Machine Learning-Driven Personalized Risk Prediction: Developing an Explainable Sarcopenia Model for Older European...

Area of Science:

• Gerontology
• Artificial Intelligence
• Clinical Medicine

Background:

• Sarcopenia poses a significant health risk for older adults, particularly those with arthritis.
• Early and accurate screening is crucial for timely intervention and management.

Purpose of the Study:

• To develop and validate an explainable machine learning (ML) model for predicting sarcopenia risk in older European adults with arthritis.
• To create a practical tool for early and precise screening in clinical settings.

Main Methods:

• Utilized data from the English Longitudinal Study of Aging (ELSA) and Survey of Health, Aging and Retirement in Europe (SHARE).
• Identified nine core predictors using ensemble feature selection and compared six ML algorithms.
• Evaluated model performance using Area Under the Curve (AUC) and calibration, with interpretability via SHapley Additive exPlanations (SHAP).

Main Results:

• A Decision Tree ML model achieved high performance with an AUC of 0.921 (internal) and 0.958 (external validation).
• Key predictors included stroke history, BMI, HDL, loneliness, walking speed, disease duration, age, recall summary score, and total cholesterol.
• SHAP analysis provided clear visualization of individual risk factor contributions.

Conclusions:

• Developed a high-performance, explainable, and lightweight ML model for sarcopenia risk prediction.
• An online tool using nine clinical indicators enables individualized risk assessment for early sarcopenia identification.
• Facilitates precision interventions for older European arthritis patients through enhanced decision support.