Racial Disparities in Preventable Adverse Events Attributed to Poor Care Coordination Reported in a National Study of Older U.S. Adults

Abstract

Background:

Previous work found that Black patients experience worse care coordination than White patients.

Objective:

To determine if there are racial disparities in self-reported adverse events that could have been prevented with better communication.

Research Design:

We used data from a cross-sectional survey that was administered to participants in the Reasons for Geographic and Racial Differences in Stroke (REGARDS) study in 2017–2018.

Subjects:

REGARDS participants aged 65+ years of age who reported >1 ambulatory visits and >1 provider in the prior 12 months (thus at risk for gaps in care coordination).

Measures:

Our primary outcome was any repeat test, drug-drug interaction, or emergency department visit or hospitalization that respondents thought could have been prevented with better communication. We used Poisson models with robust standard error to determine if there were differences in preventable events by race.

Results:

Among 7,568 REGARDS respondents, the mean age was 77 years (SD 6.7), 55.4% were female, and 33.6% were Black. Black participants were significantly more likely to report any preventable adverse events compared to Whites (adjusted risk ratio [aRR] 1.64; 95% confidence interval [CI] 1.42, 1.89). Specifically, Blacks were more likely than Whites to report a repeat test (aRR 1.77; 95% CI 1.38, 2.29), a drug-drug interaction (aRR 1.76; 95% CI 1.46, 2.12), and an emergency department visit or hospitalization (aRR 1.45; 95% CI 1.01, 2.08).

Conclusions:

Black participants were significantly more likely to report a preventable adverse event attributable to poor care coordination than White participants, independent of demographic and clinical characteristics.

Introduction

Racial and ethnic minorities are known to experience worse care coordination than their White counterparts.¹ According to responses from Medicare beneficiaries who completed the Consumer Assessment of Healthcare Providers and Systems (CAHPS) Survey, compared to Whites, Blacks and Hispanics reported that: their providers were less likely to have reviewed their medical records at appointments, they had more difficulty getting test results, they received less help managing their care, and they were less likely to discuss medications with their providers.¹ Black and Hispanic patients are also known to report worse experiences with their hospital care than White patients.² Further, racial minorities are known to be at an overall increased risk for poor health outcomes including hospitalization and ED use.³ However, no studies have used patient self-reported data to describe and quantify racial differences in preventable outcomes.^4–6

Prior studies have found that highly fragmented care (that is, care spread across multiple providers without a dominant provider) puts patients at risk for poor outcomes including increased risk of hospitalization⁷, emergency department (ED) visits⁸, repeat testing,^9,10 and prescription drug interactions^11,12. To date, most of these studies have focused on claims-based measures of preventable adverse events.⁷ Although claims-based measures of adverse events are important, they do not incorporate the patient perspective on why the adverse event may have occurred. A recent study found that individuals who self-reported one or more gaps in care coordination were at a 55% increased risk of having a preventable adverse event.¹³ To our knowledge, this was the first study to draw a link between self-reported gaps in care and self-reported adverse preventable events. However, this study did not examine differences in preventable adverse events by race.

Using data from the national REasons for Geographic And Racial Differences in Stroke (REGARDS) study, a racially diverse, community-based, longitudinal cohort, we sought to quantify any racial disparities in self-reported preventable adverse events. We hypothesized that Black participants would self-report more preventable adverse events than White participants. Specifically, we used responses to a REGARDS survey on experiences with healthcare, which allowed respondents to report on certain types of adverse events that they thought could have been prevented with better communication among their providers. Identifying racial disparities in these self-reported outcomes is important, because it would suggest that some participants are especially vulnerable to poor communication among their providers.

