Key result
Frailty risk prediction models for maintenance hemodialysis patients demonstrated acceptable discrimination (AUC 0.72 to 0.998) but suffered from high risk of bias and suboptimal clinical applicability.
Why the study?
To conduct a systematic review and critical appraisal of existing risk prediction models for frailty in maintenance hemodialysis patients.
Systematic Review (n=16)
Effect estimate: AUC 0.72 to 0.998
Existing frailty risk prediction models for maintenance hemodialysis patients show acceptable discrimination but suffer from high risk of bias and suboptimal methodological quality.
Existing MHD frailty models show acceptable discrimination but high bias; leaves open need for validated, low-bias tools before clinical use.
OBJECTIVE: We aimed to conduct a systematic review and critical appraisal of studies on maintenance hemodialysis (MHD) frailty risk prediction models. METHODS: A systematic search was conducted up to November 5, 2024, in the PubMed, Web of Science, Cochrane Library, Scopus, ProQuest, Embase, Cumulative Index to Nursing and Allied Health Literature, China National Knowledge Infrastructure (CNKI), Wanfang Database, Chinese Scientific and Technical Journals Database (VIP), and China Biology Medicine Database (CBM) for published articles, and data were extracted according to the Critical Appraisal Checklist and the Critical Appraisal of Research Systems for Predictive Modeling data extraction form. The risk of bias (ROB) and applicability of the included studies were assessed using the Predictive Modeling Risk of Bias Assessment Tool. RESULTS: An initial literature search of 2,578 articles was conducted, and screening resulted in the inclusion of 16 studies and 19 predictive models. A total of 18 models reported area under the curve for subject work characteristics ranging from 0.72 to 0.998, and eight models reported calibration using calibration plots and/or goodness-of-fit curves. The prevalence of MHD patients ranged from 14.5% to 59.57%. Age, serum albumin level, sex, Charlson comorbidity index, daily activity status, and Nutritional Risk Screening 2002 (NRS-2002) score were the most common predictors. All the studies had a high ROB, mainly owing to poor reporting of the outcome domain and analysis domain; five studies were of high concern regarding applicability. CONCLUSION: Although most included prediction models demonstrated acceptable discriminatory performance, the overall methodological quality and clinical applicability of existing studies remain suboptimal. Future multicenter studies with large-scale cohorts should adhere to both the Predictive Modeling Risk of Bias Assessment Tool and Transparent Reporting of a Multivariable Prediction Model for Individual Prognosis or Diagnosis guidelines to develop predictive models characterized by three critical attributes: robust clinical utility, minimized ROB, and scientific validity.
No takes yet. Share an insight, caveat, or question.
Wang et al. (2026) conducted a systematic review in Frailty in maintenance hemodialysis (n=16). Frailty risk prediction models was evaluated on Model discrimination (AUC) and calibration (AUC 0.72 to 0.998). Frailty risk prediction models for maintenance hemodialysis patients demonstrated acceptable discrimination (AUC 0.72 to 0.998) but suffered from high risk of bias and suboptimal clinical applicability.
Synapse has enriched 5 closely related papers on similar clinical questions. Consider them for comparative context: