Differences in patient perceptions of integrated care among black, hispanic, and white Medicare beneficiaries

Abstract

Objective

This study sought to identify potential disparities among racial/ethnic groups in patient perceptions of integrated care (PPIC) and to explore how methodological differences may influence measured disparities.

Data Source

Data from Medicare beneficiaries who completed the 2015 Medicare Current Beneficiary Survey (MCBS) and were enrolled in Part A benefits for an entire year.

Study Design

We used 4‐point measures of eight dimensions of PPIC and assessed differences in dimensions among racial/ethnic groups. To estimate differences, we applied a “rank and replace” method using multiple regression models in three steps, balancing differences in health status among racial groups and adjusting for differences in socioeconomic status. We reran all analyses with additional SES controls and using standard multiple variable regression.

Data Collection/Extraction Methods

Not applicable.

Principal Findings

We found several significant differences in perceived integrated care between Black versus White (three of eight measures) and Hispanic versus White (one of eight) Medicare beneficiaries. On average, Black beneficiaries perceived more integrated support for self‐care than did White beneficiaries (mean difference = 0.14, SE = 0.06, P =.02). Black beneficiaries perceived more integrated specialists’ knowledge of past medical history than did White beneficiaries (mean difference = 0.12, SE = 0.06, P =.01). Black and Hispanic beneficiaries also each reported, on average, 0.18 more integrated medication and home health management than did White beneficiaries (P <.01 and P <.01). These findings were robust to sensitivity analyses and model specifications.

Conclusions

There exist some aspects of care for which Black and Hispanic beneficiaries may perceive greater integrated care than non‐Hispanic White beneficiaries. Further studies should test theories explaining why racial/ethnic groups perceive differences in integrated care.

What is Known on This Topic

• Prior research has documented racial disparities in health care utilization, care performance, and health outcomes.

• Despite this established body of literature, there is a limited understanding of how well integrated minority patients perceive their care to be.

What This Study Adds

• We identified aspects of care that Black and Hispanic Medicare beneficiaries perceived as significantly more integrated compared to care perceived by non‐Hispanic White Medicare beneficiaries.

• Our research provides a foundation for further research to test theories explaining why these differences exist and what upstream factors influence perceptions of care.

1. INTRODUCTION

Racial and ethnic disparities pervade all aspects of health care, including inequalities in patient access and utilization, system delivery, and individual‐ and population‐level outcomes. ¹ Previous literature has extensively documented disparities in clinical care performance, health outcomes, and health utilization for minority patients. ² , ³ , ⁴ , ⁵ , ⁶ , ⁷ , ⁸ Despite this substantial body of research, there still remains a limited understanding of how minority patients themselves perceive their care. ⁹ , ¹⁰ Understanding differences in patient perceptions of care is critical for addressing persistent, well‐documented care disparities for minority populations and improving overall care delivery.

Understanding whether patients perceive their care as integrated may be especially useful for understanding and improving integrated care. Integrated care, defined as “care that is coordinated across professionals, facilities, and support systems; continuous over time and between visits; tailored to the patients’ needs and preferences; and based on shared responsibility between patient and caregivers for optimizing health,” ¹¹ has a substantial influence on patient outcomes and overall patient experience. ¹² , ¹³ , ¹⁴ Integrated care is ideally measured by patients and families who are positioned to observe the totality of care through the course of managing an illness. ¹⁵ , ¹⁶ It is especially needed for the management of patients with complex, chronic conditions, which often requires services from several specialties and nonmedical disciplines. ¹⁷ This need for integrated care is increasing as the population of Americans with complex care needs continues to rise, among both young and old. ¹⁸ Already, nearly one third of Medicare patients report a functional impairment, while more than one in five have more than five chronic conditions. ¹⁹ Integrated care is so central to patients’ overall health that in 2015, the Centers for Medicare & Medicaid Services introduced a module on patient perceptions of integrated care into its annual national survey of Medicare beneficiaries, the Medicare Current Beneficiary Survey (MCBS). ²⁰

