Identifying a Heterogeneous Effect of Atrial Fibrillation Screening in Older Adults: A Secondary Analysis of the VITAL-AF Trial

BACKGROUND: One-time atrial fibrillation (AF) screening trials in older adults have produced mixed results. In a secondary analysis of the VITAL-AF trial, we aimed to identify a subset of people in whom such screening is effective, using effect-based and risk-based approaches. METHODS: The VITAL-AF trial was a cluster-randomized trial of 1-time, 30-second single-lead ECG screening during primary care visits. It enrolled adults aged ≥65 years in 16 Massachusetts General Hospital primary care practices (2018–2019). In this secondary analysis, we tested 2 approaches to identify subgroups where screening is effective. First, we developed an effect-based model using T-learner, a causal inference approach that estimates screening effects by separately training 2 predictive models—one for screening and one for usual care—and then compares their predictions for each individual. Second, we used a validated AF risk model (Cohorts for Heart and Aging Research in Genomic Epidemiology AF) to test for heterogeneous screening effectiveness. We assessed AF screening effectiveness by quartile of predicted effect and predicted AF risk and determined their correlation. RESULTS: The study included 29 656 participants (mean±SD age 74±7 years, 59% women). In the highest quartile of predicted screening effect, AF diagnosis rates were higher in the screening versus the usual care group (4.00 versus 2.88 per 100 person-years, rate difference 1.12 [95% CI, 0.11–2.13] per 100 person-years). In the highest quartile of predicted AF risk, AF diagnosis rates were also higher in the screening versus the usual care group (5.55 versus 4.23 per 100 person-years, rate difference 1.32 [95% CI, 0.14–2.50] per 100 person-years). Predicted screening effect and predicted AF risk were weakly correlated (Spearman correlation coefficient, 0.23). CONCLUSIONS: One-time screening was associated with increased AF diagnoses in the top quartile of both predicted screening effect and predicted AF risk. Because predicted effect and risk were only weakly correlated, future AF screening efforts should include both high-effect and high-risk individuals.