Abstract
Background
Symptom driven remote monitoring programs for cardiac arrhythmias hold great promise, but scalability is limited due to high additional workload for healthcare providers. The Dutch HartWacht arrhythmia program consists of a connected single lead ECG device operated remotely by the patient, an algorithm for classification and a dedicated team of specialized nurses and cardiologists for additional remote interpretation. Correct classification as sinus rhythm (SR) by the algorithm would reduce workload of the HartWacht team, as it makes double-checking redundant.
Purpose
We investigated agreement of the ECG-classification between the algorithm and the HartWacht team and determined feasibility of the algorithm to classify sinus rhythm (SR).
Methods
We investigated the algorithm accompanying a single lead, handheld ECG-device that is integrated in the Dutch HartWacht program. We retrospectively studied the sensitivity, specificity, positive predictive value (PPV) and negative predictive value (NPV) of the algorithm for classifying SR on home measured 30-second single lead ECGs. We included all recordings that were classified as SR by the algorithm. We used the classification of the HartWacht team as a reference standard.
Results
Between April 2020 and January 2021, 1,671 patients with (suspected) arrhythmias (female = 982 (59%), mean age = 58 (±15) years, participating in the HartWacht program, recorded 53,748 ECGs, of which the algorithm interpreted 35,388 (66%) as SR, 10,899 (20%) as possible AF and 7,461 (14%) as other. All recordings were also interpreted by the HartWacht team. Compared to the classification by the team, the algorithm showed a sensitivity for SR of 0.953, specificity of 0.985, PPV of 0.996 and NPV of 0.841. A total of 137 (0,3%) ECGs from 50 (2,8%) patients showed false positive outcomes, classifying recordings as SR while the HartWacht team detected arrhythmias. In 42 of those patients, arrhythmias were detected by the algorithm in other recordings within the program. The remaining 8 (0,5%) patients made a total of 14 (<0,1%) recordings with false positive outcomes without having any other recordings with arrhythmias within the HartWacht program.
Conclusion
For classifying SR in home measured single lead ECGs, the algorithm and the HartWacht team showed a nearly perfect agreement. The recordings without agreement did not lead to relevant individual changes in diagnostic or therapeutic strategy for the patient. Therefore, the algorithm is feasible as standalone classification. With 66% of the recordings within the HartWacht program showing SR, a corresponding workload reduction can be achieved which importantly increases scalability and cost-effectiveness of remote monitoring of arrhythmia patients.
Funding Acknowledgement
Type of funding sources: None.
Keywords: Remote Patient Monitoring and Telehealth