Explicit and implicit bias based on race, ethnicity, and weight among pediatric emergency physicians

Abstract

Racial, ethnic, and weight disparities in patient management exist overall and in the emergency department (ED).1 Provider's implicit bias—unconscious negative attitudes toward specific social groups—contribute as a driver of these disparities.2 Although studies have shown biases against Black and obese individuals among physicians,3, 4 the magnitude of those biases varies among specialties.3 No known studies have investigated the coexistence of implicit biases related to race, ethnicity, and weight among physicians. Strategies to change biases exist.5 Awareness of the presence and the magnitude of one's own biases is a first step toward reducing the effect clinicians’ biases have on the provision of care.6 Our objectives were (1) to assess explicit and implicit racial, ethnic, and weight biases in a sample of pediatric emergency physicians and (2) to investigate whether racial, ethnic, and weight biases correlate with each other and the physicians’ demographics. We developed an anonymous Web-based study for all pediatric emergency physicians (attendees and third-year fellows) who worked in a large urban pediatric tertiary care ED. The physicians were invited to participate via email from September to December 2018. The Web-based study included three sequential steps: first, each participant filled out key demographic items. Then, they were requested to complete three implicit bias measures and finally three explicit bias measures. The study was implemented through Project Implicit and approved by the Boston Children's Hospital Institutional Review Board. We collected physicians’ age, sex, race/ethnicity, and height and weight to calculate body mass index (BMI). BMI was classified as underweight (BMI < 18.5), normal weight (BMI 18.5–24.9), overweight (BMI 25.0–29.9), and obese (BMI ≥ 30.0). Implicit bias based on race (White vs. Black race), ethnicity (Hispanic vs. non-Hispanic White), and weight (fat vs. thin) was assessed by three Implicit Association Test (IATs). The IAT is a well-established measurement of implicit bias. It measures relative association strengths between two pairs of concepts. Details about IAT procedures, scoring, and psychometric properties have been published.7 Stimuli representing four categories are presented one at a time in the center of the computer screen. Participants categorize these stimuli under two sorting conditions. The difference in the average response time between the two conditions is an indicator of the relative association strength or bias toward one group relative to the other. For example, faster categorization when asked to categorize White people with good words (and Black people with bad words) compared to the reverse indicated an implicit bias against Black people. The order of these conditions is randomized across participants. The IAT effect for each test was calculated according to the improved scoring algorithm.7 Scores ranged from –2 to +2, with 0 indicating no implicit bias. Positive scores indicated an implicit bias against Black people, Hispanic people, or fat people. Explicit bias was measured by three self-reported items. Physicians reported their attitudes about Black, Hispanic, and fat people by answering a 7-point Likert scale for each social group ranging from –3 (I strongly prefer Black/Hispanic/fat people to White/thin people) to +3 (I strongly prefer White/thin people to Black/Hispanic/fat people) with zero indicating no preference (I like Black/Hispanic/fat people and White/thin people equally). Descriptive statistics characterized physicians’ demographics. One-sample t-tests determined whether the mean explicit and implicit scores were different from zero. To interpret the strength of the biases, Cohen's d, was calculated for each of the explicit and implicit measures. As suggested, d of 0.2 to <0.5 was interpreted as small effect, d of 0.5 to <0.8 as medium effect, and d of ≥0.8 as strong effect. Pearson's correlation coefficient assessed the relationship between implicit and explicit measures and between each measure. Mann-Whitney U-test and Spearman's correlations assessed the association of physician's demographics with explicit and implicit measures. Participants were able to quit the study at any time; some only filled out the demographics items, some performed one or two implicit measures and some performed the three IATs, and some or all the explicit measures. Data were included in the analyses if at least one IAT was completed. Of the 143 physicians contacted, 101 performed at least one IAT (70.6%) and were included: 88 completed three IATs, six completed two IATs, and seven completed one IAT. Finally, 82 completed the three IATs and the three explicit measures (57.3%). A total of 70% were female with a median age of 40 years (interquartile range [IQR] = 36–49, range = 28–66). A total of 86% were White non-Hispanic, 12% Asian, and 2% Hispanic. Seven percent were underweight, 60% normal weight, 23% overweight, and 9% obese. On average, the explicit measures demonstrated a weak preference for Black over White people (mean = –0.44 [95% CI = –0.67 to –0.21]; Cohen's d = 0.43, p < 0.001, n = 84), a moderate preference for thin over fat people (mean = 0.67 [95% CI = 0.49 to 0.86]; Cohen's d = 0.78, p < 0.001, n = 83), and no preferences between Hispanic and White people (mean = 0.05 [95% CI = –0.07 to 0.17]; Cohen's d = 0.09, p = 0.4, n = 83). On average, the implicit measures demonstrated a strong implicit bias against Black people (mean = 0.33 [95% CI = –0.26 to 0.41]; Cohen's d = 0.92, p < 0.001, n = 97), Hispanic people (mean = 0.34 [95% CI = –0.26 