Important Risk Factors in Patients with Nonvalvular Atrial Fibrillation Taking Dabigatran Using Integrated Machine...

Area of Science:

• Cardiology
• Medical Informatics
• Pharmacology

Background:

• Nonvalvular atrial fibrillation (NVAF) necessitates anticoagulation therapy.
• Predicting vascular events and bleeding is crucial for managing NVAF patients on dabigatran.
• Existing prediction models may not fully capture complex risk profiles.

Purpose of the Study:

• To develop an integrated machine learning (ML) scheme for predicting vascular events and bleeding in NVAF patients.
• To identify significant risk factors associated with these outcomes.
• To compare the performance of ML models against traditional methods.

Main Methods:

• Post-hoc analysis of the Randomized Evaluation of Long-Term Anticoagulant Therapy trial database.
• Development of an integrated ML scheme combining logistic regression (LGR), naive Bayes, random forest (RF), classification and regression tree, and extreme gradient boosting (XGBoost).
• Feature selection using RF and XGBoost to identify key risk factors.

Main Results:

• RF and XGBoost demonstrated superior predictive performance (AUC 0.780 and 0.717 for vascular events, 0.684 and 0.618 for bleeding) compared to LGR (AUC 0.674 and 0.605).
• Identified major risk factors for vascular events: age, congestive heart failure, myocardial infarction history, smoking, kidney function, and BMI.
• Identified major risk factors for bleeding: age, kidney function, smoking, bleeding history, drug interactions, and dabigatran dosage.

Conclusions:

• Integrated ML approaches, particularly RF and XGBoost, are effective for predicting vascular events and bleeding in dabigatran-treated NVAF patients.
• The identified risk factors provide insights for personalized risk assessment and management.
• ML algorithms offer a powerful tool for analyzing complex medical data, paving the way for precision medicine.