USP CASES WITH OUTPATIENT SAFETY CHALLENGES Colleen Gillespie 1; Lisa Altshuler1; Kathleen Hanley1, 2; Adina Kalet1, 1; Amanda Watsula-Morley1, 1; Heather Dumorne1, 1; Sondra Zabar1, 1. 1NYU School of Medicine, New York, NY; 2Gouverneur Healthcare Services, New York, NY. (Control ID #2705472)

BACKGROUND: Safe, high quality outpatient care often depends on the degree to which patients understand their situation and how to follow through on physician recommendations. However, we do not know enough about how often physicians focus on ensuring that their patients have achieved these understandings by the end of the visit and whether such end-of-visit practices are associated with physicians’ communication, patient education and activating skills.

METHODS: Two Unannounced Standardized Patient cases (highly trained actors who present as real patients) were delivered to 71 internal medicine residents in two clinics: one required the physician to identify a patient’s depression and engage him in follow-up care, and the other required the physician to recognize a patient’s failure to use her asthma medicine correctly and educate her in using it properly. End-of-visit practices were: reviewing the plan; asking if further questions; giving information about follow-up care and further contact; and helping the patient navigate the system in order to follow through on next steps. Each was assessed by the SP as not done, partly done, or well done. SPs also rated physicians’ communication skills, patient activating skills, and case-specific education skills. Summary scores were calculated as % of items well done.

RESULTS: Close to three-quarters of the physicians reviewed the plan with the patient and invited further questions in the depression case and slightly more than half did so in the asthma case (56 and 60%). Patients were given complete information about follow-up care and how to navigate the system in just under half of depression visits (49 and 47%) and just over half of asthma visits (58 and 58%). On average, residents were rated as performing 61% of these 8 items well (SD 28%) across both cases. Primary care residents performed significantly better than categorical internal medicine residents (67% vs 47%, p = .004). There were no differences by physician gender. End of visit scores were significantly positively correlated with both general and case-specific clinical skills, and after controlling for the variance contributed by the program (R2 = 12%, p = .004), case-specific education scores explained 10% of the variance in end of visit score (p = .005), patient activating skills 10% of the variance (p = .002) and communication skills 13% of the variance (p = .001). With all variables in the model, only the general communication domain of patient education and counseling was independently associated with end of visit scores (Std Beta = .35, p = .015).

CONCLUSIONS: Had these patients been real patients, in one-quarter to one-half of the visits, the patient would have left not fully understanding the plan or how to follow-through on care. Resident physicians with more effective communication and patient activating skills tended to provide safer end-of-visit care, suggesting that these may reflect an outpatient safety orientation or skillset.

ENGAGEMENT IN AN INTERNET SUPPORT GROUP FOR MOOD AND ANXIETY DISORDERS IS ASSOCIATED WITH IMPROVED PSYCHIATRIC OUTCOMES Emily M. Rosenberger; Bea Herbeck Belnap; Kaleab Abebe; Scott D. Rothenberger; Bruce L. Rollman. University of Pittsburgh, Pittsburgh, PA. (Control ID #2687850)

BACKGROUND: Internet support groups (ISGs) are emerging as popular forums in which individuals with similar health conditions exchange resources and offer support by posting on online discussion boards, yet little is known about the role of engagement with an ISG on health outcomes. The Online Treatment for Mood and Anxiety Disorders Trial (OT Trial) evaluated the effect of computerized cognitive behavioral therapy (CCBT) plus access to a moderated ISG on clinical outcomes in primary care patients with mood and anxiety disorders delivered via a collaborative care model. Here, we examine whether differing levels of engagement in the OT Trial’s ISG impacted mental health related quality of life (HRQoL) and psychiatric outcomes.

METHODS: The OT Trial randomized 704 patients referred by their PCPs from 26 primary care practices in southwestern Pennsylvania with PHQ-9 and/or GAD-7 scores ≥10 to receive either: CCBT + ISG access, CCBT, or usual care in a 3:3:1 randomization ratio. We classified engagement among the 302 CCBT + ISG arm patients according to the 1% Rule (van Mierlo, 2014), a model that describes participation in online forums. We then assigned patients to engagement groups based on number of posts made on the ISG during the 6-month intervention phase of the study. We used descriptive statistics to compare outcomes across engagement groups on the MOS SF-12 and PROMIS Anxiety and Mood short forms at 12 months after randomization.

RESULTS: CCBT + ISG patients had a mean age of 43 years, and 71% had both depression and anxiety as measured by the PRIME-MD. 75% logged in to the ISG at least once; of these, 63% made at least one post. Among patients who posted, the mean number of posts per patient was 10.5. Our sample had a broader distribution of posts than predicted by the 1% Rule: Superusers (top 1%; n = 3) made 40% of posts (mean ± SD, 210 ± 118), Top Contributors (n = 30) made 38% of posts (mean,17 ± 13), Sparse Contributors (n = 110) made 22% of posts (mean, 3 ± 2), and Observers (n = 85) logged in but never posted. At 12 months, all engagement groups showed improvements in HRQoL and depression and anxiety symptoms relative to baseline assessments. Top Contributors reported the largest improvements in HRQoL (mean Δ ± SD, 16 ± 10) and anxiety symptoms (mean Δ, −12 ± 11), whereas Superusers reported the smallest improvement in HRQoL (mean Δ, 6 ± 8).

CONCLUSIONS: Differential engagement in a moderated ISG for mood and anxiety disorders is associated with varying levels of improvement in HRQoL and depression and anxiety symptoms. At 12 months of follow-up, Superusers, who were the most engaged with our ISG, reported the lowest levels of benefit, while Top Contributors, who engaged often but less than Superusers, reported the greatest benefit. While the motivations for very high levels of ISG engagement are still unclear, more work is needed to understand the minimum level and type of engagement necessary to yield benefits from participating in an ISG for mental health disorders. Reference: van Mierlo T, JMIR 2014;16(2)

ENGAGING AFRICAN AMERICAN VETERANS WITH HEALTHCARE ACCESS CHALLENGES IN A COMMUNITY-PARTNERED CARE COORDINATION INITIATIVE: A QUALITATIVE NEEDS ASSESSMENT Adriana Izquierdo 5, 6; Michael Ong4, 6; Felica U. Jones7; Loretta Jones2; David Ganz3, 6; Lisa V. Rubenstein1. 1GLA VA, North Hills, CA; 2Healthy African American Families, Los Angeles, CA; 3RAND Corporation, Santa Monica, CA; 4UCLA, Los Angeles, CA; 5University of California Los Angeles, Los Angeles, CA; 6VA Greater Los Angeles Healthcare System, Los Angeles, CA; 7Healthy African American Families II, Los Angeles, CA. (Control ID #2683713)

BACKGROUND: Little has been written about engaging potentially eligible members of a health care system who are not accessing the care to which they are entitled. Knowing more about the experiences of African American Veterans who regularly experience healthcare access challenges may be an important step towards equitable, coordinated Veterans Health Administration (VA) care.

METHODS: We partnered with a community organization to recruit and engage Veterans from South Los Angeles in three exploratory engagement workshops to understand better the experiences of African American Veterans at risk of experiencing poor care coordination. Veterans were asked to describe their VA and community care experiences. Field notes taken during the workshops were analyzed by community and academic partners using grounded theory methodology to identify emergent themes.

RESULTS: 23 Veterans participated in one or more engagement workshops. Their trust in VA was generally low. Negative themes included: functional barriers to accessing VA healthcare services; insensitive VA healthcare environment; lack of trust in the VA healthcare system; and Veteran status as disadvantageous for accessing non-VA community services. Positive themes included: Veterans have knowledge to share and want to help other Veterans; and connecting to VA services can result in positive experiences.

CONCLUSIONS: Veterans living in underserved areas who have had difficulty accessing VA care have unique perspectives on VA services. Partnering with trusted local community organizations to engage Veterans in their home communities is a promising strategy to inform efforts to improve care access and coordination for vulnerable Veterans.

EVALUATING GLYCEMIC CONTROL IN LOW-INCOME PATIENTS WITH TYPE 2 DIABETES Rosette J. Chakkalakal 3; Richard O. White1; Jonathan Schildcrout3; Brooklyn Stanley3; Ken Wallston4; Russell L. Rothman2. 1Mayo Clinic, Jacksonville, FL; 2Vanderbilt, Nashville, TN; 3Vanderbilt University, Nashville, TN; 4Vanderbilt University, Pisgah Forest, NC. (Control ID #2707610)

BACKGROUND: Long-term maintenance of glycemic control is difficult to achieve for patients with type 2 diabetes (T2DM). Few studies have described predictors of deterioration in glycemic control among low-income patients with T2DM. Here we (1) assess change in hemoglobin A1C (HbA1C) after a period of sustained glycemic control among low-income patients with T2DM and (2) identify predictors of deterioration in glycemic control in this population.

METHODS: We conducted an exploratory subgroup analysis of data from the Partnering to Improve Diabetes Education (PRIDE) study, a cluster randomized controlled trial testing a program designed to address limited health literacy and numeracy and promote effective health communication in caring for patients with T2DM at state Department of Health clinics in middle Tennessee. English and Spanish-speaking patients with T2DM were eligible for the study if their most recent hemoglobin A1C (HbA1C) was greater than 7.5%. Participants were scheduled for routine T2DM follow-up visits every 3 months for 2 years. This subgroup analysis was restricted to PRIDE participants who, at any point during the study, were found to (1) improve their HbA1C by at least 0.7% between 2 consecutive study visits, (2) sustain this improvement for at least 2 consecutive study visits (defined as the monitoring period), and (3) attend at least 2 of their next 3 scheduled study visits (defined as the follow-up period). We fit a linear regression model (base model) to assess the a priori association of change in HbA1C during the follow-up period with (1) change in HbA1C from baseline to the monitoring period and (2) mean HbA1C during the monitoring period. We then tested the association of change in HbA1C during the follow-up period with the following characteristics by including them, in turn, as independent variables in the base model: gender, body mass index, income, race, ethnicity, English proficiency, health literacy, numeracy, medications, years diagnosed with T2DM, depressive symptoms, food security, perceived diabetes self-management, and receipt of care at a control or intervention site.

RESULTS: One hundred PRIDE participants met the inclusion criteria. The base model indicated that HbA1C of low-income patients with T2DM increases after a period of sustained glycemic control (b0 = 0.38, SE = 0.14) but there was a significant inverse association between mean HbA1C during the monitoring period and change in HbA1C during the follow-up period (b = −0.32, SE 0.10, p = 0.002). The only other variables significantly associated with change in HbA1C during the follow-up period were English proficiency (b = 0.72, 95% CI 0.06-1.37, for patients who did not speak English very well) and treatment group (b = −0.74, 95% CI −1.29- -0.19, for patients receiving care at intervention sites).

CONCLUSIONS: Efforts to promote effective communication could play a key role in preventing deterioration in glycemic control among low-income patients with T2DM.

EVALUATING QUALITY OF DIABETIC CARE IN PATIENTS WITH SEVERE MENTAL ILLNESS (SMI) IN AN ACADEMIC PRIMARY CARE CLINIC Smita Y. Bakhai 1; Rujuta Katkar1; Neal Shah1; Gregory D. Gudleski2. 1SUNY at BUffalo, Williamsville, NY; 2SUNY at Buffalo, Buffalo, NY. (Control ID #2690472)

BACKGROUND: Higher prevalence of type 2 Diabetes Mellitus and cardiovascular risk factors has been reported in patients with serious mental illness. SMI patients get suboptimal diabetes care when compared to non-SMI patients, resulting into higher morbidity and mortality. The primary objective of this study was to test the hypothesis that SMI patients get suboptimal diabetes care as compare to non- SMI patients in an academic Primary Care Clinic.

