Machine learning effectively predicts bile leaks post-hepatectomy.

A machine learning model using extreme gradient boosting (XGBoost) was developed to predict bile leaks after hepatectomy for liver cancer. The model analyzed data from over 22,000 patients, achieving areas under receiver operating characteristic curves (AUROC) of 0.748, 0.719, and 0.711 in various patient cohorts. Factors such as serum alkaline phosphatase, surgical approach, and cancer diagnosis were highly predictive of bile leaks. An online calculator has been created to aid clinical decision-making in surgical settings.