What Causes Racial Health Care Disparities? A Mixed-Methods Study Reveals Variability in How Health Care Providers Perceive Causal Attributions

Abstract

Progress to address health care equity requires health care providers’ commitment, but their engagement may depend on their perceptions of the factors contributing to inequity. To understand providers’ perceptions of causes of racial health care disparities, a short survey was delivered to health care providers who work at 3 Veterans Health Administration sites, followed by qualitative interviews (N = 53). Survey data indicated that providers attributed the causes of disparities to social and economic conditions more than to patients’ or providers’ behaviors. Qualitative analysis revealed differences in the meaning that participants ascribed to these causal factors. Participants who believed providers contribute to disparities discussed race and racism more readily, identified the mechanisms through which disparities emerge, and contextualized patient-level factors more than those who believed providers contributed less to disparities. Differences in provider understanding of the underlying causal factors suggest a multidimensional approach to engage providers in health equity efforts.

Introduction

Racial inequalities in health care access and quality are a vexing issue for US health systems, organizations, and policy makers.¹ Although the causes are complex and multilevel, there is widespread consensus that health care providers contribute to racial health care inequalities and play an important role in their elimination.^2,3 In response, health care systems, such as Kaiser Permanente and the Veterans Health Administration (VHA), state and federal agencies, and professional organizations have developed an array of training activities and programs for health care providers.^4,5 For example, the Office of Minority Health recently released enhanced National Standards for Culturally and Linguistically Appropriate Services in Health and Health Care, which are supported by e-learning programs and communication tools.⁶ The MedEdPORTAL offers a host of online programs for providers on unconscious bias, disparities awareness, and cultural competence.^7-9 Understanding providers’ beliefs about health care disparities, particularly their causal attributions for disparities, is important because if providers do not see themselves as potentially contributing to disparities, they may not see themselves as part of solutions to ameliorate disparities in their practice.

This study is grounded in the conceptual framework of causal attributions based in social psychological research, which indicates that individuals’ causal attributions for problems are important contributors to their beliefs about solutions.^10,11 In particular, causal attributions have a centrally important role in Bernard Weiner’s theory of social behavior.¹² His model posits that people respond to an event by judging its cause (ie, causal attributions), which then contributes to the level of responsibility they assign, the type of emotional reaction they feel, and their ultimate response. When people perceive the causes of actions to be within individuals, for instance, rather than the social or economic environment, Weiner’s theory suggests that reactions to them will be more punitive, whereas a more sympathetic helping response results when people perceive the causal attributions to be not under an individual’s control.¹² Applying this theory to providers, then, if providers perceive that racial health care disparities result from factors internal to their patients of color, they may be less inclined to feel that they have a responsibility to address disparities than when they perceive the causes of those disparities to lie with health care providers and the larger health care system.

Past survey evidence demonstrates that providers tend to consider patient factors (such as patient preferences and behaviors) as more important contributors to racial differences in care than provider factors (such as bias or poor communication).^13-16 This reluctance to identify providers as playing a role in causing disparities is concerning, because as mentioned above, the workforce is an important target for disparity reduction efforts.^1,17 However, most past research exploring provider attribution of causal factors has been quantitative, yet qualitative approaches may provide greater insight into providers’ perspectives, and their interpretation of the meaning of causal attributions, than can structured surveys. The few qualitative studies that do exist suggest that providers may acknowledge provider bias more than survey data have shown; however, these studies continue to demonstrate that providers place strong emphasis on the behaviors of patients of color as contributors to health care disparities.¹⁸

The bulk of this past research was conducted in the mid-2000s, and there has been a great deal of attention to racial disparities—in news media, policy documents, and the research literature—since then, and particularly to the systemic factors that shape such disparities.^19,20 Providers’ views about the origins may thus have changed since the first wide-scale attention to health care disparities was first published, in the Institute of Medicine 2002 Unequal Treatment report.²¹ In addition, in spite of much policy and organizational attention, health care disparities have remained persistent.²² More work is thus needed to describe providers’ views in recent years, especially as training opportunities and interventions continue to develop.

The purpose of the present study was to understand how health care providers make attributions about the causes of racial health care disparities, and to examine variation in how they interpret and assign meaning to those causes. Given evidence of resistance among providers to discussions of racial bias in health care in the past,¹⁸ we rely on real-world narratives as a way to engage health care providers in discussion. Our research objective was to illuminate the multiple dimensions of providers’ understandings of the causes of racial health care disparities. These findings contribute important information about how health care providers conceptualize disparities in practice and illuminate factors that could contribute to their readiness to make changes at the provider, practice, and system level to reduce health inequity.

