Explainable machine learning for identifying high-risk Alzheimer's disease and key health characteristics among community-dwelling adults.

Abstract

Objective: To develop an interpretable machine learning model to identify individuals with AD8-defined high-risk status for Alzheimer's disease among community-dwelling adults, identify key health characteristics and their association patterns with high-risk status, and to inform community-based cognitive health promotion.

Methods: Based on health survey data from 6,267 community-dwelling adults, AD8-defined high-risk status for Alzheimer's disease was assessed using the Alzheimer's Disease Screening Scale (AD8). LASSO logistic regression, random forest, and decision tree models were constructed, and SHapley Additive exPlanations (SHAP) was used to analyze the importance of key health characteristics and their association patterns.

Results: A total of 6,267 community-dwelling adults were included, of whom 1,454 (23.2%) were classified into the AD8-defined high-risk group. The random forest model showed the highest discriminative performance among the three models. SHAP analysis indicated that fruit intake, age, snoring, mean grip strength, and sleep duration were important health characteristics associated with AD8-defined high-risk status for Alzheimer's disease. SHAP dependence plots further revealed a graded negative association between fruit intake and high-risk status, a U-shaped association for sleep duration, a U-shaped association for mean grip strength, and a positive graded association between snoring frequency and high-risk status.

Conclusions: Random forest combined with SHAP identified key health characteristics associated with AD8-defined high-risk status among community-dwelling adults and revealed non-linear association patterns involving health behaviors and physical function. These findings may provide useful information for community-based risk stratification and targeted cognitive health promotion.