INTRODUCTION

The US healthcare system is fraught with low-value care, with up to $101 billion in estimated waste in costs annually.1 Low-value care is defined as a healthcare service in which evidence shows little or no benefit for patients, and the potential of harm exceeds benefit.2 As a part of the American Board of Internal Medicine Foundation’s (ABIM-F) Choosing Wisely® (CW) campaign, more than 80 specialty societies have published over 500 recommendations of reducing low-value care since its launch in 2012.3

Although the campaign achieved significant clinician awareness since its inception (40%), the progress has plateaued in recent years.4 Furthermore, large-scale implementation of CW recommendations remains a challenge, and additional interventions including patient-focused strategies have been suggested.5 Surveys of physicians have included fear of malpractice and patient concern or request as the most common drivers of overuse.4, 6 The public may be skeptical of top-down efforts to curb low-value care, perceiving them as rationing.7 Thus, patient input is critical in reducing healthcare waste and an important next step for increasing uptake of CW recommendations among both clinicians and patients.8, 9

Past efforts at patient engagement in CW list development were predominantly aimed at translation and dissemination of recommendations created by clinician societies to patients.8, 10 Patients have yet to play a pivotal role in defining value from its inception. In this study, we utilized a multi-institutional, co-creational approach that included patient advocates in conjunction with clinicians in CW list development for hospital medicine. Patient advocates were selected as they are familiar with healthcare operations and have the unique experience of hearing multiple patient perspectives, in addition to being patients themselves.

METHODS

The Society of Hospital Medicine (SHM) Hospital Quality and Patient Safety Committee convened the High Value Care Committee (HVCC) in July 2017. Monthly conference calls were held from July 2017 to January 2020. The group consisted of 19 members total, some of which did not serve the entire period. All members are considered to be experts in quality and safety and were self-selected to be a part of the HVCC.

  • Phase 1: Crowdsourcing and Brainstorming

    An online questionnaire requesting examples of low-value care in adult hospital medicine was sent from October 5 to 31, 2017, to the SHM listserv, along with ABIM Foundation and affiliated social media outlets (Fig. 1). The target audience included both clinicians and non-clinicians. Upon receiving initial feedback that the questionnaire was unclear for some non-clinicians, a second patient-friendly questionnaire was created with the help of the patient advocates from the Right Care Alliance Community Engagement Council (Supplement 1). This is a grassroots organization including patients and community members whose vision focuses on a healthcare system that is universally accessible, equitable, and affordable.11

    All examples of low-value care from the questionnaire were compiled, edited, counted, and categorized into 5 domains: laboratory, imaging, medication, diagnostics, and other. Duplicate or similar recommendations were also taken into account. Recommendations in the previous SHM CW Top 5 list were removed. All items with 10 or more mentions were taken into the next phase in an effort to capture the most prominent themes.

  • Phase 2: Literature Search and Developing Recommendations

    All items brought into this phase were individually reviewed and discussed through an iterative process. Items were divided among HVCC members, and a literature search was performed in PubMed database. Focused recommendations were developed and presented to the committee for review. Edits were made according to feedback and citations were added to each recommendation. Items that were duplicative or had insufficient evidence to support the recommendation were removed, leaving 22 items. For the purposes of this step in the process, adequate evidence was defined as the presence of any existing peer-reviewed studies that validated the recommendation. These sources were referenced so that voting members could determine the strength of the evidence in the modified Delphi voting process.

  • Phase 3: Modified Delphi Voting

    For the remaining recommendations, a Delphi scoring process was utilized to reach consensus among clinicians and patient advocates.12 Online voting surveys (SurveyMonkey®) were developed. They included the semi-final recommendations, accompanied by evidence-based citations. To minimize bias, members of the HVCC who created the survey or were a part of the Right Care Alliance did not vote, which left a total of 7 HVCC members. To ensure an equal representation of the patient perspective, 7 patient advocates were identified by HVCC members from different institutions across the country representing both community and teaching hospitals in urban and non-urban settings. No other specific criteria were used to identify patient advocates.

Fig. 1
figure 1

Society of Hospital Medicine Adult Choosing Wisely List Development Process Map

Full size image

For each recommendation on the voting survey, clinician respondents were asked to rate on a 1–5 Likert scale on three criteria: (1) strength of evidence, (2) potential for avoiding patient harm, and (3) relevance to hospital medicine.

For patient advocates, a patient-friendly version of the recommendations was created for the voting survey. Due to the clinical complexity of the items, a standardized patient-friendly script was created. The script consisted of an explanation of normal state, abnormal state, rationale for recommendation, and potential harms (Supplement 2). Two HVCC members who did not participate in the clinician voting process used the scripts during calls with patient advocates, lasting 1 hour each. Once each recommendation was explained, the patient partner completed the voting survey independently. They were asked to rate each recommendation based on the same Likert scale on slightly different criteria: (1) strength of evidence, (2) potential for avoiding patient harm, and (3) relevance to patients.

