Liu AI-ECG (AF & anticoagulation)

Artificial Intelligence-Enabled ECGs for Atrial Fibrillation Identification and Enhanced Oral Anticoagulant Adoption: a Pragmatic Randomized Clinical Trial

Patient / Population Intervention / Exposure Comparison Outcome

In 17,817 emergency and hospitalised patients (8,857 intervention, 8,960 control) treated by non-cardiologists at two hospitals in Taiwan, AI-ECG alerts to the treating non-cardiologist when atrial fibrillation was detected in a patient at risk of stroke increased NOAC prescription within 90 days of discharge compared with usual care.

N
17,817
Design
Open-label cluster RCT (120 vs 113 non-cardiologists randomised), two hospitals, Taiwan
Endpoint
NOAC prescription within 90 days after discharge (co-primary with new AF diagnosis, echocardiogram arrangement and cardiologist visits)
Relevance
3
ResultNOAC prescription 23.3% vs 12.0% (HR 1.85, 95% CI 1.11–3.07); new AF diagnosis HR 1.40 (95% CI 1.03–1.90); no significant difference in echocardiograms, cardiologist visits, ischaemic stroke, cardiovascular or all-cause death. 275 vs 245 patients had AI-detected AF.
Liu WT, et al. J Am Heart Assoc. 2025;14(14):e042106. 10.1161/JAHA.125.042106
Discussion & critique

Targets the treatment gap rather than detection alone: an AI-ECG alert made non-cardiologists nearly twice as likely to start anticoagulation in AF patients at stroke risk. Randomised at the physician level and open-label in only two hospitals, with four co-primary endpoints of which two were met; the effective outcome population was the 275 vs 245 patients with AI-detected AF, and stroke and mortality were unchanged.