Hospice UK: Hospice-led innovations for end of life care (HOLISTIC)

Hospice UK and NHS England are running a service evaluation to improve the provision of end of life care in England: Hospice-led innovations for end of life care (HOLISTIC). A major component of this service evaluation is a quantitative comparison of health services utilisation by patients in the last 90 days of life, across different end-of-life care models, to which this agreement relates.

Medical and statistical purposes

The medical purpose of Hospice UK and NHS England’s service evaluation is to improve the provision of care and treatment and the management of health and social care services for end of life care, in terms of the quality and cost of care, as well as the alignment with patients’ preferences over care location.

The study is a statistical evaluation of different care models, using quantitative data to identify significant differences in various outcome measures, including the number of hospital bed-days in the last 90 days of life, that identify the best interventions for improving end-of-life care in the UK.

Background

It matters where people spend the end of their lives. People prefer to die at home (of those who express preferences, 80% prefer to die at home and 5% in hospital), but the reality is very different (50% die in hospital, and 25% at home). The costs of end-of-life care are high, amounting to over 20% of the NHS budget, and there are many home, community, hospice and care-home based end of care settings that provide high-quality, appropriate care at substantially lower costs (in comparison to admitted patient care in NHS acute hospitals).

Project overview and objective

NHS England and Hospice UK are running a service evaluation on the effectiveness of end of life care programmes in moving patients out of hospitals at the end of their lives and into other locations of care, such as their homes, care homes and hospices, and in their ability to prevent avoidable hospital admissions for these patients.

Twenty-four hospices have identified recently-introduced care models that may reduce patients’ usage of hospital-based end-of-life care. HospiceUK’s study will evaluate these models using a quantitative and qualitative methodology. The results will be presented in academic literature, publicly available reports and as guides on how to implement the most effective models nationwide.

The quantitative analysis will use aggregated and de-identified historical secondary care records of deceased patients, from the linked HES-mortality dataset, to measure the impact created by the introduction of these recent care innovations on the average amount of time patients spend in hospital. This will allow the effectiveness of each care model to be compared. The comparison measures include the (reduction in) number of hospital bed days in the last 90 days of life, the number of A&E and inpatient admissions in the same timeframe, the probability of dying in hospital, and the number of discharges into palliative care.

The qualitative phase will consist of interviews of hospice and care home staff, patients and carers, to understand the experience of care within each new care initiative, and to gain sufficient information to provide detailed guidance to other hospices on how to implement these new programmes. This phase will be performed in collaboration with St Giles Medical. Whilst it does not relate to this agreement, it is referenced here for completeness.

The study as a whole will therefore both provide a statistical examination of the most effective care innovations in reducing hospital usage at the end of life, as well as detailed implementation guides to enable care providers nationwide to adopt these practices. The quantitative phase uses pseudonymised historical secondary care records. The Health Research Authority have confirmed that their ethical review and approval is not required for this phase.

Note on project design, funding, and parties involved

The project is joint funded by NHS England and Hospice UK.

The HOLISTIC project, and the quantitative analyses that it includes, was designed by Hospice UK and NHS England. As part of this process, Hospice UK and NHS England took advice on the design of the project and analyses from McKinsey & Company, Inc. United Kingdom, (referred to from hereon as "McKinsey"). Based in part on this advice, the final decisions on study design were taken by Hospice UK and NHS England, who act as data controllers for this project.

McKinsey has further been commissioned by Hospice UK and NHS England to be their data processor, running analyses on the linked HES-mortality datasets to test the effectiveness of the different care models.

Patient record data will be processed by McKinsey alone, and the outputs will be aggregated with small numbers suppressed in line with the HES Analysis Guide. McKinsey will store the data onsite at their London office in a locked server room compliant with industry best practices. Data will only be accessed for the specific project by a named analyst, and will be destroyed within 14 days of study completion.

NHS England and Hospice UK will only view aggregated results of the analysis with small numbers suppressed, and will not be storing or processing record-level data.

