Investigating COVID-19

Text removed is struck through; text added is underlined. Unchanged paragraphs are summarised rather than repeated.

Processing activities

[7 paragraphs unchanged] The Integrated Data Service (IDS) Google Cloud Platform Analytical Platform (GCP AP) is a cloud-based Trusted Research Environment (TRE) where analysts process data for the purposes listed in this Agreement. Google UK Limited provide IT hosting services to ONS for the IDS GCP AP and will store the data as contracted by ONS as a processor. The Secure Research Service (SRS) is hosted on the iTS computing Ltd Cloud Ark Data Centres platform. The servers used to store data and to host the analysis [6 words unchanged] and National Cyber Security Centre (NCSC) Accredited (PGA) data centre, provided by iTS computing Ltd Ark Data Centres and based in the United Kingdom only. [1 paragraph unchanged] The Data will reside within Google Cloud Platform (GCP). All cloud services [104 words unchanged] GCP platform is region locked to UK IP addresses only. Access to IDS GCP AP and any data it holds is not permitted from outside the UK. Overseas connections are monitored, and connection attempts will lead to account suspension. [30 paragraphs unchanged]

Benefits reported

ONS has worked on a range of topics and outputs in line with this Agreement. These statistics have been used to inform government policy and health campaign decisions, impacts include: To date, ONS have undertaken a project investigating the impact of cardiometabolic multiple long term conditions on economic activity and employment outcomes during and after the pandemic. This has contributed to better understanding whether certain disadvantaged groups and people living with severe morbidity (i.e., disability) have been more impacted by economic inactivity and related employment outcomes on a national scale. This has also helped identify and characterise disadvantaged subgroups; quantify vulnerabilities with employment outcomes; and better monitor potentially widening inequalities within these groups. This work utilises Census 2021, HES, GDPPR and death registration data. This work is to be published as an ONS publication. Ethnic differences in COVID-19 mortality: All analyses have been conducted in line with ONS governance frameworks, ensuring that linked health and census data are used solely for the production of aggregate statistics. Outputs are anonymised and designed to inform public understanding and policy, without enabling identification of individuals or operational use by external departments. This approach addresses public concerns around data linkage while maintaining analytical integrity and independence. The ONS has used the PHDA to highlight Ethnic differences in COVID-19 mortality and help understand the drivers. Numerous ONS bulletin and articles were published in peer-reviewed journals. The results have been presented multiple times at Scientific Advisory Group for Emergencies (SAGE), including in a very recent meeting. ONS has actively contributed to the SAGE Ethnicity subgroup, and data from the PHDA were used to inform policy papers. The initial analysis, which was subsequently published in the International Journal of Epidemiology, has been widely cited in academic research, including regularly in the British Medical Journal (BMJ). Professor Peter Goldblatt, University College London Institute for Health Equity, described the analysis as “a milestone in ethnicity analysis”. It has provided the baseline for further research into the social, clinical and biological reasons for the differences in COVID-19 mortality. It was awarded the RSS Campion award for excellence in official statistics. All subsequent analyses generated high media attention and were published in internationally renowned journals. For future use: ONS has also published initial findings on the ethnicity data quality work described above. In order to better understand the quality of ethnicity coding in electronic health records widely used to estimate COVID-19 mortality rates, ONS has published person-level comparisons of ethnicity recorded in three key health data sources (HES, GDPPR) compared with ethnicity information recorded in Census data (which is widely regarded as the most robust source of ethnicity data). This analysis showed the consistency of ethnicity recording compared with Census varied substantially for different ethnic groups – and was generally lower for those in the Mixed and ‘other’ ethnic groups. This