Predicting Mortality in Atrial Fibrillation Patients Treated with Direct Oral Anticoagulants: A Machine Learning...

Area of Science:

• Cardiology
• Medical Informatics
• Machine Learning

Background:

• Atrial fibrillation (AF) is a prevalent arrhythmia associated with increased mortality, particularly in elderly populations.
• Direct Oral Anticoagulants (DOACs) are vital for stroke prevention in AF patients, surpassing traditional therapies.
• Existing risk scores (HATCH, CHA2DS2-VASc) have limitations in predicting mortality for AF patients.

Purpose of the Study:

• To develop and validate machine learning models for predicting 6-month mortality in AF patients treated with DOACs.
• To identify significant predictors of mortality within this patient cohort.
• To enhance personalized risk assessment and patient management strategies.

Main Methods:

• Utilized the MIMIC-IV database, analyzing data from 6431 AF patients.
• Employed LASSO for feature selection and built five machine learning models (Logistic Regression, Random Forest, XGBoost, LightGBM, AdaBoost).
• Applied SHAP values for model interpretability and feature importance analysis.

Main Results:

• The LightGBM model demonstrated superior performance with an AUC of 0.886, accuracy of 0.862, sensitivity of 0.913, and specificity of 0.859.
• Key predictors identified include length of hospital stay, ICU duration, and comorbidities.
• SHAP analysis provided insights into individual patient risk factors.

Conclusions:

• Machine learning models accurately predict mortality in AF patients on DOACs.
• The findings support the integration of these models into clinical practice for personalized patient care.
• This predictive capability can aid in optimizing treatment strategies and improving patient outcomes.