Abstract:

Objective This study systematically synthesizes and comprehensively evaluates the construction characteristics，predictive performance and methodological quality of existing models for predicting the risk of frailty in maintenance hemodialysis（MHD） patients，with the aim of providing a comprehensive evidence summary and reference for the optimization and clinical transformation of models in this field. Methods Systematically searched PubMed，Embase，Web of Science，Cocharne Library，CNKI，Wanfang Database，VIP Database and China Biomedical Literature Database to collect relevant studies published from the establishment of each database up to September 2025 on the prediction model of frailty risk for maintenance hemodialysis patients. Literature screening and data extraction were independently completed by two researchers，and the risk of bias and clinical applicability of all included studies were strictly assessed using the Predictive Model Bias Risk Assessment Tool （PROBAST）. Results A total of 15 studies involving 19 prediction models were included. Meta-analysis showed that the model had good overall discrimination overall（pooled AUC=0.882，95%CI：0.839-0.924），but there was a high degree of heterogeneity between studies（I²=92.57%），and sensitivity analysis confirmed that this result was robust. PROBAST assessment showed that 13 studies were at high risk of bias. Predictors were categorized into four groups：patient baseline data，disease burden，laboratory indicators and functional status. Among them，age，comorbidity index，serum albumin，depressive symptoms and physical activity level were the most frequently reported core risk factors. Conclusion Existing research on the risk prediction model for MHD patients shows the contradictory characteristic of “favorable apparent performance， but insufficient methodological rigor”. However， the core set of risk factors identified in this study can provide practical warning signals for high-risk individuals in current clinical practice. In the future， high-quality research on risk prediction models needs to be conducted， and their clinical applicability and effectiveness should be verified through actual clinical implementation.

Key words: Hemodialysis, Frailty, Prediction Model, Systematic Review, Evidence-Based Nursing