The UK Out of Hospital Cardiac Arrest Outcome (OHCAO) project

Introduction Reducing premature death is a key priority for the UK National Health Service (NHS). NHS Ambulance services treat approximately 30 000 cases of suspected cardiac arrest each year but survival rates vary. The British Heart Foundation and Resuscitation Council (UK) have funded a structured research programme—the Out of Hospital Cardiac Arrest Outcomes (OHCAO) programme. The aim of the project is to establish the epidemiology and outcome of OHCA, explore sources of variation in outcome and establish the feasibility of setting up a national OHCA registry. Methods and analysis This is a prospective observational study set in UK NHS Ambulance Services. The target population will be adults and children sustaining an OHCA who are attended by an NHS ambulance emergency response and where resuscitation is attempted. The data collected will be characterised broadly as system characteristics, emergency medical services (EMS) dispatch characteristics, patient characteristics and EMS process variables. The main outcome variables of interest will be return of spontaneous circulation and medium—long-term survival (30 days to 10-year survival). Ethics and dissemination Ethics committee permissions were gained and the study also has received approval from the Confidentiality Advisory Group Ethics and Confidentiality committee which provides authorisation to lawfully hold identifiable data on patients without their consent. To identify the key characteristics contributing to better outcomes in some ambulance services, reliable and reproducible systems need to be established for collecting data on OHCA in the UK. Reports generated from the registry will focus on data completeness, timeliness and quality. Subsequent reports will summarise demographic, patient, process and outcome variables with aim of improving patient care through focus quality improvement initiatives.


INTRODUCTION
Reducing premature death is a key priority for the National Health Service (NHS). 1 2 NHS Ambulance Services treat approximately 30 000 patients a year for out of hospital cardiac arrest. There is significant variability between ambulance services in rates of the reported successful initial resuscitation (13-27%) and survival to hospital discharge (2-12%). 3 Nichol et al identified evidence of regional variation in incidence and outcomes from OHCA in 10 North American sites. There was more than 100% variability in incidence (rates ranging from 71 to 160/100 000 population) and similar variability in the decision to start resuscitation. Of those patients where resuscitation was started by the emergency medical service (EMS) there was marked variation in survival rates (range 3.0-16.3%, with a median of 8.4% (IQR, 5.4-10.4%). 4 Differences in outcomes may occur due to random variation (so called common-cause variation) or due to non-random/special cause variation. The former is to be expected in any process or system, while the latter is a systematic or unexpected deviation from the norm and may highlight an area worthy of further investigation. Evaluation of the English ambulance services return of spontaneous circulation (ROSC) and survival to discharge rates suggests there may be special cause variation (see figure 1).
Potential explanations for special cause variability Lilford et al 5 describes a pyramid with five causes of non-random/special variation in health outcomes (data, case mix, structure, Strengths and limitations of this study ▪ Successful accomplishment of objectives highly likely to improve understanding and improve outcomes from UK population, and potential to influence national policy and procedures. ▪ This is a unique opportunity to study the impact of 'process' on national patient outcomes. ▪ The development of operational procedures, standardised data collection processes and data definitions. ▪ Reliance on already stretched National Health Service (NHS) resources.
process of care, individual). The concept behind the pyramid is that most variation arises from inconsistencies in data (hence the base of the pyramid) reducing to individual practitioner variation as the smallest contributor. Figure 2 shows these principles applied in the context of cardiac arrest. Differences in data collection processes can have a dramatic impact on reported outcomes. Since the rate of cardiac arrest survival is derived from the number of people who survive divided by the number of resuscitation attempts, consistency with the processes used to determine the number of cases (case ascertainment) and outcome verification is critical for ensuring systems compare like with like. Early exploratory work in the UK has identified five different ways through which ambulance services identify cases of cardiac arrest. 6 Each approach may identify patients with subtly different characteristics and outcomes. For example, cases identified by EMS dispatch systems as cardiac arrest have a higher rate of survival (due to telephone CPR instructions and more rapid EMS response) than cases missed. Reliance solely on EMS dispatch codes to identify cardiac arrest cases would inflate survival rates relative to systems that included cases which were missed by dispatchers. Differences in case ascertainment processes might explain the observed variation in the proportion of category A Red 1 999 calls (life threatening emergency) reported as cardiac arrest cases.
The Utstein templates 7 8 aim to provide consistency to the data definitions used by cardiac arrest registries. However, it is important that definitions are consistently applied to reduce variation. 9 Differences in the case mix of patients attended by ambulance services, for example, age, 10 11 sex, 12 body Figure 1 Funnel plot showing percentage return of spontaneous circulation (A) and survival to hospital discharge (B) against the total number of cardiac arrests where resuscitation was attempted. Each dot represents a single ambulance service. Variation within the dotted line boundaries are considered to be due to normal or common-cause variation. Those lying outside the dotted line represent cases of special cause variation. mass index, 13 race, 14 social deprivation 14 15 are known to influence outcome. The Utstein comparator group (bystander witnessed cardiac arrest who are in VF) attempts to allow some adjustment for case mix, although it likely accounts for only 40% of the observed variation. 16 More complex statistical adjustments for case mix may be helpful in reducing variation due to differences in case mix. 17 Structural factors may include geography, 18 the provision and uptake of public access defibrillation, 19 20 community initiatives. 21 Process variables include EMS response time, 22 time to first shock and likely the facilities at the receiving hospital. 23 Variation attributable to the individual care provider most likely relates to the quality of CPR 24 and thresholds for initiating resuscitation.

