Heart tracing model flagged higher risk of an irregular heartbeat
Seven-year rates of atrial fibrillation, an irregular heartbeat, were 6.4% in people flagged by an artificial intelligence heart-tracing model, versus 2.1% in those not flagged.
Our read
Biomarkers- Studied:
- Cohort study of 43,394 UK Biobank participants, median 3.6 years, against those without it
- What changed:
- Atrial fibrillation was higher: 7-year cumulative incidence was 6.4% with model evidence against 2.1% among those without it
- The catch:
- Underlying heart changes may have contributed to both the signal and later disease
- Why it’s aging research:
- It bears on staying healthy in midlife and later life.
Editor of record: Boris Djordjevic · AgingResearch.News
Among 43,394 UK Biobank participants, an observational cohort study tested whether an artificial intelligence model applied to standard heart tracings was linked to later atrial fibrillation. Participants averaged 55 years of age, and 47.4% were men. None had prior atrial fibrillation, heart failure or stroke.
The previously developed model was applied without retraining. It flagged 7,232 people as having evidence of diastolic dysfunction, or abnormal filling of the heart’s main pumping chambers. During a median follow-up of 3.6 years, 683 participants developed atrial fibrillation. The link between the model’s signal and later atrial fibrillation remained after adjustment for other factors. Adding the model’s measure to a clinical risk score improved prediction.
Why it matters
Subtle heart-filling abnormalities may reflect changes that precede atrial fibrillation. Identifying these changes could inform efforts to prevent this heart condition as people age.
Caveats
The analysis assessed a model-derived marker, not whether acting on it prevented atrial fibrillation; underlying heart changes may have contributed to both the signal and later disease. The findings were reported in a conference abstract and did not cover people with prior atrial fibrillation, heart failure or stroke.
The paper
Med AI; Seoul National Univ Hosp
AHA Scientific Sessions 2026 Abstract, Nov 2, 2026
Presented at AHA Scientific Sessions 2026, abstract 15755
Comments
Be the first to comment.