Introduction: The recently FDA-approved AliveCor Kardia ECG monitor (ALV) offers a smartphone-based one-lead ECG recording. No data have been published directly comparing an ALV recording to a 12-lead ECG. We compared the interpretability of the standard lead I recording and a modified left parasternal vector (to avoid the artifacts often seen in the ALV lead I recording) with the standard 12-lead ECG as gold standard.

Methods: Consecutive inpatients in our cardiology ward were included in this patient series. In all patients a routine 12-lead ECG, a lead I ALV recording and a modified left parasternal ALV recording were obtained (see figure). The recordings were blinded and evaluated separately by two experienced electrophysiologists. Signal quality of the smartphone monitor was analyzed by measuring the area under the curve (AUC) for the p- and T-waves and the QRS amplitude using DataInf ® Measure.

Results: 69 ECG from 23 patients (18 male, age 62,6±14,5 years, BMI 27,3±5,8 kg/m²) were analyzed. The average 12-lead ECG had a p-wave AUC of 3,9±3,7mVms, a QRS amplitude of 0,6±0,3mV and a T-wave AUC of 11,5±8,2mVms, respectively. The averages for an ALV lead-I recording were p-wave AUC 4,7±2,6mVms, QRS amplitude 0,8±0,3mV and T-wave AUC 12,9±8,8mVms. For the parasternal ALV recordings the averages were p-wave AUC 2,4±1,5mVms, QRS amplitude 0,7±0,4mV and T-wave AUC 9,9±6,0mVms. Using the 12-lead ECG, 18 patients were in sinus rhythm and 4 patients had atrial fibrillation. In 1 patient, there was disagreement between the observers (κ=0,87) in the 12-lead ECG, therefore this patient was deleted from the smartphone analysis. In the ALV lead I group, interobserver agreement was  κ=0,81 (Spearman correlation to the 12-lead ECG diagnosis 0,99, p=0,001) and in the ALV parasternal lead group κ=0,53 (Spearman correlation 0,36, p=0,046). There was no association between age or BMI and the likelihood of interobserver agreement (p>0,05). The sensitivity for the detection of an abnormal rhythm by the electrophysiologists in the ALV lead I was 94,4% and the specificity 50%. Respectively, the sensitivity in the ALV parasternal lead was 86,1% and the specificity 62,5%. Comparing the diagnoses suggested by the ALV software to the 12-lead ECG diagnoses by the electrophysiologists, the sensitivity of the ALV lead I was 72,2% and the specificity 100%. In the parasternal lead, sensitivity was 55,6% and specificity 75%.

Conclusion: In this patient series, there was a high correlation between 12-lead ECG diagnoses and a smartphone-based lead I ECG. A parasternal vector may be used when there is artifact or low signal quality in the ALV lead I, but led to a higher rate of disagreement between the observers with a comparably reduced correlation to the 12-lead ECG diagnoses. While the sensitivity of the electrophysiologists for detection of an abnormal rhythm was very high, the QRS tracking algorithm of the ALV software may be more specific, especially when screening for atrial fibrillation.

Figure: The AliveCor Kardia ECG monitor being used to record a) a lead I as recommended by the manufacturer and b) a modified parasternal lead.