New Contribution

This study advances the limited existing research on providers’ perspectives about health care disparities by using mixed-methods in our approach to assessing and evaluating their causal attributions. In addition, we leverage a novel communication device—narratives—to prompt challenging conversations among providers. Narratives have demonstrated promise in facilitating conversations about topics that people might otherwise resist or reject.^23,24 Narratives used in this study were drawn from real first-person provider accounts of encountering health care disparities, to maximize the likelihood that these narratives resonate with providers more than hypothetical prompts.

Methods

Setting

This study is part of a larger sequential mixed-methods study, conducted in the VHA health care system, aimed at developing and testing communication strategies for motivating providers to engage in actions to reduce health care disparities in their own practice. The VHA is the largest integrated health care system in the United States.²⁵ Currently, about 9.1 million of the 22 million US Veterans are enrolled in the VA health care system,²⁶ with black and Hispanic Veterans more likely than white Veterans to be enrolled. Approximately 23.5% of Veteran VA health care enrollees are members of a racial/ethnic minority group, which is somewhat less diverse than the US population, although the VA is projected to become increasingly more diverse.²⁶ The VA health care system is an equal access system that removes the obstacles of insurance and costs, and for certain metrics, racial and ethnic disparities are smaller among Veterans users compared with nonusers.²⁷ Moreover, extensive quality improvement efforts have led to the reduction and elimination of certain disparities.²⁸ Nonetheless, racial and ethnic disparities persist in many areas of VA care,^26,29 and researchers have identified key contributors, including provider-level factors such as lack of cultural competence and unconscious bias.²⁹

Sample and Recruitment Procedures

Study participants are 53 health care providers working at 3 sites in the VHA (see Table 1). The 3 sites were located in 3 different states in the Southern and Midwestern United States, selected based on their different racial demographics of the patient population and differing racial attitudes of people living across these geographic regions.^30,31 Nurse practitioners, physicians, and physician assistants were recruited through email and site visits to participate in a quantitative survey that 240 completed between January 6, 2014, and December 8, 2014. For the survey, we obtained a list of all physicians, nurse practitioners, and physician assistants from the 3 facilities from the VHA intranet (N = 637). We initially recruited participants through email. Given low response rates, we changed our recruitment approach to include in-person recruitment, in which the investigator at each site provided information about the study at staff meetings and invited providers to complete the survey, on paper or online. Subsequently, 134 participants completed the survey online and 106 completed a paper survey. The response rate for one of the sites was 59%. The response rates for the other 2 sites were lower (31% and 27%) because we were not able to track the surveys that were not deliverable. All surveys were self-administered and took 10 to 15 minutes to complete.

Table 1.

Characteristics of Sample of 53 Veterans Health Administration Health Care Providers, by Survey Classification of Causal Attribution to Providers.

	Full sample (N = 53)	Low in causal attributions to providers (n = 17)	High in causal attributions to providers (n = 36)a
	n (%)	n (%)	n (%)
Gender
Male	26 (49.1)	5 (29.4)	21 (58.3)*
Female	27 (50.9)	12 (70.6)	15 (41.7)
Age, mean (SD)	5.9 (11.2)	49.61 (12.14)	51.54 (10.86)
Race
White	46 (86.8)	14 (82.4)	32 (88.9)
Black	2 (3.8)	0 (0)	2 (5.6)
Latino	1 (1.9)	1 (5.9)	0 (0)
Native American	1 (1.9)	1 (5.9)	0 (0)
Asian	3 (5.7)	1 (5.9)	2 (5.6)
US born
No	5 (9.4)	3 (17.65)	2 (5.56)
Yes	48 (90.6)	14 (82.35)	34 (94.44)
Attended medical school in the United States
No	7 (13.5)	4 (23.53)	3 (8.57)
Yes	48 (90.6)	13 (76.47)	32 (91.43)
Professional degree
NP	14 (26.4)	6 (35.29)	8 (22.22)
MD	37 (69.8)	10 (58.82)	27 (75)
PA	1 (1.9)	0 (0)	1 (2.78)
Other	1 (1.9)	1 (5.88)	0 (0)

^aDifferences between high provider attribution and low provider attribution participants were tested for statistical significance using a 2-sample t test or Pearson chi-square depending on the variable. None were statistically significant except for gender.

^*P < .05.

