University of Oxford

Academic Institute

No

DARS-NIC-378657-B8F3K-v0.16

The University of Birmingham is requesting data from NHS Digital in order to help them to determine the impact of hospital-level variation in organisational and clinical approaches to acute care delivery at the hospital/community interface during waves of COVID 19 (e.g., prescribing strategies, staff redeployment, integrated community care planning, etc.) on indicators of healthcare resilience such as (a) operational outcomes (e.g. acute care flow, discharge rates), (b) clinical outcomes for COVID-19 related conditions (e.g. mortality, readmission, rates of pulmonary embolism), and (c) clinical outcomes for non-COVID-19 health conditions (e.g. rates of new onset heart failure, stroke). The data requested will make it possible to study five comparative periods of analysis: (i) Pre-COVID-19, no winter pressures, (ii) Pre-COVID-19, winter pressures, (iii) COVID-19 outbreak peaks, (iv) Post-COVID-19 peaks, and (v) Concurrence of COVID-19 and winter pressures (season 2020-2021). The surge of COVID-19 has had a profound impact on the management and delivery of acute healthcare. To tackle the epidemic, trusts have redesigned organisational models with changes in processes of assessment and care delivery, redeployment of staff, new pathways of care, and different prescribing strategies. These changes have been implemented to provide a rapid increase in acute care assessment and treatment capacity across a system of care for patients with COVID-19-related symptoms, whilst also trying to maintain delivery of care for patients with non-COVID-19 healthcare needs. The purpose of this agreement is to determine the optimal design of the acute care interface with the community, by correlating hospital-level care delivery approaches elicited by the Society for Acute Medicine Benchmarking Audit data (SAMBA) and hospital and patient outcomes from HES data before, during, and after the COVID-19 periods. The data requested will support the achievement of the aim of the project through the construction and analysis of indicators of hospital and healthcare resilience, which is defined as (1) the ability to deliver acute care for COVID-19, and (2) the ability to provide standard care for non-COVID-19-related conditions that can present with acute complications. Examples of resilience indicators include readmission rates, length of stay, mortality, intensive care unit admission rates, number of specialist visits, number of elective and emergency hospital admissions (for COVID-19), rates of heart failure (for non-COVID-19-related conditions). The datasets from NHS Digital will allow the University of Oxford (University of Birmingham's sole Data Processor) to construct indicators of hospital resilience for COVID-19, and non-COVID-19-related conditions that can evolve and develop complications that require acute care (such as, e.g., heart failure, stroke, cancer) by: - Following patients across different types of health services that they use before, during, and after COVID-19 outbreak periods; - Accounting for multiple episodes of hospital attendance/admission and study readmissions for COVID-19 and non-COVID-19-related symptoms; - Estimate out-of-hospital mortality for patients using data from the Civil Registry (Deaths) - Secondary Cut. The requested data (years 2018 to 2021) will allow the data processor (University of Oxford) to study five comparative periods of analysis: (i) Pre-COVID-19, no winter pressures (ii) Pre-COVID-19, winter pressures (iii) COVID-19 outbreak peaks (iv) Post-COVID-19 peaks (v) Possible interactions between COVID-19 and winter pressures (season 2020-2021). Due to the novel setting and disease that this project studies, and the as yet unknown COVID-19 and non-COVID-19-related medical complications that the current pandemic may cause, there is a major exploratory element to this study. The uncertainty related to the object of investigation requires access to multiple sources of data such as HES critical care, A&E, Outpatients and Inpatients, emergency care (ECDS), as well as the civil registry of deaths (secondary cut). The data processor, the University of Oxford, will use the pseudonymised code provided by NHS Digital to follow patients across the different NHS Digital products that are requested in this agreement. The University of Oxford will use the hospital code in HES to complement the analysis with information at the hospital and catchment area level from publicly available datasets and the Society for Acute Medicine's SAMBA survey of practice, which provides information regarding the size and staffing organisation of each acute medical department in the UK, alongside strategies for care delivery as well as methods of interaction with community care providers. In particular, the project will use SAMBA data from the 2018 and 2019 Winter version, and the 2020 COVID-SAMBA survey. The Society for Acute Medicine (SAM) is the national representative organisation for acute health care staff. Formed in 2000, the Society now has over 1000 affiliates, the majority of which are doctors training or specialising in acute medicine. SAM delivers annual SAMBA audits to assess acute medicine approaches and the sharing of good practices. These are England-wide surveys at the hospital level and, in the UK, they are recognised by the Healthcare Quality Improvement Partnership. The study that is subject