Variation in Healthy Life Expectancy Throughout Childhood and Adulthood in England

The University College London (UCL) Institute of Health Informatics Research is a hub for facilitating the improvement of healthcare in the NHS underpinned by rigorous research methods of complex health data. The Institute has a strong commitment towards developing a culture for sharing innovative methods and outputs with the aim of maximising the impact and visibility of research using linked health data. Building on the success of the Farr Institute, the Medical Research Council (MRC) has established Health Data Research UK (HDR UK), a multi-funder UK institute for health and biomedical informatics research.

The Farr Institute was a UK-wide research collaboration involving 21 academic institutions and health partners in England, Scotland and Wales. Publicly funded by a consortium of ten organisations, led by the Medical Research Council, between 2013 and 2018, the Institute was committed to delivering high-quality, cutting-edge research using ‘big data’ to advance the health and care of patients and the public. The Farr Institute did not own or control data but analysed data to better understand the health of patients and populations.

The Farr Institute’s five years of funding came to a close in October 2018 with the newly established Health Data Research UK stepping into position as the country’s national health data science institute. HDR UK is a joint investment led by the Medical Research Council, together with the National Institute for Health Research (England), the Chief Scientist Office (Scotland), Health and Care Research Wales, Health and Social Care Research and Development Division (Public Health Agency, Northern Ireland), the Engineering and Physical Sciences Research Council, the Economic and Social Research Council, the British Heart Foundation and Wellcome. There are 6 geographically placed centres, one of which is HDR London which includes Imperial College London.

University College London will undertake four studies on variation in healthy life expectancy throughout childhood and adulthood in England. This programme proposes 4 clinically relevant research studies investigating the relationship between age at which people develop morbidities or disability requiring hospital admission and subsequent survival. A commonly accepted criterion for prioritising health interventions is not solely to prolong life but to keep people healthy longer (Objective 4 NHS Mandate). Substantial inequalities exist in disability-free and healthy life expectancies across cross-sections of the population, particularly in groups with deprived socio-economic characteristics. The NHS Constitution for England states that the NHS has a ‘social duty to promote equality through the services it provides and to pay particular attention to groups or sections of society where improvements in health and life expectancy are not keeping pace with the rest of the population’ (DH 2015). The dual challenge researchers will tackle is, therefore, to understand how specific conditions lead to an overall degradation of health and even to death, and how this burden of illness is distributed across geographical areas and patient characteristics. Such evidence is used by health organisations such as Public Health England (PHE) and the National Institute for Health and Care Excellence (NICE) to issue clinical and policy guidelines.

Using Hospital Episodes Statistics (HES) linked to death registrations, the researchers propose to develop ways of measuring how long people live for without significant ill health (or morbidity). The researchers refer to this as healthy life expectancy. The researchers will assess how healthy life expectancy varies between different groups according to risk factors such as deprivation, ethnicity, sex, and their health conditions. The researchers will analyse patient risk factors such as; age when the researchers start to follow them from a similar point in their life course in HES (the researchers call this their inception to the cohort), their health condition, and underlying chronic conditions, that the researchers measure from coded hospital records. The researchers will also measure hospital contacts (including inpatient, outpatient and A&E) and GP registration. The researchers will take into account factors reflecting the service for patients such as the local authority and hospital they attend when the researchers start follow up. Outcomes related to healthy life expectancy will be defined as time to death and time to indicators of loss of healthy life e.g. occurrence of complications or morbidity identified in subsequent hospital presentations.

The researchers will not undertake any analyses outside the aims and objectives specified for the four studies listed below.

As part of work within Health Data Research UK (HDR-UK), UCL will harmonise and standardise methods for analysing cohorts across the life course, range of conditions and demographic indices. UCL aim to ensure consistent validation and use of algorithms across conditions and age groups to enable comparability of the studies. The four studies will also develop new tools to advance policy and research into healthy life expectancy and for use in outputs by other agencies (e.g. Public Health England (PHE)).

The approach of using information from the whole HES record across the life course represents a major advance over current estimates based on potential years of life lost from chronic conditions such as heart disease or COPD recorded on death registrations and for single conditions. These methods fail to capture conditions and risk factors earlier on in the life course, which often do not appear on death certificates and might offer opportunities for intervention. Processing of hospital records linked to death records can inform clinicians and patients about the future risks for patients given a point such as first diagnosis, when information about their risk of death or other serious adverse outcomes (the researchers call this prognosis) is crucial for patients and clinicians deciding on future treatment plans. Further, the methods will provide population-based information on the burden of illness in the population of people who are still alive. UCL aim to inform policy about the growing gap between life expectancy and healthy life expectancy: people are living longer but spending more years with disability. How can years of healthy life be increased, in whom, and at what point in disease trajectories is intervention likely to have most impact? The studies will develop methods to guide patients, clinicians and policy makers, and will inform key policy programmes such as the English Burden of Disease (EBD) Study led by PHE. No data will be shared with PHE.