Of these survey participants, 93 responded affirmatively to a survey item asking whether they would be willing to complete a qualitative interview. We purposively sampled from this 93 to achieve a diverse group of health care providers for the interview stage. Ultimately, 53 providers were interviewed (46 in person, 7 via telephone) between April 18, 2014, and October 30, 2014, at which point we determined we had reached saturation.

Survey Data

The quantitative survey instrument included our core measure that assessed providers’ perceptions of the causes of racial health care disparities, modeled after previous work³²: “It has been documented that minority patients in the VA, on average, receive lower quality health care than white patients. In your opinion, how much does each of the following factors contribute to these racial differences in health care quality: patient behavior, provider behavior, or the social and economic conditions in which patients live.” Endorsement of each factor was measured on a 7-point scale (0 = not at all, 6 = a great deal). The survey also included items on the extent to which specific patient, provider, and organizational factors such as providers’ biases, and patients’ health behaviors contribute to differences in health care quality for minority patients.

Quantitative Analysis

First, we calculated the means of the 3 factors (patient behavior, provider behavior, and the social and economic environment) for the 53 participants who participated in both phases of the study. We used bootstrapping methods to compare the means statistically, obtaining 95% confidence intervals (CIs). Next, we used the survey items described above on several specific patient, provider, and organizational factors to validate a measure of respondents’ attribution of disparities to providers, by comparing the responses with these items for respondents who rated as high provider attribution (HPA, ranking provider behavior as a 3 or more on the 5-point scale) or low provider attribution (LPA, ranking provider behavior as lower than 3). This stratification was then applied to our interview data, as described below. Validation of this HPA versus LPA classification is described in Appendix A.

Qualitative Interviews

Interviews were conducted using a semistructured interview guide. The interviews (46 in-person and 7 by telephone) were conducted by the senior author and the site principal investigator (PI) for one of the facilities and lasted approximately 30 to 45 minutes. Both interviewers were white, and had previously conducted research on race and ethnicity in health care. The interviewer obtained written, informed consent prior to the interview.

Providers were randomized to read one provider-focused and one patient-focused narrative, selected from a pool of 10 narratives. These 10 narratives were all first-person accounts of real health care providers discussing their experiences encountering health care disparities in their practice. They were either published in medical journals as perspective pieces (eg, JAMA, Health Affairs) or solicited through interviews with the study team, and then all were modified to be a consistent length but did not change the words the providers used. The narratives were chosen because they conveyed 2 types of causal attributions for health care inequality: provider-focused narratives, which focused on factors such as providers’ conscious and unconscious racial bias as the source of inequality, and patient-focused narratives, which focused on problems associated with the patient’s race, such as his or her mistrust of conventional medical treatment. Details of narrative selection and brief descriptions of the narratives are described in Appendix B and in another paper published by the study team (Burgess, 2017).³³ The order in which each type of narrative was presented was systemically varied. Participants interviewed in person were presented with paper copies of the narratives; participants interviewed by phone were emailed copies of the narratives prior to the interview, at the time the interview was scheduled.

After reading each narrative, providers were asked about their overall response to the narrative, how the story resonated with their own experiences providing care in the VHA, what role they think providers might play in racial health care disparities, and what they think VHA providers could do to improve the care of minority patients or patients from different cultural backgrounds. Interviews were audio-recorded, transcribed verbatim by a professional transcription service, and imported into NVivo 10 for qualitative content analysis.³⁴

Qualitative Analysis

Interviews were coded using both deductive and inductive approaches,³⁵ blind to the participants’ survey classification. Five authors served as primary coders, working together in an iterative process to construct a coding instrument to identify emergent themes. Two of the coders identified as white, 2 identified as African American, and one identified as Asian American. For the present analysis, we focused on codes identified a priori that captured patient-related, provider-related, health care system–related, and social structural (including structural racism) contributors to health care disparities. In our coding, we did not distinguish language describing identification of causes of disparities and language identifying solutions to disparities because these ideas were not conceptually distinct in how providers discussed them (for instance, providers identified “health literacy” as both a causal factor and as a solution to disparities).

The purpose of the analysis was not simply to document the major categories of factors that providers identified, however, but to illuminate differences in how providers reason about, interpret, and assign meaning to these factors. To identify this variation, we used a comparative method³⁶ to compare coded text across 2 groups of providers—those classified as high or low in attributing racial health care disparities to providers, as described above. We also separately identified outliers, those who were categorized as low or high in provider attribution in the survey phase, but whose language and reasoning differed from their group. To engage in-depth with the data and observe emergent differences in participants’ meaning and interpretation, 2 authors prepared and discussed memos synthesizing the coded text across the 2 groups of providers.

