Social care data in the UK; current landscape, challenges, and future recommendations

ABSTRACT

Objectives

Social care in the United Kingdom (UK) refers to care provided due to age, illness, disability, or other circumstances. Social care provision offers an intermediary step between hospital discharge and sufficient health for independent living, which subsequently helps with National Health Service (NHS) bed capacity issues. UK Health Technology Assessments (HTAs) do not typically include social care data, possibly due to a lack of high-quality, accessible social care data to generate evidence suitable for submissions.

Methods

We identified and characterized secondary sources of UK social care data suitable for research (as of 2021). Sources were identified and profiled by desk research, supplemented by information from custodians and data experts.

Results

We identified twenty-one sources; six high potential (three national, three regional data sources), five future potential, seven limited potential, and three not considered further (outdated or lacking social care data).

Conclusion

Despite identifying numerous sources of social care data across the UK, opportunities and access for researchers appeared limited and could be improved. This would facilitate a deeper understanding of the clinical and economic burden of disease, the impact of medicines and vaccines on social care, enable better-informed HTA submissions and more efficient allocation of NHS and local council social care resources.

KEYWORDS:

• Accessibility of data
• availability of data
• future recommendations for social care data
• health and social care linkage
• quality of data
• social care
• social care data
• utility of data for research

1. Introduction

Social care in the United Kingdom (UK) generally refers to care provided for persons who require such care due to age, illness, disability, or other circumstances [1]. This care can be administered in community (e.g. community nursing), residential (e.g. care homes and nursing homes), or domiciliary (e.g. care at home, support workers, and occupational therapists) settings and may include services such as assistance with activities of daily living (such as dressing and washing) [2] and reablement (supporting independent living after a period of illness or injury) [3]. Social care plays a crucial role in care pathways by keeping people well for longer outside of hospital and enabling more rapid and safer discharges to home. However, public spending on adult social care in the UK has fallen by 9.9% between 2009/10 and 2016/17 [1], although spending increased 4.8% in real terms from 2019/20 to 2020/21 in response to the COVID-19 pandemic [4]. Despite this increase, a survey of health-care leaders across the National Health Service (NHS; survey closed in July 2022) concluded that there is unsustainable pressure on health and care services, driven in part by capacity challenges that affect social care. Most respondents to this survey believed that lack of a social care pathway was the primary cause of delayed discharge of medically fit patients, and that inadequate social care capacity had a significant impact on addressing the elective care backlog [5]. Accordingly, social care is critical for ensuring UK NHS capacity and its ability to provide high-quality and safe care [5]. Unlike services provided by the NHS, which are free at the point of access, access to publicly funded social care depends on the person’s needs and assets. Social care is funded by those who pay for their own care, local government, or a combination [6]. However, the NHS provides short-term reablement services and funding to local authorities for continuing health care which is available for eligible persons with serious disability or illness [7].

Health technology assessments (HTAs) are a multidisciplinary process that seek to determine the value of a health technology at different points in its lifecycle [8]. Cost-effectiveness analyses conducted by HTA bodies in the UK, including The National Institute for Health and Care Excellence (NICE) in England, impact access to health technologies such as medicines and vaccines. By improving health or preventing disease, medicines, and vaccines can reduce the clinical and economic burden of disease, which in turn may reduce the burden on the NHS and social care services.

Despite the importance of social care to the NHS, social care data are rarely used in HTAs in the UK. Social care resource use can be included in the base case of a cost-effectiveness model for NICE if the costs are directly incurred by the NHS and Personal Social Services (PSS; social work or social care services outside the remit of health services, such as residential care homes) [9]. Other costs that may be considered in a scenario analysis include those to government bodies other than the NHS (may be exceptionally included if agreed with Department of Health and Social Care [DHSC]); costs paid by patients not reimbursed by the NHS or PSS; productivity costs; and care by family members, friends, or a partner that might otherwise have been provided by the NHS or PSS. As local authorities fund most social care, with some services falling under other government bodies, e.g. Ministry of Housing, Communities and Local Government, instead of the DHSC, it remains unclear whether all types of social care would be included in the base case. A brief informal review conducted by the authors identified two recent NICE appraisals on therapies for hereditary transthyretin amyloidosis (HST9 and HST10) that included social care evidence in the base case [10,11]. The Scottish Medicines Consortium (SMC) does not have explicit requirements for inclusion of social care costs when making assessments of new medicines [12]. The SMC states that costs should relate to those under the remit of NHS in Scotland and social work (equivalent to PSS in England); thus, the SMC adopts a similar approach to NICE [13]. The SMC also solicits patient and carer input for its decision-making process, which may indirectly describe social care costs [14]. Furthermore, the SMC acknowledges that finding social service costs for Scotland is challenging and providing English data is acceptable [13]. The All Wales Medicines Strategy Group adopts NICE guidance when available and assesses the remaining new medicines not considered by NICE [15]. The Department of Health in Northern Ireland is formally linked with NICE and reviews NICE guidance for local applicability [16].

