Age-based screening for atrial fibrillation (AF) in adults aged 65 years or older has shown modest diagnostic yield. A secondary analysis of the VITAL-AF cluster-randomized trial, published in the Journal of the American College of Cardiology, examined whether validated clinical and electrocardiogram-based artificial intelligence (AI) risk models could better identify individuals most likely to benefit from screening.

VITAL-AF included patients aged 65 years or older across 16 primary care practices. Screening sites used single-lead electrocardiography. Among 30,630 participants without prevalent AF, 16,937 with prior 12-lead ECG and clinical data were included in this analysis.

Risk was estimated using the Cohorts of Heart and Aging Research in Genomic Epidemiology-AF (CHARGE-AF) score, an ECG-based AI model (ECG-AI), and a combined model (CH-AI). For 2-year incident AF, AUROC values were 0.711 for CHARGE-AF, 0.784 for ECG-AI, and 0.788 for CH-AI. Average precision values were 0.0952, 0.132, and 0.133, respectively.

A screening effect was observed in the highest CH-AI risk decile, where AF diagnosis rates were higher in the screening group than controls (10.07 vs 7.76 per 100 person-years; P < .05). The absolute difference was 2.32 per 100 person-years (95% CI, 0.01-4.63), corresponding to a number-needed-to-screen of 43 per year.

These findings suggest risk-based screening may improve efficiency, although greater efficiency may come at the cost of screening fewer individuals.