AIMS AND OBJECTIVES OF THE PROJECT
The aim of the project is to (1) establish the epidemiology and outcome of OHCA, (2) explore sources of variation in outcome and (3) establish the feasibility of setting up a national OHCA registry as a quality improvement and research tool.

METHODS/DESIGN
This is a prospective observational study set in UK NHS ambulance services. UK ambulance services serve a population of 63 270 000 people. 25 Each ambulance service operates at least one emergency operations centre which coordinates all ambulance activity. In 2013 UK ambulance services received 9.1 million 999 calls, 7 million of these required an emergency response and of these 2.7 million were classified as needing an 8 min response. These calls generated 5 million journeys to emergency departments, of which there are approximately 247 emergency departments in the UK. 26 27 Clinical treatment protocols follow guidelines from the Association of Ambulance Chief Executive, 28 Intensive Care Society 29 and Resuscitation Council (UK). 30 The target population for the project will be adults and children sustaining an OHCA who are attended by an NHS ambulance emergency response and resuscitation is attempted. The data collected will broadly be characterised as system characteristics, EMS dispatch characteristics, patient characteristics and EMS process variables. The main outcome variables of interest will be ROSC and medium-long-term survival (30 days to 10-year survival). The project will work to standardise definitions used across ambulance services and to align them with the Utstein recommendations for Out of Hospital Cardiac Arrest. 8 See table 1 for the collected variables.
Cardiac arrest event rate, patient characteristics, setting, clinical variables, process variables and outcomes will be presented using descriptive statistics. We will use multiple logistic regression models to examine the effect of prognostic factors on binary outcomes such as ROSC on arrival at hospital and survival to hospital discharge. The Kaplan-Meier or Cox regression model will be used to identify factors that may predict patient survival.

Detailed study description
The study will be split into three phases: (1) initial feasibility, (2) data collection and (3) analysis and reporting.  The specific location where the event occurred or the patient was found. Knowledge of where cardiac arrests occur may help a community to determine how it can optimise its resources to reduce response intervals. A basic list of predefined locations will facilitate comparisons. Local factors may make creation of subcategories useful Note: 'Blank' entries will be assumed as 'Not Recorded' Note: When multiple entries occur, please refer to the 'Primacy guidance' that accompanies this document The specific location where the event occurred or the patient was found. Knowledge of where cardiac arrests occur may help a community to determine how it can optimise its resources to reduce response intervals. A basic list of predefined locations will facilitate comparisons. Local factors may make creation of subcategories useful Note: 'Blank' entries will be assumed as 'Not Recorded' Note: When multiple entries occur, please refer to the 'Primacy guidance' that accompanies this document Event continued Full location of emergency medical services occurrence (Core) A cardiac arrest that is seen or heard by another person or is monitored. EMS personnel respond to a medical emergency in an official capacity as part of an organised medical response team. Bystanders are all other groups. By this definition, physicians, nurses or paramedics who witness a cardiac arrest and initiate CPR but are not part of Continued According to the CAD system, was there an AED available at the incident location Note: OHCAO project will overwrite data from AED event form submission Bystander automated external defibrillator (AED) use (Core) Bystander AED use Note: OHCAO project will overwrite data from AED event form submission Note: When multiple entries occur, please refer to the 'Primacy guidance' that accompanies this document Primary assessments Was a ROSC noted on arrival of EMS staff? (Supplemental) Occasionally when a bystander witnesses a cardiac arrest and starts CPR, the victim will regain signs of circulation by the time EMS personal arrive. If the bystander verifies that the victim had no signs of circulation and the CPR was performed, a registry record should be initiated, EMS personnel do not need to verify that a cardiac arrest occurred for this case to be included in the registry Initial aetiology of cardiac arrest (Core) Includes cases where the cause of the cardiac arrest is presumed to be cardiac, other medical (eg, anaphylaxis, asthma, GI bleed, Respiratory), and where there is no obvious cause of the cardiac arrest First monitored rhythm (Core) Victim is found submersed in water without an alternative causation Do not attempt resuscitation (DNAR) order in place? (Supplemental) A valid DNAR order was in place and observed Note: There may be a need for initial treatment to commence whilst a valid DNAR is confirmed and treatment then withdrawn Note: If a valid DNAR order is in place, any 'Blank' 'Date of Death' will be transformed from date of incident Emergency medical services chest compressions (Supplemental) During the resuscitation, were there mechanisms or processes in place to measure the quality of CPR being delivered? Continued Before the cardiac arrest, the patient was able to perform all activities of daily living without the assistance of caregivers Comorbidities (Supplemental) The patient has a documented history of other disease conditions that existed before the cardiac arrest Ventricular assist device (Supplemental) The patient is supported by any form of ventricular assist device to augment cardiac output and coronary perfusion Cardioverter-defibrillation in place (Supplemental) The patient has an internal or external cardioverter-defibrillator The time and setting where targeted temperature control was initiated Targeted temperature control (C) Date of discharge to home or a lesser rehabilitation centre Targeted oxygenation/ ventilation (Supplemental) After ROSC, was targeted ventilation applied?

