We examined reported racial/ethnic healthcare discrimination in a diverse population of insured diabetes patients in an integrated healthcare delivery system, focusing on characteristics of PCPs and their relationships with patients. Those seeing nurse practitioners and those with better provider communication ratings reported less discrimination, and there was a positive association between number of visits and patient-reported discrimination. The fully adjusted model adjusted for variables that might influence how individuals perceive and/or report discrimination, such as patient ratings of provider communication, to attempt to reduce confounding among individuals who dislike their provider or consistently rate their healthcare experience poorly. Because all patients were insured with access to the same healthcare, there was likely less confounding due to insurance status compared to previous population-based studies.
While our population of individuals with chronic illness may be higher users of healthcare than the general population, reports of healthcare discrimination were uncommon (5% in our sample, and 3% when standardized to the racial/ethnic distribution of the Diabetes Registry). The sample facilitated examination of several racial/ethnic minority groups, and we were able to examine a number of provider-level characteristics not available in other studies (42
). Previous studies have suggested that female providers engage in more patient-centered care (14
); however, our findings suggested no association with discrimination, similar to a previous study from this same population examining provider gender and patient satisfaction (43
). In addition, we found no difference in reported discrimination comparing those seeing primary care vs. specialist physicians. Another study reported that providers with primary care training were rated as more participatory (15
), and all providers in our sample provided primary care to diabetes patients regardless of specialty. Furthermore, we found that those with more primary care visits reported more discrimination. Although we originally hypothesized an association in the opposite direction, our findings suggest more exposure to the healthcare setting could be related to an increased opportunity to experience unfair treatment. While more visits could also indicate poorer health status, we ran a sensitivity model (results not shown) also adjusting for the physical functioning SF-8 score, and the association between visits and patient-reported discrimination persisted.
LEP, depression, and CAHPS provider communication were also all strong predictors of reported healthcare discrimination in this adjusted analysis. Based on our previous work examining patient correlates of discrimination in this cohort (44
), LEP as a measure of acculturation – as opposed to immigrant status, which has been shown to be influential in previous studies (45
) – was the strongest predictor of perceived discrimination. Language-speaking ability may represent a clear obstacle to open communication between patients and providers, especially since English is not the first language for a growing population of patients. This is consistent with our finding that provider communication plays an important role in perceptions of unfair treatment. In addition, depression was a significant predictor of reported discrimination, consistent with the large body of literature on perceived discrimination (46
). Finally, the CAHPS provider communication score was related to patient reports of discrimination, which may be expected given that those who feel that they have experienced unfair treatment are then more likely to report that their provider did not listen carefully to them or explain things clearly (or vice versa). Disentangling causality among these factors and patient-reported discrimination in future studies would provide additional insight.
Our adjusted findings also did not
suggest that longer patient-provider relationships were related to reported discrimination, contrary to previous work examining ratings of interpersonal care (16
). Longitudinal studies examining if and when patients switch providers due to perceptions of unfair treatment could help determine the causality of this relationship. Our post-hoc examination of those who switched providers from the time they were matched to their PCP (i.e., 2004) until they completed the survey (i.e., 2005–2006) demonstrated a higher prevalence of reported discrimination: of the 2,024 individuals who switched providers, 7.6% reported discrimination in the previous 12 months, compared to only 4.2% of 10,100 individuals who did not switch. While dropping these patients who switched did not change the associations found in our primary analyses, it reduced the magnitude of the relationships.
Racial concordance was also not associated with decreased patient-reported healthcare discrimination once the models were adjusted for patient race/ethnicity, and there were no significant interactions with patient race/ethnicity. Although we did not examine language concordance in this study, a previous study of this population found that LEP patients seeing a language-discordant provider reported poorer interactions with their provider, while LEP respondents seeing a language-concordant provider reported similar interactions as English-proficient respondents (48
) – suggesting that language might be more influential in patient-provider interactions in this patient population.
