Development of a Visualized Risk Prediction System for Sarcopenia in Older Adults Using Machine Learning: A Cohort Study Based on CHARLS

Jinsong Du ^1,2 Xinru Tao ¹ Le Zhu ¹ Heming Wang ³ Wenhao Qi ⁴ Xiaoqiang Min ^2,5 Shujie Wei ⁶ Xiaoyan Zhang ⁷ Qiang Liu ^8*

• ¹ Zaozhuang University, Zaozhuang, China
• ² Shandong Coal Health School, Zaozhuang, China
• ³ Jilin University, Changchun, Hebei Province, China
• ⁴ Hangzhou Normal University, Hangzhou, Zhejiang Province, China
• ⁵ Shandong Healthcare Group Xinwen Central Hospital, taian, China
• ⁶ Zaozhuang Municipal Hospital, Zaozhuang, China
• ⁷ Shandong Healthcare Group Zaozhuang Central Hospital, zaozhuang, China
• ⁸ The First Affiliated Hospital of Nanchang University, Nanchang, Jiangxi Province, China

The elderly are at high risk of sarcopenia, making early identification and scientific intervention crucial for healthy aging. This study utilized data from the China Health and Retirement Longitudinal Study (CHARLS), including a cohort of 2,717 middle-aged and elderly participants. Ten machine learning algorithms, such as CatBoost, XGBoost, and NGBoost, were used to construct predictive models. Among them, the XGBoost model performed the best with an ROC-AUC of 0.7, and it was selected as the final predictive model for sarcopenia risk. To enhance model interpretability, SHAP technology was used to visualize the prediction results, and the system was implemented on a web platform. The system provides the probability of sarcopenia onset within four years based on input variables and identifies critical influencing factors, facilitating understanding and use by medical professionals. This system supports early identification and scientific intervention for sarcopenia in the elderly, with significant clinical value and application potential.