METHODS: Utilizing electronic medical records, we conducted a retrospective cohort study of 183 SMI and 185 non -SMI patients between June 1, 2013 and May 31, 2015 in a hospital based primary care clinic, consisted of 40 residents. SMI was defined as schizophrenia or bipolar disorder diagnosed by psychiatrist. Inclusion criteria included patients between age 18–75 years and diabetes patients who were managed by a primary care provider at the ECMC Primary Care Clinic for a minimum one year with at least two visits and two HbA1c levels drawn during the study period. Diabetes patients managed by endocrinologist were excluded. Variables of interest included; 1) National Committee for Quality Assurance (NCQA) diabetes recognition program (DRP) 2015 based outcome measures included HbA1c, Blood pressure, LDL, eye and foot exam, nephrology assessment and smoking status 2) Pharmacological therapy; ACE/ARB, oral anti-hyperglycemic agents and insulin 3) Health care utilization measures ;clinic visits attended and continuity of care with the same primary care provider 4) vaccinations included pneumonia, influenza and hepatitis B.

RESULTS: Results were analyzed using univariate (chi-square) and multivariate (logistic regression) models. Patients who were on insulin therapy were less likely to achieve HbA1c < 7 without any significant difference between SMI and non-SMI groups (12.5% vs 11.8%). However SMI patients who were not on insulin were more likely than non-SMI patients to achieve HbA1C < 7 (75.3% vs 63.2%, OR = 3.26, 95% CI: 2.05–5.19, p < .001). No significant between group difference was observed in remaining NCQA parameters of Nephropathy assessment (p = .519), Blood pressure (p = .934), Foot Exam (p = .079), Smoking Cessation treatment offered (p = .107) except for eye Exam (OR = 1.90, 95% CI: 1.08–3.34; p = .027). Pneumovax administration was significantly higher in SMI group (OR = 3.02, 95% CI: 1.58–5.56; p = .001) There was no significant difference in above parameters after adjusting for Antipsychotics use. There was no significant difference between the groups for number of clinic visits or continuity of primary care provided.

CONCLUSIONS: Improved control of diabetes and cardiovascular risk factors in SMI patients was found in our primary care clinic when compared with non SMI patients. Further studies assessing patients by living in a supervised group home facility status and measuring individualized HbA1c target as dynamic variable are needed for understanding these outcomes.

EVALUATING RISK OF MAJOR BLEEDING AND THROMBOEMBOLIC EVENTS FOR RIVAROXABAN VERSUS WARFARIN IN A REAL WORLD SETTING Giavanna Russo-Alvarez 2; Kathryn A. Martinez2; Megan Valente2; James Bena2; Bo Hu2; Jennifer Luxenburg2; Andrei Brateanu1; Michael B. Rothberg2. 1CCF, Cleveland, OH; 2Cleveland Clinic, Cleveland, OH. (Control ID #2701591)

BACKGROUND: Randomized trials demonstrate the non-inferiority of rivaroxaban compared to warfarin for the prevention of stroke and venous thromboembolism (VTE) in atrial fibrillation (AF) and for the treatment of VTE and pulmonary embolism (PE). Yet doubts regarding validity of some original trial findings have surfaced. Studies examining the effectiveness and safety of these drugs in clinical practice are therefore critical. The objective of this study was to assess the relative effectiveness and safety of rivaroxaban versus warfarin in a large integrated health system.

METHODS: We conducted a retrospective cohort study in the Cleveland Clinic Health System (CCHS). Eligible individuals, identified via the electronic health record (EHR), included patients 18 years or older who initiated warfarin or rivaroxaban for treatment of VTE or non-valvular AF between January 2012 and March 2014. The primary measure of effectiveness was thromboembolic events, defined as DVT, PE, atrial thrombus, or ischemic stroke. The primary measure of safety was incidence of major bleeding events, defined as any clinically overt sign of hemorrhage that required an intervention to stop or treat bleeding and that led to hospitalization or increased level of care. Outcomes were first ascertained via the EHR and then confirmed through medical record review by two clinical pharmacists. Appropriateness of anticoagulant dosing was evaluated for each event that occurred while on therapy. Rivaroxaban and warfarin patients were propensity matched 1:1 based on demographic and health history measures from the EHR. Drug groups were then compared using Cox proportional hazards models. To evaluate whether the effectiveness and safety of the drugs varied by time in therapeutic range (TTR), the aforementioned analysis was then stratified at the median TTR of the warfarin patients.

RESULTS: The cohort consisted of 531 propensity-matched pairs. Mean age was 68.9 years (SD: 14.4), and 42% had private insurance. Eighty-two percent of patients were treated for AF and the mean CHADS2 score was 1.94. Median TTR for warfarin patients was 51%. There were no significant differences in patient characteristics or indication for anticoagulation between the rivaroxaban and warfarin groups. Compared to warfarin, rivaroxaban was associated with significantly greater risk of thromboembolic events (HR:4.16, p = 0.034), but there was no significant difference in bleeding risk. Compared to warfarin patients with >51% TTR, rivaroxaban was associated with more major bleeding events (p = 0.017).

CONCLUSIONS: In clinical practice, warfarin was associated with fewer thromboembolic events and did not increase bleeding risk compared to rivaroxaban. While the risk of thromboembolic events was similar, rivaroxaban was associated with more major bleeding than warfarin among patients with a high TTR. For patients who are able to maintain a high TTR, warfarin appears safer than rivaroxaban and no less effective.

EVALUATING THE AGREEMENT BETWEEN SELF-REPORTED AND DOCUMENTED ANALGESIC USE IN OLDER VETERANS WITH OSTEOARTHRITIS Alexander Domanski 2, 3; Matthew J. Bair1, 5; Ruth Balk4; Cynthia Brandt6; Abraham A. Brody3; Rachel Dismore5; Vera Gaetano6; Melissa Garrido3; Dorian Gittleman3; Robert Kerns6; Erin E. Krebs4; Erin Linden4; R. S. Morrison3; Diana Natividad5; Joan Penrod3; Anthony Rinaldi6; Lee Stefanis3; Daniel Sun3; Ula Hwang3, 2. 1Center for Health Information and Communication, Indianapolis, IN; 2Icahn School of Medicine at Mount Sinai, New York, NY; 3James J. Peters VAMC, New York, NY; 4Minneapolis VAMC, Minneapolis, MN; 5Richard L. Roudebush VAMC, Indianapolis, IN; 6VA Connecticut Health System, West Haven, CT. (Control ID #2673143)

BACKGROUND: Studies using prescription claims data demonstrate higher rates of adverse events with opioids compared to NSAIDs for treating pain in older adults. The validity of these results, however, relies on reimbursement data as an indicator of analgesic exposure, despite evidence of the contrary. We compared health record documented prescriptions versus self-reported analgesic use in older veterans to evaluate administrative data as a proxy for analgesic exposure.

METHODS: This was a cross-sectional study of the first 14 months of data (3/2015-5/2016) collected in an ongoing prospective telephone survey of veterans from 4 VA hospitals ≥50 years of age with ≥2 inpatient or outpatient diagnoses of knee or hip arthritis. Subjects were analgesic-free (based on prescription records) for ≥180 days prior to starting a new analgesic (opioid, NSAID, other [e.g. acetaminophen]) or control (filling a non-analgesic prescription) and surveyed ≤30 days of prescription filling. Comparisons were calculated between self-reported analgesic lists (ref.) and health record analgesic prescriptions filled using k coefficients, sensitivity, and specificity for analgesic classes. This study received CIRB approval.

RESULTS: 595 subjects were recruited. Mean age was 66y (±8.9y), 93% were male. Agreement between self-reported analgesics versus health record analgesic prescriptions over 30 days was fair (k = 0.26). This varied by drug class: opioids (k = 0.56), NSAIDs (k = 0.41), other (k = 0.37). Sensitivity was higher for opioids (0.71) than NSAIDs (0.57) or other (0.39), while specificity across drug classes was similar (0.89, 0.84, 0.93 respectively).

CONCLUSIONS: Discordance exists between health record versus self-reported analgesic use in older veterans. Caution should be exercised when interpreting pharmacoepidemiological safety studies that use administrative health record data to measure analgesic exposure.

EVALUATING THE EFFECTIVENESS OF A HEALTHY FOOD SHOPPING TOUR IN LOW-INCOME AFRICAN-AMERICAN NEIGHBORHOODS IN CHICAGO Ashley K. Haddad; Adam H. Lewis; Nyahne Q. Bergeron; Daniel J. Rowell; Kathryn E. Gunter; Yolanda O’Neal; Sandra Ham; Marshall Chin; Monica E. Peek. University of Chicago, Chicago, IL. (Control ID #2705597)

BACKGROUND: Diet-related chronic diseases cause significant morbidity and mortality, and are disproportionately prevalent amongst African-Americans. Disparities in diet-related chronic diseases are complex, with multiple contributors including individual preferences, social relationships, and the built environment. Nutrition education has the potential to address these disparities if tailored to an individual’s specific behavioral, social, and environmental context.

METHODS: We evaluated the effectiveness of a 1-hour, in-store, group nutrition education tour (led by dieticians and diabetes educators) in 4 low-income, predominantly African-American neighborhoods in Chicago. We used a mixed-method approach to evaluation. We qualitatively analyzed 2 focus groups (n = 25) using a framework approach. Sessions were transcribed verbatim and analyzed using ATLAS.ti software. We also quantitatively analyzed pre/post food knowledge surveys (n = 104) using chi-squared, t-tests, and longitudinal regression modeling, and analyzed food purchase receipts (n = 99) using Wilcoxon ranked sign tests. We are obtaining grocery store control data to compare food purchases made by non-tour participants.

RESULTS: A majority of surveyed participants were female (80%) and African-American (95%).The mean age was 58.4 years, and 36% of participants (n = 37) attended more than one tour. Focus group participants noted barriers to healthy eating including the influence of family (“It’s difficult for me when I go to my sister’s house…she’s an old-fashioned cook”) and an unhealthy food environment (“sugar is in everything!”). Motivators for tour participation included knowledge gained (“It’s not about $10… it’s about the information that we get and how we can use it to better ourselves”), support from other participants (“The important thing about this tour is that all of us are in the same boat”), and the interactive learning environment (“Me, I’m a literal person so…being in the store and actually seeing the product you understand more”). Participants demonstrated improved knowledge after a single tour (6.1 vs 5.8 of 10 questions correct, p = 0.18); for those who participated in multiple tours, knowledge improvements were statistically significant (6.5 vs 5.5 correct questions, p < 0.001). Additionally, the number of tours that a participant attended was a significant predictor of knowledge gained (β = 0.03, p < 0.001). After the tour, 34% of participant food purchases were fruits or vegetables (vs. 19% nationally) with no difference between one-time and repeat participants (36% vs 28%, p = 0.24).

CONCLUSIONS: Our findings suggest that in-store group nutrition education tours in low-income, minority neighborhoods can improve participant knowledge regarding healthy eating and potentially increase the consumption of fruits and vegetables, compared to national averages. Additionally, it is important to consider behavioral, social, and environmental context when designing interventions to overcome barriers to healthy eating.

