NHS Digital Data Release Register - reformatted

The Health Foundation

Project 1 — DARS-NIC-15411-C9Z9L

Opt outs honoured: N

Sensitive: Non Sensitive

When: 2017/06 — 2017/08.

Repeats: One-Off

Legal basis: Health and Social Care Act 2012

Categories: Anonymised - ICO code compliant

Datasets:

  • Hospital Episode Statistics Admitted Patient Care
  • Hospital Episode Statistics Accident and Emergency
  • Hospital Episode Statistics Critical Care
  • Hospital Episode Statistics Outpatients

Benefits:

The Health Foundation has strong links with NHS teams, national policymakers (e.g., NHS England) and patient advocacy groups. Examples of these links are: • Senior members of Health Foundation staff regularly meet with senior representatives from across government, including the Treasury, Department of Health and Arms-Length Bodies (e.g. Monitor, CQC, NHS England, HEE). • The Health Foundation is currently working on projects for, or jointly with, organisations including NHS England. For example, we have jointly funded an evaluation of the Patient Activation Measure in the NHS, and work together on the ‘5000 Safety Fellows Programme’ • People across the Health Foundation regularly engage with policy makers at all levels on a range of topics where the Health Foundation has particular expertise: policy, data analytics, economics, patient safety and person-centred care. The Health Foundation views are regularly sought on health policy and practice, meaning that the findings from these HES-based analyses will be communicated directly with policymakers. • The Health Foundation has a long history of funding programmes across the NHS which help to improve the quality of health care. For example, the Health Foundation have funded work on the relationship between patient flow, costs and outcomes in two NHS hospital trusts, which is related to the new project on understanding the drivers of A&E attendances. • The Health Foundation have an active audience of professionals working in the NHS, many of whom are fellows sponsored by the Health Foundation, award-holders or part of the Health Foundation’s alumni. The Health Foundation provides leadership and advice on quality improvement as well as commentary on health care policy. The Health Foundation’s analysis of HES data will inform these activities. The Health Foundation assess their impact using objective measures (e.g., number of publication downloads, publication citations and attendances at events and seminars) as well as record specific instances where their work has informed decision making for the NHS and improved the quality of care ultimately delivered to patients. Due to these strong links with the health service, the Health Foundation is in a good position to reach as much beneficiaries as possible for each of the four projects. Each project is expected to have the following impact: 1. The funding pressures facing health care in England for the next 15-20 years, and how service transformation can lead to greater sustainability Following the analysis The Health Foundation will discuss the results of the scenarios of different models of delivery to inform policy makers on the impact that different decisions would have. These will mostly be discussions of the national situation, with some regional analysis at a level similar to government office regions. To maximise the benefit of the work on this project, it is important that information provided to the Department of Health and Monitor is based on the latest available data. Health Foundation will update the existing work on the basis of the latest available data, that is up to and including 2015/16, providing a comprehensive overview of NHS funding challenges and transformation programmes. Progress to date: Work on this project is ongoing, and the final modelling is expected to be completed in the summer of 2017. In the meantime our analysts have been liaising with experts in the Department of Health and Monitor, to make sure that the analysis maximises potential benefit. Discussions with these experts have led to a better understanding of costing the data using reference cost data and PbR Tariffs. 2. Phenotyping English Hospitals This project will help to identify issues surrounding the quality of care of providers for elderly patients via mapping the relationship between length of stay and readmission rates. Identifying poorly performing providers will be of interest for the Care Quality Commission as it may inform the selection of trusts for inspection. On the other hand, identifying providers who perform well, offers the opportunity to conduct qualitative research to understand reasons for such good performance. Good performance processes can be collected and inform best-practice. This work will also use the latest financial years of data requested (i.e. 2013/14-2015/16). Health Foundation believe this would add additional value and make our findings more relevant for policy. The demographics for England are rapidly changing towards an older and frailer population. Subsequently, demand for health services within this subset of the population is rising, putting increasing pressure on hospitals. The value of receiving more updated data is to allow us to investigate how healthcare providers responded to this change in demographics. Similar to the workload project, by using the most recent data, Health Foundation will be able to make a stronger contribution to the ongoing debate about care for the elderly and frail. Progress to date: Experts from Nuffield Trust, Dr. Foster Unit and health Economics Research Unit at Imperial College have also been contacted to maximise policy relevance for this study.The original deadline of the work was November 2016. However, the research team is still conducting the analysis and no final output has been produced till date. The main reasons for delay relate to the extensive research into technical aspects of the methodology. Preliminary work informed an expansion of research into the relationship between readmission rates and publicly available patient reported outcomes data at the Trust-level. Health Foundation performed a panel data analysis and a research paper will be submitted for peer-review to Medical Care in February 2017. Health Foundation aim to produce the final findings of the study for September 2017, and to write a research paper aimed for publication on peer-reviewed international journals by the end of 2017. Due to the length of the peer-review process the Health Foundation will require to retain the data for at least two years, in order to ensure the replicability and validation of the results and, ultimately, their publication, 3. Penalising readmission: success or failure The penalisation of readmissions has been in place since 2011/2012, yet, has rarely been studied. The Health Foundation’s study will inform policy makers about the likely effectiveness of the chosen financial tool and draw comparisons to the Affordable Care Act in the US. This will inform Monitor and other organisations involved with the policy debate about changes to the way that hospitals are reimbursed for emergency care in the NHS. Progress to date: This project is now completed. Early findings were presented at the American Health Economics Association in Philadelphia in July 2016; to internal seminars at The Health Foundation and to external events with the Dr. Foster Unit. The research paper has been submitted to a best student paper competition at the International Health Economic Association in November 2016. The outcome of this competition will be announced in March 2017. At the beginning of March 2017, the authors will submit the research paper to the Journal of Health Economics. Due to the lengthy process of peer-reviewing articles, we expect to require an extension for data access of at least two years. As a result of the work on penalisation of readmission, we would like to further investigate the effect of factors related to workload. This work will use the latest financial years of data requested as part of this application. The reason behind this request is that over recent years hospital trusts have reported increasing levels of bed occupancy rates, especially during the winter months. For many trusts, occupancy rates have reached critical levels of above 90%, which anecdotally leads to extreme strains for individual providers and potential adverse consequences for patients. Furthermore, high occupancy rates can also be linked to increasing waiting lists and not meeting A&E targets, which has been a popular subject within the national media. Investigating the relationship between workload and the risk of re-admissions is therefore a very timely concern, and we are particularly interested in the effect that workload pressures have on readmission rates since 2013. As a result, looking at the most recent HES data would feed directly into the current debate about bed pressures. 4. Analysis of factors associated with the performance of A&E departments in England This project will provide guidance that is urgently needed by NHS England and other policy makers regarding which factors impact most on A&E performance, in particular waiting times. This is needed because an increasing proportion of departments have not met waiting time targets. The project will provide insight about how performance can be improved. Progress to date: This project is near completion. Initial analysis identifying factors associated with A&E performance was presented to the NHS England analytical team in October 2015. Following the meeting with NHS England, a number of refinements were identified. The findings will be published in a working paper series by Spring 2017. Aiming to submit the paper for peer-review later in the year.

