Digital biomarkers for atrial fibrillation
The clearest success story in consumer digital health, and a case study in what happens when detection outruns the clinical pathway.
%20(1).png)
Atrial fibrillation is the clearest success story in consumer digital health measurement, and also a case study in what happens when detection outruns the clinical pathways behind it. Wrist worn devices can now record a single lead electrocardiogram and screen continuously for irregular rhythm, and large prospective studies have shown that they identify previously undiagnosed atrial fibrillation in the general population.
The measures used in this area go beyond detection. Once monitoring is continuous, the more useful quantity is burden: what proportion of a monitored period a person spent in atrial fibrillation, and how episodes are distributed. Burden can rise and fall with treatment, which makes it usable as a response endpoint, whereas a binary detection cannot change once it has occurred.
Alongside rhythm, this category includes the rate and physiological measures used to characterise the condition and its consequences, since heart rate control and symptom burden are both treatment targets in their own right and are measured continuously by the same devices.
Digital biomarkers used in atrial fibrillation research
How these measures are used
Three research uses dominate. Screening studies test whether population level monitoring finds clinically relevant undiagnosed atrial fibrillation, and whether finding it improves outcomes, which is a separate and harder question. Post procedure monitoring quantifies recurrence after ablation or cardioversion, where continuous burden measurement is markedly more sensitive than intermittent recording. And anticoagulation strategy trials use burden to test whether treatment can be tied to the amount of arrhythmia rather than given continuously.
In practice these measures are reported alongside a symptom and quality of life instrument. The relationship between detected arrhythmia and how a person feels is weak: many episodes are asymptomatic and many symptoms occur without arrhythmia. Reporting rhythm data alone would misrepresent the patient experience, and reporting symptoms alone would miss the events that carry stroke risk.
What the evidence supports today
Detection is the strongest evidence in this category. Consumer single lead electrocardiography has cleared regulatory review in multiple markets, irregular rhythm notification features have been evaluated in large prospective cohorts, and an atrial fibrillation history feature has been qualified by the United States Food and Drug Administration as a medical device development tool, which is a meaningful step towards regulated trial use.
Burden measurement is less settled. It depends on how much of the time the device was actually monitoring, and consumer devices sample intermittently rather than continuously, which means burden estimates from a smartwatch and from an implanted monitor are not equivalent quantities even when reported in the same units.
The open clinical question is not whether these devices detect atrial fibrillation. It is what to do about brief, asymptomatic, device detected episodes, where the evidence on whether anticoagulation improves outcomes is still developing and screening at scale generates findings that clinical pathways are not yet designed to absorb.
Common questions
Can a smartwatch diagnose atrial fibrillation?
It can flag a rhythm consistent with atrial fibrillation and prompt assessment. It does not diagnose. Confirmation requires clinical review, typically with a conventional electrocardiogram or a period of ambulatory monitoring, and a notification on its own is not a basis for starting treatment.
What is atrial fibrillation burden?
The proportion of a monitored period spent in atrial fibrillation, rather than whether an episode was ever detected. It is more useful as a research endpoint because it can increase or decrease in response to treatment, which a one time detection cannot.
Are burden figures from a smartwatch comparable to those from an implanted monitor?
No. Implanted devices monitor continuously; consumer wearables sample intermittently and only when worn. Both may report a percentage, but the denominators differ, and treating the two as interchangeable will misstate the result.
Does finding more atrial fibrillation improve outcomes?
That is the honest open question. Screening reliably finds previously undiagnosed cases, but whether treating brief, asymptomatic, device detected episodes reduces stroke without adding bleeding risk is still being established. The measurement problem has been solved considerably faster than the clinical decision problem behind it.
Run a study on these measures
WeGuide captures wearable data and patient reported outcomes in one platform, from screening through to analysis.