Recent studies examining patient perceptions of care across racial and ethnic groups has produced mixed results; in some analyses of patient‐reported experience, minority patients rate their care as better than do White patients, while in other measures they rate their care as worse. ²¹ , ²² , ²³ In this paper, we define racial disparities following the National Academy of Medicine (NAM) as disparity in service provision due to health system‐level operations, the existing regulatory climate, as well as individual‐level provider discrimination. ¹ To conceptualize how racial disparities may influence care delivery and perceived integrated care in conjunction with other variables, we adapted a version of Andersen's Access to Care framework. ²⁴ , ²⁵ In Figure 1 below, variables with asterisks highlight areas where potential inequalities align with the NAM definition of racial disparities. Andersen's framework suggests that environmental factors impact population characteristics, which impact health behavior and, in turn, health outcomes. ²⁶ Our conceptual model focuses on individuals and regards integrated care delivery and perceived integrated care as the health outcomes of interest. The model suggests that environmental factors and individual characteristics impact integrated care delivery and that the delivery of integrated care influences how individuals perceive integrated care. For example, an individual characteristic like high patient activation may lead to more patient involvement with care plans, influencing care delivery and perceived integrated care. Our model also acknowledges that expectations about integrated care may intervene in the relationship between integrated care delivery and patient reports of integrated care. ²³ , ²⁷ The model suggests that these expectations about integrated care develop based on environmental and individual characteristics as well as prior experience of integrated or nonintegrated care. For example, undisclosed mistrust in health care providers could influence patient expectations. While these expectations may not change care, they could positively or negatively influence perceptions of care, depending on whether care provided is better or worse than expected.

FIGURE 1.

Adapted Andersen's Access to Care Framework. Asterisks indicate variables that align with elements from the NAM definition of racial disparities

Prior methodological approaches used for understanding racial disparities have primarily reported bivariate and multiple variable analyses that do not fully adjust for differences in population characteristics that frequently cluster with race—including predisposing factors such as health status and treatment preferences. ²⁸ , ²⁹ , ³⁰ Critics have noted that while socioeconomic variables influence disparities in care, differences in population characteristics such as health status and differences in treatment preferences should not be included in disparity adjustments. ³¹ , ³² , ³³

Investigators, most notably, Cook et al, ³⁴ have developed methods that offer a rigorous approach for addressing the overlap between racial/ethnic disparities from socioeconomic factors and health status. Their approach, called “rank and replace,” allows users to pose the hypothetical question, “how much X would Black or Hispanic beneficiaries rate, or receive, if they had the same health status as non‐Hispanic White beneficiaries, but retained their own racial and socioeconomic characteristics?”

In this paper, we explore racial and ethnic disparities in patient‐perceived integrated care using this method. We use data from the Patient Perceptions of Integrated Care (PPIC) module of the 2015 Medicare Current Beneficiary Survey to examine relationships between beneficiary race and ethnicity and their perceptions of integrated care. We hypothesized that Black and Hispanic beneficiaries would perceive integrated care as lower than would non‐Hispanic White beneficiaries, if Black and Hispanic beneficiaries had similar population characteristics as White beneficiaries. We employed a modified version of the rank and replace methodology introduced by Cook and colleagues and also tested alternative model specifications, to increase confidence in our findings and derive new insights about the nuances of racial disparities related to care delivery.

2. METHODS

2.1. Survey instrument

The MCBS is administered annually by the Centers for Medicare and Medicaid Services (CMS) to a nationally representative population of Medicare beneficiaries. The MCBS documents beneficiaries’ health care use, costs, and care satisfaction. ³⁵ In 2015, the MCBS expanded the survey to include a portion titled Patient Perceptions of Integrated Care (PPIC) for beneficiaries living in community living arrangements. In this section, respondents were asked a series of questions related to how well they felt their care was integrated across providers, facilities, time, and with patient/family needs and preferences. ³⁶ For our analysis, we focused on data collected from this section of the MCBS survey.

2.2. Study sample

For the 2015 MCBS, CMS interviewed a total of 14 068 beneficiaries living in both community and facility living arrangements. For our study, we included all respondents that had completed the PPIC section of the community survey and had continuous eligibility for and receipt of Medicare Part A benefits for an entire year. Beneficiaries were enrolled in traditional fee‐for‐service Medicare and Medicare Advantage, and some also received Part B and/or D benefits. We excluded respondents that had missing survey weight data. Our total analytical sample included 11 978 respondents.