analysis lays the foundations for future work which will look at solutions and methods analysts can use to produce more reliable estimates despite the differences between sources. The dataset/analysis will continue to be used to expand and deepen analysis of the relationship between long-term health conditions, multimorbidity, and employment outcomes, building on the original pandemic-focused remit and assess and monitor any widening inequalities. It may also support further projects under other approved Data Sharing Agreements. It may also be used for responses to the COVID inquiry on this topic, where required. Once the project is complete and publications are finalised, data will not be kept for longer than necessary and will be destroyed in keeping with data sharing agreement. The final publication was published in May 2024 for the project which assessed the quality of NHS ethnicity information for NHS England: https://www.ons.gov.uk/peoplepopulationandcommunity/healthandsocialcare/healthinequalities/articles/understandingconsistencyofethnicitydatarecordedinhealthrelatedadministrativedatasetsinengland2011to2021/may2024 Qcovid validation: ONS was commissioned by the CMO to conduct the validation of the QCovid risk model - an algorithm predicting the risk of COVID-19 death based on clinical records. The validation was conducted at pace and the results published in the Lancet Digital Health. As as a result, the algorithm was successfully used to update the shielding list and inform the prioritisation of the vaccination campaign. The overall project won the RSS Nightingale award COVID-19 mortality and vaccination: ONS used the PHDA to analyse the links between COVID-19 vaccination and COVID-19 mortality. First, ONS conducted a study on vaccine effectiveness (VE) against mortality using a regression discontinuity design. The study was praised by Sir David Spiegelhalter as a ‘clever application of an innovative method’ and was used to assess the potential bias in the regular VE estimates produced by the UKHSA. ONS also used the PHDA to monitor COVID-19 vaccination rates by vaccination status. The findings were widely reported in the media and used by the Government and the NHS to promote the vaccination campaign. ONS has also published experimental statistics relating to the risk of hospital admission for coronavirus (COVID-19) and death involving COVID-19 by vaccination status by age group. This work is also currently undergoing peer-review for publication in the International Journal of Epidemiology. Media coverage: This recent piece with the National Statistician gives an overview of ONS’s role in the pandemic, although the specifics do not include much on linked data analysis: https://www.newstatesman.com/encounter/2022/02/counting-the-cost-of-covid-with-ian-diamond Since the start of 2024, the ongoing work is investigating the economic impact of the pandemic on disadvantaged groups and involves linking GDPPR and other health data with benefits and income data. The following outputs and benefits have been generated to date: https://blog.ons.gov.uk/2023/12/05/using-the-power-of-linked-data-to-understand-factors-preventing-people-from-working/ https://www.ons.gov.uk/employmentandlabourmarket/peoplenotinwork/outofworkbenefits/bulletins/benefitrecipientsduringthecoronaviruscovid19pandemicengland/november2019tomarch2021 ONS has also conducted an observational, retrospective, matched cohort study of individuals admitted to hospital with covid-19. ONS used the Hospital Episode Statistics Admitted Patient Care records for England up to 31 August 2020 and the General Practice Extraction Service Data for Pandemic Planning and Research (GDPPR) up to 30 September 2020. GDPPR is an extract of primary care records collected from surgeries by NHS Digital for pandemic research and analysis (supported by the British Medical Association and the Royal College of General Practitioners), including over 56 million individuals registered at NHS England general practice surgeries and updated fortnightly. The extract includes a subset of about 35 000 clinical codes, selected for potential use in pandemic related analysis. Death registrations from the Office for National Statistics were linked for deaths up to 30 September 2020 and registered by 7 October 2020.

Unchanged: Objective for processing, Expected output, Expected measurable benefits.

Text removed is struck through; text added is underlined. Unchanged paragraphs are summarised rather than repeated.

