Which medical school applicants will become generalists or rural-based physicians?

The hypotheses that data, available at the time when a medical school admission decision is made, can be used to predict generalist specialty choice and rural practice location were tested. Applicant data, available to admissions committee members at the University of Louisville in 1986 and 1987 about the classes of 1990 and 1991 respectively, were correlated with specialty choice and practice location in a retrospective cohort study. Data collected from 1994 to 1996 about the 1990 and 1991 graduates were used to develop a mathematical model to predict specialty choice and practice location using stepwise logistic regression. These models were more accurate in predicting which applicants would not select a generalist career (negative predictive value = 80.7%) than those who would (positive predictive value = 42.7%) and in predicting those who would not practice in a rural location (negative predictive value = 91.9%) than those who would (positive predictive value = 37.8%). We conclude that applicant data, available at the time admission decisions are made, are of limited value for identifying those who will eventually become generalist physicians or practice in a rural area. However, the data are useful for identifying those who will not.