An explainable machine learning-based prediction model for sarcopenia in elderly Chinese people with knee...

Area of Science:

• Gerontology
• Musculoskeletal Health
• Biostatistics

Background:

• Sarcopenia, an age-related muscle disease, causes significant decline in function and increased mortality.
• Knee osteoarthritis (KOA) is a prevalent degenerative joint disease in the elderly, often co-occurring with sarcopenia.
• Early identification of sarcopenia risk in KOA patients is critical for timely interventions and improved health outcomes.

Purpose of the Study:

• To develop and validate an interpretable predictive model for sarcopenia risk in elderly individuals with symptomatic knee osteoarthritis.
• To identify key clinical and demographic factors associated with sarcopenia risk in this population.
• To facilitate early screening and risk assessment for sarcopenia in KOA patients.

Main Methods:

• Utilized data from the China Health and Retirement Longitudinal Study (CHARLS) on 95 variables for patients aged 65+ with symptomatic KOA.
• Employed LASSO and logistic regression for feature selection, followed by eight machine learning algorithms for model construction.
• Validated models using internal cross-validation and an independent test set, with SHAP analysis for interpretability and a web application for clinical use.

Main Results:

• Identified six significant predictors of sarcopenia risk: body mass index, upper arm length, marital status, total cholesterol, cystatin C, and shoulder pain.
• The CatBoost model demonstrated superior performance with high accuracy (0.8902), F1 score (0.8627), and AUC (0.9697) on the independent test set.
• The model showed strong generalization ability, with predicted probabilities closely matching actual occurrence rates, indicating reliability.

Conclusions:

• An interpretable sarcopenia risk prediction model was developed using routine clinical data for public health and aging perspectives.
• The model is suitable for early screening and risk assessment of sarcopenia in symptomatic KOA patients.
• This tool can aid health departments and clinicians in early detection and follow-up, enhancing elderly quality of life and health outcomes.