Bespoke Extract - HES/Civil Registration Mortality Extract

Imperial College London Doctor Foster Unit (ICL DFU) requires Hospital Episode Statistics (HES) Civil Registration data to identify measures of quality and safety in healthcare. This agreement is an extension for ICL DFU to continue holding data, as well as receiving the data approved under DARS-NIC-383203-Q8B9L-v3.4 (the prior version of this agreement).

Imperial College London Dr Foster Unit (ICL DFU) receive Civil Registration (Deaths) data and then flow the civil registration data received from NHS Digital to ICL DFU to be linked to pseudonymised HES Data held under agreement, DARS-NIC-12828-M0K2.

ICL DFU is the sole data controller and processor of all data received under this agreement. The original data or any record-level data will not be shared with any external collaborators or the unit’s funder Dr Foster Limited. Access to the data requires approval from the unit’s director. Dr Foster Limited have no influence or decision-making authority over the use of the mortality data. All data under this agreement is pseudonymised by NHS Digital and remains pseudonymised when processed or used in any way by the unit.

Imperial College are processing the data being accessed under this agreement as part of their public task around research under Article 6(1)(e) and 9(2)(j) of the GDPR. Imperial College are using the data to identify measures of quality and safety in healthcare, this is important to support the health and social care system to reduce variation in outcomes, and to improve the quality and safety of healthcare in England and is therefore research is in the public interest.

ICL DFU requires continued use of mortality data already held and used for its work. Additionally, ICL DFU requires mortality data for 2018/2019 plus 30 days (up to April 30th 2019) in order to capture all deaths in hospital and in the community within 30 days of admission or procedure. This will provide the full 30 days of follow up data for 2018-2019.

Data subjects and cohort groups

Mortality data is required for all patients in HES. This is firstly because casemix-adjustments are run for 259 diagnosis groups and 200 procedure groups as part of the well-established national hospital mortality monitoring system that flagged Mid Staffordshire NHS Trust and other hospitals with problems. Death as an outcome is used within the varied programme of work, which covers a number of medical and surgical specialties. For example, analysis of all emergency admissions showed a higher risk of death in early August, when new junior doctors start work, than in late July (Jen MH, Bottle A, Majeed A, Bell D, Aylin P. Early in-hospital mortality following trainee doctors' first day at work. PLoS One 2009;4(9):e7103).

Minimum data required

The Dr Foster Unit at Imperial College hold data dating back to 2000 for several reasons:

• To examine historical trends of treatment practice (e.g. Faiz et al. Traditional and Laparoscopic Appendectomy in Adults Outcomes in English NHS Hospitals Between 1996 and 2006, ANNALS OF SURGERY 2008;248:800-806) and the historical impact of changes in policy (e.g. Balinskaite V, Bottle A, Shaw LJ, Majeed A, Aylin P. Reorganisation of stroke care and impact on mortality in patients admitted during weekends: a national descriptive study based on administrative data. BMJ Qual Saf 2018;27(8):611-618.).

• To properly identify incident cases e.g. the first hospital admission for new patients with heart failure, where “new” means not admitted to hospital for at least five years (e.g. Bottle A, Goudie R, Bell D, et al. Use of hospital services by age and comorbidity after an index heart failure admission in England: an observational study. BMJ Open 2016;6:e010669; also as yet unpublished work on long-term hospital use in new heart failure patients)

• To increase the power of predictive models for rare diseases, procedures and events (e.g. Dr Foster Unit at Imperial build standard casemix adjustment models for 259 diagnosis groups and 200 procedure groups which includes some rarer conditions: these models are part of our national hospital mortality monitoring system).

ICL DFU undertakes its research and analysis to provide measures of the quality of healthcare delivery by healthcare providers. For certain healthcare specialities or areas, variations can be shown by provider to support the management information for the NHS.

This work also aims to:

• Compare hospital mortality rates for in-hospital deaths with rates for all deaths to evaluate the effect of differential discharge policies

• Calculate total post-operative mortality rates, e.g. when comparing operative techniques such as laparoscopy and open approaches

• Assess potential quality of care issues by comparing the cause of death with the reason(s) for admission, e.g. for surgical patients who are discharged within 30 days of the procedure but who die at home, and whether the death is related to their disease process or to complications of treatment

Under this Agreement ICL DFU will continue the unit’s long-running work and its principal themes of:

• Developing and validating indicators of quality and safety of healthcare, particularly by consultant and hospital

• Showing variations in performance by unit and sociodemographic stratum

• Predicting risk and adjusting risk of indicators and variations and any other methodological aspects as they arise

Imperial College London has an Imperial managed area within the Virtus data centre that is only used to host equipment for secure data processing. ICL DFU servers are hosted in a restricted access ‘secure enclave’ that is designed and managed for processing sensitive data: this secure enclave is part of Imperial’s GDPR compliance strategy.