The limited consideration of social care data by HTA bodies in the UK may reflect a lack of sufficient evidence exacerbated by the limited availability of high-quality and accessible social care data across the UK suitable for research to inform appraisals by HTA bodies. For example, one of the appraisals for hereditary transthyretin amyloidosis described above would have considered the cost of residential care in the base case had sufficient evidence been available [10].

Given the financial pressures on adult social care and the importance of social care for NHS capacity, social care remains a key policy concern in the UK. Lack of, or limited consideration of, social care data by HTA bodies may preclude quantification of the clinical and economic burden of certain diseases on social care, which may ultimately determine if patients can access certain medicines or vaccines free of charge. The lack of social care data may also hinder efficient decision-making on preventive care and health care provided by the NHS. Understanding the availability and accessibility of social care data in the UK to researchers, including collaboration with the pharmaceutical industry for public health benefit, may facilitate evidence generation from a social care perspective by national and international researchers for both publicly and privately funded studies, help inform and support policy on health and social care provision at a government level, and improve the efficiency of limited health and social care resources. The aim of this research was to identify and characterize secondary sources of social care data in the UK suitable for research purposes to enable social care evidence generation for meaningful consideration in UK HTA appraisals.

2. Patients and methods

A protocol-driven, four-step approach was employed to identify and characterize potential sources.

2.1. Research planning

Search terms (Table S1) were developed based on key care settings and data requirements (contents) as defined in the protocol; search strings were developed from these search terms. Key care settings for the search were community, residential, and domiciliary care; key data requirements were health-care resource use, demographics, and diagnoses (Table 1). These key care settings and data requirements were considered by the authors to be essential variables required for research purposes. Beneficial care settings and data requirements represented desirable elements of a data source if found or available. The social care settings considered were intentionally broad so as to maximize the number of data sources initially identified whereby elements of social care are recorded, whilst HTA bodies may consider a wide range of social care services, and costs should data and evidence of sufficient quality exist.

Table 1. Source classification requirements.

Key data requirements	Beneficial data requirements
HCRU Social care admissions, referrals, or both Admission or referral to all care settings, e.g. hospital or primary care Length of social care use Service use post-hospital discharge Social care service type for all facets of HCRU Demographics Age, sex, ethnicity, geographic location Diagnosis Non-specific	Specific diagnosis Demographics (comorbidities) Post-hospital discharge home visits HCRU (by service) Linkable to other RWD sources
Key care settings	Beneficial care settings
Community care (Community) Nursing Day care Residential care Care homes Hospice care Nursing homes Residential homes Domiciliary care Health visiting teams Support workers Supported independent living	Community care Community palliative care Occupational Residential care Disability care Rehabilitation center Other social care services Community support Childcare Therapy

HCRU, health-care resource use; RWD, real-world data.

2.2. Identification

Potential data sources were first identified from March to April 2021 via desk research of published and gray literature. Google Search and Google Scholar were used as the primary platforms for the identification and characterization of data sources. Additional platforms were used as necessary in a secondary capacity, including those directly describing specific data source characteristics (e.g. National Institutes of Health Data Sharing Resources/Repositories, NHS Digital Data Collections and Data Sets, Academic Research Data Repositories) and those indirectly describing data source characteristics through published research (e.g. PubMed, The Cochrane Library). A flexible approach was employed, and new search platforms of relevance discovered during this step were further investigated to identify additional data sources. A pragmatic and targeted approach with iterative improvement was employed; iterative improvements of the search strings were used until relevant results were no longer returned. Stopping rules for a given search engine or platform were pre-defined (Appendix S1).

2.3. Profiling, evaluation, and prioritization

Sources were assessed for social care settings and data attributes captured. A template and associated dictionary (Table S2) constructed by the authors based on previous research experience with secondary health/social care data were used to extract key information to determine the research potential. It was anticipated that not all fields would be retrieved/completed for each source. Once the key characteristics of each data source were extracted from the initial searches, the sources were classified into three tiers (level 1, high value; level 2, low priority; level 3, exclude) to inform the next phase of data custodian/expert outreach to obtain additional information, based on their likely utility and accessibility for research.

Level 1 sources met all key requirements to possibly enable social care research (Table 1) and were investigated further; level 2 sources met most of the key requirements and were a potential source for further exploration; level 3 sources met few of the key requirements and were not further explored. As the social care data landscape in the UK was largely unknown at the start of this research, a flexible approach to the tiered classification criteria was employed; i.e. a source that met most key requirements but had an abundance of other beneficial requirements may be classified as level 1. A pragmatic approach was employed for level 2 sources, whereby further discussion among the authors reclassified level 2 sources as either level 1 (further investigation) or level 3 (exclusion). The level 1–3 classification scheme is subjective and based upon initial, likely, or perceived value (or combinations thereof) by the authors.

2.4. Interviews and gap resolution

Once data sources deemed of potential value (from initial searches only) were identified, they were extensively profiled based upon information extracted from the initial search and any additional searches. Subsequently, data custodians and experts associated with these data sources were contacted to enquire whether they could provide details on the characteristics of the database and to elicit outstanding information, validate desk research findings, and discuss potential for access or collaboration, including with the pharmaceutical industry. Conversations with data custodians were guided by a pre-specified set of questions. Multiple contacts per data source were contacted, up to a maximum of three times, to increase the likelihood of response. Custodian or expert knowledge supplemented the gaps in literature, which subsequently informed the value of key sources. Sources were subjectively classified as limited, high, or future potential for research purposes after discussion among authors and consideration of the knowledge obtained from the desk research and custodians or experts, including, but not limited to, care settings covered, data points captured, time period covered, data latency, accessibility, eligibility for linkage to other data sources, and previous research conducted.

3. Results

3.1. Search results

Thirty unique search strings (Table S3) were used, which yielded 24 data sources. Some data sources were identified as ultimately being the same parent source and were therefore grouped. Of the 24 sources, 18 sources were prioritized for full profiling, i.e. likely of high value, after discussion among the authors (Figure 1), with the remaining six sources excluded (Table S4).

Figure 1. Source selection process. Initial desk research identified 24 sources, of which 14 were initially classified as high priority, 4 as low priority, and 6 excluded. Following discussion among the authors, among the 4 low-priority sources, 2 were upgraded to high priority and the remaining 2 were combined and considered a single high priority source. Another source (Kent Integrated Care) was identified and included by the authors after the initial desk research and was considered a high-priority source. Eighteen high-priority sources remained at the conclusion of this process.

CRIS, Clinical Record Interactive Search; FNP, Family Nurse Partnership; HES, Hospital Episode Statistics; NHS, National Health Service; PARIS, Patient-Reported Indicator Surveys.

Availability of desirable data fields varied by data source based upon this initial search. Fields typically missing or unknown following the search included data refresh lag (approximate time from data entry to availability for analysis) and access route and fees. Fifty-eight custodians from 17 potential high-value sources (custodians from one high-value source not contacted due to existing relationship) were identified and contacted, of which 22 responded. Data discussions were held with custodians of 14 data sources; some custodians responded via e-mail with useful information, but a data discussion was either not warranted or forthcoming. There were 36 non-responders and five rejections.

A total of 21 sources remained at the completion of this process. From 18 initial sources, one source was upgraded from a grouped entry to an independent entry (DataLoch), and two further sources were identified with information from custodians (Care City Cohort, Research Data Scotland). These 21 sources were classified into three categories based on the authors’ subjective assessment of their potential suitability for research; five had future potential, seven had limited potential, and six had high potential (Figure 2). Three sources were not pursued further as they were unlikely to have future research potential. Of these, one source did not actually collect or hold record-level data (the Health Foundation), and the remaining two sources were outdated based on information acquired from custodians (discontinued sources, Figure 2(b)). The full characterization of all data sources identified in this research is available from the authors on request.

Figure 2. (a) overview and (b) breakdown of sources.

CRIS, Clinical Record Interactive Search; CQC, Care Quality Commission; CSDS, Community Services Data Set; DaSH, Data Safe Haven; GGC, Greater Glasgow and Clyde; GNCR, Great North Care Record; HIC, Health Informatics Centre; HSC, Health and Social Care; KERNEL, Kent Research Network for Education and Learning; KID, Kent Integrated Dataset; LHCRE, Local Health and Care Record Exemplar; LSC, Lancashire and South Cumbria; NHAIS, National Health Application and Infrastructure Services; NWEH, NorthWest EHealth; SAIL, Secure Anonymised Information Linkage; SC, Social Care; SCHC, Scottish Care Home Census; SUS, Secondary Uses Service; TVS, Thames Valley and Surrey; WCR, Wessex Care Records.