For each recommendation, a mean composite score was calculated by accounting for strength of evidence, patient harm, and relevance. This was calculated for patient advocate, clinicians, and combined. Data was presented numerically by absolute value and included a standard deviation.

RESULTS

The initial questionnaire was sent through SHM on October 12, 2017, to 28,906 individuals through an email with a subject heading “Provide Us Your Feedback.” Additionally, the questionnaire was disseminated via emails and social media feeds from the society, ABIM-F, and HVCC members.

There were 285 respondents that completed the questionnaire with1234 items (Supplement 3). The respondents included 194 (68%) hospitalists, 16 (6%) residents, 15 (5%) nurse practitioners, 14 (5%) patients, 8 (3%) nurses, 6 (2%) fellows, 6 (2%) medical students, 6 (2%) physician assistants, 3 (1%) patient advocates, and 17 (6%) others.

The second questionnaire was made in conjunction with patient advocates from the Right Care Alliance. It consisted of 4 questions assessing if any medical care was unnecessary. This resulted in 31 additional items for review.

A total of 1265 items were reviewed. After accounting for similar suggestions, there were 283 items, of which 73 were in medication domain, 65 in laboratory testing, 43 in imaging, 25 in diagnostic, and 77 in other domain. Twenty-seven of these recommendations had 10 or more mentions, and advanced into phase 2. The council reviewed these recommendations and eliminated 5 recommendations based on lack of evidence or inclusion in previous Choosing Wisely guidelines, leaving 22 items for review.

During phase 3, a total of 7 hospitalists and 7 patient partners completed the online survey. Results were tallied and the mean response scores based on the three categories were compiled for patient advocates, clinicians, and scores combined (Fig. 2). The aim was to have five recommendations; however, two recommendations received identical scores, leaving six total recommendations.

Fig. 2
figure 2

Choosing Wisely 2.0 recommendations: combined clinician/patient partner voting

Full size image

RECOMMENDATIONS

  1. 1.

    Avoid using opioids for treatment of mild, acute pain. For moderate to severe acute pain, if opioids are used, it should be in conjunction with non-opioid methods with the lowest effective dose for the shortest required duration.

  2. 2.

    Do not maintain a peripheral capillary oxygen saturation (SpO2) of higher than 96%, when using supplemental oxygen, unless treating for carbon monoxide poisoning, cluster headaches, sickle cell crisis, or pneumothorax.

  3. 3.

    Do not wake patients at night for routine care; redesign workflow to promote sleep at night.

  4. 4.

    Do not order creatine kinase (CK) or creatine kinase-myocardial band (CK-MB) in suspected acute coronary syndrome or acute myocardial infarction.

  5. 5.

    Do not order daily chest radiographs in hospitalized patients unless there are specific clinical indications.

  6. 6.

    Do not routinely prescribe VTE prophylaxis to all hospitalized patients; use an evidence-based risk stratification system to determine whether a patient needs VTE prophylaxis. If they do warrant prophylaxis, use a bleeding risk assessment to determine if mechanical rather than pharmacologic prophylaxis is more appropriate.

Recommendation 1: Avoid using opioids for treatment of mild, acute pain. For moderate to severe acute pain, if opioids are used, it should be in conjunction with non-opioid methods with the lowest effective dose for the shortest required duration.

The use of opioids has serious risks including opioid use disorder and overdose. If opioid therapy is required, pain management with short acting opioids should be the lowest effective dose for the shortest required duration, often no more than 1 week 13. A trial of non-opioid and non-pharmacological alternatives is recommended for opioid naïve patients. If opioids are used, they may be used in conjunction with non-opioid methods, when clinically appropriate 13.

For patients already on opioids for chronic pain, it is not recommended to abruptly stop or taper opioid therapy to avoid withdrawal, mental health crisis, and overdose 14. Individualized treatment plans should be made with the patient, and outpatient clinicians, whenever possible.  It is important for the clinician to assess potential biases that may affect treatment of pain 14.

Recommendation 2: Do not maintain a peripheral capillary oxygen saturation (SpO2) of higher than 96%, when using supplemental oxygen, unless treating for carbon monoxide poisoning, cluster headaches, sickle cell crisis, or pneumothorax.

