Correlates of muscle strengthening exercise in adolescents: A machine learning based analysis

Background: Muscle-strengthening exercise (MSE) is a critical component of adolescent health, yet its correlates remain less understood than those of aerobic activity. This study aimed to identify key correlates of meeting MSE guidelines among U.S. adolescents using an explainable machine-learning approach.

Methods: This cross-sectional study used data from the 2023 National Youth Risk Behavior Survey (YRBS), a nationally representative sample of U.S. high school students (n = 20,103). An eXtreme Gradient Boosting (XGBoost) classifier was developed to predict adherence to the MSE guideline (≥ 3 days/week) using sociodemographic, behavioural, dietary, and psychosocial predictors. Model performance was evaluated using the area under the curve (AUC) and accuracy. SHapley Additive exPlanations (SHAP) and partial dependence plots were employed to interpret feature importance and functional relationships.

Results: Overall, 61.8% of participants in the analytic sample met the MSE guideline. The XGBoost model demonstrated robust predictive performance (AUC = 0.835; Accuracy = 0.769). Feature importance analysis identified moderate-to-vigorous physical activity (MVPA), sex, and fruit intake as the top predictors. SHAP summary plots revealed that higher MVPA, male sex, and healthier dietary behaviours were associated with a higher probability of meeting the guideline, while poorer mental health was linked to lower adherence.

Conclusion: MSE participation is not an isolated behaviour but is strongly clustered with aerobic activity and a broader healthy lifestyle profile. The

Findings: highlight significant sex disparities and the role of psychosocial well-being in strengthening behaviours. Explainable machine learning provides a useful framework for identifying and prioritising correlates associated with MSE participation, which may help inform future hypothesis-driven research and potential public health strategies.