EVALUATING THE EFFECTIVENESS OF LUNG CANCER SCREENING IN PATIENTS WITH COPD Minal Kale; Bart Ferket; Juan Wisnivesky. Mount Sinai School of Medicine, New York, NY. (Control ID #2707373)

BACKGROUND: Chronic obstructive pulmonary disease (COPD) is a highly prevalent disease that shares a common risk factor, tobacco exposure, with lung cancer. Moreover, studies show that COPD is an independent risk factor for lung cancer and that lung cancer is a major cause of death in COPD patients. Thus, patients with COPD are important candidates for lung cancer screening. However, results from the National Lung Screening Trial (NLST) are difficult to generalize to this population due to differences in lung cancer risk, increased likelihood of complications from diagnostic work-up, and potential ineligibility of receiving full resection due to limited lung function. Thus, there is a need to assess the benefits and harms of lung cancer screening in this population.

METHODS: We created an individual level state transition model using decision analysis software that simulated the experience of individuals with all stages of COPD, undergoing lung cancer screening either with annual chest radiograph (CXR) or low dose computed tomography (LDCT) for 3 cycles with a 10-year time horizon. We estimated model parameters using a combination of published data and analysis of NLST data. The screening cohort was composed of data from the NLST cohort and consisted of individuals with spirometry findings consistent with COPD based on the American Thoracic Society and European Respiratory Society Task Force. Parameters regarding prevalence of cancer, screening performance (sensitivity and specificity) and diagnostic work-up derived from the NLST. Overall survival for individuals with lung cancer was estimated based on Surveillance, Epidemiology, and End Results data. Background mortality was based on United States 2011 life tables accounting for sex and stage of COPD. Model outcomes included difference in life expectancy with CXR vs. LDCT screening, overall survival, and complication rate.

RESULTS: We found that that lung cancer screening with LDCT in individuals with COPD leads to a mean increase in life expectancy within a 10-year time horizon of 0.65 months. We found a reduction in overall mortality at 10 years by 1% (29% vs 28%). We also found a lower rate of non-fatal complications; 172 in 4655 individuals screened with LDCT compared to 178 in those screened with CXR.

CONCLUSIONS: Our microsimulation model suggests that LDCT screening leads to a lengthening of life expectancy and reduction in overall mortality in individuals with COPD. We also found that LDCT screening was associated with non-fatal complications in individuals with COPD. Future simulation modeling is needed to characterize the effect of screening on lung cancer mortality, the benefit of screening across different stages of COPD, and the optimal screening criteria.

EVALUATION OF A CHOOSING WISELY™ INTERVENTION TO REDUCE LOW VALUE PREOPERATIVE CARE FOR PATIENTS UNDERGOING CATARACT SURGERY AT A SAFETY NET HEALTH SYSTEM John N. Mafi 1, 2; Patricia Godoy-Travieso3; Eric Wei3; Jesse Berry3; Rudy Amaya3; Brandon Wong4; Carmen A. Carillo5; Laura Sarff3; Lauren P. Daskivich6; Sitaram Vangala7; Emmett Keeler2; Cheryl Damberg2; Catherine Sarkisian5, 7. 1David Geffen School of Medicine at UCLA, Los Angeles, CA; 2RAND Corporation, Santa Monica, CA; 3LAC + USC Medical Center, Los Angeles, CA; 4LAC + USC, Los Angeles, CA; 5University of California, Los Angeles, Los Angeles, CA; 6Harbor-UCLA Medical Center, Los Angeles, CA; 7UCLA and Greater Los Angeles VA, Los Angeles, CA. (Control ID #2699282)

BACKGROUND: Because pre-op testing for cataract surgery provides virtually no patient benefit, we evaluated an initiative to eliminate it.

METHODS: Design: pre-post quasi-experiment comparing pre-op cataract surgery care at LAC + USC Medical Center (intervention site) vs. Harbor-UCLA (control site). Data: administrative data to identify pts undergoing cataract surgery between 10/15/14-4/15/16. Intervention: a quality officer at LAC + USC 1) reviewed cataract surgery pts’ charts, 2) presented data on overuse to anesthesia/ophthalmology chairs—gaining their support for the initiative, 3) recruited a resident champion and 4) empowered nurses to stop scheduling pre-op visits. On 9/30/15, the team emailed pre-op guidelines to physicians/trainees/staff, calling to eliminate routine pre-op testing for cataract surgery by 10/13/15. Outcomes: pre-op visits, labs, EKGs, and wait-time between cataract diagnosis and surgery. Analysis: difference-in-differences (DinD) comparing utilization between sites using logistic regression adjusting for pt characteristics.

RESULTS: We identified 1,009 intervention and 959 control pts undergoing cataract surgery during the study period. Baseline mean age/sex (61 yrs/53% female) was similar between both groups. The proportion of pre-op visits, labs, and EKGs declined more for intervention than control pts; intervention pts: 77, 91, and 74% before the intervention vs. 20, 39 and 27% after the intervention respectively; control pts: 62, 40, and 66% before vs. 86, 72, and 86% after the intervention respectively (DinD −81%, p < 0.001, −83%, p < 0.001, −67%, p < 0.001 respectively). Median surgical wait-time declined more for intervention pts (245 days before vs. 64 days after the intervention) than for control pts (27 days before vs. 22 days after the intervention), DinD −176, p < 0.001.

CONCLUSIONS: This intervention substantially reduced low value pre-op care and surgical wait-times among pts undergoing cataract surgery.

figure ac

Pre-Op Visits at LAC + USC (intervention) vs. Harbor-UCLA (control), 10/15/14-4/15/16

EVALUATION OF A COACHING BY TELEPHONE INTERVENTION FOR VETERANS AND CARE TEAM ENGAGEMENT (ACTIVATE) Eugene Oddone 2, 3; Laura Damschroder1; Jennifer Gierisch2; Maren Olsen2; Linda L. Sanders2; Angie Fagerlin4; Carrie L. May2; Jordan Sparks1; Felicia A. McCant2. 1Ann Arbor VA Center for Clinical Management Research, Ann Arbor, MI; 2Durham VA Medical Center, Durham, NC; 3Duke Medicine, Durham, NC; 4Salt Lake City VA Medical Center, Salt Lake City, UT. (Control ID #2702001)

BACKGROUND: A large proportion of deaths and many illnesses can be attributed to three modifiable risk factors: tobacco use, overweight/obesity, and physical inactivity. Health risk assessments (HRAs) help individuals define their risk. HRA’s are widely available online, but have not been consistently employed in healthcare systems to as a tool to activate patients’ participation in behaviorally-focused prevention programs. Health coaching is an emerging method for promoting behavior change that relies on theories of human development, social psychology, and adult learning. There is growing evidence that coaching can improve health outcomes by aligning patient preferences, values, and patient-determined goals with behavior change strategies. This study was designed to determine if a brief telephone coaching intervention, when coupled with results from an HRA, could activate patients to address their modifiable risk by enrolling in a prevention program.

METHODS: We conducted a 3 site randomized controlled trial among veterans (n = 417) enrolled in primary care. To be eligible, veterans must have had at least one modifiable risk factor: BMI >30, less than 150 min of physical activity per week, and/or current smoker. All veterans completed the Veteran Affairs Health System’s web-based HRA. Controls (n = 209) were encouraged to share HRA results with their primary care provider. Intervention participants (n = 208) received two telephone calls from a trained health coach: one week, and four weeks after completing the HRA. Coaches worked with veterans to collaboratively set a specific and actionable goal about enrolling in, and attending a structured prevention program designed to reduce their modifiable risk. The co-primary outcomes were enrollment in a prevention program by 6 months, and change in behavioral activation as measured by Patient Activation Measure (PAM). Logistic regression was used to estimate the treatment difference in enrollment. General Linear Models with unstructured covariance and parameters of time and group × time interaction were used to estimate the treatment group difference in the PAM.

RESULTS: Most veterans were male (85%), white (50%), or African American (40%) with a mean age of 56. Veterans were eligible because their BMI was > 30 (80%), were physically inactive (50%), or current smokers (39%). 58% had more than one eligibility criteria. From baseline to 6 months, intervention participants reported higher rates of enrollment in a prevention program (51% vs 29%, P < 0.0001) and greater mean increase in PAM (4.8 vs 2.4, p = 0.03) compared to controls.

CONCLUSIONS: Brief health coaching, informed by HRA results, improves patient activation and can help facilitate veterans’ enrollment in effective prevention programs to address health behaviors and health risk.

EVALUATION OF A COMMUNITY-BASED SAFE FIREARM AND AMMUNITION STORAGE INTERVENTION Joseph A. Simonetti 2, 4; Ali Rowhani-Rahbar3, 5; Cassie King1; Elizabeth Bennett1; Frederick P. Rivara5, 3. 1Seattle Children’s Hospital, Seattle, WA; 2University of Colorado School of Medicine, Denver, CO; 3University of Washington School of Public Health, Seattle, WA; 4Denver VA Medical Center, Denver, CO; 5University of Washington, Seattle, WA. (Control ID #2702749)

BACKGROUND: Firearms injured nearly 115,000 Americans in 2014. Safe firearm storage practices, such as storing them locked and unloaded, are associated with a lower risk of unintentional and self-inflicted firearm injury and death among household members. However, few community interventions have been developed to promote such practices and a large proportion of U.S. firearms remain unsafely stored. The aim of this study was to assess the effectiveness of a community-based firearm safety intervention.

METHODS: We performed a before-after evaluation of two firearm safety events in Washington State in 2015. Events were held at sporting goods’ stores that included firearm retail and were promoted through social and traditional media. Participants received a brief safety message, their choice of a free firearm safety device (firearm trigger lock or lockbox), demonstration on using both devices, and demonstrated their ability to lock and unlock a mock firearm using each device. We included all participants 18 years or older who spoke English or Spanish, completed baseline and follow-up telephone surveys, and signed legal release forms to participate. We used McNemar’s test for matched pairs to assess whether changes in four study outcomes from baseline to follow-up were statistically significant, including whether all household firearms were stored locked, all were unloaded, all ammunition was locked, and a composite measure assessing whether all firearms were locked and unloaded and all ammunition was stored locked. We performed a subgroup analysis of households with children less than 18 years of age.

RESULTS: Of 415 participants, 404 completed baseline surveys, 313 consented to the follow-up, and 206 (65.8%) completed follow-up surveys and were included. Sixty-one percent were male, 23.3% were Veterans or active military, and 53.4% had children less than 18 years in their household. Eighty-seven percent preferred the firearm lockbox rather than the trigger lock. At baseline, 63.7% stored all household firearms locked, 62.7% stored all firearms unloaded, 54.6% stored all ammunition locked, and 32.9% reported storing all firearms locked and unloaded and all ammunition locked. At follow-up, a significantly greater proportion reported storing all firearms locked (+13.7% [95%CI: 5.6-21.9]) and unloaded (+8.5% [95%CI: 2.3-14.7]) and a non-significantly greater proportion reported storing all ammunition locked (+6.3 [95%CI: −1.2-13.7]). A significantly greater proportion reported practicing all three safe firearm and ammunition storage practices (+12.6% [95%CI: 4.5-20.6]). Findings were unchanged among households in which children were present.

CONCLUSIONS: This community-based intervention that included distribution of a free, participant-selected locking device was effective in increasing safe firearm storage practices. Differences in participant preferences for devices suggest that a “one size fits all” approach may be inadequate in affecting population-level storage practices.