Results

Quantitative Survey Data

In response to the 3 items asking study participants to endorse the causes of health care disparities, participants attributed these disparities significantly more to social and economic conditions (M = 4.08, 95% CI = 3.53-4.57) than to patient behavior (M = 2.74, 95% CI = 2.25-3.25) or provider behavior (M = 2.96, 95% CI = 2.51-3.42). Overall, 37 participants were classified as HPA participants because they ranked provider behavior as a 3 or more on the 5-point scale; 17 were classified as LPA participants because they ranked provider behavior as 2 or lower.

Qualitative Data

Qualitative analysis identified examples of factors that providers noted in their discussions of attributions for health care disparities: patient-level factors, provider factors, health care system factors, and social structural factors. Examples of each of these factors and quotes illustrating themes at each level of attribution are provided in Table 2. Within the category of patient factors, providers discussed patient adherence, self-care and lifestyle behaviors, language and literacy, and that some patients of color are more mistrustful of the healthcare system or sensitive to racial dynamics. Within the category of provider factors, providers identified provider bias, labeling of patients, language barriers on the part of providers, and, in some cases, overt racism by providers. Many providers identified health care system factors that lead to disparities, such as lacking a diverse workforce, lack of interpreters, poor access to care, time constraints, and systematic factors that lead to differences in quality of care delivered (such as differences between public and private hospitals). Finally, providers described socioeconomic and structural factors that also contribute to health care differences between white patients and patients of color, such as poverty, discrimination, and structural racism. No one category of factors dominated the discussion of causal attributions.

Table 2.

Qualitative Analysis of Attributions for Racial Health Care Disparities Among Veterans Health Administration Providers (N = 53).

	Salient examples identified	Key quotation(s)	Key differences in articulation, interpretation, and meaning across provider typesa
Patient factors	Patient adherence to treatment, self-care behaviors, lifestyle behaviors, language and literacy, patient own sensitivity to race and/or mistrust	Patient compliance: “And I think you just have to realize and in fact we had so much trouble with that in the Hispanic population not being compliant with diabetic treatment.”	More blame associated with patient-level factors like mistrust and socioeconomic status rather than attributing to context for LPA providers More emphasis on patient sensitivity to racism for LPA HPA providers more often affirmed the role of race as a patient-level factor
Provider factors	Provider bias, patient labeling, lack of time, language barriers, racism	Labeling of patients: “Somehow it really requires less explanation when you say . . . African American alcoholic or Native American alcoholic. There is certainly . . . a mental image that I think people have in their minds that they conjure up quickly, when you put those kinds of labels on. And I do think those people sometimes get suboptimal care. Not necessarily bad care, but suboptimal care, and these people don’t particularly get the benefit of the doubt . . . [goes on] I think people would be more forgiving of the white patient.”	More discussion and detail on the process of providers’ role in producing disparate outcomes for HPA HPA providers more often affirmed the role of bias and stereotyping based on race
Health care system factors	Workforce issues (eg, diversity, lack of interpreters), access to care (eg, insurance, public/private), time constraints on relationship-building, systematic differences in quality	Access/insurance: “Payment and your insurance may dictate the level of, the speed with which you’ll get needed care . . . I see that as part of the problem, and was the delay because of ethnicity? [or] Was the delay because of payment?” Quality differences: “Was it just because it was a public hospital and the other was private? Cause the populations, at least race-distribution-wise weren’t that dissimilar. But I know that if the things that I saw happening in that public hospital had happened in a private hospital, no one would have ever let that happen in a predominantly upper class white neighborhood. The delays . . . difficulty in access to medications would never, people wouldn’t have allowed those to occur.” Time constraints: “And it all depends on how much time you have. And in a production-based environment, it really dehumanizes that relationship. Particularly in a difficult one, or a minority, or when you’re just not connecting, you just don’t have the time.” “I mean they don’t really give us time to find the human component on what’s going on with the patient . . . And I think a lot of it is just the pressure to see a number of people within a specific amount of time, and caring is not part of that equation.”	HPA providers more often affirmed the role of structural racism in discussing and acknowledging health care system factors
Social structural factors and systemic racism	Socioeconomic status, discrimination, systemic racism	Socioeconomic status: “It’s definitely frustrating to see the way that we as a country have treated [people] and I definitely think race is a part of it. But I think money comes into this too and uh, social status, and there is so much that . . . that is mixed in together.” Structural inequalities: “But then the reason black people are in poverty and Hispanic people are in poverty is because they were suppressed. Like black people were very suppressed, if you think about it. It’s only been maybe, you know, the Civil rights movement occurred in the 60s.”	HPA providers more often affirmed the role of race as a definitive factor More emphasis on patient sensitivity to racism for LPA More explicit acknowledgment of racism among HPA

Note. HPA = respondents who endorsed high levels of provider causal attribution; LPA = respondents who endorsed low levels of provider causal attribution.

