Machine learning stratifies disability risk in older adults

In a study of older adults aged 65 and older, researchers developed and externally validated a machine learning model to predict disability risk. The training and internal validation cohorts included 3,599 individuals from the China Health and Retirement Longitudinal Study, while an independent survey of 2,066 older adults served as an external validation cohort. Disability was measured by basic activities of daily living.

LASSO regression identified seven key predictors. Among five tested approaches, an Extreme Gradient Boosting model showed the best performance, yielding an area under the curve of 0.737 internally and 0.782 externally. Model interpretation identified age, comorbidity burden, and lower limb mobility as the primary predictors. Stratifying individuals into three tiers using risk cutoffs of 33.17% and 65.23% aligned with significant gradient increases in observed disability incidence.