AI to Improve Outcome From Traumatic Cardiac Arrest

Using Artificial Intelligence to Improve Outcome From Traumatic Cardiac Arrest in Denmark

This study evaluates data from patients in The Danish Medical Service electronical registry over a 6-year period from 2016 to 2021 with traumatic cardiac arrest. The objective of this study is to use artificial intelligence to evaluate reversible causes and relevant circumstances regarding traumatic OHCA in order to improve treatment and survival.

Study Overview

• Status: Active, not recruiting
• Conditions: Trauma; Out-Of-Hospital Cardiac Arrest

Detailed Description

Background

Traumatic cardiac arrest (TCA) is the leading cause of death among young individuals, but in cases where Return of Spontanous Circulation (ROSC) can be achieved, outcome appears to be more favorable than in other causes of cardiac arrest. Data from registry based studies have shown varying survival rates between 1.5 - 31.7%. This wide range could be reflected by heterogeneity in inclusion, study design and health care systems in different countries or regions. Further, this regional diversity might be driven by lack of evidence regarding treatment on-scene, transportation, but also within the prehospital decisions of with-holding treatment. Reversible causes of TCA includes hemorrhage, tension pneumothorax, asphyxia and pericardial tamponade, with uncontrollable hemorrhage accounting for around 50% of the insults. Specific algorithms has been established by the European Resuscitation Council in order to address and handle the reversible causes of TCA. Management of TCA is very time-dependent and depends on advanced prehospital treatment and further specialized care in the setting of a trauma centre.The Danish Emergency Medical System (EMS) introduced a nationwide registry of electronic medical reports in 2016.This report system allows electronic searches and thereby the opportunity to identify subgroups of OHCA's.

With aid of machine learning, the hypothesis is that advances text searches will lead to improvement of quality of data from the Danish registry of Out-of-Hospital Cardiac Arrest. With this pioneering approach, this project might contribute to amend the management of TCA.

Further, we speculate that this novel data from the EMS reports provide new and central data on reversible causes, which presumably are linked to enhanced survival of TCA. Thus, this study aims to:

• Investigate descriptives of traumatic out-of-hospital cardiac arrest
• Assess outcomes of cardiac arrests of traumatic origin
• Evaluate inter- and intrasystem comparisons

Through artificial intelligence, this study proposes an innovative, inexpensive, high-quality approach to substitute the manual validation of the Danish Cardiac Arrest registry, Whereas the registry at its present format requires manual perusal of 9000 cases per year in order to ensure the validity and quality of the national registry.

Materials and Methods

This registry-based follow-up study includes data from patients in The Danish Medical Service electronical registry over a 6-year period from 2016 to 2021 with traumatic cardiac arrest.

OHCA data

Data on OHCA's with attempted resuscitation in Denmark have been collected in the electronic based Danish Medical Service reporting system since 2016. The registry covers detailed data including the EMS report. The data consist of executive entries and advanced text searches of prehospital charts in conjunction, augmenting the identification and collecting all OHCA's in Denmark. All cases have been through an elaborate validation process of which all identified events were read through manually. This was conducted by an external verification team to corroborate high quality of data throughout the approximately 5200 cases of OHCA in Denmark annually. Within this practice of verification, supplementary sources of data have been linked to each individual case of OHCA. Further, data consists of death certificates and autopsy records from the Danish cause of death registry. Information of certain interest was survival, localization, initiation of bystander CPR, actions from EMS personnel and cause of death.

Identification of Traumatic Cardiac Arrest

Within this diverse entity of OHCA's further investigation of subgroups of trauma are required. Based on traditional machine learning, this study targets on extracting defined features from natural language, from the national registry of OHCA using a bag-of-words model. This approach is a natural language processing method for converting text to numbers. The text will be prepared since the method requires conversion into lowercase letters, removal of stop words and punctuation and further standardisation of used acronyms. Finally, words will be reduced to the root, often by removal of suffixes. Using the pre-defined trigger-words this processing allows text mining and thus the ability to derive high-quality information on OHCA related to trauma from the registry. The identified cases will be coupled to the national registry after external manual validation.