Methods

REGARDS Cohort Study:

REGARDS is a national prospective cohort study designed to understand racial and geographic disparities in stroke mortality. Between 2003–2007, REGARDS recruited 30,239 community-dwelling English-speaking individuals ≥45 years of age and continues to follow these participants today.¹⁴ At enrollment, a 45-minute computer assisted telephone interview (CATI) collected demographic and medical history information from each participant. Approximately one month later, participants underwent an in-home visit, which included a physical exam, medication inventory, and lab tests. Approximately 10 years later, participants completed a follow-up examination, which included a survey and in-home visit including physical exam, medication inventory, and laboratory tests. All participants provided written informed consent to participate. This study was approved by the Institutional Review Boards at the University of Alabama at Birmingham and Weill Cornell Medical College.

Survey on care coordination and adverse events:

We used data collected from a one-time cross-sectional survey on experiences with healthcare that was supported by an ancillary study. This survey was administered to REGARDS participants aged 65+ years between August 2017 and November 2018.¹³ This survey was in addition to the surveys described above. The ancillary study survey development and implementation have been previously described.¹³ In brief, the survey included 22 items covering domains for care coordination (8 items), preventable adverse outcomes (8 items), and ambulatory care utilization (6 items).

Study Design:

Our study used baseline REGARDS data from 2003–2007 and 10-year follow-up data from 2013–2017 to operationalize the primary exposure (race) and various demographic and clinical covariates (described below). The current study’s outcome was assessed using a cross-sectional survey that was administered between 2017–2018.

Study Sample:

Consistent with prior studies, this study sample included REGARDS respondents to the healthcare experiences survey who reported having more than one ambulatory visit with more than one provider over the previous 12 months.¹³ These criteria were selected because individuals with only one ambulatory visit to only one provider in a 12-month period are not at risk for gaps in care coordination, as their care is not being coordinated across multiple visits and providers.

Key Independent Variable:

Self-reported race (Black or White) collected on the baseline REGARDS survey.

Primary Outcome:

Our primary outcome was a binary indicator of any preventable adverse event in the prior 12 months. To construct this variable, we considered responses to four separate questions from the ancillary study survey. Survey respondents were asked to reflect on the previous 12 months and report if they experienced: 1) a repeat a blood test, x-ray, or other test that they had had in the past two weeks, because their doctor did not have the results of the first test; 2) a problem with medications, because different doctors prescribed medications that did not go well together; 3) an ED visit that the respondent believed could have been prevented by better communication among their doctors or other healthcare providers; and 4) a hospital admission that the respondent believed could have been prevented by better communication among their doctors or other healthcare providers.¹³ Endorsement of any of the four questions was considered an adverse event. Due to a limited number of events for ED visits and hospitalizations in the sample, which was also stratified by race, these two events together (ED visit or hospitalization that could have been prevented by better communication) were considered together. In the current study, the term “provider” is used to refer to a personal doctor or other healthcare provider, such as nurse practitioner.

Covariates:

Guided by Andersen’s Behavioral Model of Health Services Use,¹⁵ potential confounders included socio-demographics and clinical covariates that fell into three buckets: pre-disposing factors, enabling factors, and evaluated need factors. Biological sex (predisposing factor) was captured on the REGARDS baseline survey. Enabling factors including low education (defined as <high school), residing in a rural region (rural urban commuting area [RUCA] codes 9 and 10), and the Southeast Stroke Belt region (defined as North Carolina, South Carolina, Georgia, Tennessee, Mississippi, Alabama, Louisiana and Arkansas; or non-stroke belt) were also obtained from the REGARDS baseline survey. Age (predisposing) and low income (defined as annual household income <$35,000/year) (enabling) were obtained from the REGARDS 10-year follow-up survey. Given that all participants were eligible for Medicare, we did not consider insurance status as a covariate. Clinical covariates that reflect evaluated need (from the REGARDS 10-year follow-up and from adjudicated events since baseline) included binary indicators of: hypertension (self-reported diagnosis with use of antihypertensive medication, systolic blood pressure ≥140 mm Hg, or diastolic blood pressure ≥90 mm Hg), hyperlipidemia (self-reported or use of lipid-lowering medication, total cholesterol ≥240 mg/dL, low-density lipoprotein ≥160, or high-density lipoprotein ≤40 mg/dL), diabetes (self-reported diagnosis with use of oral glucose-lowering medication or insulin, fasting glucose ≥126 mg/dL, or non-fasting glucose ≥200 mg/dL), myocardial infarction (MI) (self-reported, evidence on an electrocardiogram or having an adjudicated event since baseline), stroke (self-reported at baseline or having an adjudicated event since baseline), 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 an electrocardiogram at baseline or follow-up).¹⁶ We combined comorbid conditions (i.e., hypertension, hyperlipidemia, diabetes, MI, stroke, kidney disease, and atrial fibrillation) to create a comorbidity count variable. Finally, from the healthcare experiences survey, we used responses from the questions related to perceptions of care coordination to create a binary variable for 1 or more gaps care coordination (“any gap,” out of 7 possible gaps) vs. none.¹³

