End of life care analysis: describing our methodology

The quality of care people receive at end of life and inequalities in experiences of care are critical policy concerns. We describe here our approach to examining changing trends in service use towards the end of life. This builds on an earlier phase of work which looked at the services used by people who died at home in England, before and during the first year of the pandemic. 

Acknowledgements

We are very grateful for the support received from the TPP Technical Operations team throughout this work, and for generous assistance from the information governance and database teams at NHS England and the NHS England Transformation Directorate.

Members of the The OpenSAFELY Collaborative include: Sebastian CJ Bacon, Lucy Bridges, Benjamin FC Butler-Cole, Simon Davy, Iain Dillingham, David Evans, Louis Fisher, Amelia Green, Ben Goldacre, Liam Hart, George Hickman, Peter Inglesby, Steven Maude, Amir Mehrkar, Thomas O’Dwyer, Rebecca M Smith, Pete Stokes, Tom Ward, Jon Massey, Milan Wiedemann, Christopher Bates, Jonathan Cockburn, Sam Harper, Frank Hester, John Parry. 

We would like to thank our OpenSAFELY co-pilot, Andrea Schaffer, for providing an introduction to OpenSAFELY and for providing advice and support throughout the project.

We would like to thank the members of our advisory group for their support of this project: Dr Sarah Mitchell (National Clinical Director (interim) for Palliative and End of Life Care, NHS England), Brian MacKenna & Richard Metcalfe (NHS England), Fliss Murtagh (Professor of Palliative Care, and Director of the Wolfson Palliative Care Research Centre), Karen Chumbley (Chief Clinical Officer Clinical Lead in End-of-Life Care St Helena Hospice Suffolk and North East Essex), Joanne Greengrass (Essential Standards of Care Quality Improvement Lead NHS Frimley Frimley Health and Care ICS), Katie Griffin, Ann-Marie Wilson & Jude Beng (Patient and Public Involvement representatives).  

About OpenSAFELY 

With the approval of NHSE (https://www.opensafely.org/policies-for-researchers/#information-governance-and-ethical-approval) we used the OpenSAFELY-TPP platform, which allowed us to analyse the individual electronic health records of people registered at GP practices in England using TPP SystmOne software. We received approval from NHS England to conduct this secondary phase of analysis as an extension of the application we submitted for our earlier phase of work. Primary care records managed by the GP software provider, TPP, were linked through OpenSAFELY to Office for National Statistics (ONS) death registration data and hospital data (on admitted patient care, emergency care and outpatient activity) from NHS England’s Secondary Uses Service (SUS). All data were linked, stored and analysed securely within the OpenSAFELY-TPP platform (https://opensafely.org). Data include pseudonymised data such as coded diagnoses, medications and physiological parameters. No free text data are included. Using this platform meant we had no access to patient records: our analysis code was developed using dummy data before executing on the real data using a remote server. There is a public record of all the analysis that we carried out (https://jobs.opensafely.org/deaths-at-home-during-covid-19/end-of-life-carequality/). Aggregated results were checked for whether they were disclosive, before being released. Counts less than or equal to 7 were redacted, and all counts are rounded to the nearest 5. All analytical code is shared openly for review and re-used under an MIT open license (https://github.com/opensafely/end-of-life-carequality). 

Analysis cohort

We used date of death from the ONS death registrations to identify a cohort of people of any age, registered with a TPP practice on the day they died. The cohort covered people who died between 1 March 2019 and 31 August 2023. Practices that use TPP software to manage electronic health records cover around 24 million people in England (currently registered), representing just under 45% of the population. The analysis cohort represented around 40% of ONS-published deaths. 

Describing trends in service-use over time 

Hospital and GP records were used to explore changes in the services used towards the end of life since the first year of the pandemic. We looked at service use in the last 30 days of life. We focused on comparing the average number of events per person and the proportion of people with at least one service-use event (for example, at least one emergency admission in the month before death). 

The hospital service-use measures were: 

• Accident and Emergency (A&E)
• admissions (elective and emergency)
• outpatient appointment attended (only available from April 2020, with some missing data for this month)

We looked at the end of life care received by just over 975,000 patients who died. ¹  In our previous study we compared our cohorts of patients who died at home in the pre-pandemic and pandemic period with ONS published deaths and found that it was broadly representative by age group, sex and place and cause of death. There were some distinctions by region, with lower representation in the North West (5% of ONS-published deaths). 

Table 1 provides a breakdown of different places of death across the study period as a whole. These are similar to deaths across England and Wales.

Table 1: Place of death: study population (patients registered with a TPP practice on the day they died) (March 2019 - August 2023) and England and Wales 2023

Source: OpenSAFELY; Office for National Statistics 

Figure 1 shows the trend in place of death in the study population between March 2019 and August 2023. The trend is similar to that seen in the population as a whole, although the ONS data was published in August 2023, which is likely to mean underestimated number of deaths in the most recent month of data due to registration delays. The proportion of people dying at home has remained above pre-pandemic levels in every month since March 2020. There was a notable increase in the number of people dying at home in December 2022, which coincides with high levels of flu and Covid-19 across the population.