Reperfusion attempted (Core)
Was coronary reperfusion attempted? Extracorporeal life support (Supplemental) When was extracorporeal life support used? Intra-aortic balloon pump (Supplemental) Was an Intra-aortic balloon pump used? pH (Supplemental) What was the first pH recorded after ROSC? Lactate (Supplemental) What was the first lactate recorded after ROSC? Continued Phase 1: initial feasibility We will survey all 12 UK ambulance services to establish which patient, process and outcome variables, relevant to cardiac arrest are collected, how they are stored and what the data security systems are for each ambulance service. The feasibility questionnaire will be followed up by a one-to-one conversation with the identified lead to ensure data completeness and to seek clarification of any areas of uncertainty. We will request copies of existing data dictionaries related to cardiac arrest variables where these exist. We will request anonymous samples of key cardiac arrest variables from ambulance services where these exist in electronic format which will be securely transferred to the Coordinating Centre.
These data will be used to produce a map of current processes for case identification, outcome verification and measurement/reporting of key cardiac arrest variables. We will explore the feasibility of changing to a unified approach of data management processes within UK ambulance services.
We will present the output from these surveys to the Steering Committee (SC) who will endorse which core and supplementary outcome variables will be recommended for collection in the main study.
Core variables will be prioritised based on importance and feasibility of data collection. Core variable selection will be informed by the Utstein recommendations for OHCA reporting 8 31 and will capture case mix, structure, process and outcomes.
Phase 2: data collection Screening for eligibility Case records of patients with suspected cardiac arrest will be identified by ambulance service personnel through the following screening processes: ▸ Search case records for clinical or treatment variables that are likely to occur in cardiac arrest, for example, zero pulse/zero respiratory, defibrillation ▸ Search case records for cardiac arrest ▸ Search 999 call database/dispatch systems for cardiac arrest dispatch codes During the conduct of the project we work to achieve a standardised process for case identification.

Enrolment
Inclusion criteria: 1. Out of hospital cardiac arrest 2. Resuscitation is attempted (Advanced or Basic Life Support) commenced/continued by ambulance service Exclusion criteria: 1. Arrest during inter-hospital transfer or on acute NHS hospital trust premises The Out of Hospital Cardiac Arrest Outcome (OHCAO) registry system is an Extract Transform Load (ETL) web application and database for aggregating and processing data obtained from the UK's Ambulance Services. The set up and management of this database will also comply with Warwick Clinical Trials Unit (WCTU) Standard Operating Procedures (SOPs) on data security and data management and the University of Warwick's data security policy. The system comprises a SQL Server database for storing data obtained from each ambulance service and an ASP.NET web application hosted on an IIS 6 web server. An additional SQL Server database is used to host a replicated copy of the registry for analysis and reporting. SQL Server Reporting Services (SSRS) is used for all reporting requirements. The web application prohibits users from viewing the import history from other ambulance services.

