End of life data project

The purpose of this project is to:

- provide a whole system overview of current end of life care and need within Wolverhampton;

- understand how and when people are identified as being at the end of their life and who dies an “expected” death without having been identified for end of life care;

- understand what services people use at the end of their lives and how much is spent on these;

- discover where there are opportunities to redesign services to better meet the needs of Wolverhampton residents, ensure a good patient experience with financial sustainability, and reduce health inequalities .

People are often in touch with a range of services throughout the last year of their lives, including residential or nursing homes, community services, palliative care team, acute hospital admission or A&E attendance, and their GP. Currently information about service usage is held by each service area in disparate datasets, so there is limited understanding about the flows of people through the system and how resources could be redistributed to improve patient outcomes and experience.

There is some evidence from published research that quality of experience and costs can be improved by the use of community provision and proactive care planning, but a limited understanding of how and where improvements can be made locally and what level of resource is currently spent on people at the end of their lives.

Currently each organisation providing care at the end of life only has a piece of the data puzzle, and this project would bring all that data together in one place i.e the Local Authority, so that analysis can be focussed around the patient, rather than the service. The NHS Digital Primary Care Mortality Data on deaths registrations is key to this project, because it provides a “gold standard” record of all deaths, that other datasets (in this case GP data, Social Care, Hospital, Palliative Care and Community data) can then be linked back to, by using the NHS number . This is vital to ensuring that information about all deaths and end of life experiences are captured, not just those of people who were proactively identified as end of life, or those who died in hospital.

This information has been requested to facilitate the work of the Local Authority operating, on this issue, as part of an Integrated Care Alliance (ICA); a partnership forum established in 2017 between health and social care organisations in Wolverhampton, including GPs, Clinical Commissioning Group, Council, NHS Trust, Compton Care (palliative care provider) and others. The end of life workstream is just one of a number of priority areas that are being developed under the ICA, and it is intended to use this project as an exemplar for how data could be used in the future to promote Population Health Management across all workstreams; focusing on delivering outcomes for population groups, rather than activity for specific conditions.

The ICA works together on a number of workstreams. Organisations within this partnership may contribute to suggestions on which questions of interest would be beneficial to explore. However, the use of the data for the purposes outlined in the agreement is solely decided upon by the Local Authority under their role as Data Controller. Additionally, the Local Authority Public Health team has a legal basis for conducting needs assessments on a population level.

The proposal is to take one year of data on deaths for Wolverhampton residents (those that occurred in 2017 from the

PCMD data) and link this to data from key services within the City; GP record, in-patient, out-patient and A&E hospital activity data, NHS community and specialist palliative care services, social care client data, and Compton Care palliative care services.

The DSCRO will create a pseudonymisation key using Open Pseudonymiser (www.openpseudonymiser.org) which will be unique to this project. This key will be shared with each agency providing data, who will apply the key to the NHS number in a pre-defined dataset. Only those fields which are necessary for this analysis will be requested, rather than complete datasets. Any patient identifiable data (date of birth, address, name) will be removed or replaced with non-identifiable proxy, for example date of birth would be replaced with age band. These steps will ensure that patient confidentiality is maintained.

This pseudonymised data from the GPs, the Trust, the Local Authority and Compton Care will then be shared via secure email (nhs.net or GSX) with the CSU. The DSCRO will provide pseudonymised SUS and PCMD datasets to the CSU as the processor, to conduct the record linkage using pseudo-NHS number, so that only those records which relate to deaths in the specified time period would be in the linked dataset, thereby ensuring that the data analysed is the minimum amount necessary to achieve the objectives.

Once the data has been linked using the pseudo-ID number, the data on those who died within the specified year linked to the other datasets will be further pseudonymised by the CSU by removing the pseudo-NHS number and replacing with a new unique identifier before leaving the CSU, thus avoiding the potential for reidentification of any individual by Public Health analysts.

The linked dataset will be shared by the CSU via secure email with specified analysts who are employees in Public Health at the City of Wolverhampton Council, who will then produce the specified analysis at an aggregate level, in line with the HES Analysis Guide . The aggregate outputs of the analysis will be shared with all included partner organisations in the Integrated Care Alliance, but the dataset will remain securely at the Council. None of the other organisations providing data will receive the linked dataset and therefore will not be able to re-identify any individual.

No attempt will be made by Public Health to perform any further matching of the dataset or to re-identify any individuals. Access to patient/client level identifiable Adult Social Care data is on a role-based access basis within City of Wolverhampton Council. Analysts within the Public Health department (who are not part of the wider Council business intelligence team) do not have the ability to view or access these records.