2.3. Defining domains of integrated care

In accordance with our definition of integrated patient care, we measured multiple dimensions of patient‐perceived integrated care using questions from the PPIC survey. ³⁷ These included six psychometrically derived factors and two indices. The factors measure staff knowledge about patient's medical history, provider support for patient's self‐directed care, test result communication, provider knowledge of the patient, provider support for medication and home health management, and specialist knowledge about the patient's medical history. We used two additional indices to encompass integrated care between care episodes: provider support and knowledge following a patient's hospitalization and provider support outside of office visits. All domains used a four‐point scale. Not all domains applied to all survey respondents; for example, not all survey respondents had been hospitalized or had interacted with a specialist. The effective sample size for each domain varied accordingly.

2.4. Defining racial/ethnic inequalities in health care

To understand differences in patient‐perceived integrated care across racial/ethnic groups, we adapted the definition of racial/ethnic disparity in health care from the seminal Institute of Medicine (now NAM) report, Unequal Treatment: Confronting Racial and Ethnic Disparities in Care. ¹ In this report, racial/ethnic disparities in care are defined as differences in service provision due to health system‐level operations and the existing regulatory climate, as well as individual‐level provider discrimination. We did not consider differences in care due to clinical need, appropriateness of interventions, and patient preferences as part of racial disparity. Accordingly, we attempted to distinguish disparities in care that derived from structural and provider discrimination using analogous methods.

To describe our sample population and adjust for variations in “Individual Characteristics” listed in our framework, we used measures from the MCBS for age, gender, and health status. We also included marital status, whether or not a proxy was used to complete the MCBS interview, and 40 chronic conditions, including hypertension, chronic lung disease, and depression. A full list of health covariates used for our primary analysis are included in Appendix S1. We measured “Enabling Factors” using income range according to federal poverty level, patient education level, and Medicare insurance type. We measured variations in “Environment, Community Factors” using census region and city type.

To measure individual response tendency toward more positive or negative evaluations, we used six items from the PPIC survey that comprise the Revised Life Orientation Test (LOT‐R), considered an indicator of individual respondent optimism. ³⁸ We included patients’ level of optimism as a control variable because past literature has demonstrated a positive relationship between individuals’ intrinsic optimism with patient expectations and perceived integrated care. ³⁹ , ⁴⁰ , ⁴¹ However, we distinguish an individuals’ intrinsic levels of optimism from their expectations about integrated care. As our conceptual model suggests, we view the latter as being shaped by a multitude of intrinsic and extrinsic factors.

We measured patient activation using a categorical measure (low, medium, high) based on a prior measure developed from MCBS data. ⁴² Patient activation measures patients’ knowledge, engagement and confidence in decision making about the care they receive. ⁴³ , ⁴⁴ Past research suggests that higher patient activation is related to better health outcomes. ⁴⁵ , ⁴⁶

2.5. Analysis

We first computed descriptive analyses and evaluated unadjusted PPIC scores using Stata 15. In our unadjusted analyses, we used Adjusted Wald tests to compare differences between mean PPIC scores and utilized the “svy” command in Stata to account for the complex MCBS survey sampling design. Next, we modified the rank‐based replacement method developed by Cook et al, ³² to adjust and measure disparities using three steps. In the first step, we ran multiple regression models, adjusting for health status, SES, and optimism. In each model, we treated one of the eight PPIC domains in turn as the main dependent variable. For each individual, we developed a single coefficient for health status by adding up the coefficients of each health variable and characteristic variables such as age and sex that are predictive of health status. In the second step, we ranked individuals within each self‐identified racial group by the magnitude of their health status coefficient. We then replaced the health status coefficient of racial minorities with the health status coefficient of non‐Hispanic White (hereafter described as “White”) counterparts with equivalent rankings. This step served to “equalize” the health status of individuals across different racial groups.

As our third step, we calculated adjusted PPIC scores by adding the rank‐and‐replaced health status coefficients and the SES and optimism coefficients from the original multiple regression model we ran in our first step. We then subtracted the adjusted PPIC scores of minority patients from adjusted scores of White patients, to develop estimates of disparities. By applying this three‐step method, we attempted to distinguish differences in care due to health care need, clinical appropriateness, and existing health system operations, from those considered racial disparities.

