Explainable Machine Learning Reveals Pattern-Specific Drivers of Depression in Middle-Aged and Older Adults with...

Area of Science:

• Gerontology
• Computational Psychiatry
• Public Health

Background:

• Depression is a major cause of disability, particularly in aging populations with multimorbidity.
• China's rapidly aging demographic faces complex interactions between multiple chronic conditions and depression.
• Traditional models struggle to capture these non-linear pathways, necessitating advanced analytical approaches.

Purpose of the Study:

• To identify pattern-specific determinants of depression in middle-aged and older adults with multimorbidity in China.
• To apply explainable machine learning techniques to large national datasets.
• To stratify analysis by empirically derived multimorbidity patterns for precise insights.

Main Methods:

• Cross-sectional analysis of 14,981 participants from the China Health and Retirement Longitudinal Study (CHARLS) 2020.
• Utilized interpretable machine learning (Gaussian process classifier, generalized additive model) with SHAP for feature importance and network analysis.
• Employed LASSO regression, Boruta algorithm, and logistic regression for predictor screening; SMOTE for class imbalance.

Main Results:

• Higher depressive symptom prevalence (42.96%) in the multimorbidity group compared to controls (27.01%).
• Key universal predictors included Instrumental Activities of Daily Living (IADL) disability, abnormal sleep, Activities of Daily Living (ADL) disability, rural residence, and lower education.
• Pattern-specific analyses revealed unique drivers, such as marital status for cardio-metabolic clusters and IADL disability for digestive-joint clusters.

Conclusions:

• Interpretable machine learning effectively identified universal and pattern-specific determinants of depression in multimorbid older Chinese adults.
• Functional disability and sleep disturbances are critical universal drivers.
• Phenotype-tailored insights offer a roadmap for precision screening and targeted interventions in aging populations.