Pseudonymisation of social care records would be performed by a social care analyst who has role-based access permissions. Likewise Public Health datasets are stored in a secure area of a separate server, which no one beyond the identified Public Health analysts has access to, and therefore there is no potential for the social care records and the final linked dataset to be viewed concurrently or by the same individual, thus avoiding any possibility of re-identification of an individual from the data.

A local data processing and sharing agreement for the project covering all providers supplying data, the Local Authority as the Data Controller and the CSU as the

processor has been written in accordance of IG rules for each organisation, with input from specialist IG staff across these

organisations.

No patient level data will be linked other than as specifically detailed within this agreement and data will only be used to carry out purposes specified in the Data Sharing Agreement. Data will only be shared with those parties listed and will only be used for the purposes laid out in the application/agreement. The data to be released from NHS Digital will not be national data, but only that data relating to the specific locality of interest of the applicant.

The processing activities carried out by NHS Wolverhampton CCG are purely regarding IT support, involving storage, backup and server management of the data. Access to data by NHS Wolverhampton CCG is restricted to IT support staff in fulfilment of the IT support service commissioned by the Local Authority, and the CCG may not use or access the data for any other purpose.

Analysts in Public Health undertake compulsory Information Governance training every year. All staff understand their responsibilities with regard to receiving, storage, processing and handling of data. All access to data is managed under Roles-Based Access Controls.

Data will be used to produce reports, charts and dashboards to analyse the data for the purposes listed. Any outputs shared outside the Council will be at aggregate level with small number suppression.

Outputs will take the form of reports and presentations, which will be made available to all organisations in the Integrated Care Alliance, as described above. Furthermore, a summary of key findings will also be disseminated to the wider network of organisations working on Integrated Care across the country as well as other key regional and national networks such as the West Midlands Population Health Management Network and NHS England’s Population Health Analytics Network.

The specific lines of enquiry are detailed in the table below and will be published approximately two to six months following DARS application approval.

1.Describe "expected deaths" (ONS case definition*) in terms of demographics

- By combining demographics from all datasets, no duplicates, death registration as gold standard

2.Describe palliative care identification; when does this occur, are those identified different (demographics) from those who have an expected death without identification

- By combining PCMD, primary care and demographics

3.Describe non-elective activity at end of life for expected deaths; when (including time of day, day of week, in relation to date of death), why (presenting condition/symptom), how (A&E, ambulance) are patients admitted, length of stay and discharge destination, level of need as described by SPC for those in receipt

- By using expected deaths subset from PCMD, SUS data description, SPC

4.Does the above vary depending on whether patient has palliative care flag, admission avoidance flag, advanced care plan recorded before admission

- By combining primary care and SUS

5.Describe any "unexpected deaths" for people on an end of life register e.g. accident, suicide

- Those that meet exclusion criteria as per ONS definition but have palliative care flag in EMIS

6.Describe costs associated with palliative care for those identified and not identified

- By examining costs from in-patient, A&E, outpatient, community (NB GP costs not possible)

7.Where preferred place of death is captured, describe place of death and matches

- By combining Death registration and primary care

8.Equity audit of specialist palliative care services

- By examining demographics of those in receipt of Compton or in patient palliative care services, compared with expected deaths cohort, does this vary by GP practice

9.Describe timelines of health and social care usage - identify services received in parallel

- Plotting dates in relation to death for contacts in all datasets

The outputs will form the Local Authority’s input into the ICA discussions. The Local Authority will define the use of the data, but the outputs will then be shared with ICA who may act upon those outputs or consider them alongside other evidence in their decision-making processes. These evidence-based decisions about integrated care provision and resource distribution via the ICA programme will also provide a baseline upon which the success of any service development can be evaluated .

The three overarching aims of the ICA are to improve patient experience, to improve population level outcomes, and to make the health and social care system financially sustainable. For end of life, the intention is to increase the number of patients experiencing a good quality death, to increase the number of people dying in their home or preferred place of death, and to achieve value for tax payers’ money by redistributing resource to where it achieves the biggest impact, for example by providing additional community services where these would result in a comfortable death in the home, rather than an expensive emergency admission.

The additional intended benefit is to use this project as an exemplar project to demonstrate the value of using a Population Health Management approach to service redesign – this would involve sharing a summary of the

methodology and headline findings from the data analysis with the wider network of organisations working on

Integrated Care across the country. Learning and headline findings will be disseminated across the regional West Midlands Population Health Management network, led by Public Health England and NHS England, as well as the national NHS England Population Health Analytics network, in the hope of increasing the understanding of what is possible in terms of information governance and population level analytics using linked datasets.

Yielded Benefits is not a requirement for new applications.