Datasets: + Improving Access to Psychological Therapies (IAPT) v2

Objective for processing

The Office for National Statistics (ONS) requires access to NHS England Data for the purpose of COVID-related analytical purposes. Processing of NHS England data will fall under one analytical purposes outlined below: in line with functions set out in the Statistics and Registration Service Act 2007 • Support the production of statistics on the short, medium and long-term impacts of having had COVID-19, treatment for COVID-19 or a COVID-19 Vaccine – uses include statistics on the direct impact of COVID-19 infection on mortality (acute infection) and morbidity (COVID-19 sequelea and so called long-COVID), and how these are mediated by factors such as personal characteristics (e.g. ethnicity, age, sex), vaccination status and pre-existing conditions. ONS are permitted to process Data for the following purposes: • Support production of statistics on the coronavirus pandemic and the associated social, economic and environmental impact on health and well-being - uses include statistics on the indirect impact of COVID-19 on mortality (e.g. non-COVID-19 excess mortality), morbidity (e.g. mental health), impact on vulnerable and disadvantaged groups and economic and social wellbeing. 1) To fulfil the purpose under “IMPACT OF THE PANDEMIC ON DISADVANTAGED GROUPS” • Supporting numerical analysis on the quality of the underlying Data itself and statistics produced in the purposes mentioned, so that the Data can be used appropriately by others, and so that the statistics can be interpreted appropriately. ONS has been asked by DHSC to investigate the impact of the pandemic on disadvantaged groups, notably the disabled, by taking into account employment and benefits outcomes in addition to health outcomes. For example, is there evidence of increased rates of unemployment or benefits uptake in certain groups, with a consequence potentially being widening inequalities and how does this interact with health events; for example, evidence of negative economic outcomes following infection or hospitalization from COVID-19? This DSA approves use of the Data listed for the production of statistics in the three purposes described. In doing so, the following criteria and conditions will apply: The DHSC/DWP joint work health analysis unit are stakeholders in this area of the programme. ONS are aware of possible public concern in this area – i.e. the linkage of health and economic data. Therefore, when health data are linked to benefits an income data, they will only be used to produce aggregate statistics (in this case, on impact and inequalities) by substantive ONS staff. They will never be used for operational purposes, and no DWP or DHSC officials can access the linked data and ONS retains independence and control over what analyses it conducts. A separate Data Sharing Agreement between ONS and DWP covers the use of benefits and income data for the production of official statistics. • The statistics produced will meet a definitive user need or driven by commissions from the Scientific Advisory Group for Emergencies (SAGE), the Chief Medical Officer (CMO), the British Government or the Department for Health and Social Care (DHSC). These commissions are received via the National Statistician or the Director of Health and Pandemic Insight in ONS, working on his behalf; 2) To respond to queries from the COVID inquiry • The statistics produced will be released as publicly available statistics on the ONS website in line with the UK code or practice for statistics; 3) To reuse HES Critical Care Data and GDPPR Data supplied under this Data Sharing Agreement (DSA) where these Data refer to and are required for purposes approved under other DSA's. • The Data will never be used for operational purposes that could affect an individual – ONS is only interested in processing and linking the Data (even if this is linking at an individual level) to then enable the production of aggregate statistics that describe broad patterns in the data, therefore providing insight on what is happening in the real world; • These aggregate statistics will be disclosure controlled and will never identify an individual; • The number of ONS staff who access identifying data will be kept to the minimum necessary and there will be a separation of duties between those using identifying information for linkage purposes and those doing the analysis based on integrated attribute data after the identifiers have been removed; • ONS analysts will never seek to reidentify an individual and it is worth noting that ONS analysts do not actually scrutinize individual rows of data; rather they run models on large scale datasets with millions of rows and then scrutinize the outputs and aggregate results; • No other organisation will process the Data under this DSA. The Data shared with ONS will not be onwardly disseminated or shared under this DSA except as disclosure controlled aggregate statistics and/or analysis as aggregated data with small numbers suppressed, in line with the Hospital Episode Statistics Analysis Guide; • The Data will be used only for purposes which have approval