Association between A Body Shape Index and Frailty in Older US Adults: The Mediating Role of Physical Activity and Machine Learning-Based Prediction

Background Frailty is a common syndrome in the elderly population, significantly impacting their quality of life and prognosis. A Body Shape Index (ABSI), as a novel body shape index reflecting body fat distribution, has recently gained attention. However, epidemiological studies investigating the association between ABSI and frailty remain limited. Methods We used data from the 2007–2018 National Health and Nutrition Examination Survey (NHANES) to examine the relationship between ABSI and frailty among adults aged ≥ 60 years. Frailty was defined using the frailty index (≥ 0.25). Weighted multivariable logistic regression, subgroup and interaction analyses, and generalized additive models (GAM) were used to explore associations and nonlinear patterns. Mediation analysis assessed the role of physical activity (PA). Additionally, we performed feature selection using univariate analysis, LASSO regression, and the Boruta algorithm. Nine machine learning models were built to predict frailty risk, with Shapley additive explanations (SHAP) analysis and nomograms enhancing interpretability. Results ABSI and frailty risk were shown to be significantly positively correlated. Specifically, a 0.1-unit increment in ABSI was associated with a 32% increase in the odds of frailty after full adjustment for covariates (OR: 1.32, 95% CI: 1.14,1.53; p < 0.001). When participants were stratified by ABSI categories, those in the highest ABSI group exhibited a significantly elevated risk of frailty compared to individuals in the lowest ABSI group (OR: 1.28, 95% CI: 1.08,1.51; p < 0.001). The GAM analysis also showed a clear threshold effect and a nonlinear association; for ABSI values exceeding 0.83, each 0.1-unit increase corresponded to an 84% rise in frailty prevalence. Mediation analysis demonstrated that PA accounted for approximately 17.4% of the connection between ABSI and frailty. The Extreme Gradient Boosting (Xgboost) model demonstrated the best predictive capability, achieving an area under the curve (AUC) of 71.8%. Conclusion ABSI is independently associated with frailty in older US adults, with PA playing a partial mediating role. These findings suggest ABSI may be a useful marker for frailty risk assessment, warranting validation in prospective studies.