Frailty is a common geriatric syndrome associated with increased risk of adverse health outcomes, highlighting the need for early identification and appropriate management. Although machine learning (ML) models have shown promise in classifying frailty, research on sex-specific classification performance and its contributing factors remains limited. This study aimed to develop sex-specific interpretable ML classification models and identify sex-specific contributors to frailty using the SHapley Additive exPlanations (SHAP). A total of 1,540 community-dwelling older adults (mean age 76.2 ± 3.8 years; women 48.6%) from the Korean Frailty and Aging Cohort Study were included. Frailty status was assessed using the Fried Frailty Phenotype and classified into frail and robust groups. Sex-specific ML models were developed using 73 variables across 8 domains. Model performance was evaluated using stratified cross-validation, and SHAP analysis was applied to interpret feature contributions and explore sex differences in frailty-related factors. Sex-specific ML models demonstrated high performance in frailty classification. The best model for males was logistic regression (F1-score: 0.89; accuracy: 0.94), whereas for females, both logistic regression and XGBoost achieved F1-scores of 0.92 and accuracy of 0.95. The SHAP analysis revealed sex-specific key predictors, with physical, biological, and psychological indicators being the most influential. The sex-specific frailty prediction models developed in this study demonstrated outstanding performance and revealed sex-dependent differences in key predictive factors, thereby providing an essential foundation for the development of personalized prevention and management strategies.
Jung et al. (Fri,) studied this question.