We completed several additional analyses using alternative specifications by introducing additional controls variables in sequential models. First, we adjusted for health care need based on five clinically meaningful subgroups of health care need, defined by Joynt et al ⁴⁷ As a second alternative specification, we added patient activation as a control variable in addition to health care need and other variables. In our third alternative analysis, we used multivariate regression instead of the rank and replace method to conduct our analysis. In all specifications, we adjusted significance levels to account for multiple comparisons using a Benjamini‐Yekutieli adjustment. ⁴⁸

3. RESULTS

3.1. Sample characteristics

Select characteristics of the overall sample are stratified by race and ethnicity and included in Table 1. Within our sample, 9794 individuals identified as White (81.77%), 1309 individuals as Black (10.93%), and 404 individuals as Hispanic (3.37%). We grouped 471 individuals (3.93%) as an “Other” category. Within this group, 131 individuals identified as Asian, Native Hawaiian or Pacific Islander, 78 individuals as North American Native, and 262 as other or unknown. Overall, 54.84% of our sample was male, 53.00% were married, and 57.59% were living above the federal poverty line. The vast majority of individuals lived in urban areas (92.73%) and did not require a proxy to complete the survey (92.4%). The mean age of individuals was 71.10, and the mean optimism score was 17.30 out of a maximum score of 24. Across racial/ethnic groups, the mean optimism score was approximately one‐point lower for Black and Hispanic individuals than White individuals (White: 17.51, Black: 16.66, Hispanic: 16.40). Other differences among groups included percentage of interviews completed by proxies and the prevalence of some chronic health care conditions. Among White beneficiaries, 6.45% used a proxy to complete their interview. In contrast, 11.00% of Black beneficiaries and 14.64% of Hispanic beneficiaries used a proxy. There were also several differences by race/ethnicity in the prevalence of chronic conditions: More Black beneficiaries had high blood pressure (79.90%) and a history of stroke (13.92%) than did White and Hispanic beneficiaries (High blood pressure: 64.50% and 67.70%; Stroke: 10.19% and 10.76%, respectively). More Hispanic beneficiaries reported depression (48.99%) than did Black (27.93%) or White (28.76%) beneficiaries. Hispanic and White beneficiaries had more osteoporosis (25.69% and 19.43%, respectively) and skin cancer (10.15% and 22.97%, respectively), compared with Black beneficiaries (osteoporosis: 11.82% skin cancer: 0.66%).

TABLE 1.

Sample characteristics, overall, and by racial/ethnic group

	Full sample N = 11 978	White N = 9794	Black N = 1309	Hispanic N = 404	Other N = 471
Age, mean (SE)	71.10	71.80	67.53	67.24	69.28
Age group
Under 65	16.42%	14.25%	29.45%	30.90%	15.97%
65‐74	47.28%	47.33%	43.22%	41.92%	58.02%
75‐84	26.19%	27.75%	20.44%	15.46%	18.92%
85 plus	10.11%	10.66%	6.90%	11.70%	7.10%
Female	54.84%	55.04%	57.52%	57.20%	44.69%
Married	52.98%	55.07%	36.02%	40.33%	62.11%
Education: Some college or more	46.58%	48.95%	34.66%	12.70%	51.53%
Income > 200% FPL	57.59%	62.70%	32.00%	15.50%	51.70%
Region
Northeast	18.12%	18.87%	16.07%	18.88%	18.73%
Midwest	22.04%	23.31%	13.57%	6.38%	14.26%
South	38.43%	37.24%	6.22%	27.70%	26.67%
West	20.15%	19.69%	7.52%	38.74%	38.16%
Puerto Rico	1.12%	0.10%	0.01%	8.29%	2.18%
Rural	7.27%	8.06%	4.70%	1.20%	3.30%
Medicare type
FFS not	54.60%	54.78%	56.65%	45.38%	52.41%
FFS in A	15.40%	15.93%	11.02%	15.00%	16.44%
Medicare	30.00%	29.31%	32.34%	39.62%	31.16%
LOT‐R Score, mean	17.30	17.51	16.66	16.40	16.42
Proxy completed interview	7.57%	6.45%	11.00%	14.64%	14.19%
Select chronic conditions
High blood pressure	66.50%	64.50%	79.90%	67.70%	65.60%
High cholesterol	65.00%	64.80%	64.90%	71.10%	64.09%
Stroke	10.60%	10.19%	13.92%	10.76%	10.22%
Arthritis	52.00%	52.70%	49.05%	50.38%	48.16%
Diabetes	31.60%	29.34%	40.46%	43.70%	42.90%
Depression	29.00%	28.76%	27.93%	48.99%	25.33%
Emphysema/asthma/COPD	21.30%	22.20%	19.06%	16.96%	14.65%
Skin cancer	19.40%	22.97%	0.66%	10.15%	5.80%
Osteoporosis	18.60%	19.43%	11.82%	25.69%	15.47%
Broken hip	3.18%	3.30%	2.45%	3.38%	2.69%
Patient activation
Low	24.75%	23.49%	26.34%	46.85%	34.27%
Medium	37.27%	37.89%	35.66%	32.60%	31.77%
High	37.97%	38.62%	38.00%	20.55%	33.96%

Values in table are percentages unless otherwise indicated.