3.2. High-potential sources

Three national (Community Services Data Set [CSDS], Scottish Care Home Census [SCHC], Source Social Care [Source SC]) and three regional (Care City Cohort, Discover-NOW, linked Kent Integrated Dataset & Kent Research Network for Education and Learning [KID & KERNEL]) high-potential sources were identified. Details of each source are shown in Table 2 (national sources) and Table 3 (regional sources). Possible barriers and challenges associated with these sources included minimal/no published evidence of prior use, limited scope of data capture/lack of linkage to additional datasets, and access to data (Table 4)

Table 2. National level 1 (high potential) sources.

Data type	Data period	Data lag	Data access	Linkages	Social care settings	Datasets	Measurements	Value
Community Services Data Set A patient-level data set for all people who are in contact with NHS-funded Community Health Services. Data are collected from clinical systems, acute trusts, mental health trusts, community health-care trusts, care trusts, social enterprises, independent sector providers, and local authorities.
A+L	Oct 2017-	1–2 mo	Direct (internal)^a	NHS digital datasets likely	District nursing Community bed-based care Home-based care Health visiting Respite care Care contacts/activities that occurred due to referral	Care Activities Care Plans Coded Scored Assessments Demographics Diagnoses Group Sessions Vaccinations Referrals	Comprehensive demographics Diagnoses (ICD-10, Read code, SNOMED) Detailed service use and pathways Vaccination status and date Treatments, procedures Referrals (source, discharge to, reason)	All NHS funded community health services captured National, comparable, standardized data about Community Health Services that are being delivered Detailed reference cost information
Scottish Care Home Census A yearly census using data collection forms, covering all adult care homes in Scotland. Data are collected in two parts; part 1 contains aggregate-level for the care homes and Part 2 contains individual-level for long-term residents (those who intend on staying ≥6 weeks on admission). Data collected includes basic resource use.
C+CS	2003- (linkable from 2013)	Annual	Direct (external)^b,c	National datasets (via CHI number)	Personal care (Y/N) Nursing (Y/N) Nursing care (Y/N)	N/A	Admission and discharge pathway and reasons Basic demographics Detailed service use and costs for long-term residents Funding type Select diagnoses (e.g. dementia)	Detailed resource use and cost information for long-term residents Highly linkable to external datasets
Source Social Care A revised social care dataset based on routine collection that can be linked with routinely collected health data (e.g. acute, mental health, and geriatric long-stay hospital admissions, A&E attendances, hospital outpatient appointments, community prescribing, and community health). The linked dataset can provide the basis for analysis of HCRU and outcomes across social and health care from individual to national level. Includes all Scottish social care that is at least partially publicly funded. Reviewed in 2018 and aligned with the Scottish Government’s Social Care Survey, resulting in a single data source.
A+L	Social care from financial year 2017/2018	6–12 mo	Direct (external)^b	National datasets (via CHI number)	Alarms and telecare Care home District nursing Home care Self-directed support	GP out of hours Homelessness Hospital activity (in/outpatient, day cases, delayed discharges, A&E) National Records Scotland Deaths Primary care	Comprehensive demographics Costs captured in Self-directed support dataset Detailed service use Diagnosis, vaccination, and treatment through linkage to primary care records	Designed to provide the basis for analysis of HCRU and outcomes across social and health care from individual to national level The only national social care database collected by Public Health Scotland

A, administrative; A&E, accident and emergency; C, census; CHI, community health index; CS, cross-sectional; GP, general practitioner; HCRU, health-care resource use; L, longitudinal, mo, month; N/A, not applicable.

^aVia NHS Digital’s Data Access Request Service; a legal basis is required.

^bVia Electronic Data Research and Innovation Service; data must be stored/processed externally through a Safe Haven.

^cAnalysis of ‘stay-level’ outputs must be conducted in-house by Public Health Scotland (i.e. indirect).

Table 3. Regional level 1 (high potential) sources.

