INTRODUCTION

Failures of care coordination in the outpatient setting are a major problem in American healthcare.1 Previous interventions to improve care coordination have had limited success.2,3,4 These interventions have tried to improve care coordination overall,2,3,4 which is appropriate, but they have not tried a more targeted approach of addressing gaps in care coordination as they occur. Other studies have shown that patients can report gaps in care coordination,5,6,7 but it has not been clear whether the gaps in care coordination that patients report indicate inconvenience and inefficiency, or if they also indicate clinically significant problems that reflect the potential for harm.

We sought to determine the frequency of patient-reported gaps in care coordination, the frequency of patient-reported preventable adverse outcomes, and any association between gaps in care coordination and preventable adverse outcomes. If patient-reported gaps in care coordination are associated with preventable adverse outcomes, it would suggest that future interventions could be designed to leverage patients’ observations to identify and address gaps in care coordination as they occur.

METHODS

Overview

We conducted a cross-sectional study, administering a survey between August 2017 and November 2018 to participants in the nationwide Reasons for Geographic and Racial Differences in Stroke (REGARDS) study who were ≥ 65 years old at the time of the survey. The study protocol was approved by the Institutional Review Boards of Weill Cornell Medicine and the University of Alabama at Birmingham.

Setting

Between 2003 and 2007, the REGARDS study enrolled 30,239 community-dwelling African-American and white adults aged ≥ 45 years.8, 9African-Americans and individuals living in the Southeastern US were oversampled, because the study was designed to elucidate reasons for racial and geographic differences in stroke mortality.9 All participants provided written informed consent. Baseline data collection included computer-assisted telephone interviews (CATI) about health status and medical history, as well as in-home visits by trained health professionals that included physical examinations, blood and urine collection, electrocardiograms (ECGs), and a medication inventory by pill bottle review. REGARDS participants, or their proxies if they cannot be contacted, are contacted by telephone every 6 months to detect potential study endpoints. A second in-home visit was conducted approximately 10 years after the baseline visit.

Survey Instrument

We developed a survey module on experiences with healthcare (Appendix 1), which consisted of 22 questions, many of which have been previously validated, as described below.

Perceptions of Care Coordination

Eight questions (questions 3, 4, 5, 6, 7, 8, 13, and 14 in Appendix 1) related to perceptions of care coordination, and these were combined to generate the main exposure variable in our study, as described in the “Statistical Analysis” section. These eight questions included the six questions that comprise the Care Coordination Measure from the Consumer Assessment of Healthcare Providers and Systems (CAHPS).10 These CAHPS questions ask the respondent to reflect on their experience with healthcare over the past 6 months and rate the frequency of care coordination processes (such as “getting the help you needed from your personal doctor’s office to manage your care among different providers and services”) on a 4-point scale (Never, Sometimes, Usually, or Always).10 The other two care coordination questions asked the respondent to reflect on their experience with healthcare in general and report: whether they think the doctors they see communicate with each other about their care (Yes, No, or I Don’t Know)11 and how they would describe the coordination among their providers (Excellent, Very Good, Good, Fair, or Poor).5

Preventable Adverse Outcomes

Another 8 questions (questions 15–22 in Appendix 1) related to respondents’ perceptions of four outcomes that could have occurred over the past 12 months: (1) a test that was repeated because the result of the first test was not available5; (2) a problem with the respondent’s medication, because “different doctors prescribed medications that did not go well together;” (3) an emergency department (ED) visit that the respondent believed could have been prevented by better communication among doctors; and (4) a hospital admission that the respondent believed could have been prevented by better communication among doctors. We considered these as outcome measures in our study, as described in the “Statistical Analysis” section.

Ambulatory Utilization

The remaining 6 questions (questions 1, 2, 9, 10, 11, 12) related to ambulatory utilization, covering whether the participant has a personal doctor that he or she usually sees,12 how many ambulatory visits the participant had in the past 12 months,13 how many different doctors a participant saw for his or her ambulatory visits, and the perceived appropriateness of the number of visits. We used these questions to narrow the study sample to those at risk for gaps in care coordination and for descriptive statistics, as described in the “Statistical Analysis” section.