^aProviders were compared across 2 groups based on their response to the following survey question: “How much does [provider behavior] contribute to racial differences in health care quality?”

However, the meaning and interpretation of factors within these categories differed a great deal across different types of providers. The last column of Table 2 identifies 4 major interpretive differences that emerged from the analysis between HPA and LPA participants across these levels of causal attributions. These included ascribing characteristics to individuals versus the larger context, openness to race as a definitive contributing factor, highlighting the processes through which disparities arise, and acknowledgment of racism. Each distinction, along with illuminating quotes, is described below.

Ascribing characteristics to individuals versus context

While both LPA and HPA groups described patient-level factors to a similar extent, LPA participants described these factors as attributes of individuals rather than interpreting them in a context. For instance, many providers discussed the concept of “mistrust” among their patients of color. LPA participants described mistrust as an individual psychological characteristic, an aspect of individuals, not the larger social context. One such provider said, “Sometimes I feel like people don’t always make—or patients don’t always make—the right decisions sometimes because of fear or education or mistrust.” Such providers implicitly assigned blame to patients for this mistrust. For HPA participants, in contrast, mistrust was identified as stemming from historical/structural factors:

Like we have heard about Tuskegee and all these things where the health care system wasn’t doing really the ethical thing, and so the skepticism on the part of the patient is totally understandable and I think reasonable . . . it like actually probably makes sense even though it’s hard for you as a doctor to deal with. But it’s not like your patient’s crazy . . . they know the history, they know what the world is like.

Providers applied similar interpretations to socioeconomic status or insurance status. Some participants (mainly LPA) referred to these patient-level factors as attributes of individuals (eg, having few financial resources as a justification for why a patient may not be able to afford a drug), whereas others (mainly HPA) explicitly identified the upstream cause of the particular patient-level factor, interpreting these individual-level patient factors as resulting from systemic or historical conditions (eg, not having equal employment opportunities).

Openness to the role of race as a definitive, contributing factor to disparities

Providers mentioned numerous contributing factors to differences in health care across patient groups, as displayed in Table 2, ranging from language abilities to insurance status to mental illness to the composition of the health care workforce. The key differences that emerged were in providers’ degree of openness to identifying race as a contributing factor. HPA participants were more likely to indicate that race was as plausible as other factors, using language emphasizing multiplicity (ie, a “both/and” rhetoric regarding race and other factors). For instance, one HPA participant noted the likelihood of provider bias and stereotyping as a result of patients’ educational attainment or their race: “I suspect that the problems are exacerbated . . . [and the provider is] more likely to be judgmental if they feel like, if the person is pigeonholed into a group where they’re considered to be less well-educated or maybe it could just be a race issue.” Other HPA participants articulated more definitively that race played a role in differential treatment: “She was, you know, in pain and vomiting. She was absolutely treated [differently] because of her race. No doubt about it.”

In contrast, LPA participants used definitive language to rule out patient race as a reason for differential treatment, reducing the importance of race in favor of behavioral explanations, social class, or health care system–related factors. Many suggested that racial disparities in care were attributable (often entirely attributable) to differential access to care and uninsurance. Others reduced racial differences to behavioral attributes, as in this example: “I haven’t necessarily seen it in regards to race in particular. It’s more like with particular behaviors. Say for example, if someone has . . . alcoholism or drug dependence as a disease. Uh, [doctors] will kind of have a bias towards them.” Another provider similarly acknowledged provider stereotyping, but indicated that these biases stem from patient attributes other than race:

I feel like the bigger issue, honestly is socioeconomic status. And that more providers, I think, are more likely to jump to conclusions about patients who are homeless or, you know, severely mentally ill, as opposed to making conclusions based on somebody’s race or ethnic background.