None of the collaborating teams or funders listed in this application will have access to the data. The full research database will be accessible only to researchers who are substantive employees within UCL, or PhD students (with Honorary Contracts) on UCL MSc and doctorate courses under the supervision of UCL substantive employees.

The four research studies are as follows:

Study 1. Population-based indicators of healthy life expectancy related to COPD (Chronic Obstructive Pulmonary Disease).

A study focused on COPD will develop methods to use HES and death registration records to produce population health indicators to assess variations in healthy life expectancy. Hospital and death records provide survival data and information that can be used to predict patient-reported health and disability status. Focusing on COPD as an exemplar condition, the study will use longitudinal HES-mortality data for patients with any diagnoses related to COPD (using the years of hospital records for an individual patient, the researchers use this to characterise when and what health problems occurred that required presentation to hospital, these longitudinal trajectories, combined from all patients to estimate ‘population at risk’ statistics for each area), based on patterns of attendance and survival. Researchers will use this approach to infer the size of the population at risk of ill health (or severe degradation of their health) leading to emergency admission or death. The study will estimate the size of populations at risk of an acute exacerbation broken down into age, sex, ethnicity and either local authority district or main treatment hospital catchment. Estimates of population-at-risk sizes for COPD/other related conditions (heart failure, asthma) may be used to inform health organisations across England of who may benefit from a health intervention for an ambulatory care-sensitive condition to improve service delivery to help people recover from poor health or stay in good health longer.

Data minimisation has been applied in the following ways for this specific project:

The study will include all adult patients (aged 18y+ from 01.04.1997) with at least one of the following:

(a) an inpatient/day case hospital emergency admission with diagnosis codes J40, J41, J42, J43, J44 or J47 (COPD) or J45 (asthma) or I11, I25, I42, I50 (heart failure).

(b) registered cause of deaths J40-J44, J47, I11, I20-I25, I42, I50. It will include the Admitted Patient Care, the Outpatients and the Critical Care HES data.

It will include all deaths both linked and unlinked. In order to perform survival and health expectancy analyses, a long time period is required to model the length of time from first diagnosis to death. Month and year of death (rather than full date of death) is being supplied. It is not possible to minimise by geography and patients move residence across their lifetime and the researchers are required to follow them up.

Funder: HDR-UK.

Study 2. Reproductive health

Pregnancy and admission for delivery is an opportunity for health services to identify and address underlying chronic health, pregnancy complications and psychosocial needs in mothers and to plan future care for the mother and child. Maternal mortality during delivery is rare, but mortality in the long-term can be high for some groups (e.g. drug or alcohol misusing pregnant mothers). Likewise, adverse birth outcomes such as very pre-term birth and congenital malformations may impact on the future health and life expectancy of the child and impact maternal health. Better information on healthy life expectancy, including expected disability years of life for mothers, children and young people could inform more proactive healthcare with benefits to the NHS. This study will focus on mortality up to 15 years after delivery for mothers with high risk characteristics (e.g. underlying chronic conditions and/or psychosocial needs) and children with risk factors at birth (e.g. preterm birth, congenital malformations) compared with unaffected populations. Longitudinal HES records will provide a measure of onset of conditions/procedures (e.g. diagnosis of cerebral palsy or epilepsy in children, admissions for mental illness in mothers). UCL will use published external evidence on health states to estimate years lived with disability.

Data minimisation has been applied in the following ways for this specific project:

The data will be minimised to children and young people aged < 25 years (ie the oldest one with follow up would be 45 – and aged 29.9 in 1997). Along with all women with any codes indicating a live or stillbirth (from 1.4.97 onwards). It will include the Admitted Patient Care, the Outpatients, the Accident and Emergency and the Critical Care HES data. It will include all deaths both linked and unlinked for the children and young people <25 and linked only for the maternities. The data will be further minimised to include month and year of death (rather than full date of death). However, full date of death is required for infants. Data is required for all available years from 1997 onwards in order to evaluate questions about changes over time, this is to determine the shift from health to chronic conditions, to death. The data will be used to evaluate questions about changes over time.

Funder: UCL GOS Institute of Child Health and Institute of Women’s Health and Epidemiology and Health Care, HDR-UK and Great Ormond Street Hospital (GOSH) Biomedical Research Centre (BRC).

