PREDICTING USMLE 1 PERFORMANCE: CORRELATION OF INTERNAL AND EXTERNAL FACTORS WITH STEP 1 SCORES

Carol F. Whitfield* and Richard J. Simons, Penn State College of Medicine, Hershey, PA, U.S.A.

Purpose
The purpose of this study was to determine which of the available external and internal assessments of student performance would best predict performance on USMLE 1 on first try.

Methods
Data including MCAT, College GPA (CGPA), and internal course grades were used in the analysis.  Data from the classes of 2007 (N=125) and 2008 (N=135) were analyzed by multiple regression and other standard methods. 

Results

MCAT and CGPA were somewhat predictive of USMLE 1 scores. R² was 0.190 and 0.158 for MCAT vs USMLE 1 for the class of 2007 and 2008 respectively, and a model containing MCAT and GPA produced R² of 0.203 and 0.253 respectively.  The best predictors were the average course grade in medical courses (R²  of 0.662 and 0.795) over 2 years, and the Comprehensive Basic Science Exam (CBSE) (R2 of 0.568 and 0.634).  The R² for average course grade in year 1 was 0.568 and 0.463.  A model including average course grade over 2 years, CBSE, Med GPA for year 1, college GPA and PS section of MCAT, developed from 2007 data predicted 79% of 2008 USMLE 1 failures. A model developed for 2008 predicted 86% of failures for 2008. The at-risk population had an average course grade below 75% for years 1 plus 2. 

Conclusions

The best predictor of student success in USMLE 1 is our internal course grades.  Students in the at-risk group will be advised to use extra study time and/or take a preparatory course before taking USMLE 1.