If future clinical trials confirm its effectiveness, this AI could profoundly improve patient care by more accurately identifying high- versus low-risk patients for stroke and bleeding.

Mount Sinai researchers have developed an AI model that personalizes anticoagulation decisions for atrial fibrillation (AF) patients by analyzing their entire electronic health record to recommend whether to use blood thinners to prevent stroke.

This AI model, trained on data from 1.8 million patients and validated on additional datasets, can recommend against anticoagulants for up to half of AF patients who would otherwise receive them under standard guidelines, potentially reducing unnecessary bleeding risks.

This innovative approach signifies a paradigm shift in AF treatment, offering dynamic, personalized recommendations that improve shared decision-making between patients and clinicians.

Traditional methods struggle with the complex risk profiles of AF patients, but this AI-driven model aims to overcome those limitations with more accurate, individualized assessments.

The algorithm employs a causal forest machine learning framework to estimate heterogeneous treatment effects, addressing the shortcomings of traditional subgroup analyses.

The AI model can update recommendations dynamically before appointments, easing clinicians' manual risk assessments and providing clearer, personalized risk profiles.

Experts from Mount Sinai highlight the model’s ability to adapt recommendations based on comprehensive patient data, facilitating shared decision-making.

Future plans include testing the AI in prospective clinical trials, refining it with additional data, and integrating it into electronic health records to support bedside decision-making and improve patient outcomes.

This approach could significantly reduce unnecessary anticoagulant use and associated bleeding risks while maintaining stroke prevention, marking a potential paradigm shift in AF management.

Given that AF affects roughly 59 million people globally, this AI-driven decision-making tool could enhance clinical outcomes by providing real-time, individualized risk assessments.

The model was trained on data from over 744,000 adult patients treated between 2015 and 2019, identifying that approximately 30% could benefit from left atrial appendage occlusion, especially among older adults with higher comorbidities.