Existing cognitive frailty models show good discrimination in older adults with diabetes but suffer high bias risk.

ObjectiveCognitive frailty (CF) represents a significant geriatric issue closely linked to diabetes. Although multiple CF risk prediction models exist for older adults with diabetes, their methodological quality and clinical utility remain unclear. This systematic review evaluates the predictive performance and risk of bias of existing models to provide inform clinical practice.MethodsA systematic search was conducted in PubMed, Embase, Web of Science, Cochrane Library, CINAHL, Sinomed, CNKI, and Wanfang from inception to September 2025. Two researchers independently performed literature screening, data extraction, and quality assessment. Study and model characteristics were summarized descriptively; pooled AUC values were analyzed using Stata 17.0. PROBAST was used to evaluate risk of bias and applicability.ResultsEight studies involving 2,947 diabetic patients were included. CF prevalence ranged from 12.1% to 40.0%. Predictors encompassed sociodemographic, disease-related, psychological, and lifestyle factors, with age, depression, diabetes duration, nutritional status, and regular exercise being most frequently reported. The models showed good discrimination (AUC: 0.790-0.975) but exhibited high overall bias risk.ConclusionExisting CF prediction models demonstrate acceptable discrimination but are limited by high bias risk and poor applicability. Future research should prioritize developing rigorously designed models with multicenter external validation to enhance prediction accuracy. The study was reported in accordance with the Preferred Reporting Items for Systematic Reviews and Meta-analyses (PRISMA) guidelines (17). The study protocol was registered on PROSPERO (CRD420251054250).Systematic Review Registrationhttps://www.crd.york.ac.uk/PROSPERO/view/CRD420251054250, identifier CRD420251054250.