Patient-Physician Racial Concordance and the Perceived Quality and Use of Health Care FREE

NUMEROUS studies have demonstrated racial inequalities in health care in the United States. Specifically, minority populations have less access to care,^1- 6 use fewer health care resources,^6- 8 and are less satisfied with the care they receive,^3,5 than the majority white population. Differences in health insurance coverage do not fully explain these disparities.^1,4,8

Racial inequalities in health care may be partly attributable to racial, cultural, and communication barriers between minority patients and white health care providers.^9- 12 Such barriers might arise from cultural or linguistic incongruity between patient and physician, from lack of mutual trust, or from racial discrimination. If these barriers existed, one might expect patients and physicians of similar racial or ethnic background to have better communication and more salubrious relationships than those of dissimilar background. Better relations might in turn lead to greater patient satisfaction and more effective use of the health care system. While this reasoning seems plausible, little empirical evidence exists to support it. We therefore sought to determine the extent to which racial concordance between patient and physician affects patients' ratings and reported use of health care.

We analyzed data from the Commonwealth Fund's Minority Health Survey, a telephone survey of noninstitutionalized adults in the 48 contiguous United States. Interviews were conducted between May and July 1994 in 6 different languages, using an interviewer-administered, computer-assisted telephone interviewing system. Data from this survey contained information regarding individuals' health care access and utilization, regular physicians, health status, and demographics.

Sampling was designed to obtain approximately 1000 white, 1000 black, and 1000 Hispanic subjects. Random digit dialing using a 3-stage, stratified-sampling process was used to ensure proper representation of households in different regions of the country and in central-city, suburban, and rural areas. After achieving the desired number of white respondents in initial cross-sections, oversamples of black and Hispanic Americans were obtained by screening additional national cross-sections and interviewing only members of those minority groups, until the desired sample sizes had been achieved. Additionally, interviews were conducted with Chinese, Vietnamese, and Korean Americans drawn from list samples, based on surnames listed in telephone directories.¹³

Of 10,576 individuals contacted, 5776 (55%) agreed to participate. The eligibility status and characteristics of nonparticipants were not known. Of those agreeing to participate, 1684 (29%) were ineligible either because they were not among the targeted racial groups or because the quota for their racial groups had been filled. Of 4092 respondents known to be eligible, 3789 (93%) completed interviews.

The 626 Asian respondents consisted, by sampling design, almost exclusively of Chinese, Vietnamese, and Korean Americans, and were not representative of Asian Americans in general. They were therefore excluded from analysis, as were the 19 American Indians and 24 respondents who did not identify their race.

After classifying themselves as Hispanic or non-Hispanic, respondents were asked to indicate their race as well as their national origin. Respondents were also asked if they had a regular physician to whom they usually went when they needed health care. Those who responded affirmatively identified their physicians' race. Based on these responses, we categorized individuals and their physicians into the following groups: non-Hispanic white, non-Hispanic black, and Hispanic. We considered racial concordance to be present when respondent and physician were in the same group.

We examined the association of racial concordance with responses to questions in the following 4 categories: ratings of physician, satisfaction with health care, receipt of preventive care, and receipt of needed medical care (Table 1).

We examined potential confounding variables in 4 domains: sociodemographic factors, access to care, sources of care, and health and well-being. Sociodemographic variables included age, sex, marital status, urbanicity, geographic region, education, employment status, household income, home ownership, receipt of public assistance, primary language (English vs other), birthplace (United States vs other), and number of years in the United States (for immigrants). Access variables included health insurance type, convenience of physician's office hours and location, and major barriers faced when seeking medical care: transportation, cost, waiting time, difficulty getting an appointment, and poor access to specialty care. Variables related to sources of care included usual care site, ability to choose one's care site, health maintenance organization membership, number of physicians seen over the last year, physician specialty (generalist vs other), ability to choose one's regular physician, physician sex, and patient-physician sex concordance. Variables related to health and well-being included self-perceived health status, number of health care visits during the last year, hospitalization within the last year, and a psychological score derived from 5 questions measuring psychological distress and well-being.¹⁴

For bivariate comparisons of respondents with racially concordant vs nonconcordant physicians, we used t tests for continuous variables and Pearson χ² test for binary and categorical variables. Ordered response variables were dichotomized between the highest rating and all other ratings (eg, excellent vs other).

