Research on Health and Ageing using English Longitudinal Study of Ageing (ELSA) data linked to NHS Digital data

The English Longitudinal Study of Ageing (ELSA) is a well-established, on-going, multi-disciplinary cohort study involving a collaboration between University College London (UCL), the Institute for Fiscal Studies (IFS), the University of Manchester (UoM), and NatCen Social Research (NatCen).

NatCen is the lead organisation for the ELSA study, with several collaborators including IFS and UCL. Each collaborator brings their own specialism to the analysis of ELSA data, working in a distinct field and under their own instruction. An overarching agreement between NHS England and NatCen is in place (DARS-NIC-311182-N0L1Y) with this Agreement (DARS-NIC-32854-Y8P8B) relating specifically to NatCen sharing data with IFS. There is also a third Agreement in place relating specifically to NatCen sharing data with UCL as well (DARS-NIC-30493-Y0C0K). In summary:

> DARS-NIC-311182-N0L1Y: Agreement between NHS England and NatCen, where data was disseminated from NHS England to NatCen. NatCen run the field work to collect participants’ personal information for linkage with NHS England data.

> DARS-NIC-32854-Y8P8B: Agreement between NHS England , NatCen and IFS, to permit data provided under NIC-311182 to be pseudonymised by NatCen and sent to IFS for purposes which focus on the economics regarding health and social care.

> DARS-NIC-30493-Y0C0K: Agreement between NHS England , NatCen and UCL, to permit data provided under NIC-311182 to be pseudonymised by NatCen and sent to UCL for purposes which focus on epidemiological research.

Under this Agreement, IFS and NatCen are the Data Controllers, and IFS process the data.

The UK GDPR legal basis for processing are Article 6(1)(f) (processing is necessary for the purposes of the legitimate interests pursued by the controller) and 9(2)(j) (processing is necessary for archiving purposes in the public interest, scientific or historical research purposes or statistical purposes). As a research organisations, IFS and NatCen have an interest in understanding ageing and how that affects individuals, society and the economy. The processing is necessary for scientific research purposes because it is the only source of data which allows detailed analysis of the use of health services. It is a reasonable and proportionate way of achieving this purpose and no more data than what is required is held. This analysis is in the public interest because the research aims to help people in government, charities, academics and other influential people, when thinking about planning health services and making policies which affect people aged 50 and over.

Since its inception in 2002 it has provided valuable insights into a range of social, health and economic issues. Traditionally, data have been collected biennially face-to-face via interview and clinical examination. Participants in ELSA are selected from the Health Survey for England (HSE). All participants who met the age criteria (i.e. all those aged 50 or over) and who agreed to be re-contacted at the end of their HSE interview were invited to take part in ELSA. Each biennial data collection is referred to as a “wave”. Wave one data collection took place in the period 2002-03, wave 2 in 2004-05 and so on. The 2020 wave 10 was disrupted by Covid and data collection for this wave began in November 2021 and is due to take around nine months. Each wave, all individuals who previously took part are re-contacted to ask if they would be willing to be re-interviewed. All those who agree take part along with any new partners they may have. Additionally, in wave 3, 4, 6, 7 and 9, the sample has been “refreshed” with new members to avoid small sample sizes at the younger ages as the cohort ages, In each wave, participants are reminded of the permission they gave for data linkage and are given the opportunity to revoke that permission. While this approach has been very useful and will continue, linkage of study members in ELSA to routinely-collected data offers not only additional rich, complementary information about their health which cannot be gathered using these methods (e.g., valid data on diagnosis and prognosis of common chronic diseases such as cancer and depression) but, crucially, data which come at no burden to the study members. Participants are invited to re-consent every 2 years when study members are re-interviewed.

The Institute for Fiscal Studies (IFS) require linked pseudonymised Hospital Episode Statistics (Admitted Patient Care, Outpatient, and Accident & Emergency), Cancer registration data and ONS Mortality as part of their research obligations as part of the ELSA research group. This agreement will permit NatCen (under NIC-311182-N0L1Y subject to an active DSA and supporting purpose) to share linked pseudonymised HES, ONS Mortality and Cancers in order for IFS (under this agreement) to carry out their obligations.

The requested data will be used for a programme of research on health and ageing in England. This is a long-standing and on-going programme of work which aims to improve understanding of the ageing process, and how the use of health care affects this ageing process and the evolution of health over the lifecycle.

Linking NHS England data with ELSA will allow IFS to combine detailed information on health outcomes; the use of hospital services; the quality of health care and the identification of trends in health that will impact on future demands for health care with wider characteristics of the elderly population. The proposed linkage of ELSA to administrative health data will provide novel data for research on ageing in England. Existing studies on ageing, and in particular the use of health and social care services of individuals as they age, has been restricted by extremely limited data on the use of these services. Studies on the evolution of health at older ages using administrative health records has also been limited by a lack of information on the socio-economic and wider health characteristics of individual. Linking the data together therefore provides a rich dataset which enables research in this crucial policy area.

