Does the transition from paediatric to adult healthcare lead to increased healthcare usage for young people with a life limiting condition? A quasi-experimental study

The research aims to determine whether there is an increase in healthcare use (particularly emergency healthcare use) when children with life limiting conditions transition from children's to adult services. The term Life limiting conditions (hereafter LLC) includes both life limiting conditions (these are conditions that invariably lead to premature death, e.g. Duchenne muscular dystrophy) and life-threatening conditions (these are conditions that may lead to premature death but may be cured, e.g. cancer). Examples of included conditions (the 282 most commonly recorded life limiting diagnoses in a previous extract of Hospital Episodes data) are provided to NHS Digital.

The research is in the public interest under Article 6 1(e) and Article 9 2(j) as it aims to quantify the effects of current healthcare practices and identify the impact on those affected. The research has the potential to improve healthcare for these individuals by identifying best practices.

The risks from dissemination are potential identification of individuals due to small numbers of individuals in some possible groupings (some conditions considered are very rare). This will be mitigated by following NHS Digital disclosure guidelines - either combining small groups of fewer than ten persons or censoring results for these groups. All dissemination for LLC will be at the level of diagnostic groups (e.g. cancer, neurological conditions) rather that at individual diagnosis level.

The requested data are essential for achieving the research objectives - outpatient data are needed to determine the point of transition from children's to adult services. Inpatient and A&E data are required to measure healthcare use. Death data (month and year of death only) are needed to remove individuals from the cohort after death.

The research is part of an NIHR funded fellowship looking at the transition from children to adult services for children with life limiting conditions, which commenced in 2019. The proposed analyses are a key part of the programme of research.

The broad objectives of the fellowship are:

1. Review the evidence for (quantitative evidence) a change in emergency hospital care and GP consultations during and after transition and (qualitative evidence) the reasons for this.

2. Quantify changes in care during and after the transition for young people with life limiting conditions, as compared to those with other chronic or no long term conditions:

a. in hospital care

b. in primary care

3. Determine what factors are associated frequency of GP consultation and emergency hospital care for young people with LLC during the transition.

4. Quantify the costs of any increase in emergency hospital care at the transition, for providers and patients and their families and compare these with costs of GP consultations.

Objectives 2a and 4 will be addressed using the data requested in this application. Objective 1 has been addressed through a systematic review of the literature and objectives 2b and 3 will be addressed by data requested from the Clinical Practice Research Datalink (CPRD), discussed in more detail further down in this agreement.

CPRD is a real-world research service supporting retrospective and prospective public health and clinical studies. CPRD is jointly sponsored by the Medicines and Healthcare products Regulatory Agency and the National Institute for Health Research (NIHR), as part of the Department of Health and Social Care. CPRD collects fully-coded patient electronic health records from GP Practices using the Vision® or EMIS® software systems. More detail on CPRD can be found here - https://www.cprd.com/services

The objectives addressed using data requested in this application can be further split as follows:

(i) determine basic demographic information - sex, age in each year, ethnic group, deprivation category and geographical region

(ii) Divide the population in HES experiencing transition from paediatric to adult care into three groups: those with a life limiting condition, those with a non-life limiting chronic condition and those without a long term condition. For those with a life limiting condition, category of condition will also be determined (e.g. cancer, neurological condition). Life limiting conditions will be determined from an ICD-10 (International Classification of Diseases, 10th Edition) LLC coding framework comprising 777 diagnoses.

(iii) Determine the age of transition of individuals in these groups

(iv) quantify changes in healthcare use at the transition, including (a) A&E visits per person per year, (b) Emergency inpatient admissions per person per year, (c) all inpatient admissions per person per year, (d) length of stay and (e) bed days per person per year and compare these between the groups identified in (i) and (ii)

(v) estimate the change in costs, to the NHS and to individuals/families of after transition compared to before the transition and compare these between the groups identified in (i) and (ii)

DATA REQUESTED

Pseudonymised data from the Inpatient Hospital Episodes Statistics (HES), Outpatient and A&E HES will be extracted by NHS Digital. This dataset which will include records for all individuals who were aged between 12 and 23 years at any point between 1 April 2007 and 31 March 2019. The University of York will only request data for those individuals in financial years 2006/07 (2007/08 for A&E) to 2018/19 and when they were aged 0-23 years at the first record in the year (see provided figure on cohort logic). Data should be provided in complete years - i.e. if an individual has records aged both 23 and 24 within a single financial year then all those records should be included (records when aged 24 years, within that financial year should not be excluded).

