Software Annotation of Defibrillator Files: Ready for Prime Time?

INTRODUCTION: High quality chest compressions are associated with improved outcomes after cardiac arrest. Defibrillators record important information about the quality of chest compressions during CPR and can be used in quality-improvement programs. Software made for reviewing defibrillator files can automatically annotate and measure chest compression metrics. However, evidence is limited regarding the accuracy of such measurements. OBJECTIVE: To compare chest compression fraction (CCF) and rate measurements made with software annotation vs. manual annotation vs. limited annotation of defibrillator files recorded during Out-of-Hospital Cardiac Arrest (OHCA) CPR. METHODS: This is a retrospective, observational study from the Dallas Fort-Worth site of the Resuscitation Outcomes Consortium. We reviewed chest compression waveforms from the bioimpedance channel of defibrillator recordings (Physio-Control Lifepak 12 and 15) of 100 OHCA patients enrolled in the DFW Cardiac Arrest Registry from 9/8/2018 to 3/9/2019. Included cases were ≥18 years, had presumed cardiac cause of arrest, and continuous chest compressions. We assessed chest compression waveforms from the time of initial CPR until the time the defibrillator was removed. A trained reviewer revised the software annotations in two separate ways: completely manual annotations and limited manual annotations, which required minimal revising. Software, manual, and limited annotation measurements were compared for CCF and rate using intraclass correlation coefficient (ICC) statistical analysis. RESULTS: Mean patient age was 63 years with 59% male. The mean (±SD) duration of CPR was 30.4 ± 10.6 min. Case mean CCF for software, manual, and limited annotation was 0.64 ± 0.19, 0.86 ± 0.07, and 0.81 ± 0.10, respectively. ICC for manual vs. limited annotation was good to excellent. Case mean rate for all three methods was between 108.1-108.6, with no significant difference between the methods. The software misidentified epochs before the start of chest compressions, after resuscitation ended, and after return of spontaneous circulation, resulting in low ICC for CCF when compared with manual and limited annotation. The ICC was excellent for compression rate because the software only counted epochs where chest compressions were actually given. CONCLUSIONS: Software annotation performed very well for chest compression rate. With respect to CCF, the difference between manual and software annotation measurements was clinically important, while manual vs. limited annotation compared favorably.