Outputs:

As outlined above, outputs for all projects will be in line with best practice guidelines on statistical disclosure control and privacy protection. The outputs for each project are as follows: 1. The funding pressures facing health care in England for the next 15-20 years, and how service transformation can lead to greater sustainability The primary output of this project will be a Health Foundation report, similar in style to the report “A Decade of Austerity” Nuffield Trust 2012. It will provide an update to the funding pressures that are expected over the next 15-20 years. This report would be made available through the Health Foundation’s website by 2017. Depending on the results, The Health Foundation will look to publish the results of the historic trends of hospital admission for chronic conditions in a public health journal. 2. Phenotyping English Hospitals The study will be completed by the end of 2017 and ultimately published in peer-reviewed international journals. These will be a mixture of health services research journals (e.g. Health Services Research and Policy) and economics journals (e.g. the Journal of Health Economics). These journals are read by policy makers, nationally and internationally, who wish to identify and classify hospitals according to the level of quality of care that they provide. The research will also be presented at conferences and events aimed at policy makers. The Health Foundation will present internally at The Health Foundation and also to statutory bodies such as the Department of Health and Monitor. The Health Foundation will publish a summary of the research on the Health Foundation’s website. 3. Penalising readmission: success or failure This project serves a similar audience as the previous project, and the same type publication will be pursued. The main work on this study has now been completed and a research paper for publication on peer-reviewed journals has been produced. While conducting the analysis the researchers developed a deeper understanding of the data and further important research to be derived by this study. The additional research looks into variation of 30-day readmission rates across commissioning groups and investigates how those changes developed over time. Health Foundation hypothesis that a decrease in variation measured via the standard component of variation implies overall improvements in quality of care in the English NHS. Health Foundation are therefore planning to write up and submit a research paper for BMK Open in summer 2017. In addition to an academic research paper, the results will form part of a PhD thesis chapter. 4. Analysis of factors associated with the performance of A&E departments in England The findings of this study will be disseminated through meetings with senior policy makers and NHS leaders, plus a short policy-focused report on the Health Foundation’s website. The Health Foundation will also submit the findings to a peer-reviewed journal as above. The study will be completed by December 2015. Each of the projects listed in the application will produce a number of publications. These publications typically take the form of: - Reports aimed at policy makers, disseminated through the Foundation’s website - Peer-reviewed journal articles - Blogs on the Foundation’s website or others (e.g. Health Service Journal) - Conferences and presentations - Press releases The Health Foundation’s approach to dissemination includes not only publications but also engagement with national policy makers, practitioners and researchers. Each project has member of the Foundation’s Communications team leading on dissemination of findings. The Foundation works closely with key stakeholders and has strong link with NHS teams, national policymakers (e.g., NHS England) and patient advocacy groups. Examples of these links are: - The Health Foundation is currently working on projects for, or jointly with, organisations including NHS England. For example, The Health Foundation have jointly funded an evaluation of the Patient Activation Measure in the NHS, and work together on the ‘5000 Safety Fellows Programme’ - People across the Health Foundation regularly engage with policy makers at all levels on a range of topics where The Health Foundation has particular expertise: policy, data analytics, economics, patient safety and person-centred care. The Health Foundation’s views are regularly sought on health policy and practice, meaning that the findings from these HES-based analyses will be communicated directly with policymakers. - The Health Foundation has a long history of funding programmes across the NHS that help to improve the quality of health care. For example, The Health Foundation have funded work on the relationship between patient flow, costs and outcomes in two NHS hospital trusts, which is related to the new project on understanding the drivers of A&E attendances. - The Health Foundation have an active audience of professionals working in the NHS, many of whom are fellows sponsored by the Health Foundation, award-holders or part of The Health Foundation’s alumni.

Processing:

