Abstract 10289: An Automatic Arrhythmia Episode Classification Algorithm for Implantable Cardioverter Defibrillators Achieved Comparable Performance to an Electrophysiologist Panel
Introduction: Remote monitoring systems for implantable cardioverter defibrillators (ICD) collect large quantities of data that present opportunities and challenges to patient management and clinical research. Approximately 30% of ICD shocks are for non-ventricular arrhythmias (non VT/VF). Individual adjudication of millions of ICD treated arrhythmia episodes in thousands of patients is important and labor intensive. The objective of this study was to evaluate an automatic classification algorithm for adjudication of ICD treated arrhythmia episodes.
Methods: A total of 131 dual chamber ICD shock episodes from 127 patients were previously adjudicated by 7 electrophysiologists (EPs) from 4 institutions. The patients were followed on the LATITUDE (Boston Scientific) remote monitoring system. Episodes were classified by panel consensus as VT/VF or non VT/VF, with the resulting classifications used as the reference. Subsequently, each episode was assigned to 3 EPs with atrial electrogram (EGM) and to 3 EPs without atrial EGM (as a model for single chamber ICD classification). Those episodes were also classified by the automatic algorithm with and without atrial EGM. The algorithm utilizes EP domain knowledge and an advanced pattern recognition method for adjudication. Agreement with the reference was compared between the algorithm and the consensus of 3 EPs. A general estimating equations model was utilized to adjust for covariates, as well as account for correlation between adjudications for the same episode.
Results: Of the 131 episodes, 84 were VT/VF and 47 were non VT/VF based on reference. The overall agreement with the reference was similar between 3 EP consensus and the algorithm for both with atrial EGM (94% vs. 95%) and without atrial EGM (90% vs. 91%). After adjusting for covariates, the odds of accurate adjudication did not significantly differ (algorithm vs. EP: 1.02, 95%, confidence interval: 0.97 - 1.06).
Conclusions: It is possible to implement an algorithm with sufficient intelligence to perform as well as EPs in the adjudication of arrhythmia episodes treated by dual and single chamber ICDs. This type of algorithm has the potential to greatly advance our understanding of how ICDs are delivering therapy to enhance future programming.
- © 2011 by American Heart Association, Inc.