To adjust for covariates we used logistic regression analysis. We selected a uniform set of covariates for all responses within a given category (eg, ratings of physician) through the following model-building strategy. We built regression models for the association between racial concordance and each response variable within each racial group. All models included age, sex, insurance type, primary care site, and self-perceived health status. Other covariates were included if their elimination from a model containing other variables in the same covariate domain (eg, sociodemographic factors) changed the odds ratio (OR) estimate for the association being examined by 10% or more. Covariates that changed the OR estimate for one response variable within a category were used in the final models for all response variables in that category, and covariates that changed the estimate for one racial group were used for all racial groups. Interactions of racial concordance with respondents' education, income, language, and nationality were examined. All analyses were stratified by respondents' race.

We also conducted further analyses to explore potential reasons for observed associations. We first repeated all our analyses while adjusting for respondents' reported experience of racial discrimination within and outside medical care settings. Next, we repeated our analyses after excluding those respondents who stated that their choice of regular physician was influenced by the physician's race or ethnicity. We conducted these analyses to determine whether respondents who were "race-conscious," and who therefore might be expected to have better relationships with physicians of their own race, accounted for observed associations between racial concordance and our response variables.

To address the possibility that any observed associations were confounded by differences in health plans or in physicians' office staffs, we repeated our analyses adjusting for respondents' satisfaction with their health plans and with the helpfulness of their physician's office staff.

Finally, to address the concern that significant associations between racial concordance and our ordered response variables might be a result of our dichotomizations, we used linear regression analysis to test the relationship between racial concordance and our ordered measures modeled as continuous variables. All data were analyzed using SPSS for Windows (Release 7.0; SPSS Inc, Chicago, Ill).

A total of 3120 white, black, and Hispanic individuals completed the survey. Nine percent of black respondents were of Caribbean heritage. Hispanics were primarily of Mexican (53%) and Puerto Rican (16%) descent.

A total of 2331 respondents (75%) had a regular physician. Whites were more likely than nonwhites to have a regular physician (82% vs 71%; P<.001). Among the 2201 respondents who were able to identify their physician's race, whites were much more likely than non whites to have a racially concordant physician (88% vs 22%; P<.001) (Figure 1). Nonconcordant physicians for blacks and Hispanics were primarily white (82%) and Asian (14%).

Within each racial group, respondents with racially concordant and nonconcordant physicians were similar with regard to most of our covariates (Table 2). We used multivariate analysis to adjust for differences.

In unadjusted comparisons, blacks with racially concordant as opposed to nonconcordant physicians more often rated their physicians as excellent in providing health care, in treating them with respect, in explaining their medical problems, in listening to their concerns, and in being accessible (Table 3). Similar results were seen for other measures of satisfaction with physician. White respondents gave higher ratings to white physicians than to nonwhite physicians on treating them with respect, explaining medical problems, and listening to their concerns. Hispanics with Hispanic physicians were found not to differ significantly from those with non-Hispanic physicians in rating their physicians, but were more likely to be very satisfied with their health care overall. Multivariate analyses revealed similar patterns, although after adjusting for confounders, racial concordance among whites was no longer significantly associated with excellent ratings of physicians in treating patients with respect or in explaining patients' medical problems (Table 4).

In unadjusted analyses, blacks with racially concordant physicians were more likely than those with nonconcordant physicians to report receiving preventive care and always receiving needed medical care (Table 3). These findings persisted after adjusting for confounders (Table 4). Racial concordance was not significantly associated with differences in delaying care seeking for any of the racial groups.

Restricting the sample to respondents who said that their choice of regular physician was not influenced by the physician's race or ethnicity altered our results in 2 ways. The effect of racial concordance among whites on rating physicians as excellent in listening to concerns was substantially reduced (OR, 1.34; 95% confidence interval [CI], 0.80-2.24), as was the effect among blacks on receiving preventive care (OR, 1.21; 95% CI, 0.65-2.25). Adjusting for respondents' reported experience of discrimination did not materially change our results.

Adjusting for respondents' satisfaction with their health plans also had little effect on our results. However, adjusting for satisfaction with the helpfulness of physicians' office staffs reduced the magnitude of the effect of racial concordance on satisfaction with health care among Hispanics (OR, 1.40; 95% CI, 0.78-2.51).

