Liaison Psychiatry: Measurement and Evaluation of Service Types, Referral Patterns and Outcomes (Workstream 2 - Phase 1)

The overall aim of LP-MAESTRO is to evaluate the cost-effectiveness and efficiency of particular configurations of liaison psychiatry services for specified target populations. To do this, an innovative approach based upon linking routinely collected patient-level data and using economic modelling with the resulting aggregated data will be developed and evaluated.

Workstream 2 – Phase 1 (WS2P1) focuses on 11 acute hospitals in England that were identified within Workstream 1 of LP-MAESTRO and the Liaison Psychiatry Survey England 2015 as not having a Liaison Psychiatry service.

Of those 11 hospitals, 8 gave approval (see approval letters associated with this application) for their data to be included within this study as follows:

-- St Mary’s Hospital (Isle of Wight NHS Trust)

-- James Paget University Hospital (James Paget University Hospitals NHS Foundation Trust)

-- Diana, Princess of Wales Hospital (North Lincolnshire and Goole NHS Foundation Trust)

-- Scunthorpe General Hospital (North Lincolnshire and Goole NHS Foundation Trust)

-- Rotherham Hospital (Rotherham NHS Foundation Trust)

-- Pilgrim Hospital (United Lincolnshire Hospitals NHS Trust)

-- Lincoln County Hospital (United Lincolnshire Hospitals NHS Trust)

-- County Hospital (University Hospitals of North Midlands NHS Trust) - previously Stafford Hospital (Mid Staffordshire NHS Foundation Trust)

The NHS trusts for the following hospitals did not provide approval within the required timescale and are therefore not included within this application:

-- Basingstoke and North Hampshire Hospital (Hampshire Hospitals NHS Foundation Trust)

-- Royal Hampshire County Hospital (Hampshire Hospitals NHS Foundation Trust)

-- Luton and Dunstable Hospital (Luton and Dunstable University Hospitals NHS Foundation Trust)

For patients with an A&E attendance or Inpatient episode (Index Episode) at one of the 8 included hospitals within the defined index period (1st April 2013 – 31st March 2014), A&E attendances, Inpatient admissions and Outpatient appointments will be obtained from Hospital Episode Statistics (HES) for the period from 1st April 2012 - 31st March 2015 (i.e. 1 year prior to the index period, the index period itself and 1 year following

the index period).

For patients included within the HES data, the study will determine whether primary care data exists for the patients within TPP’s ResearchOne database. If so, a defined set of primary care data will be obtained for these patients for the period prior to 1st April 2014. Additionally, a mapping file will be produced by NHS Digital and supplied to the University of Leeds to enable the HES data for a specific patient to be linked to the

corresponding primary care data (if present).

From the linked data, pathways for patients attending the 8 hospitals without a Liaison Psychiatry service will be constructed.

Workstream 2 – Phase 2 (WSP21) focuses on hospitals in England that were identified within Workstream 1 of LP-MAESTRO and the Liaison Psychiatry Survey England 2015 as having a particular configuration of Liaison Psychiatry service. To enable pathways to be constructed for patients attending these hospitals that consider the interaction of patients with the Liaison Psychiatry services in these hospitals, data is additionally

required from the NHS trusts that run these services. A variation to the linkage methodology proposed for WS2P1 would be required to support the use of these additional data sources. Accordingly, WS2P2 is not covered by this application and will be subject to separate approval and application.

For the avoidance of doubt, this application relates ONLY to WS2P1.

Patient pathways constructed within WS2P1 and WS2P2 will be tracked and the cost of each to the health care sector calculated using national data sources. A whole system perspective will be adopted in order to determine if there is an association between the presence and configuration of liaison services and health care utilisation by patients. Metrics will include emergency admissions, occupied bed days and length of stay.

Each metric will be analysed by age band.

Aggregate figures on A&E and Inpatient admissions (for the named hospitals) on which to estimate sample size are not publicly available. This data is required in order to estimate the numbers that can be expected in the study. General figures on the number of A&E attendances and inpatient admissions in a given year are published by HSCIC at the level of the trusts that operate these hospitals (see http://www.hscic.gov.uk/catalogue/PUB16728/acci-emer-atte-eng-2013-14-pla.xlsx and http://www.hscic.gov.uk/catalogue/PUB16719/hosp-epis-stat-admi-prov-leve-2013-14-tab.xlsx). These figures provide an indicative upper bound on the number of A&E attendances and inpatient admissions that can be expected at any of the hospitals operated by that trust in a given year.