The secure area has DSP Toolkit (EE133887-SPHTR) and ISO27001 compliance (16170-ISN-001). Virtus do not have access to the data or servers; other Imperial systems and users do not have access to the data or servers. ICL DFU also has DSP Toolkit (8HL46-FOM-SPH). The area is physically and logically separate from other organisations. Physical access is by swipe card and monitored by Imperial College. The enclave is an isolated environment within the Imperial College network.

The secure enclave system can only be accessed using Imperial registered terminals. These terminals are only used to access the enclave and cannot browse the internet or otherwise communicate; standard Imperial desktops do not have access to ICL DFU servers. All processing occurs within the enclave on virtual machines that cannot communicate externally.

Data supplied under this agreement (DARS-NIC-383203-Q8B9L) can only be linked with de-identified data from agreement DARS-NIC-12828-M0K2 using the bridge file supplied by NHS Digital. No other data linkage will be attempted or carried out.

Data will not be processed or stored at any organisation not specified in the previous processing and storage locations.

HES data supplied under agreement DARS-NIC-12828-M0K2 and mortality data supplied under this agreement are transferred from NHS Digital to ICL DFU. These data are used to identify measures of quality and safety of healthcare.

The HES data includes bespoke identifiable data (NHS Number and local patient identifier) under the separate DARS-NIC-12828-M0K2. The data sets are:

• Admitted patient care

• Outpatient

• Critical care

• A & E

The HES data supplied to Imperial College London under agreement ref DARS-NIC-12828-M0K2 is supplied as 2 separate files, one which flows with identifiers under a section 251 support and the other which flows as pseudonymised data under the Health and Social Care Act. The data which flows under Section 251 has objections upheld. The pseudonymised data does not. It is the pseudonymised data which the mortality data being supplied under this agreement will be linked to.

Imperial College holds 2 databases to store data – A Research database and a Patient Identifiable database. The patient identifiable database is used to provide a re-Identification service to NHS provider trusts for patients that are / were under their care. The patient identifiable data is only used for the re-identification service and no other purpose. The release of data for this purpose is done under agreement ref DARS-NIC-12828-M0K2. The patient identifiable database only contains NHS Number, local patient identifier, and a generated pseudo identifier.

Patient identifiers are stored separately to the unit’s research database which holds the HES extracts (including sensitive fields). ICL DFU researchers have no access to identifiable fields. Therefore, the data which the Civil registration data will be linked to will be the pseudonymised data not the identifiable.

Mortality data supplied under this agreement may be linked with HES Data Supplied under DARS-NIC-12828-M0K2 for the purposes of cross HES-civil registration mortality analysis. Data will be linked using Encrypted_HESID. Linkage of the mortality data to identifiable HES data provided to Imperial College London under this agreement is not permitted.

All individuals with access to the data are substantive employees of Imperial College London. All outputs (including those shared with collaborators) are aggregated with small numbers suppressed in line with the HES Analysis Guide.

No record level data provided under this agreement will be shared with any third party, commercial company or Dr Foster Limited.

There will be no attempt re-identify data by the applicant or by any third party.

NHS Digital reminds all organisations party to this agreement of the need to 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).

Outputs detailed are ongoing. It is anticipated that the main outputs over the next 12 months (i.e. until December 2021) will be a continuation/update of those already achieved. Two new projects are also listed.

All outputs are aggregated with small numbers suppressed in line with the HES Analysis Guide.

Research into variations in quality of healthcare by provider: background to proposed work

The Dr Foster Unit at Imperial College use hospital administrative data in the form of HES/civil registration Mortality data to provide measures of quality and safety of delivery of healthcare by provider, or in some instances, by area or time. The unit’s work focuses on quality of care and patient safety, including healthcare acquired infections and safety indicators. Collaborative projects with clinical colleagues have helped develop and validate healthcare quality indicators other than mortality, including bariatric surgery, primary angioplasty rates, indicators for stroke care, obstetric care, orthopaedic redo rates and returns to theatre.

Proposed work will continue with the Unit's principal themes:

i) developing and validating indicators of quality and safety of healthcare, particularly by consultant and hospital;

ii) show variations in performance by unit and sociodemographic stratum;

iii) risk prediction and risk adjustment of such indicators and variations and any other methodological aspects as they arise.