Study 3. Inequities

Health inequities are unfair, avoidable differences in health that occur across the gradient of social deprivation. There is a lack of large-scale research in the UK that has examined whether deprivation has more impact on health outcomes within certain ethnicities. Some groups (e.g. homeless people/drug users) experience much more extreme health inequities. With increasing levels of homelessness and alcohol- and drug-related deaths it is important that the public response to health inequities encompasses the gradient across all groups and more extreme health inequities. Recent research has demonstrated the importance of overlapping risk factors for extreme health inequalities including homelessness and drug use – an area UCL are calling inclusion health. Estimating the extent and nature of hospital contact for inclusion health populations is fundamental to understanding the need for preventive services in secondary care. In the most comprehensive assessment of NHS homeless hospital care utilisation to date, the Department of Health estimated need using the ‘No Fixed Abode’ (NFA) code in HES data. NFA is a proxy indicator for single people sleeping rough or in a hostel. This work is over 10 years old and UCL will address the recent lack of research delineating the impact of such extreme social exclusion on healthy life expectancy and access to health care. Current population health assessments guiding policy and practice do not address the extremes of social exclusion or interactions between social deprivation and exclusion, ethnicity and health. This hinders the development of targeted approaches to prevention and uptake of services to address inequities in healthy life expectancy. The researchers will address this by comparing key socially excluded groups (homeless people, injecting drug users) and ethnic minority groups against the general population across different strata of social deprivation. Looking at this issue across deprivation categories and including extreme exclusion will allow researchers to identify the need and rationale for prevention opportunities in hospital settings for all deprived groups.

Data minimisation has been applied in the following ways for this specific project:

The data will include all patients (aged 18y+ from 01.04.1997 onwards) with at least one of the following: a) Substance use disorders - all patients with diagnosis codes: SUD: F11 F14 b) Homeless – all patients with any use of code: z59 c) Homeless: No Fixed Abode marked used for discharge and postcode of patient, or in any other field d) Ethnic groups: all HES records for all ethnic ethnic groups, e) IMD groups: all records for all IMD quintiles. The datasets used will Include Admitted Patient Care, Outpatients, Critical Care, and Accident and Emergency HES data and all death records linked and unlinked. This amount of data is requested to capture past history and long term follow up to measure morbidities and mortality and produce results on reductions/exacerbations of inequalities in healthy life expectancies. Only month and year of death have been requested, as has year of birth or age rather than full date of death and full date of birth.

Funders: Expected component of forthcoming UK-Prevention Research Partnership research application, NIHR PhD Fellowship (Luchenski), Health Data Research UK, North Thames ARC (Applied Research Collaboration) - funding is in place until 2024).

Study 4. Multimorbidity: detecting high risk clusters

The number of people living with multiple diseases is increasing. The Academy of Medical Sciences report ‘Multimorbidity: a priority for global health research’ recently listed numerous evidence gaps in multimorbidity research and called for more research on the scale and nature of multimorbidity. The first aim is to answer research priority 5 from the report: “What strategies are best able to maximise the benefits and limit the risks of treatment among patients with multimorbidity?”. To do this, the researchers need first to address research priority 1: “What are the trends and patterns in multimorbidity?”. The researchers will (a) identify the most common multimorbidity clusters at the population level using cluster and network analysis (b) follow changes in these clusters over the lifecourse (c) distinguish clusters which are associated with functional deficits, disability, or mortality. The second aim is to develop tools to determine which diseases cluster together more often than expected by chance.

Data minimisation has been applied in the following ways for this specific project:

Data requested includes all patients with a HES admission or death registration record, in all age groups from 1.4.97 onwards. The data requested will Include Admitted Patient Care, Outpatients, and Critical Care HES data and all death records linked and unlinked. The data requested is pseudonymised. It is not possible to reduce the number of years as the study requires as long a follow up as possible. Data for all of England is required to follow patient admissions and deaths across different locations in England. The data has also been restricted to adults only. Month and year of death will be disseminated, rather than full date of death. Year of birth or age is also sufficient rather than full date of birth.

Funders: Wellcome Trust clinical PhD studentship; Rutherford Fellowship-MRC, UCLH BRC, HDR-UK.

Programme level minimisation:

Data has been requested to allow the completion of each specific project, without the need for NHS Digital to disseminate data multiple times. Only datasets required for each specific study have been requested. Only pseudonymised data will be disseminated (e.g. year of birth rather than full date of birth). A large number of years have been requested to allow the projects to complete their work (e.g. long term follow-up to asses changes over time). The data is also not able to be minimised by geography for the same reason. The data cannot be restricted by age as each project requires different age ranges. Only those fields that are pertinent to each project have been requested.

The legal basis for processing personal data for this purpose data at UCL falls under Article 6(1)(e) of the General Data Protection Regulations (GDPR), i.e. “a task carried out in the public interest”. It also falls under Article 9(2)(j), “processing is necessary for archiving purposes in the public interest, scientific or historical research purposes or statistical purposes”. It is in the public interest because the programme of research will facilitate the conduct of clinically relevant research into variation in healthy life expectancy in children and adults in England.