When we used linear regression to model our ordered response variables as continuous measures, the results of significance tests were generally similar to those obtained in our logistic regression analyses (data not shown). The only differences were that (1) among whites, the association between racial concordance and ratings of physicians on listening to concerns was not significant; and (2) among blacks, the associations between concordance and ratings of physicians on explaining problems and on listening to concerns were also no longer significant.

No significant interactions were found between racial concordance and education, income, language, or nationality.

Our findings indicate the importance of racial and cultural factors in the patient-physician relationship. They also suggest that the importance of these factors differs across racial groups. We found that blacks were more satisfied with the care they received from black, as opposed to nonblack physicians. This finding was consistent across several measures of satisfaction. In addition, blacks with black physicians were more likely to report receiving preventive care and necessary medical care than blacks with other-race physicians. Whites rated white physicians more often than nonwhite physicians as excellent in listening to patients' concerns, but not in the overall provision of health care. Finally, Hispanic individuals with Hispanic, as opposed to non-Hispanic, physicians were more likely to be very satisfied with their health care overall, but not with their physicians. All these findings persisted after controlling for sociodemographic characteristics and factors related to health care access, sources of care, and health status.

Prior studies^15- 18 have demonstrated that minority populations receive a disproportionately large amount of their care from racially concordant physicians. Our data corroborate this finding. For instance, although black physicians account for less than 5% of the total US physician workforce,¹⁹ they served as regular health care providers for 23% of black individuals in our sample (Figure 1). While this observation may simply reflect the fact that many black individuals live in areas with high concentrations of black physicians,¹⁷ our findings suggest that black patients prefer the care they receive from black physicians. Racial matching between minority patients and physicians may result, therefore, not only from the geographic distribution of minority physicians but also from the nature of the care they provide as well.

We believe that our findings reflect important differences in health care quality and access for individuals with racially concordant vs nonconcordant physicians. Patient satisfaction with interpersonal care as assessed in our study is an important aspect of health care quality that influences whether a patient continues to see a particular physician or remains enrolled in a health care plan.^20- 22 Use of preventive services and the ability to obtain needed care are recognized indicators of access to health care.^23- 24 While self-reported preventive care probably overestimates actual use,²⁵ unless the reports of respondents with racially concordant physicians were more or less accurate than those with nonconcordant physicians, our results are likely to represent conservative estimates of true differences in utilization.²⁶

We dichotomized our satisfaction variables between the highest and all other ratings. This raises the concern that our findings may reflect differences of subtle, perhaps unimportant, degrees of patient satisfaction. However, we believe our dichotomizations were appropriate for several reasons. First, most survey respondents, including ours, express high levels of satisfaction with their health care, possibly because those who are not satisfied tend to shop around for different physicians or health plans until they are satisfied.²⁷ Most detectable differences in satisfaction, therefore, fall between the highest rating and all others. Second, differences between excellent and lower ratings of health care are associated with important patient behaviors, such as remaining with one's physician.²⁸ Finally, we repeated our analyses using linear regression to eliminate the effects of dichotomization, and the results were largely unchanged.

To add further perspective to our statistically significant results, we compared unadjusted estimates of the effect of racial concordance on some of our response variables to the effect of other predictors within our data set. Among blacks, the effect of racial concordance on ratings of physicians as excellent overall was similar to the effect of respondents' ability to choose their physician (absolute difference [AD], 16% vs 19%; OR, 1.94 vs 2.14), a factor associated with high levels of patient satisfaction.²⁹ The effect among blacks of racial concordance on receiving preventive care was similar to the effect of uninterrupted health insurance coverage during the previous 2 years (AD, 7% vs 6%; OR, 1.64 vs 1.75). The effect of racial concordance for blacks on receiving needed care was similar to the effect of having any health insurance at all (AD, 5% vs 6%; OR, 2.23 vs 2.56). Among Hispanics, the effect of racial concordance on overall satisfaction with health care was also similar to the effect of having any health insurance (AD, 10% vs 11%; OR, 1.56 vs 1.56).

Why might racial concordance between patients and physicians affect patients' assessments of their health care? Our analyses provide some insight. Among blacks, the strongest association between racial concordance and the several response variables measuring patient satisfaction was with respondents' ratings of their physicians in treating them with respect. This finding suggests that black physicians may have more harmonious interpersonal relationships with their black patients than do physicians of other races. Better relations may be a product of cultural and experiential similarities that promote mutual understanding and trust.³⁰ Alternatively, since many black physicians see large numbers of black patients, it is possible that better relationships are due to cultural competence acquired through practice rather than to factors more directly attributable to racial concordance.

