Back

Study protocol for The Heart Watch Study - Prognostic performance of smartwatch 12-lead ECG with advanced ECG analysis in consumer self-screening for cardiovascular disease

Al-Falahi, Z. S.; Schlegel, T. T.; Kozor, R.; Lindow, T.; Ugander, M.

2025-03-06 cardiovascular medicine
10.1101/2025.03.05.25323393 medRxiv
Show abstract

BackgroundIt is now possible to acquire a fully diagnostic standard 12-lead electrocardiography (ECG) using a smartwatch, a smartphone app, and no additional hardware. This study aims to evaluate the predictive performance of advanced ECG (A-ECG) analysis applied to a smartwatch 12-lead ECG (SWECG) acquired using an Apple Watch in identifying cardiovascular risks among asymptomatic adults performing consumer self-screening in the general population. Methods and DesignThe Heart Watch Study is a prospective, Australia-wide observational cohort study. The study will recruit 30,000 participants aged 20-79 years without prior known cardiovascular disease. Participants will enrol and consent electronically, and download a dedicated study iPhone app that enables a directed 12-lead SWECG acquisition using an Apple Watch. The recordings will undergo both conventional and A-ECG analysis. Follow-up will be performed through linkage with administrative health datasets in Australia, including hospital visits and mortality. The primary combined endpoint is all-cause mortality, hospitalization for cardiovascular causes, and incident cardiovascular disease (arrythmia, ischemic heart disease, heart failure) as predicted by existing A-ECG analyses. ConclusionThis study evaluates the feasibility of performing a fully diagnostic standard 12-lead SWECG using only a smartwatch and a smartphone application, and addresses the prognostic value of A-ECG machine-learning based analysis in consumer self-screening in a large, diverse and apparently healthy population.

Matching journals

The top 6 journals account for 50% of the predicted probability mass.

50% of probability mass above

"Similar papers" are the closest papers from that journal in the model's embedding space. They show what the match is built on, but the ranking comes mostly from a classifier over the whole training set, not from these examples alone.