Objective: To identify predictors of attrition during graduate medical education (GME) in a single medical school cohort of contemporary US medical school graduates. Design: Retrospective cohort study. Setting: Single medical institution. Participants: Recent US allopathic medical school graduates. Main Outcome Measure: Attrition from initial GME program. Results: Forty-seven of 795 graduates (6%) did not complete the GME in their initial specialty of choice. At bivariate analysis, attrition was associated with election to the Alpha Omega Alpha Honor Medical Society, being an MD-PhD degree holder, and specialty choice (all P < .05). Attrition was not associated with graduation year (P = .91), sex (P = .67), or age (P = .12). In a multivariate logistic regression model, MD-PhD degree holder (odds ratio, 3.43; 95% confidence interval, 1.27-9.26; P = .02), election to Alpha Omega Alpha (2.19; 1.04-4.66; P = .04), choice of general surgery for GME (5.32; 1.98-14.27; P < .001), and choice of 5-year surgical specialty including those surgical specialties with a GME training requirement of 5 years or longer (2.74; 1.16-6.44; P = .02) each independently predicted greater likelihood of attrition. Conclusion: Academically highly qualified graduates and graduates who chose training in general surgery or in a 5-year surgical specialty were at increased risk of attrition during GME.