Colorblind racial ideology and physician use of race in medical decision-making

Abstract

Background:

Colorblindness is an ideology that minimizes the role of systemic racism in shaping outcomes for racial minorities. Physicians who adhere to colorblindness may be less likely to interrogate the role of racism in generating health disparities and less likely to challenge race-based treatment. This study evaluates the association between physician colorblindness and the use race in medical decision-making.

Methods:

This is a cross-sectional survey study, conducted in September 2019, of members of the Minnesota Academy of Family Physicians. The survey included demographic and practice questions and two measures: Color-blind Racial Attitudes Scale (CoBRAS; measuring unawareness of racial privilege, institutional discrimination, and blatant racial issues) and Racial Attributes in Clinical Evaluation (RACE; measuring the use of race in medical decision-making). Multivariable regression analyses assessed the relationship between CoBRAS and RACE.

Results:

Our response rate was 17% (267/1595). In a multivariable analysis controlling for physician demographic and practice characteristics, CoBRAS scores were positively associated with RACE (β=0.05, p=0.02). When CoBRAS subscales were used in place of the overall CoBRAS score, only unawareness of institutional discrimination was positively associated with RACE (β=0.18, p=0.01).

Conclusions:

Physicians who adhere to a color-blind racial ideology, particularly those who deny institutional racism, are more likely to use race in medical decision-making. As the use of race may be due to a colorblind racial ideology, and therefore due to a poor understanding of how systemic racism affects health, more physician education about racism as a health risk is needed.

Introduction

Health care providers use race to evaluate kidney and lung function; estimate the risk of atherosclerotic cardiovascular disease and the success of vaginal delivery following cesarean section; determine prostate cancer screening; and select antihypertensive medications [1–3]. However, as race is a social construct, its continued use as a biological variable within the medical community requires further investigation. Colorblindness is a racial ideology characterized by the minimization of racism—defined as a system of social and institutional power that advantages Whites and oppresses people of color [4]. Colorblind individuals dismiss racism by proclaiming that that they do not “see color” (an impossibility) and denying the existence of a racial hierarchy and structural racism [5]. Colorblindness functions to maintain the current racial hierarchy through advancing the belief that our society is just and people are treated fairly. Race, as a biological phenomenon, provides an explanation for racial health disparities that excludes discussions of racism whereas race, as a social construct, does not. By not acknowledging the social context of race, colorblind providers may ignore the important structural factors that drive health disparities. Physicians ascribing to a colorblind ideology may be less likely to interrogate the role of systemic racism in shaping health outcomes, less likely to challenge the notion of biological race and, therefore, more likely to engage in race-based medical treatment.

Racial categories have changed over time and vary between societies [6]. Furthermore, race is not a benign construction as biological race has played a significant role in the justification of abhorrent treatment of racial minorities [7]. Despite these truths, there has been relatively little discussion regarding the appropriateness of using race as a marker of biological risk [2, 8–10]. Controversy regarding race-based medical calculations and the use of race in medical guidelines is recent and little is known regarding physician motivations for using race to guide care. Studies evaluating physician use of race have found a positive association between the belief in the biological basis of race and greater use of race in medical decision making [11]. Colorblindness, and the desire to avoid confronting systemic racism, may explain why race continues to be used as a biological risk factor in medicine without greater examination of its origin, evolution, and relevance.

Colorblind racial beliefs have not been well evaluated in physicians. Thus far, there has only been one qualitative study that explicitly addressed colorblind beliefs in medical providers [12]. The aim of that study was to evaluate healthcare worker’s perception of the etiology of racial health disparities. The researchers found that healthcare personnel’s colorblindness informed their understanding of the causes of and solutions for health inequities.

In the current investigation, we used a cross-sectional quantitative survey to investigate the association between colorblindness and the use of race in medical decision-making among Minnesota family medicine physicians. We hypothesized that physicians who endorse colorblindness will be more likely to use race to guide medical treatment.

Methods

Study Design

In September 2019, members of the Minnesota Academy of Family Physicians (MAFP) were sent an email inviting family medicine physicians to participate in the study. A link to the survey was included in the email, and members who did not open the email were sent one reminder email a week later. Participants who completed the survey had the opportunity to enter a raffle for one of ten $100 gift cards. The study was evaluated by the University of Minnesota Institutional Review Board and considered exempt.