Previously approved processing activities:

UCL are requesting month of death (date for infants) and causes of death. This level of detail is essential to be able to use death registration to identify morbidities in people who die and to determine age at death and time from key events such as birth or first diagnosis. Timing of death is also important to examine changes in external factors such as flu season, environment (e.g. temperature), or services. The studies also require unlinked death registrations to identify individuals with conditions of interest who were not seen in hospital or were missed links to HES.

UCL also require hospital spell number and augmented care period local ID to produce rates of health care utilisation (emergency admissions, elective admissions, 30-day relapse admission, critical care admission) for the specific cohorts defined from an inception event such as a first diagnosis or a first admission.

UCL are requesting data since 1997 (a) to enable the inclusion of sufficient numbers of patients in cohorts for less common conditions, for complex conditions or combinations of conditions; (b) to gain certainty regarding the time of first diagnoses/admissions took place (which requires several years of data prior to the first diagnosis); (c) to allow adequate follow-up periods: cohort studies on respiratory disease or diabetes usually require 15 years of follow up when it comes to mortality; (d) to use past admissions to characterise underlying risk factors (e.g. birth characteristics or prior admission with a chronic condition); (e) to examine evidence of changes over time and across birth/diagnosis cohorts in coding, hospitalisation practices; (f) to document changes over time in healthy life expectancy.

Data extracts are transferred from NHS Digital to the UCL Data Safe Haven by an authorised data scientist within the Institute of Health Informatics (IHI) using Secure Electronic File Transfer in the form of de-identified extracts on a yearly basis. They come with encrypted pseudonymised identifiers (pseudonymised HESID) which indexes records relating to any given patient.

UCL Institute of Health Informatics data scientists manage the dataset within the UCL Data Safe Haven and enrich the data with suitable metadata:

• NHS Classifications ICD-10 and OPCS-4 tables,

• Organisational Data Service tables referencing NHS trusts and hospitals,

• Data Quality Maturity Index information from NHS Digital reports,

. geographical coordinates of census Lower layer Super Output Areas (LSOAs) and

relation with other health and administrative geographical units,

• Area-level aggregate characteristics for LSOAs and larger geographical units, including environmental exposure, indicators of the physical, social and built environment such as climate, deprivation, health prevalence or housing density.

Linkage with data other than these area- or organisation-level data is not allowed. There will be no linkage to other record level datasets.

Extracts will be prepared for the 4 studies to minimise the amount of data seen by researchers. For example, the data scientist will provide extracts related to relevant age ranges, or conditions, as required for the study.

Data scientists at UCL will use standardised data cleaning methods, algorithms and coding clusters for defining cohorts, exposures and outcomes, so that healthy life expectancy can be compared across different age groups and different condition-specific or age-related cohorts. Methods for defining phenotypes or condition-specific cohorts, and their validation across age and condition groups, constitute considerable added value brought by UCL to the HES-mortality dataset. Researchers will undertake univariate and multivariate analyses using standard statistical packages (R and Stata) to determine associations between a range of demographic, clinical and environmental factors (e.g. season, area) and life expectancy, constructing relevant control groups and making appropriate adjustments for confounders. HES-mortality data will be used to determine a range of clinical outcomes, to estimate health states and determine cause-specific mortality, as part of analyses of healthy life expectancy.

No record-level data will be exported outside the Data Safe Haven and no data will be shared with any third-party organisation or user. No aggregate data with small cell size in breach of NHS Digital requirements will be exported from the UCL Data Safe Haven.

Aggregate data outputs for export can only be exported through the established disclosure control procedure by a data scientist or PI. Data scientists authorised to export aggregate outputs will control outputs by scrutinising aggregate tables and figures to assess whether the outputs meet the following requirements:

• the HES analysis guide

• the ICO Anonymisation Code of Practice

• the Anonymisation Standard for Publishing Health and Social Care Data Specification (ISB1523)

• the ONS Disclosure control guidance for birth and death statistics

• the SLMS Health Informatics – Pseudonymisation ISO/TS 25237:2008 Overview for audit purposes

Outputs deemed to have an excessive re-identification risk will not be released and will instead be returned to the researcher for further processing.

Authorised researchers will hold a substantive contract with UCL, apart from up to eight PhD students (two per study - see further information under amendment request below). Researchers and PhD students analysing the data will only have access to minimised extracts. Only designated staff within the UCL Data Safe Haven who are responsible for managing the data (cleaning, validating, downloading and extracting data) will have access to the full dataset across all years. The data requested will only be used for the purposes described in this application.

