Early Prediction of Medical Student Performance on Initial Licensing Examinations

Abstract

Many medical schools can successfully predict overall performance on the initial licensing examination by the end of the second year. Unfortunately, the time left to remediate students before students take exams is short. The goal of this project was to determine the feasibility of predicting licensing examination scores at the end of the first year of training. Data from three cohorts of medical students, each made up of about 230 individuals from the graduating classes of 2013 to 2015 were used. A series of prediction equations for licensing examination scores, based on class performance, was developed from the first cohort of students. The best predictive equation was applied to the second cohort of students to generate a predicted examination score which was then compared to the actual examination score. This approach was cross-validated with another cohort of students. The R-values for the selected predictive equations ranged from 0.691 to 0.814, accounting from about 48 to 66 % of the variance in actual scores. The mean difference between predicted and actual examination scores was 19 points on a three-digit scale ranging from 200 to 800 for the second cohort and 25 points for the third cohort. The data show that initial licensing examination scores can be successfully predicted 1 year before students take the exam. Success at this task would give medical schools the ability to utilize their available resources efficiently to provide additional assistance to those students in need.