EVALUATION OF AN ALGORITHM TO CLASSIFY EMERGENCY DEPARTMENT UTILIZATION IN THE CONTEXT OF PATIENTS WITH ESTABLISHED PRIMARY CARE Sharon Rikin; Emilia Hermann; Jessica R. Singer. New York Presbyterian Hospital - Columbia University Medical Center, New York, NY. (Control ID #2704841)

BACKGROUND: Identifying factors associated with emergency department (ED) utilization for patients with established primary care may reveal problems within the primary care system or patient barriers to accessing timely care. The NYU ED Algorithm has been used by health service researchers to distinguish between four categories of ED visits: non-emergent, emergent but primary care treatable (PCT), emergent requiring ED care but primary care preventable, and emergent requiring ED care that are not primary care preventable. This study sought to validate the NYU ED algorithm from the perspective of primary care physicians in order to improve health care utilization of patients with established primary care.

METHODS: This nested, case–control study took place at an academic medical center in Manhattan, NY. We identified ED visits made by primary care patients during two 7-day study windows. Two reviewers independently classified each visit as PCT or non-PCT following an adapted NYU ED Algorithm. Visits were categorized based on the patient’s initial complaint, vital signs, medical history, treatment/tests provided, and final diagnosis. Frequency of ED and primary care visits and chronic health conditions including psychiatric illness were also collected. Cohen’s kappa was used to evaluate inter-rater agreement. Multivariable logistic regression was used to evaluate associations between characteristics and odds of PCT visit.

RESULTS: Among the 92 ED visits identified, 48.91% were PCT. Inter-rater agreement for acuity of complaint was high (Cohen’s kappa 0.68, 95% CI (0.45, 0.90)), moderate for PCT vs. non-PCT determination (Cohen’s kappa 0.52, 95% CI (0.36, 0.68)), and fair for determination if medical complaint was potentially preventable (Cohen’s kappa 0.35, 95% CI (0.06, 0.64)). The average number of primary care and ED visits in the previous 12 months was 4.27 (SD 2.98) and 4.21 (SD 4.57), respectively. There was no difference in the number of primary care or ED visits between those with PCT vs. non-PCT visits (OR 0.92, 95% CI (0.79, 1.09) and OR 1.02, 95% CI (0.90, 1.17)). Patients with PCT visits were more likely to have <2 vs. ≥2 chronic conditions (OR 5.19, 95% CI (0.89, 30.39).

CONCLUSIONS: There was only a moderate level of inter-rater agreement in determination of PCT vs. non-PCT visits, likely because our adaptation of the algorithm relies on provider evaluation of the clinical scenario as compared to the original algorithm which utilizes ICD coding to determine appropriate level of care. Frequency of healthcare utilization did not differ between patients with PCT vs. non-PCT visits. Those with fewer chronic conditions were more likely to have PCT visits while those with more chronic conditions were more likely to have non-PCT visits. This study highlights the need for primary care physician and patient perspectives in future programs designed to evaluate health care utilization in order to adequately address barriers to obtaining prompt, non-emergent medical care.

EVALUATION OF AN INTERDISCIPLINARY COMMUNICATION TOOL CALLED TEMPO (TEAM ENGAGEMENT TO MONITOR PATIENT OUTCOME). Alok Arora 1; Idrees Mohiuddin2; Larissa Verda3. 1Weiss Memorial Hospital, CHI, IL; 2Weiss Memorial Hospital, Park Ridge, IL; 3Weiss Memorial Hospital, Chicago, IL. (Control ID #2705058)

BACKGROUND: Weiss Memorial Hospital is in a healthcare provider shortage area and 40% of our patients are from long term acute care hospital (LTACH) and skilled nursing facility (SNF) with multiple co-morbidities and high readmission rates, such background demands synchronized medical care to ensure safe and timely discharges. TEMPO board rounds were introduced in our hospital in November 2015 as an enhanced visual tool to coordinate and promote interdisciplinary communication. TEMPO involves a large white board on medical floors which is manually populated with data retrieved from various medical information systems and includes patient’s ‘geometric mean length of stay’ (GMLOS), discharge disposition and an anticipated discharge date which drives the TEMPO discussion. Morning TEMPO rounds includes a medical resident, named nurse, social worker and a case manager. Patient status relative to discharge date is indicated by ‘red, yellow and green’ triangles and act as a cue to identify ‘barriers to discharge’ and address key patient specific needs. Early recommendations from physical and occupational therapy help with rehabilitation and placement planning for ‘fall risk/stroke’ patients.

METHODS: We performed an anonymous survey for the efficacy of the TEMPO rounds process. Random cohort of t participants were selected which included 10 case managers/social workers, 14 RNs and 28 medical residents (N = 52). Respondent commented on time management (consumption/saving), communication (improved/worsened), patients care (improved/worsened) and discharge planning (facilitates/complicates).

RESULTS: The visibility of the information during TEMPO rounds was widely appreciated as compared to the old style ‘sit around the table’ collaborative case rounds. Most respondents reported a high level of satisfaction with TEMPO rounds. 76% (N = 40) reported improvement in delivery of optimal patient care due to TEMPO rounds. 75% (n = 39) of respondents expressed satisfaction with time savings associated due to improved interdisciplinary communication and early identification of discharge barriers. A better formulated discharge strategy following TEMPO was reported by 70% (N = 36) respondents; but 40% (N = 21) also thought that the TEMPO process did not lead to an actual reduction in the GMLOS. We think the large LTACH/SNF demographics we cater to can explain this variance.

CONCLUSIONS: TEMPO delivers an effective discharge strategy by promoting communication and participation between various stake holders. An interdisciplinary TEMPO board has empowered our staff to bring in changes and has improved health care delivery at our hospital by addressing barriers to a safe discharge.

Impact of TEMPO on patient safety and outcomes

Specific attention to readmissions Early identification of frail and de-conditioned for early placement
Improved compliance of CAUTI and CLABSI toolkits Increased adherence with care bundles.
Switching IV to PO antibiotics and stopping IV fluids early Higher flu and pneumococcal vaccination rates
Keeping track of Medicare days Contact precautions compliance
Improving ‘core measure’ compliance Identifying ‘falls risk’ patients

EVALUATION OF BURNOUT AMONG RESIDENTS IN AN ACADEMIC PRIMARY CARE CLINIC SETTING Caroline Falker 3, 1; Frank D. Buono3; Faith Harrington1; Rebecca Kosowicz2, 1; Sumit Kumar2, 1; Rachel Laff1; Yungah Lee2, 1; Kenneth Morford4, 1; Destiny M. Printz3; Jonathan Stock2, 1; Rebecca Brienza1, 2. 1VA Connecticut Healthcare System, West Haven, CT; 2Yale New Haven Hospital, Newington, CT; 3Yale School of Medicine, New Haven, CT; 4Yale University School of Medicine, New Haven, CT. (Control ID #2706022)

BACKGROUND: Physician burnout is a work-related syndrome involving emotional exhaustion, depersonalization, and a sense of reduced personal accomplishment. Burnout is prevalent among healthcare providers, especially in primary care settings. Studies show that the prevalence of burnout is near or exceeding 50% in physicians. Medical residents are especially vulnerable to burnout. However, there are few studies focusing on the evolution of burnout among residents over time. Additionally, most studies of resident burnout focus on inpatient settings rather than outpatient primary care settings. Identifying the presence of burnout and designing residency curriculum to minimize those factors would benefit overall health and wellbeing of resident trainees, and potentially improve quality of patient care.

METHODS: We utilized the Tedium Measure-21, which is a validated questionnaire to assess burnout. TM-21 comprises of total 21 questions that assess three domains: physical, emotional and mental exhaustion. Each question is weighted with a scale from 1 (never) to 7 (always). The Tedium mean score is calculated for each survey to measure critical stress (mean score between 3 and 4) or burnout (mean score greater than 5). TM-21 was administered to Internal Medicine residents (PGY1-PGY3) and Nurse Practitioner residents over the course of the academic year 2016–2017 at the end of each ambulatory block. Each group of residents completed the TM-21 once every three months for a total of four survey submissions throughout one academic year.

RESULTS: We collected a total of 63 TM-21 surveys from Internal Medicine and Nurse Practitioner (NP) residents over the first two (out of four) data collection periods. Surveys were completed by 24 (38%) PGY-1 residents, 21 (33%) PGY-2 residents, 13 (21%) PGY-3 residents, and 5 (8%) NP residents. Of the total surveyed, 52 residents (82.5%) met criteria for critical stress, 10 residents (15.9%) met criteria for burnout and 1 resident (1.6%) met criteria for neither. Higher rates of burnout were observed in PGY-2 residents as compared to other training levels although preliminary results have not reached significance. Higher rates of critical stress were noted in PGY1 residents.

CONCLUSIONS: We found that a significant proportion of Internal Medicine and Nurse Practitioner residents meet criteria for burnout according to the validated TM-21 survey. Critical stress and burnout varied across different training levels and this may reflect a need for more targeted interventions. Furthermore, studying rates of critical stress and burnout overtime can provide valuable information regarding burnout prevention. Further characterizing burnout into the three subdomains (physical, mental, emotional) may help provide clues to identify factors associated with burnout. We acknowledge that the Maslach Burnout Inventory is accepted as a gold standard for burnout, but we used TM-21 because it has been shown to better assess burnout in healthcare providers with test-retest reliability

EVALUATION OF HOUSE STAFF BURNOUT AND WORK ENVIRONMENT: HOW CAN WE HELP? Mitchell A. Izower; Johanna Martinez; Andrew Yacht. Northwell Health, Manhasset, NY. (Control ID #2702835)

BACKGROUND: Burnout is a long term stress reaction that adversely affects house staff psychological health, patient care and satisfaction, and increases self-perceived errors. The ACGME has increasingly focused on burnout management and mitigation, but specific methods to decrease burnout are needed. We sought to evaluate and delineate sources of, and means to improve, house staff burnout.

METHODS: A cross-sectional study of Northwell Health house staff was performed. 1652 house staff received a ten-item Mini-Z survey regarding stress, burnout, and work conditions, as well as 2 open-ended questions on work life and wellbeing. A formal content analysis was performed for open-ended question responses.

RESULTS: Among 1652 house staff surveyed, 611 responses were received (37%). High stress was present in 52%, with 24% reporting burnout. 47% reported a very busy or chaotic workplace. 32% felt they did not have satisfactory control of their work. 33% felt they did not have sufficient time for documentation. 11% described too much home EMR time. Responses to “What suggestions do you have that would improve your well-being?” were classified by suggestion type. Requested activities included more physical activity (59%) and social events (21%). Services included improved nutrition (29%) and an on-campus gym (25%). Support/work-life changes included decreased work hours (46%). Work flow/dynamic suggestions included EMR improvements (28%), more clinical support staff (23%), and better team dynamics (17%). Training/education suggestions included more educational time (35%). Responses to “If you could change one thing to improve your work life, what would it be?” were classified by suggestion type. Requested work flow/dynamic changes included EMR improvements (44%) and more efficient workflow (19%). Program changes included reduced work hours (34%) and schedule improvements (20%). Services included improved nutrition (27%). Personal wellness suggestions included more exercise (34%) and sleep (19%).

CONCLUSIONS: Given the deleterious effects of burnout on house staff and patient care, programs should seek to reduce house staff burnout. The Mini-Z survey is an efficient method to determine burnout and burnout sources, and house staff feedback should be solicited to address burnout. In our sample, nearly one-quarter of house staff were experiencing burnout, and over half felt highly stressed. Almost half reported a very busy or chaotic workplace, while one-third felt they did not have satisfactory control of their work or sufficient time for documentation. House staff felt their wellbeing and work life would improve with changes to their amount of physical activity and exercise, nutrition, social activities, educational time, scheduling and work hours, clinical support staff availability, team dynamics, workflow, and EMR functionality. Programs should optimize these factors to decrease house staff burnout.