McKinsey's support to Hospice UK and NHS England, both in the provision of advice before the project design was confirmed, and in their provision of data processing activities, is on a pro-bono basis. McKinsey has no commercial interest in any aspect of this study, and in particular has no commercial interest in the processing of the data or publication of results relating to this project.

St Giles Medical have been contracted to perform qualitative data collection and analysis related to the qualitative research. St Giles medical will also assist with final report writing, but will only view aggregated outputs with small numbers suppressed.

The data requested for this study relates specifically to the following cohorts:

1) Intervention cohorts

2) Pre-treat cohort

3) Control cohorts.

Each of these is described in detail below. All data relates to the deceased only.

1) INTERVENTION COHORTS

The following cohorts all contain deceased people who were affected by an intervention of end of life care. They have been divided in to four cohorts as the most appropriate method of evaluation depends on which type of intervention influenced their end of life care.

• Hospice service users

• Care home residents

• Patients that received hospital-based interventions

• Patients in the hospice catchment area

All cohorts will be minimised to patients who were aged over 18 years of age as well as to episodes no greater than 90 days prior to the date of their death.

Intervention cohort 1- Hospice service users

This cohort includes 7 hospices. The individuals included in this cohort are all those who used the hospice service from its start date (range January 2013-November 2016) to April 2018 and died prior to April 2018. Hospices will only upload identifiers of those individuals who are deceased.

This cohort includes intervention types: hospice at home, community services for non-cancer patients and nurse-led beds for less complex patients. These services are in this cohort because the individuals in receipt of these interventions can be identified through records kept by the hospices.

To select records associated with the intervention cohort the 7 hospices in the evaluation will collect a set of patient identifiable information (the NHS number, sex, date of birth and home postcode of each patient) indicating the complete cohort of patients that have been involved in the studied care intervention, over the past 6 years. The individuals included in this cohort are all those who used the hospice service from its start date (range January 2013-November 2016) to April 2018 and died prior to April 2018. Hospices will only upload identifiers of those individuals who are deceased.

Each hospice will securely upload the above identifiers to NHS Digital of any individuals referred to their intervention from the point of service implementation to NHS Digital. NHS Digital will associate these data with the appropriate APC HES and mortality data and flag these patients as belonging to a specific hospice in the pseudonymised dataset provided to McKinsey & Company, Inc. United Kingdom, for further data processing. This will enable McKinsey to identify which Hospice the patient is associated with.

Intervention cohort 2- care home residents

This cohort relates to 5 hospices and includes the 80 care homes that they have worked with. The individuals in this cohort are residents living at the care homes since hospice service start date (range April 2015-July 2016), and who died before April 2018.

This cohort includes the hospice interventions that have involved working with care homes (termed care/nursing home education in this study).

To select the individuals in this cohort, Hospice UK will provide the names and addresses of the care homes. NHS Digital will associate the relevant APC HES and mortality data based on recorded addresses of the care home residents.

Intervention cohort 3- patients that received hospital-based interventions

In total this cohort relates to 5 hospices and involves 8 hospitals (some of the hospices work with more than one local hospital). The individuals in this cohort are those who died prior to April 2018 and interacted with one of these 8 hospitals in their last 90 days of life, after the hospice implemented their service (range January 2013-October 2016).

This cohort includes the hospice interventions that involved working with a local hospital. These are: hospital-based discharge service, hospital-based discharge service delivering social care and nurse-led beds for less complex patients.

Hospice UK will provide NHS Digital with the name, address and NHS trust code of the hospital. They will use this information to extract APC HES and mortality data of people who have interacted with this hospital in the 90 days prior to death.

It should be noted for one of the interventions in this cohort , nurse-led beds for less complex patients, involves two hospices and is being measured through both intervention cohort 3 and intervention cohort 1. This is because, due to hospice records, it is possible to measure at the level of identifier-upload by the hospice (cohort 1). However, in terms of the wider system, the impact is likely to be seen within the hospital population as less-complex patients will now be transferred to these beds instead of perhaps remaining in hospital. It is for this reason that these two interventions will be measured through two cohorts. Triangulating the data between the two cohorts also presents an opportunity to critically assess whether variation in data is due to the method of evaluation.