of this agreement is part of a broader project, which has three operational tiers: (i) The first part includes literature reviews, engagement with stakeholders and a survey of healthcare delivery practices of UK acute medicine units at the hospital level, based on the Society for Acute Medicine Benchmarking Audit (SAMBA). This part of the project will not use NHS Digital data. The Principle Investigator (PI) of the overall project is an active member of the SAM (Society for Acute Medicine) network, has delivered three previous national surveys through the SAMBA network, and has published peer reviewed papers analysing key points from previous audits. (ii) The second part of the project is the empirical analysis of hospital resilience based on the NHS Digital data that the University of Birmingham (Data Controller) is requesting in this agreement. This part will rely on developing quantitative econometric analyses of indicators of healthcare resilience for COVID-19 and non-COVID-19 diseases with acute complications constructed from the HES data. Examples of indicators of healthcare resilience include mortality rates, readmission rates, rates of pulmonary embolism, average length of stay in intensive care units, rates of new onset heart failure or stroke, rates of A&E attendances and emergency admissions for heart attack and stroke/transient ischaemic attack (TI), during and after the first COVID-19 wave. Hospitals will be grouped by common approaches to organisation of care from the SAMBA survey (see (i) above). The trust/hospital/deliverer-level variables that describe care delivery approaches elicited from SAMBA will constitute the main explanatory variables. The analyses will control also for patients' demographics and comorbidities, and data on pre-COVID-19 organisational practices and healthcare needs of the patients and of the population in the trust/hospital/deliverer catchment area. The analysis will deliver aggregate-level results that do not identify individuals, and the publications will not identify hospitals. All outputs will be aggregated with small numbers suppressed in line with the HES analysis guide. (iii) The third part of the overall project will develop a qualitative study to learn about healthcare seeking behaviour among patients with non-COVID-19 severe disease. For example, this part of the project will develop focus groups to understand the reasons behind the reorganisation/postponement and delay of diagnoses (e.g., for cancer-related screenings and the screening and treatment of heart failure). This final part of the project will also include qualitative work based on site visits (or remote interviews) in well performing systems of care, to understand how novel structures and organisational contexts were successfully implemented and embedded. This part of the study will not involve analyses of NHS Digital data nor any linkage to NHS Digital data. The analysis will control for underlying health conditions, healthcare needs, and characteristics of the population in acute care units and in their catchment areas. Information on different care delivery approaches at the hospital level will be elicited from a national survey of organisation and delivery of acute care, the Society for Acute Medicine Benchmarking Audit (SAMBA). The SAMBA dataset is described below. Importantly, SAMBA contains information at the hospital level and does not entail patient-level linkages. The findings from this programme of research will enable policy makers within the Department of Health and Social Care and NHS England to determine how best hospitals and community systems should organise and deliver care during and after waves of COVID-19. The GDPR legal basis for processing data for this research comes under Article 6(1)E – 'task in the public interest' The data processing will provide evidence to help (a) policymakers make evidence-based policy decisions, (b) hospital managers to develop evidence-based decisions on the organisation of acute medical services, and (c) acute care clinicians to understand which practices have improved the resilience of acute care services. The public interest that justifies the processing of this data relates to the improvements that can be made to health-care provision within the NHS as a result of the findings. The University of Birmingham is proposing to process data under point (j) of Article 9(2). The 2020 version of SAMBA for COVID-19 (COVID-SAMBA) collects hospital-level information about variations in organisational and care delivery approaches during the COVID-19 outbreak, such as the degree of integration across acute/community healthcare providers (e.g., discussion of guidelines and common planning for the referral and management of patients with ambulance services, primary care providers, and care homes), novel care pathways (e.g. prescription and patient screening strategies, staff redeployment), and novel structures/systems of care (e.g. home-based hospitalisation). With regards to the analysis for which the University of Birmingham is requesting access to NHS Digital data, the SAMBA surveys will provide information at the hospital level on care delivery approaches. SAMBA data will be linked to HES data using hospital site codes and not at the patient level. After the onset of the current pandemic, the Department for Health and Social Care (DHSC) asked the project team to analyse pressures as a consequence of COVID-19, as