Abstract: Objective  To investigate the prevalence and determinants of social frailty among maintenance hemodialysis (MHD) patients, and to establish a theoretical framework for its prevention. Methods  From November 2023 to April 2024, MHD patients from the hemodialysis centers of Binzhou Medical University Hospital and Shanghai East Hospital were selected as study subjects. Variables were screened based on univariate analysis, and an optimized eXtreme Gradient Boosting (XGBoost) tree model was constructed using the XGBoost library to predict the degree of social frailty.  Results  A total of 354 patients were included, with a social frailty prevalence of 26.27% (93/354). Based on the optimal feature subset, an optimized XGBoost classification model was constructed. The Shapley Additive exPlanations (SHAP) importance ranking (mean absolute SHAP values) was as follows: physical exercise（0.447）, social participation(0.358), sleep conditions (0.328), among others. Conclusion  The prevalence of social frailty among MHD patients in this study was relatively high. Clinical healthcare providers can early identify high-risk groups and implement effective interventions based on the factors influencing social frailty in MHD patients.

Key words: Maintenance hemodialysis, Social frailty, Influencing factor