Highlighting the processes through which disparities arise

Third, providers differed in how they discussed disparities, particularly whether references were made to the pathways through which disparities arise versus just referencing the existence of disparities. As prompted by the narratives they read, nearly all providers across both groups acknowledged differing health outcomes across different types of groups. However, HPA participants described the process that produced these outcomes in more detail more often, including team dynamics, implicit or explicit bias or stereotyping, labeling patients based on their behaviors, and identifying how socioeconomic status serves as a mechanism producing disparities. For instance, one provider described how racially homogenous staffing in hospitals and clinics can create the conditions for health care disparities: “You have to have staff of different backgrounds. So if a white male physician is working only with other white male physicians, stereotypes, cultural biases are likely to be perpetuated.” Another articulated the process through which white patients may be given the benefit of the doubt, providing an overt example of the type of stereotypical reasoning in which providers might engage:

One that I certainly know is that there are differences in the way that black or Hispanic or patients who are foreign born . . . they’re treated differently than other more majority patients. Ones who probably look like your sort of typical hardworking American. Yeah, like the patient you can look at and know they’re going to do well, they’re going to do everything they possibly can. And in some others, you look at them and you’re like “they’re not going to be compliant, I don’t know why I’m bothering.”

Differences in acknowledgment of racism

Fourth and finally, differences emerged in how often providers acknowledged the role of racism in contributing to racial health care disparities. Providers across both groups identified racism as a contributor, but it was more often mentioned by HPA participants: 8 of the 17 (47%) LPA participants identified racism, whereas 27 of 36 (75%) HPA participants referenced racism. Importantly, the meaning they assigned to racism also differed. HPA participants were most likely to describe racism as a present struggle in society and/or the health care system, whereas LPA providers offered examples of racism in the past (ie, Civil Rights movement, Tuskegee). In addition, while both types of providers sometimes referred to patients of color having heightened sensitivity to race in their dealings with the health care system, LPA participants more often resisted the idea that racism exists in health care and were more likely to dismiss patients’ experiences of racism as illegitimate. As one provider recounted,

Long story short, the gentleman is pretending to be blind and trying to get benefits. So, he’s accusing the woman who’s the coordinator of the vision impairment program of being racist. Because she won’t sign off on a piece of paper that says that he’s blind. When he’s not. So, he’s saying that she’s racist and that’s why he’s not getting the care he feels entitled to.

While the previous example is extreme (given the claim of a falsified illness), others suggested that patients’ sensitivity to or concerns about racial discrimination in the VHA can interfere with the doctor-patient relationship.

Similarities across causal attributions

The qualitative analysis also revealed many strong similarities in how providers attributed the cause of disparities, regardless of their survey-measured prior beliefs. Providers expressed great empathy for patients in general and their patients in particular in the stories they recounted. In addition, most providers discussed access to care as an important determinant of racial health care disparities, and most also readily identified many other socioeconomic and structural factors that contribute to racial health care disparities.

Discussion

This study reveals important information on how providers attribute the causes of racial health care disparities. Study participants were more likely to indicate social and economic conditions as playing a role compared with patient behaviors, in contrast to earlier studies.^13-16 Providers rarely offered reductionist explanations that located disparities solely with patient behaviors, instead describing numerous multilevel pathways through which social and economic conditions, provider behavior, and the health care system shape patient outcomes. These findings demonstrate growing provider recognition of the literature on racial disparities and the social determinants that influence patient care, and are in line with the conclusions of commentators who argue for multilevel interventions to promote health equity.³⁷

Yet our results examining differences across subgroups illuminated striking heterogeneity in the ways in which health care providers interpreted and understood the causes of disparities. Most notably, providers differed in how much they acknowledged race and racism as playing a role. For some providers, behavioral and social attributes dwarfed or even eliminated the direct role of race in health care disparities. Factors like access to care, insurance, health literacy, and mental illness were all cited by some providers as the most plausible rationales for any racial or ethnic group differences that they observed, more plausible than racial discrimination and bias. This attribution suggests that efforts to directly name and confront race and structural racism in health care³⁸ may be met with resistance from the group of providers who either do not acknowledge racism or who judge the other social determinants as more valid explanations for inequalities.

This study further reveals that quantitative research studies assessing providers’ attributions about health care disparities are limited, because they do not incorporate the complex multidimensional reasoning in which providers actually engage, when given the opportunity. Quantitative assessments do not allow providers to identify what they assert to be the “causes of causes”—how they interpret the multiplicity of factors that produce disparities. Survey measures also cannot distinguish how providers assign meaning to the factors. For instance, study participants viewed the same patient-related factors (such as patients’ mistrust or unhealthy behaviors) in very different ways, with HPA participants much more likely to identify the social structural factors that produce these patient attitudes and behaviors, while LPA participants described these factors in ways that connoted more individual-level blame.