Statistical Analyses:

First, we described sample characteristics by race (Black vs. White) and tested differences using t-tests for continuous variables that were normally distributed, Wilcoxon rank-sum test for skewed continuous variables, and chi-square tests for categorical variables. Next, we calculated unadjusted rates of any preventable adverse events for Blacks and Whites, separately. Finally, as our primary analysis, we used multivariable modified Poisson models with robust error variance¹⁷ to estimate the association between race and risk of experiencing any preventable adverse outcome (binary outcome), adjusting for potential confounders that were found to be statistically significantly different between Blacks and Whites in Table 1. As a secondary analysis, we also estimated separate Poisson models for repeat tests, drug-drug interactions, and the combined ED visits/hospitalization outcome, separately. We calculated adjusted risk ratios (aRR) and 95% confidence intervals (95% CI) for all estimates. Rates of missingness were relatively low, with income (18% declined to report) being the most frequently missing variable. All other covariates had less than 10% missing. However, to reduce bias introduced with complete case analyses, we conducted both complete case analyses and models using multiple imputation by chained equation to handle missing covariates.¹⁸ Statistical analyses were conducted using SAS (version 9.4, Cary, NC) and R (version 3.4.1, Vienna, Austria). P-values below <0.05 were considered to be statistically significant.

Table 1:

Survey Cohort Characteristics by Race

	White	Black
N	5028 (66.4%)	2540 (33.6%)
Age, median [IQR]	63.0 [59.0, 68.0]	62.0 [58.0, 68.0]	<.0001
Female sex, N (%)	2537 (50.5)	1653 (65.1)	<.0001
Low income, N (%)	1259 (30.1)	1025 (48.7)	<.0001
Low education, N (%)	187 (3.7)	263 (10.4)	<.0001
Southeast residencea, N (%)	2771 (55.1)	1351 (53.2)	0.1128
Rural residence, N (%)	694 (15.4)	76 (3.2)	<.0001
Hypertension, N (%)	3151 (64.4)	2046 (83.3)	<.0001
Hyperlipidemia, N (%)	3363 (69.8)	1649 (68.1)	0.1417
Diabetes, N (%)	1115 (22.8)	1019 (41.6)	<.0001
Myocardial infraction, N (%)	788 (15.7)	342 (13.5)	0.0109
Stroke, N (%)	321 (6.4)	210 (8.3)	0.0025
Kidney disease, N (%)	1322 (26.3)	612 (24.1)	0.0393
Atrial fibrillation, N (%)	656 (14.4)	151 (6.9)	<.0001
Heart failure, N (%)	387 (7.6)	302 (12.0)	<.0001
Number of comorbidities, Median [IQR]	2 [1, 3]	2 [1, 3]	<.0001
Self-reported gaps in care coordinationb, N (%)	1890 (37.9)	994 (39.2)	0.2740

^*percent from the observed

^aREGARDS study oversampled residents from the stroke belt (Alabama, Arkansas, Louisiana, Mississippi, Tennessee, and the noncoastal regions in North Carolina, South Carolina, and Georgia) and the stroke buckle (the coastal regions within North Carolina, South Carolina, and Georgia).