Variables included

• Age: Age will be defined as the subject age at the time of the event.
• Gender: Gender will be defined as either male, female or undetermined, and derived from personal identification numbers. Gender will be characterized based on EMS-charts in those cases without an identification number.
• Initial rhythm: The initial rhythm will be defined as the first rhythm observed by EMS personnel, and categorized as either shockable, non-shockable (asystole), non-shockable (other) or undetermined.
• Location of Incidence: Location will be characterized as either residential area, outdoors and nature, private home or other.
• Observation of occurrence: Cardiac arrests will be defined as either unwitnessed, bystander witnessed or EMS-witnessed.
• Cardio-pulmonary-resuscitation (CPR): CPR will include bystander-initiated CPR and EMS treatment with CPR.
• Defibrillation and use automatic external defibrillators (AEDs): Defibrillation will include defibrillation by bystanders and/or EMS personnel. Use of AED's includes using the device also in the case it did not deliver shocks.
• EMS-response time: This will be defined as the time between a dispatcher receiving the emergency call and the arrival of the first EMS-personnel.
• Hospitalization: Hospitalization will be considered as either; transported to hospital or declared dead by EMS-personnel.
• Return-of-spontaneous-circulation (ROSC): ROSC will be classified as the achievement of ROSC anytime between recognition of the event and termination (defined as either hospital admission or declaration of death by EMS-personnel).
• State at hospital admission: State of the case upon arrival at the hospital will be defined as either ROSC or ongoing CPR.
• Survival: Survival will be defined as ROSC at the time of hospital admission, additionally 30-day survival will be included, this derived from data from the National Patient Registry.
• Airway management: This includes the airway manoeuvres performed on each case. It is defined by listing the use of airway adjuncts such as nasopharyngeal airway, oropharyngeal airway as well as endo-tracheal intubation, supraglottic airway device and isolated bag-valve-mask ventilation.

Analysis

All data will be pseudo-anonymized, and all analyses will be accomplished on an aggregated nationwide level. Data is collected using the STrengthening the Reporting of OBservational studies in Epidemiology (STROBE) statement. The objective of this study is to clarify etiology, descriptive data and prehospital interventions of OHCA's related to trauma. Secondary an update of the annual incidence and survival rate is carried out. Descriptive statistics includes the above-mentioned variables labelled with absolute numbers and percentages. Comparative analyses will be carried out using non-parametric testing to examine subgroups. Forward logistic regression analysis will be performed for multivariate analysis. Within this multivariate logistic regression analysis both known and unknown variables will be processed. Odds ratio for survival will be calculated stratified by etiology, localization, bystander and EMS-actions.

Data storage

Data is stored on secure drive according to the regional instructions for safe conduct of data management.

Ethical considerations

GDPR will be followed according to danish law and the study will be registered at the Danish Data Protection Agency, capital region of Denmark. Since it is a registry-based study, no ethical approval is required.

Perspectives

This study gives unique information on TCA in the general population of Denmark; the descriptive statistics provides relevant data based on a reviewed, high-quality database. Furthermore, throughout analyses, a better understanding of the preceding circumstances and etiology might contribute to improve handling this type of arrests. Lastly, it proposes improvement of quality and development of observational health research.

Publication

The final results are targeted for publication in an international peer reviewed journal. Participation as coauthors will be decided according to the Vancouver criteria or acknowledged for providing access to data. All Danish regional EMS regions will receive this manuscript prior to publication for eventual comments.

List of Abbreviations

TCA Traumatic Cardiac Arrest ROSC Return of Spontanous Circulation OHCA Out-of-Hospital Cardiac Arrest EMS Emergency Medical Service STROBE STrengthening the Reporting of OBservational studies in Epidemiology CPR Cardiopulmonary Resuscitation GDPR General Data Protection Regulation

• Study Type: Observational
• Enrollment (Actual): 31200

Contacts and Locations

Study Locations

• Denmark
  • Næstved, Denmark
    • Prehospital Center

Participation Criteria

Eligibility Criteria

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

• Sampling Method: Probability Sample

Study Population

This registry based research evaluates traumatic OHCA in the general population of Denmark based on the national registry of cardiac arrests in Denmark.

Description

Inclusion Criteria:

• OHCA related to trauma

Exclusion Criteria:

• All other causes of OHCA

Study Plan

How is the study designed?

Design Details

• Observational Models: Cohort
• Time Perspectives: Retrospective

Number of groups / cohorts

2

Cohorts and Interventions

Group / Cohort
OHCA in the general population; All Out-of-Hospital Cardiac Arrests
Traumatic OHCA; OHCA related to Trauma

What is the study measuring?

