Prediction of early and long-term atrial fibrillation in ischemic stroke patients

Abstract Background Prolonged cardiac surveillance is administered to a minority of patients who have experienced an ischemic stroke. Enhancing diagnostic yield may facilitate early intervention strategies that improve outcomes. Purpose This study aims to develop a risk score to predict incident atrial fibrillation (AF) in ischemic stroke patients without a prior history of AF. Methods This retrospective cohort study analyzed electronic health records of ischemic stroke patients at our university hospital from 2010 to 2020. Patients with AF-related strokes were identified, and predictors of AF development were evaluated using multivariable logistic regression and survival analysis. Results Among 3,406 ischemic stroke patients, 1,231 had AF before or shortly after their stroke, while 176 developed AF during follow-up. The PAFIS (Prediction of AF in Ischemic Stroke) score assigns points as follows: 2 points for age ≥ 75 years, 1 point for age 65-74, 2 points for heart failure, elevated NT-proBNP, or valvular heart disease; 3 points for pulmonary hypertension; 5 points for hyperthyroidism; and 6 points for left atrial enlargement. Scores are categorized into four classes: I (0-3 points), II (4-5 points), III (6-10 points), and IV (>10 points), corresponding to AF incidences of 2.6%, 8.7%, 14.3%, and 27.5%, respectively. Kaplan-Meier analysis indicated a significant association between the PAFIS score and incident AF, cardiovascular mortality, and overall mortality (log-rank p<0.001). The PAFIS score achieved an AUC of 0.735 [0.697-0.773] for predicting long-term AF. Conclusion The PAFIS score effectively predicts AF in ischemic stroke patients, enabling timely identification of high-risk individuals.