MCAT scores and undergraduate GPA are two screening devices that are widely used by medical schools in the U.S. to select aspiring medical students. Given their widespread use, it is important that the validities of MCAT scores and undergraduate GPA be investigated to ensure the accuracy of medical school admission decisions.
This study was therefore designed to address three important aspects of validation, namely, predictive validity of MCAT subtest scores and undergraduate GPA, differential validity, and differential prediction of the MCAT subtest scores. First, predictive validity was evaluated as an index of the relationship between the predictors, MCAT subtest scores and undergraduate GPA, and the criterion, first-year medical school GPA. Second, differential validity was assessed by comparing the magnitude of validity coefficients obtained as the correlation between first-year medical school GPA and MCAT subtest scores for men and women and for White, Black, Asian, and Hispanic medical students. Third, differential prediction was examined by testing for differences in the regression systems obtained for the different subgroups of examinees.
The sample used in the study consisted of 3,187 students drawn from 1992 and 1993 cohorts of 14 medical schools. Statistical procedures utilized in the research included regression analysis, Fisher's z transformations, F-ratio test of equality of standard errors of estimate, and ANCOVA tests of equality of regression slopes and intercepts.
Results obtained showed moderately high correlations between the predictors and the criterion. MCAT subtest scores and previous grades were individually good predictors of medical school freshman grades. The combination of MCAT subtest scores and undergraduate GPA was, however, a more powerful indicator of performance in the first-year of medical school.
Differential validity results showed that in most cases women had higher validity coefficients compared to men. With regards to differential prediction, the results implied that using common regression equations derived from a pooled sample of examinees to predict performance in the first year of medical school may result in underprediction for Whites and overprediction for Blacks and Hispanics.