Abstract Title: Investigating Potential Bias in Medical School Assessments

Background: Assessments are critical in medical education, shaping career paths from entry to progression. Despite efforts for equity, disparities persist. Underrepresented minority groups and women score lower on the MCAT and USMLE, with differences by gender and ethnicity persisting. Biases extend to narrative assessments like letters of recommendation and MSPEs, where non-white students receive fewer strong descriptors, and gendered terms characterize female students. This study investigates how gender, race, language proficiency, and immigration status influence test outcomes, aiming to inform fair assessment practices.

Methods: 54 first-year medical students completed a practice test and demographic survey at the University of Central Florida College of Medicine. Test questions underwent Angoff scoring and 12-item writing flaw rubric evaluation. Students received immediate feedback and qualitative analysis of incorrect responses. Quantitative analysis included Mantel-Haenszel chisquare tests and two-sample tests for differential item functioning. Multivariate logistic regression assessed demographic predictors of test outcomes, combining quantitative and qualitative approaches.

Results: Logistic regression showed gender alone did not predict question outcomes, but its interaction with overall performance did. Non-uniform differential item functioning was observed, confirmed by significant Chi-square results. Limited study participation precluded separate race analysis, affecting statistical power. The Nagelkerke measure quantified variation explained by DIF. Ongoing analysis will examine group performance trends to identify broader disparities.

Conclusion: This study identifies nuanced biases in medical school assessments, particularly regarding gender. While logistic regression didn't directly predict gender's impact on question outcomes, its interaction with performance suggested predictive value. Significant Chi-square results confirmed differential item functioning, with ongoing analysis aiming to reveal broader performance trends among demographic groups. This insight informs efforts to foster fairness in medical education assessments.