A predictive model for the USMLE Step 1 was created based on NBME Comprehensive Basic Sciences Self-Assessment (CBSSA) exams. All second year medical students at a southwestern university from 2014-2016 took a NBME CBSSA exam under controlled testing conditions six months (January) and three months (April) prior to their Step 1. A multiple regression analysis using practice scores and medical knowledge grades as predictive variables for Step 1 scores revealed exam scores explain 67% of the variance of students' Step 1 scores. April scores are more predictive of Step 1 performance with a significant increase from January (163) to April (187). Intervention efforts ameliorated risk of failing and explain passing scores for at-risk students.