Intervention cohort 4- patients in the hospice catchment area

This cohort involves data from 9 hospices relating to decedents who resided in the hospice catchment from date of service implementation (range April 2011-December 2016) until April 2018 and died prior to April 2018.

This cohort includes these intervention types: single point of access (SPA), 24/7 helpline for service users, ambulance staff education, palliative care outreach to rural areas and hospice inpatient capacity expansion. These intervention types are included in this cohort because they impact the entire of the hospice catchment area and there is no way of knowing which individuals benefitted from these interventions. Hospices do not keep records on who rings an advice line for example, however there is likely to be a cumulative impact on hospital usage at the population level.

Hospice UK will provide NHS Digital with postcodes at the sector level (e.g. SW9 6) that represent the catchment of the hospice. NHS Digital will select individuals from within this catchment based on recorded home addresses.

2) PRE-TREAT COHORT

To obtain baseline data, Hospice UK are looking at the relevant patient for all 24 hospices prior to intervention onset. Each intervention cohort will have its own pre-treat cohort.

• Pre-treat cohort 1 (hospice service users, 7 hospices), will include decedents who resided in the hospice catchment from April 2009 until April 2018 and died prior to April 2018. (For the hospice service users only, having data from the full time period is valuable: patients who were in the catchment area, at the same time as the intervention, but not receiving the intervention, are an important ‘pre-treat’ control group.)

• Pre-treat cohort 2 (care home residents), will include residents living at the care homes until service start date (range April 2015-July 2016), from April 2009, who died prior to April 2018.

• Pre-treat cohort 3 (patients that received hospital-based interventions, 8 hospitals) will include individuals who died after April 2009 and prior to April 2018 and interacted with one of these 8 hospitals in their last 90 days of life, before the hospice implemented their service (range January 2013-October 2016).

• Pre-treat cohort 4 (patients in the hospice catchment area, 9 hospices) will include decedents who resided in the hospice catchment from April 2009 until the service start date, and died prior to April 2018.

3) CONTROL COHORTS

For the control cohort 11 control cohorts will be created and their size will range from 1,000 to 15,000, with an average size of 9275.

There are three cohorts; appropriate to the method of evaluation for the intervention cohorts. These three cohorts allow Hospice UK to control for any general changes in palliative care that may have happened during the period being looked at in the study: April 2010-April 2018.

For the controls hospices Hospice UK do need something from them to include in our analysis and the postcode lists are least sensitive as they are not patient-identifiable (sector level rather than full postcode) and they will be looking at an aggregated locality level.

Control cohort 1- the control for intervention cohorts 1 and 4 (Hospice service users and Patients in the hospice catchment area)

This control cohort is for both intervention cohorts 1 and 4 and follows the same method described for intervention cohort 4.

This cohort involves data from 11 hospices relating to decedents who resided in the hospices catchments in April 2010-April 2018 and died in the same period.

This group involves data from 11 hospices relating to decedents who resided in the hospices catchments in April 2010-April 2018 and died in the same period. These hospices were selected through two routes. 1. In response to an email sent about service provision stability over the study period to a potential longlist of 66 hospices suitable to be controls as identified by Hospice UK leadership and 2. Hospices that had applied to be in the intervention arm, but were excluded due to their intervention start date. These two methods resulted in 14 control hospices and of those Hospice UK received data from 11.

Control cohort 2- the control for intervention cohort 2 (Care home residents)

This cohort is the control for intervention cohort 2 and relates to the same 5 hospices. In addition to supplying details of the homes that they had worked with (intervention cohort 2), they also provided names of care homes in their locality with which they had not worked. This control cohort includes those who resided at these 25 control care homes between April 2010-April 2018, and who died prior to April 2018.

To select the individuals in this cohort, Hospice UK will provide the names and addresses of the care homes to NHS Digital and NHS Digital will associate the relevant HES and mortality data based on recorded addresses.