Principal investigators of each of the four studies listed will be responsible for monitoring each specific study undertaken as part of the study. All staff and PhD students working with the data will undergo mandatory annual training in information governance run by the UCL Information Services Division (ISD), who maintain a log of course attendees.

When the pseudonymised HES and civil registration deaths data extract is available from NHS Digital, a nominated researcher will download the data and immediately transfer it into the UCL data safe haven. Once in the data safe haven, researchers based at the Institute of Child Health and Farr Institute of Health Informatics London (the researchers are all substantive employees of UCL, apart from between one to eight PhD students) will be able to access the data in the safe haven.

The IDHS safe haven operates as a walled space and researchers are not able to connect to the internet or export data from it.

Data processing for the four research studies:

Study 1. Population-based indicators of healthy life expectancy related to COPD.

This study will involve a combination of life table modelling and survival analysis which require follow-up until death and hence require past HES records and 10 to 15 years of follow-up. UCL therefore require data HES-Civil Registration and unlinked deaths from April 1997 onwards for adults aged 18 years or more with specific ICD10 codes relating to COPD, asthma or heart failure.

Study 2. Reproductive health

Data comprising the full HES record are required for all women with indicators of a live or stillbirth and for children (<25y) recorded in HES from 1997 until the most recent data available. This is to build cohort life tables describing changes in mortality across the period by age group and across clinical conditions, early risk factors (e.g. preterm birth or young age at onset of chronic conditions) and demographic risk factors. Information from HES-mortality data will be used to model the relationship between deprivation, ethnicity and specific conditions or comorbidites with mortality to determine how healthy life expectancy is changing over time. Causes of death are required to characterise clustering of health events and underlying conditions related to death). Date of death is required for infants <365 days old as age at death is strongly associated with pre and postnatal causal factors in infancy. Unlinked deaths (i.e. death registration records not linked to HES) are required to minimise biases due to the substantial proportion of children and young people whose death is not linked to a hospital record. Information on area of residence, hospital trust and GP practice, will be used to identify clustering of comorbidity and mortality, for mothers, children and within mother-child pairs. These results will be relevant to service commissioning for maternity, child health and primary care.

Study 3. Inequities

HES data enables exploration of the social gradient using the Index of Multiple Deprivation based on postcode of residence. Ethnicity codes enable identification of how social deprivation interacts with ethnicity to impact on healthy life expectancy. Groups experiencing more extreme disadvantage can be harder to identify but specific codes related to substance use disorders and homelessness as well as administrative codes such as “No fixed abode” can support this. HES and death registration data will be used to develop measures of healthy life expectancy and measures of avoidable mortality. UCL requires all HES records for adults aged 18 yrs and above linked to death records (including month and causes of death). UCL also requires unlinked death registrations to identify deaths that could identify individuals not seen in hospital or missed links to HES. UCL requires data from April 1997 to date to capture past history and long term follow up to measure morbidities and mortality and produce results on reductions/exacerbation of health inequities since 2001, when the NHS was officially mandated to address health inequities. Local authority of residence codes, CCG, GP practice and trust codes are needed to produce local level performance indicators and to link to spending measures via the Spend and Outcome Tool.

Study 4. Multimorbidity: detecting high risk clusters

UCL will use The Academy of Medical Sciences definition of multimorbidity as the co-existence of two or more of the following: non-communicable long-term conditions, mental health disorders, or chronic infections, with details of functional deficits or disabilities, frailty, states of poor health such as obesity, or health-related behaviours such as alcohol misuse. The case definition for a long-term condition will be based on codelists generated using HES diagnoses and procedural codes, and ICD codes recorded at death registration. Mortality data will be obtained from the death registration to estimate 5, 10, 15, and 20-year mortality risks. Cluster analysis will be applied to patient-disease matrices to identify disease clusters by age and sex. Comorbidity measures such as the Pearson correlation coefficient and the relative risk ratio which compare observed with expected co-occurrence of diseases will form the basis of network analysis methods which will investigate the correlation between pairwise diseases. Researchers will apply similar methods across the adult age range. UCL require HES admission data linked to mortality records and unlinked deaths from April 1997 onwards for adults (18y+). UCL will further minimise the data request by restricting to admission data and mortality only (i.e: no A&E or OP data) and by restricting to month of death.

This request is to increase the number of PhD students who can access the data from one (as per the previous approval) to up to eight.

Each separate research study will have two PhD students working on the data. UCL have requested the additional PhD students to analyse the HES data as they are core members of UCL’s research teams. Allowing PhD students to contribute to research using HES will allow UCL to work on the data faster, and build capacity for future health record research. The PhD students are an important part of the research workforce in UCL.