In case of all four projects, data will be processed by a limited number of researchers within the Health Foundation’s secure environment. All researchers with access to the data will have completed an accreditation course on data protection legislation and statistical disclosure control, completed an information security training specific to the Health Foundation’s infrastructure, and signed a non-disclosure agreement and the terms of use of the secure environment. HES Data will only be processed on the Health Foundation’s premises on 90 Long Acre in London and any publication derived from the data will be subjected to best practice guidelines on Statistical Disclosure Control (SDC) including the Code of Practice on Confidential Information, the Anonymisation Standard for Publishing Health and Social Care Data and the code of practice published by the ICO, Anonymisation: managing data protection risk code of practice before being released form the environment. All data requested will be used exclusively for the purposes stated in this application. Project specific processing of data is outlined below. 1. The funding pressures facing health care in England for the next 15-20 years, and how service transformation can lead to greater sustainability For this project, the Health Foundation require person level data linked across for inpatient (Elective, non-elective and day case), outpatient and A&E. The critical care dataset will be used separately (rather than linked to the other data). The project aims to project cost pressure over the next 15-20 years; therefore the Health Foundation needs to track historic trends over a similar period for the whole population. The Health Foundation will initially estimate the demand pressures for each service type separately, to allow for the differences in the time periods covered by the relevant data sets. The data required within this project are: • Inpatients between 1997/98 and 2015/16 • Outpatients between 2003/04 and 2015/16 • A&E between 2007/08 and 2015/16 • Critical Care between 2008/09 and 2015/16 Since this involves modelling the evolution of health care utilisation for people with various specific health conditions in the different government office regions, it will need comprehensive data coverage for these time periods. For each service type, the Health Foundation will explore how service use is affected by factors including age, sex, residence (LSOA), treatment provider (site and trust), commissioner, diagnosis and procedure codes, treatment function, admission method and time. The Health Foundation will also explore how the level of use of one service affects demand for other services (excluding critical care). As this will require an overlapping time period, the Health Foundation can only explore these interactions for shorter periods. The Health Foundation are comfortable with this limitation as the information will be used to create scenarios for analysis on how policy decisions might impact on total cost projections in addition to the overall projections. The primary models will be independent service-specific linear, or log-linear person-level models of the trends in the level of activity for emergency inpatients, elective inpatients, outpatients, A&E and critical care. The results of the models will be used to create projections for future use of these services at a national level, and by government office regions (or other similar sized areas as appropriate). The Health Foundation will therefore apply the results of the analysis of historic trends to publicly available population projections produced by the ONS. The Health Foundation will also measure how service use differs between people with various chronic conditions, namely diabetes, COPD, asthma, coronary heart disease, cancer, arthritis, dementia, epilepsy, renal disease and stroke. The Health Foundation will identify these groups using the diagnosis codes present within the inpatient dataset. Again, as The Health Foundation plan to project the numbers of patients in these groups, The Health Foundation will explore the trend over time for inpatient admissions between 1997/98 and 2015/16. These will be done using a linear regression, with transformations applied where appropriate to ensure the best fit for each condition. By producing trends in this way, The Health Foundation is able to explore the trends for certain co-morbidities, instead of using single condition prevalence projections. However The Health Foundation will compare their estimates to national data on prevalence of these conditions where possible for assurance. The projections for costs on these services will be combined with projections for other NHS services, such as GP attendances and community pharmacies, produced using publically available data. The combining of data in this way will primarily be done at a national level, and will not be done at a level lower than government office region. Having established the models and projections, The Health Foundation will use the results to test the impact of a series of assumptions around future changes in NHS delivery. This will take the form of modelling assumptions on how service delivery might change at a national level. For example, The Health Foundation will test the potential impact on total NHS spending of a substantial investment in GP practices, which might be expected to lead to a reduction in hospital admissions. More complex policies are likely to impact on multiple hospital services for certain types of people. In these cases, The Health Foundation will need to understand the relationship between the different hospital service types. For example, if a new community diabetes service is set up that includes additional outpatient appointments, but might reduce the need for inpatient care, The Health Foundation will produce summaries of the current levels of use for these services for people with diabetes, to understand the full impact of the change. Again, these results will only be published at a level no lower than government office regions. The Health Foundation will also explore the impact of likely productivity growth on the projected growth. The Health Foundation will base this on evidence on recent and longer-term levels of productivity growth by running random effects, fixed effects and stochastic frontier analysis on weighted activity of different types of providers. As with other analysis, results will be used at a national or large regional level. 2. Phenotyping English Hospitals For this project, inpatient, outpatient and Accident & Emergency data over the last ten years will be used to inform descriptive analysis (i.e. 2003/04 – 2015/16, where available). The project requires data on older people (age 65+). This descriptive analysis will look at the variation of length of stay vs. readmission rates by hospital provider in England. Multivariate regression analysis will be used to estimate relationships between patient-characteristics and outcome variables. Risk-standardised length of stay for each hospital-year will be calculated, by adjusting for differences in patient case-mix across hospitals over time. Likewise, the risk-standardised 30- day readmission rate will be calculated. Based on these constructed measures for length of stay and readmission rates, hospital phenotypes will be identified using a group-based, semi parametric mixture modelling approach. The determination of phenotypes will depend upon the Bayesian Information Criteria index, average posterior probability of phenotype and 95% confidence intervals of adjacent trajectories. Variables covering the following areas are likely to be included in the analysis adjusting for case mix: emergency admissions, source of admission, patient characteristics, provider code and deprivation measures. Results from the analysis are will include (i) Estimated regression coefficients (and their associated p-values) relating to the associations between provider characteristics and the phenotype. This information will help understand the associations between hospital characteristics (e.g., teaching hospital status, region) and the two dependent variables (length-of-stay and 30-day readmission rates). (ii) Graphs mapping the changes in readmission rates and length of stay by Trust over a 10 year period. (iii) Depending on the number of identified phenotypes, graphics displaying the observed trajectories for each. Each trust can be attributed to one of the created graphics – making classification and groupings of trusts easier and more visual. All outputs for this project will be at Trust level and no additional sources of information will be combined with HES for this project besides publicly-available Trust-level data. 3. Penalising readmission: success or failure For this project, inpatient, outpatient and Accident & Emergency data will be needed covering the four years before and after the policy change in England (i.e., 2007 – 2012/13). Descriptive analysis will look at crude (unadjusted) readmission rates at the Trust level, as well as 30-day readmission rates by Trust when adjusted for case-mix using the method described in project (2). The consequences of the policy change will be assessed using a segmented regression analysis of interrupted time series, comparing trends before and after the intervention. The analysis will be conducted for all hospital admissions and for subsets of admissions defined by health condition (as determined by the recorded diagnosis codes). Similar to the above results, results from the analysis will include estimated regression coefficients and associated p-values by NHS Trust, over time, as well as graphs representing the 30-day readmission rate per trust on a monthly basis. We would also like to understand pressures affecting changes in readmission rates. This strand of the study will investigate the effect of factors related to workload, i.e., admission rates, occupancy rates and bed occupancy rates. This will be in addition to financial penalties employed to reduce re-admissions and results will provide a more detailed understanding of driving mechanisms behind changing readmission rates. For this project we will be needing inpatient, outpatient and A&E data covering the most recent years (i.e. 2015/16). 4. Analysis of factors associated with the performance of A&E departments in England This project focuses on A&E attendances for the whole population of England, and will cover the time period 2007 – present. In addition to A&E data, The Health Foundation will use inpatient and outpatient data at episode level to characterise patients in terms of their demographics, diagnoses, number of previous attendances, and missed appointments. This information will be used within a series of panel data models to investigate how the performance of an A&E department varies with factors related to demand and supply of health care, and the characteristics of patients.

