Feasibility of Single-Lead Apple Watch Electrocardiogram in Atrial Fibrillation Detection Among Heart Failure Patients

Central Illustration

Patients at a quaternary hospital HF clinic were approached if symptomatic palpitations, post electrical cardioversion, or AF ablation.Exclusion criteria included chronic or persistent AF.Over the 3-month study period, participants emailed AW-generated 1lead ECGs when they experienced symptoms or received an irregular pulse notification from the AW to a shared inbox.One of 3 study cardiologists reviewed each ECG within 48 hours and provided an email response on their assessment and recommendations.All ECGs were subsequently adjudicated by a blinded panel of 4 cardiologists whose consensus served as the comparator to the AW-generated interpretation in performance analysis.Statistical analyses were performed using R (version, 4.3.1).Concordance between AW interpretation and physician interpretation was assessed with Cohen's kappa.Data collection occurred at the time of study enrollment and included patient demographics, questionnaires, and electronic medical record review.Post-study, patients completed the System Usability Scale (SUS)-12 questionnaire to assess AW usability. 3hics approval was obtained from University Health Network REB panel (19-6070).
Of the 208 ECGs, additional health care was sought in 26 cases further outlined in Figure 1A.Estimated costs incurred on the health care system when using current regional fees totaled to $2478 or average $82.6 per patient over the 3-month study period (Figure 1B). 4 The SUS-12 results indicate that while the study population found the system "easy to use" (median score 6 out of 7 on SUS-12 question 2), they also felt the system was "cumbersome to use" (median score 6 out of 7 on SUS-12 question 8). 3 Pearson correlation coefficient analysis revealed a positive user experience association with younger age (r ¼ 0.87; P < 0.0005), higher education (r ¼ 0.65; P < 0.005), and greater household income (r ¼ 0.65; P < 0.005).
In an HF population at high risk of AF, all 30 participants successfully transmitted a 1-lead ECG to a shared inbox.The adherence rate was 70% (21/30 participants sent more than 1 ECG).Three cardiologists were able to promptly reply to all emails sent within 48 hours.The AW 1-lead ECG had substantial concordance for AF identification when compared to an expert cardiologist panel interpretation.Twentysix ECGs led to health care interventions, with an average health care cost of $82.6 per patient over 3 months.A 2020 study estimated routine HF care cost to be $7,100/patient per year in Canada, thus the costs incurred on the health care system due to AW implementation was <5% of the current estimate. 5r study also introduced the possibility of In conclusion, the AW 1-lead ECG holds promise as a valuable complement to HF care.However, it is crucial to continue refining the algorithm to enhance its accessibility for elderly, lower income, and less educated patients, thus ensuring equitable benefits across diverse user groups.
doi.org/10.1016/j.jacadv.2024.101051emergency department visit cost savings as majority of ECGs sent due to symptoms were addressed via email alone suggesting that this may be a costeffective means of monitoring in this patient population.Taken together, these results support feasibility for a larger randomized control trial to address whether the AW could improve HF care and reduce health care costs.To our knowledge, this is the first study to assess the feasibility, performance, health care utilization, and patient experience of the AW 1-lead ECG in detecting AF in a HF population.Limitations include the study's single-arm design given this was intended to be a pilot feasibility study.Importantly, as with any tech-driven study, the algorithm for the AW heart rate interpretation parameters shifted during our study, initially limited to interpreting ECGs with a heart rate between 50 and 120 beats/min which later changed to between 50 and 150 beats/min, leaving 43 ECGs uninterpreted and introducing system inconsistency.Most participants (87%) were Caucasian, raising concerns about wearable accuracy across skin tones.Recruitment focused on those who could send a baseline ECG, potentially limiting generalizability given the more typical HF demographic of older age and tech unfamiliarity.
Rachel H. Heo, MD, MSc a Farid Foroutan, PhD b,c Enza De Luca, RN, MN b,c Lisa Albertini, MD a,b,c Valeria E. Rac, MD, PhD a,d,e

FIGURE 1
FIGURE 1 ECGs Resulting in Healthcare Utilization and the Cost Estimate of Healthcare Utilization