Data type	Data period	Data lag	Data access	Linkages	Social care settings	Datasets	Measurements	Value
Care City Cohort A linked dataset of routinely collected, retrospective, pseudonymized data for the population of the London Borough of Barking and Dagenham. The dataset contains detailed service use information from five settings of care (hospital, primary care, community care, mental health, and social care), from CCG, Council, NHS Trust, GP, and HES records. The cohort was established by a community interest company to study the benefits of community and social care in health care.
Co+L	2011- (Social care from 2014, community/mental care from 2016.	Annual	Direct (external)^a	N/A; already extensively linked	Weekly packages: Crisis intervention Home care Residential and nursing home Supported living All community and social care captured in CCG or NHS Trust data.	Barking and Dagenham or Havering CCG Council records GP HES NELFT (community services) NHS Trust	Comprehensive demographics Detailed service use and costs Vaccinations Prescriptions Primary diagnosis from hospital data Treatment	Extensively linked primary, secondary, and social care data Detailed resource use information with costs/tariffs recorded Is the most deprived borough of London, and one of the most deprived in England, with a young population Custodians are eager to collaborate and have pharmaceutical industry requirements shape the cohort direction
Discover-NOW Discover-NOW is a linked, longitudinal, costed, patient-level database and the NHS real-world evidence generation hub of the 7 Health Data Research Hubs. Discover-NOW specializes in conducting or facilitating clinical pathway mapping and decision making, and is tailored to population health and health resource use studies. Led by Imperial College, Discover-NOW provides access to one of the largest depersonalized linked data sets in Europe (2.5 million currently) and holds preconcerted databases for patient recruitment. Data are collected from NW London.
A+MR+L	1980- (Up to 40 years for some areas; social care from 2015)	Monthly^b	Direct (external)^c	Possible (already extensively internally linked)	Service type (e.g. care home, home care, residential, nursing, short and long stay) Social care package (costs, units, and units of measure) Provider of care (location, organization) All social care and community care settings captured.	Community Health GP (events and prescriptions) Hospital (outpatient, admitted patient care episode, A&E) Mental health Pathology Data	Comprehensive demographics Diagnosis (primary and secondary settings) Health-care resource use and costs captured Primary care prescriptions Treatments, procedures Vaccinations dataset	Population matches age-sex distribution of the UK and CPRD, but with more ethnic diversity Care organizations and postcodes identifiable, thus linkable to health-care provider variables and patient pathway derivable. Contains a consent-to-contact register for prospective studies
Kent Integrated Dataset KID is a historically linked dataset of patient-level records from multiple services, including general practices, hospitals, community health services and social care. KID is now closed to further data collection but is linkable to KERNEL, a new, larger dataset in creation that will allow re-identification of patients and external analytics. Coverage include Kent and Medway.
A+MR+L	2014–2019 (Social care data up to 2018)	N/A	Direct (internal)^d	Possible (already extensively linked)	Domiciliary care Help at Home Nursing care	Community health services (NHS Foundation Trust) Mental Health Primary care (GP) Population register Social care (Kent County Council) Secondary care (Acute NHS Trusts)	Admissions, procedures Comprehensive demographics Detailed service use, costs, and pathways Diagnosis (ICD-10), diagnostic tests, and results Prescriptions Risk factors (e.g. alcohol) Vaccinations	Extensively linked primary, secondary, and social care data. Dataset is well situated for costs and morbidity groups Linkable to KERNEL, which will contain more recent data

A, administrative; A&E, accident and emergency; CCG, clinical commissioning group; Co, cohort; CPRD, clinical practice research datalink; CS, cross-sectional; GP, general practitioner; HES, hospital episode statistics; HCRU, health-care resource use; KERNEL, Kent Research Network for Education and Learning; KID, Kent Integrated Dataset; L, longitudinal, mo, month; MR, medical record; N/A, not applicable, NELFT, North East London NHS Foundation Trust.

^aResearch purpose must be in line with values and purposes of this cohort; data must be stored/processed externally through a Safe Haven.

^bSocial care updated monthly, vaccination daily.

^cData must be stored/processed externally through the Trusted Research Environment since late 2021 (previously indirect access only).

^dData can be stored/processed on researchers’ own environment (i.e. internal) or through an external secure environment.

Table 4. Barriers and challenges of high-potential sources.

