Digital health for atrial fibrillation detection: A new era in cardiac monitoring after cryptogenic stroke

Abstract

The use of smartphones, smartwatches, and accompanying mobile applications has been growing tremendously over the past years. This implies that mHealth approaches, using these technologies, provide opportunities to change how we deliver care, including detection of heart rhythm disorders such as AF detection. As such, mobile health can be used anytime, anywhere, and with any device, generating a hyperconnected digital world and accelerating the digital transformation in healthcare. In one-third of strokes, no cause can be found, this is called a cryptogenic stroke. However, it is of great importance that the underlying cause of this subtype is identified to enable appropriate treatment and as such, reduce the risk of recurrence (i.e., secondary prevention). The most important possible cause for cryptogenic stroke is the heart rhythm disorder atrial fibrillation. Currently, the guidelines recommend to use an insertable cardiac monitor in these patients. This device continuously monitors the heart rhythm of the patient, and detects arrhythmias for up to three years. These small monitors, which are subcutaneously inserted (i.e., invasive procedure), can reliably estimate the incidence and duration of atrial fibrillation episodes. The aim of this doctoral thesis was to investigate the potential of mobile health technologies in detecting atrial fibrillation in patients with cryptogenic stroke. The first part of this doctoral research found that the recommended long-term cardiac monitoring methods (7-day Holter monitoring and insertable cardiac monitors) were only used in a small portion of cryptogenic stroke patients. This is notable because these long-term methods were shown to detect atrial fibrillation more effectively. Additionally, we identified some patients where an earlier start of long-term cardiac monitoring might have prevented secondary strokes if the arrhythmia had been detected and treated sooner. In the second part of the study, we compared several mobile health applications. These included two methods that use the heart's electrical signals and two methods that measure blood volume changes in the fingertips. According to current guidelines, the latter method cannot yet be used to formally diagnose AF but is approved for AF detection. Despite this, all methods provided highly reliable results compared to a standard electrocardiogram. We further tested one of these apps, the one based on blood volume changes, in cryptogenic stroke patients, comparing its performance to an insertable cardiac monitor. Half of the participants used the app via a smartphone, while the other half used a smartwatch. This study demonstrated that the app could reliably detect episodes of AF lasting several hours, particularly when multiple consecutive measurements indicated the presence of atrial fibrillation. Additionally, it highlighted the limitations of the ‘gold-standard’ insertable cardiac monitors, notably the high number of false positives, which require time-consuming, extensive review by healthcare practitioners. In the final part of the research, we explored whether it is possible to predict which patients are likely to develop AF. Our findings showed that an artificial intelligence algorithm was most effective at making these predictions using electrocardiogram data that appears normal to a trained cardiologist. Combining patient selection to identify those at the highest risk for AF with the strategic use of non-invasive mobile health applications and insertable cardiac monitors could drive the digital transformation of care for cryptogenic stroke patients in the future.