Participants completed an online informed consent followed by an online REDCap survey consisting of demographic and practice-related questions and two surveys: the Color-blind Racial Attitudes Scale (CoBRAS) [13] and the Racial Attributes in Clinical Evaluation (RACE) scale [14]. Participants were required to complete all demographic and practice-related questions before starting the CoBRAS and RACE measures. All questions in the CoBRAS and RACE measures required a response to submit the survey, but participants were given the option of selecting “I choose not to answer.” Participants who chose not to answer any part of the survey were excluded from the analysis.

Study Population

MAFP membership includes active (N=1337) and retired (N=173) family medicine physicians, family medicine residents (N=240), medical students (N=271) and “other members” (honorary, inactive and supporting members) (N=18). To proceed with the survey, participants had to identify as a physician or resident and describe their current practice location. The MAFP subsequently confirmed that all “other members” who accessed the survey were family medicine physicians.

Measuring Race in Medical Decision-making

The RACE (scored 0–28), is a seven-item measure that assesses the use of race in medical decision-making by scoring providers response on a five-point Likert scale—ranged from 0 (none of the time) to 4 (all of the time)—to statements such as “I consider my patients race to better understand their genetic predispositions.” A higher RACE score indicates a greater self-reported use of race in medical decision making.

Colorblindness

The Color-Blind Racial Attitudes Scale (CoBRAS) is a 20-item validated measure composed of three subscales that assess level of awareness of: racial privilege (7 items, scored 7–42), institutional discrimination (7 items, scored 7–42), and blatant racial issues (6 items, scored 6–36). Each item is evaluated using a six-point Likert scale ( 1 =“strongly disagree”, 6 = “strongly agree”). Example statements from each subscale are as follows: “White people in the United States have certain advantages because of their skin” (racial privilege); “Social policies, such as affirmative action, discriminate against White people” (institutional discrimination); “Racial problems are rare, isolated incidents in the United States” (blatant racial issues). The scores from the subscales are added together to create a total CoBRAS score. Higher scores indicate greater levels of colorblindness.

Covariates

Demographic variables included age, race, ethnicity, gender and location of medical education. Physicians who identified as Hispanic/Latinx ethnicity were grouped together irrespective of their racial identity. Non-Hispanic/Latinx physicians who selected more than one race were reclassified as multiracial in our analyses. Age was analyzed as a categorical variable, (e.g., 20–29, 30–39, 40–49, 50–59, 60+). We categorized location of medical education as Midwest or non-Midwest. Physicians were classified as Midwest-trained if they completed medical school or residency in the Midwest.

Practice variables included years in practice, practice type, practice location, and minority composition of patient panel. Categorical variables included years in medical practice (i.e., still in residency, 0–4, 5–9, 10–19, 20+) and minority representation—that is, physicians selected the category which best represented the percent of racial and ethnic minorities in their practice panels (e.g., <15%, 15–29%, 30–49%, 50–69%, 70–84%, >=85%). In our analyses, we collapsed diversity categories into low (< 30%), moderate (30%−69%), and high (>=70%). Physicians were able to select multiple practice types to describe their primary practice. Practice types were recategorized as private practice (e.g. independently-owned practice), public practices (e.g., government clinics, Federally Qualified Health Centers, Indian Health Service, Rural Health Clinics), corporate practices (e.g., hospital, health-system or HMO), academic practice (e.g., academic health center, faculty practice) or “other” (e.g., workplace clinic, not listed, multiple practice types selected). Physicians who reported working at academic practices in addition to other practice types were reclassified as academic practitioner in the analyses. Physicians also defined their practice location as suburban, urban and rural.

Statistical Analysis

Descriptive statistics were used to analyze physician demographic and practice characteristics and scores on CoBRAS and RACE. Bivariate analyses were used to evaluate the relationship between CoBRAS and RACE. Multivariable linear regressions were used to evaluate the relationship between CoBRAS and RACE, controlling for age (ref category: 60+ yrs), gender, race/ethnicity (ref category: non-Hispanic/Latinx White), Midwest training, race/ethnic composition of patient panel (ref category: <30% racial/ethnic minorities), practice location (ref category: urban) and practice type (ref category: corporate practice). Practice length was not included in the model as it was considered a similar metric as age. Both the overall CoBRAS score and each individual subscale were used in separate multivariable analyses. While most studies using CoBRAS have used the overall scale, the subscales have also been evaluated independently in prior research [15]. All analyses were conducted in SAS 9.4 (SAS Institute Inc., Cary, NC USA).

Results

Seventeen percent (267/1595) of the intended recipients completed the survey. Twenty-eight participants were excluded from the analysis due to “choosing not to answer” any question in the CoBRAS or RACE scale. As detailed in Table1, the majority of survey respondents were ≥40 years of age (62%), female (62%), Midwest trained (79%) and non-Hispanic/Latinx Whites (86%) (Table 1). Forty-nine percent of respondents practiced in an urban setting, 41% had patient panels with at least 30% racial and ethnic minorities, and 24% were in academic practices. Practice length was not included in the analysis due to collinearity with age.