to 0.41]; Cohen's d = 0.94, p < 0.001, n = 90), and fat people (mean = 0.41 [95% CI = –0.33 to 0.48]; Cohen's d = 1.14, p < 0.001, n = 96). The correlations between the implicit and the explicit measures are shown in Table 1. None of the implicit measures were correlated with any explicit measures. Significant correlations were found between all implicit measures (racial and ethnic implicit biases, r = 0.48, p < 0.001, n = 88; racial and weight implicit biases, r = 0.33, p < 0.01, n = 94; ethnic and weight implicit biases, r = 0.31, p < 0.01, n = 88). Overweight or obese physicians had a weaker implicit bias against Hispanic people than physicians with normal or underweight BMI (mean = 0.21 vs. mean = 0.39, respectively, p < 0.05, n = 87) but a stronger explicit preference for White people compared to Hispanic people (mean = 0.24 vs. mean = –0.04, respectively, p < 0.05, n = 81). Female physicians showed a stronger explicit preference for Blacks compared to male physicians (mean = –0.58 vs. mean = –0.15, respectively, p < 0.05, n = 84). Other associations between physicians’ age, sex, race/ethnicity, or weight status and explicit or implicit measures were not significant. Regardless of physician age, sex, race/ethnicity or BMI, strong implicit biases based on race, ethnicity, and weight exist among pediatric emergency physicians in this study. These implicit biases are potentially important to recognize and address recognizing that high-pressure situations may activate biases that negatively impact care.8 Our findings are consistent with previous studies demonstrating bias against Black patients among ED physicians.9 We add to the literature by demonstrating the presence of strong bias against Hispanic and obese individuals as well. To the best of our knowledge, this is the first study investigating those additional biases against minority patients in the pediatric ED.2 Our study is congruent with other studies that demonstrated an implicit bias against Hispanic patients across primary care physicians10 and against obese people among physicians overall.4 Because biases and their magnitude may vary across specialties,3 bringing to light the existence of these biases among pediatric emergency physicians is critical to increase awareness of this issue. EDs can use this information to critically evaluate the role of biases in health inequities and guide efforts to address negative effects of bias on patient care. Implicit biases against different social groups are rooted in systems of oppression (e.g., racism, sexism, fat-shaming, etc.) and the normalization of negative social stereotypes against marginalized communities. These stereotypes, attitudes, and beliefs underlie implicit biases that can affect physician's ability to provide equitable care; multiple studies demonstrate that implicit biases are associated with poor patient–provider communication and lower quality of care among marginalized groups.2 As implicit biases are more likely to favor advantaged groups, their negative effects on patient care manifests as structural inequities in health care quality experienced by marginalized groups. The correlation between implicit biases suggests that physicians with one implicit bias have a higher likelihood of bias against other marginalized social groups. Our finding that biases based on race, ethnicity, and weight are correlated suggests an underlying cognitive process on the individual level that may make an individual more susceptible to implicit biases against marginalized groups. This is reflected in the concept of essentialism, the cognitive process of believing that categorization in the world is a reflection of natural order, which has been linked to prejudice and rationalizes the marginalization of different social groups.11 Finally, we found no correlations between explicit measures and implicit biases. The discrepancy between explicit and implicit bias has been shown previously and is one of the main reason to assess implicit biases even if one reports no conscious/explicit biases.12 This study has the following limitations: First, this is a single institution study with a limited sample size, which limits generalizability. However, the data should encourage other EDs to consider implicit biases self-screening to increase awareness of personal implicit bias. Second, the lack of diversity among our sample restricts our ability to assess the association between physician race/ethnicity and implicit biases. Third, we restricted our sample to physicians in a highly dynamic environment where care is provided by nurses, clinical assistants, and others whose explicit preferences and implicit biases were not captured and may affect clinical care. In conclusion, we found that pediatric emergency physicians in our sample displayed implicit bias against Black, Hispanic, and obese people. Moreover, implicit biases were correlated, suggesting that some physicians may be at risk of multiple biases. Efforts to screen for implicit biases among ED providers, assess their consequences on patient care, and test interventions to decrease them are required. The authors have no potential conflicts to disclose. Romain Guedj, Maddalena Marini, Joe Kossowsky, Charles Berde, Eric W. and Fleegler conceptualized and designed the study. Romain Guedj, Maddalena Marini, and Eric W. Fleegler built the survey. Camila Mateo, Maddalena Marini, and Romain Guedj performed the statistical analyses. Romain Guedj drafted the initial manuscript, and all authors (Joe Kossowsky, Charles Berde, Camila Mateo, Eric W. Fleegler, Maddalena Marini) contributed substantially to its revision. Romain Guedj takes responsibility for the paper as a whole.

Topics & Concepts