A novel artificial intelligence (AI) model applied to a 12-lead electrocardiogram (ECG) image was able to identify individuals at elevated risk of heart failure (HF) across multinational cohorts without HF at baseline, according to a study published Jan. 13 in European Heart Journal.

Lovedeep S. Dhingra, MBBS, Harlan M. Krumholz, MD, SM, FACC, et al., evaluated whether an AI-ECG model developed to detect signatures of left ventricular systolic dysfunction on an ECG image at baseline could predict patients at high risk of new-onset HF, measured as the first HF hospitalization, across three geographically and clinically distinct populations: the Yale New Haven Health System (YNHHS) cohort, the UK Biobank (UKB) cohort and the Brazilian Longitudinal Study of Adult Health (ELSA-Brasil) cohort.

The YNHHS cohort consisted of 231,285 patients (median age 57 years, 56.6% women, 37.0% non-White). Over a median follow-up of 4.5 years, 4,472 (1.9%) had a primary HF hospitalization, 9,645 (4.2%) had a primary HF hospitalization or an echocardiogram with an LVEF <50% on subsequent echocardiogram and 17,380 (7.5%) died.

The UKB cohort included 42,141 individuals (median age 65 years, 51.7% women, 96.6% White). Over a median follow-up of 3.1 years, 46 (0.1%) had an HF hospitalization event and 346 (0.8%) died. The ELSA-Brasil cohort comprised 13,454 participants (51 years, 54.6% women,15.8% Black and 28.0% Brazilian ‘Pardo’). A total of 31 individuals developed HF and 229 died over a median follow-up of 4.2 years.

Results in the YNHHS cohort showed that 17,868 (7.7%) screened positive with the AI-ECG model, and this was associated with a >6.5-fold higher risk of incident HF (hazard ratio [HR], 6.51; 95% CI, 6.11-6.93). A positive AI-ECG screen, compared with a negative screen, was associated with a nearly four-fold risk of incident HF,  after adjusting for differences in age and sex (adjusted HR [aHR], 3.88; 95% CI, 3.63-4.14), as well as additionally accounting for differences in baseline HF risk factors for hypertension and diabetes (aHR, 3.73; 95% CI, 3.50-3.99). When adjusting for death, age and sex, a positive screen was associated with an aHR of 3.54 (95% CI, 3.30-3.79) for incident HF. Of note, among patients younger than 65 years, there was an eight-fold higher adjusted risk of incident HF associated with a positive screen.

The findings were consistent across all demographic subgroups. Similar results were seen across the UKB and ELSA-Brasil cohort, with a positive AI-ECG screen associated with an 18-fold and 24-fold higher risk of incident HF, respectively.

The authors write that, "As a digital biomarker of HF risk that requires just an ECG image, this AI-ECG approach can enable scalable and efficient screening for HF risk."

In an accompanying editorial comment, Charalambos Antoniades, MD, PhD, and Kenneth Chan, MBBS, discuss the growing role of AI for screening, diagnosis and risk stratification in HF. They note that "the ability to predict future HF from ECG waveforms, as presented by Dhingra, et al., is a major step towards implementing low-cost population screening in search of the hidden signs of HF."