Presented By: William Alley, Wake Forest School of Medicine
Co-Authors: Janet Tooze, Wake Forest School of Medicine
Andrea Vallevand, Wake Forest School of Medicine
Catherine Wares, Atrium Health Carolinas Medical Center

Purpose
Unstructured oral examinations (OE) have been criticized for lack of reliability and the potential for bias, but structured OEs may be valuable tools for evaluating clinical skills and reasoning and mitigate subjectivity and unintended bias towards URM students. We aim to retrospectively investigate Emergency Medicine (EM) OE scores between students self-categorized by gender as well as Underrepresented Minority (URM), Asian Pacific Islander (API), or White for evidence of unintended bias.

Methods
Multiple clinical cases for two common chief complaints were developed by board-certified EM faculty. Each case is highly structured and standardized, including the final diagnosis, time allotted, and critical components. Faculty assessors were oriented to the cases and assessment tool. Every EM clerkship student is required to take the OE, which consists of two 15-minute cases. Scores converted to a 0.0-4.0 scale. OE data were analyzed using factorial analysis of covariance and Levene's test (independent variables: race and gender; covariates: Step 1 score and rotation number).

Results
Data from seven cohorts were analyzed (n = 806: range 96 to 135/cohort). Self-reported race and gender demographics: API = 164(20.3%); URM = 131(16.3%); White = 511 (63.4%); Male = 396(49%); Female = 411(51%). Step 1 performance was significantly related to OE scores on six of seven analyses (all p ? .03), EM rotation timing on two of seven analyses (p = .002 and .035). There were no statistically significant differences for the main effect of race. The main effect of gender was significant for one cohort (p ? .011, estimated mean: female 2.88 and male 2.72).

Conclusion
Integrating structured, standardized oral examinations can provide a valuable method for assessing EM clerkship students without introducing unintended bias.