To provide some illustrative examples:

• An individual turns 23 on 20 January 2006. This person was not aged 12-23 at any point in financial years 2007/08 to 2018/19 and no data is requested for this person and he is not included in the study.

• An individual turns 12 on 1 April 2007. This person was aged 12-23 for the whole 2007/08 to 2018/19 study period (aged 23 in the 2018.19 financial year). Her records are requested for years 2007/08 to 2018/19 in A&E data and from 2006/07 to 2018/19 in APC and outpatient datasets (as she was included in the cohort and was aged 11, within ages 0-23 in 2006/07). Outcomes (objective iv and v) are analysed in all years 2007/08-2018/19, when she was aged 12 to 23. Categorisation under objectives (i to iii) takes place from years 2006/07 to 2018/19 when she was aged 11 to 23.

• An individual turns 12 on 30 May 2000. She was aged 12-23 years of age for the years 2007/08 to 2011/12. Data are requested for her for years 2007/08 to 2011/12 in A&E and from 2006/07 to 2011/12 in APC and outpatient data. Outcomes (objective iv and v) are analysed in years 2007/08-2011/12, when she was aged 19-23. Categorisation under objectives (i to iii) takes place from years 2006/07 to 2011/12, when she was aged 18 to 23.

• An individual turns 12 on 1 Junes 2018. He was aged 12 to 23 years for the 2018/19 year only. Data are requested for him for years 2007/08 to 2018/19 in A&E and 2006/07 to 2018/19 in APC and outpatient data. Outcomes (objective iv and v) are analysed in 2018/19 only, when he was aged 12. Categorisation under objectives (i to iii) takes place from years 2006/07 to 2018/19, when he was aged 0-12.

The study period of 1 April 2007 onwards has been chosen to match availability of A&E data (A&E visits are a key outcome). An additional preceding year of data (2006/07) is requested for APC and outpatient datasets as this will enable identification of diagnoses and transition points prior to the start of the study period.

Objective (i) requires all data possible for individuals included in the study, to pick demographic information that is not recorded in each record or often recorded inconsistently (e.g. ethnic group). Therefore data prior to age 12 are requested for included individuals.

For objective (ii), the additional data requested (i.e. data prior to age 12) are required to put individuals in one of the three groups identified, based on presence or absence of diagnostic codes and also patterns of hospital care - for example, an individual lacking diagnoses defined a-priori as indicative of a LLC or chronic condition but nonetheless having frequent outpatient appointments or inpatient admissions may be excluded from the group with no long term conditions. To form the three groups, data on all individuals in HES within the requested age range during the study period are required.

Objective (iii), determining transition points from paediatric to adult care, requires data from the APC and outpatient datasets, as transitions will be determined from the main speciality of consultants providing care. A year of data (2006/07) is required prior to the main study period to identify whether individuals in adult care in 2007/08 have just undergone transition or transitioned at an earlier point.

Objectives (iv) and (v) require records from the APC and A&E datasets as the outcomes of interest (hospital admissions and A&E visits are contained within these datasets). The 12-23 year age group is chosen to provide at least four years of data on each side of transition, assumed to take place from 16 to 19 years of age.

The years requested will enable the identification of any trends over time and also maximise power in the analyses in the smallest group, those with life limiting conditions.

Data are required at record level as there is a need to determine transition points for individuals and also to count individual events over time so that individual points of transition can be linked to individual hospital healthcare use. It is not necessary to identify individuals, but pseudonymised data are necessary to link healthcare events for the same individuals within and between the datasets requested. All records, even those that may initially be considered less relevant, such as maternity records for the cohort, are relevant to objectives (i) and (ii) and can also be used to determine presence of individuals in England within a year for time at risk calculations in the analyses.

