Inter-rater agreement of paramedic rhythm labeling

Abstract

Study hypothesis:Substantial inter-rater agreement is present in the labeling by paramedics of ventricular fibrillation and asystolic rhythms.Design:Prospective, cross-sectional study.Type of participants:One hundred five practicing paramedics from nonvolunteer agencies who are advanced cardiac life support certified.Methods:Five static cardiac arrest rhythm strips, classified by Cummins' average peak amplitude method, were arranged into five different orders of presentation and placed into five booklets. The paramedics were instructed to label each rhythm ventricular fibrillation or asystole based on rhythm recognition, not on treatment plan.Results:The overall [kappa] value for labeling the five rhythms was .63, indicating a moderate degree of inter-rater agreement. However, as the rhythm's amplitude decreased, the amount of inter-rater agreement also decreased. When the amplitude was approximately 1 mm, agreement was no different than chance; the proportion of paramedics labeling the rhythm ventricular fibrillation was .46 (95% confidence interval, .36, .56). Only a flat line (0 mm) demonstrated perfect inter-rater agreement, with no paramedic labeling the rhythm ventricular fibrillation.Conclusion:Inter-rater agreement of ventricular fibrillation rhythm labeling by paramedics in this emergency medical services system was amplitude dependent. An analysis of ventricular fibrillation rhythm data that does not address the degree of inter-rater agreement of rhythm labeling cannot ensure uniform reporting of out-of-hospital cardiac arrest data.