^bGaps in care coordination were determined from 8 questions on the healthcare experiences survey. Six questions came from the Care Coordination Measure from the Consumer Assessment of Healthcare Providers and Systems (CAHPS).³⁷ One question came from the Consumers’ Views of Patient Safety and Quality Information report³⁸ and the other from the Patient Perceptions of Integrated Care survey³⁹

Results

There were 15,817 REGARDS participants alive and engaged in the longitudinal cohort study at the time of the 10-year follow-up of whom 11,138 (88%) participants aged 65+ years completed the care coordination survey. We further limited our analytic sample to the 7,568 individuals with more than 1 visit and more than 1 provider in the preceding 12 months, as these participants were considered to be at risk for gaps in care coordination. Characteristics of individuals with >1 visit and >1 doctor compared to characteristics of those with ≤1 visit or ≤1 doctor have been previously reported.¹³ Overall, these individuals were of similar age, more likely to be Black, have an annual household income <$35,000, and have less than a high school diploma.¹³ Past medical history between these patients and those in our study did not differ considerably.¹³

Sample Characteristics:

The mean age of our study sample was 77.0 years (SD 6.7 years), 55.4% were female, and 33.6% were Black (Table 1). Blacks were more likely to have an annual household income <$35,000 (48.7%. vs. 30.1%, p<0.0001) and have less than a high school education (10.4% vs. 3.7%, p<0.0001) compared to Whites. Blacks were also more likely to have hypertension (83.3% vs. 64.4%, p<0.0001) and diabetes (41.6% vs. 22.8, p<0.0001) compared to Whites. White participants were more likely to reside in a rural area (15.4% vs. 3.2%, p<0.0001) and have a history of atrial fibrillation (14.4% vs. 6.9%, p<0.0001) compared to Black participants. We did not observe a statistically significant difference in self-reported gaps in care coordination between Blacks (39.2%) and Whites (37.9%, p=0.27). There were no significant differences in self-reported ambulatory care visits between Blacks (median 5.0 [IQR 3.0–8.0]) and Whites (median 5.0 [IQR 3.0–8.0]).

Unadjusted associations between race and preventable adverse events:

Compared to White participants, Black participants were significantly more likely to report any preventable event (13.6% vs. 7.9%, p<0.0001) (Table 2). Specifically, Black participants were significantly more likely to report having a repeat test because their provider did not have access to the first test results (4.7% vs. 2.6%, p<0.0001) compared to Whites. Blacks were also more likely to report a drug-drug interaction than Whites (8.9% vs. 4.7%, p<0.0001). Blacks compared to Whites reported having more ED visits (1.9% vs. 1.2%, p=0.01) and hospitalizations (1.1% vs. 0.6%, p=0.01) that they felt could have been prevented with better communication between their providers.

Table 2:

Unadjusted Frequencies of Preventable Adverse Events Attributed to Poor Care Coordination by race

Preventable events	White	Black	Chi-square p-value
	N / N eligible (%)
Repeated a blood test, X-ray, or other test that the participant had had in the past 2 weeks because his or her doctor did not have the results of the first test (Repeated tests)	128/4,915 (2.6)	116/2,470 (4.7)	<.0001
Had a problem with his or her medications, because different doctors prescribed medications that did not go well together (Drug-drug interaction)	228/4,896 (4.7)	219/2,471 (8.9)	<.0001
Had an emergency department visit that could have been prevented by better communication across his or her different providers and other healthcare professionals (ED visits)	60/4,987 (1.2)	48/2,500 (1.9)	0.0141
Had a hospital admission that could have been prevented by better communication across his or her different providers and other healthcare professionals (Hospitalizations)	28/5,006 (0.6)	27/2,506 (1.1)	0.013
Either ED visit or hospitalization	75/5,021 (1.5)	59/2,530 (2.3)	0.009
Any of the above	397/5,028 (7.9)	346/2,540 (13.6)	<.0001

ED: emergency department

Adjusted associations between race and preventable adverse events:

The statistically significant association between Black race and increased risk of reporting any preventable adverse event persisted after adjustment for potential confounders (aRR 1.64; 95% CI 1.42–1.89) (Table 3). When considering specific adverse events, Black participants were more likely to report a repeat test (aRR 1.77; 95% CI 1.38–2.29), drug-drug interaction (aRR 1.76; 95% CI 1.46–2.12), and an ED visit or hospitalization (aRR 1.45; 95% CI 1.01–2.08) than White participants.

