Diagnostic accuracy of the Confirm-Rx atrial fibrillation detection algorithm in real-world patients

Abstract

Abstract Funding Acknowledgements Type of funding sources: None. Introduction Continuous rhythm monitoring with implantable cardiac monitors (ICMs) is commonly used to detect and characterise atrial fibrillation (AF) episodes. The Confirm Rx™ with SharpSense™ technology offers four new discriminators and second pass analysis aimed at enhancing detection and improving accuracy for cardiac arrhythmias. This study sought to investigate the diagnostic accuracy of the Confirm Rx™ AF detection algorithm in a UK cohort of ‘real-world’ patients. Methods This multicentre retrospective study included patients with Confirm Rx™ and at least one episode of AF>6 minutes from August 2018 to August 2021 across 9 UK hospitals. Each episode had a corresponding 120-second electrogram (EGM) and heart rate scatterplot. One cardiologist manually adjudicated all EGMs as ‘True-AF’ or ‘False-positive. To assess for intra and inter-observer variability, 10% of the EGMs were reviewed. Diagnostic accuracy was determined by calculating the raw and patient-averaged positive predictive value (PPV) for AF episode of different durations and implant indications. Results During the study 232 patients met inclusion criteria with a total of 315 patient-years of follow-up. 16,320 individual AF episodes were adjudicated; intra- and interobserver variability was excellent (Cohen’s kappa 0.85 and 0.86, respectively). The rate of ‘True-AF’ detection was 3.19 episodes per month corresponding to a raw PPV of 74.5% for the whole cohort. The highest number of episodes per months was observed in patients with a Confirm-Rx for palpitations (5.1) and suspected AF (5.8) but only approximately half of these represented ‘True-AF’ episodes (Figure 1). Patients with known AF had the lowest rate of AF episodes (1.6 episodes per month) but the highest proportion of ‘True-AF’ episodes (PPV of 95.5%). A clear trend of improving diagnostic accuracy was seen with longer AF episodes (Table1). AF>3 hours had a PPV above 94% and all episodes lasting longer than 24 hours were ‘True-AF’. For AF episode of short duration, the PPV varied with the population being monitored; however, for longer AF episodes the PPV increased significantly and irrespective of implant indication (Figure 2). Conclusion Overall, the Confirm Rx™ ICM diagnostic accuracy was modest for all AF episodes lasting longer than 6 minutes (74.5%) but improved considerably for longer AF episodes irrespective of implant indication.