Development and validation of an ADL dependency model for Centenarians: a CHCCS-based cross-sectional cohort study

Objective This study aimed to investigate the influencing factors of activities of daily life (ADL) among centenarians and to develop and validate a prediction model of ADL dependency for this population. Methods A total of 952 eligible centenarians from the China Hainan Centenarian Cohort Study (CHCCS) were included. An ADL score below 90 was defined as ADL dependence. Participants were randomly divided into development (70%) and validation (30%) groups. Univariate and multivariate logistic regression analysis (LRA) of the development group were used to identify independent risk factors related to ADL dependency. The selected variables were employed for modeling and nomogram construction. The model's performance was assessed using the receiver operating characteristic (ROC) curve, calibration plots, net reclassification index (NRI), and integrated discrimination improvement (IDI) scores. Decision curve analysis (DCA) was utilized to evaluate the clinical utility of the model. Results The development group comprised 668 participants, and the validation group included 284. After variable selection via univariate and multivariate logistic regression analyses, eight factors—residential type, chronic pain, incontinence, weight loss, napping, social participation, BMI, and albumin—were incorporated into the prediction model. The area under curve (AUC) the ROC curve was 0.796 (95% CI: 0.763–0.829) for the development group and 0.800 (95% CI: 0.750–0.851) for the validation group. Calibration plots, NRI, and IDI indicated a good fit of the model in both groups. The DCA demonstrated clinical effectiveness. Conclusions Factors such as living alone, experiencing chronic pain, incontinence, weight loss, absence of napping, lack of social participation, low BMI, and low albumin levels were identified as risk factors for ADL dependency among centenarians. The tailored prediction model encompassing these eight factors is suitable for early identification and prediction of ADL dependency in extremely elderly individuals.