Abstract 239: Evaluation Of A Shock Advisory Algorithm In An Automated External Defibrillator Using Real-world Data

Abstract

Objective: Studies have shown that survival rate can be increased with AED deployment, but delivering a shock to patients who may not need one can be harmful. Thus, it is important to evaluate AED shock advisory algorithms in real-world use. This study reports the retrospective performance from out-of-hospital uses of the Philips SMART Analysis algorithm in the FR3 AED. Methods: A representative sample of 1,619 de-identified patient uses from four emergency medical services agencies in the United States, between 2015 and 2017, were included in this study. In each case, the first analysis by the SMART Analysis algorithm was always extracted, and a second analysis was randomly selected among the subsequent analyses (if any), resulting in a total of 2,739 analyses. All analyses were independently annotated by three clinical experts who were blinded to AED advisories (shock advised/no-shock advised). Based on the annotations, analyses with undeterminable rhythms or without an AED advisory were excluded. The recorded AED advisories were compared against the rhythm annotations. Results: Among the 2,739 analyses, a total of 2,616 (95.5%) were included in this study. Based on the consensus among the experts, there were 12.4% (324/2,616) annotated as shockable rhythms, 84.3% (2,205/2,616) non-shockable and the remaining were intermediate rhythms (benefits of shock are uncertain). The overall sensitivity and specificity (n/N, 90% one-sided LCL) were 96.3% (312/324, 95.0%) and 99.9% (2,202/2,205, 99.8%), respectively. The algorithm performance in each rhythm category is shown in Table 1. Conclusions: This study demonstrates that the Philips SMART Analysis algorithm in FR3 AEDs has a high sensitivity to detect shockable rhythms while maintaining a near 100% specificity for non-shockable rhythms. The algorithm’s real-world performance exceeds the requirements in the American Heart Association (AHA) guidelines and shows the safety and effectiveness of the algorithm.