For 2013-2014, these figures were the following:

- Isle of Wight NHS Trust: A&E Attendances - 59,494, Inpatient Admissions - 28,263

- James Paget University Hospitals NHS Foundation Trust: A&E Attendances - 67,726, Inpatient Admissions - 58,144

- North Lincolnshire and Goole NHS Foundation Trust: A&E Attendances - 137,841, Inpatient Admissions -107,403

- Rotherham NHS Foundation Trust: A&E Attendances - 74,458, Inpatient Admissions - 74,313

- United Lincolnshire Hospitals NHS Trust: A&E Attendances - 144,788, Inpatient Admissions - 146,845

- University Hospitals of North Midlands NHS Trust: A&E Attendances - 119,709, Inpatient Admissions - 157,605

A summary of the data flows, which are performed using either: i) NHS Digital's Secure File Transfer (SEFT) or ii) University of Leeds LICTR's Secure File Transfer (SFT), is provided below:

1. University of Leeds provides TPP (SystmOne) with the cryptographic salt (via SFT)

2. University of Leeds provides NHS Digital with the cryptographic salt (via SFT)

3. NHS Digital provides the University of Leeds with HES A&E, Inpatient and Outpatient data according to specified criteria with patients referenced by HESID (via SEFT)

4. NHS Digital provides TPP (ResearchOne) with a set of patient pseudonyms (via SEFT).

5. TPP (ResearchOne) provides University of Leeds with primary care data for patients identified from the set of patient pseudonyms supplied by NHS Digital with patients referenced by ROID (via SEFT).

6. TPP (ResearchOne) provides NHS Digital with a mapping from each patient pseudonym to a unique ROID (via SEFT)

7. NHS Digital provides University of Leeds with a mapping file that maps a unique HESID to a unique ROID identifier (via SEFT)

The data which will be accessed by the University of Leeds will be pseudonymised they will have no access to identifiers. They are not permitted to use to data provided to try and re-identify any individual.

ResearchOne has been separated out in the data flow diagram and the summary above to TPP (ResearchOne) and TPP (SystmOne). Project documentation refers only to ResearchOne to emphasise that the project will use only data from the ResearchOne database. The Phoenix Partnership (TPP) are responsible for SystmOne, ResearchOne, and for ensuring the necessary separation of duties for staff working on SystmOne and ResearchOne. We leave it to NHSD to liaise with TPP regarding the data flows between the two parties, and how these data flows can be performed via SEFT in a manner that is consistent with the necessary separation of duties. The ResearchOne database is stored and processed inside the data centres which host the SystmOne production environment. It is not possible to disclose the location of these data centres due to governance requirements but they are Tier 3 data centres that have been accredited by the NHS for storing and processing clinical data. There is a disaster recovery site (hosted to exactly the same standards as the live site) which is used as a back-up location in the event of any failures at the main site. Documentation relating to ResearchOne can be found at: http://www.researchone.org/documentation/.

University of Leeds generates and supplies a salt key that is used in the generation of pseudonyms by NHSD and TPP (SystmOne). Further details regarding the pseudonym generation process are provided in the "LPMAESTRO - WS2P1 - Pseudonym Generation Process.pdf" evidence document associated with this application. To ensure that the identifier used to generate the pseudonyms (NHS number) cannot be recovered by the University of Leeds, the University of Leeds does not receive any pseudonyms generated using the supplied salt key. The University of Leeds receives the following: i) HES data referenced by a HES ID, ii) ResearchOne data referenced by a RO ID, and iii) a mapping file from HES ID to RO ID. The linkage methodology has been developed so that the University of Leeds is responsible for the generation and supply of the salt key, and subsequent data flows to the University of Leeds do not contain any data that could be re-identified by the knowledge of the salt key. Additionally, the salt key is used only a one-time basis, such that any pseudonyms generated cannot be matched across data sets obtained for different usages. Approvals for the project (including NHS Ethics) have been obtained on the basis of the proposed linkage methodology.

NHS Digital will identify patients based on the presence of an A&E attendance or Inpatient episode (Index Episode) at one of the 8 included hospitals within the defined index period (1st April 2013 – 31st March 2014), A&E attendances, Inpatient admissions and Outpatient appointments will be obtained from Hospital Episode Statistics (HES) for the period from 1st April 2012 - 31st March 2015 (i.e. 1 year prior to the index period, the index period itself and 1 year following the index period). This data will be provided directly to the University of Leeds with patients uniquely referenced by a unique HESID. Additionally, NHS Digital will generate a pseudonym for each included patient, according to the methodology described above, and supply this list of pseudonyms to TPP (ResearchOne).

TPP will generate pseudonyms for all patients in the ResearchOne database by applying the same methodology as NHS Digital and using the salt key supplied to TPP (SystmOne). TPP (ResearchOne) will then determine those patients whose pseudonyms are present in the list provided by NHS Digital. Selected primary care data will be extracted from ResearchOne for these patients for the period prior to and including

31st March 2015.

Primary care data will only be obtained where:

- the GP practice to which the patient is registered using TPP's SystmOne clinical information system and

- has opted-in to ResearchOne (and the patient has not individually opted-out).