The survey instrument was reviewed and pilot tested for clarity by the University of Alabama at Birmingham (UAB) Survey Research Unit.14

Data Collection

Trained staff at the UAB Survey Research Unit administered the survey instrument to REGARDS study participants during their routine follow-up CATI between August 2017 and November 2018. This survey instrument on healthcare was administered to REGARDS study participants ≥ 65 years old, as a proxy for Medicare eligibility, in anticipation of a future study that would link survey responses to Medicare claims. The REGARDS study does not offer monetary incentives for participation in follow-up calls.

Data from the REGARDS Study

We used the following self-reported variables collected by the REGARDS study at baseline: gender, race, educational attainment, and history of stroke. We used baseline variables for region of the USA and type of census tract (rural, suburban, or urban) that were derived by the REGARDS study from the participants’ addresses. We used the following self-reported variables collected by REGARDS at the second in-home visit: age and annual household income. We defined clinical covariates using data from the second in-home visit: hypertension (self-reported, self-reported use of antihypertensive medication, systolic blood pressure ≥ 140 mmHg, or diastolic blood pressure ≥ 90 mmHg), hyperlipidemia (self-reported, use of lipid-lowering medication, total cholesterol ≥ 240 mg/dL, low-density lipoprotein ≥ 160, or high-density lipoprotein ≤ 40 mg/dL), diabetes (self-reported, use of oral glucose-lowering medication, use of insulin, fasting glucose ≥ 126 mg/dL, or non-fasting glucose ≥ 200 mg/dL), myocardial infarction (self-reported or evidence on the electrocardiogram), kidney disease (self-reported kidney failure, or estimated glomerular filtration rate < 60 mL/min/1.73 m2), and atrial fibrillation (self-reported, or evidence on the electrocardiogram). We supplemented this information with REGARDS’ adjudicated events for myocardial infarction and stroke.15, 16

Statistical Analysis

We restricted the analysis to participants who reported having ≥ 2 ambulatory visits and ≥ 2 ambulatory providers in the past 12 months, as those with ≤ 1 visit or ≤ 1 provider in the past 12 months were not considered eligible for gaps in care coordination. We used descriptive statistics to characterize the sample.

For each of the 8 survey questions on perceptions of care coordination, we dichotomized the response scales to identify problems with care coordination, which we refer to as “gaps” in care coordination. As per CAHPS scoring rules, we combined the responses to two similar questions that asked about how often and how quickly a respondent received test results.10 This yielded 7 distinct questions on perceptions of care coordination. We calculated the frequency of participants who reported a gap in each of these 7 areas and the frequency of reporting any gap. Any gap in care coordination was the primary exposure.

We calculated the frequency of participants who reported each of the 4 preventable outcomes and the frequency of participants who had any of the 4 outcomes.

Using logistic regression, we determined the bivariate association between having a gap in care coordination and any of the outcomes. We then adjusted for participants’ demographic and clinical characteristics, using a dummy variable for each covariate. We conducted a complete case analysis and then an analysis using multiple imputation by chained equation to handle missing covariates.17 The most frequently missing covariate was income, which was missing for 18% of the sample. If patterns of missingness were not random, the complete case analysis might be biased, so we considered the model with multiple imputation to be the final model.

To determine if the association between a gap in care coordination and any preventable outcome varied with the numbers of visits and providers a participant saw in the previous year, we conducted a sensitivity analysis comparing those ≥ 75th percentile for visits (≥ 8 visits) and ≥ 75th percentile for providers (≥ 4 providers) to the rest of the sample. We chose this approach to approximate the approach used in other studies that have identified “highly fragmented care” based on the top quartile or quintile.18

To determine the representativeness of participants included in the analysis, we used CAHPS scoring rules to aggregate the responses to the 6-item Care Coordination Measure within and then across participants.10 We then compared these data to national benchmarks for this measure.

We conducted a supplemental analysis using Poisson regression to determine if any participant characteristics were associated with having any gap in care coordination.

Analyses were conducted with SAS (version 9.4, Cary, NC) and R (version 3.4.1, Vienna, Austria). We considered p < 0.05 to be statistically significant.

