Discrepancies between physician review and algorithmic detection of the zoll rescuenet post-cardiac arrest case review

Abstract

Background

Accurate measurement of CPR quality metrics is critical for improving cardiac arrest outcomes. Impedance based automated devices have demonstrated limitations. Zoll RescueNet CaseReview, rather, uses accelerometry to analyze chest compressions and automatically provides code feedback, including CPR pause number, length, and chest compression fraction. However, the reliability of these automated measurements compared to manual physician review remains uncertain.

Methods

We conducted a retrospective observational cohort study at a tertiary academic medical center, analyzing 212 in-hospital cardiac arrest cases recorded between July 1, 2023, and July 1, 2024. The study compared CPR metrics generated by the Zoll RescueNet CaseReview algorithm to manual physician review of raw defibrillator data, focusing on pause durations and chest compression fraction (CCF) using Bland-Altman plots.

Results

Bland-Altman plots indicated overestimation of individual pause times (mean difference 4.00 s), max pause time per arrest (mean difference 24.57 s) total pause time per arrest (mean difference 0.73 min), and average number of pauses per arrest, with corresponding underestimation of CCF (mean difference 8.33%). Substantial variability was present for all variables with increased disagreement for longer pause times.

Conclusion

The Zoll RescueNet CaseReview algorithm estimates longer CPR pause durations than manual physician review, thereby lowering the chest compression fraction estimate. These findings support manual review of raw data and improved algorithmic detection of compressions to ensure feedback to resuscitation teams is reliable.