Portable AI-Assisted Electrocardiograms May Successfully Predict Heart Failure

A recent multinational cohort study has indicated that wearable, single-lead electrocardiograms (ECGs) utilizing noise-resilient AI models can potentially be used to predict heart failure (HF).

Heart failure is one of the greatest public health issues in the world, affecting roughly 64 million people. This number is predicted to rise as the world population ages. However, despite the accessibility of therapies to modify the disease trajectory, strategies for heart failure stratification are still limited.2

Artificial intelligence has been proven in previous studies to be capable of detecting otherwise hidden cardiovascular disease signatures from ECGs. However, the portable ECGs utilizing these models are prone to noise introduction, which can limit the effectiveness of AI unless it is specifically built to account for this noise. This study makes use of AI models developed in the presence of randomized noise, thereby accounting for real-world noise levels.1

“While several clinical risk scores have been proposed to identify individuals at high risk, these strategies often require clinical evaluation and blood testing. This limits their scope to patients with established access to health care services,” wrote Lovedeep Singh Dhingra, MBBS, Yale School of Medicine, and colleagues. “In contrast, our AI-based approach using single-lead ECGs may offer a means for HF risk stratification outside clinical settings.”1

Dhingra and colleagues utilized 3 cohorts between the US, UK, and Brazil, collecting participants who had undergone a conventional ECG. A total of 255,604 participants had at least 1 outpatient ECG – 47,720 were excluded in the model development population and 11,954 with prevalent HF. 1590 patients with LV dysfunction and 1673 with an NT-proBNP level >300 pg/mL were excluded as well, leaving the study with 192,667 participants from the US, 42,141 from the UK, and 13,454 from Brazil.1

The first screening saw 42,775 patients marked as positive at baseline, which was associated with more than a 5-fold increase in HF development risk (hazard ratio [HR], 5.05: 95% CI, 4.73-5.39). After calculating for risk factors such as hypertension, type 2 diabetes, and prior ischemic heart disease (aHR, 2.81: 95% CI, 2.63-3.01), as well as the competing risk of death (aHR, 2.73: 95% CI, 3.10-3.54), this association remained statistically significant.1

Among the UK patients, 5513 screened positive with the AI-ECG model; a positive screening was associated with a 7.5-fold higher HF development risk (HR, 7.52: 95% CI, 4.21-13.41). From the Brazil cohort, 1928 had a positive scan, indicating a 9-fold higher HF risk (HR 8.74: 95% CI, 4.13-18.48). Over a median 4.6-year follow-up, around 3697 individuals were hospitalized for HF, 7514 had an HF hospitalization or an LVEF <50% on a later echocardiogram, and 10,381 died.1

When the team utilized the AI-ECG predictor alongside existing methods of detecting HF, such as PREVENT and PCP-HF, the AI-ECG exhibited an increased discrimination improvement of .091 vs PCP-HF’s .205 and .068 to PREVENT’s .192. It also showed a net reclassification improvement of 18.2% to 47.2% vs PCP-HF and 11.8% to 47.5% vs PREVENT.1

Dhingra and colleagues indicated that the portability of the single-lead ECG, coupled with its ease of use and quick acquisition, could facilitate a non-laboratory-based strategy. The team pointed out that its inherent scalability could also be adapted to local health promotion strategies for those who are less likely to seek out preventative care.1

“With the increasing availability of single-lead ECGs transmitted from portable and wearable devices, future studies are required to determine if this AI-ECG-based noninvasive digital biomarker can enable improved stratification of HF risk across communities,” Dhingra and colleagues wrote.1

References

1. Dhingra LS, Aminorroaya A, Pedroso AF, et al. Artificial Intelligence–Enabled Prediction of Heart Failure Risk From Single-Lead Electrocardiograms. JAMA Cardiol. Published online April 16, 2025. doi:10.1001/jamacardio.2025.0492

2. Shahim B, Kapelios CJ, Savarese G, Lund LH. Global Public Health Burden of Heart Failure: An Updated Review. Card Fail Rev. 2023;9:e11. Published 2023 Jul 27. doi:10.15420/cfr.2023.05