Scottish Out-of-Hospital Cardiac Arrest data linkage project: initial results

Data linkage process

Permissions process

This project was dependent on securing a number of permissions before data could be shared and linked. The three stages of permissions were:

1. Approval from the Scottish Government's Analytical Leadership Group

The Scottish Government has provided funding for 60 Scottish Government/public sector linkage projects to be supported through SILC ( Scottish Informatics and Linkage Collaboration). In order to be selected as one of the 60 projects by the Analytical Leadership Group, the project had to meet a number of criteria including:

• providing a proportionate balance between privacy, mitigating the risk of identification, and realising the public benefits
• demonstrating public benefit and the ethical and secure use of data
• demonstrating the data is processed lawfully
• demonstrating alignment with Scottish Government priorities
• feasibility (ie the project is likely to succeed)
• outlining appropriate methodology
• demonstrating how the project addresses the Guiding Principles for Data Linkage
• identifying potential for significant impact on Scottish Government policy

Further information about Scottish Government Funded Data Linkage projects can be found here: http://www.gov.scot/Topics/Statistics/datalinkageframework/Funded-Linkage-Projects.

The OHCA Data Linkage Project secured this permission in August 2015.

2. Approval from the Scottish Intensive Care Society Audit Group ( SICSAG)

The Scottish Intensive Care Society Audit Group ( SICSAG) authorises (in conjunction with the Public Benefit & Privacy Panel for Health and Social Care ( PBPP) for national data requests) the use of intensive care and high dependency care data which is collected from Scottish health boards.

Further information about SICSAG is available here: http://www.sicsag.scot.nhs.uk/research/main.html.

The OHCA Data Linkage Project secured this permission in April 2016.

3. Public Benefit and Privacy Panel for Health & Social Care

The PBPP is a governance structure of NHS Scotland, exercising delegated decision-making on behalf of NHS Scotland Chief Executive Officers and the Registrar General. The panel assesses the public benefit, privacy, confidentiality, and information governance in relation to NHS Scotland or National Records of Scotland originated data.

Further information about the PBPP can be found here: http://www.informationgovernance.scot.nhs.uk/pbpphsc/

The OHCA Data Linkage Project secured this permission in May 2016.

Data sources

This first report of findings from the OHCA Data Linkage project relates to data from 2011-2014/15. Figure 13 and Table 4 provide information about the data sources included in the linked dataset:

Fig 13: A schematic overview of the linkage process and data sources

Table 4: Data sources

Dataset	Time period	Source
Ambulance	2011-2014/15	Scottish Ambulance Service/Unscheduled Care Datamart, NHS National Services Scotland
A&E	2011-2014/15	Unscheduled Care Datamart, National Services Scotland
Hospital inpatient data	2011-2014/15	Unscheduled Care Datamart, National Services Scotland
Intensive care	2011-2014/15	Scottish Intensive Care Society Audit Group
High dependency	2011-2014/15	Scottish Intensive Care Society Audit Group
Scottish patients at risk of readmission	2011-2014/15	National Services Scotland
Deaths	2011-2014/15	National Records of Scotland
Scottish Index of Multiple Deprivation	2012	Scottish Government

Data linkage process

Figure 14 gives a schematic overview of the data linkage process.

The starting point of this process is the list of 46,471 calls received by the Scottish Ambulance Service which were initially coded as possible OHCA. From this list, 14,895 OHCA cases between January 2011 and December 2015 were identified by SAS as OHCA cases where resuscitation was attempted ('worked arrests').

The next step is to convert the list of worked arrest ambulance service incidents into patient records by linking each patient with their Community Health Index ( CHI). In this dataset only one OHCA had a CHI number immediately available.

For the other cases, SAS incidents were initially linked by interrogating the Unscheduled Care Datamart ( UCD). The UCD is a collaboration between Public Health Information, Information Services Division ( ISD) and unscheduled care providers NHS 24 and SAS. A probabilistic matching process was used to link unscheduled care records from NHS 24 and Scottish Ambulance Service to inpatient data (including Emergency Department, Acute, Mental Health and Deaths) by matching patient identifiers in incident data to those in CHI linked records.

Comparing the worked arrest list of incidents to the CHI linked incidents in the UCD, 66% of OHCA cases were found. For the remaining cases, additional probabilistic linkage was carried out. This type of linkage makes use of unique personal identifiers or combinations of unique personal identifiers. For this data linkage the following identifiers were used: parts of surname, first and second initial, sex, date of birth and postcode. A detailed description of the probabilistic linkage process, advantages and disadvantages can be found in reference 48. Using this process, an additional 9.7% of CHI numbers were retrieved. The probabilistic linkage was done in a trusted environment and all personal identifiers were removed before the datasets were made available for the researchers.

Following CHI linkage, the eData Research and Innovation Service team ( eDRIS) at the Farr Institute (Scotland) were able to join the worked arrest data to the data sources outlined in Table 4. A diagram of the data linkage process can be found in Figure 14.

Figure 14: A flow chart of the linkage process.

Data confidentiality and security

The linked dataset was stored in the NHS National Services Scotland ( NSS) safe haven environment and was accessible to authorised researchers only after compliance with specific information governance, data stewardship and system security requirements ^[49] .

Statistical disclosure control has been applied to the analysis prior to any outputs leaving the safe haven environment. Statistical Disclosure Control ( SDC) covers a range of ways of changing data which are used to control the risk of an intruder finding out confidential information about a person or unit (such as a household or business). This publication has used the following methods where there are under five patients in a particular category:

Suppression of possibly disclosive cells (eg where the value is small) which means that the value for that cell in the table is not given and secondary suppression of cells which means at least one other value in the row or column is also not given to ensure that disclosive cells cannot be deduced through subtraction;

Table redesign and recoding, where cells are grouped together to protect small value cells.

Further information about Statistical Disclosure Control is available here: http://www.gov.scot/Topics/Statistics/About/Methodology/Glossary