All those involved in the processing of the data are substantive employees of UCL, or students on UCL MSc and doctorate courses under the supervision of UCL substantive employees. The work undertaken by the students is only for the purpose stated in this Purpose section.

All UCL students are expected to undertake annual training on handling highly confidential information. All Trainees and students register for and complete NHS Digital’s Data Security Awareness (NHSD) course provided by e-Learning for Health. The course covers data security awareness, the law, threats to data security, breaches and incidents, and the General Data Protection Regulation.

UCL has a specific data protection and information security policy, which applies to all staff and students when processing personal data on behalf of UCL. All UCL students working on the study are bound by this policy, and that they will face potential sanctions in the event of a breach of the policy.

All students sign up to the UCL's Academic Manual. The Student Academic Misconduct section of the 2019-2020 manual Section 9.1, item 3 states "All instances of Research Misconduct whether by taught students, research students or members of staff will be investigated under UCL’s Procedure for Investigating and Resolving Allegations of Misconduct in Academic Research".

UCL will provide a signed Honorary Contracts (Student) to NHS Digital for a PhD student (up to a maximum of eight) before the student can access data under this data sharing agreeme

Knowledge resulting from this programme will be communicated via HDR-UK meetings and other mechanisms, including publication in high impact journals, presentations at key conferences and events, and teaching and training activities.

Publications and reports include:

Multiple reports will be produced for publication for each of the four studies. These reports will be disseminated at different stages. Initial outputs will be preliminary reports to promote discussion and critique by clinicians, service providers, public and scientists. Feedback from these discussions then informs peer reviewed and other publications.

By 12 months: Preliminary analyses for specific studies to develop the four studies, for example, reporting on the development of methods for phenotyping clinical conditions and computational methods, will be presented at a series of meetings within one year of receipt of the data, including Faculty or Institute seminars open to UCL staff, academics outside of UCL and NHS clinicians, clinical/health informatics interface meetings; North Thames ARC/UCL Partner activities; and HDR-UK seminars and conferences. These fora will provide opportunities for discussion of preliminary results throughout the programme.

By 24 months: The applicants will submit abstracts for presentation of early findings from the 4 studies at key national and international clinical and data linkage conferences, e.g. Informatics for Health, the International Population Data Linkage conference, Medical Informatics Europe and Public Health Informatics. The applicants will feedback findings from individual studies to NHS clinicians through their involvement in Biomedical Research Centres at GOSH and UCL Hospital, through UCL Partners and links to AHSNs networks across HDR-UK London, and through working with policy makers through four policy research units based at UCL, and through presentations, meetings and dissemination of working papers through HDR-UK.

Tools created include:

Within 24 months: New phenotypes, analytic scripts, tools and algorithms developed as part of the programme will be published on the UCL Institute of Health Informatics data portal, a resource made available for researchers to promote the transparent and scalable use of linked health data for research and benefits realisation for the NHS.

By 36 months, UCL will have produced papers for publication on all four studies and published tools on the website of UCL Institute. These working papers will be presented to relevant NHS bodies such as Public Health England, and NHS Digital’s methodological review panels as well as through HDR-UK. Each study will be expected to have at least one research report submitted for publication in key scientific journals by 36 months after the receipt of data. All information allowed to be removed from the safe haven will be anonymised and in aggregate form only, which will be checked to comply with NHS Digital guidelines for publications. UCL will publish papers in high impact, open access scientific journals such as the Lancet, PLoS One, BMJ Open and Arch Dis Child, Heart, PlosMed, Journal of Public Health. Reports of findings will be shared with funders e.g. NIHR, MRC, and other stakeholders as appropriate. UCL will publish summaries of findings in newsletters disseminated through the IHI website, the North Thames ARC, relevant clinical groups within UCL Partners, and to NHS organisations such as Public Health England, NHS Digital, Department of Health and NHS England.

Information for the public include:

From the outset, a public list of approved protocols and publications (with lay summaries) will be maintained on the UCL Institute of Health Informatics website via the UCL Faculty of Population Health Sciences. Findings will also be shared with patient groups (e.g. National Children’s Bureau, Generation R, Great Ormond Street Hospital (GOSH) young people advisory group) and UCLH ‘About Me’ public engagement group.

It is anticipated that the public will be involved in these research studies from the outset. Case study examples will be developed to show how patient health data is used by researchers to inform decisions to improve health, with examples of how data are handled, as part of the applicants mission to promote understanding of the use of health data among the public (see above).

All outputs and publications contain aggregated data with small numbers suppressed in line with the HES Analysis Guide.

Outputs specific to the 4 research studies:

Study 1. Population-based indicators of healthy life expectancy related to COPD.

