Our analysis reveals race-, ethnicity-, and migration-based disease associations in the UWorld QBank. These associations have potential for systematic harm, including promoting inaccurate race-based associations and upholding cultural conventions of White bodies as normative. Here, we discuss the implications of three key findings: (1) the low frequency yet unequal distribution of race, ethnicity, and immigration status; (2) the associations with disease conditions in minoritized racial and ethnic groups; and (3) the misrepresentation of immigration status as a risk factor for disease.
Low Frequency yet Unequal Distribution of Mentions of Race, Ethnicity, and Immigration Status
Mentions of race, ethnicity, and immigration status occurred infrequently, representing only 3.5% of the questions we examined. This contrasts with prior research analyzing the UWorld Step 1 QBank, which found that mentions of race and ethnicity occurred in 20.6% of questions.13 This finding may reflect differences between USMLE Step 1 and Step 2 CK or conscious efforts to remove mentions of race and ethnicity from the exam. In our analysis, non-White race and specific ethnic identifiers (e.g., Mediterranean, Ashkenazi Jewish, Middle Eastern) occurred much more frequently than mentions of White race. The invisibility of race and ethnicity (and immigration status) risks promoting color-blind racism and White supremacy due to hegemonic societal conventions of White, cisgender bodies as unmarked.14,15 In other words, to describe a 45-year-old woman with fatigue and muscle weakness is to imagine a 45-year-old White, cisgender woman with these symptoms. We therefore discourage the removal of race and ethnicity from the QBank and instead emphasize the need for racial and ethnic categories to remain descriptive rather than to suggest disease risk or guide management.
Mentions of race/ethnicity varied significantly by UWorld system, indicating that the embeddedness of race-based associations may differ among specialties. Recent discussions of race-based medicine describe harmful practices in cardiology, nephrology, obstetrics, and oncology.3,4 This suggests that race-based associations with disease conditions and management may be more pervasive in certain specialties relative to others. Accordingly, interventions to reform the use of race, ethnicity, and immigration status in UWorld may yield greater benefit if targeted toward test question contributors from specific specialties. Rather than using race, ethnicity, or immigration status as proxy for genetics, healthcare status, infectious disease exposures, or behavior, individualized details (e.g., family history, immunization status, diet) should be included by specialists and question authors to help students develop a more inclusive and holistic approach to medical care.
Associations with Disease Conditions in Minoritized Racial and Ethnic Groups
White race had nearly 10 times greater odds of occurring in the question stem only relative to presence in both the question stem and answer explanation. This indicates that race-based associations with disease are less likely to be made with patients of White race. Our qualitative analysis echoes this finding: mentions of White race without clear rationale or link to the answer choices or explanation emerged 16 times.
By contrast, Asian race had 6.4 times greater odds of occurring in the answer explanation only relative to the question stem only. Although the odds ratios for Black and Latinx were not significant, our qualitative analyses suggest that race-based associations are evident across minoritized racial groups. These included epidemiologic associations and descriptions of increased disease risk within racial/ethnic/immigrant groups (Table 2). By presenting these associations without socio-structural context, these diseases are implied to be biological in nature. This is further codified through geneticization, or the attribution of “social, behavioral, or physiological problems” to genetic variation by race.16 These models build on descriptive analyses in the UWorld Step 1 QBank,13 quantifying the tight associations between minoritized racial groups and disease conditions.
Although recent, individual ancestry may be genetically meaningful when traced to a narrowly circumscribed geographic population of origin, clinicians and researchers alike frequently elide the concepts of ancestry and race through continental descriptors (e.g., “African ancestry”) that lack precision.17,18 Mentions of “ancestry” in our analysis commonly corresponded to broad geographic areas (e.g., “Asian,” “Northern European”). The persistence of race and continental ancestry as ready substitutes for more explicit bio- or genetic markers of disease demonstrates how assumptions about the intrinsic, biological nature of race remain unchecked in biomedicine.
Our analyses of the “race-based management” and “distraction” codes demonstrate how race-based associations contribute to racially tailored care and may even potentially lead to harm. In one question, a Black patient with histoplasmosis is initially suspected of sarcoidosis and mismanaged with high-dose corticosteroids, which exacerbates his condition. Here, the test authors tacitly underscore the risks of race-based associations, noting how inappropriately narrow differential diagnoses can lead to iatrogenic harm.