Abbreviation: LOT‐R, Life Orientation Test—Revised.

3.2. Unadjusted patient‐perceived integrated care

PPIC means were mostly mid‐range on the scale of 1 to 4. On average, perceived integrated care was highest among beneficiaries for test result communication (Mean = 3.67, Confidence Interval = 3.65‐3.70) and provider support for medication adherence and home health management lowest (Mean = 1.99, CI = 1.96‐2.02). Comparing across racial/ethnic groups, there were significant differences between racial groups in means for a majority of PPIC domains, including staff knowledge about their medical history (P < .01), provider support for patient's self‐directed care (P < .01), provider knowledge of the patient (P < .01), provider support for medication adherence and home health management (P < .01), and specialist knowledge about the patient's medical history (P < .0001). Black and Hispanic beneficiaries perceived higher integrated care than White beneficiaries did in all of the PPIC domains described above. Unadjusted PPIC means are included in Table 2.

TABLE 2.

Unadjusted PPIC Scores by Racial/Ethnic Group

	Full sample N = 11 978	White N = 9794	Black N = 1309	Hispanic N = 404	Other N = 471
Provider knowledge of the patient	3.513 (0.01)	3.525 (0.01)	3.450 (0.03)	3.521 (0.04)	3.433 (0.04)
Staff knowledge of the patient's medical history	3.133 (0.027)	3.137 (0.03)	3.195 (0.06)	3.422 (0.09)	2.741 (0.16)
Provider support for the patient's self‐directed care	2.356 (0.02)	2.326 (0.02)	2.516 (0.05)	2.639 (0.08)	2.373 (0.06)
Provider support for medication and home health management	1.988 (0.02)	1.938 (0.02)	2.231 (0.04)	2.405 (0.08)	2.091 (0.72)
Specialist knowledge of the patient's medical history	2.103 (0.01)	2.083 (0.02)	2.241 (0.04)	2.240 (0.08)	2.245 (0.05)
Test result communication	3.674 (0.01)	3.686 (0.01)	3.644 (0.03)	3.593 (0.06)	3.566 (0.07)
Outside of office visit support	3.277 (0.02)	3.273 (0.02)	3.330 (0.04)	3.323 (0.06)	3.200 (0.07)
Provider support and knowledge following a hospitalization	2.616 (0.04)	2.639 (0.04)	2.529 (0.06)	2.400 (0.14)	2.498 (0.20)

Table displays the mean score for each PPIC with linearized standard errors in parentheses. Bolded values indicate where there is a significant (P < .05) positive difference in mean scores across subgroups using an Adjusted Wald test.

3.3. Adjusted differences in patient‐perceived integrated care

In our adjusted analyses using the rank and replace method, we found several differences between Black versus White and Hispanic versus White beneficiaries. Adjusted results resembled our unadjusted results. Mean perceptions for self‐directed care among Black beneficiaries were 0.14 (SE: 0.06) higher than among White beneficiaries (P =.02). Mean perceptions among Black beneficiaries for their specialists’ knowledge of their past medical history were significantly higher than among White beneficiaries (Mean difference: 0.12, SE: 0.06, P = .01). Black and Hispanic beneficiaries also each reported on average 0.18 greater support for medication and home health management than did White beneficiaries (P < .01 and P < .01).

After applying the Benjamini‐Yekutieli adjustment for multiple comparisons, only differences Black versus White and Hispanic versus White beneficiaries for provider support for medication and home health management remained significant. On average, Black beneficiaries reported 0.18 greater provider support than did White beneficiaries (P = .03, CI: 0.06‐0.30) while Hispanic beneficiaries reported provider support 0.37 higher than did White beneficiaries (P = .02, CI: 0.13‐0.61).