Previous work has been published in various journals, including the British Medical Journal (BMJ) Quality and Safety, Health Services Research and specialty-specific journals (orthopaedics and cardiology), and ICL DFU intend to continue such work. Further, ICL DFU plan on presenting further work at conferences, such as HSR UK and International Society for Quality in Healthcare (ISQua).

With a team of orthopaedic surgeons and anaesthetists, DFU will analyse the trends and outcomes for admissions with periprosthetic fracture. This condition seems to be on the increase and represents a significant burden to the NHS and challenge for staff specialist training. This will use HES data linked to ONS mortality records to capture total 30-day death rates. The work has been submitted to a journal, “Injury” and part of this was presented at the Health Service Research UK (HSRUK) conference in 2019. For the resubmission of this, ICL DFU need to run further mortality analysis (Kaplan-Meier plots). With colleagues from the National Heart and Lung Institute (NHLI), ICL DFU compared NHS hospital activity for heart failure patients with and without cardiac devices in order to estimate treatment costs and the impact of device implantation, including subsequent complications. This used HES linked to ONS mortality, the latter to provide five-year total death rates and allow activity and cost estimates to be adjusted for variable follow-up lengths due to death. A full report has been written up and given to the funder, and ICL DFU are currently writing this up for a cardiology journal. This will involve further mortality and hospital use analysis on patients with left ventricular assist devices, reiterating the current renewal application.

Benefits detailed are ongoing. It is anticipated that the benefits over the next 3 years will be a continuation of those already achieved.

1) Ongoing benefits from previous work

The Imperial Unit’s methodological research forms the basis of a near real-time monitoring system, currently used by 70% of English NHS Acute Trusts to assist them in monitoring a variety of casemix-adjusted outcomes at the level of diagnosis and procedure groups. The unit works with the Care Quality Commission (CQC), contributing to its surveillance remit using the same methods and data. From the Unit's monitoring system, monthly mortality alerts are generated, based on high thresholds, which they have been running since 2007. This was pivotal in alerting the then Healthcare Commission (HCC) to problems at the Mid Staffordshire NHS Foundation Trust between July and November 2007. The resulting Public Inquiry recognised the role that the ICL DFU’s surveillance system of mortality alerts had to play in identifying Mid Staffs as an outlier. Key recommendations, reflecting the unit’s work, are that all healthcare provider organisations should develop and maintain systems which give effective real-time information on the performance of each of their services, specialist teams and consultants in relation to mortality, patient safety and minimum quality standards. ICL DFU’s monitoring system continues to generate mortality alerts, which is sent to hospitals, copying in the CQC, to aid their quality improvement efforts.

A further recommendation is that summary hospital-level mortality indicators should be recognised as official statistics. With continued access to the data, this monitoring tool from Dr Foster Unit at Imperial College that detected Mid Staffs will continue to monitor patient outcomes at acute hospitals and be ready to detect any future outliers. ICL DFU will be able to assist the investigation of variations in outcomes at a local level by providing a set of fields from the analyses to authorised users within trusts to enable reconciliation with local information systems and the instigation of clinical audits and case note reviews. The Unit's mortality outlier outputs are used by CQC within their Hospital Inspection framework.

As a result of Dr Foster Unit at Imperial College leading role in the development of hospital mortality measures, in 2010 they were invited to contribute to a Department of Health (DoH) Commissioned expert panel (Steering Group for the National Review of the Hospital Standardised Mortality Ratio) to develop a national indicator of hospital mortality. The resultant Summary-level Hospital Mortality Indicator (based in part on the Unit's HSMR methods) is now a public indicator used by all acute trusts. A relatively “poor” SHMI should trigger further analysis or investigation by the hospital Board. The review (published in July 2013) into the quality of care and treatment provided by 14 hospital trusts with consistently high mortality in either measure led to 11 out of the 14 trusts identified being immediately placed on special measures.

The review also informs the way in which hospital reviews and inspections are to be carried out with the recommendation that mortality is used as part of a broad set of triggers for conducting future inspections. ICL DFU continue to advise NHS Digital as part of the technical group on methodological issues around the Summary­level Hospital Mortality Index (SHMI).

The Unit’s research on specific aspects of care has received a high media profile and has been highly cited. The research on weekend mortality in emergency care, analysis of mortality associated with the junior doctor changeover and work on elective procedures and mortality by day of the week resulted in front page broadsheet coverage, and radio and TV interviews. ICL DFU are currently rerunning the junior doctor changeover work with up to date data to see what has changed.

