Large-Scale Assessment of a Smartwatch to Identify Atrial Fibrillation
Level 3 - non-randomized controlled study
Prospective single-arm cohort study without a randomized or concurrent unexposed comparison group.
OpenAlex W2987985090 · doi:10.1056/nejmoa1901183
What was done
In a prospective, single-arm, pragmatic screening study, 419,297 self-reported atrial fibrillation-free participants used a smartwatch app to monitor for irregular pulses over a median of 117 days. When an irregular pulse notification was triggered, participants underwent a telemedicine visit and received a mailed 7-day single-lead ECG patch to confirm arrhythmias. Follow-up surveys were administered at 90 days. Key endpoints included the proportion of notified participants with atrial fibrillation detected on subsequent ECG patches and the positive predictive value of smartwatch alerts concurrent with ECG patch recordings.
What was found
Across 419,297 participants, 2,161 (0.52%) received irregular pulse notifications. Among 450 notified participants who returned analyzable ECG patches (applied on average 13 days post-notification), atrial fibrillation was identified in 34% (97.5% CI, 29 to 39) overall and in 35% (97.5% CI, 27 to 43) among those 65 years or older. The positive predictive value was 0.84 (95% CI, 0.76 to 0.92) for observing atrial fibrillation on simultaneous ECG during a subsequent irregular pulse notification, and 0.71 (97.5% CI, 0.69 to 0.74) for irregular tachograms. Of 1,376 notified participants returning 90-day surveys, 57% contacted healthcare providers outside the study.
Why it matters
This trial established the feasibility of large-scale, siteless digital health screening using consumer smartwatches, demonstrating high diagnostic concordance between real-time irregular pulse alerts and simultaneous ECG-confirmed atrial fibrillation.
Limits
Substantial loss to follow-up occurred, with only 450 of 2,161 notified participants (20.8%) returning analyzable ECG patches. The study lacked an unnotified control group, precluding calculation of sensitivity, specificity, and false-negative rates. Additionally, baseline atrial fibrillation status was self-reported, and a median 13-day delay occurred between pulse notification and patch application.
Cited by
- supports Smartwatches can detect atrial fibrillation and measure electrocardiograms.