Impact of closing Emergency Departments in England

This agreement is for the retention of data following the completion of the study.

The University of Sheffield was awarded funding by the National Institute for Health Research (NIHR) to identify the impact of closing Emergency Departments (ED) in England. The mission statement of NIHR is to “maintain a health research system in which the NHS supports outstanding individuals, working in world-class facilities, conducting leading-edge research focused on the needs of patients and public”.

The study was funded following a researcher led call from the Health Services and Delivery Research (HS&DR) Programme. Prior to the award being given, this study was reviewed by a HS&DR funding panel, external reviewers, and the HS&DR commissioning board.

In recent years a small number of EDs in England have closed for all or part of the day, usually because it has not been viable to provide senior cover 24/7. This is set against a period of increasing demand for emergency care in terms of rising ED attendances and NHS ambulance calls. This research will therefore focus on a key area of current health service re-organisation in the UK, addressing a highly topical and important question. There is considerable controversy in relation to the reconfiguration, and closure of EDs. Closing EDs is not viewed positively by the public and patients, as demonstrated by campaign groups which have formed to prevent these closures. However, closures may not have the negative impact on patient outcomes that campaign groups suggest. Evidence on the effects of ED closures is essential to inform public debate and policy given that further closures are planned.

In recent years a number of Emergency Departments (EDs) have closed, or been replaced by another facility such as an Urgent Care Centre. Currently, there is little research evidence to inform decision making about these closures. This study looked to identify if local populations and emergency care providers are affected by such closures, focusing on five EDs which closed between 2009 and 2011.

The University of Sheffield was awarded funding to identify the impact of closing Emergency Departments (ED) in England. Five EDs which closed (or were downgraded) between 2009 and 2011 are the focus of the analysis. The purposes of the analysis was the following:

- To identify any changes in the pattern of mortality in the local population following closures

- To identify any changes in local emergency care service activity, and performance following closures

The University of Sheffield requested pseudonymised data sets with the exception of Civil Registration Data. Findings have been presented in research reports, at academic conferences, and peer reviewed journal publication is also planned. All dissemination activity will report aggregated results across patient episodes and not individual patient episodes. In any reports, small numbers will be suppressed in line with the HES Analysis Guide.

All reasonable steps in processing were taken to avoid re-identification of deceased patients from within the HES data. The full Date of Death was used to derive vital status at 7, and month of death only. There is no other requirement for Date of Death apart from the derived vital status output listed. For the purpose of producing these derivations only Date of Admission from the HES data will be required so therefore there is no requirement for Date of Death to be linked with the HES data. Date of Death (full) will not be linked to the HES data and the two will not be stored in a database as linked data. Date of Admission (HES) will be the only field linked to Civil Registration.

All organisations party to this agreement must comply with the Data Sharing Framework Contract, including requirements on 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 received by University of Sheffield will not be used for any purpose other than to meet objectives as stated in this Data Sharing Agreement and will not be shared with any other third party or organisation.

The Data Controller should ensure appropriate data processing agreements with all data processors contracted to undertaking work referenced within this agreement.

This data is processed in accordance with General Data Protection Regulation Article 6 (1) (e) and General Data Protection Regulation Article 9 (2) (j) as it is in the interest of the public to see if the closure of emergency departments has an adverse effect on care.

To enable analysis, five control EDs were selected.

There were a number of stages involved in creating the outputs:

Stage 1: calculation of catchment populations

In the first instance, a resident catchment population was identified for each target ED (both intervention and control). Catchment populations were identified from Hospital Episode Statistics (HES) Accident & Emergency (A&E) attendance/ and (or) Ambulance Service data and were be defined as the lower super output areas of residence from which the majority of attenders at any ED during the study period, used the target ED.

Stage 2: Calculation of indicators

The next stage of processing was the use of routine data, to produce a series of population and emergency care indicators for each resident catchment population. In total there were 17 indicators relating to ‘death’ and ‘risk of death’, ‘A&E attendances’, ‘emergency hospital admissions’, ‘condition severity’, and ‘ambulance service performance’. Data was drawn from Civil Registration, HES A&E attendance, HES Admitted Patient Care, and NHS Ambulance Service Computer Aided Despatch datasets. Data from NHS Ambulance Services was drawn directly from each NHS Ambulance Service Trust and was not linked to any other dataset.

The information below describes each indicator in detail (all numbers and proportions, etc. relate to residents of the catchment areas) and identifies the data source required.

Data source - Civil Registration data and HES-Civil Registration linked mortality data

• The number of deaths from conditions (as identified by ICD code) identified as rich in avoidable deaths

• Case fatality ratio for conditions (as identified by ICD code) identified as rich in avoidable deaths

For patients in the University of Sheffield’s HES APC data who also hold a Civil Registration deceased flag, the Civil Registration Date of Death (along with encrypted HESID) was linked to Date of Admission (HES Admitted Patient Care) to derive vital status after 7 days, and Month and year if the calculation exceeds 30 days, by calculating the days lapsed from Date of Admission to Date of Death. For those patients that hold a Civil Registration status but do not appear in the HES APC dataset, only month and year of death was derived.

The Civil Registration database (with full Date of Death) was stored separately and unlinked and a updated Civil Registration database with the derived vital status details (and for removal of doubt, does not contain full Date of Death) was linked to the HES data.

Data source - HES Accident & Emergency attendance

• Total ED attendances

• Total ED attendances by mode of arrival (i.e. by patients brought in by ambulance and those identified as having an ‘other’ mode of arrival)

• The number of arrivals at ED discharged without treatment or investigations(s) that required hospital facilities.