3.4. Alternative specifications

In contrast to our initial hypothesis, our primary analysis suggested that Black and Hispanic minority patients perceive some aspects of integrated care as higher than do White patients. We considered the possibility that our primary model was over‐specified for chronic health conditions, affecting our results. Thus, we reran our models using categories of health care need developed by Joynt et al ⁴⁷ : healthy patients, patients with simple, minor complex, and major complex chronic conditions; frail patients above the age of 65; and patients under the age 65 with disability. We used these categories in our rank and replace model to rank individuals according to their health care need and replace health care need coefficients between race/ethnicity groups. In our second alternative specification, we included patient activation as an additional population characteristic. In both of these analyses, our findings were similar to our primary model. After correcting for multiple comparisons, Black and Hispanic beneficiaries perceived their support for medication and home health management as 0.18 and 0.37 higher, respectively, than did White beneficiaries (P =.008 and P =.003). Both of these alternative specifications, and our primary specification, are included in Table 3.

TABLE 3.

Models of PPIC Scores

	Provider support and knowledge following hospitalization	Outside of office visit support	Provider knowledge of patient	Provider support for patient's self‐directed care	Specialist knowledge of patient's medical history	Staff knowledge of patient's medical history	Provider support for medication and home health management	Test result communication
A. Primary Specification
Black‐White difference	−0.05 (0.12)	0.03 (0.04)	−0.05 (0.03)	0.14* (0.06)	0.12* (0.06)	−0.06 (0.11)	0.18** (0.06)	−0.01 (0.04)
Hispanic‐White difference	−0.17 (0.24)	0.11 (0.06)	0.02 (0.06)	0.20 (0.11)	0.20 (0.11)	0.35 (0.19)	0.37** (0.12)	0.05 (0.08)
Other‐White difference	−0.29 (0.30)	−0.09 (0.10)	−0.09 (0.06)	−0.06 (0.09)	0.20 (0.08)	−0.38 (0.20)	0.07 (0.08)	−0.10 (0.09)
B. Sensitivity analysis: Inclusion of Joynt indicators of health care need
Black‐White difference	−0.05 (0.12)	0.03 (0.04)	−0.05 (0.03)	0.14* (0.06)	0.12* (0.06)	−0.06 (0.11)	0.18** (0.06)	−0.01 (0.04)
Hispanic‐White difference	−0.17 (0.24)	0.11 (0.06)	0.02 (0.06)	0.20 (0.11)	0.20 (0.11)	0.35 (0.19)	0.37** (0.12)	0.05 (0.08)
Other‐White difference	−0.29 (0.30)	−0.09 (0.10)	−0.09 (0.06)	−0.06 (0.09)	0.20 (0.08)	−0.38 (0.20)	0.07 (0.08)	−0.10 (0.09)
C. Sensitivity analysis: Inclusion of patient activation variable
Black‐White difference	−0.23 (0.19)	0.08 (0.05)	0.00 (0.04)	0.17* (0.09)	0.17* (0.08)	−0.01 (0.14)	0.31** (0.08)	−0.03 (0.06)
Hispanic‐White difference	−0.27 (0.33)	0.14 (0.09)	0.06 (0.08)	0.34 (0.16)	0.06 (0.13)	0.54 (0.21)	0.34** (0.19)	0.05 (0.11)
Other‐White difference	−0.95 (0.24)	−0.03 (0.15)	−0.05 (0.08)	0.00 (0.12)	0.15 (0.12)	0.00 (0.21)	0.08 (0.12)	−0.03 (0.08)

Standard errors in parentheses. *P < .05 without multiple comparisons correction, **P < .05, with Benjamini‐Yekutieli correction; controls in the primary specification include: urban/rural, geographic region, education, income, health insurance type, and LOT‐R. A rank and replace estimate of health status was also used in the primary specification, which included chronic conditions described in Appendix S1 as well as age, sex, marital status, and whether a proxy completed the survey interview.

As our third alternative specification, we conducted multiple variable regression in lieu of the rank and replace model. Full results, including coefficient estimates and adjusted p‐values can be found in Appendix S2. The results of our regression analysis were similar to our unadjusted primary model and other specifications; Black beneficiaries perceived their providers’ support for medication and home health management and their providers’ support for self‐directed care significantly higher than did White beneficiaries. (Coefficient: 0.12, SE: 0.05; Coefficient 0.13, SE: 0.05, respectively) Hispanic beneficiaries perceived their providers’ support for self‐directed care and their support outside of office visits significantly higher than did White beneficiaries (Coefficient: 0.31, SE: 0.11; Coefficient: 0.29, SE: 0.11, respectively).