Implications for Practice and Policy

These findings suggest an openness among providers to organizational, provider-focused, and systemic changes to reduce racial health care disparities, especially when these policy approaches are contextualized in terms of the social determinants of health. Study participants articulated many ways in which unequal health care access, insurance differences, limited time, lack of interpreters, and—among some providers—bias, labeling, and stereotypes contribute to racial differences in patient experience and treatments. As with previous research, they also see patient characteristics as having a role in contributing to differences in care received. Like other health policy areas (such as obesity),³⁹ framing causes in a dichotomous way as either the environment (eg, health care systems) or individuals (eg, patients) is likely a less productive approach to engaging health care providers than is emphasizing the broader factors that shape patients’ experience. In addition, providers consistently emphasize access to care as an important policy lever for improving equity. These key findings suggest that interventions in the health care setting that rely on health care providers’ participation should acknowledge multiple ways that health care disparities emerge. Moreover, while this study was not designed explicitly to test communication approaches to assess which are more or less acceptable to providers, results from a related study suggest that certain narratives may be more acceptable and less likely to produce resistance, especially those that emphasize providers’ self-efficacy in addressing patients’ structural barriers (thus acknowledging the multiplicity of pathways at the provider and patient level that can lead to unequal care).³³

Our findings dovetail with recent recommendations from policy makers and health care leaders to eliminate health care disparities. For example, the VHA Health Equity Action Plan (HEAP) recommends that VHA leadership “assess health equity impact for all policies, memos, handbooks, procedures, directives, action plans, and governance,”^40(pS512) and “increase education and training on health equity, cultural competency to include unconscious bias, microinequities, diversity, and inclusion.”^40(pS512) Likewise, the Department of Health and Human Services Action Plan to Reduce Racial and Ethnic Health Disparities (HHS Disparities Action Plan), includes this strategy: “Increase the ability of all health professions and the health care system to identify and address racial and ethnic health disparities.”^2(p1824) These recommendations are consistent with many providers’ belief that organizational and systemic changes are needed to reduce disparities. However, our findings also suggest that a minority of providers do not (yet) accept the idea that racism is a direct contributor to health inequity. Additional research, grounded in social psychology and communication science, is needed to better understand whether providers with such predisposing beliefs would accept or reject policy changes framed around racism in health care and what additional strategies are necessary to offer these providers insights into the experiences of their patients of color.

Finally, any consideration of how these findings might be incorporated into real-world health disparities reduction efforts must acknowledge that medical care is just one—likely small—input into the bigger problem of health disparities. It is well known that health disparities result from the “upstream” social determinants of health, such as living and working conditions that are in turn influenced by economic opportunity, public policy, and political choices.²⁰ Interventions in the health care setting are still necessary to reduce the likelihood of patients of color having poor access to care or receiving inappropriate treatment, but are not sufficient to advance health equity. Reducing population health disparities will also require interventions that target the social policy factors that contribute to the social inequalities that lead to health outcomes, such as education, housing, and income.⁴¹

Limitations

Our study should be interpreted with certain limitations. First, like most studies drawing participants from a larger population, there may be differences between those who chose to respond to the surveys and participate in interviews, compared with the other VHA health care providers, and these differences may relate to how participants understand racial health inequities. Because we sampled for individuals willing to be interviewed from a survey that was about health care disparities, those who cared more about the issue may have been more likely to agree to be interviewed. Second, we used narratives to elicit discussions focused on particular individual patients and providers. In the language of social psychology, these narratives were “episodic,” in that they focused on specific individuals, rather than contextual factors and presenting systemic data about a problem.¹¹ This format of the narratives may have primed participants to offer more individualized assessments of both causes and solutions than they might have otherwise.¹¹ Third, racial diversity was limited among our sample, and past research suggests that providers’ attributions of disparities differ by providers’ race.⁴² Finally, our results may not be generalizable to patients beyond the VHA. While VHA patients are similar to Medicare beneficiaries, they are less similar to the Medicaid population or to privately insured patients.⁴³

Conclusions

This study provides insight into how providers attribute the causes of racial health care disparities and provides a foundation for understanding how to engage providers, with varying predisposing beliefs, in efforts to reduce health care inequality. Our findings are aligned with recent recommendations to reduce disparities in the health care setting, and point to the importance of a multidimensional approach, encompassing consideration of both patient circumstances and provider- and system-level factors in efforts to reduce health care disparities.^37,40

Appendix A

Validation of the Screening Question Among Survey Participants (N = 240)