National data are required to (a) maximise the sample size for those with life limiting conditions, enabling analyses of subgroups, (b) identify any regional variations that may indicate areas of best practice to inform future provision of services and (c) capture any regional variations in conditions - focussing on a particular region may give results that are not representative of the national situation. The results will be used to inform care nationally.

There are no other data available that provide information on hospital care for individuals in the number required (CPRD can provide HES data, but in much smaller quantities, with marginal power to meet the objectives (iii) and (iv) and insufficient power to analyses particular conditions or categories of LLC).

DATA MINIMISATION

The data requested have been minimised in the following ways:

Cohort definition: data are only requested for individuals aged 12-23 years at some point during the study period. This is the minimum age range needed to provide individuals aged to provide sufficient data either side of transition (i.e. 4 years either side of transition, which is expected to take place between 16 and 19 years of age for most individuals).

Datasets: the requested data sets are only those essential to the research: admitted patient care and A&E data as these contain the outcomes of interest (healthcare use as inpatient admissions, A&E visits etc), admitted patient care to identify life limiting and chronic conditions, and outpatient and admitted patient care datasets to determine the point of transition. Mortality data are requested as it is essential to know when individuals are no longer alive to calculate time at risk for the various outcomes - omitting these data would mean that any individuals dying outside of hospital would be retained in the analyses even though they had no further risk of inpatient admission or A&E visit. Other datasets, such as adult critical care, have not been requested to minimise data. Anonymised data are not suitable as it is necessary to link data for individuals between datasets - i.e. to link A&E and inpatient admissions for an individual with that individuals date of transition based on outpatient and inpatient data. Pseudonymised data have been requested to minimise data as the research does not require knowledge of whom each pseudonymised individual is, nor does it require linkage to datasets beyond those requested (i.e. no linkage will be made to other data, for which full identification would have been necessary).

Years/Study period: this is restricted to 2007/08-2018/19, i.e. not extending the study period back further than the point where all outcomes are available as these years would have lower research value due to limited outcome data (A&E data only available from 2007/08). An additional year of data (2006/07) is requested for APC and outpatient datasets to enable grouping of individuals for objectives (i) and (ii) prior to analysis of outcomes for those individuals within the study period (if data for 2006/07 were not requested, data from the first year of the study period, 2007/08 would have to be used to group individuals into the LLC and comparator* groups). Any reduction in years requested would reduce the statistical power to identify differences between subsets of individuals based on condition or geographical differences, one of the objectives of the research, important to identify areas of good practice with a lower impact of transition.

*Comparator groups are defined as (i) those with long term but not life limiting conditions to provide a comparison group for changes in healthcare at transition ages for conditions for which the transition is believed to be well managed and (ii) those with no known long term conditions to determine any changes in healthcare use not related to transition (e.g. due to other life changes such as transition from education to employment or from secondary education to further/higher education).

Filtering: The data cannot be narrowed by geography as the research question is relevant to care across England and one of the objectives is to identify any regional variations which may point to best practice that can be applied elsewhere. Restrictions have been applied by demographics, specifically age, limiting the data requested to that on individuals within four years of age of the expected range of transition ages - all these individuals can contribute useful data to the analyses, those younger or older cannot and their data have not been requested. The data cannot be minimised by clinical factors such as diagnosis as an objective is to compare care use around the transition for those with life limiting conditions, those with chronic conditions and those with no known long term condition - i.e. whole population in the relevant age ranges.

Episodes: The data have been minimised by only requesting episodes within the study period and up to age 23 years (in combination, these provide for an absolute limit of 13 years of data for any person; many included persons will have fewer years of data due to either being first observed in Hospital Episodes Statistics later than 2006/07 or reaching age 24 prior to 2018/19). All types of episode are required as they can be used to determine diagnoses - these include maternity episodes as some of the life limiting conditions are commonly recorded pre- or immediately post-birth (and these records also provide additional sources of demographic information, e.g. ethnic group, and evidence of presence in England at that time). Elective and non-elective episodes are of interest for determination of diagnoses, determination of transition points and as outcome data (evidence of healthcare use).

Data fields requested: Only ͚general͛ data items have been requested (i.e. no ͚high risk͛ or ͚identifiable͛ data fields) and possible duplicate information/fields with similar function have been removed in consultation with NHS Digital.

