P2840Discrimination between sinus rhythm and atrial fibrillation/flutter: reliability of seven different built-in automatic-diagnostic computer programs in a cohort of >2000 12-lead electrocardiograms

Abstract

Abstract Introduction Atrial fibrillation and flutter (AF/AFl) are the more common sustained arrhythmias in the elderly. ESC guidelines underline the need for large-scale screening strategies, especially to improve primary prevention of thromboembolic complications. However, the current gold-standard for identification of AF/AFl is a 12-lead ECG reviewed by an appropriately trained physician. The availability of automatic discrimination between AF/AFl and sinus rhythm (SR) by automatic- diagnostic computer programs (ACP) implemented in current 12-lead ECG recorders is a possible solution to improve this process. Aim To assess the reliability and agreement of the main world-wide available ACPs implemented in current 12-lead ECG recorders in discriminating between AF/AFl vs. SR in a large dataset of real-world ECGs. Methods We assessed seven ECG interpretation programs from seven different manufacturers (GE 12SL, Glasgow, MEANS, Midmark, Mortara VERITAS, Philips DXL and Schiller). We created a large set of representative ECGs converted from previously recorded digital ECGs acquired with equipment that complied with the requirements of International Electrotechnical Commission standard IEC 60601–2-51:2003 and were representative of those in hospital settings. We excluded ECGs from pacemaker carriers. We used a specific device for playing back ECGs to 12-lead ECG recorders implementing the seven programs. Each statement from automatic diagnosis provided by each device was recorded and combined appropriately for the purpose of this analysis: identification of AF/AFl vs. SR. Gold standard was built by independent re-assessment by three different reviewers. Results We collected 2064 10s 12-lead ECGs with SR (1882) or AF/AFl (182) that were analyzed by seven different ACP. ECG's with other arrhythmias were excluded for this analysis (to increase transferability of the results). All seven programs agreed on SR in 1645 (87.4%) and AF/AFl in 139 cases (76.4%) (Figure 1, panel A). In 280 cases (13.6%), at least one program did not agree with the others. After revision by cardiologists 237 were found to be SR and 43 AF/AFl. Sensitivity for AF/AFl ranged between 90%-97% and false positive diagnosis ranged between 3.4% and 0.4%. Notably, the chance of obtaining at least a wrong diagnosis from one device was 280/2064 (13.6%), with a number of possible false AF/AFl greater than real prevalence of AF/AFl (Figure 1, panel B). Figure 1 Conclusions Despite a general good reliability of each single ACP for AF/AFl recognition the chance of between-device discordance is not negligible and the risk of false positive automatic diagnosis of AF/AFl should be considered when managing real-world patients especially when deciding to start oral anticoagulation.