Introduction: The use of consumer-grade electronic devices capable of recording an electrocardiogram (ECG) is increasing but there is a lack of comparative data between technologies. We initiated this unsponsored study to compare three consumer-grade devices to the 12-lead (12L) ECG.
Methods: A total of 11 healthy volunteers had an ECG recorded consecutively using four devices: Apple Watch Series 6, KardiaMobile 6L, Withings ScanWatch and 12L. The 44 ECGs were anonymised and analysed for heart rate (HR), PR interval, QRS duration (QRSd), uncorrected QT interval and quality (1 – clear trace, 2 – low-frequency variation, 3 – high-frequency variation). Each ECG was analysed by a medical student, senior house officer, cardiology registrar and cardiology consultant who were all provided with a rate ruler. ECG interpreters were blinded to automated machine readings and to whom the ECG belonged but knew which device had been used.
1. Consumer-grade device ECGs were compared with same individual’s interpretation of the corresponding 12L ECG.
2. Intra-observer variability was assessed with the intraclass correlation coefficient (ICC) statistic.
Results: For PR interval, Apple and Withings ECGs measured within 20 ms of the corresponding 12L in 75% of cases and the Kardia in 70.9% (p=0.81). At a 40 ms cut-off, Apple and Kardia ECGs achieved this in 90.9% of cases vs 84.1% of Withings ECGs (p=0.52). Finally, 11.4% of Withings traces measured >80 ms vs 0% of Apple (p=0.05) and 2.3% Kardia (p=0.2).
For QRSd, all devices were similar, with 84.1% of Apple and Withings traces ≤20 ms of the corresponding 12L ECG and 79.5% of Kardia traces (p=1). Both Kardia and Apple had 1 ECG (2.3%) >60 ms QRS discrepancy.
For QT interval, 63.6% of Withings ECGs and 61.4% of Kardia and Apple ECGs were ≤20 ms. At ≤40 ms, the Kardia improved to 81.8%, the Apple watch to 93.2% and the Withings to 88.6%. Finally, 18.2% of Kardia ECGs had a QT discrepancy of >60 ms vs 6.8% Apple (p=0.19) and 11.4% Withings (p=0.71). Overall, measurements ≤40 ms for PR/QT and ≤20 ms for QRSd were achieved in 84.1% of Kardia, 89.4% Apple and 85.6% Withings traces (p=0.43). The ICC for Apple ECGs was PR 0.77, QRSd 0.22, QT 0.70, HR 0.79; for Kardia ECGs, PR 0.78, QRSd 0.31, QT 0.27, HR 0.89; for Withings ECGs, PR 0.5, QRSd 0.55, QT 0.52, HR 0.94; for 12L ECGs, PR 0.78, QRSd 0.41, QT 0.42, HR 0.97 (p<0.05). The mean ECG quality score (QS) was 1.4 ± 0.7 for Apple, 1.9 ± 0.5 for Kardia, 3.0 ± 0.3 for Withings and 1.0 ± 0.0 for 12L (p<0.001). Mean discrepancy was 17.1 ± 10.7 ms for QS1 traces, 22.4 ± 12.9 ms for QS2 traces and 22.3 ± 11.8 ms for QS3 traces (p=0.21).
Conclusion: In this small sample size, the Apple Watch QT measurements were shown to be inaccurate when compared with the 12L ECG, with poor (<0.5) or moderate (0.5–0.75) intra-observer variability amongst all devices including the 12L, highlighting the challenges of accurate assessment amongst non-specialists. Apple and Kardia ECG traces were the clearest traces, while all Withings traces showed high-frequency variation. ❑
Figure 1
