# Comparison of End-Tidal Carbon Dioxide Values in ICU Patients with and Without In-Hospital Cardiac Arrest

## TL;DR

This study found that lower end-tidal carbon dioxide levels in ICU patients may predict in-hospital cardiac arrest, suggesting potential for improved early warning systems.

## Contribution

The study demonstrates that ETCO2 trends can serve as a predictive marker for in-hospital cardiac arrest in mechanically ventilated ICU patients.

## Key findings

- ETCO2 values were significantly lower in patients who experienced in-hospital cardiac arrest compared to those who did not.
- A threshold of less than 23 mmHg ETCO2 showed 67% sensitivity and 71% specificity for predicting cardiac arrest.
- The ROC analysis indicated moderate predictive reliability with an AUC of 0.687 for ETCO2 as a marker.

## Abstract

Objective: The purpose of this study was to evaluate the utility of end-tidal carbon dioxide (ETCO2) values as a predictive marker of in-hospital cardiac arrest (IHCA). This was achieved by comparing the trends of ETCO2 values in mechanically ventilated ICU patients that experienced an IHCA versus patients that did not. Methods: A single-center, retrospective, observational, and comparative cohort study at an academic medical center. Mechanically ventilated adults in the ICU who received continuous ETCO2 monitoring were included. Patients who were transferred to our facility already intubated, experienced an out-of-hospital cardiac arrest, or had a do-not-resuscitate order were excluded. Extracted data points included demographics, comorbidities, vitals, labs, and outcomes. Patients were grouped into IHCA and non-IHCA cohorts, and the trends of ETCO2 values were compared at multiple time points for 48 h before the IHCA or after intubation (time zero) for the groups, respectively. An ROC curve was constructed to determine the predictive value of ETCO2 for IHCA. Results: A total of 207 patients were included, of which 104 (50.2%) had an IHCA and 103 (49.8%) did not. There were no differences in the mean SOFA scores at the initiation of mechanical ventilation (8.5 vs. 7.6). The ETCO2 values were decreased in the IHCA cohort, and significantly different at each time point analyzed from 300 min until immediately prior to the arrest (p < 0.001). The ETCO2 values were a mean of 20.0 mmHg in the IHCA cohort at the index time vs. 34.7 mmHg in the non-IHCA cohort (p < 0.001). The ROC analysis demonstrated moderate reliability, with an AUC = 0.687 (p < 0.0001, 95% CI 0.613–0.761) and with an ETCO2 of less than 23 mmHg, demonstrating a 67% sensitivity and a 71% specificity, as well as a 70% PPV for predicting the IHCA from our sample. Conclusions: Patients typically have rapid clinical deteriorations prior to cardiac arrest, and monitoring ETCO2 is easily achieved at the bedside while aiding in clinical decision making. The ETCO2 values in our study were significantly decreased in the IHCA cohort prior to cardiac arrest compared to the stable values in those that did not experience an IHCA, indicating that ETCO2 monitoring may have utility in predicting cardiac arrest. Further study is warranted to evaluate if predictive models utilizing ETCO2 can be constructed to predict IHCAs in mechanically ventilated ICU patients.