EXPERIENCES OF LESBIAN, GAY, BISEXUAL, TRANSGENDER AND QUEER (LGBTQ) PEOPLE OF COLOR IN SHARED DECISION-MAKING WITH HEALTHCARE PROVIDERS ABOUT INTIMATE PARTNER VIOLENCE Fanny Y. Lopez 1; Kathryn E. Gunter1; Scott Cook1; Justin Jia1; Arshiya A. Baig2, 1. 1The University of Chicago, Chicago, IL; 2University of Chicago, Chicago, IL. (Control ID #2704939)

BACKGROUND: High-quality Shared Decision-Making (SDM) has been positively associated with patient satisfaction, quality of care, and health outcomes. However, SDM has been infrequently studied among minority populations, especially LGBTQ people of color (POC). Successful SDM is especially important for survivors of intimate partner violence (IPV) where sharing information, deliberating options and making decisions collaboratively between patients and providers can result in outcomes ranging from life-saving to catastrophic. However, no research studies have investigated how IPV survivors who are LGBTQ POC engage in SDM with providers. We aim to describe LGBTQ POC’s SDM experiences with providers (i.e. physicians, nurses, counselors) to address IPV and provide recommendations to improve SDM with this patient population.

METHODS: From December 2015 to December 2016, we conducted one-on-one, semi-structured interviews with LGBTQ POC, 18 years of age or older, who reported past history of IPV with same-sex, transgender or genderqueer partners. We explored participants’ definitions of IPV, experiences discussing IPV and decision-making with providers, and their preferences regarding tools to improve communication and SDM about IPV. Data extracted from audio recordings were transcribed verbatim. To ensure internal consistency, multiple reviewers developed a codebook through an iterative process. We used a modified template approach for data analysis to identify barriers and facilitators that LGBTQ POC experience when discussing IPV and making decisions with providers.

RESULTS: We conducted 36 interviews and coded 24 transcripts. From preliminary analysis, we found that 12 of the 24 participants discussed IPV with healthcare providers. Those who did not discuss IPV cited barriers, including lack of time and trust, as well as concerns about provider’s assumptions about IPV based on the patient’s race/ethnicity, sexual orientation and/or gender identity. Those who discussed IPV cited facilitators, such as providers’ LGBTQ identity, as well as sensitivity and understanding of LGBTQ POC’s life stories and experiences of IPV within same-sex, transgender and genderqueer relationships. Participants recommended that providers become culturally competent in LGBTQ and racial/ethnic minority health, ask about IPV in a sensitive way, and address the unique needs of LGBTQ POC to improve IPV SDM.

CONCLUSIONS: The majority of study participants are open to providers asking about IPV and providing culturally tailored resources. Participants recommended that providers use tools, such as face-to-face conversation, written materials, phone and web-based communication, and support groups to facilitate SDM about IPV. Future research should design and evaluate tools that improve SDM with LGBTQ POC experiencing IPV.


BACKGROUND: Neighborhood crime can have deleterious health effects that lead to or worsen chronic conditions, such as obesity and hypertension. Prior studies have predominantly attributed these health effects to adverse psychological responses to crime, such as negative perceptions about neighborhood safety and chronic stress. However, few studies have examined both the direct experience of crime and adverse psychological responses, in order to corroborate theoretical pathways to downstream health consequences. The purpose of this study is to examine the relationships between experiences of neighborhood crime, adverse psychological responses, and obesity and hypertension status in a high-risk, urban population.

METHODS: An address-based probability sample of 267 participants (ages ≥ 35 years) was conducted to query residents on the South Side of Chicago about neighborhood crime. During hour-long, in-person interviews, participants were asked about their prior experience of neighborhood crime (theft or property damage), psychological responses to crime (perceived neighborhood safety and stress due to crime), demographics, and health status. Objective biological measures (height, weight, and blood pressure) were obtained at the time of interview. Multivariable logistic regression models were used to assess obesity and hypertension status as a function of neighborhood crime and psychological response measures, adjusting for demographic characteristics, self-reported health status, and interaction effects for perceived safety and stress.

RESULTS: Overall, 431 eligible residents were invited to participate and 267 (62%) completed the survey. People were primarily non-Hispanic black (68.9%) or Hispanic (19.3%), female (62.9%), and between the ages of 35–70 (86.5%). More than half (54.9%) were obese and almost half (49%) had hypertension. Prior experience of neighborhood crime was associated with higher adjusted odds of perceived neighborhood safety concerns (Adjusted odds ratio [AOR] = 2.3; 95% CI, 1.3–4.0; P = 0.003) and stress due to crime (AOR = 2.9; 95% CI, 1.6–5.2; P = 0.001). However, perceived neighborhood safety and stress due to crime were not associated with obesity or hypertension status. Any prior experience of neighborhood crime was independently associated with 87% higher adjusted odds of obesity (95% CI, 1.07–3.26; P < 0.05) and 74% higher adjusted odds of hypertension (95% CI, 0.96–3.17; P = 0.07).

CONCLUSIONS: Psychological measures, such as perceived neighborhood safety or stress due to crime, may be insufficient to fully examine the relationship between neighborhood crime and chronic conditions in a high-risk, urban population. Alternative pathways, such as community-level processes (e.g., breakdown of social ties may lead to inadequate social support for self-management activities) should be explicitly examined for a more robust understanding of the intersection between neighborhood crime and chronic disease.

FACILITATING HIGH VALUE CARE THROUGH THE ACCURATE DIAGNOSIS OF SKIN DISEASE: A SYSTEMATIC REVIEW Syril Keena T. Que5; Sara-Megumi L. Naylor1, 2; Robert Dellavalle3, 4. 1Division of Primary Care at the VA Greater Los Angeles Healthcare System, Los Angeles, CA; 2David Geffen School of Medicine at UCLA, Los Angeles, CA; 3Denver VA Medical Center, Denver, CO; 4University of Colorado, Denver, Denver, CO; 5Brigham and Women’s Hospital, Boston, MA. (Control ID #2705634)

BACKGROUND: Primary care physicians (PCPs) are frequently the frontline providers of care for skin disease. Dermatologic complaints make up 6-22% of PCP visits, with only a third of these patients referred to dermatologists for further work-up or management. Accurate diagnosis and optimized care coordination between PCPs and dermatologists is a high priority as it influences subsequent management decisions, including prescriptions, biopsies, referrals, and patient outcomes. It is unclear if current education on dermatologic conditions is sufficient for PCPs to accurately diagnose common skin conditions. In this systematic review, we assess the diagnostic accuracy of PCPs and dermatologists for adult skin conditions encountered in the outpatient setting. To our knowledge, this is the first systematic review that examines diagnostic accuracy for a broad spectrum of dermatologic conditions, including neoplastic, infectious, and inflammatory skin conditions.

METHODS: A systematic review of observational studies and randomized controlled trials was conducted using PubMed and EMBASE from database inception to August 2016. Bibliographies of retrieved studies were further analyzed. Two independent reviewers identified studies and abstracted data. Studies were assessed using a priori inclusion and exclusion criteria. Our aim is to compare the diagnostic accuracy of PCPs and dermatologists. Primary outcome measures are sensitivity and specificity.

RESULTS: A total of 25 studies were included, with data from more than 1,953 PCPs, 866 dermatologists, and 12,446 lesions. Only three (12%) studies reported both sensitivity and specificity. Dermatologists showed higher accuracy when compared to PCPs in diagnosing all skin conditions, which reached statistical significance for basal cell carcinoma, seborrheic keratosis and psoriasis. Pooled weighted sensitivity for basal cell carcinoma was 0.46 (95% CI 0.36, 0.55) for PCPs and 0.92 (95% CI 0.84, 1.00) for dermatologists. For seborrheic keratosis it was 0.23 (95% CI 0, 0.48) and 0.79 (95% CI 0.60, 0.98), respectively. For psoriasis it was 0.56 (95% CI 0.43, 0.68) and 0.97 (95% CI 0.94, 1.00), respectively. PCPs exhibited high sensitivity for certain inflammatory and infectious skin conditions including acne— 0.88 (95% CI 0.7, 0.99) and verruca vulgaris— 0.79 (95% CI 0.67, 0.91).

CONCLUSIONS: Dermatologists showed higher diagnostic accuracy when compared to PCPs. Educational initiatives for current and future PCPs should focus on common diagnoses with gaps in diagnostic accuracy, which are highlighted in this systematic review. Meanwhile, increased involvement of PCPs in the management of conditions like acne and verruca allows for improved access to care and reduced health care utilization. Increased collaboration between providers and improvements in diagnostic accuracy can ultimately decrease the number of unnecessary or inappropriate prescriptions, biopsies, referrals, and other indicators of low value care.

FACILITATORS AND BARRIERS OF INTERDISCIPLINARY TEAM FUNCTION IN PRIMARY CARE: SIX YEAR POST-PCMH IMPLEMENTATION Linda Kim 1; Karleen Giannitrapani2; Alexis K. Huynh1; Danielle Rose1; Alison B. Hamilton1; Susan E. Stockdale1; Lisa V. Rubenstein1. 1VA Greater Los Angeles Healthcare System, Los Angeles, CA; 2VA Palo Alto Healthcare System, Palo Alto, CA. (Control ID #2706118)

BACKGROUND: Findings from earlier implementation studies provide some understanding on how elements of the Patient Centered Medical Home (PCMH) model of care delivery affect primary care providers (PCPs), nurses, and administrative staff; however, more work is needed to understand what factors promote or hinder PCP and staff performance and team function. In this study we aim to describe provider mentioned facilitators and barriers to team function, six years post-PCMH implementation, as the initial step in a longitudinal study evaluating changes in team function over time.

METHODS: Data were collected through semi-structured interviews of 36 frontline staff (PCPs, RNs, LVNs, and administrative staff) in 6 VA primary care practices, as part of a larger PCMH implementation study. To ensure we captured a range of perspectives we employed a quota sampling approach to randomly select key informants from each teamlet member role. All sites had implemented Patient Aligned Care Teams (PACT), the VA’s version of PCMH, four to six years prior to the interviews. Data were analyzed using a content analysis method of constant comparison to capture provider identified facilitators and barriers to team function.

RESULTS: Factors that facilitated team function include: 1) increased familiarity with team members’ roles: “we still come across some kinks here and there, butWe know whats expected of each of us to make the teamlet;” 2) participatory decision making: “And if theres something, we all kind of brainstorm together what should we do. Its a team effort;” 3) effective communication: “We have just talked amongst ourselvesif something comes up that we need to talk abouttheres never a problem really with our teamlet;” and 4) having various communication methods available within teams to facilitate coordination of care and between teams to share strategies regarding change processes. Barriers include: 1) staffing shortages/coverage: “So I think thats one of the issues with PACT that we have here. It may have to do with staffing or just the way coverages need to be coordinate;” and 2) poor conflict resolution skills: “… the only time that she would talk to me is if theres a problem and then she will tell me how to fix the problem;” and 3) systems/processes that hinder workflow: “You have to cut and paste the noteand thento an addendumIt seems like a lot of extra work.”