this is an overwhelming national priority in acute care. In particular, following the research focus commissioned by the DHSC, this project defines the concept of hospital resilience as the ability to meet the acute healthcare needs of the population during COVID-19. The researchers will assess which organisational and care delivery practices are associated with improved healthcare delivery performance. Other parts of the study, which are not based on the requested data and do not include data analysis, involve literature reviews, consultations with stakeholders, health professionals and patients, and qualitative work in a sample of acute hospitals. The University of Birmingham’s data processor, the University of Oxford, will process the NHS Digital data received and conduct a quantitative analysis for this project. The University of Birmingham holds the main NIHR research contract for the overall study and has entered into an honorary contract with the Chief Investigator (CI). The CI will formulate hypotheses to be tested and help to interpret the findings of the overall study. Hence, the University of Birmingham is the Data Controller. It will not, however, be involved in processing the data. In its capacity as the University of Birmingham’s data processor, the University Oxford team will hold and process the NHS Digital data. The University of Birmingham are determining the means and purpose of the processing of the personal data and the University of Oxford are providing their expertise as the data processor but have no role in determining the means and purpose of the processing. The University of Warwick, where the CI now resides, will not be involved in any decisions about the data nor the analysis of the data. The University of Leicester employs the researchers undertaking the qualitative component of the wider study, that is the third part of the study as described above. The University of Leicester team will not access, process nor control the data. Department for Health and Social Care (DHSC) has no role in the conduct of the study. It is providing the funding (through the NIHR) and will receive the outputs. It is not involved in deciding which analyses should or should not be conducted. The overall project, including the collection of SAMBA data and the cross-mapping of SAMBA with HES data, received ethical approval. Since the study is not an evaluation of a specific intervention, the research approach is not based on a distinction between treated and control groups. Rather, the empirical design relies on correlations between hospital-level care delivery approaches and health outcomes. More specifically, the analysis will correlate indicators based on patient clinical information, mortality data from civil death registry with hospital-level indicators that identify relevant elements in the organisation of acute care delivery during and after COVID-19 outbreaks, elicited from the SAMBA survey. As the data processor on behalf of the University of Birmingham, Oxford will inform the analysis using data from all attendances at A&E specialist or outpatient clinics or admissions between January 2018 and September 2021. This project requires information on all patients attending/admitted to the hospital, with information on the referral status, the cause of attendance/admission, inpatient/outpatient visits and outcomes, the length of stay, and the clinical health outcome for each episode/service use. The analysis will be conducted with pseudonymised data and no individual patient data will be released. The purpose of this project is to understand which acute care delivery approaches developed and implemented before, during and after COVID-19 outbreaks translate into better acute care and health outcomes for the population, and to identify the practices best able to make hospitals more resilient when there is an outbreak of a disease such as COVID-19 or the winter flu. Combining SAMBA and HES data will allow the data processor, the University of Oxford, to achieve this aim. While SAMBA contains all the information relating to the processes of care implemented by English hospitals, HES data make it possible to investigate how these processes affect patients and hospitals. This project requires data from the following data sets: Emergency Care Data Set (ECDS) Hospital Episode Statistics Accident and Emergency (HESA&E), non sensitive data Hospital Episode Statistics Admitted Patient Care (HESAP), non sensitive data Hospital Episode Statistics Critical Care (HESCC), non sensitive data Hospital Episode Statistics Outpatients (HESO) HES: Civil Registration (Deaths) - Secondary Care Cut link The ECDS data requested is not currently within the TRE dat offering and thus this request can not at this point in time be fulfilled by the NHS Digital TRE service. Using the pseudonymised identifiers provided by NHS Digital to bridge the products requested, the data analysis will connect patient's admission episodes across the HES products (inpatient, outpatient, critical care) and with (i) readmissions, and (ii) out-of hospital mortality (through the Death Civil Registry). As the data processor, the University of Oxford will analyse this information also in conjunction with hospital-level care delivery approaches from the Survey of Acute Medicine Benchmark Audit (COVID-SAMBA and 2019, 2020 SAMBA - please see point 4 of this section and the attached documents for a description) and