Sensitive Detection of Atrial Fibrillation in Acute Stroke Patients by Short-Term Bedside Electrocardiography Monitoring Software Analysis

Background: Atrial fibrillation (AF) is an important cause of stroke. Continuous electrocardiography (ECG) monitoring with software-based analysis algorithms has been suggested to enhance the AF detection rate. We investigated the ability of stroke risk analysis (SRA) in the detection of AF in acute stroke patients. Methods: Consecutive stroke patients numbering 1,153 were screened. Patients with cardioembolic stroke related to AF (n = 296, paroxysmal n = 63, persistent n = 233) and patients with cryptogenic stroke (n = 309) after standard diagnostic work-up (bedside ECG monitoring, ultrasound, transesophageal echocardiography, 24 h Holter ECG) received SRA during their stay at the Stroke Unit. Determination of AF risk by SRA in the patients with AF and in the patient group with cryptogenic stroke was assessed and compared. Results: Median SRA monitoring analysis time was 16 h (range 2-206 h, interquartile range 10-36). In AF patients, SRA also detected a possible or definitive AF in 98%. The overall sensitivity of SRA to detect possible or definitive AF in patients with proven AF by standard diagnostic work up and cryptogenic stroke was 98%, specificity 27%, positive predictive value 56%, and the negative predictive value (NPV) was 92%. Area under ROC curve was 0.622. Conclusion: SRA was found to be highly sensitive to detect possible or definitive AF in clinical routine within a short monitoring time. However, low specificity and poor accuracy do not allow diagnosing AF by SRA alone, but with the high NPV compared to current diagnostic standard, it is a valid diagnostic tool to rule out AF. Thereby, SRA is a contribution to clarify stroke etiology. (c) 2018 S. Karger AG, Basel