Control cohort 3- the control for intervention cohort 3 (Patients that received hospital-based interventions )

The 8 hospitals in intervention cohort 3 were matched by deprivation score and size (number of beds used as a proxy) to 5 hospitals appropriate to be controls.

Hospice UK will provide NHS Digital with the name, address and NHS trust code of the hospital. NHS Digital will use this information to extract HES and mortality data of individuals who interacted with this hospital in their 90 days of life.

The individuals in this cohort are those who died prior to April 2018 and interacted with one of these 5 hospitals in their last 90 days of life, between April 2010-April 2018.

For each patient cohort, NHS Digital will provide a pseudonymised extract from the HES APC linked with mortality data, labelled by cohort. No direct patient identifiers will flow from NHS Digital.

No direct patient identifiers will be revealed outside the organizational boundaries of the care providers or of NHS Digital. The data flow between care providers and NHS Digital is covered by section 251 support.

Each record will be given a unique pseudonymised ID. To enable association of each record with a hospice and intervention, the year of death, month of death, and the first half of the patient’s postcode (or other indicator of which hospice and intervention the patient links to), will also be recorded.

OUTCOME VARIABLES

The principal outcome measure for this study is the reduction in the number of hospital bed days for an average patient involved in an intervention. This is measured as the reduction in the number of nights spent in hospital during the last 90 days of life.

Additionally, secondary outcome measures are being requested form NHS Digital. These are: service of place of death (hospital / hospice / home), number of A&E admissions in last 90 days of life, number of emergency admissions in last 90 days of life, number of inpatient admissions in last 90 days of life, number of outpatient admissions in last 90 days of life, number of transfers to palliative care and social care services in the last 90 days of life

ADDITIONAL EXPLANATORY VARIABLES

Each record will include the following control variables: ICD-10 code for diagnosis at death, ICD-10 code for additional diagnoses (including cancer), age, sex, ethnicity, Index of Multiple Deprivation to measure socio-economic deprivation at the home postcode, an urbanisation indicator based on home postcode, distance from home postcode to nearest hospital, distance from home postcode to nearest hospice.

It is the intention that each of these variables should be included as patient demographics within the model.

USAGE OF THE DATA SUPPLIED BY NHS DIGITAL

As data processors, McKinsey will test the effects of the various care interventions on the patient cohorts in the final 90 days of life.

The quantitative analysis at McKinsey will consist of a separate quasi-experimental longitudinal study for each end of life care initiative being evaluated. These analyses will test the differences in hospital utilization in the last 90 days of life for similar cohorts of patients, both before and after the introduction of each initiative. These differences will be compared to control cohorts of patients, sampled over similar time points, that did not have access to the new care initiatives. This is, in effect, a “difference of differences” analysis. Other outcome measures will include numbers of A&E admissions, together with the place of death (such as hospital, home or hospice). No direct flow of identifiers will flow from NHS Digital.

The results of the analysis will be a statistical comparison of the impact of each care initiative on the hospice utilisation of patients involved in these care programmes, controlling for site-specific differences and for other trends in hospice utilisation over time.

The assessment of hospital utilisation will be made using pseudonymised patient secondary care records, drawn from the HES and mortality databases. These linked records will provide a detailed picture of patients’ interactions with the hospital system in the last 90 days of their lives and are matched to dates and places of death.

Only results constructed from aggregate data with small numbers suppressed in line with the HES Analysis Guide will be shared with Hospice UK or St. Giles Medical for the purposes of writing the final report. For example, these outputs might effect size estimates with standard errors; or mean and variance for the number of bed days patients spend in hospital in the last 90 days of life, by hospice location and year, after removing the variability attributable to patient demographics.)

OTHER DATA FLOW CONSIDERATIONS

No data will be shared with any 3rd party organisations other than outputs that are aggregated with small numbers suppressed in line with the HES Analysis Guide. The requested data cover a sensible time-period for testing recent hospice-led care innovations: 2009-2018, with most records relating to patients who died in the last 3-4 years.