The Dr Foster Unit completed a two-year National Institute for Health Research (NIHR)-funded project looking at predictors of readmissions and one-year mortality (in and out of hospital) in patients with chronic diseases (heart failure and COPD). This work followed on from DFU’s previously published studies on readmissions in heart failure patients. Mortality acts as a competing risk͟ for readmission, and it is therefore essential to know whether a patient has been discharged alive but subsequently dies and is therefore no longer at risk of readmission. This report was published by the funder in 2018 at: Bottle A, Honeyford K, Chowdhury F, Bell D, Aylin P. Factors associated with hospital emergency readmission and mortality rates in patients with heart failure or chronic obstructive pulmonary disease: a national observational study. Southampton (UK): NIHR Journals Library; 2018 Jul. After approval, DFU will publish peer reviewed papers from this work. The first two of these came out this year: Honeyford K, Bell D, Chowdhury F, Quint J, Aylin P, Bottle A. Unscheduled hospital contacts after inpatient discharge: A national observational study of COPD and heart failure patients in England. PLoS One 2019;14(6):e0218128.

Honeyford K, Aylin P, Bottle A. Should Emergency Department Attendances be Used With or Instead of Readmission Rates as a Performance Metric?: Comparison of Statistical Properties Using National Data. Med Care 2019;57(1):e1-e8.

2) Expected benefits from the proposed work

For future research, it is important to be able to capture deaths occurring following discharge from hospital to assess the full mortality burden relating to that hospitalisation. Out of hospital deaths are particularly useful for surgical outcomes, e.g. for the calculation of total 30-day post-operative death rates, as the effect of premature discharge (in terms of mortality) would otherwise go unnoticed. Longer-term follow-up of hospitalised patients, e.g. using one-year survival, necessitates being able to capture all deaths, not just those occurring in hospital. For this reason, the Summary Hospital-level Mortality Indicator (SHMI) specification requires out of hospital post discharge deaths. As described above in relation to the project on heart failure and COPD, mortality is a “competing risk” for important non-fatal outcomes such as readmission. Accurate prediction of the risk of these other outcomes will help with risk stratification and health service planning, and is not possible without total mortality.

Knowledge of the cause of death is particularly important for quality improvement. The relation between the cause(s) of death and the reason(s) for admission is of particular interest too. The place of death, including whether it was an NHS institution, is necessary to monitor end-of-life services. Further, another interest is the proportion of patients who die at home. As part of ICL DFU’s work on heart failure, the analysis is expected to extend for the funder to cover cause of death in patients with and without a cardiac implanted device. If there is a difference in mortality rates, it is useful to know whether they are explained by deaths from cardiac or from non-cardiac causes.

Ongoing analysis of the mortality alerting system will allow the Dr Foster Unit to improve the alerting process, reducing the number of false positives and unnecessary effort spend by hospitals investigating them. It will also provide advice to hospitals who receive the mortality alerts on how to follow them up and learn, for example, which are the key contributing factors in the alerts.

Benefits detailed in the measurable benefits section are ongoing, and it is anticipated that continued data access will lead to a continuation of the benefits already achieved in the next 36 months. However, there are some key, yielded benefits to note here from previous work.

The national mortality monitoring system mentioned identified problems at the Mid Staffordshire NHS Foundation Trust between July and November 2007 using in-hospital mortality. The resulting Public Inquiry recognised the role that the unit’s surveillance system of mortality alerts had to play in identifying Mid Staffs as an outlier. Key recommendations, reflecting the unit’s work, are that all healthcare provider organisations should develop and maintain systems which give effective real-time information on the performance of each of their services, specialist teams and consultants in relation to mortality, patient safety and minimum quality standards. A further recommendation was that summary hospital-level mortality indicators should be recognised as official statistics

ICL DFU conducted a piece of collaborative working with the University of Manchester, supported by the CQC. This work aimed to improve understanding of the Unit mortality alerts and to evaluate their impact as an intervention to reduce avoidable mortality within English NHS hospital trusts by focusing on two conditions commonly attributed to mortality alerts: acute myocardial infarction and septicaemia. The Dr Foster Unit at Imperial College provided a descriptive analysis of all alerts, their relationships with other measures of quality and their impact on reducing avoidable mortality. This report was approved by the funders in late 2016. Our analysis of the cause of the high-mortality alerts, the mortality falls following them and of hospitals’ responses to receiving them provide important knowledge of how a monitoring system works in practice, such as which parts of the system are used and valued by hospitals. All record-level data processing was done by the unit at Imperial College and the unit controlled how these data were processed. Three papers and the HS&DR report for the funder, NIHR, have been published:

The Unit’s “Out of hours” work has been a key driver in moving NHS towards 7/7 care. Headlines include, “NHS Services – open seven days a week: every day counts” and, “Sunday Times Safe Weekend Care”. As a result of the Unit's published research into the junior doctor changeover, a new protocol was introduced: a week's shadowing where newly qualified doctors worked alongside more senior ones for a week before commencing work in August. The Academy of Medical Royal Colleges published proposals (16th April 2014) suggesting all Foundation Year 1 posts should begin on the first Wednesday in August as has always been the case, but other training posts should begin in September.

