To identify the risk factors associated with frailty among older adults in China and develop a predictive model for assessing their frailty risk. Secondary cross-sectional analysis. The 2018 Chinese Longitudinal Healthy Longevity Survey (CLHLS) provided data for this study. A total of 9006 participants were included in the analysis. Their general demographic, socioeconomic status and health behaviour risk factors were collected in the CLHLS. Frailty was assessed using the Frailty Index. A visual nomogram model was constructed based on independent predictors identified using multivariate analysis. The nomogram's discrimination and calibration capabilities were evaluated using the C-statistics and calibration curves. A 1000-times resampling enhanced bootstrap method was performed for internal validation of the nomogram. The results showed that living in rural settings, having a primary education level, having a spouse, having basic living security, smoking, drinking, exercising and social activities were protective factors against frailty. Increasing age, being underweight or obese, adverse self-assessed economic status and poor sleep quality were risk factors of frailty. The AUC values of the internal validation set were 0.830. The calibration curve was close to ideal. The Brier score was 0.122. The above results showed that the nomogram model had a good predictive performance. A simple and fast frailty risk prediction model was developed in this study to help healthcare professionals screen older adults at high risk of frailty in China. The frailty risk prediction model will assist healthcare professionals in risk management and decision-making and provide targeted frailty prevention interventions. Screening high-risk older adults and early intervention can reduce the risk of adverse outcomes and save medical expenses for older adults and society, thereby realising cost-effective planning of health resources and healthy ageing. No patient or public contribution. This study was a cross-sectional, secondary analysis of the CLHLS data.
Read full abstract