Machine Learning Assisted Recognition of Out-of-Hospital Cardiac Arrest During Emergency Calls.

Can a Machine Learning Recognise of Out-of-Hospital Cardiac Arrest During Emergency Calls and Assist Medical Dispatchers

Emergency medical Services Copenhagen has developed a machine learning model that analyzes the calls to 1-1-2 (9-1-1) in real time. The model are able to recognize calls where a cardiac arrest is suspected. The aim of the study is to investigate the effect of a computer generated alert in calls where cardiac arrest is suspected.

The study will investigate

1. whether a potential increase in recognitions is due to machine alerts or the increased focus of the medical dispatcher on recognizing Out-of-Hospital cardiac Arrest (OHCA) when implementing the machine
2. if a machine learning model based on neural networks, when alerting medical dispatchers will increase overall recognition of OHCA and increase dispatch of citizen responders.
3. increased use of automated external defibrillators (AED), cardiopulmonary resuscitation (CPR) or dispatch of citizen responders in cases of OHCA on machine recognised OHCA vs. medical dispatcher recognised OHCA.

Study Overview

• Status: Completed
• Conditions: Out-Of-Hospital Cardiac Arrest
• Intervention / Treatment: Other: Alert on dispatchers screen 'Suspect cardiac arrest'

Detailed Description

Chances of survival after out-of-hospital cardiac arrest decrease 10% per minute from collapse until CPR is initiated. dispatcher assisted telephone CPR will be initiated only in cases where the dispatcher recognizes the cardiac arrest.

In a previous project "Can a computer through machine learning recognise of Out-of-Hospital Cardiac Arrest during emergency calls" (supported by TrygFoundation), the investigators found, it was possible to create a Machine Learning (ML) model, which could recognise OHCA with higher precision than medical dispatchers at the Emergency Medical Dispatch Center (EMDC-Copenhagen).

In this study the model andt is effect is to be documented in the EMDC-Copenhagen. For this purpose, a computer server running the ML-model are created. This server is integrated in the network at EMDC-Copenhagen, making it possible to push alerts to the medical dispatcher, when a cardiac arrest is recognised by the model.

With aid of machine learning, the hypothesis is, that recognition of OHCA is improved, and happen both more frequent and faster than present.

An instruction for the medical dispatchers is developed, which guides the medical dispatcher in instance of an alert from the machine.

• Study Type: Interventional
• Enrollment (Actual): 5242
• Phase: Not Applicable

Contacts and Locations

Study Locations

• Denmark
  • Danmark
    • Ballerup, Danmark, Denmark, DK-2750
      • Emergency Medical Services Copenhagen

Participation Criteria

Eligibility Criteria

• Ages Eligible for Study: Child; Adult; Older Adult
• Accepts Healthy Volunteers: No
• Genders Eligible for Study: All

Description

Inclusion Criteria:

• Call regarding a cardiac arrest registered in the national Danish Cardiac Arrest Registry
• OHCA is recognized by machine-learning model
• Call originates from 1-1-2

Exclusion Criteria:

• OHCA Emergency Medical Services - witnessed
• Call is from another authority (police or fire brigade)
• Call is a repeat call
• Call has been on hold for conference

Study Plan

How is the study designed?

Design Details

• Primary Purpose: Diagnostic
• Allocation: Randomized
• Interventional Model: Parallel Assignment
• Masking: Triple

Number of Arms

2

Arms and Interventions

Participant Group / Arm	Intervention / Treatment
Experimental: Machine alert; These cardiac suspected cardiac arrest will have had an alert generated by the machine learning model in addition to standard Emergency Medical Services response.	Other: Alert on dispatchers screen 'Suspect cardiac arrest'; Alert on dispatchers screen 'Suspect cardiac arrest'
No Intervention: Usual care; These suspected cardiac arrests will receive standard Emergency Medical Services response.

What is the study measuring?

Primary Outcome Measures

Outcome Measure	Measure Description	Time Frame
Dispatcher recognition of cardiac arrest	Dispatcher recognition of out-of-hospital cardiac arrest is the primary outcome. Recognition is reported by a questionnaire filled in by a group of auditors listening to recordings of all included calls. The questionnaire is a modified CARES protocol for the calls and consists of 21 questions whereby the quality of the call is evaluated. The questionnaire is validated and has been used in other studies.	During call to emergency Medical Services, up to 15 minutes from call start.

Secondary Outcome Measures

Outcome Measure	Measure Description	Time Frame
Time to recognition	Time from call-start until dispatcher recognition of cardiac arrest	During call to emergency Medical Services, up to 15 minutes from call start.
Dispatcher assisted telephone CPR	Does the dispatcher ask caller to initiate CPR.	During call to emergency Medical Services, up to 15 minutes from call start.
Time to T-CPR	Time from call-start until dispatcher starts guiding caller in cpr	During call to emergency Medical Services, up to 15 minutes from call start.

Collaborators and Investigators

• Sponsor: Emergency Medical Services, Capital Region, Denmark
• Investigators: Study Director: Freddy Lippert, MD, Copenhagen Emergency Medical Services

Publications and helpful links

General Publications

• Blomberg SN, Folke F, Ersboll AK, Christensen HC, Torp-Pedersen C, Sayre MR, Counts CR, Lippert FK. Machine learning as a supportive tool to recognize cardiac arrest in emergency calls. Resuscitation. 2019 May;138:322-329. doi: 10.1016/j.resuscitation.2019.01.015. Epub 2019 Jan 18.
• Blomberg SN, Christensen HC, Lippert F, Ersbøll AK, Torp-Petersen C, Sayre MR, Kudenchuk PJ, Folke F. Effect of Machine Learning on Dispatcher Recognition of Out-of-Hospital Cardiac Arrest During Calls to Emergency Medical Services: A Randomized Clinical Trial. JAMA Netw Open. 2021 Jan 4;4(1):e2032320. doi: 10.1001/jamanetworkopen.2020.32320.

Study record dates

Study Major Dates

• Study Start (Actual): September 1, 2018
• Primary Completion (Actual): April 1, 2020
• Study Completion (Actual): April 2, 2020

Study Registration Dates

• First Submitted: December 27, 2019
• First Submitted That Met QC Criteria: January 3, 2020
• First Posted (Actual): January 7, 2020

Study Record Updates

• Last Update Posted (Actual): April 16, 2020
• Last Update Submitted That Met QC Criteria: April 15, 2020
• Last Verified: April 1, 2020

More Information

Terms related to this study

• Keywords: Cardiovascular Diseases; Artificial intelligence; Heart Diseases; Heart Arrest; Machine learning; Dispatcher assisted telephone CPR
• Additional Relevant MeSH Terms: Pathologic Processes; Heart Diseases; Cardiovascular Diseases; Disease Attributes; Emergencies; Heart Arrest; Out-of-Hospital Cardiac Arrest
• Other Study ID Numbers: F-35101-01

Plan for Individual participant data (IPD)

• Plan to Share Individual Participant Data (IPD)?: No
• IPD Plan Description: Data will be available upon reasonable request by mail to primary investigator.

Drug and device information, study documents

• Studies a U.S. FDA-regulated drug product: No
• Studies a U.S. FDA-regulated device product: No