Race, Ethnicity, and Immigration Status in a Medical Licensing Educational Resource: a Systematic, Mixed-Methods Analysis

Abstract

Background Medical students preparing for the United States Medical Licensing Exam (USMLE) Step 2 Clinical Knowledge (CK) Exam frequently use the UWorld Step 2 CK Question Bank (QBank). Over 90% of medical students use UWorld QBanks to prepare for at least one USMLE. Although several questions in the QBank mention race, ethnicity, or immigration status, their contributions to the QBank remain underexamined. Objective We conducted a systematic, mixed-methods content analysis to assess whether and how disease conditions might be racialized throughout this popular medical education resource. Design We screened 3537 questions in the QBank between May 28 and August 11, 2020, for mentions of race, ethnicity, or immigration status. We performed multinomial logistic regression to assess the likelihood of each racial/ethnic category occurring in either the question stem, answer explanation, or both. We used an inductive technique for codebook development and determined code frequencies. Main Measures We reviewed the frequency and distribution of race or ethnicity in question stems, answer choices, and answer explanations; assessed associations between disease conditions and racial and ethnic categories; and identified whether and how these associations correspond to race-, ethnicity-, or migration-based care. Results References to Black race occurred most frequently, followed by Asian, White, and Latinx groups. Mentions of race/ethnicity varied significantly by location in the question: Asian race had 6.40 times greater odds of occurring in the answer explanation only (95% CI 1.19–34.49; p < 0.031) and White race had 9.88 times greater odds of occurring only in the question stem (95% CI 2.56–38.08; p < 0.001). Qualitative analyses suggest frequent associations between disease conditions and racial, ethnic, and immigration categories, which often carry implicit or explicit biological and genetic explanations. Conclusions Our analysis reveals patterns of race-based disease associations that have potential for systematic harm, including promoting incorrect race-based associations and upholding cultural conventions of White bodies as normative.

Academic Surgery in the Era of Large Language Models

“…A 2022 study analyzed the resources used to prepare for the USMLE Step 2 Clinical Knowledge examination to see if and how diseases may be racialized. 52 The analysis demonstrated patterns of race-based disease associations with potential for bias, promoting false associations, and upholding cultural conventions as normative. Additionally, LLMs run the risk of generating AI hallucinations, which are similar to neuropsychiatric confabulations, where the output text is delivered with confidence but may be an inaccurate synthesis leading to false information.…”

“…An example can come from medical education and the resources used to study and evaluate learners. A 2022 study analyzed the resources used to prepare for the USMLE Step 2 Clinical Knowledge examination to see if and how diseases may be racialized . The analysis demonstrated patterns of race-based disease associations with potential for bias, promoting false associations, and upholding cultural conventions as normative.…”

Race-based medicine in the point-of-care clinical resource UpToDate: A systematic content analysis

“… 2 , 3 , 4 Biological, racial ‘types’ emerged during European colonization as a tool to divide and control populations worldwide. 2 Such embedded notions of biologized race pervade pre-clinical and clinical training, 4 , 5 , 6 medical licensing examinations, 7 , 8 , 9 and clinical guidelines. 2 , 3 , 10 , 11 Such racial essentialism may exacerbate racialized health inequities.…”