RESULTS

Sample

Of the 15,817 REGARDS participants active in the cohort at the time of the survey, 13,232 (84%) completed their routine follow-up call (Fig. 1). Of those participants, 12,600 (95%) were willing to participate in the survey module on experiences with healthcare. Of those, 1462 were not eligible due to being < 65 years of age. The remaining 11,138 participants completed the survey module. The sample for this analysis consisted of those who reported having ≥ 2 visits and ≥ 2 providers in the previous year (N = 7568). Appendix 2 shows how the characteristics of individuals with ≥ 2 visits and ≥ 2 providers compared to the characteristics of those with ≤ 1 visit or ≤ 1 provider.

Figure 1
figure 1

Derivation of the study sample.

Full size image

Sample Characteristics

Overall, the average age of the sample at the time of the survey was 77.0 years (SD 6.7 years). More than half of the sample (55%) was female, and 34% were African-American. Nearly one-third (30%) had an annual household income < $35,000. Approximately one-fourth (27%) had a high school diploma or less. Approximately half lived in the Southeastern US (55%), and 10% lived in a rural area. The most common chronic medical condition was hypertension (69%) (Table 1).

Table 1 Characteristics of Participants Who Completed the Care Coordination Survey and Reported Having ≥ 2 Ambulatory Visits and ≥ 2 Ambulatory Providers in the Previous Year*
Full size table

Overall, 98% of participants reported having a personal doctor that they usually see for a checkup or sick visit, and 85% reported having seen their personal doctor in the previous 6 months. The median number of ambulatory visits reported in the prior year was 5 (25th percentile 3; 75th percentile 8). Those visits were distributed across a median of 3 different providers (25th percentile 2; 75th percentile 4). Overall, 92% of participants reported that that they thought the number of visits they had was appropriate. Half (51%) reported that the number of visits they had was the result of shared decision-making with their doctors, whereas 14% reported that they asked for most of their visits and 35% reported that their doctors asked for most of their visits.

Self-Reported Gaps in Care Coordination

Reflecting on care over the past 6 months, the most frequently reported gap in care coordination (reported by 14.7% of participants) was that doctors sometimes or never talked about all of the prescription medications they were taking (Table 2). Other gaps in care coordination processes were reported by 2.9% to 10.9% of participants. Reflecting on care in general, 11.8% reported that they think that their doctors do not communicate with each other, and 8.3% rated care coordination among all of their health professionals as “fair” or “poor.” Overall, 2884 participants (38.1%) reported experiencing ≥ 1 gap in care coordination (Table 2).

Table 2 Self-Reported Frequency of Gaps in Care Coordination (N = 7568)
Full size table

The average raw composite score for the items from the CAHPS Care Coordination measure in this study was 3.7 (standardized score 87.5). The average raw composite score in the national benchmark for Medicare fee-for-service beneficiaries is 3.6 (standardized score 86.0).

None of the participant characteristics was independently associated with having any gap in care coordination (Appendix 3).

Outcomes

Overall, 244 (3.2%) of participants reported having a repeat test because the result of the first test was not available, and 447 (5.9%) reported having a problem with medications, because different doctors prescribed medications that did not go well together. Of those who had an ED visit (N = 1824), 108 reported that the ED could have been prevented with better communication. Of those who had a hospital admission (N = 1477), 55 reported that their admission could have been prevented with better communication. Overall, 743 participants (9.8%) reported ≥ 1 preventable outcome (Table 3).

Table 3 Self-Reported Outcomes, Among Those with at least 2 Different Doctors in the Previous 12 Months (N = 7568)
Full size table

Association Between Gaps in Care Coordination and Outcomes

Among those who reported a gap in care coordination (N = 2884), 356 (12.3%) reported having any preventable outcome, compared with 384 of the 4645 (8.3%) who did not report a gap in care coordination (unadjusted p < 0.0001).

Having a gap in care coordination was associated with an increase in the odds of each preventable outcome, as well as an increase in the odds of any preventable outcome (Table 4). These associations were present after adjustment for demographic and clinical covariates, using multiple imputation to handle missing covariates. In the final model, having a gap in care coordination was associated with an increase in the odds of any preventable outcome (odds ratio 1.55; 95% confidence interval 1.33, 1.81).