Table 1:

Demographic and practice characteristics of survey participants

Characteristics	(N=239)
Age, %
20–29	12.1%
30–39	26.4%
40–49	21.3%
50–59,	18.4%
60+	21.8%
Gender, %
Female	62.3%
Male	37.2%
Nonbinary	0.4%
Race, %
White	87.0%
Asian	5.4%
Black	2.9%
Other race	4.6%
Hispanic/Latinx, %	0.8%
Midwest-trained, %	79.1%
Racial/ethnic minorities in patient panel, %
<30%	59.0%
30–69%	24.3%
>=70%	16.7%
Practice location, %
Urban	48.5%
Suburban	30.5%
Rural	20.9%
Practice type, %
Corporate	42.3%
Academic	23.8%
Private	10.9%
Government	10.5%
Other	12.6%

The mean RACE score was 16.4 (range 5–28) and the mean CoBRAS score was 44.1 (range 20–106) (Table 2). The mean scores for the CoBRAS subscales were: 15.8 (range 7–41) for unawareness of racial privilege, 18.2 (range 7–40) for institutional discrimination and 10.0 (range 6–31) for blatant racial issues.

Table 2:

The mean, SD and range of RACE and CoBRAS survey scores

Survey Scores (N=239)	Mean	SD		Range
RACE	16.4	5.1		5–28
CoBRAS	44.1	16.4		20–106
Racial Privilege	15.8	6.9		7–41
Institutional Discrimination	18.2	7.0		7–40
Blatant Racial Issues	10.0	4.3		6–31

In the unadjusted (β=0.11, 95% CI 0.07 to 0.14) and adjusted (β=0.05, 95% CI 0.01 to 0.09) models, CoBRAS score was positively associated with RACE (Table 3). When the CoBRAS subcale scores were examined, unawareness of institutional discrimination subscale was the only subscale positively associated with RACE in both the unadjusted (β=0.26, 95% CI 0.13 to 0.40) and adjusted (β=0.18, 95% CI 0.04 to 0.31) models (Table 4). Physicians under 40 years of age scored lower on RACE than physicians who were 40 years of age and older (Table 4). Physicians with patient panels with high percentage of racial/ethnic minorities scored lower on RACE than those with low percentage of racial/ethnic minorities (β= −2.17, 95% CI −4.09 to −0.25). Physicians working at clinics in rural locations (β=2.20, 95%CI 0.53 to 3.87) scored higher on RACE compared to those working at urban locations.

Table 3:

Unadjusted and adjusted associations (beta coefficient, 95% CI, p-value) between RACE and CoBRAS using a univariate and multivariable linear regression

Factors	Model I	P-value	Model II	P-value
CoBRAS *	0.11 (0.07,0.14)	<0.001	0.05 (0.01,0.09)	0.015
Age,				0.002 *
60+ (ref)
20–29			−2.96 (−5.17,-0.76)	0.009
30–39			−3.06 (−4.84,-1.28)	0.001
40–49			−0.25 (−2.03, 1.53)	0.781
50–59,			−0.81 (−2.65, 1.04)	0.389
Gender,				0.263*
Female (ref)
Male			−1.01 (−2.26, 0.23)	0.110
Nonbinary			1.27 (−7.57,10.11)	0.777
Non-Hispanic/Latinx White
No (ref)
Yes			−1.56 (−3.43, 0.31)	0.103
Midwest-trained
No (ref)
Yes			1.17 (−0.30, 2.65)	0.118
Racial/ethnic minorities in patient panel				0.052*
<30% (ref)
30–69%			−0.38 (−1.94, 1.18)	0.634
>=70%			−2.31 (−4.22,-0.39)	0.018
Practice location				0.038 *
Urban (ref)
Suburban			0.52 (−0.91,1.96)	0.475
Rural			2.16 (0.48, 3.83)	0.012
Practice type				0.240*
Corporate (ref)
Academic			−1.40 (−3.00, 0.20)	0.085
Private			1.09 (−0.84, 3.02)	0.266
Government			−0.47 (−2.52, 1.58)	0.653
Other			−0.51 (−2.39, 1.37)	0.597

^*Overall p-value for categorical variables with three or more levels

Table 4:

Unadjusted and adjusted associations (beta coefficient, 95% CI, p-value) between RACE and CoBRAS subscales using multivariable linear regressions

