Explainable Machine Learning Approaches Predict Frailty and Adverse Outcomes in Older Adults: Development and...

Area of Science:

• Gerontology
• Artificial Intelligence in Healthcare
• Public Health

Background:

• Early identification of frailty in older adults is crucial for preventing adverse health outcomes.
• Existing frailty prediction models often lack reliability, interpretability, and generalizability.
• Machine learning offers potential for developing more accurate and robust frailty prediction tools.

Purpose of the Study:

• To develop and validate a reliable and interpretable machine learning model for predicting frailty in older adults.
• To identify key predictors of frailty, including lifestyle and psychological factors.
• To assess the association between predicted frailty and adverse clinical outcomes.

Main Methods:

• Utilized data from the Chinese Health and Retirement Longitudinal Study (CHARLS) for model development (n=3419) and the Chinese Longitudinal Healthy Longevity Survey-Heart Failure (CLHLS-HF) for external validation (n=1017).
• Applied six machine learning models, with XGBoost selected as the best-performing model.
• Employed the SHapley Additive exPlanations (SHAP) method for model interpretability and evaluated clinical outcomes in predicted frail vs. non-frail individuals.

Main Results:

• The XGBoost model demonstrated strong performance in both internal (AUC=0.934, F1=0.712) and external validation (AUC=0.792, F1=0.702).
• Key predictors of frailty included instrumental activities of daily living, BMI, self-rated health, and depression.
• Individuals predicted as frail exhibited significantly higher risks of falls (OR=2.11), hospitalization (OR=1.75), and disability (OR=1.42).

Conclusions:

• The developed XGBoost model serves as a robust and interpretable digital tool for predicting frailty in older adults.
• The model effectively identifies individuals at increased risk for adverse clinical outcomes.
• This approach enhances the potential for timely interventions to mitigate frailty-related complications.