Background: Atrial fibrillation (AF) is the most common cardiac rhythm disorder associated
with stroke. Increased risk of stroke is the same regardless of whether the AF is permanent
or paroxysmal. However, detecting paroxysmal AF is challenging and resource intensive. We
aimed to develop a predictive model for AF in patients with acute ischemic stroke, which could
improve the detection rate of paroxysmal AF.
Methods: We analyzed 10,034 adult patients with acute ischemic stroke. Differences in clinical
characteristics between the patients with and without AF were analyzed in order to develop a predictive
model of AF. The associated factors for AF were analyzed using multivariate logistic regression
and classification and regression tree (CART) analyses. We used another dataset, which
enrolled 860 acute ischemic stroke patients without AF at baseline, to test whether the developed
model could improve the detection rate of paroxysmal AF. Among the study population, 1,658 patients
(16.5%) had AF.
Results: Multivariate logistic regression revealed that sex, age, body weight, hypertension, diabetes
mellitus, hyperlipidemia, pulse rate at admission, respiratory rate at admission, systolic blood
pressure at admission, diastolic blood pressure at admission, National Institute of Health Stroke
Scale (NIHSS) score at admission, total cholesterol level, triglyceride level, aspartate transaminase
level, and sodium level were major factors associated with AF. CART analysis identified NIHSS
score at admission, age, triglyceride level, and aspartate transaminase level as important factors for
AF to classify the patients into subgroups.
Conclusion: When selecting the high-risk group of patients (with an NIHSS score >12 and age
>64.5 years, or with an NIHSS score ≤12, age >71.5 years, and triglyceride level ≤61.5 mg/dL) according
to the CART model, the detection rate of paroxysmal AF was approximately double in the
acute ischemic stroke patients without AF at baseline.