Delivering oxygen to prevent hypoxemia must be balanced with the possible harms of over-oxygenation including cytokine production, decreased cardiac output, coronary vasoconstriction, and acute lung injury.15 In a systematic review and meta-analysis of 25 studies and over 16,000 patients, liberal oxygen saturation levels above 94–96% increased in-hospital mortality by 20%. The overuse of supplemental oxygen has been shown to increase mortality in numerous studies of patients with critical illness including cardiac arrest, stroke, trauma, and emergency surgery.16 An important caveat exists to this recommendation as studies published after these guidelines were created demonstrate a higher incidence of occult hypoxemia, defined as an arterial oxygen saturation of less than 88% with a pulse oximetry measurement of 92 to 96%, in black patients compared to white patients17.

Recommendation 3: Do not wake patients at night for routine care; redesign workflow to promote sleep at night.

Inadequate sleep in hospitalized patients has been associated with poor outcomes including high blood pressure, hyperglycemia, immune dysfunction, and delirium. Environmental factors (noise, light disruptions), care-related factors (blood draws, vital signs), and patient factors (illness, pain) all contribute to sleep disruption. It is recommended that non-pharmacologic interventions be the first line of prevention.18 Although data is limited and the quality of evidence is low, multifaceted interventions targeting modifiable factors including nighttime interventions to decrease noise and light, relaxation techniques, and minimizing unnecessary patient contact may improve sleep quality and duration.19 Non-pharmacologic sleep aids including earplugs and eye masks can be utilized and may provide some benefit.20.

Recommendation 4: Do not order creatine kinase (CK) or creatine kinase-myocardial band (CK-MB) in suspected acute coronary syndrome or acute myocardial infarction.

The American College of Cardiology Foundation and American Heart Association’s guidelines for management of myocardial infarction recommend using cardiac troponin as the lab test of choice to diagnose acute coronary syndrome or acute myocardial infarction. Troponin is highly sensitive for AMI and more specific than CK-MB.21 Despite the superiority of troponin compared to CK-MB, many centers continue to use CK-MB. This is thought to be a result of clinicians mistakenly believing that CK-MB is more useful in certain clinical scenarios. CK-MB yields no incremental diagnostic value even in patients with chronic kidney disease. No study has shown that CK-MB is superior over troponin in detecting re-infarction.22.

Recommendation 5: Do not order daily chest radiographs in hospitalized patients unless there are specific clinical indications.

Patients in the intensive care unit have historically had daily chest x-rays as part of routine management to detect complications with endotracheal tubes and central venous catheters. The use of daily chest x-rays leads to excessive test utilization, unwarranted exposure to radiation, and additional cost. A systematic review and meta-analysis of 9 studies with close to 40,000 chest x-rays showed no difference in mortality, ICU length of stay, or mechanical ventilation days when comparing a restrictive x-ray strategy with routine, daily chest x-rays.23.

Recommendation 6: Do not routinely prescribe VTE prophylaxis to all hospitalized patients; use an evidence-based risk stratification system to determine whether a patient needs VTE prophylaxis. If they do warrant prophylaxis, use a bleeding risk assessment to determine if mechanical rather than pharmacologic prophylaxis is more appropriate.

Venous thromboembolism (VTE) is a major cause of morbidity and mortality in hospitals. Pharmacologic prophylaxis has been shown to reduce the risk of clinically significant VTE. While VTE prevention should be considered for every hospitalized patient, excess VTE prophylaxis—prophylaxis inappropriately administered either to patients at low risk of VTE or to high-risk patients with contraindications—may be harmful. The CHEST antithrombotic guidelines recommend objective risk stratification for VTE prevention in hospitalized medical patients.24.

The top recommendations of clinicians were minimizing overuse of oxygen (4.52), opioids (4.43), improving sleep (4.38), reducing CT head in syncope (4.38), and reducing CK-MB use (4.29) (Table 1). The top recommendations from patient partners were decreasing daily CXR (4.76), reducing CKMB use (4.71), opioids (4.71), improving sleep (4.67), appropriate VTE ppx (4.62), and reducing unnecessary IV fluids (4.62).

Table 1 Mean Response Scores with Mean Confidence Intervals for Recommendations for Patient Advocates, Clinicians, and Combined Scores. All Results Are Based on Responses from a 1-to-5 Likert Scale
Full size table

Three common selections among patient advocates and clinicians were included in the final recommendations. After combining answers, the top recommendations were opioids (4.57), improving sleep (4.52), minimizing overuse of oxygen (4.52), reducing CK-MB use (4.50), appropriate VTE ppx (4.43), and decreasing daily CXR (4.43). The lowest-rated recommendations included reducing inappropriate physical therapy consults (3.50), folate (3.79), reducing inappropriate consults in general (3.81), reducing urine electrolyte usage (3.83), and decreasing docusate (3.86).