Hospice UK is an umbrella organisation that represents hospices in the UK. While 7 Hospices are sending in the identifiers to NHS Digital to facilitate the linkage of datasets, no identifiable or record-level data associated with this application will flow to or from Hospice UK.

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

By signing the Data Sharing Agreement, the Data Controller confirms that the Data Processors listed within this agreement have each:

- Confirmed that they understand their roles and responsibilities on behalf of the Data Controller as defined within the Data Sharing Agreement.

- Confirmed that the Processing Activities described within the Data Sharing Agreement are accurate and achievable in terms of the particular Data Processors’ processing.

Outputs from the analysis

The output from the analysis of the HES-ONS data will contain only aggregate data with small numbers suppressed in line with HES analysis guidance.

The use of these data will be solely to produce the outputs for NHS England and Hospice UK’s service evaluation, and not for any other purpose, organisation, company or project.

Outputs from NHS England’s and Hospice UK’s service evaluation

In disseminating the findings, NHS England and Hospice Uk will seek input and guidance from key stakeholders, including hospice services, commissioners, and patients and families.

Alongside a final report to NHS England (by November 2019) and publication of the quantitative evaluation as an open-access article through submission to a peer-reviewed academic journal (by November 2019), Hospice UK will also be developing ‘how to’ guides for hospices and for commissioners (written and presented by November 2019) to help them put the findings in to practice. These guides will draw out the practical steps that those key stakeholders can take based upon the study conclusions, and will be developed in partnership with representatives of those key audiences. The guides will be supplemented by workshops and webinars, and will sit alongside other resources developed by Hospice UK to help support service improvement based on population need, including our Population Needs Assessment Tool (PopNAT). The project team will also be exploring the use of Project ECHO methodology to establish communities to practice to implement the findings.

Hospice UK will also discuss the findings with our People in Partnership group, which is made up of people with lived experience of terminal illness. They will ask the group to advise us on how best to communicate the findings of the study to patients and families, for example via resources that help describe the role that hospice care can play in supporting people and their families.

The overarching objectives of the proposed analysis of end-of-life care innovations is to improve care for patients at the end of life, in terms of the quality and cost of care, as well as the alignment with patients’ preferences over care location. Analysis of the linked HES-ONS data will allow Hospice UK and NHS England to test the ability of recent care interventions to reduce the number of bed days spent in hospital by patients at the end of life, to determine how to scale up the best approaches, and to develop evidence-based end of life pathways for commissioners.

The expected benefits from the recommended care innovations to patients include:

- Reduced lengths of stay for end-of-life patients in hospital

- A shift of up to 30,000 patients to end of life care in other settings than hospital: supported at home, in care homes, or in hospices. (This represents a movement of 10% of the patients who currently end their lives in hospital.)

- Improved patient and caregiver/family satisfaction, as more end-of-life care is provided in situations that align with patient preferences (only 5% of surveyed patients that expressed a preference over place of care at the end of life said that they preferred to die in hospital).

The expected benefits to commissioners and NHS England include:

- Clear evidence-based guidance on the number of hospital bed days saved by the surveyed interventions, and the potential for the number of saved bed days nationally if the best interventions are adopted

- An understanding, based on this evidence, which end-of-life care innovations reduce hospital costs, align with patient preferences, and are scalable to regional and national levels

- Guidance on how best to commission and sustain the recommended models

- With this evidence, understanding and guidance, commissioners will change end of life care in the UK. This could realise savings of around £500 per patient in the last days of life by following alternatives to hospital palliative care, and provide more care that satisfies the NHS Patient Experience Framework’s respect for patient preferences

The expected benefits to hospices and care homes include:

- Evidence-based recommendations on new care innovations that increase the usage of alternatives to hospital services at the end of life

- Detailed guidance on how to interact with local commissioners and other funding sources, and on how to operate the recommended models at scale

- As a result of this evidence and guidance, hospices and care homes will be enabled to implement the most effective care innovations over the next 1-4 years.

Not stated in the register.