Objectives:

As part of this application for an extension to the data retention period, the Health Foundation would like to add the latest financial years of the same datasets (and same variables specification) to be used only in the ongoing strands of analysis. More specifically, the latest data series will be used for the following work packages: 1. The funding pressures facing health care in England for the next 15-20 years, and how service transformation can lead to greater sustainability 2. Phenotyping English Hospitals 3. Penalising readmission: success or failure (only for the sub-strand analysis on “effect of factors related to workload”, as the rest of the work is now completed) This will provide further value to the work produced until now and add a more recent evidence base to the ongoing analysis packages (further details are provided in section 5.d) The Health Foundation is an independent charity working to improve health and the quality of health care in the United Kingdom. The Health Foundation is requesting access to data for four research projects that aim to inform public discussions about the focus, design and effects of policies intended to improve the quality of health care in the United Kingdom or reduce costs. The projects will inform policy makers and the NHS about the variability in quality and costs of health care in England, and thus help to identify priority areas for improving health and social care These four projects are: 1. The funding pressures facing health care in England for the next 15-20 years, and how service transformation can lead to greater sustainability: To create an economic model of the person level factors that determine use of hospital services to i) estimate of how spending pressures on these services will grow in the future, and ii) estimate the potential impact of policies to reduce these pressures. Progress to date: Work on this project is on-going, and the final modelling is expected to be completed in the summer of 2017. Health Foundation analysts have been liaising with experts in the Department of Health and Monitor, to make sure that the analysis maximises potential benefit. Discussions with these experts have led to a better understanding of costing the data using reference cost data and PbR Tariffs. 2. Phenotyping English Hospitals: Variations in hospital performance and quality of care are significant. This project aims to classify English hospitals with respect to trends in length of stay and 30-day readmission rates for elderly and frail patients. The study follows the phenotyping approach used by Xu et al. (2014). Progress to date: The original deadline of the work was November 2016. However, the research team is still conducting the analysis and no final output has been produced to date. The main reasons for delay relate to the extensive research into technical aspects of the methodology. 3. Penalising readmission: success or failure: Readmission rates have been increasing over the past decades. Various policies have been implemented to revert this trend, including the introduction of financial penalties on hospitals for re-admissions from 2011. However, its effectiveness has never been evaluated. Thus, this project will assess changes in readmission rates following the policy implementation period. Progress to date: This project is now completed. As a sub strand of this study we would like to understand pressures affecting changes in readmission rates (as factors related to workload, i.e., admission rates, occupancy rates and bed occupancy rates) by using inpatient, outpatient and A&E data covering the most recent years (i.e. 2015/16). 4. Analysis of factors associated with the performance of A&E departments in England: A&E departments have been under pressure, and this could be because of changes in demand, supply, the resilience of departments, or wider contextual factors (e.g. health policy). The Health Foundation will use person-level HES data to assess changes in the characteristics of patients attending A&E departments over time. The Health Foundation will also assess the relationships between A&E activity and the volume of patients admitted as inpatients. The Health Foundation will use patients’ clinical information to derive risk-adjusted variables at the Trust level, which will be used within a panel data model that relates these to variables relating to the supply and demand of health care. Progress to date: This project is near completion. Initial analysis identifying factors associated with A&E performance was presented to the NHS England analytical team in October 2015. Final findings will be published in a working paper series by Spring 2017. The Health Foundation is committed not to combine the requested HES data with any other data source that might result in increased re-identification risk. The Health Foundation is happy to formalise this commitment in the Data Sharing Agreement (DSA) for this extract. The Health Foundation’s secure processing environment holds dedicated projects folders that will be used for projects using HES data, as specified in The Health Foundation’s information security policy. Users have access only to data that relates to their project, and the content of all project folders is reviewed regularly to make sure The Health Foundation deliver on this commitment. The process for moving additional data sources on to the secure environment is a carefully controlled process monitored by the Data Manager, part of this process is to review the purpose statement of each project and restrictions on the use of specific datasets. The only data sources that may be used in combination with HES data will not contain any detail that might lead to increased identification risk, but rather add contextual information at an aggregate level (e.g. contextual geography). The Health Foundation has limited the amount of data requested for each project to the minimum amount. For example, the analysis of phenotyping English hospitals is limited to records for patients aged 65 and over, and to ten years rather than the full duration. Although all HES data will be held on The Health Foundation’s secure environment, researchers will only access and analyse the required subset of the data on a day-to-day basis. The largest amount of data is being requested for ‘The funding pressures facing health care in England for the next 15-20 years, and how service transformation can lead to greater sustainability’ in terms of geography, time coverage and population base. Each of the dimensions is discussed below in the context of this project. - Geography – healthcare utilisation displays a lot of regional variation. In order to produce accurate projections England wide all geographical area’s need to be covered in the research data. - Time coverage – the aim of the project it to predict long term healthcare use (15-20 years). In order to accurately predict variation over time (including seasonal variation, business cycles and long term trends) rich historic data is required. - Population – the projects aims to cover hospital utilisation for all conditions. Whole population data gives good coverage (in terms of numbers) on rare conditions. It would be inequitable to narrow the research down to only certain population groups because these groups would then not be considered in important research to inform future budget decisions about the NHS. Reference Xu, X., Li, S.-X., Lin, H., Normand, S.-L. T., Kim, N., Ott, L. S., … Krumholz, H. M. (2014). “Phenotyping” Hospital Value of Care for Patients with Heart Failure. Health Services Research, 1997, 1–17. doi:10.1111/1475-6773.12197


Project 2 — DARS-NIC-35820-Y3H1M

Opt outs honoured: N

Sensitive: Non Sensitive

When: 2016/12 — 2017/02.