with aggregate metrics of general health and population characteristics in the acute department's catchment area from publicly available data sources. Due to the as yet unexplored and as yet unknown context of a novel disease outbreak, and because this project studies how COVID-19-related care as well as non-COVID-19-related conditions relate to different healthcare provision approaches, information on all symptoms and causes of hospital admission is necessary. As features of acute illness are often non-specific (e.g. confusion, generalised functional decline, reduced mobility among older adults), the project requires all available health information without restriction to specific conditions. In addition, there is no guidance yet as to which groups of patients have had fewer admissions due to COVID-19 and its overall effect on hospitals' ability to deliver care: therefore looking at all hospital admissions is the most inclusive and correct approach. Hospital Episode Statistics Accident and Emergency data will allow the data processor, the University of Oxford, to identify whether patients that attend A&E are discharged or admitted, and to classify the cause of attendance (COVID or non-COVID related). HES A&E (and the ECDS, once a code will be developed), HES-Outpatient, HES-Inpatient, HES-Critical Care will make it possible to: - Follow patients that attend A&E/ the hospital/ trust in the subsequent stage (i.e., inpatient / outpatient / discharged), record their process of admission and outcome (e.g., length of stay and clinical health outcome); - Control for the utilisation of primary care before and after an acute illness that requires A&E attendance or inpatient/outpatient admission; - Construct and correlate indicators of acute care resilience with organisational changes and care delivery practices during and after COVID-19 outbreaks (from the SAMBA hospital-level data). Civil Registration Deaths - Secondary Cut will allow the data processor at the University of Oxford to link attending/admitted patients with out-of-hospital mortality outcomes. In particular, the University of Birmingham is requesting the following groups of variables: - Admissions - Period of care (e.g., method, source, date, waiting time) to control for different circumstances and procedures of admission in the analysis of the correlation between hospital-level acute care delivery approaches and average health outcomes, and group patients’ health outcomes by heterogeneous characteristics; - Augmented/critical care period variables, with information such as time, period, outcome, source, discharge, status, intensive care, high dependency of patient’s admission episodes, to construct outcomes for the analysis (e.g., average time in intensive care, mortality, probability of high dependency case), controlling for further clinical and admission characteristics; - Clinical information with date of operation, cause of admission, primary and additional diagnosis codes, operation status, and durations, to control for these elements in the analysis, construct health outcomes by specific circumstances/causes/etc. of admission, and duration of the episode(s); - Clinical information regarding patient classification and consultant/treatment specialty, and Practitioner/Referring organisation codes, to categorise patients’ health outcomes according to specific treatment groups either by own classification or consultant specialty or practitioner; - Diagnosis codes and Alcohol Attributable Fraction; - Discharge dates and methods (and flags), to control for length of admissions and cross-validate precision of the duration, and study time lags between readiness for discharge and actual discharge, and their trends before, during, and after peaks of acute care activity; - Episodes and spells (Period of care) data, such as dates, durations, types, ward types, and Patient Pathway information, to form groups of similar episodes and to control for such characteristics in the analysis of the correlation between care delivery approaches and health, mortality, and readmission outcomes; - Geographical codes (e.g., CCG, area, region, site code of GP practice, treatment, residence areas, ONS electoral ward codes), Healthcare resource groups (HRG), Organisation codes/information, and Socio-economic indicators (location-based IMD indexes), to control for/group health outcomes by locations, and associate health outcomes to other local-level information from publicly available data at the trust/catchment area level; - Patient demographic data, to group patients by categories or control for patient characteristics in the empirical analysis of the correlation between hospital-level care delivery approaches and indicators of health care resilience from patient health outcomes; - System Data to verify validity of assignment of patient/CDS/SUS codes. The specification of a COVID-19 diagnosis for patients will be based on the ICD-10 code. This project only requires pseudonymised data, because the analysis will follow patients in the different services/units. The analysis requires patient level records to analyse health outcomes by different patterns of use of the healthcare services and to be able to group/control for demographic characteristics, waiting times, diagnosis, procedures, etc. Furthermore, patient record data will allow the empirical estimations to follow patients/episodes of care across the various NHS Digital