Improving the selection of family medicine residents through development of multi-dimensional policy models

The annual cost of training a single family medicine resident may now exceed $50,000. This, together with the fact that normally only a small fraction of those applying for family medicine residency is selected for admission, creates a decision problem of enormous import to affected institutions. Despite these considerations, the applicant evaluation and selection process remains relatively subjective, with results often disappointing. In the current paper, a preference-based approach is proposed that first models the evaluation/selection process on the basis of well-defined cognitive and noncognitive criteria. It is suggested that validation of this model be based on future performance levels of both the accepted and rejected cohorts during and following their residencies. Discrepancies between future success levels and predicted outcomes may then be translated into appropriate control actions designed to: (1) improve the definition/measurement of selection criteria; (2) enhance the evaluation/selection policies and decisions of the admissions committee; and (3) better inform potential applicants of the department's program and selection philosophies. The approach is applied to two recent, accepted cohorts of the University of North Carolina Department of Family Medicine. Preliminary results indicate that the procedure is capable of improving the in-residency success levels of selected applicants, and that these levels can be better predicted than when no formal, i.e., analytic, process is followed.