Wearable-Grade Lead Reduction Disproportionately Degrades ECG AI Performance in Elderly Patients: Evidence from PTB-XL and MIT-BIH

The study examined the impact of reducing the number of electrocardiographic leads on artificial intelligence performance in cardiac disease diagnosis across different age groups. The research analyzed 21,091 ECG recordings from the PTB-XL database using neural networks with various lead configurations (12, 6, 2, and 1 lead). With the full 12-lead configuration, model accuracy declined from 84.5% in patients under 40 years to 66.2% in patients over 75 years. When reduced to a single lead, accuracy decreased by 14.1 percentage points in the 75+ age group but only 0.4 percentage points in the under 40 group, representing approximately a 40-fold difference. Older patients exhibited greater diagnostic complexity with multiple conditions. Results were confirmed on the independent MIT-BIH database. The study recommends that age-stratified performance reporting should be a minimum standard in wearable ECG device validation and regulatory assessment.