Abstract
Background
Stroke survivors combined with frailty have high rates of complications, mortality, disability, and readmission. Given that frailty is an early stage of disability that is reversible and preventable, a reliable frailty risk identification model should be developed. This study aimed to develop and validate a stroke frailty risk identification model using information collected from the China Health and Retirement Longitudinal Study (CHARLS) database.
Methods
Data were obtained from the CHARLS. Stroke survivors were selected from the database and analyzed for 30 relevant indicators, including socio-demographic variables, physical status, psychological, cognitive, and social factors. The data were divided by year, with 2013 and 2015 as the development set and 2018 and 2020 as the validation set. Screening was performed using least absolute shrinkage and selection operator (LASSO) regression analyses. Logistic regression risk identification models were developed based on the results of univariate analyses and LASSO variable screening. Factors associated with frailty in stroke survivors were explored and identified. A nomogram was constructed for modelling risk identification. Calibration curves and decision curve analysis were used to determine the fit of the model and test the discriminatory power of that model, respectively.
Findings
A total of 2,188 stroke survivors from the CHARLS database collected at follow-up in 2013, 2015, 2018, and 2020 were included in the final analysis. About 68% stroke survivors had symptoms of frailty. We found statistically significant differences in age, marital status, living alone, hypertension, and self-reported health status (all with p < 0.05). Age, sleep quality, balance, nervousness and anxiety, and living alone were independent risk factors for the development of frailty in older stroke survivors. The area under the receiver operating characteristic (ROC) curve of the column line graph for the development and validation sets was 0.833 and 0.838, respectively.
Interpretation:
Frailty risk identification models for stoke survivors built using CHARLS data have better discriminatory performance than models built using raw data collected from small samples in the literature. Thus, this work has an implication for the clinical practice of identifying those high-risk populations for frailty.