products such as, e.g., deaths registry data, outpatients, etc., to measure healthcare outcomes, before, during and after the pandemic. The University of Birmingham does not request any identifiable or 'high risk' variable, and the estimation outcomes and findings of this project will be produced solely in aggregate form, with small numbers suppressed in line with the HES analysis guide. Nonetheless, the results of the quantitative analyses will only be included in the study outputs and communicated at an aggregate level. There will be no way to identify individual or critically small/selected groups of people from the results of the study all outputs will be aggregated in line with the HES analysis guide. The estimations will only deliver coefficients of correlation between care delivery practices and aggregate categories of health outcomes and indicators (e.g., total A&E admissions, mortality rate, total admissions in cardiology, ICU admissions, average length of stay by non-identifiable demographic characteristics such as age groups). This data request is limited to the years between 2018 and 2021 inclusive. This will allow the University of Birmingham to study five comparative periods of analysis: Pre-COVID-19, no winter pressures (2018-2019, spring-summer) Pre-COVID-19, winter pressures (2018-2019, winter) COVID-19 outbreak peaks (2019-2020 winter and spring) Post-COVID-19 peaks (e.g., July-August 2020) Possible interactions between COVID-19 and winter pressures in the winter season of 2020-2021. The quantitative analysis will compare the outcomes of patients in different hospitals and acute care units across England and, as such, it needs data concerning all English hospitals. There exists no possibility other than via HES to construct and analyse variables that are based on following patients across different units of care, multiple episodes of admission, and out-of-hospital mortality at the hospital/acute care unit aggregate level. This project requires patient-level information also to account for patients' demographic characteristics, and prevalence of as yet not know preconditions and co-morbidities in the reference population that may contribute to determining the success and failure of hospital/acute care unit care delivery organisational approaches and practices in terms of both COVID-19 and non-COVID-19 related care. The University of Birmingham has minimised the request in the time dimension. In particular, the required data is limited to the years between 2018 and 2021, ending with the release of September 2021. Due to the exploratory nature of the project and as yet unknown consequences of COVID-19 and care delivery approaches during the current pandemic, the request is not restricted to specific health conditions and causes of admission. The aim of this proposal makes it necessary to request and explore individual-level data because this study is the first of its kind, and the context of the COVID-19 pandemic is as yet unexplored. This analysis will request and explore all possible conditions, causes of admission, and demographic characteristics. It is not possible to pre-aggregate and request health outcomes at the hospital level. This is required to understand the pathways of each individual in the use of the health system, in response to the COVID-19 pandemic, and to group outcomes by (or control for) demographic characteristics, waiting times, diagnosis, and procedures in the analysis. The ethnic category variable is requested because there is evidence that people from BAME communities are the most affected by the COVID-19 pandemic and the analysis needs to control for this factor. This project requires only pseudonymised data and the request does not include any identifiable or 'high risk' variable. The results of the quantitative analyses will only be communicated and included in the study outputs at an aggregate level, further suppressing critically small/selected groups of people. The request is further minimised by excluding data concerning maternity and psychiatry. The University of Birmingham is the sole Data Controller. University of Birmingham are determining the means and purpose of the processing of the personal data and the University of Oxford are providing their expertise as the data processor but have no role in determining the means and purpose of the processing. The University of Oxford operates under specific protocols for processing of data directed by the University of Birmingham. The University of Warwick is not carrying out joint data controllership activities, in light of the Chief Investigator holding an honorary contract with the University of Birmingham, but being a substantive employee of the University of Warwick. The University of Birmingham will remain the only Data Controller, according to its original contract with DSHC and NIHR . University of Leicester employs the researchers undertaking the qualitative component of the wider study. They are not involved in the NHS Digital data processing. The Department of Health and Social Care (DHSC) is the commissioner of this project. DHSC has no direct influence over the analysis performed and will have access to a final report of the findings but not the data used. The project funder is National Institute fir Health Research (NIHR).

16/07/2021

De-Personalised

Health and Social Care Act 2012 - s261 - 'Other dissemination of information'

Not applicable

Recurring

TRE