The request for death records has been limited to date of death, at month and year level only to minimise data requested and the risk of identification. This is because it is only necessary to be able to remove individuals from the analyses on death (and it is sufficient to do this from the start of the month of death) not to know the precise date of death.

The research is in support of a NIHR funded doctoral fellowship (DRF-2018-11-ST2-013) and so has further benefits in research training and increasing future research capacity in the field of research into children and young people with life limiting conditions, particularly around service provision and economic evaluations for this group.

RELATIONSHIP TO WIDER RESEARCH (INCLUDING CPRD DATA REQUEST) AND HANDLING OF DATA

As set out above, the research using the requested data is part of a wider fellowship looking at the transition for children and young people with life limiting conditions. It forms a vital part of this work, particularly for objectives 2a and 4.

To address objective 2b and 3 of the fellowship, a data request has also been submitted to CPRD for all primary care records in their GOLD dataset for individuals aged 12-23 years within the study period and linked pseudonymised HES data (linked by CPRD using their own pseudonymised HES extracts; not linked to the data presently requested from NHS Digital). The data requested from CPRD serves a particular purpose in meeting objectives 2b and 3 as, unlike the data presently requested from NHS Digital, it includes primary care records. Previous work has suggested that regular contact with the same GP is associated with reduced emergency hospital care and it is intended to test whether this is true for the population going through transition.

CPRD and NHS Digital data will be held separately in separate secure databases and will be analysed entirely separately with no attempt to link the two datasets.

The research is entirely contained within the University of York Department of Health Sciences, they are the sole Data Controller who also process data. There are no other organisations involved in the data processing.

Young people, their parents and their carers at Martin House Children's Hospice will be involved in the research through patient and public involvement through the Martin House Research Centre's Family Advisory Board. Members of this group have already fed into the research proposal through an earlier research prioritisation exercise and will be consulted before and after data processing to help focus the analyses on outcomes of particular interest and to interpret the findings. This involvement will be at the level of discussing research plans and aggregated outputs and the Family Advisory Board members will have no access to the datasets. Martin House Research Centre and Martin House Children's Hospice are not involved in any activity which could lead them to being considered a joint data controller. They were not involved in decisions on how or why the data provided by NHS Digital will be processed, they will only be involved through patient and public involvement to help focus the finalized analysis into meaningful outputs.

Martin House Children's Hospice is a charity providing family-led care for children and young people with life-limiting conditions, supporting families from across West, North and East Yorkshire. More details can be found here - https://www.martinhouse.org.uk/About-Us.

Martin House Research Centre is a multi-disciplinary centre for research on the care and support of children and young people with life limiting conditions or medical complexity, their families and the workforce that care for them. The Centre is holistic in its scope, recognising that the care and support needs of children and families span clinical/medical, social, psychological, parenting/caring, spiritual, financial and practical domains. More details can be found here - https://www.york.ac.uk/healthsciences/research/public-health/projects/martinhouse/

The research is funded by the NIHR under a Doctoral Research Fellowship. The NIHR had no input into the research design.

BENEFITS

The purpose of the research is to provide a number of benefits to young people with life limiting conditions and the organisations that provide their care:

• Under objective (iv) the research will identify changes in healthcare use at transition. This could, for example, be a sharp increase in A&E use among individuals with a LLC at the transition with high associated costs of care (objective v). This would make the case to commissioners to explore alternative forms of service provision, such as extending the paediatric model of care beyond the current age of transition.

• Under objectives (i and iv), the research will also identify groups that are more or less impacted by transition and so enable local services, such as Children’s and adult Hospices to better target their resources to assist these groups.

• Under objectives (ii and iv) , the research will highlight whether there are condition groups for which the transition appears to have fewer impacts, identifying possible best practice that can be replicated by service providers elsewhere

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

The data will be accessed and processed only by employees of the data controller (University of York) who have undergone data protection training. No other individuals will have access to the data. Individuals accessing the data are all substantive employees of the University of York (none on honorary contracts). One individual is currently registered as a PhD student at the University as they are completing this work within an NIHR Doctoral Research Fellowship, but they are also a substantive employee of the University (and hold a prior PhD and have substantial experience of using HES data). The data will be accessed at the University from offices located behind locked doors on password-protected machines or via password protected virtual private network remote desktop access with data located on network drives and/or databases. No further linkage will take place beyond that performed by NHS Digital in creating the data extract (i.e. linkage between APC, A&E, outpatient datasets and death data). There will be no matching to publicly available data, other than comparisons at aggregate level. There will be no attempt to identify individuals.

