Machine learning algorithms predicts risk of urinary incontinence after transurethral prostate surgery: a retrospective cohort study

Purpose This study aimed to develop a machine learning (ML) model to predict postoperative urinary incontinence (UI) risk after transurethral prostate surgery, identify key influencing factors, and provide a data-driven tool for personalized risk assessment. Methods Retrospective analysis of data from 1135 patients diagnosed with benign prostatic hyperplasia who underwent transurethral prostate surgery. Postoperative UI were assessed at 2 weeks (early UI, EUI) and 2 months (late UI, LUI). 23 potential influencing factors were included in the study.Significant influencing factors were identified through LASSO regression. Subsequently, eight ML algorithms were employed to construct a predictive model for UI following surgery. Model performance was evaluated by calibration curves and decision curve analysis (DCA). SHapley Additive exPlanation (SHAP) values were used to explain the importance of features in the model. Results 136 patients (12.0%) occurred EUI, with 13 significant features (P < 0.05).39 patients (3.4%) experienced LUI, associated with 10 features. We employed eight ML algorithms to construct the models after LASSO regression analysis. The results showed that the eXtreme Gradient Boosting (XGBoost) model outperformed other models in predicting EUI and LUI, with AUCs of 0.766 (95% CI: 0.656–0.876) and 0.892 (95% CI: 0.763–0.996), respectively. The DCA curves further demonstrated that the XGBoost model provides significant advantages over all models. SHAP analysis revealed the contribution of each feature to the XGBoost model. Conclusion Machine learning models developed from clinical data effectively predict UI following transurethral prostate surgery, identifying critical predictors to inform personalized risk stratification and optimize postoperative clinical management.