Determining patient outcomes
Resuscitation is terminated at the scene of the cardiac arrest in approximately 30% of cases. 32 The remaining 70% are transferred to hospital of which approximately two-thirds have resuscitation efforts terminated in the emergency department. 32 Of those patients who initially survive and are admitted to hospital, only approximately half survive to go home. Tracking these patients to determine their outcome is complex and time consuming because it involves manual follow-up from the 14 ambulance services with over 220 acute NHS Trusts. We propose to explore the possibility to standardise and streamline the process for outcome verification for those patients who did not die in the care of the ambulance services, to determine whether or not these survivors died subsequently (and if so, why and when). We will attempt to match patients who are known by the ambulance service to obtain a ROSC with data held by the Health and Social Care Information Centre (HSCIC). We will also sample 10% of patients across all ambulance services where resuscitation is attempted. We will utilise the HSCIC Data Linkage and Extract Service to establish survival status of these patients at 30 days following cardiac arrest. We will use the flagging service to follow long-term survival. We will measure the proportion of patients where it is possible to obtain a match and compare 30 day survival status with the survival to hospital discharge information provided by the ambulance service.
Once a match has been obtained with HSCIC, we will delete non-essential patient identifiable information, retaining only the study unique ID to allow later updating of death status. Patient's NHS number, date of birth and postcode will also be retained to allow future data linkage for further assessment of sources of variation (ie, intensive care management, cardiovascular interventions) that influence survival rates.

Ethical considerations
We have carefully considered the data that are required to examine the epidemiology and outcome of OHCA. Ethics committee permissions were gained from South Central-Oxford C Research Ethics Committee (reference 13/SC/0361). The study also has received approval from the Confidentiality Advisory Group (CAG) Ethics and Confidentiality committee (ECC 8-04(C)/2013), which provides authorisation, on behalf of the Secretary of State, to lawfully hold identifiable data on patients without their consent. We will comply with the common law duty of confidentiality owed by health professionals in regard to information provided by patients in the course of clinical care, and the principles of the Data Protection Act 1998, which apply to the processing of data by Research Databases in the same way as to specific research projects. The project has received approval from the CAG for permission to implement Section 251 of the NHS Act 2006 (originally enacted under Section 60 of the Health and Social Care Act 2001), which allows identifiable patient information to be used without consent in very specific circumstances. The CAG approval also provides the SC with the authority to provide other researchers access to anonymised data in specific circumstances.

Phase 3: analysis and reporting
We anticipate having data on at least 35 000 cardiac arrests by the end of the project. The study statistician will develop and present a detailed statistical analysis plan to the Steering Committee for approval prior to data analysis.
We anticipate using descriptive statistics to summarise patient characteristics, clinical variables, EMS dispatch characteristics, EMS process variables, location and cardiac arrest event rate. Data will be presented for the entire population, the Utstein comparator group (witnessed arrest, bystander CPR, shockable rhythm) and broken down by ambulance service.
We are interested in the outcomes of ROSC and patient survival to hospital discharge. Both outcomes are binary variables where in the latter variable, the dichotomy is whether the patient survives to be discharged from hospital or not.
Potential factors that may explain the binary outcome will be identified using logistic regression model with ambulance services as random effects. Factors that have been identified will be included in a multiple logistic regression model with ambulance services as random effects to assess their inclusion in the risk prediction model.
We will also explore which factor may predict patient survival with either the Kaplan-Meier or Cox regression model. Factors that are relevant will be investigated further in a multiple Cox regression model. Both univariate and multivariate modelling will be adjusted by ambulance services. Survivors at time of analysis will be treated as censored cases.
However, as some prognostic factors may be correlated, we will assess multicollinearity to avoid including prognostic factors that are highly correlated in the same model.
As submitting data to the database is not compulsory, and there is a large variability in data quality of individual patient data, the data may be incomplete because of missing observations for the outcomes or the prognostic factor. This was previously the experience of other databases such as MINAP (Myocardial Ischaemia National Audit Project). Missing data may also follow some pattern, which would lead to biased results if appropriate methods are not used.

Quality improvement reports
The main risk prediction modelling will be based on complete case analysis. We will assess the pattern of missing data and consider multiple imputation.
In collaboration with the Steering Committee and collaborating ambulance services we will agree on the content of reports that will be provided for ambulance services. It is envisaged that initial reports will focus on data completeness, timeliness and quality. Subsequent reports will summarise demographic, patient, process and outcome variables. It is anticipated that data will be presented in summary form and broken down by ambulance service. Identification of ambulance service in any reports will be by unique code (known only to the ambulance service concerned). Reports will be sent to the principle investigator at each ambulance service and members of the Steering Committee.

SUMMARY
Improving patient outcomes from OHCA is a key priority for the NHS. To identify the key characteristics contributing to better outcomes in some ambulance services, reliable and reproducible systems need to be established for collecting data on OHCA in the UK. The aim of this project is to establish the epidemiology and outcome of out of hospital cardiac arrest, explore sources of variation in outcome and establish the feasibility of setting up a national OHCA registry.