Misrepresentation of Immigration Status as a Risk Factor for Disease
Seven answer explanations involving immigration status described endemic regions beyond continental boundaries, often including large areas of the globe outside of Europe and North America. The term “developing country” occurred in six answer explanations as an indication of risk for conditions such as rheumatic fever and echinococcus. Infectious disease risk varies within regions, countries, and across continents. Although “developing country” may imply conditions of limited public health infrastructure, these vary widely by geographic and socioeconomic position and cannot be inferred by an individual’s country or continent of origin. This may promote harmful biases against immigrants from low- and middle-income countries, particularly those racialized as Black or Brown.
Among references to immigration status, nearly half discussed migration from areas in which specific infectious diseases are endemic. Many immigrant source countries face high burdens of infectious disease, largely relating to environmental and socioeconomic conditions.19,20 The Global Fund for tuberculosis, HIV, and malaria highlights this: these diseases are largely due to limitations of public health infrastructure and subtropical environments.21 In these cases, migration history may serve as an important indicator of risk.
However, our analysis of the “distraction” code suggests that immigration status was often used to imply disease risk in a misleading way. This exemplifies a key pitfall of conflating immigration status with ethnicity,22 and of using ethnicity (or race) as a proxy for genetic background. Migration from a malaria-endemic region may suggest an individual has an increased likelihood of carrying traits for sickle cell disease or thalassemia; however, careful family history should inform pre-test probability rather than migration history or ethnicity alone.
A robust literature exists on migration as a social determinant of health.23,24,25,26 This research examines the effects of healthcare policy, authorization status, and language barriers on immigrant health. Across these structural dimensions, immigration status is a real predictor of health outcomes and should affect management decisions; however, our analysis did not reflect this complexity, instead linking immigrants with disease and adverse health beliefs and behaviors. QBanks and multiple choice examinations may not be a fitting modality through which to teach and assess structural competency or respond to the institutions and social conditions that affect health.27
Our analysis suggests that immigration status was often used to inappropriately suggest inadequate health maintenance, behavior, culture, and ethnicity; in these instances, appropriate management involved vaccination, nutritional supplementation, or antibiotic treatment. Assumption that immigrants, solely due to their immigration status, have not received routine health maintenance or are not up-to-date with immunizations may contribute to harmful stereotypes of immigrants as “victims” or as being unfamiliar with healthcare routines.28,29 Immigration status cannot be used as a proxy for behavior: culture is not static, nations are pluralistic, and individuals are intersectional.30,31 For instance, information about high intake of corn or nitroso compounds could be obtained through a careful dietary history of a patient suspected of niacin deficiency or gastric cancer, respectively.
Our study carries limitations. First, we limited our analysis to one study tool for the USMLE Step 2. We recognize that medical students use a variety of study materials and future research might analyze the use of race, ethnicity, and immigration status in other such tools. Second, we were only able to examine the questions available during the capture period of May 28 through August 11, 2020. Repeated analysis as questions update may yield new findings. Third, copyright restrictions prevented direct use of QBank quotes, limiting the specificity of our qualitative analysis. Fourth, although our mixed-methods analyses allow us to draw inferences regarding the function of mentions of race and ethnicity, we are unable to speak to the motivations of the QBank developers for including race, ethnicity, and immigration status. Finally, our analysis of this QBank does not represent the USMLE Step 2 CK exam, although the QBank claims to closely approximate the USMLE.9 Despite these limitations, we provide a robust, systematic analysis of the use of race and ethnicity in the UWorld Step 2 QBank.
Our study identifies several problematic associations between racial, ethnic, and immigration categories and disease conditions in the UWorld Step 2 CK QBank. We emphasize the need to account for individual patient details (e.g., family history, immunization status, diet) rather than using race, ethnicity, or immigration status as proxy for genetics, healthcare status, infectious disease exposures, or behavior. Reform of race-based medicine in medical licensing educational resources—and in medical education more broadly—can contribute to health equity.