Algorithmic Accuracy in Arrhythmias? Human Review of Ambulatory ECG Recordings Outperforms Computer Algorithms

Ambulatory electrocardiographic monitors (AECG) are widely utilized in the detection of clinically significant arrhythmias, but does the rate of arrhythmia detection differ when analysis relies first on human review when compared to a computer algorithm? Findings from a recent study show that human based interpretation of the Carnation Ambulatory Monitor long-term continuous electrocardiograms (LT-ECG) identified significantly more arrhythmias when compared to the BioGuardian MCT/CEM algorithm-based mobile cardiac telemetry (MCT).

Cardiac monitors have evolved significantly since the invention of the Holter monitor in the mid-20th century, and as technology continues to improve, the amount of raw electrocardiogram (ECG) data has increased.1 Device manufacturers utilize algorithms to detect and report arrhythmias given the large amount of data, but literature on the diagnostic accuracy of each algorithm is scarce. 2-3 Most raw electrocardiographic data collected by AECGs is not available, leaving referring clinicians to rely on reports when making conclusions about study results. The sensitivity of devices to detect arrhythmias and the reliability of devices to provide clinically relevant tracings for review has become a subject of significant interest, especially now that consumer devices utilizing proprietary algorithms are becoming increasingly popular.

In a study presented at Heart Rhythm Society (HRS 2021), Mark Willcox, MD, and colleagues enrolled 50 patients prescribed an AECG to simultaneously wear MCT and LT-ECG (4 patients failed to wear both and were excluded from analysis). In the study devices, arrhythmias are detected on full disclosure ECGs by the MCT device and followed by human review, while the full disclosure of ECGs obtained by the LT-ECG device are reviewed by humans. The readers of both devices were blinded to the fact that patients were enrolled in a study, and for analysis all reports were adjudicated by two electrophysiologists. Significant arrhythmias were detected in 11/46 patients in the MCT group and 23/46 patients in the LT-ECG group (P=0.018). The number of individual arrhythmia episodes detected in the LT-ECG group was also significantly higher than the MCT group (61 versus 19, P<0.001). “Because every cardiac monitor employs different reporting methods to process recorded rhythms, it can be challenging for the physicians receiving and interpreting the final report.” Dr. Willcox said. “We found technology, paired with key human oversight and input, proved to be the most accurate in detecting critical arrhythmias”.

Although this study investigated only two available AECGs and was performed in a small population, it raises several important questions about the reliability of algorithm-based arrhythmia detection. Future work is needed to determine the accuracy of algorithm-based detection in all available devices, especially given the rising use of consumer-based products. Ultimately, the authors concluded that human review of LT-ECG was superior to algorithm-dependent MCT.

References

  1. Kennedy, HL. The Evolution of Ambulatory ECG Monitoring. Progress in Cardiovascular Diseases. 2013. www.sciencedirect.com/science/article/abs/pii/S0033062013001503?via%3Dihub
  2. Steinberg JS, et al. 2017 ISHNE-HRS Expert Consensus Statement On Ambulatory ECG and External Cardiac Monitoring/Telemetry.” Heart Rhythm. 2017;doi:10.1016/j.hrthm.2017.03.038
  3. Bansal A, Rajnish J. Portable out‐of‐Hospital Electrocardiography: A Review of Current Technologies. Journal of Arrhythmia. 2018;34(2):129–138. doi:10.1002/joa3.12035