Interpretable Machine Learning-based Predictive Model For Fall Risk In Older Adults Receiving Maintenance Hemodialysis

IntroductionFalls are common and disabling in older adults receiving maintenance hemodialysis (MHD), yet conventional risk assessment tools often show limited predictive accuracy in this population. This study aimed to develop and validate an interpretable machine learning (ML) model to predict fall risk in older patients undergoing MHD.MethodsIn this prospective study, 1,248 older adults receiving MHD were followed for 6 months. Participants were randomly divided into training and testing sets. To optimize feature selection and reduce multicollinearity, a dual-algorithm strategy combining the Least Absolute Shrinkage and Selection Operator (LASSO) and Boruta algorithms was employed. Nine ML algorithms were constructed and compared. Model performance was assessed using the area under the receiver operating characteristic curve (AUC), calibration plots, and decision curve analysis (DCA). Model interpretability was evaluated using Shapley Additive exPlanations (SHAP), and the final model was implemented as a Shiny-based web application.ResultsDuring follow-up, 487 participants (39.0%) experienced at least one fall. Among the nine algorithms, the Categorical Boosting (CAT) model showed the best overall performance, with an AUC of 0.865 (95% confidence interval [CI]: 0.828–0.902), the highest accuracy and F1-score, and good calibration (Brier score = 0.145). DCA demonstrated that the CAT model yielded the greatest net clinical benefit across a range of threshold probabilities. SHAP analysis identified frailty, use of walking aids, and older age as the strongest contributors to fall risk.ConclusionWe developed and validated a robust and interpretable CAT-based model for predicting falls in older adults receiving MHD. By highlighting major clinical and functional risk factors and providing an accessible web-based calculator, this model may support early risk stratification and individualized fall prevention strategies in clinical practice.