TPP (ResearchOne) will provide primary care data directly to University of Leeds. Each patient will be uniquely referenceable in this data using a unique ROID. No pseudonyms will be included in the data provided by ResearchOne to the University of Leeds. TPP (ResearchOne) will also provide NHS Digital with a file that maps each pseudonym present in the list provided by NHS Digital and in the ResearchOne database to its corresponding ROID. NHS Digital will use the mapping file provided by TPP (ResearchOne) to map the HESID of each patient in the HES data to the ROID identifier of each patient in the ResearchOne data. NHS Digital will supply the resulting mapping file to the University of Leeds who will use it to match the data for a particular patient in HES data to the data for that patient in the RO data (and vice-versa).

Based on the HES data and ResearchOne data, patients with admissions to these hospitals (without an LP service) will be matched to referred and non-referred cases from hospitals with an LP service, either by matching for co-variates or by propensity scoring will be determined and implemented. As described in Section 5a (Objective for Processing), WS2P2 of the LP-MAESTRO project will focus on the linkage and

supply of data for patients attending hospitals with an LP service. Application and approvals for WS2P2 will be undertaken separately. Relevant variables for matching will include demographic (e.g. age, carer support, Index of Multiple Deprivation (IMD) calculated from postcode) clinical (e.g. diagnosis, long-term medication) and health service utilisation (e.g. inpatient days, GP visits, major procedures, A&E/ED visits, and Medical and Surgical outpatient appointments). One of the novelties of this approach is the use of these characteristics identified from HES and from the primary care database to tackle the utilisation in the financial year before referral, as a way of ensuring that outcomes in the financial year after referral are not attributable to easily identifiable pre-existing characteristics (case complexity) that are confounded with likelihood of referral.

Analysis will use the linked dataset and will estimate a standard regression model for estimating the relation between health care utilisation and key variables capturing the configuration of liaison services. The dependent variable in this model is therefore cost of health care utilisation derived from factors such as inpatient days, readmission rates, Emergency Department attendances and General Practitioner visits. The variables that will be used as determining factors relate to the configuration of liaison services, sourced from the data collected in WS1. The quantum of the liaison service provision will be captured by WS1 data related to structure and process e.g. staffing levels and contact time after referral. The standard regression approach rather than frontier-based approaches will be used as the focus is on how the explanatory variables impact

on the dependent variable, secondary care utilisation, rather than on the efficiency of certain providers.

All persons accessing the record level data provided by NHS Digital under this agreement are substantive employees of University of Leeds.

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 ie: employees, agents and contractors of the Data Recipient who may have access to that data).

There will be not data linkage undertaken with NHS Digital data provided under this agreement that is not already noted in the agreement.

Data will only be accessed and processed by substantive employees of University of Leeds and will not be accessed or processed by any other third parties not mentioned in this agreement.

- The study will publish an academic paper (describing the findings in relation to patient outcomes and cost effectiveness) in a reviewed health services research and probably also in a mental health journal.

The audiences will be [a] academics with an interest in liaison psychiatry services [b] academics with an interest in use of routine data to answer health services questions.

These papers will be submitted by the end of the project. Due to delays in data access a project extension request was submitted to NIHR. Intention to approve a project extension has now been received from NIHR. Once formally approved by the Department of Health, the end date of the project will be 30th June 2019. The study will submit to open-access journals to maximise coverage.

- The funders (NIHR) require a final report of main project in the form of a monograph, at the end of the project. The monograph will be published in the NIHR library, which is widely accessed by academics with an interest in applied health research.

- The study will write two discussion papers for health service commissioners by the end of the project – describing the findings in relation to evidence on the effectiveness of liaison psychiatry services, and our findings on the use of routine data to evaluate services.

- A final methodological paper will describe the project as proof of concept for future research projects using linked data from primary and secondary care settings by the end of the project.

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

The results of this application will be used to determine the cost-effectiveness and efficiency of different liaison psychiatry configurations. Based on these results, the study will prepare two types of report to inform the commissioning and provision of general hospital mental health services (expected date: 2019). One report will be aimed at local commissioners, to aid them in making decisions about local service funding. The other will be aimed at national bodies and especially NHS England, where there is an active interest in such services.

The benefits will be measured through changes in the profile of mental health services, in line with report recommendations, as determined by the yearly national survey of such services currently funded by NHS England and Health Education England. These results should influence provision of mental health services in every acute hospital in the NHS. Since a typical mental health service of that sort sees five thousand or so cases a year, the benefits should accrue to many hundreds of thousands of patients.

Local benefits might be seen as soon as a year after publication. National benefits are more likely to be seen over a 2-5 year timescale.

Local benefits might be seen as soon as a year after publication. National benefits are more likely to be seen over a 2-5 year timescale.

Requested HES data was made available by NHS Digital on 16th November 2018. However, the study are currently awaiting additional required data from both NHS Digital and The Phoenix Partnership before analysis can commence.