Machine learning-based model to predict composite thromboembolic events among Chinese elderly patients with atrial...

Area of Science:

• Cardiology
• Medical Informatics
• Geriatrics

Background:

• Accurate survival prognosis is crucial for clinical decision-making in elderly patients.
• Atrial fibrillation (AF) increases the risk of various thromboembolic events.

Purpose of the Study:

• To develop and validate a machine learning model for predicting composite thromboembolic events (CTEs) in elderly patients with AF.
• Identify key risk factors associated with CTEs in this population.

Main Methods:

• Retrospective study of 6,079 elderly patients (≥75 years) with AF.
• Utilized random forest imputation for missing data.
• Trained and validated four machine learning models (logistic regression, decision tree, random forest, XGBoost) on training and validation datasets.

Main Results:

• The incidence of CTEs was 19.53%.
• A random forest model achieved high performance (AUC: 0.927), outperforming other models.
• Key predictors identified included history of ischemic stroke, high triglycerides, high total cholesterol, elevated plasma D-dimer, and age.

Conclusions:

• A highly accurate random forest model can stratify elderly AF patients at high risk for CTEs.
• History of ischemic stroke, age, lipid profile, and D-dimer levels are significant correlates of CTEs.