Barrier or challenge	Data sources affected	Details
Data access	Discover-NOW	Indirect access through Discover-NOW team only; looking to allow direct access through Trusted Research Environments
	Care City Cohort	Direct access through data safe haven only; studies must have public benefit for population of Barking and Dagenham specifically
	Kent Integrated Dataset	Direct access through secure environment or researcher’s own environment; lawful basis required with public benefit
	Source Social Care	Indirect access through Source Social Care team or directly via data safe haven
	Community Services Data Set	Direct access through NHS Digital’s Data Access Request Service; requires data governance standard to be met and a legal basis for processing; previous releases were for councils, academia, and health informatics companies
	Scottish Care Home Census	Direct access through data safe haven only; application via Electronic Data Research and Innovation Service; requires approval from the Public Benefit and Privacy Panel.
Linkage to health data	Scottish Care Home Census	No routine linkage
	Source Social Care
	Discover-NOW
	Kent Integrated Dataset	Linkage not yet established (Kent Research Network for Education and Learning has plans to link to data from NHS Digital)
	Care City Cohort	External linkage not possible (already well linked internally)
Limited published research	Community Services Data Set	N/A
	Discover-NOW
Incomplete data capture	Kent Integrated Dataset	Social care data capture ended in 2018; Kent Research Network for Education and Learning to integrate social care data from 2022)
	Scottish Care Home Census	Data on short-term patients available at aggregated level only
Limit score of data captured	Scottish Care Home Census	Data collected for adult care homes only
Data quality issues	Source Social Care	Patients can be lost to follow-up; extensive missing/unknown data

N/A, not applicable.

The barriers and challenges identified may also affect other data sources not shown here. It is possible that a given barrier or challenge affects different sources to different degrees, and the information provided herein was dependent on the volume and quality of information provided by data custodians. For example, more sources may have incomplete data capture (beyond the Kent Integrated Dataset and Scottish Care Home Census), but this was not identified as a key barrier or challenge during data custodian discussions.

3.3. Future potential sources

Sources considered to be of future potential were DataLoch, Research Data Scotland, KERNEL, Wessex Care Records, and other Local Health and Care Record Exemplars (Table S5). These sources were attempting to increase the availability of regional or national social care data to researchers, integrate social care with health-care data (i.e. established linkages), and/or widen the scope of social care data captured. These sources were generally still in development. However, they were considered to have future potential based on their objectives as described by the data custodians. Once the data availability, access route, and/or scope of social care data captured have been established, these sources may be suitable for research purposes in the future.

4. Discussion

This research identified six social care data sources in the UK (three in England; three in Scotland) with current high potential to facilitate research studies that were accessible to researchers in general, if the research need met the data governance requirements. Several other sources are anticipated to be of value in the coming years. All devolved nations had at least one national health-care source, albeit with varying social care coverage. When compared to those in England, data custodians of national sources in Scotland were more open to collaboration with industry (not limited to academic or public health institutions only) and were responsive and informative regarding opportunities for collaboration, and the data sources had transparent access routes. However, regional sources in England were eager to collaborate with industry and generally had flexible access (i.e. determined in a case-by-case basis) and extensive, established linkages. Regional sources in Scotland consisted of isolated pools of data in Safe Havens (platforms for the use of NHS electronic data in Scotland) [17] with limited social care scope. Given existing research conducted by non-industry researchers, Care City Cohort may be one source ready for immediate collaboration with industry given its extensive linkages, established requirements for working with industry, and datasets that include hospital episode statistics.

Although this search identified numerous sources of social care data, the full utility of these sources for research purposes is unknown given the limited use in existing literature. While access routes may be transparent, there may also be challenges when accessing these sources in practice. For example, access to CSDS is via the NHS Digital Data Access Request Service, which can be a lengthy and comprehensive process that requires a legal basis (among other requirements) for a limited scope dataset. This may explain in part why no published studies using this source were identified in this search. Improving access is of benefit to researchers (primary research is both costly and time-consuming), patients (research findings inform patient care), and society (data can be more readily considered by government bodies). Difficulty in accessing social care data may in part explain the limited consideration of such data by the UK HTA bodies. Greater integration of social care data in HTA processes may not only better capture the social care impact on patients but also improve access to medicines and vaccines.

