Race against time : performance of emergency medical dispatch centres in out-of-hospital cardiac arrest

Background: Out-of-hospital cardiac arrest (OHCA) is an acute medical condition where the heart suddenly stops beating, thus causing blood circulation to cease in the body with irreversible brain damage and death if not treated. Every year, about 6000 people in Sweden experience OHCA and the survival rate in 2022 was 12%. OHCA is the most urgent medical situation handled by emergency medical dispatch centres (EMDCs). However, EMDCs can greatly affect the person’s chance of survival through early recognition, rapid dispatch of Emergency Medical Services (EMS) and other resources, initiation of dispatcher-assisted cardiopulmonary resuscitation (DA-CPR) and referral to the nearest available automated external defibrillator (AED). The chances of survival decrease by approximately 10% per minute that passes without cardiopulmonary resuscitation (CPR) and defibrillation, therefore continuous optimization of time delays at the EMDC may be crucial for survival.

Aim, methods and results: The overall aim of this thesis was to evaluate the handling and performance of EMDCs and dispatchers during emergency 112 calls concerning OHCA in accordance with the American Heart Association (AHA) performance goals for DA-CPR and to investigate whether a machine learning (ML) model could be used to recognize OHCA in emergency calls.

The aim of study I was to investigate whether an ML model can recognize OHCA to a greater extent and at an earlier stage compared with dispatchers. Before the study, the ML model was trained to understand Swedish and to recognize OHCA in emergency calls. In this study, the dispatcher's performance was compared with the ML model for 851 OHCA calls. Of all cases of OHCA recognized, ML recognized 36% (n = 305) within 1 min compared with 25% (n = 213) by dispatchers. The proportion of recognized OHCA was 86% for ML and 84% for dispatchers. The median time to recognition was 72 s for ML and 94 s for dispatchers. Comparing the time to OHCA recognition identified by both ML and dispatchers, ML was on average 28 s (p < 0.001) faster than dispatchers.

In study II, the aim was to describe the performance of EMDCs in OHCA calls in accordance with the AHA performance goals for EMDCs regarding OHCA call handling and estimate the probability of 30-day survival related to time to recognition, time to dispatch, and time to first DA-CPR directed chest compression. The study included 936 OHCA calls, of which 79% (AHA goal, 75%) were recognized by the dispatchers. Of all cases of recognizable OHCA, 85% were recognized by dispatchers (AHA goal, 95%). DACPR chest compressions were provided in 61% (AHA goal, 75%) of OHCA calls. The median time to recognition was 113 s (AHA goal, <60 s) and the median time to dispatch of an ambulance was 88 s (AHA goal, <60 s). First DA-CPR directed chest compressions were performed at a median time of 240 s (AHA goal, <90 s). The calculations estimating additional survivors of OHCA showed that if recognition had taken place within 60 s, an estimated 25 more lives could have been saved. If an ambulance had been dispatched within 60 s, an estimated 10 more lives could have been saved. If the first DA-CPR chest compressions had been performed within 90 s, an estimated 73 more lives could have been saved.

The aim of study III was to investigate the ability of Swedish EMDCs to answer medical emergency calls and dispatch an ambulance in the event of an OHCA in accordance with the AHA performance goals in a 1-step (immediate triage at the EMDC) and a 2- step (call transferred from initial EMDC to a regional EMDC for triage) procedure over 10 years and to assess whether delays are associated with 30-day survival. A total of 9,174,940 medical emergency calls were answered with a median answer delay of 7.3 s. A total of 45,367 OHCA calls were answered with a median answer delay of 7.2 s (AHA goal, <10 s). There was no significant difference in 30-day survival regarding the answer delay. For OHCA (1-step), an ambulance was dispatched after a median time of 111.9 s (interquartile range, 81.7–159.9 s). Thirty-day survival was 10.8% (n = 664) when an ambulance was dispatched within 70 s (AHA goal, high-performance) compared with 9.3% (n = 2174) after >100 s (AHA goal, minimal acceptable) (p = 0.0013). Outcome data in the 2-step procedure were not available for analysis.

In study IV, the aim was to compare Swedish dispatchers’ performance during OHCA calls after dispatchers were trained and supported by a new decision support system STEP (safety, security, efficiency, precision) versus a previous criteria-based dispatch (CBD) system. The study included 958 OHCA calls of which 82% were recognized in STEP compared with 79% in CBD (p = 0.550, AHA goal, 75%). Of the cases of recognizable OHCA, 92% were recognized in STEP compared with 85% in CBD (p < 0.001, AHA goal, 95%). In 75% (AHA goal, 75%) of cases, DA-CPR directed chest compressions were given in STEP compared to 61% in CBD (p < 0.001). The median time to recognition in STEP was 94 s compared with 113 s in CBD (p = 0.001, AHA goal, <60 s), the median time to ambulance dispatch was 110 s in STEP versus 88 s in CBD (p < 0.001, AHA goal, <60 s). First DA-CPR directed chest compressions were performed at a median time of 200 s in STEP compared with 240 s in CBD (p < 0.001, AHA goal, <90 s).

Conclusions: The results from study I showed that an ML model was able to recognize a larger proportion of OHCAs within the first minute compared with dispatchers and thus has the potential to be a decision support tool during live emergency 112 calls regarding OHCA. Study II showed that improvement measures are needed at Swedish EMDCs to achieve the AHA goals for OHCA call handling. More lives can potentially be saved if EMDCs optimize handling of OHCA calls in accordance with the AHA goals. Study III showed that most emergency calls were answered within the AHA's performance goals. When an ambulance was dispatched within the AHA's high-performance goal in response to OHCA calls, survival rates were higher compared with calls when dispatch was delayed. The results from study IV showed that when supported by STEP, the new decision support system, dispatchers performed better in five of six performance measures compared with when dispatchers were supported by a CBD system. Highperformance OHCA handling is a continuous process with room for improvement over time.

List of scientific papers

I. Byrsell F, Claesson A, Ringh M, Svensson L, Jonsson M, Nordberg P, Forsberg S, Hollenberg J, Nord A. Machine learning can support dispatchers to better and faster recognize out-of-hospital cardiac arrest during emergency calls: A retrospective study. Resuscitation. 2021;162:218–26.
https://doi.org/10.1016/j.resuscitation.2021.02.041

II. Byrsell F, Claesson A, Jonsson M, Ringh M, Svensson L, Nordberg P, Forsberg S, Hollenberg J, Nord A. Swedish dispatchers’ compliance with the American Heart Association performance goals for dispatch-assisted cardiopulmonary resuscitation and its association with survival in out-of-hospital cardiac arrest: A retrospective study. Resuscitation Plus. 2022;9:100190.
https://doi.org/10.1016/j.resplu.2021.100190

III. Byrsell F, Jonsson M, Claesson A, Ringh M, Svensson L, Riva G, Nordberg P, Forsberg S, Hollenberg J, Nord A. Swedish emergency medical dispatch centres’ ability to answer emergency medical calls and dispatch an ambulance in response to out-of-hospital cardiac arrest calls in accordance with the American Heart Association performance goals: An observational study. Resuscitation. 2023;189:109896.
https://doi.org/10.1016/j.resuscitation.2023.109896

IV. Byrsell F, Claesson A, Ringh M, Svensson L, Jonsson M, Riva G, Nordberg P, Forsberg S, Berglund E, Hollenberg J, Nord A. A new Swedish decision support system increase dispatchers’ performance in out-of-hospital cardiac arrest calls compared with a previous criteria-based dispatch system. [Manuscript]