Models for identifying cognitive frailty in hospitalized older adults with chronic heart failure in China: a systematic review

Abstract

Objective To systematically evaluate models for identifying cognitive frailty (CF) among hospitalized older adults with chronic heart failure (CHF) in China and assess their intended use, performance, and methodological quality. Methods A systematic search was conducted in PubMed, Web of Science Core Collection, Embase, the Cochrane Library, China National Knowledge Infrastructure, Wanfang, VIP, and the Chinese Biomedical Literature Database from inception to March 31, 2026. Two reviewers independently screened the records according to the predefined eligibility criteria and extracted data using the Checklist for Critical Appraisal and Data Extraction for Systematic Reviews of Prediction Modeling Studies. For each included study, the operational definition of CF, the reported approach to dementia exclusion, and the timing of predictor and outcome assessment were rechecked. Models in which predictors and CF were assessed during the same hospitalization were classified as diagnostic or screening models for prevalent CF rather than prognostic models for future CF onset. Risk of bias and applicability were independently assessed using the Prediction Model Risk of Bias Assessment Tool. Results Eight cross-sectional studies reporting 20 models were included. All were conducted among hospitalized older adults with CHF in China, and none predicted incident CF during follow-up. Development sample sizes ranged from 246 to 645, and CF prevalence ranged from 17.8% to 50.5%. Six studies performed internal validation, four performed external validation, and one reported no validation; three studies performed both internal and external validation. The area under the receiver operating characteristic curve in external validation ranged from 0.770 to 0.852. One study did not report calibration. The most frequently retained predictors were age, New York Heart Association functional classification, depression, nutritional status, and sleep-related factors. All studies had a high overall risk of bias, whereas applicability concerns were generally low. Conclusion Existing evidence concerns concurrent identification of prevalent CF rather than prediction of future onset. Current models are not ready for routine clinical decision-making because of high risk of bias, incomplete reporting, limited external validation, and lack of clinical-impact evidence. Longitudinal model development, multicenter external validation, comprehensive calibration and clinical-utility assessment, and prospective impact evaluation are required before implementation. Systematic review registration https://www.crd.york.ac.uk/PROSPERO/view/CRD420261413101, CRD420261413101.