CONCLUSIONS: Six years after PCMH implementation, there were several elements that helped to promote team function and enhance communication within and between teams; however we also provide insight into areas that need further improvement. Findings from this study can help inform future efforts evaluating changes in team function over time, which can then be used to guide development of evidence-based improvement strategies aimed at optimizing team function.

FACTORS ASSOCIATED WITH BASELINE HYPERTENSION CONTROL IN HEART HEALTH NOW Samuel Cykert 1; Kamal Henderson2; Jacqueline Halladay2. 1University of North Carolina, Chapel Hill, NC; 2University of North Carolina School of Medicine, Chapel Hill, NC. (Control ID #2706870)

BACKGROUND: Heart Health Now (HHN) is the North Carolina Cooperative for Evidence NOW (EN). EN is an AHRQ funded initiative designed to measure the impact of practice facilitation on the implementation of patient centered evidence in small primary care practices. The current focus of HHN and EN is cardiovascular risk reduction for the adult population cared for by participating practices. This report addresses baseline hypertension control in the first 96 practices reporting and factors associated with higher percentages of control.

METHODS: Practice sites qualified for HHN if the provider staff did not exceed 10 full time equivalents and used an electronic health record (EHR). Note that 90% of primary care practices in NC utilize EHRs. Independent practices and practices that are distant from their affiliated organization were prioritized for enrollment. Data for HHN measures were extracted from electronic health records (EHRs). Given that these practices used 18 different EHR vendors, once data were extracted, they needed to be normalized using continuity care documents from the various vendors then processed into uniform data sets and dashboards. At baseline, clinicians in each practice completed a practice member survey that included a series of questions on the practice’s adaptive reserve (ARS). The primary outcome for this analysis is the proportion of patients with controlled hypertension defined as systolic blood pressure (BP) < 140 mmHg and diastolic BP < 90 mmHg. To identify factors that were associated with practices with higher percentages of patients with BP control, we performed linear regression. Independent variables incorporated in the model included practice ownership/affiliation, daily patient visits per provider, number of clinicians per site, payer mix, PCMH recognition status, the use of practice data discussions, and ARS scores.

RESULTS: EHR baseline data from 96 practices of an eventual 246 have been obtained (practice start dates were randomized). An average of 5.3 providers practiced per site. 65% of practices were clinician owned, 4% hospital owned, and 31% were federally qualified or rural health centers. The average payer mix was Medicare 25%, Medicaid 16%, dual 9%, uninsured 12%, commercial insurance 31%, “other” 6%. 65% of practices were PCMH recognized. The average proportion of patients diagnosed with hypertension who met the definition of control per practice site was 59.5% (SD +/− .17). Regression analysis yielded an adjusted R-squared of 0.36. Factors that were significantly associated with higher percentages of BP control were PCMH recognition (p = 0.04) and higher ARS scores (p = 0.04).

CONCLUSIONS: Given the correlation between adaptive reserve scores and baseline BP control, a close examination of the effect of practice facilitation on adaptive reserve will likely add to the theoretical understanding of how practice facilitation can influence practice performance. Neither payer mix nor practice affiliation was associated with BP control.

FACTORS ASSOCIATED WITH BOWEL RESECTION IN ACUTE MESENTERIC ISCHEMIA David Mossad 2; Drew Triplett1; Ronald J. Markert1; Sangeeta Agrawal1. 1Wright State University, Dayton, OH; 2Wright State University Boonshoft School of Medicine, Dayton, OH. (Control ID #2705755)

BACKGROUND: Acute mesenteric ischemia (AMI) is a serious and urgent condition. Morbidity and mortality in these presenting patients are significant, with intestinal resection at times being a major necessary intervention. The purpose of this study was to evaluate patient demographics, comorbidity burden, and hospital characteristics in patients with AMI undergoing bowel resection.

METHODS: We used ICD-9 diagnosis codes from the 2001–2010 National Hospital Discharge Survey to identify cases of AMI. ICD-9 procedure codes were then used to identify patients who underwent bowel resection. Comparisons on demographics, comorbidities, hospital length of stay (LOS), and selected health system characteristics were made between those who underwent bowel resection and those who did not. SPSS was used for chi-square and t test analysis at alpha = 0.05.

RESULTS: Among 3,441 cases of AMI, 1,393 underwent bowel resection and 2,048 did not. The resection group was similar in age (64.4 vs 65.7 years, p = 0.11) but had a longer LOS (18.0 vs 7.9 days, p < 0.001). Males were more likely to undergo resection (43.1% of males vs 38.9% of females, p = 0.015). No differences were seen in resection rate based on race (p = 0.55). Patients with hypertension (25.5% vs 44.8%, p < 0.001), coronary artery disease (CAD) (28.5% vs 41.6%, p < 0.001), and diabetes (30.3% vs 41.5%, p < 0.001) had lower resection rates. Patients who suffered acute kidney injury (AKI) (51.4% vs 38.6%, p < 0.001), and sepsis (61.3% vs 38.9%, p < 0.001) during their hospitalization were more likely to undergo resection. Those undergoing surgery were more often being discharged to a short term (37.9%) or long term (58.6%) care facility compared to going directly home (32.4%) (p < 0.001). Patients needing resection had a higher mortality rate (23.0% vs 20.4%) but this did not reach statistical significance (p = 0.067). Patients in the Northeast (35.5%) were less likely to undergo resection compared to the Midwest (42.9%), South (41.7%), and West (41.6%) (p = 0.007). Neither hospital size (p = 0.36) nor ownership (p = 0.29) were associated with resection. Principal source of payment was not associated with resection (p = 0.21). Emergent admissions (42.2%) were more likely to undergo resection compared to urgent (35.8%) and elective (37.4%) admissions (p = 0.013).

CONCLUSIONS: Bowel resection is often a necessary step in the management of AMI. This study shows that factors such as gender, LOS, regional distribution, and type of admission were associated with undergoing resection. Comorbidities potentially prohibiting safe intervention such as CAD, diabetes, and hypertension were linked to lower resection rates. Indicators of disease severity, including sepsis and AKI, pointed to a greater likelihood of resection. Those undergoing resection also appeared to have worse functional outcomes as they were less likely to be discharged home.

FACTORS ASSOCIATED WITH WELL PERSON EXAMINATIONS AMONG ADOLESCENTS AND ADULTS WITH DOWN SYNDROME Kristin M. Jensen 1, 2; Elizabeth J. Campagna2; Elizabeth Juarez-Colunga2, 2; Desmond K. Runyan2; Allan V. Prochazka1. 1University of Colorado School of Medicine, Aurora, CO; 2University of Colorado, Aurora, CO. (Control ID #2705371)

BACKGROUND: Persons with Down syndrome (DS) now live well into adulthood and are at risk for comorbidities both congenital and acquired that necessitate regular follow-up across their lifespan. Yet, little is known about their preventive healthcare patterns as they age. In this retrospective study, we evaluate factors that are associated with routine well-person care in adolescents and adults with DS.

METHODS: Using Medicaid claims data (2006–2010) in California (CA), Colorado (CO), Michigan (MI), and Pennsylvania (PA), we defined our cohort as DS patients ≥12yo, enrolled in Medicaid for ≥45/60mo without concurrent Medicare (n = 629) for a total of 3951 patients. DS cases were identified by ICD-9 = 758.0 for non-obstetric claims. Well-person examinations were identified by billing codes. The odds of receiving a well examination were modeled using logistic regression controlling for patient demographics, comorbidities, and health care utilization. The model retained all covariates regardless of significance.

RESULTS: Our cohort was 53% male, with 20% adolescent (12–17yo), 31% transition age (18–25yo), and 49% adult (≥26yo). Only 37% of the cohort received at least one well examination during this 5 year study (18% had 1 well examination, 19% had ≥2). Factors predictive of well examination included younger age (Odds Ratio [95% Confidence Limits]: 12–17yo vs ≥26yo = 1.6[1.3–2.0]; 18–25yo vs ≥26yo = 1.5[1.2–1.7]), female gender (1.7[1.4–1.9]), rural residence (1.5[1.2–1.8]), involvement in neurologic (1.5[1.2–1.8]), genitourinary (1.3[1.1–1.5]), or dermatologic organ systems (1.3[1.1–1.5]), mental health diagnoses (1.3[1.1–1.5]), the lack of respiratory (1.2[1.0–1.4]) or hematologic disorders (1.2[1.0–1.5]), and residence outside of CA (MI 4.0[2.5–6.6]; CO 3.2[2.5–4.2]; PA 4.2[3.1–5.7]). Subgroup analysis excluding CA data showed increased rates of well person examination to 64% (19% had 1 well examination, 44% had ≥2). Predictors of well examinations in the non-CA subgroup included age <26 years, female gender, at least one sick encounter, involvement in neurologic or dermatologic organ systems, and lack of hematologic disorders. Neither emergency department nor inpatient encounters were predictive of receiving well examinations, nor were the presence of comorbidities commonly-associated with DS: congenital heart disease, hypothyroidism, or obstructive sleep apnea.

CONCLUSIONS: The majority of adolescents and adults with DS do not receive well examinations routinely. Although the factors of age and gender correspond to national trends in preventive healthcare, the DS population is inherently more medically complex with risks of acquired comorbidities throughout their lives and thus requires regular follow-up. Our findings represent a significant opportunity to improve delivery of primary care to persons with DS to actively screen for and identify comorbidities with known associations in DS.

FACTORS INFLUENCING PATIENT DECISIONS TO SELF-TITRATE MEDICATIONS FOR DIABETIC PERIPHERAL NEUROPATHY TREATMENT Somalee Banerjee 2, 3; Lin Ma1; Alyce S. Adams1. 1Kaiser Permanente, Oakland, CA; 2Kaiser Permanente Oakland, San Francisco, CA; 3University of California, Berkeley, Berkeley, CA. (Control ID #2705361)

BACKGROUND: Treating diabetic peripheral neuropathy (DPN), one of the most common complications of diabetes, has been challenging due to limited effectiveness of medications, and the need to titrate to balance symptom relief and side effects to prevent medication non-compliance. This study assesses patient self-titration of DPN medications within a larger randomized trial, the Diabetes Telephone Study (DTS), to improve treatment outcomes through automated monitoring of symptoms and treatment side effects.

METHODS: Self-reported data were collected from patients in the intervention arm of DTS between September 2014 and July 2016 and integrated with demographic and clinical data from the electronic health record. The primary outcome was patient initiated titration, as reported via 3 interactive voice response calls over 8 months. Chi square tests were used to examine associations between each of the titration outcomes and patient demographics, drug coverage benefits and clinical characteristics. We used multiple logistic regression models to identify independent predictors of patient titration.

RESULTS: Of the 605 patients studied, 195 or 32% of these patients reporting self-titration during the 8 months following treatment start. Self-titration was associated with gabapentin or nortriptyline use, higher baseline pain interference and being an English speaker, but age, sex, race, chronic pain diagnoses, diabetes severity and other related complications were not. After controlling for baseline demographic and clinical features, factors associated with higher odds for self-titration included: higher baseline pain interference scores (OR: 1.02; 95% CI: 1.00, 1.04); age between 65 and 74 as compared to those older than 75 (OR: 1.62; 95%CI: 1.02, 2.58), speaking English (vs. Spanish) (OR: 5.08; 95%CI: 1.41, 18.30) and more months of full drug coverage benefits (OR: 1.79; 95%CI: 1.01, 3.18). Compared to those taking gabapentin, patients taking nortriptyline (OR: 0.58; 95%CI: 0.38, 0.89) had lower odds of self-titration.

