The PAFIS score for risk stratification of atrial fibrillation detected after stroke: a multicenter cohort study

Abstract

Background Atrial fibrillation detected after stroke (AFDAS) is clinically important, but AFDAS-specific risk tools for patients without known atrial fibrillation (AF) remain limited. We developed and temporally validated the Prediction of AF in Ischemic Stroke (PAFIS) score. Methods We retrospectively analyzed ischemic-stroke patients from the National Taiwan University Hospital Integrative Medical Data Center. The development cohort included patients hospitalized in 2010–2020 (n = 3406), and the temporal validation cohort included those hospitalized in 2021–2023 (n = 1366). Known AF (KAF) was defined as AF documented before stroke or within 14 days after stroke; AFDAS was defined as newly documented AF beyond 14 days among patients without KAF. Multivariable logistic regression restricted to KAF-free patients was used for score derivation. Discrimination, calibration, and time-to-AFDAS risk stratification were assessed. Results Among KAF-free patients, AFDAS was detected during routine clinical follow-up in 176 of 2175 (8.1%) in the development cohort and 148 of 1366 (10.8%) in the validation cohort. The final PAFIS score included age ≥ 75 years, female sex, valvular heart disease, left atrial diameter ≥ 40 mm, and tricuspid regurgitation peak gradient ≥30 mmHg. AUCs were 0.72 (95% CI, 0.68–0.76) in development and 0.65 (95% CI, 0.60–0.70) in validation. PAFIS outperformed CHA₂DS₂-VASc, HAVOC, and AF-ESUS, but not Brown ESUS-AF. Observed AFDAS detection rates increased across risk groups in both cohorts. Conclusions PAFIS provides a simple AFDAS-specific tool for selective post-stroke rhythm monitoring. Because AF ascertainment was based on routine clinical care without standardized prolonged monitoring, PAFIS predicts AF detection under routine practice rather than true AFDAS incidence. Prospective multicenter validation with standardized monitoring is warranted.