A RETROSPECTIVE STUDY OF MEDICAL SCHOOL PERFORMANCE OF 1975-1981 GRADUATING CLASSES: COGNITIVE AND NONCOGNITIVE VARIABLES THAT PREDICT SUCCESS (ADMISSION CRITERIA)

The purpose of the study was to determine to what extent selected cognitive and noncognitive variables, known upon application, are related to success in medical school as measured by two criterion variables: basic science and clinical grade-point averages. Of the 33 variables on which data were collected for seven classes matriculating into at a large Northeastern medical school from 1971 through 1977 (N = 1103), eleven were selected for entry into a stepwise multiple regression analysis, using forward inclusion. Double cross validation was done where appropriate. The significance level for selection of variables for the development of regression equations was set at .01, with incremental validity of at least two percent being also required. For the basic science grade-point average, the regression analysis was performed separately for two subsets of the seven classes, 1971-1974 and 1975-1977. Undergraduate science grade-point average and MCAT science score accounted for most of the criterion variance (24%). For the clinical grade-point average, the results were problematic. Inequalities in regression lines with regard to slope and/or intercept were detected for race and gender; thus separate regression analyses were done for the appropriate subsets of classes. Overall, the undergraduate grade-point averages and MCAT verbal and general information scores entered alternately among the various separate regressions by gender and race. Overall, less than ten percent of the variance was accounted for in the clinical criterion. Of interest was the entry at the .01 level of two Astin indices, the number of non-science hours, and college major; however, the incremental validity was minimal. Additional regression analyses were done, entering the medical school basic science grade-point average as a predictor of clinical performance. Separate regression analysis was done for dominant and minority group students and for men and women. The basic science grade-point average entered first and increased the percent of variance accounted for in the clinical criterion by a substantial amount. Conclusions for the development of predictive models and recommendations for schools\u27 current admissions policies must be framed within the limitations of the study; namely, revisions in the MCAT, modifications in a school\u27s curriculum, and the fallibility of the criteria