Automated Data Abstraction of Cardiopulmonary Resuscitation Process Measures for Complete Episodes of Cardiac Arrest Resuscitation. (27th September 2016)

• Record Type: Journal Article
• Authors: Lin, Steve; Turgulov, Anuar; Taher, Ahmed; Buick, Jason E.; Byers, Adam; Drennan, Ian R.; Hu, Samantha; J. Morrison, Laurie
• Editors: Yadav, Kabir
• Is Part Of: Academic emergency medicine. Volume 23:Number 10(2016)
• Page Start: 1178
• Page End: 1181
• Publication Date: 2016-09-27
• Subjects: Emergency medicine -- Periodicals; 616.02505
• DOI: 10.1111/acem.13032
• Languages: English
• ISSNs: 1069-6563

Abstract

Abstract: Background: Cardiopulmonary resuscitation (CPR) process measures research and quality assurance has traditionally been limited to the first 5 minutes of resuscitation due to significant costs in time, resources, and personnel from manual data abstraction. CPR performance may change over time during prolonged resuscitations, which represents a significant knowledge gap. Moreover, currently available commercial software output of CPR process measures are difficult to analyze. Objective: The objective was to develop and validate a software program to help automate the abstraction and transfer of CPR process measures data from electronic defibrillators for complete episodes of cardiac arrest resuscitation. Methods: We developed a software program to facilitate and help automate CPR data abstraction and transfer from electronic defibrillators for entire resuscitation episodes. Using an intermediary Extensible Markup Language export file, the automated software transfers CPR process measures data (electrocardiogram [ECG] number, CPR start time, number of ventilations, number of chest compressions, compression rate per minute, compression depth per minute, compression fraction, and end‐tidal CO2 per minute). We performed an internal validation of the software program on 50 randomly selected cardiac arrest cases with resuscitation durations between 15 and 60 minutes. CPR process measures were manually abstracted and transferred independently by two trained data abstractors and by the automated software program, followed by manual interpretation of raw ECG tracings, treatment interventions, and patient events. Error rates and the time needed for data abstraction, transfer, and interpretation were measured for both manual and automated methods, compared to an additional independent reviewer. Results: A total of 9, 826 data points were each abstracted by the two abstractors and by the software program. Manual data abstraction resulted in a total of six errors (0.06%) compared to zero errors by the software program. The mean ± SD time measured per case for manual data abstraction was 20.3 ± 2.7 minutes compared to 5.3 ± 1.4 minutes using the software program (p = 0.003). Conclusions: We developed and validated an automated software program that efficiently abstracts and transfers CPR process measures data from electronic defibrillators for complete cardiac arrest episodes. This software will enable future cardiac arrest studies and quality assurance programs to evaluate the impact of CPR process measures during prolonged resuscitations.