Analysis of frailty status and its influencing factors in maintenance hemodialysis patients based on the health ecological model

Abstract Background Frailty is a significant public health concern in maintenance hemodialysis (MHD) patients.Previous studies have predominantly focused on isolated risk factors, lacking a comprehensive framework to address its multifactorial nature. This study aims to systematically investigate the prevalence and determinants of frailty in MHD patients using the Health Ecological Model (HEM), which integrates five ecological tiers—individual traits, behavioral characteristics, interpersonal networks, living and working conditions, and policy environment—and to develop a backpropagation neural network (BPNN) prediction model for early frailty identification. Methods From January to June 2025, a cross-sectional study was conducted using convenience sampling. A total of 2,224 adult patients on maintenance hemodialysis for ≥3 months were recruited from 10 centers in northeastern Sichuan, China. Frailty was assessed using the Fried Frailty Phenotype scale. Independent variables based on the five tiers of the HEM were analyzed using binary logistic regression. A BPNN model was employed to identify the primary predictors of frailty. The performance of the BPNN model was validated and compared against Random Forest and eXtreme Gradient Boosting (XGBoost) algorithms. Results The prevalence of frailty among MHD patients was 30.08%. Key risk factors included weekly exercise < 150 min, depression,poor self-rated health, ≥ 3 chronic comorbidities, age ≥75 years, higher education level, sleep disorders, and divorced/widowed marital status. Protective factors included social support, and urban employee basic medical insurance. The BPNN model identified weekly exercise < 150 min,depression and poor self-rated health as primary predictors of frailty (AUC = 0.944). Conclusion The HEM provides an innovative multidimensional perspective for unraveling the complexity of frailty in hemodialysis patients. The BPNN offers a robust tool for precise risk stratification.