High provider attribution (HPA) was defined as a score of greater than or equal to 3 on the question about providers’ contribution to racial health care disparities, whereas low provider attribution (LPA) was a score of 2 or lower. Quantitative analyses demonstrated the validity of this screening question and the use of this particular cut-point. Specifically, HPA participants were more likely than LPA participants to endorse specific provider-level and system-level factors as contributing to health care disparities including difference in provision of specialty referrals (P < .0001), provider workforce diversity (P < .0001), provider attitudes and beliefs about minorities (P < .0001), lack of time/resources to address social issues (P < .0001), patient social/economic circumstances (P < .0001), differences in prescribing medication (P < .0001), poor provider communication (P < .0001), and provider biases in decision making (P < .0001). HPA participants were also more likely to endorse the statements—minority patients receive lower quality health care than white patients (P < .0001) and the Veteran Affairs health care system treats people unfairly based on race/ethnicity (P < .0001), although endorsement of this latter question was low among both groups. Classification as high versus low was not significantly associated with having had cultural competency training related to health care disparities.

Classification was not significantly associated with gender, age, being US born, or current professional status, percentage of patients who are nonwhite, or provider race/ethnicity. However, 39% of white providers were classified as low, whereas 0% of black, 29% of Hispanic, and 36% of Asian providers were classified as low. The 2 providers who identified as American Indian also were classified as low.

Appendix B

Narrative Selections for Interviews

To engage participants, we selected narratives to use in the study in which the narrators were similar to the participants (ie, were health care providers), were told in the first-person perspective,^44-46 and were real stories recounted by the provider who had experienced the situation.⁴⁷ A literature review was conducted to identify published narratives told from the point of view of a provider confronting, recognizing, and/or resolving issues in his or her practice related to race. We sought to identify narratives that broadly have 2 types of causal attributions for health care inequality embedded within: provider-focused narratives, which focused on factors such as providers’ conscious and unconscious racial bias as the source of inequality, and patient-focused narratives, which focused on problems associated with the patient’s race, such as his or her mistrust of conventional medical treatment. The senior author identified an initial pool of approximately 20 first-person provider narratives through literature reviews of medical journals that publish first-person narratives of health care providers’ experiences, on a range of topics including race and racism. These included those published in the “A piece of my mind” section of the Journal of the American Medical Association,^48-50 the “Narrative Matters” section of Health Affairs,^51-55 the “Editor’s Choice” section of the American Journal of Public Health,⁵⁶ the “On Being a Doctor” section in the Annals of Internal Medicine,^57-61 and the “Perspectives” section in the Annals of Internal Medicine.⁶² A subset of the research team classified them as either provider-focused or patient-focused.

Because this initial strategy led to more provider-focused than patient-focused narratives, we supplemented these published works by soliciting narratives through interviews with a purposive sample of 4 physicians, who were known by the study team to be reflective about their experiences caring for patients of color. These physicians were prompted with a standardized prompt (“Can you tell me a story about a time when you realized that a patient of color had certain behaviors or attitudes that, in your view, were going to lead them toward poorer health outcomes. What did you do to address this?”

Using an iterative process during which the team discussed and categorized the narratives based on several criteria (eg, how well they fit our criteria of provider-focused and patient-focused, authenticity, length), the team identified 10 narratives for use in the study, 6 of which were provider-focused and 4 of which were patient-focused. Five of these were drawn from the published literature,^{48,51,56,59,60} and 5 were drawn from our interviews. Narratives were shortened so that they were of similar length.

Provider-focused narratives included stories of (1) a white physician describing how his relationship with an African American patient makes him aware of his own racial biases and bias in health care⁵¹; (2) an African American physician who describes receiving racially biased treatment in the emergency room for back pain⁵⁶; (3) a physician who reflects on his experiences witnessing racism in health care⁵⁹; (4) a physician who reflects on a situation in which an American Indian patient accused her of denying pain medication to him due to his ethnicity; (5) a physician whose African American patient died after being denied curative surgery for his cancer based on his appearance (which the surgeon deemed as healthy) even though his objective findings indicated that he was a fit candidate; and (6) a physician who misjudged the teenage son of an African American woman, hospitalized for a drug overdose.⁶⁰ Patient-centered narratives included stories of (1) a white physician who describes being unable to convince his older African American patient to get treatment for cervical cancer due to her mistrust in the medical system⁴⁸; (2) a physician who took the time to get to know a Mexican American patient perceived as difficult by the house-staff, and how this improved their relationships; (3) a physician who describes taking a patient-centered approach to helping African American smoker with heart problems, who was resistant to pharmacotherapy, to quit smoking and lose weight; and (4) a physician who was initially annoyed at a African American patient who accused clinic staff of being racist, but then established a relationship and learned about patient’s experiences of racism.