There are no data flows into NHS Digital. Data will flow out of NHS Digital to the University of York at record level (pseudonymised). These will be health data. Subsequent flows of data will be only within the University of York's systems (these will also be record level pseudonymised health data).

All data storage/access/processing and analyses will be undertaken in the University of York and no data will be provided to third parties that isn't fully aggregated with small number suppression in line with NHS Digital guidelines. All outputs will be at aggregate level with small numbers suppressed in line with NHS Digital guidelines.

DATA MANAGEMENT

The following section makes reference to objectives set out in the 'Objective for processing' section.

In line with objective (i) demographic data will be determined for the cohort. The start age recorded at the first hospital episode in each year will be used to assign the age category for each individual. Sex will be coded as male, female or not known. Individuals with more than one recorded sex will be assigned the most commonly recorded sex. Ethnicity is reported by census groups in the HES data. Individuals with more than one ethnicity will be assigned the most commonly reported ethnicity unless the most common ethnicity is ͚not known͛͘. For statistical analysis the diagnoses for the group with LLC will be categorised into sub-groups based on the main ICD10 chapters (ICD10 is the coding system currently used to record diagnoses in HES and is organised into chapters). Diagnoses for those with non-LLC long term conditions will be also be grouped. Each individual will be assigned a local authority and government office region (GOR) based on the lower super output area (LSOA, a small geographical area used by the government to group individuals for assessments of deprivation and other measures) of residence. These assignments will be done per year and if an individual moved address within that year the first local authority/GOR reported that year will be used. This will allow the individual to be assigned new geographical area over the time period but not within a year. An index of multiple deprivation score (the government's preferred measure of deprivation in England) will be assigned to each individual based on the LSOA of residence.

In line with objective (ii) individuals will be assigned to one of three groups – those with life limiting conditions (LLC, using a previously defined coding framework), those with other long term conditions (based on diagnoses and/or numbers of outpatient/inpatient episodes) and those with no known long term conditions (those not in the other two groups, possibly with further restrictions on numbers of hospital records).

In line with objective (iii) the point of transition from paediatric to adult services will be determined for each individual (based on consultant main specialty in outpatient and APC datasets - consultant main specialty indicates the primary competence of the physician and includes adult and paediatric specialties). For the group with no long term conditions (with sparse records in HES) the transition will be assumed to take place at 16 years of age.

STATISTICAL ANALYSES

The number of A&E visits, inpatient admissions, emergency inpatient admissions, length of stay and bed days will be determined per person per year and compared across the groups determined in objectives (i) and (ii).

The above measures of healthcare use will then be compared across the transition for the three groups identified in objective (ii). These comparisons will use statistical methods for comparing levels of a measured variable on either side of a cut-off in time – such as an intervention such as transition – or in other variables such as age. As not all individuals will transition at the same age, techniques will be used that account for this. These analyses will be repeated, split by the groups identified in objective (i). These analyses together will meet objective (iv).

Finally, costs of care will be estimated. This will be done using details of patient conditions and actions performed recorded in the inpatient data and using standardised costs for A&E visits. In this way, costs of care per person per year can be estimated and also analysed for differences across the transition using the techniques outlined above.

OUTPUTS

Outputs will be based around the main objectives of the present study. will summarise age of transition (objective (iii) numbers of A&E and emergency and all inpatient admissions, length of stay and bed days and changes in these at transition (objective iv) and estimated costs (objective v). These will all be at aggregate level (grouped with no fewer than ten individuals in each group) with grouping into condition types (objective ii) and demographics (objective i). Coefficients of regression models (i.e. the techniques discussed above: regression discontinuity, interrupted time series and difference in difference) may also be disseminated.

The research will aim to provide the following outputs:

• At least two conference abstracts by December 2021, presenting the results of objectives (iv) and (v) to palliative care and general medical audiences. The targeted conferences will encompass clinical and academic audiences: the Royal College of GPs’ annual conference and the European Association of Palliative Care Congress .

