Comparative effectiveness of atrial fibrillation screening guided by age versus estimated risk: a decision-analytic model

Abstract Background Atrial fibrillation (AF) is a common arrhythmia linked to increased stroke and heart failure risks. Since AF may be asymptomatic and is commonly episodic, the European Society of Cardiology guidelines recommend AF screening for individuals aged ≥65 years. However, comparative studies have shown mixed results in unselected older individuals and using approaches guided by estimated AF risk (e.g., clinical scores or artificial intelligence AI-based methods) may be more efficient. Since AF screening trials are resource-intensive, decision-analytics modeling can help inform optimal approaches for prospective studies. Purpose To compare the clinical effectiveness of AF screening approaches guided by age versus novel risk-informed approaches. Methods We developed an individual-level multistate decision-analytic model simulating 30 million individuals aged ≥50 years, with demographic and comorbidity profiles matching the United States population, tracking them until death or age 100. Among individuals with a guideline-based indication for anticoagulation by the CHA2DS2-VASc score, we assessed two AF screening modalities: 1) pulse palpation followed by 12-lead electrocardiogram (ECG) ("traditional") and 2) continuous monitoring with wrist-worn wearable devices ("contemporary"), plus a no-screening scenario. Each modality was applied in three contexts: (a) age-guided (screening individuals aged ≥65 years), (b) risk-guided (screening individuals with 5-year AF risk ≥5%), and (c) combined age- and risk-guided. AF risk was estimated using the test characteristics of the validated Cohorts for Heart and Aging Research in Genomic Epidemiology (CHARGE-AF) score (traditional) and a comparably performing AI-based ECG model (contemporary). Primary effectiveness outcomes were ischaemic strokes and major bleeding events, with AF true and false positive rates as secondary measures. Results Compared with no screening, all screening strategies resulted in a lower incidence of stroke and higher AF true positive rates but also a higher risk of major bleeding and higher AF false positive rates. Across both screening modalities, strategies with a risk-guided component consistently had lower stroke rates (traditional: 7.485-7.491 per 1,000 person-years; contemporary: 7.449-7.472 per 1,000 person-years) but also higher major bleeding rates (traditional: 3.585-3.604 per 1,000 person-years; contemporary: 3.591-3.634 per 1,000 person-years) (Table). Regarding clinical events, the most favorable screening strategy appeared to be contemporary screening using the combined age- and risk-guided approach (with stroke 7.449 and major bleeding 3.634 per 1,000 person-years). Conclusions In this simulation study, AF screening strategies utilizing estimated AF risk to select screening candidates appear more effective than current guideline-endorsed approaches utilizing age thresholds alone. Prospective assessment of risk-guided AF screening is warranted.