Artificial Intelligence-Assisted ECG in a Hub-and-Spoke Network in India: Real-World Performance in Acute Coronary...

Acute coronary syndrome (ACS) remains a leading cause of morbidity and mortality in India, with delayed diagnosis contributing to poor outcomes. The 'Heart Beat' project - an AI-assisted hub-and-spoke model was implemented to improve early and accurate ACS detection across diverse healthcare settings. In this retrospective, real-world observational analysis, we leveraged data from this initiative to (1) assess the prevalence of cardiovascular diseases (CVD) in the Indian population using AI-powered electrocardiogram (ECG) interpretation and (2) evaluate the turnaround time (TAT) for ECG performance and diagnosis.

This report utilized data from the Heart Beat project, collected between January 2020 and December 2022. We included patients presenting with chest pain at primary healthcare centers (spokes), linked to centrally located hospitals (hubs) equipped with a catheterization laboratory (Cath-lab). AI-assisted 12-lead ECGs were performed at spokes, and data were transmitted to hubs for centralized analysis. ECG findings were categorized as Abnormal, Borderline, Critical, Normal, Otherwise Normal, or Pacemaker based on AI interpretation. Descriptive statistics were employed to evaluate (1) the prevalence of CVD patterns and (2) the TAT for ECG acquisition-to-diagnosis. Findings were analyzed using descriptive statistics.

A total of 45,488 ECGs were obtained from 66 spokes, which were connected to 10 hubs across six Indian states. The most commonly reported ECGs were Normal (n=22987, 50.53%), Abnormal (n=19396, 42.64%), and Critical (n=2992, 6.58%). Most of the CVD conditions were diagnosed from Critical and Abnormal ECGs. Out of the seven CVD conditions analyzed from the dataset, left ventricular hypertrophy (LVH; n=3993, 8.78%) was the most frequently reported. Higher ST-elevation myocardial infarction (STEMI) was identified in 231 cases (0.51% of total ECGs), whereas non-STEMI (NSTEMI) was identified in 22 cases (0.05% of total ECGs) among the total ECGs recorded. The mean (SD) TAT for ECGs of all patients was 5.12 (9.61) minutes, with the minimum mean TAT recorded for Critical (2.91 minutes) and the maximum for Normal (5.61 minutes) ECGs.

The findings highlight a higher prevalence of STEMI over NSTEMI cases, emphasizing the need for adequate resources at healthcare centers for timely and effective management. An average lower ECG TAT (5 minutes) further supported AI-assisted ECG's potential in the early diagnosis of ACS patients.