The effect of black patient-physician racial concordance on reported use of preventive care diminished after the exclusion of respondents whose choice of physician was influenced by physician race. This suggests that for the majority of blacks, physician race did not appear to influence the likelihood of receiving preventive care. However, for a substantial minority of blacks, race was important, affecting not only their choice of physician but perhaps also their likelihood of using physician services.

Our findings among Hispanic respondents were more difficult to explain. It is puzzling that patient-physician racial concordance affected patients' satisfaction with health care without affecting their satisfaction with physicians. However, it is noteworthy that adjustment for satisfaction with the physician's office staff significantly reduced this effect. It is possible that Hispanics are more satisfied with their health care when the office staff, including the physician, provide a more culturally and linguistically congruent setting. Future studies should address this possibility.

Our results should be viewed in light of several limitations. First, participation in the survey was incomplete, raising the concern of selection bias. However, for our results to be biased in this respect, the effects of self-selection would have to be unevenly distributed between those with racially concordant and nonconcordant physicians.³¹ Second, nonparticipation resulted in a sample of subjects with higher levels of education, income, and insurance coverage than the population from which it was drawn (Table 2).³² Also, the Hispanic group had few recent immigrants, limiting the generalizability of our results. Third, we did not have detailed information on the ethnicity, or nationality, of physicians. We were therefore not able to test the effects of finer degrees of patient-physician ethnic concordance on our response variables. This may explain the paucity of observed associations among Hispanics, who were more ethnically diverse than the black respondents in our sample. It is possible, for instance, that Hispanic individuals of Puerto Rican heritage are more satisfied with physicians who are also Puerto Rican, but not necessarily with physicians who are Mexican American. We could not address this hypothesis with our data. Fourth, we relied on respondents' assessments of physician race, which may not always accurately reflect the physician's true race. However, it is likely that the mechanisms by which racial concordance affects patient satisfaction and use of health care depend more on patients' perceptions of the physician's race than on the physician's true race.

Finally, because of the observational nature of our study, the possibility exists that our findings do not reflect true associations but are the result of confounding. We adjusted for numerous variables associated with satisfaction and use of health care, including age, sex, education, income, insurance, language, primary care site, physician sex, access barriers, and psychological distress.^20,33- 40 Furthermore, we attempted to adjust for illness severity through variables in our data set, including health status, number of health care visits, and hospitalization. It remains possible, however, that unmeasured differences across our comparison groups in illness severity or other factors may have partially accounted for our findings.

These limitations notwithstanding, we believe our results confirm the importance of racial and cultural factors in the patient-physician relationship. The finding that racial concordance affected patients' perceptions of their health care in measurable ways suggests that there is room for improvement in the relationships that physicians have with patients of nonconcordant backgrounds and supports efforts to increase cultural competence among physicians caring for diverse populations.

Our findings may also help predict the likely effects of policies influencing the recruitment of minorities into the US physician workforce. In the wake of recent trends in policies regarding affirmative action and international medical graduates, a decline in the proportion of physicians from racial and ethnic minorities is expected.^41- 42 Studies^15- 18 showing that minority physicians provide care for a large proportion of the rapidly growing minority communities in the United States suggest that with fewer minority physicians, the physician supply for many already underserved communities may diminish even further. Selectively admitting medical students with an interest in primary care of underserved communities may mitigate this problem. However, in light of our finding that physician race may affect black and Hispanic patients' satisfaction with and use of health care, such admissions policies, without the specific aim of recruiting black and Hispanic students, would serve as only a partial solution.

Our findings point to unique benefits that black and Hispanic individuals may experience when receiving health care from black and Hispanic physicians. When considering policy changes likely to reduce the number of underrepresented minorities in the physician workforce, governments and educational institutions should be mindful of the detrimental impact that such changes may have on health care for minority populations.

Accepted for publication September 3, 1998.

This work was supported by the Robert Wood Johnson Clinical Scholars Program and the Department of Veterans Affairs.

The views expressed in this article are those of the authors and are not necessarily the views of the Robert Wood Johnson Foundation, the Department of Veterans Affairs, or the Commonwealth Fund.

Reprints: Andrew B. Bindman, MD, Primary Care Research Center, University of California, San Francisco, Box 1364, San Francisco, CA 94143-1364 (e-mail: bindman@itsa.ucsf.edu).