Background <p>Detecting concealed paroxysmal atrial fibrillation after stroke requires elaborate electrocardiographic monitoring. We evaluated previously established predictors to quantify the individual risk of detecting atrial fibrillation within six months in the Find-AF_RANDOMISED-trial.</p> Methods <p>We analyzed 200 patients ≥ 60 years with acute ischemic strokes in the intervention arm of the Find-AF_RANDOMISED-trial. Patients received three ten-day Holter-electrocardiograms within six months. Regression analyses and receiver-operator-characteristics were used to select promising biomarkers and assess predictive performance. We applied previously established cut-offs for the most promising markers to determine those at a high risk of underlying atrial fibrillation.</p> Results <p>27/200 patients (13.5%) had atrial fibrillation after six months. The left atrial diameter, atrial premature beats, supraventricular runs and brain natriuretic peptide were associated with atrial fibrillation, whereas the established markers age and suspected stroke etiology were not. Atrial premature beats differentiated best between those with and without atrial fibrillation (area-under-the-curve = 0.75). Only brain natriuretic peptide ≥ 100pg/ml and supraventricular runs ≥ 20 beats independently predicted atrial fibrillation in multivariable models.</p> Conclusions <p>Supraventricular runs and brain natriuretic peptide were the most promising predictors to define a high risk of underlying atrial fibrillation after stroke in our study. Future screening strategies for atrial fibrillation in stroke patients should focus on these parameters rather than the suspected stroke etiology.</p> Trial registration <p>clinicaltrials.gov NCT01855035 registered 05132013 https//www.clinicaltrials.gov/study/NCT01855035?tab=table.</p>