Developing a machine learning-based predictive model for depression risk in patients with cardiovascular diseases

Area of Science:

• Medical Informatics
• Public Health
• Machine Learning in Healthcare

Background:

• Cardiovascular diseases (CVD) often co-occur with depression, negatively impacting patient outcomes.
• Accurate depression risk assessment is crucial for managing comorbid conditions in CVD patients.

Purpose of the Study:

• To develop and validate a machine learning (ML) model for predicting current depression risk in patients with CVD.
• To identify key predictors for early depression risk identification in this population.

Main Methods:

• Utilized data from the China Health and Retirement Longitudinal Study (CHARLS).
• Trained and tested eight ML models (AdaBoost, XGBoost, RF, etc.) on 2020 data, with temporal validation using 2018 data.
• Evaluated model performance using ROC/PR curves, calibration, DCA, and interpretability via SHAP analysis.

Main Results:

• The Adaptive Boosting (AdaBoost) model demonstrated superior predictive performance.
• Key predictors identified by SHAP analysis included life satisfaction, instrumental activities of daily living (IADL), sleep duration, and self-rated health.

Conclusions:

• A validated ML model was developed to estimate depression risk in CVD patients.
• This model can assist in the early identification of individuals at high risk for depression, facilitating timely interventions.