Abstract: Objective  To systematically search and evaluate the frailty risk prediction models for maintenance hemodialysis (MHD) patients.  Methods  The relevant literatures on frailty risk prediction models for MHD patients published on Web of Science, PubMed, CINAHL, Cochrane Library, Embase, CNKI, Wipnet, Wanfang and Chinese Biomedical Literature Database were systematically searched. The search period is up to May 1, 2024. Two researchers independently screened the literatures according to the inclusion and exclusion criteria, extracted data, evaluated model performance and quality of the literatures.  Results  A total of 16 studies with 18 models were included. Six studies conducted internal validation, one conducted external validation, and one conducted a combination of internal and external validation. The differentiation of 15 studies was ＞0.7. Models applicability is generally good, but the risk of overall bias is high, and mainly concentrated in the field of statistical analysis. The predictors of frequency ≥5 were age, comorbidities, albumin, sex, nutrition and sports training.  Conclusion  Most of the existing frailty risk prediction models for MHD patients have methodological defects and higher bias. In the future, frailty risk prediction models should be constructed and verified internally and externally on the basis of standardized research design and reporting procedures.

Key words: Maintenance hemodialysis, Frailty, Prediction model, Systematic evaluation