Table 4 Associations Between Any Self-Reported Gap in Care Coordination (vs. None) and Four Outcomes That Participants Reported Could Have Been Prevented
Full size table

This association was observed both in the subset of participants with ≥ 8 visits and ≥ 4 providers and in the rest of the sample, although the magnitude of the effect size was larger in the subset with ≥ 8 visits and ≥ 4 providers (Appendix 4).

DISCUSSION

In this national survey of adults 65 years and older, self-reported gaps in care coordination were common, affecting more than one-third (38.1%) of those who had ≥ 2 ambulatory visits and ≥ 2 ambulatory providers in the past year. Nearly 1 in 10 participants (9.8%) reported having experienced an outcome that they thought could have been prevented by better communication across their healthcare providers. Participants who reported a gap in care coordination had a 55% greater adjusted odds of experiencing any preventable outcome than those who did not experience a gap in care coordination.

In general, previous studies have considered patient-reported gaps in care coordination as one aspect of patient satisfaction and have used average results across patients as a population-level measure of quality.7, 19 While these approaches are appropriate, the current study suggests that patients’ reports can be used not only as a measure of satisfaction and quality but also as a measure of patient safety. Patient safety experts have noted that the patient perspective has often been discounted in favor of the provider perspective, even though patients have a distinct vantage point.20 Indeed, patients may be the first people able to detect breakdowns in communication among providers.20 This suggests that the results revealed by our study could inform the design of novel patient-centered interventions to improve care coordination.

Previous interventions designed to improve care coordination, such as the Patient-Centered Medical Home21, 22 and electronic health information exchange,23, 24 have generally focused on providers. Electronic patient health records (PHRs) have been increasing in frequency and do facilitate patient-provider communication, but they are mostly designed to allow patients to update their medication lists, check their laboratory results, or send messages to their physicians.25 Although all of these interventions are appropriate, they do not systematically elicit and act on patients’ observations about gaps in care coordination.

Interventions to elicit and act on patients’ observations about gaps in care coordination could take several forms. They could include educational interventions for patients to inform them about the importance of identifying and reporting gaps in care coordination. They could also include the development of electronic tools to facilitate reporting, like the tool developed for OpenNotes that specifically asks patients to report inaccuracies they find in their medical records.26 An evaluation of the OpenNotes patient reporting tool found that 64% of the safety concerns reported by patients were validated upon clinician review and that 57% of confirmed problems resulted in a change in patient care.26 Interestingly, the rate of patient-reported safety concerns in OpenNotes exceeded the rate of provider-reported safety concerns by several fold.26 Although the OpenNotes feedback tool was not designed to elicit gaps in care coordination across providers, it provides a template for what future patient-reported tools for gaps in care coordination could look like.

This study has notable strengths, including its national reach, diverse sample, and use of many previously validated survey questions. It also has some limitations. First, because this is a cross-sectional study, we cannot infer causality, and our results may have been affected by unmeasured confounding. Second, it is possible that patients do not accurately recall their experiences with healthcare; patients may perceive gaps in care coordination when in fact coordination occurred, or they may report preventable outcomes that either did not occur or could not have been prevented. However, as the OpenNotes study suggests,26 many problems that patients identify are likely to be valid and should be taken seriously. Third, of the survey questions related to gaps in care coordination, five had a 6-month recall period.10 Although this 6-month recall period is recommended by the Centers for Medicare and Medicaid Services,27, 28 it may underestimate the frequency of gaps in care coordination compared with longer recall periods. Fourth, the recall period for outcomes was 12 months, because these outcomes are relatively rare.29 Because the recall period for outcomes is longer than the recall period for the exposure, we cannot be certain that the exposure preceded the outcome.

In summary, in this national survey of adults 65 years and older, 38.1% reported experiencing a gap in care coordination and 9.8% reported experiencing a repeat test, drug-drug interaction, ED visit, or hospital admission that they thought could have prevented with better communication among their providers. Experiencing a gap in care coordination was associated with an increased odds of experiencing an event that could have been prevented by better communication among providers. Future interventions should be designed to incorporate what patients observe, as patients may be able to detect and help prevent or ameliorate the consequences of failures of care coordination.