Some recommendations as to how these data sources could be improved for research purposes may be considered. Firstly, if data custodians were to ensure sources of social care data were more visible online, this would enable researchers to identify and comprehensively assess candidate data sources more quickly. Some data sources appeared promising but would be of limited value without linkage to other data sources for elements that would be of interest to HTA bodies, such as primary and secondary care data and comprehensive clinical data; of the six high-potential data sources, linkage to established external health data was feasible for four of the sources, with another source planning to accept linkages to such data in the future. Ready access to such integrated data would provide a more thorough picture of the patient journey and what services are accessed and when; this access would inform HTAs and consequently inform policy, ultimately improving care pathways. For example, the ability to link anonymized patient data collected from general practitioners to a range of secondary care and other health and area-based datasets is a feature of the UK Clinical Practice Research Datalink (CPRD) [18]. CPRD data have been used extensively for research purposes (over 3000 peer-reviewed publications over 30 years), and new linked datasets are added regularly to facilitate further research, such as those of the COVID-19 Hospitalisation in England Surveillance System and COVID-19 intensive care admissions of the Intensive Care National Audit and Research Centre [18,19]. HTA processes, access to medicines and vaccines, and ultimately UK society would benefit from reduced fragmentation across the social care data landscape, standardized, and simplified access procedures, connectivity between health and social care data, and transparency in permissions, costs, and time required to use and access such data. Standardized and transparent access schemes, such as those of the UK CPRD [18], NHS Trusted Research Environment [20], or NHS Hospital Episode Statistics [21] may be suitable models; access to each of the six high-potential data sources identified here requires a lawful basis and clear indication of public benefit, suggesting access by the pharmaceutical industry is feasible. Routine data collection would minimize researcher and recall bias. The DHSC could be tasked with guaranteeing access and connectivity between social and health-care data and creating a standardized framework to social care data collection, which would facilitate incorporation of social care data in HTAs. This would enable a more holistic assessment of the burden of disease and the impact of new health technologies across health and social care settings. Given the current limitations in the quality and the availability of social care data, future NICE guidelines should propose acceptable proxies that could be used instead (to increase inclusion or consideration of social care data). These guidelines should also advise on how best to integrate social care data in HTAs once high-quality data are more available.

Several additional recommendations from a commentary on gaps in UK care home data may also be applicable here. These recommendations include use of standardized identifiers to allow linkage between data sources from different sectors; investment in large-scale, anonymized linked data analysis in social care in collaboration with academics; and a core national dataset developed in collaboration with key stakeholders to support integrated care delivery, service planning, commissioning, policy, and research [22]. Recommendations from the Goldacre report on using health data for research and analysis, such as standardized data curation and coding practices, also warrant consideration [23].

Integrated care systems (ICSs) were formally established in England on 1 July 2022 with the Health and Care Act 2022. Forty-two ICSs currently exist. ICSs are partnerships of organizations that plan and deliver joined health and care systems, with the goal of improving outcomes in population health and health care reducing inequality in outcomes, experience, and access, and enhancing productivity and cost efficiency [24]. The regions covered by some ICSs overlap with some of the sources presented here. For example, the One London LHCRE is a collaborative of the five ICSs of London and the London Ambulance Service [25]. One London aims to include electronic health records, imaging and pathology records, mobile and wearable data, and local authority and social care records. It is anticipated that several years will be required for this source to be functional and accessible [26]. Given the remit of ICSs, it is crucial that they are well informed based on local, high-quality health and social care data. Collaboration with researchers, including industry, may provide suitable opportunities to accelerate proper data collection to better support decisions to improve health and economic growth. Moreover, a centralized data collection framework developed by the DHSC would ensure the standardization of all social care data collection by ICSs, thus increasing the research potential of these data.

The barriers and challenges of the sources identified in this research reflect some of the difficulties encountered by other researchers using UK social care data over the last decade. Knapp et al. had access to detailed electronic clinical records of over 3000 individuals with Alzheimer’s disease in London to identify predictors of care home and hospital admissions and associated costs. However, analyses were limited by what was available in the records-derived data set; the authors did not have comprehensive data on primary or community health or social care use [26]. Bardsley et al. used data from a convenience sample of five primary care trust areas in England and their associated councils with social services responsibilities to construct models to predict intensive social care use. The limited accuracy of their models may have been due in part to the insufficient quality and completeness of the routine data collected. In addition, the social care data sets used were limited to assessments and services recorded by the local authority [27]. Furthermore, in a study on antibiotic prescribing in UK care homes, Smith et al. analyzed data from administrative care home systems that were not designed for research (Four Seasons Health Care & Boots UK; Limited potential sources, Figure 2(b)). Data on specific comorbidities, relevant risk factors, and antibiotic indication were not available [28]. Moreover, both Witham et al. and Atherton et al. explored access to social care data and the value of linkage to health-care data in Scotland, but acknowledged the barriers that need to be overcome, such as data provenance and quality of data, as well as processes for data sharing and access [29,30]. The DHSC recently published a new health in data strategy aiming to reshape health and social care with data – in part to clear the backlog and apply learnings from the COVID-19 pandemic – which includes principles such as, improving data for adult social care, empowering researchers with the data they need to develop life-changing treatments and diagnostics, and developing the right technical infrastructure [31]. However, it remains to be seen whether this strategy will result in higher quality or more accessible social care data in the UK for research.

