Prediction of the fall risk at home in maintenance-hemodialysis patients

Background This study aimed to identify the key risk factors for falls in the home environment among maintenance hemodialysis (MHD) patients and predict the fall risk. Methods This single-center retrospective cohort study included 365 MHD patients. Multidimensional data encompassing demographic characteristics, clinical indicators, home environment, and daily activity behaviors were collected. Variables were selected using the Least Absolute Shrinkage and Selection Operator regression. A Random Forest prediction model was constructed, and the Synthetic Minority Over-sampling Technique was applied to address class imbalance. Model performance was evaluated using the area under the receiver operating characteristic curve, accuracy, sensitivity, and specificity. Clinical utility was assessed via Decision Curve Analysis. Shapley Additive Explanations (SHAP) analysis was employed to elucidate the direction and magnitude of the contribution of each variable to individual fall risk. Results The RF model demonstrated excellent predictive performance on the test set, with an AUC of 0.94 (95% CI: 0.89–0.98), accuracy of 0.909, sensitivity of 0.922, and specificity of 0.850. SHAP analysis revealed the top six predictors as history of falls, serum potassium level, use of walking aids, absence of anti-slip devices in bathrooms, intradialytic hypotension, and history of diabetes mellitus. Conclusions The RF model, constructed based on the PEO framework, effectively predicts the risk of falls in the home environment in MHD patients, demonstrating strong discriminative ability and clinical applicability. This model may provide healthcare professionals with a personalized risk-assessment tool to support the development and implementation of fall-prevention strategies.