DFU worked with the University of Leicester on thoracic aortic disease (TAD) looking at variations in rates of surgery and mortality between centres. There seems to be wide variations in the rates of treatment for this condition, but it is unclear how this impacts on outcomes. In-hospital mortality only captures part of the effect. With the recent growth in the number of endovascular procedures (TEVARs), post-discharge deaths are vital to assess the impact of these procedures and of TAD services in general. The first paper as a result of this work was published in January 2017 (Bottle et al, JAHA 2017 – see list below). This paper is being used by Prof Sir Muir Gray, in his assessment of how to reorganise services for these patients; ICL DFU await the publication of his team’s report. All record level data processing was done by the unit at Imperial College and the unit controlled how these data were processed.

Published analyses on the use of NHS services by patients with heart failure was cited in a 2018 NICE guideline and in the 2019 NHS Long-term Plan in the section on better management for patients with long-term conditions.

Examples of key published research that have used HES/civil registration data include:

Cecil E, Wilkinson S, Bottle A, Esmail A, Vincent C, Aylin PP. National hospital mortality surveillance system: a descriptive analysis. BMJ Qual Saf 2018;27(12):974-981.

Cecil E, Bottle A, Esmail A, Wilkinson S, Vincent C, Aylin PP. Investigating the association of alerts from a national mortality surveillance system with subsequent hospital mortality in England: an interrupted time series analysis. BMJ Qual Saf 2018;27(12):965-973.

Aylin P, Bottle A, Burnett S, Cecil E, Charles KL, Dawson P, D’Lima D, Esmail A, Vincent C, Wilkinson S, Benn J. Evaluation of a national surveillance system for mortality alerts: a mixed-methods study. Southampton (UK): NIHR Journals Library; 2018 Feb.) (Updated Sep-2019)

Bottle A, Honeyford K, Chowdhury F, Bell D, Aylin P. Factors associated with hospital emergency readmission and mortality rates in patients with heart failure or chronic obstructive pulmonary disease: a national observational study. Health Services and Delivery Research 2018; 6(26).

Rao A, Bottle A, Darzi A, Aylin P. Sequence Analysis of Long-Term Readmissions among High-Impact Users of Cerebrovascular Patients. Stroke Res Treat 2017: 7062146.

Bottle A, Ventura CM, Dharmarajan K, Aylin P, Ieva F, Paganoni AM. Regional variation in hospitalisation and mortality in heart failure: comparison of England and Lombardy using multistate modelling. Health Care Manag Sci 2018;21(2):292-304.

Bottle, A., Mariscalco, G., Shaw, M. A., Benedetto, U., Saratzis, A., Mariani, S., Murphy, G. J. Unwarranted Variation in the Quality of Care for Patients With Diseases of the Thoracic Aorta. J Am Heart Assoc 2017; 6(3):66.

Bottle A; Goudie R; Cowie MR; Bell D; Aylin P. Relation between process measures and diagnosis-specific readmission rates in patients with heart failure. Heart 2015;101(21):1704-10.

Bottle A, Aylin P, Bell D. Effect of the readmission primary diagnosis and time interval in heart failure patients: analysis of English administrative data. Eur J Heart Fail 2014; 16(8): 846-853.

Aylin P; Alexandrescu R; Jen MH; Mayer EK; Bottle A. Day of week of procedure and 30 day mortality for elective surgery: retrospective analysis of hospital episode statistics. BMJ 2013;346:f2424.

Palmer WL; Bottle A; Davie C; Vincent CA; Aylin P. Dying for the Weekend: A Retrospective Cohort Study on the Association Between Day of Hospital Presentation and the Quality and Safety of Stroke Care. Arch Neurol 2012;69:1296-1303.

Aylin P, Yunus A, Bottle A, Majeed A, Bell D. Weekend mortality for emergency admissions. A large, multicentre study. Qual Saf Health Care 2010;19:213-217.

For full publication list see unit website: http://www.imperial.ac.uk/dr-foster-unit/publications/