Abstract
Background
Atrial fibrillation (AF) is a common cause of ischaemic stroke and a major global health burden, but is often undetected in routine clinical practice. Detection of AF is essential for secondary stroke prevention, but it remains unclear which patients who had a stroke are at highest risk of previously unrecognised or newly detected AF, limiting targeted detection strategies. The aim of this study was to determine the prevalence of AF and risk factors that are associated with AF among patients who had an ischaemic stroke.
Methods
In this retrospective cohort study, all patients who had an ischaemic stroke at the Radboud University Medical Center (Nijmegen, the Netherlands) from January 2019 to October 2024 were included. We applied descriptive statistics, logistic regression analyses and generative topographic mapping (GTM), a probabilistic machine learning method for clustering and data visualisation.
Results
Of 1836 patients (median age 73 years (IQR 64–80), 44.1% female), 30.5% had AF, of which 19.7% were previously known and 10.8% were newly detected after stroke. The most significant factors associated with AF were older age, increased stroke severity, history of ischaemic stroke and heart failure, mitral regurgitation, increased left atrial volume and elevated peak E wave velocity. GTM identified a phenotype characterised by older age, severe strokes, high prevalence of history of ischaemic stroke, hypertension, diabetes, high cardiac burden with history of myocardial infarction, heart valve abnormalities and heart failure, impaired kidney function, high premature atrial and ventricular contraction burden, and signs of severe left atrial remodelling to be most associated with AF (prevalence 47.9%).
Conclusions
The prevalence of AF and stroke severity among AF patients in this study highlights the relevance of optimising monitoring strategies. Stroke severity, history of ischaemic stroke, high cardiac burden and left atrial remodelling signs were most strongly associated with AF. These findings suggest potential areas for targeted screening in high-risk clinical phenotypes.