Validation and comparison of a brief instrument vs a single-item screen to predict entry to family medicine at matriculation to medical school

BACKGROUND AND OBJECTIVES: A strong US primary care workforce is necessary to meet health care needs, yet fewer than 9% of allopathic medical students choose family medicine each year. No validated instrument exists to identify students likely to enter family medicine upon medical school matriculation. METHODS: A subset of a larger survey at the University of Washington School of Medicine (UWSOM) was used to create the Family Medicine Interest Survey (FMIS), a 15-item instrument to predict eventual practice in family medicine for a 2003-2007 matriculating cohort. A single-item screen asking about top specialty choice was administered at UWSOM for the same cohort and for a 2006-2012 matriculating cohort of students at Oregon Health & Science University (OHSU). Test performance measures including D (discrimination) and Cronbach α were calculated. Logistic regression determined whether FMIS score or reporting family medicine as the top specialty choice predicted family medicine practice for 601 UWSOM graduates or family medicine residency match for 744 OHSU graduates. RESULTS: The FMIS is reliable (Cronbach α=0.76). Both tests significantly predicted the probability of entering family medicine. Listing family medicine as the preferred specialty choice yielded a 47% predicted probability for UWSOM graduates entering family medicine. OHSU graduates listing family medicine first had an eightfold odds of matching to family medicine residencies. Combining the two instruments for UWSOM graduates showed a dose-response curve for predicted probability of entering family medicine with increasing levels of interest.