Cardiothoracic Surgery Training Attrition (2014-2023): Predictors, Trends, and Delay-Adjusted Rates Across Traditional and Integrated Pathways

Resident attrition disrupts training pipelines and threatens cardiothoracic (CT) surgical workforce sustainability. This study quantifies national attrition rates and identifies predictors using trainee data from the Association of American Medical Colleges (AAMC), while accounting for reporting delays.

We analyzed de-identified records of 1,476 CT surgery residents entering ACGME-accredited programs from 2014 to 2023. Attrition was defined as departure before expected completion. Rates were stratified by year, gender, citizenship, and training pathway: integrated (TS-I) versus traditional (TS). Multivariable logistic regression, LASSO, Random Forest, and XGBoost with nested cross-validation were used to identify predictors. A parametric nowcasting model corrected for administrative reporting lag.

The overall attrition rate was 3.9% (57/1,476), significantly higher in the TS-I than the TS pathway (7.0% vs. 2.6%, p<0.001). On multivariable logistic regression, the TS-I pathway was the strongest predictor (aOR 7.11, 95% CI 3.16-16.01, p<0.01). Age was also independently associated with attrition, with higher odds in both the 31-36 group (aOR 2.89, 95% CI 1.28-6.54, p=0.04) and the >36 group (aOR 6.05, 95% CI 1.87-19.62) compared with those ≤30 years. Gender and citizenship were not significant after adjustment. Model performance was consistent across methods (AUC 0.67-0.72). Delay-adjusted analyses corrected an apparent decline in recent cohorts, revealing a stable underlying rate with a 3-4-year reporting lag.

Training pathways are the strongest predictor of CT resident attrition, with TS-I trainees at markedly higher risk. Correcting reporting delays is critical for accurate workforce planning. These findings support targeted retention interventions for at-risk trainees.