Predicting social isolation in maintenance hemodialysis patients using machine learning methods: a cross-sectional study

Background: This study aims to develop and validate a machine learning-based risk prediction model for social isolation in maintenance hemodialysis (MHD) patients, and at the same time determine the key risk factors. Method: 362 patients with MHD were recruited from a tertiary hospital in Shanghai and randomly divided into the training group and the detection group. We implemented and compared seven machine learning algorithms: Random Forest (RF), Decision Tree (DT), K-Nearest Neighbor (KNN), Logistic Regression (LR), Resilient Network (EN), Extreme Gradient Boosting (XGB), and Support Vector Machine (SVM). Result: In our MHD cohort, the incidence of social isolation was 45.856%. The comparative analysis shows that RF is the best prediction model (AUC = 0.95). Feature importance analysis identified significant predictors: Place of residence (1.277), Heart failure (HF) (0.559), Anxiety (0.306), Monthly household income (0.269), Age (0.255) Sleep condition (0.138). Conclusion: The prediction model based on RF has a good effect in identifying the social isolation risk of MHD patients. These findings enable clinicians to stratify high-risk populations and implement timely and targeted intervention measures, effectively reducing the risk of adverse consequences. Future multicenter studies should validate these results in larger cohorts.