A Comparison of Statistical Criteria for SettingOptimally Discriminating MCATand GPA Thresholds in Medical School Admissions

Abstract

Background: Using Medical College Admission Test-grade point average (MCAT-GPA) scores as a threshold has the potential to address issues raised in recent Supreme Court cases, but it introduces complicated methodological issues for medical school admissions.Purpose: To assess various statistical indexes to determine optimally discriminating thresholds for MCAT-GPA scores.Methods: Entering classes from 1992 through 1998 (N = 752) are used to develop guidelines for cut scores that optimize discrimination between students who pass and do not pass the United States Medical Licensing Examination (USMLE) Step 1 on the first attempt.Results: Risk differences, odds ratios, sensitivity, and specificity discriminated best for setting thresholds. Compensatory versus noncompensatory procedures both accounted for 54% of Step 1 failures, but demanded different performance requirements (noncompensatory MCAT-biological sciences = 8, physical sciences = 7, verbal reasoning = 7-sum of scores = 22; compensatory MCAT total = 24).Conclusions: Rational and defensible intellectual achievement thresholds that are likely to comply with recent Supreme Court decisions can be set from MCAT scores and GPAs.