Enhanced prediction of cholecystectomy using obesity-modified TyG indices: a machine learning and SHAP-based study

TyG-derived composite indices, particularly TyG.WHtR, are strong predictors of both gallstone prevalence and the history of cholecystectomy. The XGB model demonstrated the best performance in predicting cholecystectomy risk, with the De Long test confirming its superior AUC-ROC compared to other models. The combination of strong predictive performance, good calibration, and high interpretability makes XGB a valuable tool for clinical decision-making in managing gallbladder disease risk.