Accurate risk assessment for thromboembolic events in atrial fibrillation remains a clinical challenge. Data presented at the European Society of Cardiology Congress 2025 demonstrate that an artificial intelligence-driven electrocardiogram system outperforms traditional CHA2DS2-VA scoring. 

The study analyzed sixty thousand four hundred thirteen electrocardiograms from ten thousand one hundred eighty-one patients alongside longitudinal clinical records. The artificial intelligence model effectively distinguished low- and high-risk patients, achieving a predictive accuracy with an area under the receiver operating characteristic curve of 0.923. 

Kaplan-Meier analysis showed significant separation in survival for major adverse cardiac events. Compared with the CHA2DS2-VA score, artificial intelligence-based electrocardiogram analysis provided superior precision across risk categories, offering a more refined stratification of thromboembolic risk.

These findings highlight the potential of computational electrocardiogram interpretation to guide anticoagulation therapy and improve patient outcomes in atrial fibrillation.