Table 3.

Adjusted Associations Between Race and Preventable Adverse Events Attributed to Poor Care Coordination

Preventable adverse event	aRRa,b (95% CI)	p-value	aRRa,b (95% CI)	p-value
	Complete case		Multiple Imputations
Repeated a blood test, X-ray, or other test that the participant had had in the past 2 weeks because his or her doctor did not have the results of the first test (Repeated tests)	1.75 (1.32, 2.32)	0.0001*	1.77 (1.37, 2.28)	<.0001*
Had a problem with his or her medications, because different doctors prescribed medications that did not go well together (Drug-drug interaction)	1.69 (1.37, 2.08)	<.0001*	1.74 (1.45, 2.10)	<.0001*
Had an emergency department visit that could have been prevented by better communication across his or her different providers and other healthcare professionals (ED visits)	1.43 (0.93, 2.19)	0.10	1.45 (0.97,2.17)	0.07
Had a hospital admission that could have been prevented by better communication across his or her different providers and other healthcare professionals (Hospitalizations)	1.29 (0.68, 2.44)	0.44	1.63 (0.91,2.90)	0.09
Either ED visit or hospitalization	1.39 (0.95, 2.06)	0.09	1.43 (1.00, 2.05)	0.05
Any of the above	1.57 (1.34, 1.84)	<.0001*	1.63 (1.41, 1.88)	<.0001*

^aEach fully adjusted model adjusts for age, sex, Southeastern region, comorbidities count (continuous variable), low education, low income.

^bPoisson regression with robust error variance¹⁵

^*Statistical significance, p<0.05

ED: emergency department

Discussion

In this report of a national survey of U.S. adults 65 years and older who had >1 visits to >1 provider in the past year, nearly 40% of Black and White participants reported experiencing at least one gap in care coordination in the prior 12-months. The current study was designed to assess patient perceptions of provider-provider communication. Compared to White participants, Black participants were 64% more likely to report an adverse event that they thought could have been prevented by better communication between their providers (i.e., repeat test, drug-drug interaction, ED visit or hospitalization). These results adjusted for a host of confounders including socio-demographics and medical conditions.

Numerous prior studies have documented that Black Medicare beneficiaries are at increased risk for preventable hospitalizations and ED visits.^{4–6,19–24} Further, racial/ethnic disparities in preventable hospitalizations have increased over the last 20 years.⁴ To our knowledge, our study is the first to examine racial disparities in self-reported preventable ED visits and hospitalizations. Patient-reported experiences in general have been validated against measures of clinical safety and effectiveness (i.e., mortality, length of hospital stay, and adherence to treatments)²⁵. As such, our work, which documents racial differences in these self-reported adverse events, makes a novel contribution to the literature.

Although racial disparities in claims-based preventable hospitalizations and ED visits have been well described, few studies have examined racial disparities in repeat testing or drug-drug interactions. A study of Medicare beneficiaries found that Blacks had higher rates of computed tomographic (CT) scans (inefficient use of CT scans is a negative quality measure according to the Centers for Medicare and Medicaid Services)²⁶ than Whites.²⁷ However, that study did not measure repeat testing. A recent REGARDS study found that Black males (compared to White males) were more likely to experience drug-drug interactions (as assessed by the Beers criteria²⁸) due to potentially inappropriate medication use.²⁹ We are unaware of any studies that have previously reported on racial disparities in patients’ awareness of repeat testing or drug-drug interactions.

