AI Model Identifies AF Patients Requiring Blood Thinners to Prevent Stroke

Atrial fibrillation (AF) is the most common abnormal heart rhythm, affecting around 59 million people globally. It increases stroke risk because quivering in the upper heart chambers allows blood clots to form and travel to the brain. Blood thinners are the standard treatment, but can also trigger dangerous bleeding events. Now, a new artificial intelligence (AI) model can make individualized treatment recommendations, balancing stroke prevention with bleeding risks.

This AI system, developed by researchers at the Mount Sinai Health System (New York, NY, USA), is called a graph neural network model and is designed to guide anticoagulation decisions for AF patients. Unlike current population-level risk tools, this model draws on each patient’s complete electronic health record. It calculates personalized probabilities of stroke and major bleeding, producing individualized treatment recommendations tailored to the patient’s clinical features.

The model was trained on 1.8 million patients, covering 21 million doctor visits, 82 million notes, and 1.2 billion data points. Validation was performed on 38,642 AF patients at Mount Sinai and further tested on 12,817 patients from Stanford datasets. Results showed that the model aligned treatment recommendations with both stroke and bleeding risk, reclassifying up to half of patients away from unnecessary anticoagulant use while maintaining safety.

By moving beyond generalized risk scores, this AI system provides clinicians with clearer, patient-specific recommendations that can reduce cognitive burden and improve decision-making. It offers a potential paradigm shift in AF care by minimizing stroke while avoiding unnecessary bleeding events. Researchers now aim to test the model in randomized clinical trials, with the hope of transforming anticoagulation strategies worldwide and improving long-term patient outcomes.

“This work illustrates how advanced AI models can synthesize billions of data points across the electronic health record to generate personalized treatment recommendations,” said co-corresponding author Girish Nadkarni, MD, MPH. “By moving beyond the ‘one size fits none’ population-based risk scores, we can now provide clinicians with individual patient-specific probabilities of stroke and bleeding, enabling shared decision making and precision anticoagulation strategies that represent a true paradigm shift.”

Related Links:
Mount Sinai Health System