This research has some limitations. The search strategy was not a systematic literature review or an exhaustive search of all possible combinations of relevant terms; it is possible that the search strategy employed would miss relevant sources. The classification of sources was based on subjective discussions, and some sources may be of greater or lesser value for research purposes than suggested by the classification provided. Some historical but nevertheless useful sources may have been missed by our search strategy. Given the non-response rate of the contacted custodians and data experts, it is likely that there remain some gaps and uncertainties in the sources presented here. As such, our findings are subject to publicly available information on social care data sources and the responsiveness, availability, and willingness of data custodians to share knowledge on each source for collaboration with the pharmaceutical industry. Nevertheless, this research provides an extensive overview of the social care data landscape in the UK, with desk research conducted to an approved search protocol. Information extracted via desk research was also validated and/or supplemented through interviews with data custodians, thus resulting in a more comprehensive profiling and understanding of each data source and associated potential.

5. Conclusion

This study identified several social care data sources that are currently suitable for research purposes or that may be suitable in the future should data availability, access route, and/or scope of social care data captured be established, and data custodians’ aspirations are achieved. However, there is still a need to improve the availability and the accessibility of social care data to researchers. Given the extreme cost and resource pressures on social care and the interdependence between social and health care, the ability to perform studies with comprehensive, high-quality, and readily accessible social care data may improve the efficiency and use of social care in the UK, allow for better integration of social care data in HTA processes, improve assessment of the impact of medicines and vaccines on social care, and ultimately save very limited financial and other resources.

Article highlights

• Social care data are rarely used in HTAs in the UK, despite the importance of social care to the NHS

• This may reflect the limited availability of high-quality and accessible social care data across the UK suitable for research

• We employed a protocol-driven approach to identify, characterize, and appraise secondary sources of UK social care data suitable for research purposes

• We found six data sources of social care data (three national; three regional) with high potential for research, and five sources with future potential (one national; four regional)

• Key barriers or challenges to conducting research using the high potential sources included data access processes/methods, limited published evidence of prior use, and limited scope of data capture/lack of linkage to health and other data

Declaration of interest

D Mendes and S Collings are employees of Pfizer and hold stock or stock options of Pfizer.

R Wood and M Seif are employees of Adelphi Real World; Adelphi Real World received funds from Pfizer to conduct this research.

The authors have no other relevant affiliations or financial involvement with any organization or entity with a financial interest in or financial conflict with the subject matter or materials discussed in the manuscript apart from those disclosed.

Reviewer disclosures

Peer reviewers on this manuscript have no relevant financial or other relationships to disclose.

Author contributions

All authors were involved in project conception and design, implementation of research, interpretation of findings, manuscript development, and the decision to submit the article for publication.

Acknowledgments

The authors thank Eunmi Ha, Joe Thomas and Vicky Banks (all former employees of Adelphi Real World) for their contributions to the initial research and interpretation of findings; Tendai Mugwagwa and James Campling (employees of Pfizer Ltd.), and Susan Donaldson and Siobhan Ainscough (both former employees of Pfizer Ltd.) for their contributions to the conceptualization and interpretation of findings; Rachel Russell (of Pfizer Ltd.) for her contributions to the conceptualization, interpretation of findings and critical review of this paper; Derek Ho for drafting and finalizing the paper.

Supplementary material

Supplemental data for this article can be accessed online at https://doi.org/10.1080/14737167.2023.2274843

Additional information

Funding

This study was funded by Pfizer Ltd. Medical writing and editorial support were provided by Derek Ho PhD (ScriboMedica Ltd) and was funded by Pfizer Ltd.

Appendix

Stopping rules Search terms were modified if the current search term did not yield a new source or information of interest within the top 30 publications published since 2016 (a filter was applied to discard publications prior to 2016). This stopping rule was implemented for each search engine or platform. Thus, approximately 30 abstracts or journal articles were evaluated before a search string was discarded (if it was clear after 20 publications that the search term yielded irrelevant results, the search term was discarded prior to the 30-publication point). The next combination of search terms was then subsequently employed.