Characterization of Models for Identifying Physical and Cognitive Frailty in Older Adults With Diabetes: Systematic Review and Meta-Analysis

Background: A growing number of risk prediction models have been developed to estimate the risk of frailty in individuals with diabetes. However, the methodological quality and clinical utility of these models remain uncertain, limiting their application in both clinical settings and future research. Objective: This systematic review and meta-analysis aimed to evaluate and characterize existing prediction models for frailty in patients with diabetes. Methods: A comprehensive search of PubMed, Embase, Web of Science, and the China National Knowledge Infrastructure (CNKI), Wanfang, and VIP was conducted from their inception until March 2025. Relevant studies were identified, and extracted data included study design, data sources, outcome definitions, sample sizes, predictors, model development methods, and performance metrics. The Prediction model Risk Of Bias Assessment Tool (PROBAST) was used to assess the risk of bias and applicability of each included study. Results: A total of 3,080 records were identified through database searches. After screening, 25 studies (reporting 29 prediction models) were included. The prevalence of frailty in these patients with diabetes ranged from 10.1% to 62.1%. The most frequent predictors were depression, age, and regular exercise status. Among the 29 models, the reported the area under the receiver operating characteristic curve (AUC) values ranged from 0.630 to 0.983. All studies were found to have a high risk of bias, mainly owing to limitations in the analytical methods. The meta-analysis yielded a pooled AUC of 0.85 (95% CI: 0.83–0.87). Conclusions: Although the included studies demonstrated reasonable discriminative ability in predicting frailty among patients with diabetes, the overall methodological quality was suboptimal, with a high risk of bias identified in all studies. To enhance the reliability and generalisability of future prediction tools, further research should prioritise the development of models using large-scale, multicentre prospective cohorts, with rigorous external validation across diverse populations. Clinical Trial: Not Applicable