CONCLUSIONS: In summary, self-titration of DPN pain medications in this population was associated with the type of medication prescribed and higher baseline pain levels. Importantly, patients with more continuous drug coverage had nearly twice the odds of self-titration, underscoring the importance of medication cost to patient decision-making. Also, while being an English (vs. Spanish) speaker dramatically increased one’s odds of self-titration, the estimates lacked precision due to the small number of Spanish speaking patients in this sample. Further exploration using data from clinical settings is needed to confirm these findings and explore potential barriers to optimal pharmacologic treatment for Spanish-speaking patients.

FACTORS PROMOTING RECOVERY FROM BURNOUT AMONG INTERNAL MEDICINE RESIDENTS: A QUALITATIVE STUDY Nauzley C. Abedini; Shobha W. Stack; Jessie L. Goodman; Kenneth P. Steinberg. University of Washington, Seattle, WA. (Control ID #2673813)

BACKGROUND: Burnout has been well documented among physicians at all career stages, yet little is known about how physicians recover from and avoid recurrent burnout. Residents are particularly vulnerable, and burnout rates for internal medicine residents are among the highest for all specialties. Using qualitative methods, we sought to identify factors promoting recovery from and avoidance of recurrent burnout among residents in the University of Washington Internal Medicine Residency Program.

METHODS: Between June and August 2016, 25 in-depth semi-structured, 60-min interviews were conducted with a convenience sampling of PGY2, PGY3, and recent graduates of the University of Washington Internal Medicine Residency Program who both experienced and recovered from burnout during residency. Residents were deemed ineligible for participation if they were currently experiencing burnout, as determined by a validated single-item burnout measure. Interviews were audio-recorded and transcribed verbatim. Three investigators independently performed coding in an iterative fashion consistent with Grounded Theory. Investigators compared open and axial codes and reached a consensus regarding major themes.

RESULTS: Participating residents identified factors that put them at risk for burnout, including: uncertainty in their professional role and expectations; lack of connection with patients and co-workers; lack of awareness of burnout syndrome; personal challenges outside of work; workplace challenges; and physical exhaustion. Participants cited multiple factors that contributed to their recovery and avoidance of recurrent burnout, namely (1) professional identity formation through guided reflection, role models, future career planning, setting boundaries, and/or improving medical knowledge; (2) connection with patients, co-workers, family and friends; (3) awareness of burnout and feeling validation of their experience from others; (4) time away from clinical responsibilities; and (5) resolution of personal challenges. The majority of participants reported deriving meaning in medicine from relationships with patients and co-workers, and that burnout was often linked to a loss of that meaning.

CONCLUSIONS: Residents utilize a number of different strategies to recover from burnout. Thus, interventions to alleviate burnout might reflect these strategies. We hypothesize that efforts should be made to (1) raise awareness of burnout and create opportunities for dialogue to validate resident experiences; (2) improve systems and circumstances that lead to fatigue and work compression, and promote self-care; (3) facilitate professional identity formation through reflection and mentorship; and (4) foster connection and relationships with patients and colleagues. In particular, helping residents identify where they derive meaning in medicine and supporting them in building that meaning into their work may be protective against burnout.

FAILURES IN REPEAT DIABETES SCREENING IN PATIENTS WITH PREDIABETES IN AN URBAN SAFETY-NET HEALTH CARE SYSTEM Brooks Brodrick; Hua Lin; Ethan Halm; Michael E. Bowen. Univ of TX Southwestern Med Ctr, Dallas, TX. (Control ID #2706186)

BACKGROUND: National diabetes screening guidelines recommend repeat screening every 12 months for those with prediabetes and every 36 months for those with normal screening results. However, little is known about frequency and timing of repeat diabetes screening in real world clinical practice.

METHODS: We conducted a retrospective cohort study in a large, integrated safety-net healthcare system using electronic medical record (EMR) data from 2010 to 2014. Eligible patients were age 18–65, non-pregnant, and had an index, outpatient visit and a resulted gold standard diabetes screening test (fasting glucose or hemoglobin A1C) between January 2010 and June 2013. Eligible patients had ≥1 additional office visit before December 2014. Patients with diabetes were excluded using ICD-9 codes and test results consistent with diabetes. We describe the frequency of rescreening and examine characteristics associated with rescreening in patients with initial test results in the normal and prediabetes range. We used Cox proportional hazards regression to describe the time to repeat diabetes testing in normal vs. prediabetes patients. Multivariate models adjusted for age, sex, race, BMI, hypertension, hyperlipidemia, and family history.

RESULTS: A total of 18,323 patients met inclusion criteria. On average, patients were 48 years old and had a BMI of 31.7. Overall, 62% were female, 83% were non-white, 53% had a family history of diabetes, and 57% had hypertension. Only 37% of patients completed repeat screening (mean time to follow-up time was 398 days). Rescreening was more common among those that were older, female, had a higher BMI, prediabetes, hypertension, and hyperlipidemia (p < 0.01). Those with prediabetes on their initial screening (N = 6585) were more likely to be rescreened compared with those with normoglycemia (HR = 1.33) in multivariate analysis. However, at 12 months, only 25% of those with prediabetes had been rescreened, and by 18 and 36 months, repeat screening improved to only 34 and 41% respectively. Those with high-risk prediabetes (A1C ≥ 6% or FBG ≥ 110 mg/dL) were more likely be rescreened than those with low risk prediabetes at 12 months (34% vs 18%), 18 months (42% vs. 28%), and 36 months (49% vs. 36%) (p < 0.01 for all). Among those who were rescreened, 17% transitioned from normal to prediabetes, 3% transitioned from normal to diabetes, and 9% transitioned from prediabetes to diabetes.

CONCLUSIONS: Although patients with prediabetes are more likely to be rescreened than those with normoglycemia, only slightly more than one third of patients with prediabetes completed repeat diabetes screening at 18 months. Even among those with high risk pre-diabetes, no more than half were rescreened within 3 years. Better, more systematic, population health approaches to promote timely rescreening in patients with prediabetes are needed.

FEASIBILITY AND ACCEPTABILITY OF A GROUP MEDICAL VISIT INTERVENTION TO IMPROVE HCV TREATMENT UPTAKE AMONG PERSONS WHO INJECT DRUGS (PWID) IN A PRIMARY CARE SETTING Brianna L. Norton 4; Mary Gover4; Cara McMurry3; Joseph Deluca4; Chinazo Cunningham2; Alain H. Litwin1. 1Albert Einstein College of Medicine, Bronx, NY; 2Albert Einstein College of Medicine & Montefiore Medical Center, Bronx, NY; 3Montefiore, Brooklyn, NY; 4Montefiore Medical Center, Bronx, NY. (Control ID #2707549)

BACKGROUND: Throughout the US, the majority of people infected with HCV (hepatitis C virus) are persons who inject drugs (PWID), yet very few have ever initiated treatment. We set out to investigate the feasibility and acceptability of implementing a primary care based HCV Group Evaluation and Treatment Uptake (HCV GET-UP) intervention to improve HCV treatment uptake among PWID.

METHODS: Patients were recruited from a primary care clinic in the Bronx, NY. Participants were considered potentially eligible if they had a positive HCV antibody in the electronic health record. Patients were then called for study screening, and if eligible, asked to attend a baseline visit where patients signed written consent, were asked baseline demographic information, and provided urine for drug toxicology screening. The study intervention consisted of 4-weekly group medical visits with a physician where HCV-related medical evaluation/work-up, education, support, and skill building were provided. All visits were billed to insurance as a 1-hour group medical visit. At the end of the intervention, patients were offered an expedited appointment within 2 weeks for onsite HCV treatment at the primary care clinic. We assessed feasibility via recruitment and retention rates, and acceptability via brief post-group visit surveys with 5-point Likert scale (not helpful-very helpful).

RESULTS: Among 40 patients determined to be preliminarily eligible (HCV Ab+) for the study, phone contact was made with 27 (67.5%). Of the 27 patients reached, 13 (48.1%) persons agreed to be screened, and 7 were ultimately enrolled. Among the seven participants, 71.4% were male, all were either African American or Latino/Hispanic, with a median age of 54.5 (41.0-63.0). The majority were on public insurance (85.8%), over half received food stamps (57.1%), and one person was homeless (14.3%). Nearly all were on opioid agonist treatment (85.7%), and three (42.9%) were actively using drugs at baseline. Six participants made it to at least one group visit (participant range 4–6 at each group). Acceptability was high among participants; all components were found to be very helpful (median Likert scores for medical evaluation: 5, education: 4.8 (4–5), skill building: 5, group activity (support): 5. Of the 6 participants who attended at least one visit, 5 attended their follow-up appointment for HCV treatment and 4 have initiated treatment.

CONCLUSIONS: These findings indicate that a primary care based HCV Group Evaluation and Treatment Uptake (HCV GET-UP) intervention for PWID is both feasible and acceptable. We will now commence a randomized control trial to test this intervention for improving HCV treatment uptake and cure among PWID in primary care. This intervention has potential to improve individual health outcomes, while reducing the prevalence of HCV in a community with continued transmission.

FEMALES IN ACADEMIA WITH THE FEMALE MENTORS HAVE A DECREASED LIKELIHOOD OF PROMOTION: RESULTS FROM A NETWORK ANALYSIS OF ACADEMIC GENERALISTS Krisda H. Chaiyachati 1; Joshua M. Liao1; Gary E. Weissman1; Rebecca Hubbard2; Anna U. Morgan1; Anna Buehler3; Judy A. Shea1; Katrina Armstrong4. 1The Perelman School of Medicine at the University of Pennsylvania, Philadelphia, PA; 2University of Pennsylvania, Philadelphia, PA; 3UC San Diego Medical School, San Diego, CA; 4Massachussetts General Hospital, Boston, MA. (Control ID #2706314)

BACKGROUND: While mentorship is critical for academic careers, the association between academic promotion and mentorship attributes is poorly understood. Moreover, we hypothesized that mentor attributes may affect the known gender disparities in promotion. We used network analysis methods to explore the relationship between mentorship and promotion among academic generalists trained in internal medicine, pediatrics, and family medicine.

METHODS: Through online searches and contacting programs, we identified graduates of generalist-oriented fellowships (e.g., general medicine, RWJ Clinical Scholars Program) between 2002–2007. We manually extracted PubMed publication data and electronically surveyed graduates to obtain information about grants, academic rank, and their top 10 most influential career mentors at the 5-year mark post-fellowship. With respondent mentees and nominated mentors represented as vertices, and mentor relationships as edges, we constructed a generalist community graph and generated 2 network measures: percent of common gender dyads (i.e., gender concordant mentee-mentor pairs) and mentor centrality (i.e., how connected a mentor is to other network mentees). Using multivariable logistic regression adjusted for respondent and training institution characteristics, we evaluated the impact of common gender dyads, mentor centrality, and the interaction between the two on promotion to associate professor or higher, 5 years post-fellowship. We also calculated “gender assortativity” for the community graph, a composite measure of mentee preference for the same gender mentors [range: −1 (pref. for opposite-gender) to 1 (pref. for same-gender)].