The top-rated recommendations that would reduce patient harm include reducing opioids (4.7), improving sleep (4.43), reducing antibiotic usage for viral infections (4.43), reducing unnecessary IV fluids (4.43), and reducing overuse of oxygen (4.29). Three of these recommendations were included in the final recommendations.

Top-rated recommendations for strength of evidence included reducing overuse of oxygen (4.71), ordering chest x-rays daily (4.71), reducing CK-MB use (4.71), CT for syncope (4.64), and use of carotid imaging for syncope (4.62). Again, three of these recommendations were included in the final recommendations.

CONCLUSION

Through a collaborative process with patient advocates, we successfully co-developed six CW recommendations to reduce low-value care in adult hospital medicine. We built upon CW campaign’s principal aim of promoting conversation between clinicians and patients by integrating the concept of co-production in healthcare.25 Many clinicians are more motivated by focusing on patient-centered concerns and patient safety, rather than cost alone. Through this approach, we hope to engage more clinicians in implementing the CW recommendations.26 Moreover, the vast majority of published efforts in CW list development engaged patients after the lists were fully developed.8, 27 For example, the CW partnerships with Consumer Reports in the US or Altroconsumo in Italy facilitated dissemination with public-friendly language. Models for incorporating patient input during list development are lacking. Exceptions include the Canadian Rheumatology Association and Canadian Psychiatry CW lists that incorporated only one patient advocate into their working groups.28 Additionally, the 28 perinatal organizations responsible for the CW Newborn Medicine list incorporated 15 family representatives (from 1047 total individuals) who responded to a survey for suggestions.29.

Building upon the model from the Right Care Alliance’s Hospital Medicine Council Top 10 list development, we formally integrated patient advocates into the Delphi voting process.12, 30 Patient advocates were chosen rather than patients as they can represent a broader patient perspective being exposed to a variety of patient concerns. To overcome the known barriers in clinical knowledge that prevented full participation30, we utilized a patient-friendly script. This allowed a standardized method of communication with patient partners to allow meaningful understanding and input into the CW list. We also selected equal number of patient advocates to clinicians to provide equity in the decision-making process. This represents a shift in paradigm from paternalistic medicine and simply finding better ways to explain recommendations developed from clinicians to patients. Integrating patients’ views from inception to completion is critical for determining if these CW recommendations are truly “low-value,” as the patient is ultimately the end-consumer. Interestingly, three of the final recommendations were shared among patients and clinicians, indicating common priorities.

We note a key limitation and learning point, which was the degree of public engagement. Through the first questionnaire outreach, we elicited only 3 responses from patient/patient advocates. Even with assistance from Right Care Alliance and ABIM foundation, through their email listservs, websites, and social media, as well as our HVCC members’ efforts in reaching out to their contacts, we obtained only 48 total responses, albeit the highest number of patient engagement in CW list development. It should also be noted that one of the top recommendations of reducing nighttime awakenings was provided as an example on the patient-friendly survey for initial suggestions, which may have inadvertently promoted this recommendation. Harrison et al. outlined some barriers of patient partner engagement in research in their qualitative study, including researchers’ lack of training in appropriate engagement.31 Moreover, the main method of patient outreach in this study was through social media, which may not be representative of all patients cared for by hospitalists.

There were other limitations. We observed a small overall response to the 1st questionnaire. The main goal of this questionnaire was to reach a broad audience to promote brainstorming of ideas for CW. It was not intended as a survey in the classic research sense, where we would calculate and aim for a high response rate. In order for this to occur, we would need a denominator, which was not possible given that this questionnaire was posted on the ABIM-F website, sent over social media, and forwarded through many emails. Finally, this complete process took over 2 years to complete. The clinician and patient advocate voting surveys were administered approximately 1 year apart, and a greater 2 years from the initial questionnaire to the patient advocate voting survey. Research and media coverage on healthcare issues may have shifted during this time, including notably the US Health and Human Services (HHS) declaration of the opioid epidemic, and Hospital Consumer Assessment of Healthcare Providers and Systems (HCAHPS) survey change from pain management to pain communication questions. Chronological changes such as these may have skewed survey responses that otherwise may have been different had it been completed at the same time. This process may have been accelerated with financial support, incentives to complete the survey, and a broader outreach through partnering with other organizations. Additionally, a more extensive literature review utilizing the PRISMA guidelines was not conducted, which may have affected the strength of evidence section.

Utilizing focused categories of strength of evidence, patient harm, and relevance for voting results in recommendations that have large clinical impact and actionability. The clinical feasibility with a positive impact of several of these recommendations has already been demonstrated in several studies. 32–34 The breakdown of these categories, along with utilizing patient-friendly language allows patients to meaningfully contribute to these important recommendations. We hope that this will serve as a model for CW in the future.