The work will be carried out by researchers at the IFS and is funded by the Economic and Social Research Centre (ESRC). ESRC funding is provided through the ESRC Centre for Microeconomic Analysis of Public Policy (CPP) 2020-2025, which aims to improve existing data sources through the linkage of survey data with high quality administrative data.

Below are the projects that will take place as part of this programme of work:

(1) To understand the extent to which variation exists in the use of NHS hospital services among the older population that is not explained by differences in need? IFS will examine variation in the use and cost of hospital care across the socio-economic gradient, differences in cognitive abilities, and across geographic areas.

(2) To examine how the pattern of hospital care use changes in the final year(s) of life, and to examine whether it is proximity to death, as opposed to age, that determines healthcare utilisation (controlling for other characteristics captured in the ELSA data)

(3) To understand the extent to which individuals can substitute between different types of social care and hospital care? For example, IFS will examine whether reduced availability of publicly funded social care (as a result of cuts to local authority spending) has resulted in an increased use of NHS hospitals.

(4) To compare the risk of survival following the onset of different health conditions across demographic and socioeconomic groups within the older population in England, and between similar groups in England and the US (US data will be obtained separately). IFS will use the information on cause of death from mortality data to find out who has had an onset of a condition prior to their death, so that IFS can work out the probability of survival among those who experience (e.g.) a heart attack. IFS has missing survey information on those who die before they are able to report a new onset, and the cause of death information allows them to fill in the gap.

The requested data is used solely for research purposes, in line with the research aims stated above.

As specified, ELSA consist of NatCen, IFS, UCL, and UoM working in collaboration with NatCen being the lead collaborator. NatCen are the holders of the ELSA cohort and thus only NatCen hold the identifiable data in association with this cohort. IFS, UCL, and UoM only have access to a pseudonymised version of the ELSA data with only the pseudonymised study ID as a form of identifier. For the purpose of this agreement IFS will obtain pseudonymised data from NatCen directly. The data shared by NatCen contains both ELSA data (pseudonymised) and data provided by NHS England under NIC-311182-N0L1Y (also pseudonymised). The NHS England data shared will be restricted to the fields and identifiability specified in this agreement.

NatCen sent NHS England NHS Number, Postcode, Date of Birth, Gender and study ID for a cohort of approximately 15,000 participants and this is linked to the HES, cancer and mortality data requested in NIC-311182-N0L1Y only before flowing to NatCen. All identifiers were stripped (or converted to pseudonymised format) before NatCen onwardly shared with IFS under this Agreement. The shared data has been restricted to the fields and pseudonymised as specified in this Agreement and in that of NatCen's Agreement under NIC-311182-N0L1Y. No more data needs to be shared under this Agreement. All data that is needed for the purpose described has already been shared. This Agreement is simply to retain the data previous disseminated.

The data received from NHS England has been converted by NatCen into a pseudonymised format before onward sharing to IFS by removing identifying data. Date and Birth, Date of Death, Date of Inquest, and Date of Registration has been converted to MM/YYYY format. Cancer registration number was downgraded to the first 6 digits. Only pseudonymised data has been shared with IFS and IFS may only receive, process and retain the data with an active NHS England Data Sharing Agreement in place.

The data are stored on an encrypted network that is protected by strong individual passwords and is accessible to named individuals only, all of whom are substantive employees of IFS. The IFS offices are located within a building owned by CILIP (the library and information professionals). IFS offices have their own security and are not accessible by CILIP (other than by their caretaker). All data stored on these premises is stored on served owned and managed by IFS. CILIP do not have access to the data or the servers holding the data. The servers are located in a locked room that is accessible only to ICT personal and the Head of Data Services.

All organisations party to this agreement must comply with the Data Sharing Framework Contract requirements, including those regarding the use (and purposes of that use) by “Personnel” (as defined within the Data Sharing Framework Contract i.e.: employees, agents and contractors of the Data Recipient who may have access to that data).

No data will be shared with 3rd parties.

Persons accessing the data are direct employees of IFS or contracted to IFS, and who are named ELSA collaborators. IFS terms and conditions will be adjured to. Currently, the only individual with access to the data held under this Agreement who is not substantively employed by IFS is an employee of University of Manchester who has an honorary contract with IFS.

The Data will only be used for the purposes described in this agreement.

IFS do not require identifiable data, nor will they attempt to re-identify this data. The data will not be linked to any other dataset.