Previous studies examining racial concordance in relation to reports of discrimination have also produced similar mixed and sometimes counterintuitive results. In a national patient survey, Asians were less likely to report being treated unfairly due to race/ethnicity in a concordant relationship, yet Latinos were more
likely to report being treated with disrespect in a concordant relationship (10
). Another study using the same survey data reported that concordance decreased perceptions of racial bias in healthcare treatment among white respondents, but not among minority respondents (11
). Finally, analysis of another survey found that black patients in concordant relationships were more likely to report being treated with respect, while this was not significant for Hispanic patients in concordant relationships (49
). Furthermore, when examining outcomes beyond patient ratings of care, a recent review found inconsistent evidence between racial concordance and improved health and/or healthcare outcomes (42
), and a study within the same population of diabetes patients at Kaiser Permanente found no relationship between racial concordance and medication intensification (50
). It is important to note that cultural competency training has been a mandatory component for continuing medical education at Kaiser for several years, ensuring that providers receive training and education on the topics of diversity, language, and improved health care delivery. The training and a healthcare system that embraces cultural sensitivity could have influenced our null findings with racial concordance. In addition, the racial concordance category for Asians may be too crude. Because of the large amount of heterogeneity between Asian subgroups and inability to disaggregate the provider Asian category, our measure would have classified Chinese patients seeing an East Indian PCP as “concordant.” Additional studies are needed to examine Asian provider subgroups to explore these relationships further.
This is the first study to our knowledge that examined provider type in relation to reported discrimination, although it is consistent with previous literature on nurse practitioner care. A review found that patients were more satisfied with care from nurse practitioners compared to physicians, and that nurse practitioners had longer visits with more investigations (51
). In addition, black patients seeing nurse practitioners in another study had increased trust compared to those seeing physicians (38
). These findings also support previous evidence that nurse practitioners focus on listening and asking questions, participatory decision-making, and “whole-person care” in their practice (52
). In particular, patients could feel that the interpersonal elements of care were more satisfied when seeing a nurse practitioner.
This study has additional limitations to acknowledge. We were not able to account for patient preferences in choosing their provider, which may also influence how individuals perceive discrimination. Previous studies (including within this diabetes population) found that minority patients are more likely to choose a provider of the same race/ethnicity (54
), and that black patients who reported a preference for a concordant physician were more likely to rate their physician as excellent (56
). Other research has suggested that patients who perceive discrimination in race-discordant relationships were more likely to prefer a provider of the same race (57
). Previous research has suggested that racial concordance is most important for this smaller group of patients that prefer a provider of the same race/ethnicity (58
). Thus, studies accounting for patient choice of provider might provide more insight into these relationships, especially if those who prefer a racially concordant provider have a heightened sensitivity to perceiving discrimination.
Moreover, our survey item captured discrimination from any provider at Kaiser and patients could have been reporting discrimination from a different provider (or even other staff interactions) rather than the PCP when responding to the questionnaire. Given that patients are likely to consider the full range of interactions with the health care providers rather than focusing only on those with the PCP, there is some potential misclassification in our analyses. However, it is possible that the experiences with the PCP, as the coordinating healthcare provider, may be psychologically buffering and potentially override experiences with other care providers. In addition, we used a single-item for capturing a personal experience with discrimination in the past year, which has been shown to underreport the true prevalence and variance in the population (31
) – perhaps influencing our findings conservatively. Moreover, while a broader measure (assessing differential healthcare treatment for racial/ethnic minorities overall) might have resulted in a higher proportion of patients reporting discrimination (60
), we were most interested in personal experiences with discrimination since we matched these reports to characteristics of the respondents’ PCPs.
Finally, there were missing survey data, primarily for the discrimination outcome. While the missing outcome data could bias the marginal estimates of self-reported discrimination in our population, it is unlikely to substantively impact the associations (61
), especially given that we controlled for variables associated with missingness in our adjusted regression models (62
). Additional examination of the survey response patterns revealed there was survey “fatigue” in answering the discrimination item, which was near the end of the 184-item questionnaire, but there were no other noticeable patterns suggesting a response bias compared to other, similar questions near the end of the survey.