Repeats: One-Off

Legal basis: Health and Social Care Act 2012

Categories: Anonymised - ICO code compliant

Datasets:

  • Hospital Episode Statistics Admitted Patient Care
  • Hospital Episode Statistics Accident and Emergency

Benefits:

The Health Foundation has strong links with NHS teams, national policymakers (e.g., NHS England) and patient advocacy groups. The Health Foundation provide leadership and advice on quality improvement as well as commentary on health care policy. The Health Foundations analysis of HES data will inform these activities. The Health Foundation assess the impact using objective measures (e.g., number of publication downloads, publication citations and attendances at events and seminars) as well as record specific instances where the work of The Health Foundation has informed decision making for the NHS and improved the quality of care ultimately delivered to patients. Due to these strong links with the health service, the Health Foundation is in a good position to reach as much beneficiaries as possible for this project. This study aims to evaluate the “weekend effect” by optimising risk adjustment and exploring factors that potentially contribute to it. Moreover, variation in provider performance between trusts will be explored, both cross-sectionally and over time. Identifying time trends in mortality rates between providers will shed light on the possible performance trajectories in improving quality of care at the weekend and any trade-offs in weekday quality and relate these to provider characteristics. This novel approach will provide useful information in light of a general policy move towards seven-day working. In addition, identifying providers who perform well or/and improve over time offers the opportunity to conduct qualitative research in the future that will help to understand the reasons behind these differences. Good performance processes can be collected and form part of a best-practice policy. The analysis for the first phase of the project is expected to be completed by the end of 2016, with the aim to realise the benefits through publication by the Spring of 2017. The subsequent second phase will start in the first quarter of 2017, aiming for completion by the end of the year.

Outputs:

Outputs for this project will be in line with best practice guidelines on statistical disclosure control and privacy protection. The aim is to publish the study and its findings in peer-reviewed international journals. These will be a mixture of health services research journals, and health economics journals. The health services journals the Health Foundation will target are: the BMJ, Health Services Research, Health Policy. The target economic journals include: the Journal of Health Economics, Health Economics, European Journal of Health Economics and the Journal of Health Services Research and Policy. The Health Foundation will aim to submit for publication in spring 2017. The aforementioned journals are read by policy makers, nationally and internationally, who wish to identify and classify providers according to the level of quality of care that they provide. Further engagement with key stakeholders will take place through presentation at conferences and events aimed at policy makers. Throughout the analysis period the Health Foundation will actively engage with statutory bodies such as the Department of Health and Monitor, and international organisations involved in healthcare policy (e.g. OECD, WHO, the World Bank). All published material will be in line with HSCIC’s ‘Code of Practice on Confidential Information’, the ‘Anonymisation Standard for publishing Health and Social Care Data’ and ICO’s guide to ‘Anonymisation: managing data protection risk code of practice’. All outputs will be aggregated with small numbers suppressed in line with the HES Analysis Guide. All results are manually checked by two independent analysts. All outputs (graphs, tables, results) that are released from the secure environment (and therefore being considered for publication) are subject to a process called Statistical Disclosure Control. This process ensures that re-identification of the outputs is not possible. In order to ensure this, the outputs themselves are considered, but also in combination with other information available in the public domain (this is called secondary disclosure). The Health Foundation follow best practice guidelines that are in line with NHS requirements. All outputs are considered by at least two members of staff, prior to being released from our secure environment.

Processing:

