Discussion
We evaluated the use of a telemetry-derived ECG in the evaluation of individuals admitted to the ICU with COVID-19 infection. The population studied had a high burden of comorbid cardiac disease, and the majority were treated with hydroxychloroquine and azithromycin, requiring close cardiac monitoring of QT interval for concern of risk of torsade de pointes. Our results suggest that telemetry derived ECGs cannot replace a standard 12-lead ECG, both for interval measurements and morphologic assessment.
The mean QT or QTc intervals for the entire group as measured on a 12-lead ECG, 7-lead derived ECG, or single-lead tracing were similar, which is not unexpected given small differences in measurements in a small population. Though there was a strong linear correlation between the QT interval as measured on the different tracings, the correlation was less robust when using the single-lead tracing. The method of using Pearson correlation only measures the magnitude of a relation between variables, and not the agreement between two methods. Accordingly, one would expect a high magnitude of correlation when making measurements of the same QT intervals. The Bland-Altman method of agreement plots is an established method for quantifying agreement between two quantitative measurements. When we evaluated the agreement between measurements of the QT intervals on a standard 12-lead ECG and telemetry-derived tracings, a positive bias was found, suggesting that the telemetry tracings tend to underestimate the “true” QT interval, as defined by the manually-measured QT on a standard 12-lead ECG. This finding was also demonstrated when we averaged the individual differences between the corrected QT intervals on derived leads and the manual 12 lead measurement, with an underestimation of QT interval by both derived measurements and a significantly lower value by the single lead method. We observed differences in measurements often between 25 and 50 msec, and up to 110 msec when using the single lead tracing.
There was significantly decreased QT dispersion when using the telemetry derived 7-lead ECG, which is likely explained by fewer leads available for measurement. This finding may account for the tendency to underestimate the QT interval when using the telemetry-derived ECGs, and is consistent with other studies that have attempted to identify alternative means to monitor patients during the COVID pandemic. When comparing the use of a handheld ECG device to a standard 12-lead ECG to assess QT intervals, Cheung, et al. concluded that while QT was similar when the device was used across multiple positions, interpretation of a single lead consistently led to underestimation of the QT interval.10
Expert guidance has suggested that the risk of treatment with hydroxychloroquine and azithromycin may outweigh the benefit in patients with a QT interval longer than 500msec.11 Given a mean baseline QTc interval of 467msec, variance in the measured QT interval of the magnitude suggested in our cohort may lead to inappropriate treatment with QT-prolonging medications.
Other studies have suggested that mobile cardiac outpatient telemetry may be used for QT and arrhythmia monitoring during the COVID pandemic.12, 13 However, these studies did not compare data to the standard 12-lead ECG, and our findings are consistent with a prior study comparing QTc measurements using single lead portable ECG devices and 12-lead ECG.14 A multi-lead tracing may allow for increased diagnostic accuracy. Prior comparisons of telemetry-derived ECGs and 12-lead ECGs also found moderate correlation between QTc measurements using these modalities, though had conflicting conclusions as to whether or not the telemetry-derived ECGs offered an acceptable alternative to the standard ECG.15, 16 None of these comparisons assessed the ability of these alternative ECG monitors to assess other ECG parameters such as morphology or ST segment changes.
In our analysis, there was acceptable agreement in diagnoses among the different ECG modalities, though with some notable limitations. Although in general there was agreement in the QRS axis, 3 (9.09%) subjects with normal axis were incorrectly identified as having a rightward or northwest axis using the 7-lead derived ECG. This may be attributable to inconsistent telemetry electrode placement leading to systematic error in the mathematical derivation.
There was a suggestion of decreased sensitivity for identifying low voltage QRS complexes on the 7-lead and single lead ECGs, which may be due to the lack of precordial leads for interpretation. Conversely, a larger number of subjects were identified as having T wave inversion or ST segment changes on the 7-lead or single-lead ECG than on the standard 12-lead ECG. This finding is not consistent with prior comparisons of the ESAI configuration with a standard 12-lead ECG, which found comparable incidence and location of ST segment changes in the setting of acute MI.17
Our findings suggest that single lead tracings are quite limited and may not be an adequate replacement for traditional ECG monitoring given the inability to define axis or morphology, a high likelihood of underestimating the QT interval, and overestimation of ST segment changes. Although utilizing a 7-lead tracing may allow additional accuracy, there were still significant limitations in interval measurement and the identification of morphologic changes. We did not evaluate serial changes on telemetry derived ECGs, however our results call into question whether changes in QT intervals can be reliably monitored with telemetry-derived ECGs. Based on our results, it appears prudent that a baseline 12-lead ECG should be performed, as well as a new 12-lead after any dose change of QT prolonging drug or significant clinical event.
Our study has several strengths in its methodology. We rigorously evaluated the ECG measurements by having two independent readers with disagreements adjudicated by a third reader. This study was performed in a real-world setting; we included consecutive admissions to two cardiac care units and did not exclude any 12-lead ECGs or telemetry ECGs if they were readable. Our statistical approach used multiple validated methods for assessing the agreement between our measurements, including use of the Bland-Altman agreement plots to quantify and visualize differences between pairs of measurements. While this study was motivated by a novel use of the ICU telemetry system during the time of a pandemic, it has broader implications for the use of the telemetry system as a replacement for routine electrocardiograms in a wide array of cardiovascular and acute care conditions.
There are several limitations to this study. Though attempts were made to collect ECG and telemetry strips at close intervals, not all of the tracings were performed simultaneously, which may lead to minor differences in morphology and intervals among ECG and telemetry strips. Additionally, variation in telemetry electrode placement may lead to error in the mathematical formulas used to derive the telemetry ECG, and correct electrode placement could not be verified although ICU staff are well trained in correct lead placement. Finally, the current study only assesses for differences between ECG parameters at one point in time. We did not assess serial ECGs in the same patient, thus our results cannot be extrapolated to the utilization of telemetry-derived ECGs to monitor for changes in intervals or morphology.