Methodological working papers will be disseminated to the RCP, PHE, NHS Digital’s Methodology Review Panel and NICE and to clinicians through the North Thames CLARHC. This knowledge transfer exercise would serve to cross-examine data quality questions before working papers are submitted for publication. Further engagement with NHS Digital and the Government Statistical Service could lead to the development of new official statistics methodology.

Study 2. Reproductive health

The study will initially describe variation in maternal healthy life expectancy following pregnancy outcome, over time, and by area, age and specific risk groups. Healthy life expectancy for children will be reported for specific high-risk groups (e.g. according to gestational age at birth, chronic conditions or congenital malformations). Research papers will report results of clustering of maternal and child morbidity and discuss relevance for health interventions to families, mothers and children.

Study 3. Inequities

Outputs of this research will include development of a ‘Toolkit’ report to translate research findings into prevention practice and online resources, working in collaboration with Pathway (national and local teams), The Faculty for Homeless and Inclusion Health and Pathway, University College London Hospital (UCLH), and collaborators from Health Data Research UK and the UCLH NIHR Biomedical Research Centre (BRC). Findings will be published in research papers describing the inequities. The research will be used to develop a series of indicators to monitor health inequities in these populations at national and local level. Developments will be fed back to health organisations (e.g. PHE, NHSD, NHS England) to consider for their own public health statistics production. By providing burden estimates specific to these vulnerable populations, the work will also contribute to the GBD Estimates, which are highly influential in guiding policy.

Study 4. Multimorbidity: detecting high risk clusters

This study will generate tools to inform clinicians, healthcare providers, and researchers of the components and progression of high-risk multimorbidity clusters. These tools will be made available on the UCL IHI website, shared with relevant organisations (e.g. PHE, NHSD, commissioning groups), and submitted for publication. The study will generate information on multimorbidity prevalence rates and associated risk factors for multimorbidity clusters at high risk of mortality or reduced healthy life expectancy. Findings will be relevant to policy and service provision and for developing trials (e.g. relevant to industry and NIHR).

The programme of research will facilitate the conduct of clinically relevant research into variation in healthy life expectancy in children and adults in England. Expected measurable benefits have been outlined under each of the four research themes. More broadly, these benefits are of three kinds:

(i) Methodological benefits in improving the potential uses of HES and mortality data: UCL will evaluate new ways of analysing hospital data cohorts based on inception events such as occurrence of a first diagnosis or admission, and approaches for defining specific conditions and comparators. These benefits will be disseminated to the research community through research publications. In addition, UCL has planned very specific research in statistical inference and modelling that will enable the production of new public health indicators (e.g. study 1). Widening the range, frequency and quality of public health indicators will benefit a range of statistical and health organisations in monitoring the health of the population, shaping policy, and allocating resources. UCL will ensure these benefits are delivered by engaging with NHS bodies (e.g. Public Health England) and NHS Digital’s Methodological Review Panel to seek peer review and examine opportunities for implementation in statistics production. UCL will also engage with patients early on to generate evidence around what they understand from survival and healthy life expectancy estimates, and whether it can support them in making individual care decisions. Public engagement is relevant to all 4 studies.

(ii) there will be benefits to NHS systems and services nationally and patient outcomes across a wide range of clinical disease areas, including, but not limited to, cardiovascular diseases, cancers, renal diseases, respiratory diseases and mental health in adulthood. In childhood, priorities will include mortality in early childhood compared with other developed countries (recognised as a priority for the NHS), outcomes for children with rare or chronic conditions, vulnerable adolescents (e.g. those admitted for injury related to self-harm, drug or alcohol use or violence), and the impact of transition from paediatric to adult services on service use and healthy life expectancy. The range (richness) and breadth (from 1997 to latest available) of data will provide information on healthy life expectancy across the age range from childhood to early adulthood, and from early adulthood middle to old age, and for a range of conditions, comorbidities and patient circumstances defined by area indicators (e.g. socioeconomic status, or local authority) or indicators of vulnerability (e.g. ethnic minorities, the homeless). The information from the programme can be used by policy makers and service commissioners to address disparities in healthy life expectancy.

(iii) Findings from the 4 studies on healthy life expectancy will shape policy by feeding into national programmes. The potential to increase the quality and breadth of risk factors examined as part of the Global Burden of Disease study in collaboration with Public Health England. UCL researchers currently contributing to the Global and English Burden of Disease studies will apply methodological results from study 1 to propose methodological improvements to the studies. This will result in higher quality disease prevalence and health burden information required to better inform the prioritisation of health policy across a widened range of risk factors than is currently possible. For more specific diseases and health care procedures, highlighting variation in outcomes and services for rare diseases and for maternal and child health and health inequalities, will inform the provision of health care by determining which patient groups are most at risk of poor healthy life expectancy; how outcomes vary across regions and hospital trusts, and where risk assessment and interventions should be targeted.