Factors	Model I	P-value	Model II	P-value
Racial Privilege	0.02 (−0.12,0.15)	0.828	−0.04 (−0.17, 0.10)	0.588
Institutional Discrimination	0.26 ( 0.13, 0.40)	<0.001	0.18 ( 0.04, 0.31)	0.011
Blatant Racial Issues	−0.01 (−0.23,0.21)	0.946	−0.02 (−0.23,0.19)	0.832
Age,				0.003 *
60+ (ref)
20–29			−2.73 (−4.94,-0.52)	0.016
30–39			−2.84 (−4.63,-1.05)	0.002
40–49			0.02 (−1.77, 1.82)	0.981
50–59,			−0.79 (−2.63, 1.04)	0.395
Gender,				0.181*
Female (ref)
Male			−1.18 (−2.43, 0.08)	0.066
Nonbinary			0.35 (−8.50, 9.21)	0.938
Non-Hispanic/Latinx White
No (ref)
Yes			−1.57 (−3.44, 0.30)	0.100
Midwest-trained
No (ref)
Yes			1.13 (−0.34, 2.60)	0.131
Racial/ethnic minorities in patient panel				0.081*
<30% (ref)
30–69%			−0.50 (−2.06, 1.05)	0.525
>=70%			−2.17 (−4.09,-0.25)	0.027
Practice location				0.032 *
Urban (ref)
Suburban			0.50 (−0.94, 1.94)	0.496
Rural			2.20 ( 0.53, 3.87)	0.010
Practice type				0.301*
Corporate (ref)
Academic			−1.28 (−2.87, 0.32)	0.117
Private			1.12 (−0.81, 3.04)	0.254
Government			−0.49 (−2.54, 1.55)	0.635
Other			−0.26 (−2.16, 1.63)	0.784

^*Overall p-value for categorical variables with three or more levels

Discussion

To our knowledge, this is the first quantitative study to examine the relationship between colorblindness and the self-reported use of race in medical decision-making. We hypothesized that, compared to their peers, physicians who embraced colorblindness may be less likely to think critically about the impact of racism on health, and were therefore more likely to use race to guide medical treatment. Our findings show that physicians who endorse a colorblind racial ideology, particularly those who are unaware of institutional racism, are more likely to use race in their screening and treatment decisions. We also found that older, rural practitioners, and those who served a less diverse patient population were more likely to endorse the use of race in medical decision-making.

Previous studies that used RACE all relied on a 2010 nationwide survey of internal medicine physicians (N=787) [11, 16–18]. The average RACE score for internal medicine physicians was 13.5. In comparison, the mean RACE score for Minnesota family medicine physicians was 16.4, suggestive of an increased use of race in medical decision making. Internal medicine and family medicine physicians receive the same medical school education but differ in their scope of training—with family medicine physicians providing pediatric and obstetric care in addition to adult medicine. The difference in RACE score may be due to scope of training if there is a greater use of race-based medicine in obstetrics and pediatric care. However, this is unlikely. The score difference could also be due to regional differences in the use of race in medical schools and residencies, as the study of internal medicine physicians represented medical providers throughout the country and may not be a fair comparison to our sample of Minnesota family physicians. Though our study did not find a relationship between region of training and the use of race, it is still possible that medical training programs in the Midwest use race at higher rates than the rest of the country. Given that the majority of our study participants trained in the Midwest, and that our sample size was relatively small, our study may not have been sufficiently powered to observe such a distinction.

Like most other studies, we found that younger physicians were less likely to use race and we did not find a relationship between provider race and their use of race in medical decision-making [11, 19, 20]. Two studies posited that minority physicians were more likely to use race than their White peers [18, 21] when choosing medications [21] and when considering social determinants that impact their patients’ health [21]. The low number of non-White physicians in our study may have adversely impacted our ability to assess the impact of provider race. That said, it may be that ways in which physicians think about race are more powerfully influenced by other factors, such as socioeconomic background, undergraduate major, or professional socialization.

Our finding that greater racial/ethnic diversity in patient panels was negatively associated with the use of race to guide medical treatment differs from a previous study showing that higher numbers of minority patients was associated with a greater use of race in medical decision making [11]. This could be due to differences in how minority patients were categorized. In our study, we created three groupings for minority patients, comparing providers with the largest and smallest percentage of minority patients. Other studies used two categories, comparing providers with a small percentage of minority patients to all other providers. We also found that physicians working in urban areas are less likely to use race, which disagrees with another study that found that physicians working in urban areas were more likely to value race in clinical decision-making [22]. It is possible that physicians working in urban areas—which have histories of redlining, white flight, ghettoization of minority communities—may be more attuned to the impact of systemic racism on health inequalities.

