Impact of artificial intelligence–enabled ECG implementation on diagnostic and longitudinal surveillance guideline adherence in hypertrophic cardiomyopathy

Background Adherence to guideline-directed surveillance in hypertrophic cardiomyopathy (HCM) is critical to optimize risk stratification and reduce clinical outcomes. Artificial intelligence-enabled electrocardiogram (AI-ECG) can identify individuals who are subsequently diagnosed with HCM. However, its role in optimizing care and longitudinal surveillance in HCM patients is not well defined. Purpose Evaluate the impact of AI-ECG alerts in longitudinal diagnostic surveillance care in accordance with HCM clinical guidelines. Methods Between March and September, 2024, 957 consecutive HCM patients were alerted for suspected HCM by an FDA-approved AI-ECG algorithm (Viz HCM, Viz.ai). Among these, we reviewed 75 consecutive patients with known HCM and no extensive missing chart data (54 diagnosed prior to AI-ECG alert and 21 after AI-ECG alert was implemented). Guideline adherence was determined based on timeline recommendation for diagnostic surveillance and compared between tertiary center and community center. Results Among 75 patients, median age was 65 (50-74), 38 (51%) were female, 48 (65%) presented obstructive phenotype, 48 (70%) were symptomatic at initial evaluation, mostly in NYHA class II. At follow-up, only 36% ambulatory monitoring, 20% stress echocardiogram and 21% cardiac MRI were performed as recommended by guidelines. (Table1). HCM patients evaluated in community centers were younger and underwent less initial diagnostic testing including baseline ECG, echocardiogram, ambulatory monitoring and cardiac MRI. Longitudinal surveillance showed lower annual clinical assessment (61% vs 87%, p=0.029) and a trend to lower annual ECG for those evaluated in the community compared to tertiary center. Time from diagnosis to surveillance testing did not differ significantly between groups, although a trend toward longer time to annual testing was observed among patients evaluated in community centers compared with the tertiary center. Following AI-ECG implementation, median time from diagnosis to annual ECG and echocardiography was shorter compared with the pre-AI cohort (Table 2). Conclusions Our study demonstrates contemporary gaps in longitudinal guideline-directed diagnostic surveillance. These findings highlight actionable opportunities to optimize patient care, including systematic re-engagement of patients lost to follow-up, standardized surveillance intervals for imaging, and ambulatory monitoring. An AI-ECG alert system may serve as a scalable quality surveillance tool to proactively identify care gaps and facilitate timely, guideline-based management.