IFS require data from 1997/98 to the current day (or as far back as each product allows) in order to provide the longest time series possible over which IFS can track the hospital use of ELSA respondents. This is for two principal reasons. First, this will provide the longest history of NHS service use for individuals. IFS are interested in the use of services by individuals both in a given period of time and over the lifecycle. Using a long time series therefore allows IFS to more accurately proxy lifetime use of NHS services. Second, using more years of data will maximise sample size. This is crucial in boosting the statistical power of the research, helping to accurately identify and estimate effects. As a result, these data requirements are essential in allowing IFS to carry out their proposed research.

When turning the supplied data to outputs IFS will be using record level data to do the following:

(1) Produce hospital utilization measures

(2) Create a mortality indicator as a patient outcome measure

(3) Create indicators of cause of death

(4) Use (1), (2) and (3) to run regressions, correlations and produce descriptive statistics to study the relationship between these measures, and better understand the ageing process.

(5) Create indicators of hospital admissions at the local authority level

(6) Match the data with local authority level measures of public spending on social care using data from the Chartered Institute of Public Finance Accountants (CIPFA)

(7) Run regressions of hospital utilization (recorded in ELSA) on self-reported data on social care receipt (recorded in ELSA)

The analysis will be used to produce a range of outputs.

Three types of written output are expected:

(i) working papers, published as part of the IFS working paper series, which is available on the IFS website and read by all those who use the website including government departments and academics. The first two working papers from these projects were published in September 2019 (https://ifs.org.uk/publications/14326) and May 2020 (https://ifs.org.uk/publications/14864). IFS will produce several further working papers, with expected publication dates from Summer 2022 onwards.

(ii) peer-reviewed journal article submitted to peer-reviewed economics and social science journals. For example, outputs will be submitted to the Economic Journal, an international peer-reviewed economics journal with an impact factor of 2.587 and over 900,000 article downloads in 2014. The principal audience for this type of output is economics academics who will read and cite the paper. So far, IFS have published a paper in the Milbank Quarterly (Impact factor 4.911, https://onlinelibrary.wiley.com/doi/10.1111/1468-0009.12479) in October 2020. An initial working paper was submitted to a journal in Autumn 2019 upon completion of the working paper (see (i)). However, publication in economic journals typically takes between 1-3 years after initial submission to be published, and so the publishing process for both of the working papers noted above is ongoing. IFS would therefore also expect to see future papers published over the next year, and for newly submitted papers (in Summer 2022) to be published in 2023 and 2024.

(iii) non-technical research summaries which will be press-released and target policy makers, such as the Department of Health and NHS England. These summaries would be published at the same time as working papers and/or peer-reviewed articles are published (see, for example from Autumn 2019, https://ifs.org.uk/publications/14364). IFS will disseminate results directly to policymakers as soon as our results emerge. For example, IFS has already discussed planned projects and current findings with DHSC, HM Treasury and the ONS, and will update them on result as they emerge.

IFS researchers presented preliminary results from work on spillovers between different types of social and health care using unlinked ELSA data at the Department of Health in June 2016, and presented emerging findings in October 2018 and March 2019. IFS researchers have been invited back to present at the Department again in future.

Other outputs will include presentations at academic conferences and presentations to policy makers. Academic presentations will take place at general economics conferences (e.g., European Economists Association Annual Conference) and more specialist health economics conferences (e.g., UK Health Economists’ Study Group meetings), and will focus on receiving comments from other economists on how to improve the analysis. Presentations with policymakers will focus on disseminating results and helping to inform the government departments who are involved in planning and delivering NHS care to elderly individuals.

All outputs will only report large sample aggregate statistics and regression outputs, and small numbers will be suppressed in line with the HES analysis guide. No individual or episode level data will ever be published.

The legitimate interest in understanding ageing and how that affects individuals, society and the economy and how health services should be planned for people aged 50 and over has a number of benefits for the health and social care system in England. These benefits are further described in this section.

The twin pressures of a rapidly ageing population and a prolonged period of public spending austerity will produce unprecedented pressures on NHS services over the coming years. The English population aged 65 and over is expected to grow by more than 20% over the next decade. Meanwhile, the NHS is experiencing a period of funding freezes, with annual UK health funding increasing by 1.2% between 2010-11 and 2014-15 (compared to an average increase of 5.6% in the preceding fifteen years). Understanding how to meet these additional demands with fewer resources is therefore a key challenge for health policymakers and practioners. The importance of this challenge is reflected in the recent policy and practice debate (e.g., the Better Care Fund), and the size of the challenge has been well documented by the Dilnot Commission and initiatives such as the Quality Innovation Productivity Prevention (QIPP) programme.