The data from this study cannot fully elucidate why no racial difference in self-reported gaps in care coordination was observed, while racial differences in preventable adverse events were found. The data from this study cannot determine the accuracy of either self-reported measure. For example, it could be that White participants overreport gaps in care coordination, while Black participants underreport gaps in care coordination. If this is true, the rates of actual gaps in care coordination may be different while self-reported rates appear similar. We also cannot determine from this study the details of what respondents meant when they described an event as being preventable. That is, we do not know exactly what happened or why they think it might be preventable. Nevertheless, the possibility of harnessing self-report to improve quality is intriguing and deserves further study. Black adults appear to be especially vulnerable to preventable adverse events that occur as a result of poor communication among their providers. There are several possible patient- and provider-level explanations for why this may occur. First, racial/ethnic minorities have higher rates of low health literacy³⁰, and low health literacy has been associated with inappropriate use of health services, poor health outcomes, and higher costs.³⁰ If communication gaps between providers occur and if a patient is aware of those gaps, a patient with lower levels of health literacy may be less prepared to advocate for themselves or ask questions of their providers. For example, consider a patient who receives two prescription medications from two different providers that should not be taken together. A patient with high health literacy may think to ask their provider about possible drug-drug interactions. On the other hand, a patient with low health literacy may not think to question the conflicting prescriptions and thus may experience an adverse event due to the drug-drug interaction. This is one hypothetical example illustrating the role that low healthy literacy may play in terms of increasing a patient’s risk of experiencing an adverse event that could have been prevented by better communication across their healthcare provider. Indeed, studies have shown that racial/ethnic minorities are less likely to demand high performance from their providers.³¹ Another possible explanation is that racial/ethnic minorities also have worse patient-provider communication than White patients, which can exacerbate the harms of poor communication across providers.³² For example, Black and Hispanic patients receive less information and less support from their providers during medical encounters than White patients, which may worsen the harms associated with poor communication across providers.³³ Finally, minorities face systemic racism (e.g., implicit physician bias³¹ and discomfort navigating different cultural norms²⁹) interacting with the healthcare system, which has been shown to lead to worse quality of care for minority patients³¹ and can possibly worsen the effects of poor communication among doctors.³⁴

Provider-level explanations include that Black patients are more likely to be treated by providers with less clinical training and limited access to healthcare resources than White patients.³⁵ A national study found that primary care doctors who cared for Black patients (vs. White patients) were less likely to endorse that they could provide high-quality care for all of their patients.³⁵ Another more recent study found that providers who treat a higher proportion of minority patients have lower performance scores compared to providers who treat a lower proportion of minority patients.³⁶ If Black patients receive care from lower quality providers, this may put them at increased risk for poor communication among these doctors and thus at greater risk for preventable adverse events related to that poor communication. In addition, providers who care for Black patients have also reported barriers to referring their patients to high-quality subspecialists and diagnostic imaging.³⁵ This may lead patients to seek care outside of their provider’s informal network of known colleagues, thus making communication among doctors even more difficult. In general, communication between generalists and specialists is known to be suboptimal,^37,38 and this problem may be magnified for minority patients. Our findings underscore the importance of ensuring that strategies to improve communication are implemented among providers who treat Black patients, as these patients experience a greater burden of preventable adverse events from poor communication.

Limitations:

Preventable adverse events were self-reported, and whether respondents overreport or underreport these events is not yet fully understood. The study may be limited due to the 12-month recall period, which was chosen because the four preventable events are relatively rare; a longer recall period may have yielded higher frequencies of these events.^13,39 The study was observational, and analyses were limited to tests of associations that may not be causal. We were unable to include individuals younger than 65 years old, limiting generalizability to those younger ages.

In this national sample of older adults, Black participants were at significantly increased risk for reporting adverse events that they thought could have been prevented with better communication among their providers, even after adjustment for socio-demographics and co-morbidities. Future studies should seek to understand the mechanisms by which better communication among providers may prevent adverse events. A deeper understanding of the ways that gaps in communication among providers lead to preventable events among minority patients may allow us to leverage insights to improve care and avert preventable outcomes, thus reducing racial disparities.