Benefits of the four studies include:

Study 1. Population-based indicators of healthy life expectancy related to COPD.

The investigators will widen the range of uses researchers can make of hospital data in the UK, taking advantage of the uniquely high coverage of HES compared to similar hospital datasets internationally. They will produce working papers that are directly relevant to the NHS and to PHE English Burden of Disease estimates. Production of new outputs on population-at-risk size estimates, disability-adjusted life years and healthy life years lost can produce better and more disaggregated information, enabling clinician and public health professionals to improve the way they prioritise and target health interventions e.g. production of estimates of the size of populations at risk of an acute exacerbation of COPD for small groups and catchment areas will enable acute trusts, commissioners and organisations such as NICE to measure severity of illness in groups that would benefit from interventions such as smoking cessation or winter pneumococcal immunisation.

Study 2. Reproductive health

Findings will inform targeting of interventions by obstetric and child health services and public health policy for families where the mother and/or child is at high risk of adverse outcomes and reduced healthy life expectancy.

Study 3. Inequities

This study will generate information describing how HES-mortality data can be used to inform the development of health promotion and health service responses to health inequities at national and local levels. It will inform resource allocation, evaluation of interventions to address health inequities and monitoring of local authority and health service performance in relation to health.

This work will draw together evidence of needs and opportunities for hospital teams to implement evidence-based prevention for deprived patients. By providing evidence, access to prevention services can be improved, and with sufficient scale-up, contribute to a reduction in preventable disease, inequalities, and NHS costs.

Study 4. Multimorbidity: detecting high risk clusters

UCL researchers will create tools to identify patients with multiple morbidities that co-occur more often than expected by chance to develop preventive strategies targeted at patients in “high-risk” clusters. These tools can inform practice and policy and the selection of patients for inclusion in randomised controlled trials (RCTs). Currently, RCTs frequently exclude patients with multimorbidity and there is an urgent need to determine benefits and harms of interventions in this group and to identify high risk patients with precursor conditions that put them at risk of chronic disability or mortality who might benefit from early intervention.

Prior amendment (DARS-NIC-06527-J1Q6T-v1.3, 09/12/2020):

The additional PhD students will allow UCL to complete the detailed validation of methods and exploration of alternative approaches that is not always possible in commissioned research.

Extension request:

Due to delays in the initial release of data and disruption caused by the COVID-19 pandemic, outputs from this study have been delayed. Initial outputs described in yielded benefits below have been delivered. Additional outputs are also in progress. An extension is requested to support the ongoing delivery of outputs and upcoming publications, and any reviewer queries that require additional analyses. Follow up papers may also result from initial publications, which will all fall under existing approvals and the agreed permitted uses.

Yielded benefits to date include:

Methodological benefits and tools to improve the uses of HES and mortality data. This has involved:

• The development of methods for phenotyping clinical conditions, analysis of hospital data cohorts based on inception events, and computational methods following FAIR and open scientific data management and stewardship principles (https://github.com/UCL-CHIG).

• Regular twice annual training on the use of HES, open to the public and advancing HES use with clinicians, service providers and scientists (e.g. https://www.ucl.ac.uk/child-health/events/2021/oct/introduction-hospital-episode-statistics)

• Publications with projects and protocols (with lay summaries) maintained on the UCL website to provide information and promote awareness for the public.

Preliminary reports to promote discussion and critique by clinicians, service providers, public and scientists. For instance, engagement with NHS clinicians, applied health research networks and policy makers through the policy research units based at UCL on the ongoing:

• Mixed-method study forming an evidence-based toolkit to guide hospital care planning and practice for people experiencing homelessness. This is combining systematic reviews of hospital-based interventions, quantitative analysis to understand health needs and priorities for preventative interventions engagement with clinicians, commissioners, and people experiencing homelessness to examine implementation barriers and facilitators.

• The study of trends and patterns in survival and multimorbidity for children with chronic hereditary conditions over time (Duchenne Muscular Dystrophy (DMD)), involving multiple systems, including behavioural, cognitive, physical and psychiatric morbidities and well as respiratory, understanding the long-term risk factors and outcomes for patient groups, with PPI and clinician input defining indices of adverse health outcomes and provision of health care.

The production of research papers for publication, with research work presented at conferences and a research paper currently resubmitted for publication:

• The 'Multimorbidity patterns and risk of hospitalisation in childhood cancer survivors and children without cancer: a population study of 3.6 million children' currently under re-submission to the key scientific journal, the Lancet. This research is addressing substantial challenges in achieving personalised cancer care by creating matched patient cohorts from routinely collected electronic health records.