Atrial fibrillation (AF) is the most common abnormal heart rhythm, affecting nearly 59 million people worldwide. It significantly increases the risk of stroke, often leading doctors to prescribe anticoagulants, or blood thinners, to prevent clot formation. While effective, these medications also carry the risk of dangerous bleeding events, making treatment decisions complex and often uncertain.
Researchers at Mount Sinai have developed a groundbreaking AI model that may transform how clinicians make these decisions. Instead of relying on population-based risk scores that provide only average estimates, the model uses each patient’s entire electronic health record to generate individualized treatment recommendations.
The study, published as a late-breaking presentation, showed that the AI tool recommended against anticoagulant therapy for up to half of patients who would otherwise have received it under current guidelines. This could represent a paradigm shift in managing AF, balancing stroke prevention with reduced bleeding risk on a patient-by-patient basis.
The model was trained on over 1.8 million patient records, covering 21 million doctor visits, 82 million notes, and 1.2 billion data points. It was then validated both within the Mount Sinai Health System (38,642 patients) and with external data from Stanford (12,817 patients).
Results demonstrated that the system can dynamically update recommendations prior to a clinical visit and even break down probabilities of stroke and bleeding. This helps relieve the cognitive burden for clinicians and enables more informed conversations with patients.
Experts at Mount Sinai emphasize that this model represents a new era of precision medicine in cardiology. By moving beyond a “one size fits all” strategy, it allows for personalized anticoagulation strategies that could prevent unnecessary bleeding while still protecting patients from stroke.
If confirmed in future randomized clinical trials, the implications for global health are profound. Millions of patients could receive safer, more effective care, and clinicians could finally make decisions guided not by averages, but by the true risks of the individual in front of them.
Article written by Mount Sinai redaction team
01/09/2025
Source:
Mount Sinai