Detection of Atrial Fibrillation in a Large Population using Wearable Devices: the Fitbit Heart Study

Presented at the American Heart Association’s Scientific Sessions 2021 by: Steven Lubitz - Boston, MA, USA.

Introduction and methods

Aim of the study

Early detection of atrial fibrillation (AF) and flutter may prevent morbidity. Smartwatches and fitness trackers are commonly used and these may have optical photoplethysmography (PPG) sensors to measure heart rate. Presence of AF may be predicted by software algorithms that passively analyze PPG pulse data. Correct classification is important to limit false positives.

Study design

A novel software algorithm was developed with frequent overlapping PPG pulse techogram sampling. The positive predictive value for undiagnosed AF of this algorithm was examined in a large-scale remote clinical trial with a range of wearable devices. Individuals were electronically, remotely enrolled. Those with an irregular heart rhythm detection (IHRD) were notified, invited to schedule a visit with a telehealth provider and received an ECG patch. The ECG patch was self-applied and worn for one week. After return of the ECG patch, a second visit was scheduled to discuss the results. After 90 days, the individuals with an irregular heart rhythm detection were asked to complete a follow-up survey. All participants were asked to complete an end of study survey.

The novel irregular heart rhythm detection algorithm was developed as follows. There was continuous sampling of the pulse with 5 min techograms with a 50% overlap. If 11 of 11 consecutive techograms were irregular (≥30 minutes of irregular rhythm), this was considered an irregular heart rhythm. Data were only analyzed if the individual was inactive. And there was a reset after a negative techogram. 455,699 Participants enrolled during 5 months, 4,728 had a IHRD notification, 1,057 returned a ECG monitor with analyzable data and 340 had AF during ECG monitoring.

Primary outcome

Primary outcome was the positive predictive value for undiagnosed AF of the novel IHRD algorithm.

Main results

• 1.0% of the study population had an IHRD, with a higher percentage in men and in those ≥65 years.
• Among those with an IHRD and who wore an ECG patch, 32% had AF.
• The positive predictive value for AF of the algorithm was 98% and this was consistent across different strata.
• Median burden of AF among those with confirmed AF was 7%. The median duration of the longest AF episode on ECG patch was 7 hours.

Conclusion

A novel PPG software algorithm for Fitbit wearables had a positive predictive value for AF of 98% and may enable large-scale identification of undiagnosed AF.

Discussion

The discussant Prof. Sana Al-Khatib, MD (Duke University, Durham, NC, USA) raised a few points on the Fitbit Heart Study. She said a lot of patients were lost after an IHRD and did not have analyzable data of an ECG monitor. She wondered what happens when individuals are active, because the algorithm did not use these data and sympathetic activation can promote AF. Furthermore, the study did not test whether screening for AF resulted in management of patients and improved outcomes.

• Our reporting is based on the information provided at the American Heart Association’s Scientific Sessions 2021 -