Primary Outcome Measures

Outcome Measure	Measure Description	Time Frame
Survival to hospital admission	Status at admission.	The timeframe is at the time of hospital admission, therefore no specific time point (ie 24-hours, 1hour ect) can be stated
Longterm Survival	30-day survival	30-day

Secondary Outcome Measures

Outcome Measure	Measure Description	Time Frame
Return Of Spontaneous Circulation	ROSC at anytime. It is a categorical, binary variable.	The timeframe is from incident (at scene) to discharge from admission to hospital. As of such it is not meassured at any specific time or time point..

Collaborators and Investigators

• Sponsor: Prehospital Center, Region Zealand
• Investigators: Principal Investigator: signe A Wolthers, MD, Prehospital Center, Næstved, Region Zealand, Denmark

Publications and helpful links

General Publications

• Lott C, Truhlar A, Alfonzo A, Barelli A, Gonzalez-Salvado V, Hinkelbein J, Nolan JP, Paal P, Perkins GD, Thies KC, Yeung J, Zideman DA, Soar J; ERC Special Circumstances Writing Group Collaborators. European Resuscitation Council Guidelines 2021: Cardiac arrest in special circumstances. Resuscitation. 2021 Apr;161:152-219. doi: 10.1016/j.resuscitation.2021.02.011. Epub 2021 Mar 24. Erratum In: Resuscitation. 2021 Oct;167:91-92.
• Kleber C, Giesecke MT, Lindner T, Haas NP, Buschmann CT. Requirement for a structured algorithm in cardiac arrest following major trauma: epidemiology, management errors, and preventability of traumatic deaths in Berlin. Resuscitation. 2014 Mar;85(3):405-10. doi: 10.1016/j.resuscitation.2013.11.009. Epub 2013 Nov 25.
• Lindskou TA, Mikkelsen S, Christensen EF, Hansen PA, Jorgensen G, Hendriksen OM, Kirkegaard H, Berlac PA, Sovso MB. The Danish prehospital emergency healthcare system and research possibilities. Scand J Trauma Resusc Emerg Med. 2019 Nov 4;27(1):100. doi: 10.1186/s13049-019-0676-5.
• Leis CC, Hernandez CC, Blanco MJ, Paterna PC, Hernandez Rde E, Torres EC. Traumatic cardiac arrest: should advanced life support be initiated? J Trauma Acute Care Surg. 2013 Feb;74(2):634-8. doi: 10.1097/TA.0b013e31827d5d3c.
• Evans CC, Petersen A, Meier EN, Buick JE, Schreiber M, Kannas D, Austin MA; Resuscitation Outcomes Consortium Investigators. Prehospital traumatic cardiac arrest: Management and outcomes from the resuscitation outcomes consortium epistry-trauma and PROPHET registries. J Trauma Acute Care Surg. 2016 Aug;81(2):285-93. doi: 10.1097/TA.0000000000001070.
• Pfeifer R, Tarkin IS, Rocos B, Pape HC. Patterns of mortality and causes of death in polytrauma patients--has anything changed? Injury. 2009 Sep;40(9):907-11. doi: 10.1016/j.injury.2009.05.006. Epub 2009 Jun 21.

Study record dates

Study Major Dates

• Study Start (Actual): February 22, 2022
• Primary Completion (Anticipated): January 1, 2023
• Study Completion (Anticipated): January 1, 2023

Study Registration Dates

• First Submitted: February 24, 2022
• First Submitted That Met QC Criteria: March 16, 2022
• First Posted (Actual): March 24, 2022

Study Record Updates

• Last Update Posted (Actual): March 24, 2022
• Last Update Submitted That Met QC Criteria: March 16, 2022
• Last Verified: March 1, 2022

More Information

Terms related to this study

• Keywords: Cardiopulmonary Resuscitation; Out-of-Hospital Cardiac Arrest; Emergency Medical Service; Traumatic Cardiac Arrest
• Additional Relevant MeSH Terms: Heart Diseases; Cardiovascular Diseases; Heart Arrest; Out-of-Hospital Cardiac Arrest
• Other Study ID Numbers: AIT_2022

Plan for Individual participant data (IPD)

• Plan to Share Individual Participant Data (IPD)?: Yes
• IPD Sharing Time Frame: Upon request. Data will be available for one year from the date.
• IPD Sharing Supporting Information Type: Study Protocol; Statistical Analysis Plan (SAP)

Drug and device information, study documents

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