Huff, K.L., Koenig, J.A., Treptau, M.M., & Sireci, S.G. (1999). Validity of MCAT scores for Predicting Clerkship Performance of Medical Students Grouped by Sex and Ethnicity. Academic Medicine, 74, S41-S44.

PURPOSE: This study is intended to extend previous work by examining the validity of the MCAT for predicting performance in year three of medical school. Two research questions were addressed: (1) To what extent do MCAT scores predict performance in third-year clerkships? And (2) To what extent does the MCAT predict performance differentially for groups based on race/ethnicity and sex?

METHOD: The MCAT Predictive Validity Study followed two cohorts of students (the entering classes of 1992 and 1993, respectively) enrolled at 14 institutions through all four years of medical school. Participating schools were chosen to be representative geographically, racially, and ethnically of all U.S. medical schools. The sample included both public and private schools as well as schools with traditional, problem-based, and systems-based curricula. Data were collected for 3,249 students overall. The following preadmission data were included in five predictor sets: (1) four MCAT scores; (2) SGPA and NSGPA (undergraduate science and non-science GPA); (3) MCAT scores, SGPA, and NSGPA; (4) SGPA, NSGPA, and selectivity, and (5) MCAT scores, SGPA, NSGPA, and selectivity. The criterion variables were the performance data or clerkship grades, as reported by the school.

Prediction of Total Group Performance: Hierarchical regression analyses were used to evaluate relationships between predictor sets and clerkship performance. Regression analyses were run separately by school and by cohort, which resulted in two validity coefficients per predictor set at each school.

Analysis by Group: An alternative procedure was used that allowed for comparison of prediction errors across groups when the regression equation for the total group was used. That is, the procedure looked at the extent to which use of a common regression equation resulted in systematic prediction errors for certain groups. As with the total group analyses, regressions were run separately for each cohort at each school.

RESULTS:

Prediction of Total Group Performance: The corrected coefficients show that, on average, UGPA accounted for approximately 13% (R=.36) of the variability of clerkship grades, while MCAT scores accounted for approximately 21% (R=.46). Explained variance in clerkship grades increased to 28% when the two predictors were considered together; however, the addition of selectivity did not increase the multiple correlation. When UGPA was considered along with selectivity, approximately 18% of the variance in clerkship grades was explained. This value, compared with 28% of the variance explained when MCAT is added to the predictor set, speaks to the incremental validity of the MCAT.

Analysis by Group: The results of analyses for racial/ethnic groups are consistent with those of previous studies of the MCAT and of other admissions tests. That is, performance for racial/ethnic minorities tends to be overpredicted, while performances for Caucasians and women tend to be underpredicted.

CONCLUSION: As with the total group analyses, the by-group analyses suggest that the addition of other preadmission variables, particularly SGPA, tended to improve prediction, more so for African Americans and Hispanics than for others. These results emphasize the importance of considering multiple factors in admissions. This study was necessarily limited to certain preadmission variables, those measures common across the study schools. However, other factors might have improved prediction. Future research is needed into the roles of additional predictor variables, such as diligence, motivation, communication skills, study habits, and other relevant characteristics that are likely to be related to success in medical school.

Contact Us    © 1995-2008 AAMC    Terms and Conditions    Privacy Statement