Although the CoBRAS has been used extensively in studies assessing colorblindness, particularly in psychology, there are no published studies that use this measure to assess colorblindness in physicians. CoBRAS has been used to evaluate colorblindness in dental faculty at a Florida University [15]. Dentistry is a similar field to medicine, as both fields are in healthcare, require a four-year post-undergraduate degree, require similar prerequisites for entry into their respective training programs, and have residency programs that allow for specialization. Their survey of 48 dental faculty who did not identify as underrepresented minorities had average scores of: 23 for racial privilege, 17 for institutional discrimination, and 20 for blatant racial issues. These scores were noticeably higher than our CoBRAS scores. One possible explanation for this difference is that the dental study was conducted several years prior (2015–2016) to our own (2019) and in the intervening years there has been greater awareness in the general public of the impact of racism on the lives of racial minorities. Also, it is possible that in states with policies that emphasize collective, compared to individual, responsibility there may be greater awareness of how system-wide policies and practices impact individuals, including racial minorities. Last, there may be greater social desirability in our study population, possibly due to recent increased emphasis on the importance of addressing racism, resulting in lower colorblindness scores.

There are many reasons why physicians may rely on race to guide treatment. While physicians have the autonomy to decide the extent to which they use race in their medical practice, race is often built into medical guidelines and calculators. Physicians are trained to practice evidence-based medicine, which may result in their reliance on race-based algorithms. However, a systematic review evaluating physician overall use of clinical guidelines found that disagreement with the guideline was a common reason for lack of use [23], and studies show that physician use of guidelines is low [24]. Thus, racial beliefs, including beliefs about the biological validity of race, may impact a physician’s use of race-based guidelines. Nevertheless, clinical encounters are brief, and physicians may not have the additional time, or mental energy to consider alternatives to using race in medical calculators. Physicians who value certainty may also feel more comfortable using “trusted” guidelines and “proven” findings. One study found a positive relationship between anxiety due to uncertainty and the use of race in medical decision-making [17]. Lastly, physicians may believe that race-based medical care reduces racial health disparities, because they believe race-specific medications will work better in racial minorities.

The strengths of this study include the use of validated measures to assess both colorblindness and the use of race in medicine. This is the first quantitative assessment of physician colorblindness and the impact on RACE among family medicine physicians. While colorblindness has been assessed in physicians in one qualitative study, this is the first study to show a relationship between a colorblind racial ideology and the use of race in medical decision making.

There were several limitations to this study. First, our low response rate may have reduced our ability to detect relationships between provider characteristics and the self-reported use of race and limit generalizability. It is worth noting that although our sample had a high percent of non-Hispanic/Latinx White physicians (86%), it somewhat reflective of the racial demographics of Minnesota physicians (76% non-Hispanic/Latinx White) [25]. Second, the CoBRAS scale was developed 20 years ago. The statements used to assess colorblindness may underestimate the extent of colorblind beliefs given changing cultural norms. Further, social desirability may have affected results, with respondents underreporting their level of colorblindness. Third, we could be missing relevant confounders or effect modifiers in our model that impact the relationship between CoBRAS and RACE. Last, this is a survey about beliefs and direct behavior were not observed. The RACE scale does not address what is considered a “high” use of race and previous studies have not looked at physician behavior directly.

Conclusions

Adherence to a colorblind racial ideology, particularly unawareness of institutional discrimination, is associated with an increased use of race in medical decision-making. Experiencing racism has been shown to be positively associated with many health conditions such as depression, anxiety, decreased subjective cognitive function, and hypertension [26–28]. Colorblindness ignores the impact of racism on health and points to other explanations for racial disparities. For example, if race were biological, the higher rates of chronic disease in Black communities could be partially attributable to unknown genetic factor specific to the Black race, although Black people are not biologically real. Maintaining these beliefs hinder our investigation of societal factors that are responsible for racial disparities, and, therefore, allows them to persist.

Colorblindness may be due to a poor understanding of how systemic racism affects health. Medical educators need to teach trainees about racism as a health risk to increase awareness of the role racism plays in determining health outcomes. Additional research is needed to evaluate other racial beliefs of physicians, as colorblindness clearly affects physician use of race in medical decision-making. Future research should include a broader sample of physicians, including family physicians living in other geographic regions and clinicians practicing in other specialties. Moreover, further investigation as to how medical researchers interpret the relevance of race as a covariate in their analyses is needed, as the use of race in clinical research results in literature supporting race-based practices, contributing to physician use of race.