Impact of a USMLE Step 2 Prediction Model on Medical Student Motivations

Abstract

PURPOSE: With the transition of USMLE Step 1 to Pass/Fail, Step 2 CK carries added weight in the residency selection process. Our goal was to develop a Step 2 predicted score to provide to students earlier in medical school to assist with career mentoring. We also sought to understand how the predicted scores affected student’s plans. METHOD: Traditional statistical models and machine learning algorithms to identify predictors of Step 2 CK performance were utilized. Predicted scores were provided to all students in the Class of 2024 at a large allopathic medical school. A cross-sectional survey was conducted to assess if the estimated score in uenced career or study plans. RESULTS: The independent variables that resulted in the most predictive model included CBSE score, Organ System course exam scores and Phase 2 (Third Year Clinical Clerkships) NBME percentile scores (Step2CK= 191.984 + 0.42 (CBSE score) + 0.294 (Organ Systems) + 0.409 (Average NBME). The standard error of the prediction model was 7.6 with better accuracy for predicted scores greater than 230 (SE 8.1) as compared to less than 230 (SE 12.8). Nineteen percent of respondents changed their study plan based on the predicted score result. Themes identified from the predicted score included reassurance for career planning and the creation of anxiety and stress. CONCLUSION: A Step 2 Predicted Score, created from pre-existing metrics, was a good estimator of Step 2 CK performance. Given the timing of Step 2 CK, a predicted score would be a useful tool to counsel students during the specialty and residency selection process.