57Accuracy of automated atrial fibrillation detection of a single-lead electrocardiogram device in a pharmacy-based mass screening. (18th June 2020)

• Record Type: Journal Article
• Authors: Zink, M D; Mischke, K G; Rummey, C; Keszei, A; Marx, N
• Is Part Of: Europace. Volume 22(2020)Supplement 1
• Publication Date: 2020-06-18
• Subjects: Arrhythmia -- Treatment -- Periodicals; Cardiac pacing -- Periodicals; Catheter ablation -- Periodicals; Heart -- Physiology -- Periodicals; Electrophysiology -- Periodicals; 617.4120645
• DOI: 10.1093/europace/euaa162.262
• Languages: English
• ISSNs: 1099-5129

Abstract

Abstract: Funding Acknowledgements: The investigators received an unrestricted research grant by Pfizer/BMS. Pfizer/BMS was not involved in the planning, conduct, analysis, or interpreta Background: Recently, various devices specifically designed to automatically detect atrial fibrillation, became available. Utility of automated atrial fibrillation detection in mass-screening remains unclear, because accuracy and validation of automated AF detection were mostly tested retrospectively in preselected cohorts with high atrial fibrillation prevalence. Purpose: The aim of this analysis was to determine the accuracy of automated atrial fibrillation detection by a single-lead hand-held ECG device in a prospective population-based screening approach with several thousand subjects. Methods: For automated atrial fibrillation detection, we investigated the MyDiagnostick a 30 cm long hand-held ECG-stick. All customers aged 65 years or older in the pharmacies were eligible to participate. After informed consent was obtained, a one-minute single-lead ECG was recorded by trained staff. Five levels of human ECG expert validation were performed to determine heart rhythm. All recorded unfiltered ECGs (Figure 1) were analyzed at least by two independent, blinded investigators, in case of an automated AF detection by at least 3 independent, blinded investigators. In case of contradiction in the first three levels of heart rhythm assessment, ECGs were analyzed in level 4 by two independent cardiologists. In level 5 all single-lead ECGs were unblinded and any single-lead ECG with a contradiction between the human ECG expert results and the automated ECG-Stick result were analyzed by a rhythm consortium consisting of 3 cardiologists. Final results for heart rhythm and ECG quality were stipulated by a hierarchical approach, whereas results in higher levels of analysis vote down results in earlier levels of analysis in case of a contradiction. Results: In total, 7209 single-lead ECGs were available for analysis. The mean age was 74 ± 6 years, 42% (N = 3049) of participants were male, mean CHA2DS2-VASc score was 3.3 ± 1.4. We identified 5.3% (N = 383) patients with atrial fibrillation identified by human investigators. The ECG-Stick diagnosed 79% (N = 304) of the atrial fibrillation patients correctly. In 21% (N = 79/383) the result was false-negative, and in 1% (N = 72/7209) false-positive. In another 1% (N = 81/7209) single-lead ECG recordings were uninterpretable by expert investigators, although the automated analysis provided a result (SR: 27, atrial fibrillation: 54). Using final level results by expert ECG reading as reference, automated ECG-Stick analysis showed good prediction in receiver operating calculation of atrial fibrillation with an area under the curve of 0.89, sensitivity: 80%, specificity: 98%, and a good interclass correlation coefficient of 0.75, 95%CI (0.74; 0.76)). Conclusions: The results encourage the implementation of single-lead ECG devices for automated atrial fibrillation screening. A rhythm diagnosis by single-lead ECG seems possible for recordings with excellent ECG signal quality.