For this project data will be processed by a limited number of analysts within the Health Foundation’s secure environment. All researchers with access to the data will have completed information governance and data security training, as well as training specific to the Health Foundation’s infrastructure, and signed a non-disclosure agreement and the terms of use of the Foundation's Secure Data Environment. HES Data will only be processed on the Health Foundation’s premises in London and any publication derived from the data will be subjected to best practice guidelines on Statistical Disclosure Control (SDC) including the Code of Practice on Confidential Information, the Anonymisation Standard for Publishing Health and Social Care Data and the code of practice published by the ICO, Anonymisation: managing data protection risk code of practice before being released from the environment. Data requested: This study will use admitted patient care data and accident and emergency data (cross-section only). The Health Foundation request non-sensitive date relating to patient characteristics (including IMD and rural/urban indicators), admissions, discharges, episodes and spells, clinical data; some provider information and GP practice code. In addition, the Health Foundation are requesting the variable mental category (mentcat), as this variable will help determine, together with the diagnosis codes, the subgroup of patients with mental ill health and severe mental ill health. For the first phase of the project, the cross-sectional analysis the Health Foundation will only use data from 2012/13 to 2014/15, as the calculation of comorbidity indices used for risk adjustment requires data over the three years prior to each hospital admission. The second phase, the time-series analysis, requires data over a ten-year period to identify phenotypes. This will ensure an adequate period over which to identify changes in hospital performance (respective mortality rates) over time. In addition, as in the cross-sectional analysis, this analysis requires data over the three years prior to the initial trajectory analysis for the risk adjustment. Therefore we will need data for a 13-year period, from 2002/03 – 2014/15. Although the majority of the analysis will be at trust level, patient level data is required to adequately risk adjust for case mix. Additional data: The data supplied by HSCIC will be linked to publicly available aggregate provider level data. The data used in the analysis will be supplemented by information that is already available in the public domain. HSCIC published trust level information including trust characteristics, staffing characteristics, number of beds, etc. This information will be linked to the HES using trust identifiers regularly available on HES. Processing: Different methods of risk-adjustment will be explored in the cross-sectional analysis, building on previously established models. The descriptive analysis will look at the variation in admissions and 30-day mortality by provider (hospital site and trust) level. Risk-standardised 30-day mortality rates will be estimated for each provider, by adjusting for differences in patient case-mix. For the cross-sectional analysis, these will be estimated for each day of the week, as well as aggregated over the weekend (Sat-Sun) and weekdays (Mon-Fri). In addition, an alternative definition of weekend will be explored as a sensitivity analysis, using the A&E conclusion date and time, for patients who attend A&E. Due to lack of data for the 10-year period, the time-series analysis will concentrate on weekend (Sat-Sun) vs weekday (Mon-Fri). Risk-standardised mortality rates will be calculated quarterly and annually for the time series analysis. Multivariate logistic regression analysis will be conducted to estimate relationships between provider characteristics, the weekend effect and 30-day mortality (cross-sectional and time-series). The provider phenotypes will be identified using a group-based, semi parametric mixture modelling approach via the SAS macro PROC TRAJ. The determination of phenotypes will depend upon the Bayesian Information Criteria index, average posterior probability of phenotype and 95% confidence intervals of adjacent trajectories. Outputs: Outputs will be released at hospital or trust level. However, the Health Foundation will not identify individual hospitals in our output. The Health Foundation will produce risk-standardised 30-day mortality rates for each day of the week (cross-sectional study only) and for weekdays and weekends, at hospital site and trust level (cross-sectional and time-series analysis). The multivariate logistic regression analysis will produce regression coefficients, associated p-values and odds ratios. These data will help understand the influence and importance that different provider characteristics (e.g. if it is a teaching hospital, which region it belongs to, staffing levels) have on 30-day mortality rates and how these are associated with the observed weekend effect. Odds ratios of 30-day mortality by day of the week will be displayed and illustrated using box plots. Charts mapping the changes over time in 30-day mortality by provider across the study period will be produced. Depending on the number of identified phenotypes, graphics displaying the obtained trajectories will be created (e.g. high weekday mortality and low weekend mortality, or high weekday and low weekend mortality rates). Each provider can be attributed to one of the created graphics – making classification and groupings of trusts easier and more visual. Similar output will be created for the subgroup analyses on patient groups of particular interest: depending on route of admission (through A&E or direct admissions); for emergency surgery, myocardial infarction, stroke, heart failure; and patients with mental ill health. Minimisation efforts related to the inpatient data: Only non-sensitive data relating to patient characteristics, admissions, discharges, episodes and spells, and clinical data are requested. Out of the available socio-economic data, only rural/urban indicator, LSOA and overall IMD information is requested. This information is necessary to identify the patient population of interest, outcomes and patient characteristics such as age, gender, comorbidities and IMD scores, which will be used in the matching process and in the modelling. Also requested is the GP practice code and some organisational variables but these have been restricted to provider type and code and site code. These variables are needed for the trust level analyses. In addition, there is a request for the variable mental category (mentcat), as this variable will help determine, together with the diagnosis codes, the subgroup of patients with mental ill health and severe mental ill health. This application has not requested variables that are sensitive/identifiable, nor other variables that are not necessary to the analysis, such as items relating to Health Research Group, Maternity, Augmented care period, or psychiatry. Time period: The cross-sectional analysis requires data for one year only. However, the calculation of the comorbidity indices used for risk adjustment requires data for up to three years prior to each hospital admission. The time-series analysis requires data over a ten-year period to identify phenotypes. This will ensure an adequate period over which to identify changes in hospital performance (respective mortality rates) over time. In addition, as in the cross-sectional analysis, this analysis requires data over the three years prior to the initial trajectory analysis for the risk adjustment Minimisation efforts related to A&E data: Only non-sensitive data relating to patient characteristics (incl rural/urban indicator, LSOA and overall IMD information), attendances, clinical diagnoses and clinical treatment, some provider information and GP practice code were requested. The A&E attendances will be linked to the ACP admissions, using information on patient characteristics and dates. Information such arrival mode may be relevant to risk adjustment. The A&E discharge time will be used as proxy for admission time.

Objectives:

The Health Foundation is an independent charity working to improve health and the quality of health care in the United Kingdom. The Health Foundation is requesting access to data aiming to inform public discussions about the focus, design and effects of policies intended to improve the quality of health care in the United Kingdom or reduce costs. The project will inform policy makers and the NHS about the variability in quality of health care in England, and thus help to identify priority areas for improving health and social care A “weekend effect” on mortality following emergency admission is well established. However, there is still uncertainty if this effect is due to patient mix or organisational characteristics. This project will be in two phases: Firstly, the Health Foundation will aim to understand the weekend effect and variation across all days of the week, optimising risk adjustment (case-mix), exploring the effect of organisational and staff-related factors and examining variability in the weekend effect between providers (cross-sectional). This analysis will inform a second phase, aiming to identify mortality trends (weekday and weekend) over time, classify English hospitals by observed trajectories and investigate relationships between ‘phenotypes’ and hospital characteristics (time-series). The study follows the ‘phenotyping’ approach used by Xu et al. (2014) for patients undergoing an emergency admission.


Project 3 — DARS-NIC-90019-Q8P9K

Opt outs honoured: N

Sensitive: Non Sensitive

When: 2017/03 — 2017/11.

