Comprehensive Summary
Love et al. examined how a commercially available AI-ECG system (Viz HCM) functions when deployed across routine clinical environments rather than in a retrospective model-testing setting. Over the course of 2023, five health systems used the software of 145,848 adults with eligible ECGs, producing alerts for 3% of tracings. Clinicians reviewed roughly two-thirds of these alerts, and 217 individuals without a prior HCM diagnosis ultimately met criteria for structured evaluation. Most of these patients (84%) warranted further testing, generating more than 200 follow-up imaging or clinical assessments. When imaging led to confirmation, the typical turnaround from the AI-flagged ECG to diagnostic clarification was just over a week. Seventeen patients received a newly established HCM diagnosis, representing the actionable diagnostic yield of the workflow. Midway through implementation, the operating threshold of the algorithm was tightened, reducing the alert frequency by about half while maintaining similar proportions of patients moving forward to full evaluation.
Outcomes and Implications
The findings show that AI-enhanced ECG triage can be incorporated into the everyday practice of busy, multi-site health systems, and can reveal previously unrecognized HCM cases in a clinically practical timeframe. By prompting clinicians toward follow-up in patients who otherwise might not elicit additional scrutiny, this approach has the potential to advance earlier identification of a disease with substantial implications for sudden cardiac prevention, family screening, and exercise counseling. However, the modest number of new diagnoses and the absence of direct comparison with standard ECG interpretation limit conclusions about overall effectiveness. The study nonetheless provides an early demonstration of how AI-driven ECG tools may operate at scale and highlights the need for future work assessing cost, workflow impact, and outcome improvement.