• The proportion of attenders at ED who are admitted to an inpatient bed

Data source - HES Admitted Patient Care

• The number of emergency hospital admissions for any condition

• The number of emergency hospital admissions for conditions (as identified by ICD code) identified as rich in ‘avoidable admissions’

• Mean length of stay in hospital for those admitted as emergencies

• The numbers and proportions admitted to critical care medicine

Data source - Ambulance service Computer Aided Despatch data (routine data extracted directly from NHS Ambulance Service Trusts)

• Mean time from 999 call to ambulance on scene,

• Mean time from ambulance arriving on scene to ambulance arriving at hospital

• Mean time to hospital from time of 999 call

• Mean time from ambulance arriving at hospital to ambulance ‘clear’ time

• Total ambulance service call volumes

• Non conveyance rates

• The number of emergency hospital transfers between local hospitals

Stage 3: Primary analysis

For all the indicators, data was analysed using a time series of monthly values (for a minimum of 48 months spanning the closure or downgrading of the ED). A simple time series was fitted to the data including a linear time trend, a seasonal effect, step interruptions for any other major changes to the local emergency care system, and a step interruption at the time of the change to the ED. Control series was also used. The control catchment areas were populations in similar areas not expected to be affected by the closures either directly (as a result of ED attendances being diverted) or indirectly (via any impact on the ambulance service). Analyses estimated the step in the intervention series, and the difference in the step between the intervention series and the control series. The University of Sheffield used the estimate of the size of any step to estimate the impact on the indicators following the closure of the ED.

Data processing

The data was processed and analysed within a single department, the School of Health and Related Research (ScHARR) within the University of Sheffield. The data will not, in any circumstances, be released to a third party. The data will be accessed by a restricted number of authorised individuals within the study team by use of a networked project folder which will be password protected.

Access to the data requires authentication (username and password) on the university network; further and distinct authentication (distinct username and password belonging to specific user accounts and from specific on-site machines - sited in locked rooms - only) on a specific "virtual machine". This virtual machine is the only "entity" from which intelligible access to the data is possible.

Specific user accounts allowing access to the data requested on the ‘virtual machine' are only granted to members of the study team, who are employees of University of Sheffield.

Authorisation to the specific project system folder will be given to a strictly limited number of individuals. All these individuals work within the same department (ScHARR) at the University of Sheffield and are subject to confidentiality policies of the University of Sheffield. The data was cleaned and analysed by the data manager and statisticians who have experience in dealing with large datasets of routine data. The dataset was then analysed by the statisticians in line with the aims of the study detailed in the study protocol that was approved by the University of Sheffield Ethics.

The calculated data was reported to the team members in the form of a report detailing the aggregated results in line with the HES analysis guide.

The study took 24 months to complete (study end January 2017). The data extracts from the datasets will be removed from the network folder and destroyed five years after the study end date to allow the University of Sheffield time to complete the study dissemination period, and respond to any data queries prompted following dissemination. This activity is supported by the funder of this research, NIHR.

The data will be used to produce outputs for academic conferences and journal articles in peer reviewed health-related journals. The University of Sheffield will seek to publish ‘open access’ articles, therefore allowing their findings to be freely available. The study team have drafted a manuscript, looking at the impact on closures on the emergency and urgent care system, for submission to an academic journal (submission planned Winter 2019).

To date, one journal article has been published (Knowles E, Shephard N, Stone T, et al The impact of closing emergency departments on mortality in emergencies: an observational study Emergency Medicine Journal 2019;36:645-651).

A research report, as required by the funding body (NIHR), has been published and is now publicly available: https://www.journalslibrary.nihr.ac.uk/hsdr/hsdr06270/#/abstract

The research team shared the final report with key stakeholders, both national organisations (NHS England, Association of Ambulance Chief Executives, College of Paramedics, Royal College of Emergency Medicine, and Healthwatch) and local health communities (ie in areas affected by ED closures) to facilitate the use of evidence to inform future decision making.

The University of Sheffield have presented findings from this study to a number of conferences: HSRUK, 999 EMS Forum, Royal College of Emergency Medicine annual meeting. These conferences were attended by researchers, healthcare providers/commissioners, and patient and public groups. No further conference dissemination is planned.

All outputs will only contain results in aggregated format and as statistical summaries. It is likely that for each of the 17 indicators, summary tables/graphs will be produced in the intended outputs.

The evidence base regarding ED closures was lacking. The output (and therefore benefit) of this study has been to produce robust information about a series of indicators which can measure the effect of an ED closure on a local population, and emergency care providers, and therefore has the potential to inform future decisions about which EDs are selected for closure. These indicators may be transferable to other evaluations concerning the emergency and urgent care system, and may be particularly transferable to any evaluation of the re-defining of A&E care, proposed as part of the current review of emergency and urgent care services.

The expected impact of this study is to inform the re-organisation of emergency care in England by providing the general public, the NHS, and policymakers with the evidence to enable them to make informed decisions.

As such, there is no specific target date when measurable benefits will be apparent: benefits from the study will be applicable in relation to any proposed closure following the publication of the findings. However, given that further closures/downgrades are currently planned or under consideration it is likely that University of Sheffield's findings will be considered in the context of closures immediately after publication of the results.

In summary, the expected benefits are the production of a set of accurate, reliable, and credible indicators which can measure the effect of an ED closure (‘what’), and can be used by local populations and the healthcare community (‘whom’) when faced with decision making about potential ED closures (‘when’).

University of Sheffield have produced a set of robust indicators to measure the impact of Emergency Department closures on local populations and emergency care providers. Following the calculation of these indicators, using HES data, University of Sheffield have produced evidence which will help to inform any future re-organisation of emergency care.

Any re-organisation of healthcare will impact on the wider local health community, and it is important that findings are shared to enable commissioners and providers to build the findings into any future re-organisation, if appropriate.