AI remote ECG monitoring improves arrhythmia detection

Abstract

Abstract Background Ambulatory ECG monitoring is needed to investigate intermittent heart rhythm symptoms. Traditional monitoring approaches are resource intensive, requiring multiple patient hospital visits and time-consuming analysis, using fixed capacity monitoring equipment. Low diagnostic yield means tests are often repeated. Purpose We carried out a large pilot to compare ambulatory cardiac monitoring with an AI-led remote ECG monitoring service (Zio XT) with Holter monitoring. Patients were fitted with single-use Zio monitoring patches, which they wore for 14 days of continuous monitoring and then returned it by Freepost for outsourced analysis and reporting. Methods Over the course of 12 months, 3505 patients were involved in evaluation of Zio. Of these, we reviewed the detailed rhythm data of 200 consecutive patients, and compared this with a retrospective cohort of 200 consecutive standard Holter monitored patients, fitted prior to the evaluation period. All patients had been referred to cardiology for investigation of heart rhythm symptoms. Results The total Zio group had a mean age of 63.8 years. The 400 participants used in the analysis groups matched this (Table) and there was no significant difference in the age ranges between the two groups. Holter duration was for: 24-hr (45%), 48-hr (8%), 72-hr (20%), 5-day (2%), 7-day (26%). Significant arrhythmias (defined as SVT ≥ 30s, VT ≥ 8bts, AV Block Mob 2⁰ or 3⁰, Pause ≥ 3s, Paroxysmal AF) were detected in 49 (25.8%) Zio patients compared with 18 (9.0%) Holter patients (P<0.00001). Total number of arrhythmia findings was greater in Zio compared with Holter; P<0.05 (Table). The arrhythmia findings in Zio were comparable with those in the pilot’s total cohort. In Holter, 41% of significant arrhythmias were detected in patients monitored for 24 hrs, comparable with Zio where 46% of arrhythmias were detected within the first 24 hours of wear-time. In Zio, 15% of significant arrhythmias were detected post 7 days of the wear-time. Conclusions Arrhythmia detection with Zio was significantly greater compared with Holter; detecting significant arrhythmias in nearly 3 times as many patients. Longer detection enabled greater arrhythmia detection, with 3 in every 20 patients having episodes occurring after 7 days. Even when adjusting for wear time, arrhythmia detection within 7 days was superior in the Zio group. Over half (59%) of arrhythmias were detected in patients monitored for 48 hours or longer, suggesting that for an efficient service longer monitoring is needed, particularly when symptoms are intermittent. For Holter monitoring this has considerable resource utilisation implications, relating to staff (cardiac physiologist) time and capital equipment (Holter monitoring stock). Also, greater detection of arrhythmia across the wear time suggests a higher sensitivity than Holter for detecting cardiac arrhythmias.