Sex and racial bias in medical student EPA assessments: Findings and hypotheses for bias mitigation targets

Introduction: Despite extensive documentation of implicit bias across traditional medical education assessments, no studies have examined bias in Entrustable Professional Activity (EPA)-based assessments of medical students. This gap threatens both assessment validity and advancement of diverse trainees, as even small inequities can significantly impact career trajectories for underrepresented students. As such, we sought to detect implicit sex and racial bias in EPA-based assessments of medical students. Methods: We examined 17,061 EPA-based assessments of 279 third year medical students for evidence of sex or racial bias. We examined EPA assessment comments for an association between race/sex and comment length, sentiment, or the use of keywords associated with implicit bias. We also tested for an association between race/sex and entrustment scale rating. Additionally, three authors qualitatively coded 6% (n = 1069) of narrative comments for ‘feedback value’ and tested for an association with race/sex. Results: There was no significant difference in comment length, feedback value, sentiment, or use of keywords previously associated with bias by student sex/race. Students who identify as white were more likely to receive a higher entrustment rating than non-whites, but no sex differences were noted in entrustment ratings. Discussion: Although we did not detect sex or racial bias in EPA-based narrative assessments, we did find evidence of racial bias in entrustment scale ratings. Following these results, we consider whether EPA-based assessments are less prone to implicit bias and identify targets for future research seeking to mitigate implicit bias.