AimsTo evaluate the performance of the ABC (Age, Biomarkers, Clinical history) and CHA₂DS₂-VASc stroke scores under real-world conditions in an emergency setting.Methods and ResultsThe performance of the biomarker-based ABC-stroke score and the clinical variable-based CHA₂DS₂-VASc score for stroke risk assessment were prospectively evaluated in a consecutive series of 2,108 patients with acute symptomatic atrial fibrillation at a tertiary care emergency department. Performance was assessed according to methods for the development and validation of clinical prediction models by Steyerberg et al. and the Transparent Reporting of a Multivariable Prediction Model for Individual Prognosis or Diagnosis. During a cumulative observation period of 3,686 person-years, the stroke incidence rate was 1.66 per 100 person-years. Overall, the ABC-stroke and CHA₂DS₂-VASc scores revealed respective c-indices of 0.64 and 0.55 for stroke prediction. Risk-class hazard ratios comparing moderate to low and high to low were 3.51 and 2.56 for the ABC-stroke score and 1.10 and 1.62 for the CHA₂DS₂-VASc score. The ABC-stroke score also provided improved risk stratification in patients with moderate stroke risk according to the CHA₂DS₂-VASc score, who lack clear recommendations regarding anticoagulation therapy (HR: 4.35, P = 0.001). Decision curve analysis indicated a superior net clinical benefit of using the ABC-stroke score.ConclusionIn a large, real-world cohort of patients with acute atrial fibrillation in the emergency department, the ABC-stroke score was superior to the guideline-recommended CHA₂DS₂-VASc score at predicting stroke risk and refined risk stratification of patients labeled moderate risk by the CHA₂DS₂-VASc score, potentially easing treatment decision-making.