Repeats: One-Off

Legal basis: Health and Social Care Act 2012

Categories: Anonymised - ICO code compliant

Datasets:

  • Hospital Episode Statistics Admitted Patient Care
  • Hospital Episode Statistics Outpatients
  • Hospital Episode Statistics Accident and Emergency

Benefits:

Benefits In line with the primary objectives described above, the purpose of this project is to build an evidence base on inequality across England. This information will be used by commissioners including NHS England to identify priority areas for reducing inequality, in line with the aforementioned mandate from Government. The overall benefit of this work is that policymakers and commissioners can use the findings in targeting specific cohorts in the population where inequality is particularly high. Existing initiatives aimed at reducing inequality can be assessed, and better evidence will inform debates on inequality more widely. Specifically, objectives (i) and (ii) will inform the varying level of inequality across geographical areas and difference health care services (objective (ii) focussing primarily on the latter). Variation in inequality by health care service will help target commissioners in tackling the problem of inequality. Objective (iii) will model the cost of care for patients that are more or less deprived of access to health care (e.g. the average cost of an outpatient appointment may be higher for hard to reach groups). This will benefit commissioners and policy makers in prioritising this particular policy area. As described in the outputs section, outputs (a) and (d), as well as ongoing interaction with the Department of Health, NHS England and NHS Improvement will deliver the benefit for policy makers and commissioners, and in turn for patients. Outputs (b), (c) and the involvement of patient advocacy groups are targeted towards our objective to inform the public debate about health inequality in England. Note: the purpose of this project is to identify areas for improvement regarding inequality in access to health care services including hospital services, maternity services and access to A&E. Although the Health Foundation can help inform proposals to reduce inequality going forward, this project will not reduce inequality in and of itself. Audience As a non-profit organisation, the Health Foundation’s mission is to maximise the public benefit and the impact of the research that is produced in-house. The aim is to produce useful evidence that can inform better policy and ultimately improve health and health care. This is why The Health Foundation target specific areas of interest for policy and NHS users that are less explored and particularly complex to analyse. This is the case of health inequalities. The Health Foundation has strong links with NHS teams, national policymakers (e.g., NHS England) and patient advocacy groups. Examples of these links for previous projects are: • Senior members of Health Foundation staff regularly meet with senior representatives from across government, including the Treasury, Department of Health and Arms-Length Bodies (e.g. Monitor, CQC, NHS England, HEE). • The Health Foundation is currently working on joint projects with NHS organisations. One example is our partnership with NHS England in evaluating new models of care outlined in the Five Year Forward view (http://www.health.org.uk/programmes/projects/improvement-analytics-unit). • People across the Health Foundation regularly engage with policy makers at all levels on a range of topics where we have particular expertise: policy, data analytics, economics, patient safety and person-centred care. Health Foundation views are regularly sought on health policy and practice, meaning that the findings from these HES-based analyses will be communicated directly with policymakers. One example is the recent engagement of the Economics team with NHS Wales (http://www.health.org.uk/programmes/projects/fiscal-sustainability-nhs-wales), leading to the following report (http://www.health.org.uk/publication/path-sustainability). • The Health Foundation have a long history of funding programmes across the NHS which help to improve the quality of health care. For example, funding work on the relationship between patient flow, costs and outcomes in two NHS hospital trusts, which is related to the new project on understanding the drivers of A&E attendances. • The Health Foundation have an active audience of professionals working in the NHS, many of whom are fellows sponsored by the Health Foundation, award-holders or part of our alumni. Engagement The approach to dissemination includes not only publications but also active engagement with national policy makers, practitioners and researchers. This project will have a member of the Foundation’s Communications team leading on dissemination of findings which will include a number of alternative channels as TV, radio interviews and articles on national, local and online media.

Outputs:

As outlined above, outputs will be aggregated with small numbers supressed in line with the HES analysis guide. Results from the analysis will include: • Summary statistics (including number of observations, mean values and standard deviation values) for inequality regression analyses assessing health inequality in relation to geographical area, deprivation and ethnicity; • Estimated regression coefficients (and their associated p-values, which show the level of statistical significance of the estimated coefficient.) relating to the associations between health inequality and geographical area, deprivation and ethnicity; • Charts and maps to show changes in health inequality by geographical area, deprivation and ethnicity at the Trust or local authority level between 2003/04 and 2016/17; • The results of the regressions included in this analysis will be presented in a tabular format with an accompanying body of text describing and explaining the results. Planned outputs for this body of work are: (a) A Health Foundation report, similar to the recent report on the need for a dedicated transformation fund, Making change possible: a Transformation Fund for the NHS (the Health Foundation and the King’s Fund, 2015) ( https://www.kingsfund.org.uk/press/press-releases/making-change-happen-transformation-fund-nh). This will be accompanied by a detailed communications plan on how to target the media to ensure the results have the maximum penetration. It is expected that this report will be published in the first quarter of 2018 (January – March 2018). (b) Findings will be submitted to international peer-reviewed journals. These will be a mixture of health services research journals (e.g. Health Services Research and Policy) and economics journals (e.g. the Journal of Health Economics). These journals are read by policy makers, nationally and internationally, who wish to identify and classify hospitals according to the level of quality of care that they provide. The expectation is to submit articles for peer review no later than March 2018. (c) Interim findings will be submitted to (and if accepted presented at) various conferences to seek early feedback and to allow improvement of the work. Conferences targeted include: The Health Economist’s Study Group at University of Aberdeen (June 2017) or London City University (January 2018)), NHS Providers annual conference (November 2017), NHS Confed annual conference (June 2017 or 2018) (d) Findings will be presented at The Health Foundation to statutory bodies such as the Department of Health, NHS England and NHS Improvement. The Health Foundation expect to invite the statutory bodies by June 2018. Note: The Health Foundation meets regularly with representatives from the Department of Health, NHS England and NHS Improvement. The proposed work will inform ongoing conversation with these organisations and interim findings will be presented to their representatives in addition to the more formal outputs listed above. Similarly, The Health Foundation will engage with patient advocacy groups in order to ensure findings that may benefit NHS hospital patients, are communicated with these groups who can use them towards positive change.