• At least two papers submitted to peer reviewed journals by December 2021, presenting the results of objectives (iv) and (v) and again targeting clinical/academic journals e.g. British Medical Journal and Archives of Disease in Childhood.

• Lay summaries of the research findings, to be distributed to Children’s Hospices across England (and their representative associations such as Together for Short Lives), parents (and their representative groups such as Contact), specialist services commissioners, local NHS commissioners and the National Institute for Health and Care Excellence (NICE), also by December 2021.

All outputs will contain only data that is aggregated with small numbers suppressed in line with the HES Analysis Guide.

At least two conference abstracts by December 2021, presenting the results of objectives (iv) and (v) to palliative care and general medical audiences. The targeted conferences will encompass clinical and academic audiences: The Royal College of GPs annual conference and the European Association of Palliative Care Congress. These conferences are attended by senior commissioners and so this will provide a means to bring the research to their attention, helping to realise the first of the benefits in section 5(d) - i.e. making the case to commissioners to explore different forms of service provision. Details on groups most affected will also be presented and the European Association of Palliative Care Congress is particularly relevant here to reaching hospice practitioners, helping them to identify groups in most need of extra support at transition, helping to realise the second of the benefits in section 5(d).

At least two papers submitted to peer reviewed journals by December 2021, presenting the results of objectives (iv)and (v) and again targeting clinical/academic journals e.g. British Medical Journal and Archives of Disease in Childhood. These provide a secondary means of reaching commissioners and hospice practitioners as outlined above and will provide them with additional detail, helping to realise the first two benefits set out in section 5(d). However, the main benefits from these outputs will be forming part of the evidence base for inclusion in future Chief Medical Officer reports and NICE guidelines, the latter of which routinely reference published research as the basis of new policy (as set out in section 5(d)). Inclusion in NICE guidelines is another way of influencing healthcare commissioning and service provision and so is relevant to all three benefits set out in section 5(d). The papers are also directly relevant to the benefit of supporting a PhD research study set out in section 5(d).

Lay summaries of the research findings, to be distributed to Children’s Hospices across England (and their representative associations such as Together for Short Lives), parents (and their representative groups such as Contact), specialist services commissioners, local NHS commissioners and the National Institute for Health and Care Excellence(NICE), also by December 2021. This is the quickest, most accessible and direct route to reaching this audience and will help to realise the benefits of changes in service provision and targeting of extra support to the most affected groups and implementation of established best practice - i.e. all three benefits in section 5(d).

All outputs will contain only data that is aggregated with small numbers suppressed in line with the HES Analysis Guide.

The research will identify changes in healthcare use at the transition. If, for example, it identifies a sharp increase in A&E use among individuals with a LLC at the transition with high associated costs of care (objectives iv and v), this would make the case to commissioners to explore alternative forms of service provision, such as extending the paediatric model of care beyond the current age of transition. The conferences and, particularly, the directly targeted lay summaries will be the main way of reaching this audience, backed up by the detailed evidence contained within the journal papers. These impacts will be available from the date of distribution of lay summaries, i.e. by December 2021.

The research will also identify groups (objectives i and iv) that are more or less impacted by transition and so enable local services, such as children’s and adult Hospices to better target their resources to assist these groups. These impacts will come about mainly through the lay summaries distributed to hospices nationally and through hospice practitioner attendance at conferences. These impacts will also be available by December 2021.

Finally, in comparing the transition across different regions and different diagnostic groups within the population with LLC and with those with chronic conditions, the research will highlight whether there are any groups for which the transition appears to have fewer impacts (objectives ii and iv), identifying possible best practice. This will be highlighted in summaries to local commissioners so that they can look to adopt practices elsewhere that appear to work well. Again, this information will be available by December 2021.

The research outputs will provide quantitative evidence on the impacts of transition, the need for which has been highlighted in reports from the Chief Medical Officer, the Care Quality Commission and in NICE guidelines. It is intended that they should be included in future versions of such reports and NICE guidelines. These benefits will be available from the publication of peer-reviewed papers, expected from December 2021 or earlier.

Yielded Benefits is not a requirement for new applications.