USMLE Scores Do Not Predict the Clinical Performance of Emergency Medicine Residents

Abstract

Background: Scores on “high-stakes” multiple choice exams such as the United States Medical Licensing Examination® (USMLE) are important screening and applicant ranking criteria used by residencies.Objective: We tested the hypothesis that USMLE scores do not predict overall clinical performance of emergency medicine (EM) residents.Methods: All graduates from our University-based EM residency between the years 2008 and 2015 were included. Residents who had incomplete USMLE records were terminated, transferred out of the program, or did not graduate within this timeframe were excluded from the analysis. Clinical performance was defined as a gestalt of the residency program’s leadership and was classified into three sets: top, average, and lowest clinical performer. Dissimilarities of the initial blind rankings were adjudicated during a consensus conference.Results: During the eight years of the study period, there were a total of 115 graduating residents: 73 men (63%) and 42 women. Nearly all of them (109; 95%) had allopathic medical degrees; the remainder had osteopathic degrees. There was not a statistically significant correlation between our ranking of clinical performance and the Step 2 Clinical Knowledge score. There was a non-significant correlation between clinical performance and the Step 1 score.Conclusion: Neither USMLE Step 1 nor Step 2 Clinical Knowledge were good predictors of the actual clinical performance of residents during their training. We feel that their scores are overemphasized in the resident selec­tion process.