[Predicting atrial fibrillation through a sinus-rhythm electrocardiogram; useful or not?].
Journal:
Nederlands tijdschrift voor geneeskunde
Published Date:
Nov 28, 2019
Abstract
In patients with cryptogenic stroke, the detection of atrial fibrillation (AF) is important, since it is an indication for the prescription of oral anticoagulation, instead of anti-platelet therapy, to decrease the chance of a recurrent ischaemic cerebral event. The presence of permanent AF is easily detected by means of an electrocardiogram (ECG). However, in paroxysmal AF patients, expensive long-term rhythm monitoring might be necessary to detect the arrhythmia. In the present article, the authors describe an algorithm, constructed by artificial intelligence and big data, to detect a 'footprint' of AF on a 12-lead ECG during sinus rhythm. However, despite this promising premise, the algorithm comes from a 'black-box' calculation and, therefore, it cannot rule out possible bias from medication or patient characteristics. Furthermore, the combination of a moderate test specificity and a low prevalence of only 10% undetected AF in the post-stroke population, results in a low positive predictive value, which is not useful for initiation of anticoagulation therapy.