RESULTS: In total, 162 graduates representing 19 institutions completed our survey (51% response rate). Among the 136 who remained in academia 5 years post-fellowship, the majority were female (65%) and white (67%). At the 5-year mark, respondents reported a median of 3 mentors (IQR 3–5) and 10 peer reviewed publications (IQR 4–18). Thirty-nine percent reported < $500 K in grants and 13% achieved associate professor or higher. In multivariable analysis, neither gender nor percent of common gender dyads independently predicted promotion. However, the association between common gender dyad and promotion differed for men and women respondents with an odds ratio (OR) of 8.61 for men and 0.06 for women (interaction term p-value = 0.03). Other significant associations with promotion included mentor centrality, publications, and grant dollars. The assortativity of the mentorship graph was 0.11 (p < 0.001).

CONCLUSIONS: We identified facets of mentorship that may affect early career promotion. While more work is needed, we found the relationship between promotion and common gender dyads varied by gender—female mentees are less likely to be promoted with more female mentorship compared to males mentees with more male mentorship. Educators should consider mentorship disparities to help generalists, irrespective of gender, achieve academic promotion.

FINANCIAL IMPLICATIONS OF NON-VISIT-BASED PRIMARY CARE UNDER CAPITATED PAYMENT: A MODELING STUDY Sanjay Basu 3; Russell Phillips1; Zirui Song2; Asaf Bitton1; Bruce E. Landon1. 1Harvard Medical School, Boston, MA; 2Massachusetts General Hospital, Boston, MA; 3Stanford University, Stanford, CA. (Control ID #2701995)

BACKGROUND: Capitated payments may provide incentives for primary care practices to shift towards new care models. Our objective was to simulate whether shifting toward team- and non-visit-based care is financially sustainable for internal medicine primary care practices under traditional fee-for-service (FFS), capitated payment, or a mix of the two.

METHODS: We utilized a microsimulation model of utilization, cost and revenue under different care models at U.S. primary care practices, which sampled from data in the Medical Group Management Association national Cost and Revenue survey (N = 969 primary care practices, 2014). We externally validated the model by ensuring it was within 5% absolute error from utilization, revenue, and cost estimates (by age, sex, race/ethnicity, primary ICD-9 diagnostic code, and insurance type) among participants in the National Ambulatory Medical Care Survey (NAMCS, 2014, N = 31,229 patients) and an independent national survey of N = 2,518 practices. In the model, simulated practice revenues and costs were computed across a range of FFS and capitated payments, with the capitated proportion of payment varied from 0% (traditional FFS) to 100% (fully capitated). Revenues and costs were compared before and after substitution of low-complexity physician visits (those for low-risk diagnoses without any laboratory testing, medication prescriptions, or imaging) with team- and non-visit-based services (nurse visits, nurse/physician telephone/email visits, and care coordination with a medical assistant).

RESULTS: Substitution of low-complexity visits with team- and non-visit-based services produced financial losses for simulated practices under FFS payment, reducing net revenues from $78,288 per full-time physician per year (95% CI: $12,517, $98,552) to $35,890 (95% CI: −$42,846, $91,283). Substitution produced financial gains for simulated practices under capitated payment if >68% of annual payment was capitated. At 100% capitation, net revenues increased to $205,418 (95% CI: $140,206, $222,780) per full-time physician, due to growth in panel size. When practices received a bonus of 0.6% of total expenditures, substitution of low-complexity produced financial gains for simulated practices if >58% of annual payments were capitated.

CONCLUSIONS: A shift to capitated payment may create an incentive for practices to increase non-visit-based primary care delivery, but a high share of capitation may be required to make such a shift financially attractive.

FINANCIAL INCENTIVES FOR SMOKING ABSTINENCE IN HOMELESS SMOKERS: A RANDOMIZED CONTROLLED TRIAL Travis P. Baggett 4, 3; Yuchiao Chang4, 1; Awesta Yaqubi2; Claire McGlave2; Stephen Higgins5; Nancy A. Rigotti2, 1. 1Harvard Medical School, Boston, MA; 2Massachusetts General Hospital, Boston, MA; 3Boston Health Care for the Homeless Program, Boston, MA; 4Division of General Internal Medicine, Massachusetts General Hospital, Boston, MA; 5Vermont Center on Behavior and Health, University of Vermont, Burlington, VT. (Control ID #2707557)

BACKGROUND: Three-quarters of homeless adults smoke cigarettes, contributing to 3- to 5-fold higher rates of tobacco-attributable mortality compared with the general population. Homeless smokers want to quit smoking, but the proportion who are able to do so is one-fifth the national average and studies of conventional treatment approaches have yielded modest results. Financial incentives for smoking abstinence have shown promise in non-experimental studies of homeless smokers, but randomized trial data are lacking in this population. We conducted a pilot randomized controlled trial of financial incentives for smoking abstinence among homeless smokers in Boston, MA.

METHODS: All study procedures occurred between October 2015 and June 2016 at Boston Health Care for the Homeless Program. Eligibility criteria included age ≥18 years, current smoking (≥5 cigarettes/day, ≥100 cigarettes lifetime), current homelessness, and desire to quit smoking in the next month. Individuals were not excluded for current drug or alcohol use. Participants randomized to the financial incentives arm (N = 25) could earn escalating monetary rewards ($15-$35) for carbon monoxide (CO)-defined smoking abstinence (<8 parts per million [ppm]), assessed 14 times over 8 weeks. Control arm participants (N = 25) were given $10 at each assessment visit, regardless of their exhaled CO levels. All study payments were made onto reloadable debit cards issued to each participant at enrollment. Participants in both arms were offered nicotine patches and weekly behavioral counseling. The primary outcome was a repeated measure of point-in-time smoking abstinence (CO <8 ppm) across the 14 study visits. The secondary outcome was point-in-time abstinence at 8 weeks. Exploratory smoking outcomes were self-reported past 1-day and 7-day abstinence from 1) any cigarette and 2) any puff of a cigarette. Other outcomes included quit attempts, nicotine patch use, counseling attendance, and changes in alcohol and drug use severity scores.

RESULTS: Compared with control arm participants, financial incentive arm participants had greater point-in-time abstinence overall (OR 7.28, 95% CI 2.89, 18.3) and at the end of treatment (48.0% vs. 8%, p = 0.004). Similar effects were seen for past 1-day abstinence, but the rates of past 7-day puff abstinence were negligible (0-4%) in both arms. Incentive arm participants made significantly more past-month quit attempts at 4 weeks (p = 0.03). Nicotine patch use (p = 0.80) and counseling attendance (p = 0.43) did not differ significantly between groups. There were no significant changes in alcohol or drug use severity scores in either group.

CONCLUSIONS: Among homeless smokers, financial incentives increased brief smoking abstinence and quit attempts without worsening substance use. This approach is promising but requires modification to achieve more sustained smoking abstinence in this vulnerable population.

FINANCIAL INCENTIVES OF HIGH-DEDUCTIBLE PLAN MEMBERS AND CLINICIAN-BASED PREDICTORS OF LOW-VALUE BACK IMAGING Arthur Hong; Dennis Ross-Degnan; Fang Zhang; James F. Wharam. Harvard Medical School and Harvard Pilgrim Healthcare, Boston, MA. (Control ID #2701134)

BACKGROUND: Physicians and policy-makers often cite patient demand as a driver of unnecessary medical care, while other analysts point to unexplained variations in clinician practice. Consumer-driven health plans (CDHPs) are touted to reduce patient demand, are increasingly prevalent, and are prominent among proposals to replace the Affordable Care Act. We examined whether patient financial incentives outweigh clinician characteristics as predictors of one of the most common low-value services: low-value back pain imaging.

METHODS: Using commercial insurance claims from 2010–2014, we used multivariate logistic regression to model the likelihood of imaging for acute uncomplicated back pain, in adults aged 18–64. We used claims algorithms from the literature to define imaging (X-ray, CT scan, or MRI) after a back pain visit, excluding patients with red flag symptoms such as cancer or neurologic deficit. Patient predictors were patient demographics, CDHP enrollment (high-deductible plan with health savings account), and whether a patient had zero out-of-pocket cost for the defining clinician visit and thus likely faced no cost sharing for potential imaging. Clinician predictors were clinician type (primary care, specialist, chiropractor, physician assistant/nurse practitioner), imaging equipment ownership, and if a clinician’s prior patient received imaging. Ownership was established from claims details using validated methods. We used marginal effects to estimate adjusted imaging rates by clinician type and ownership.

RESULTS: Our sample had 1,017,440 visits seen by 103,041 clinicians. CDHP members had 0.91 [95% CI: 0.89, 0.92] lower odds of low-value image, and patients with no cost sharing for a potential image had 1.26 [95% CI: 1.23, 1.29] higher odds. If a clinician owned imaging equipment, low-value image odds were 1.97 higher [95% CI: 1.94, 2.00], and marginally adjusted rates of image were higher for all types of clinician owners (figure). If a clinician’s prior patient received imaging, odds of image were 2.57 higher [95% CI: 2.54, 2.60].

CONCLUSIONS: Clinician factors appeared to be stronger predictors of low-value back imaging than patient financial incentives. Policymakers should be aware of the limitations of increasing patient cost sharing to reduce unnecessary care.

figure ad

FIREARM STORAGE PRACTICES, RISK PERCEPTIONS, AND PLANNED SUICIDE PREVENTION ACTIONS AMONG U.S. VETERANS WITH AND WITHOUT SELF-HARM RISK FACTORS Joseph A. Simonetti 1, 2; Deborah Azrael3; Matthew Miller4. 1Denver VA Medical Center, Denver, CO; 2University of Colorado School of Medicine, Denver, CO; 3Harvard TH Chan School of Public Health, Boston, MA; 4Northeastern University, Boston, MA. (Control ID #2704086)

BACKGROUND: Veterans accounted for 18% of all U.S. adult suicides in 2014, two-thirds of which involved firearms. Despite the disproportionate use of firearms in veteran suicides and the well-established link between firearm access and suicide risk, little is known about Veterans’ firearm storage practices or their beliefs about the relationship between firearm access and suicide risk.

METHODS: Data come from a nationally representative, Web-based survey designed by the investigators and conducted in April 2015 by the survey firm Growth for Knowledge. Respondents were drawn from a group of approximately 55000 U.S. adults selected (on an ongoing basis) with an equal probability of selection. Primary outcomes were the proportion of veteran firearm owners who: 1) stored any household firearm loaded and unlocked, 2) agreed or strongly agreed that a household firearm increases suicide risk, and 3) reported they would limit firearm access were a household member to become suicidal. We categorized respondents based on the presence of any self-harm risk factor (yes/no), including chronic pain, mood or anxiety disorder, schizoaffective disorder or schizophrenia, and alcohol intake in excess of recommended gender-specific limits. We examined how outcomes varied based on risk for self-harm. We applied weights to account for oversampling and nonresponse and describe findings using weighted proportions and 95% confidence intervals (CI).

RESULTS: Of 7318 panel members invited to participate, 3949 (54.6%) completed the survey and 580 met inclusion criteria. The mean age of Veteran firearm owners was 62.8 years; 94.6% (CI 92.0,96.4) were male and 33.4% (CI 28.8,38.4) had at least one self-harm risk factor. Thirty-four percent (CI 28.8,38.7) of Veteran firearm owners reported storing at least one firearm loaded/unlocked and 6.2% (CI 4.5,8.6) agreed or strongly agreed that firearms increase household suicide risk. Findings did not differ substantially by self-harm risk factor status. A greater proportion of firearm owners with (than without) self-harm risk factors indicated they would limit firearm access if a household member were suicidal [89.3% (CI 82.2,93.8) vs. 78.6% (CI 72.3,83.8)]. Of those who would take steps to limit firearm access, 24.6% (CI 20.2,29.6) would store firearms away from home and 21.5%