Journal Watch: Early Warnings for Cardiac Arrest

Reviewed This Month

The EMS Modified Early Warning Score (EMEWS): A Simple Count of Vital Signs as a Predictor of Out-of-Hospital Cardiac Arrests

Authors: Clemency BM, Murk W, Moore A, Brown LH      

Published in: Prehosp Emerg Care, 2021 Apr 13; 1–22

About 10% of out-of-hospital cardiac arrests (OHCAs) each year are witnessed by EMS. No prehospital provider wants to be in this situation, especially when it’s unanticipated. It would be extremely helpful to have a way to identify patients who may go into arrest prior to the transfer of care. 

In fact, there are early warning scores that have been used for many years in hospitals. The Modified Early Warning Score (MEWS) is one. Unfortunately, its calculation requires assigning weights to categories of vital signs, then summing the results to generate a score. It also requires a temperature reading, and these are not always recorded in the prehospital environment. MEWS is not ideal for the prehospital setting. 

The authors of this month’s study developed and tested a tool that could be readily utilized by EMS for the identification of patients at risk for OHCA. The EMS Modified Early Warning Score (EMEWS) is calculated simply by counting abnormal vital signs, including systolic blood pressure, heart rate, respiratory rate, and AVPU. To test the performance of EMEWS, the authors retrospectively compared it to MEWS to determine if it could equally identify patients at risk for EMS-witnessed OHCA. Because temperatures were not always documented, the authors calculated a MEWS-temperature (MEWS-T) for the EMEWS comparison. 

To perform this analysis the authors utilized data from the ESO Data Collaborative. ESO is a large EMS electronic health record provider, and the ESO Data Collaborative consists of participating EMS agencies from across the United States that permit research using their deidentified records. ESO builds an annual research data set from the ESO Data Collaborative and makes it publicly available for research purposes. This is a great resource for those who have EMS research questions but are not sure where to get the data to find answers. 

The authors of this study used ESO’s 2018 data set for their analysis. In 2018 the ESO Data Collaborative included more than 7.5 million patient records from over 1,200 EMS agencies. 

All 9-1-1 records for adult prehospital patients were reviewed for inclusion. Records were excluded if there was a traumatic mechanism or the patient experienced OHCA prior to EMS arrival. The remaining cases were divided into two categories, those that reported EMS-witnessed arrests and those that did not. Patients without EMS-witnessed cardiac arrests who were transported with lights and sirens were categorized as critical/unstable and used as the comparison group. 

The authors calculated an area-under-the-curve (AUC) statistic to assess the ability of the scores to predict EMS-witnessed arrests. AUC is relatively easy to interpret—it ranges from 0–1. An AUC of 0 means the predictions are 100% wrong, while an AUC of 1 means the predictions are 100% correct. 

The authors also identified optimal cut points for each score by comparing sensitivity and specificity. Sensitivity is the probability the test will identify those with a condition if the patient has that condition. Specificity is opposite: the probability the test will identify those without the condition if the patient does not have the condition. Although screening tests ideally are both highly sensitive and highly specific, increasing sensitivity will decrease specificity, and vice versa. The optimal test cut point would be one with the highest sensitivity and specificity possible. 

Subgroup analyses examined the scores’ performance based on patients’ chief complaints, arrest timing, prearrest procedures, and crew configurations. Arrests were categorized as delayed if they occurred more than 10 minutes after EMS arrival. An unanticipated arrest was defined as one where no procedures were performed prior to the arrest. Finally, an arrest was considered paramedic-attended if the responding unit’s level of care was documented as paramedic-level. 

Results

There were 369,064 patients included in this study—364,413 patients in the critical/unstable group and 4,651 in the EMS-witnessed arrest group. The average age for patients in the critical/unstable group was 61, while the average age for those in the EMS-witnessed arrest group was 65. More than three-quarters of the patients were treated by paramedics in both groups. Patients in the critical/unstable group were less likely to have cardiac-related chief complaints (11.8% vs. 35.1%, respectively). 

Overall, MEWS-T performed better than EMEWS (AUC: 0.79 vs. 0.74, respectively). The optimal cut point for MEWS-T was a score of 3 with a sensitivity of 87% and specificity of 51%. The optimal cut point for EMEWS was a score of 2 (two abnormal vital signs or more) with a sensitivity of 81% and specificity of 54%. MEWS consistently performed slightly better than EMEWS in all subgroup analyses. However, it is unknown if this difference is clinically significant. The authors note EMEWS is much simpler to calculate and therefore likely more suitable for use by EMS. 

As with all studies, this one had limitations. The authors had to account for missing data that may have resulted in misclassification. They also noted that because scores were based on first available vital signs, they likely occurred at different times relative to the onset of OHCA. Also, because only those transported with lights and sirens were included, the results may not be applicable to the entire population treated by EMS. 

This was a very interesting study, and I congratulate the authors for publishing their work. The simplicity of the EMEWS calculation makes it an interesting candidate for EMS systems seeking to implement a tool for identifying those at risk for EMS-witnessed arrest.   

Antonio R. Fernandez, PhD, NRP, FAHA, is a research scientist at ESO and serves on the board of advisors of the Prehospital Care Research Forum at UCLA.