Creating the CUREd dataset – a real world linked data for UEC
Recent reports in the UK of severe delays to ambulance responses for critically ill patients, congested EDs and hospital wards affecting patient safety, are evidence of severe pressure on these services, with record levels of demand compromising patient experience and safety, with associated costs running into billions of pounds. 1-4 Understandably, there is considerable public pressure to find solutions to what is becoming a familiar crisis in our NHS.5-6
Emergency services, such as ambulance and emergency department (ED) are part of a complex urgent and emergency care (UEC) system, with patients frequently moving between different providers during a single episode of care and experiencing a ‘system of care’ rather than care from an individual service. Access to the system may be through different routes such as the NHS 111 phone and triage service, GP out of hours, 999. However, patients typically encounter at least two different services during a single episode, which might end in attendance at ED or inpatient hospital care.
The challenge for health services researchers has been to capture this UEC data in a way that truly assesses these complex patient journeys. In the UK, large amounts of real world, routinely collected data is available for individual services i.e. data collected for administrative purposes, such as the ambulance service, describing activity for every patient, including time and durations of care, investigations and treatments received and onward referral to other services. However, this individual service data is not routinely linked across services in the UK, and so research based on analysis of routine data in the past has not reflected real world patient journeys through the UEC system.
Curating and analysing linked UEC data is crucial to understanding how the system is used on a population level, providing insights on services used and by whom, when and what outcomes result. Taking this first step in understanding the challenges faced by the system inspired a team of researchers, data specialists and statisticians led by ARC UEC Theme Lead, Professor Suzanne Mason at the University of Sheffield to curate the first integrated data platform for UEC, the CUREd Database. The process towards achieving this and using the data has been arduous, but ultimately rewarding and the use of real world linked data such as CUREd will continue to inform solutions to the challenges facing UEC in years to come.
Curating the CUREd Database
Service provider buy-in
Creating a large linked routine dataset such as the CUREd dataset is not possible without the commitment of the services themselves to participate in the work and provide the data. The CUREd team has benefited from incredible support from the regional services involved, the Yorkshire Ambulance Service (YAS) and the 13 acute hospital trusts in the region who agreed to participate and provide the data. YAS provides both the 999 and NHS 111 service across YH, which facilitated the region wide approach to creating an UEC dataset. Collaborations such as the ARC YH undoubtedly make this collaborative working easier, with services, particularly those included as match funded research partners, able to see clear benefits for the services for contributing to the establishment of a research platform, which might answer key questions for their services.
For the CUREd dataset, we aimed to develop a regional UEC dataset for 6 years (2011-17). In practice, this meant collecting and linking data for over 25 million episodes of UEC patient care, comprising 999 ambulance, NHS 111, ED and inpatient emergency admissions. In order to link patient data across the different services required patient identifiers, such as NHS number and date of birth. The usual process of obtaining individual patient consent to provide the care and identifiable data items for this research was clearly unrealistic when managing this many ‘participants’. In addition, obtaining informed patient consent in UEC research is difficult due to severity of illness involved. Enhanced regulatory approval from the Confidential Advisory Group (CAG), which grants the legal basis for organisations to collect and process identifiable data without patient consent was therefore required. As part of the approval an ‘opt out’ basis of inclusion was adopted, whereby patients were notified of the development of the database, through notification on service websites or in public areas and given the option to have their information removed retrospectively from the database . Successfully obtaining CAG and all ethical approvals for the CUREd dataset was a challenging endeavour taking the best part of a year!
Once we obtained data from providers, the challenges of linking routine data across services was by no means a straightforward one and the expertise and capacity developed by CUREd data management staff during the creation of CUREd is highly specialised and invaluable. Led by a Systems Data Architect Tony Stone, a transparent and detailed staged process of linking data was undertaken. NHS number in particular and other patient identifiers were key in this process. Particular challenges were apparent in linking ambulance call and dispatch data to hospital episodes in cases where patients were conveyed to hospital, due to less than 15% of cases having NHS numbers. However, despite these challenges across most services, the rate of linkage was over 98% and the final database comprised over 23 million episodes of care.
How the database is being used to inform solutions to UEC
The CUREd database is utilised by a number of projects within the ARC YH UEC programme and shared with collaborating partners outside of the University of Sheffield, to widen the understanding of the UEC system and the different key patient groups who use it.
Example of the work include the following:
NHS 111 Pathways Project
This analysis examined the pathways of patients using the 111 phone advice and triage service and subsequent care in the UEC through to ED use and hospital admission. This analysis highlighted 111 triaging some demand unnecessarily to emergency services, but other drivers exist beyond 111 control, such as patients’ preferences for timely ED care and availability or knowledge of comparable alternatives.8
Non-urgent ED attendances in children
We found 21% of ED attendances in the region in children were low acuity, with the highest numbers presenting out of hours and in the under 5s. Interventions focused on this group of children in the early evening and weekend would provide most benefit in reducing strain on hospital services. 9
Widening the understanding of the UEC system
Additional linkages have also been made between CUREd and other non-UEC providers, to understand how specific cohorts of patients present to the UEC system. These have included Sheffield Health and Social Care data on patients with serious mental illness and Sheffield Hospice Care data, to follow patients at the end of life who may be have unwanted presentations to EDs.
Future developments for CUREd database
Currently the CUREd dataset contains data up to 2017. It is vital that this data is ‘refreshed’ and updated, so that our continued data analysis reflects the latest population use and service developments so that insights from the data are timely and precise. To this end, ARC YH is funding a ‘refresh’ of CUREd data. Our learning from CUREd will inform a different approach to data acquisition, particularly of hospital data. However, the skills of data processing, including ‘in house’ linking of provider data will continue to be vital. Also new analytical approaches incorporating techniques from other academic disciplines, such as machine learning from computer science, offer new ways to understand and apply the learning from real world linked data.
This blog was written by Colin O'Keefe and Prof Sue Mason, Urgent and Emergency Care theme at Yorkshire and Humber ARC.
14 December 2021References
- Lowthian JA, Curtis AJ, Cameron PA, et al. Systematic review of trends in emergency department attendances: an Australian perspective. Emerg Med J 2011;28:373–7.
- NHS Digital. Hospital accident and emergency activity 2018-19, 2019. Available: https://files.digital.nhs.uk/F5/ACF07A/AE1819_Annual_Summary.pdf [Accessed 26 Aug 2021].
- NHS England. High quality care for all, now and for future generations: Transforming urgent and emergency care services in England - Urgent and Emergency Care Review End of Phase 1 Report, Appendix 1 – Revised Evidence Base from the Urgent and Emergency Care Review, 2013. Available: https://www.nhs.uk/NHSEngland/keoghreview/documents/UECR.Ph1Report.Appendix%201.EvBase.FV.pdf [Accessed 26 Aug 2021].
- NHS England. The NHS long term plan, 2019. Available: https://www.longtermplan. nhs.uk/publication/nhs-long-term-plan/ [Accessed 26 Aug 2021].
- Knowles E, O'Cathain A, Nicholl J. Patients' experiences and views of an emergency and urgent care system. Health Expect. 2012 Mar;15(1):78-86. doi: 10.1111/j.1369-7625.2010.00659.x. Epub 2011 Jan 31. PMID: 21281414; PMCID: PMC5060599.
- Lewis J, Stone T, Simpson R, et al. Patient compliance with NHS 111 advice: analysis of adult call and ED attendance data 2013–2017. PLoS One 2021;16:e0251362
- Simpson RM, O’Keeffe C, Jacques RM, et al. Emerg Med J Epub ahead of print: [please include Day Month Year]. doi:10.1136/ emermed-2021-211431