A data linkage between HES, Cancer Registration data and ONS mortality data with ELSA would provide an important contribution to this debate. The linkage would provide detailed information on the characteristics of individuals who use health and social care services. This would allow a detailed analysis of who uses these services, and to identify any spillovers in the use of health and social care (e.g., do cuts in social care spending have negative impacts on NHS services). In particular, the ability to follow the same individuals over an extended period of time will provide information on how needs for (and use of) health and social care have changed over across cohorts. This will contribute directly to an important debate over the size of additional pressures on services as a result of an ageing population (e.g. does ‘healthy’ ageing lead to increased health spending?).

Existing IFS work with unlinked HES data (separate to this agreement) has been used to inform policy makers including Monitor, NHS England, the Department of Health, the Cabinet Office and representatives from PCTs of which academic work helps to understand the impacts of former policy and guides improvements to the existing health and social care system. In particular, the Department of Health notes that they “have no doubt that the linkage of the Hospital Episode Statistic with survey data from the English Longitudinal Study of Ageing will be a valuable source of information in understanding the variation of health care use across individual with similar medical needs but different characteristics”.

Specific projects that will make use of the linked data will have a range of direct benefits to health and social care over the coming years, including:

(1) Understanding how population health care needs are likely to change is important for both national policy makers and local commissioners, particularly given tighter NHS budgets. The Department of Health has already shown interest in this work, part of which (using unlinked HES data) IFS presented to the Department in June 2015. This work was further discussed with representatives from the Department and NHS England in February 2016. The aim is that the findings will also feed into initiatives such as the Better Care Fund and help to achieve efficiency savings set out by the QIPP initiative and NHS England’s Five Year Forward View. As noted above, this research is supported by the Department of Health, who acknowledged its vital importance and provided supporting evidence for IFS funding applications to the Health Foundation. Representatives from the NHS England Strategy Group have also indicated to IFS the importance of such work in informing future policy when meeting to discuss the work. These findings are expected to have long-lasting benefits, which would be accrued during the project and in the years following the completion and publication of the work.

(2) Work on healthcare inequities will provide evidence of the extent to which the use of NHS health care varies across individual characteristics, and help to identify where particular groups receive higher levels of spending than can easily be accounted for by differences in medical ‘need’. This will provide two benefits to the health and social care system. First, these results can be used to inform policy makers on the future allocation of funds to CCGs. This will help to better direct funds to areas with higher health needs in future. Second, it will help commissioners and practitioners to target more treatment to individuals who previously have not received sufficient care. These benefits will be accrued in the years following dissemination of the findings.

The next stage of this work has been discussed with a number of government departments recently, including the ONS and HM Treasury (letters of support are available on request), in order to both maximise the policy-relevance of the work and to feed into current methodology used by these departments. When completed, the work will provide a much more comprehensive picture of how NHS care use varies across the socioeconomic gradient (both from a static and a lifecycle perspective) and will be able to directly answer questions such as ‘who benefits most from spending increases in the NHS?’. This is particularly relevant given the recent large spending increases to attempt to tackle backlogs in the NHS caused by COVID-19 and the associated tax rises announced to fund them, and the likely potential for further spending/tax increases to be required in the near future.

(3) Work on spillovers between different types of social and health care will provide new evidence on the extent to which individuals use a different mix of informal, and state- and privately-funded formal social care, and its interactions with use of NHS hospital care. This will significantly enhance understanding of the relationship between the use of social and hospital care, and identify spillover effects of changes to social care funding on hospital use. This work will deliver significant benefits to the health and social care system, by providing key evidence on the impacts for NHS spending as a result to cuts to social care funding. This is particularly important to policymakers given the raft of recent policies to better combine health and social care funding (e.g. Better Care Fund). Benefits will accrue following the completion of the project.

Much of the work is ongoing and so the final benefits are not realised. Specific benefits so far include:

1) Results from projects 1 and 3 were presented to analysts at DHSC and NHS England in October 2018, and to the DHSC Social Care Analysis team in March 2019, with updates on the work (or newly required extensions) requested by DHSC each time. These presentations have been requested by DHSC as part of their evidence gathering on how demand for NHS and various forms of social care is changing, and how use for certain types of care are likely to impact other parts of the health and social care system (e.g. if the generosity of the publicly provided social care system is scaled back, how is this likely to impact informal carers and NHS hospitals?). In particular, the March 2019 presentation was requested by the Social Care Analysis team as part of their preparations for the 2019 Spending Review. The research will therefore potentially influence the future provision of health and social care in England through increasing the evidence base used by DHSC and NHS England in their decision making-processes.

2) IFS has widely discussed the design of the next stage of work in project 1 with various government departments and policy bodies, including HM Treasury, ONS, DHSC and the OBR. The discussions with HMT and ONS in particular have helped to highlight shortcomings with the current approaches taken by these departments in estimating who receives the most in-kind benefit from the NHS, and to evaluate going forward how these methods could be adapted. IFS' future work will feed into this, and IFS believe that the benefits yielded in this area will be substantial in future.