Advanced Diagnosis of Hypertrophic Cardiomyopathy with AI-ECG and Differences Based on Ethnicity and HCM Subtype

Background/Objective: Hypertrophic cardiomyopathy (HCM) often presents later in the disease course, with frequent misdiagnoses and population-level underdiagnoses. Underserved patients may have even greater diagnostic delays. We aimed to test the hypothesis in a retrospective cohort that artificial intelligence analysis of ECGs (AI-ECG) could have afforded the opportunity for earlier diagnosis of HCM in one health system. Methods: We collected all available ECGs from patients referred to an HCM Center of Excellence over 15 years, both before and after HCM diagnosis. We applied AI-ECG to each ECG in a blinded fashion to predict the probability of HCM. We calculated the time between each patient's AI-ECG diagnosis and clinical diagnosis. We examined the sensitivity and specificity of AI-ECG for all patients, and by septal subtype and genetic test result. Results: 3499 ECGs were analyzed in 404 patients (age 56 ± 18 years, 52% female). AI-ECG correctly identified HCM in 155 patients with a sensitivity of 67%, specificity of 95%, positive predictive value of 94%, and a negative predictive value of 69%. The AUC was similar using mean probability from all ECGs for each patient (AUC 0.91 [0.88, 0.94]) or using probability from the first ECG (AUC 0.91 [0.87,0.93]). AI-ECG diagnosed 27 patients over 1 year before clinical diagnosis, and up to 16.3 years early. Black patients were more likely than White patients to have an AI-ECG diagnosis before a clinical diagnosis (p = 0.005). Conclusions: AI-ECG offers the potential for advanced HCM diagnosis. Differences in identification timing between subgroups highlight inequities in current care and show the potential of AI-ECG for the greatest benefit in underserved ethnic groups.