4. DISCUSSION

In this study, we examined differences in patient perceived integrated care across racial/ethnic groups in a national sample of Medicare beneficiaries, using an analytical approach designed to ensure appropriate comparisons. Our results suggest several differences in perceptions of integrated care among Black, Hispanic, and White beneficiaries—each favoring the minority groups relative to White beneficiaries. Among eight domains of integrated patient care, Black and Hispanic beneficiaries reported higher integrated care for three: support for medication and home health management, specialists’ knowledge of their past medical history, and support for self‐directed care.

Our results suggest that Black and Hispanic Medicare beneficiaries perceive some aspects of their care as more integrated than do White beneficiaries. We adopted the NAM definition for racial disparities and adapted the Andersen's Access to Care conceptual model in an attempt to parse out racial disparities due to structural and provider discrimination from differences due to clinical need, appropriateness of interventions, and other individual characteristics. ¹ , ⁴⁹ All three of our alternative specifications support our primary rank and replace model, suggesting that our findings are not due to errors in model specification or our use of the rank and replace method.

There are several possible explanations for our results. One possibility is that minority patients indeed receive more integrated care than do White patients. However, this explanation is unlikely because it contradicts established past literature on racial disparities in access, delivery, and health outcomes. ⁵ , ⁶ , ⁵⁰ , ⁵¹ , ⁵² More likely, our findings highlight a potential bias in patient‐measured metrics. It is possible that minority patients perceive their care as more integrated than nonminority patients, despite differences in quality of care delivered or underlying expectations, inherent prejudices, and differences in culture. ⁵³ , ⁵⁴ Prior literature, for example, suggests that minority patients tend to rate care more positively and utilize the extremes of 10‐point rating scales more frequently. ⁵⁵ An historical precedent for discrimination against minority groups may result in lower expectations among beneficiaries from minority groups—particularly for Black beneficiaries, ⁵⁶ , ⁵⁷ hampering consistent interpretation of response to measures across groups. Prior research also suggests immigrant Hispanic and Asian populations may have greater difficulty communicating care needs, and providers may have greater difficulty recognizing verbal and nonverbal communications. ⁵⁴ Combined, these factors make it difficult for patients to communicate preferences and recognize efforts to provide integrated care, introducing potential measurement bias. ⁵⁸ , ⁵⁹ Our analysis accounted for only some of the variables in our adapted framework. A more robust data set that enables exploration of these nuances and other hypotheses drawn from critical race theory and its methods would greatly benefit this area of research. ⁶⁰

Our analysis has several limitations. By using the rank and replace method, we created hypothetical populations with equal health status, then controlled for SES and other variables, and measured differences in perceived integrated care. However, we acknowledge that health, socioeconomic status, and race interact. ⁶¹ , ⁶² To analyze more granular subsets, with greater potential for directing future policy and health care investments, requires larger population samples. Further research should also consider additional measures of expectations and individual characteristics, especially related to differences in cultural values, language barriers, and how prior personal, family, and racial history with health care may influence ratings.

Despite having a large sample overall, we had notably smaller samples of minority patients. While we were able to compare differences in patient‐perceived integrated care for Black, Hispanic, and White patients, we did not compare differences in patient‐perceived integrated care for Asian, Native American and other minority populations because of small sample size. Future research would benefit from including larger samples of minority populations in analyses of differences in perceptions of integrated care. Differences due to proxy‐assisted interview completion and the level of cognitive impairment due to prior strokes or other debilitating illnesses may have also affected individuals’ capacity to complete the MCBS interview and accurately rate integrated care. Limitations of the present study should be used to direct further research toward addressing differences in patient‐perceived integrated care.

5. CONCLUSION

In this paper, we explored racial and ethnic disparities across multiple aspects of patient‐perceived integrated care. Our findings suggest there may be aspects of care that Black and Hispanic beneficiaries perceive as more integrated than do non‐Hispanic White beneficiaries. Further research should seek to assess the source of this difference, whether real or a result of upstream factors. Developing improved methods for assessing racial and ethnic differences in integrated care will be increasingly important as the need for more integrated care continues in order to ensure equity in care delivery.

Supporting information

Author matrix

Appendix S1

Appendix S2

Ling EJ, Frean M, So J, et al. Differences in patient perceptions of integrated care among black, hispanic, and white Medicare beneficiaries. Health Serv Res. 2021;56:507–516. 10.1111/1475-6773.13637