Evaluating smartwatch-based detection of supraventricular tachycardia and atrial fibrillation in the emergency department

Abstract

Background

In this study, we categorized smartwatches into two groups: those with electrocardiogram (ECG) capability and those utilizing photoplethysmography (PPG) sensors. The aim was to evaluate the diagnostic performance of these devices in differentiating atrial fibrillation (AF) from supraventricular tachycardia (SVT) among patients presenting to the emergency department (ED) with an irregular rhythm alert.

Methods

This retrospective, single-center observational study was conducted in the ED of a tertiary hospital over a two-year period. The study included adult patients (≥22 years) who presented to the ED following an irregular rhythm alert generated by their smartwatch. Smartwatches were classified as either ECG-capable devices or PPG-based devices. Final rhythm diagnoses were confirmed by two independent emergency physicians based on 12‑lead ECG findings obtained during the ED evaluation.

Results

A total of 424 patients were included. Of these, 63.7 % (n = 270) used ECG-capable smartwatches, and 66.3 % (n = 281) received an AF alert. Among all alerts, 60.8 % (n = 258) were ECG-based and 39.2 % (n = 166) were PPG-based. Common presenting symptoms included palpitations (56.1 %), dizziness (21.7 %), and dyspnea (23.6 %), while 8.5 % of patients were asymptomatic. In total, 44.6 % (n = 189) required pharmacologic or electrical interventions. Smartwatches demonstrated moderate accuracy for AF and SVT detection. PPG-based detection showed higher sensitivity than ECG-based detection (AF: 54.3 % vs. 44.9 %; SVT: 83.6 % vs. 79.2 %). Agreement with final ECG diagnoses was poor (Cohen's Kappa = 0), with significant misclassification for both arrhythmias (McNemar's p < 0.001), except for PPG-based SVT (p = 0.072). No significant predictors of false-positive AF alerts were found (area under the curve [AUC] = 0.538). False-positive SVT alerts were more common in younger patients (odds ratio [OR] = 0.92, p = 0.039), females (OR = 0.55, p = 0.048), and less frequent in diabetic patients (OR = 0.24, p = 0.001; AUC = 0.685).

Conclusion

Among patients presenting to the ED with smartwatch-reported irregular rhythm alerts, PPG-based detection demonstrated higher sensitivity for SVT compared to ECG-based methods; however, the overall diagnostic agreement for both AF and SVT was limited.