Processing:

The requested data will be processed by a limited number of analysts within the Health Foundation’s Secure Data Environment (SDE). All researchers with access to the SDE will have completed an accreditation course on data protection legislation and statistical disclosure control, completed an information security training specific to the Health Foundation’s infrastructure, and signed a non-disclosure agreement and the terms of use of the secure environment. All researchers with access to the data are substantive employees of the Health Foundation. The data will only be processed on the Health Foundation’s premises on 90 Long Acre in London and any publication derived from the data will be aggregated with small numbers supressed in line with the NHS guidance before being released from the environment. As outlined in the objectives above, The Health Foundation would like to evaluate health inequality over time, and across different geographies within England. The aim is to investigate how changes in inequality in the most recent decade compares with changes in the preceding decade. There is a particular interest in trends in inequality and access to health care in the period before the current period of austerity and trends during the current period of austerity. For the purpose of this project The Health Foundation will consider the following periods: • 2003/04 until 2009/10 as the pre-austerity period; • 2010, as the start of the austerity; • 2010/11 onwards as the austerity period. For that purpose, the following comprehensive datasets are required: • Inpatients from 2003/04 to date • Outpatients from 2003/04 to date • A&E from 2007/08 to date For this project, The Health Foundation will combine inpatient, outpatient and A&E records over time to establish patients’ health care utilisation. The data will be further enhanced by linking in additional contextual information at GP practice level, or small area level (note: The Health Foundation will NOT link any data at patient level). Contextual data sources are typically publicly available statistics published by the Office for National Statistics, NHS Digital or NHS England (e.g. aggregated census data at small area level, GP patient survey data aggregated at GP level). As stated in the methodology, The Health Foundation wish to examine the differences in the changes in inequality over the period running up to austerity compared to during the current period of austerity. It is essential to this project to analyse data from all available years pre-austerity (from 2003/04) in order to capture all significant changes in health inequality within hospital departments in the period running up to austerity for an accurate comparison with the period of austerity. As A&E data are only available from 2007 onwards, this will restrict the analysis to a 3 year period before the period of austerity. However, The Health Foundation are expecting to complement this trend with underlying trends in hospital admissions and outpatients therefore it is imperative to this project to examine trends in inequality before austerity using the earliest available data for inpatients and outpatients (from 2003/04). The data will be used to generate descriptive statistics on inequality across different health care services, geographies within England, ethnicity and over time in line with objectives (i) and (ii). Measures of inequality that will be used include the slope index and the relative index of inequality. Objective (iii) will require the Health Foundation to apply statistical modelling to the same data, explaining inequality levels in terms of prevalence of health conditions, age, sex, location of residence, ethnicity and time.

Objectives:

The Health Foundation is an independent charity working to improve health and the quality of health care in the United Kingdom. The Health Foundation is requesting access to HES data for the “Assessment of inequality” project to inform public discussions about the focus, design and effects of policies intended to improve the quality of health care in the United Kingdom. The project will inform policymakers and the NHS about the variability in health inequality with respect to NHS hospital services and A&E waiting times in England by geographical area and ethnicity, and thus help to identify priority areas for reducing inequality. The aim of the project is to identify areas (geographical and within treatment specialities) in which healthcare inequality exists. The Health Foundation will disseminate these findings with the aim of raising awareness of existing and growing inequalities in access to healthcare and how these have changed over time and also to inform policy makers of areas in which there is potential for these inequalities to be redressed. The Health Foundation meets regularly with representatives from Department of Health, NHS England and NHS Improvement and the findings will inform ongoing conversations and interim findings will be presented to these representatives. The Health Foundation will also engage with patient advocacy groups to ensure that the findings will benefit NHS hospital patients. Objective: (i) The objective of this project is to create an evidence base that will inform policymakers in the Department of Health, NHS England and NHS Improvement about the variability in health inequality with respect to NHS hospital services and A&E waiting times in England by treatment speciality, geographical area and ethnicity. (ii) The Health Foundation will also assess health inequality within maternity services due to the topical nature of this issue and the high level of media coverage in recent years including issues identified at Morecambe Bay maternity (Bunyan, 2015). (iii) The project will also create an economics model of the determinant of health inequality by geographical area, deprivation and ethnicity. It is recognised in the literature that health inequality exists in England for example Cookson et al. (2016) highlights that residents of more deprived areas are more likely to die from treatable conditions and less likely to see a specialist than residents of less deprived areas. England is not alone in the existence of health inequality; Hart & Williams (2009) also discussed the presence of health inequality in the American setting, linking inequality to quality of life as well as health service factors such as access to care and quality of care. The NHS has a mandate from the government to reduce health inequality (Department of Health, 2015). One reason that this analysis aims to investigate health inequality in England is to assess how the health inequality level has changed in recent years, as the NHS has faced rising financial difficulty, and whether it can be expected to reduce in future years in line with the NHS’ mandate. The Health Foundation would also like to assess health inequality within maternity services due to the topical nature of this issue and the high level of media coverage in recent years including issues identified at Morecambe Bay maternity (Bunyan, 2015). The Health Foundation will use hospital episode statistics (HES) data at pseudonymised patient level to assess the relationship between geographical area and inequality using a measure of inequality such as the slope index of inequality or the relative index of inequality. Trends in inequality will be examined in the years before the current period of austerity (2003-2010) and in the current period of austerity (2010-date). The Health Foundation require a sufficient length of time to reliably compare the changes in the trends during these two periods and examine the significant differences between the intervals. This is needed in order to adequately capture the impact of austerity on inequality in healthcare. The reason The Health Foundation wish to examine these differences is based on the theory that before austerity there will have been more resources available to put towards access to health care and to allocate these resources in the most equal way may have been easier during this period than the current period of austerity. Austerity aims to reduce deficits using methods such as reducing expenditure to bring it in line with revenue. Austerity is often associated with a reduction in government spending; in times of austerity this may result in spending cuts within certain hospital departments. The Health Foundation wish to investigate whether certain groups of the population are disproportionally affected by these cuts in comparison to others.