Atrial fibrillation has historically been a challenging target for preventive cardiology. Atrial fibrillation is common, often paroxysmal, and might not present any symptoms until individuals experience complications such as stroke or heart failure. Traditional methods of detecting atrial fibrillation involve clinic-based electrocardiograms (ECGs), ambulatory monitoring and implanted devices. Although these approaches are useful, they are constrained by point-in-time assessment, limited monitoring duration, or invasiveness and are therefore poorly suited to a rhythm disorder that can appear unpredictably and resolve before clinical evaluation.

In 2019, the seminal Apple Heart Study by Marco Perez and colleagues offered a new model for atrial fibrillation screening at an unprecedented scale by investigating whether passive monitoring using consumer smartwatches could identify atrial fibrillation during daily life. In this siteless, pragmatic study, the investigators used an innovative approach to cardiovascular research by remotely enrolling a total of 419,297 individuals across the USA through the use of a smartphone application. The participants did not have a previous diagnosis of atrial fibrillation, as reported by the participants themselves. Those who received an irregular pulse notification from the smartwatch during the study period were then offered a telemedicine consultation and mailed an ECG patch for confirmatory ambulatory monitoring. These participants were followed up for 90 days after the notification, which provided important context for the number of alerts, ECG patch returns and subsequent clinical encounters. Notably, the irregular pulse detection algorithm in the smartwatch prompted an alert in just 0.52% of the participants, suggesting that large-scale heart rhythm surveillance could be feasible without creating an excessive burden of alerts. Of the 450 individuals who received notifications and returned the ECG patches with analysable information, 34% had atrial fibrillation according to the ECG patch, and the positive predictive value of the smartwatch notifications recorded during concurrent ECG patching was 0.84. Importantly, the investigators noted that most of the detected episodes lasted at least an 1 h, which revealed that the algorithm was capturing clinically meaningful rhythm abnormalities rather than trivial signal noise.