from the National Statistician’s Data Ethics Advisory Committee (NSDEC). NSDEC’s role is to scrutinize and where necessary challenge, the ethical basis for the work as new statistics are proposed. PRE-APPROVAL PROCESS FOR NEW USES: In the absence of amending this DSA ONS will complete the following steps for any new use of the Data. Any new use of the Data will be in support of one of the purposes outlined above. Before processing begins, ONS will:: i. Submit a briefing document to NHS England specifying: the new uses; the expected outputs, dissemination plans and expected benefits; what datasets and what subsets thereof will be used; and any non-NHS England datasets with which the Data will be linked. ii. Provide evidence of approval for each new use by the NSDEC. iii. Not use the Data for the proposed purpose before approval is confirmed in writing by the Data Access Request Service (DARS) team in NHS England after the above conditions are met. iv. If the new use involves either the GPES Data for Pandemic Planning and research (GDPPR) or the COVID-19 Ethnic Category Data, the briefing will be provided to Chair of NHS England’s Profession Advisory Group (PAG) and NHS England’s Advisory Group for Data (AGD), and NHS England approval will be subject to taking advice from AGD and the PAG. It will not be necessary to amend this DSA before the Data is used for the new use. [1 paragraph unchanged] • Hospital Episode Statistics (HES) Admitted Patient Care, Accident & Emergency and Outpatients – necessary because the HES data provides historical information about comorbidities relevant to negative outcomes from COVID-19, which is needed for risk factor modelling, and also provides information about the cause of secondary care contact during the pandemic (for example, being hospitalized because of acute COVID-19 infection). • Hospital Episode Statistics (HES) Admitted Patient Care, • Hospital Episode Statistics (HES) Critical Care – necessary because it allows ONS to stratify hospitalisations according to severity (ICU vs. non-ICU admission), which will be useful when defining pre-existing conditions, COVID-19 exposure, and outcomes following hospitalisation. ONS used this grading of severity in their long-COVID publication but it is not possible to replicate this using HES alone. These Data allow ONS to take into account the type and level of intensive care support patients were given. Hospital Episode Statistics (HES) Accident & Emergency and Outpatients • Emergency Care Data Set (ECDS) – necessary because HES A&E Data has been discontinued and these data have been replaced by the more comprehensive ECDS. The ECDS Data will be used for the same reasons as HES A&E described above. • Hospital Episode Statistics (HES) Critical Care • Personal Demographic Service – necessary to: • Emergency Care Data Set (ECDS) o Facilitate linkage of non-NHS England data to NHS England Data by using personal identifiers to match records in the Census, CIS or NPD populations at risk with their NHS number. This enables onward linkage to other sources where NHS number is the unique identifier on those other sources; • Personal Demographic Service o Ensure each data asset includes the most up to date residential location for the subjects included (especially the Census based asset given the Census 2011 data is 10 years out of date); in turn, this location information can be used to assign the most up to date location-based indicators that are relevant to the statistics being produced; for example: • Improving Access to Psychological Therapy (IAPT)  Determining if the current location a care home. • Birth Notifications  Determining the current Lower Layer Super Output Area of residence (and therefore deprivation score of area of residence). • General Practice Extraction Service (GPES) Data for Pandemic Planning and research (GDPPR)  Linking members of household together which is important for taking household transmission into account. • COVID-19 Ethnic Category Data • Improving Access to Psychological Therapy (IAPT) – necessary because it provides information about contact with such services and helps enable statistics on the indirect mental health impact of the pandemic. The ethnicity data included in this dataset will also help enable the quality assurance of ethnicity data on behalf of NHS England. • Birth Notifications – necessary because it includes NHS number of the mother and baby so will allow identification of the fact of a birth and allow ONS to identify babies in wider NHS England datasets. • General Practice Extraction Service (GPES) Data for Pandemic Planning and research (GDPPR) including the 'POSSPOSTCOVID_COD' cluster and the vaccinations and immunisations cluster – necessary because they provide historical information about pre-existing conditions relevant to negative outcomes from COVID-19 (acute or sequelae). The Data also provide information about diagnoses made during the pandemic relevant to indirect impacts; for example, mental health diagnoses. • COVID-19 Ethnic Category Data – necessary to assess the quality of ethnicity data in this dataset. [1 paragraph unchanged] To address the UK GDPR Principle of data minimisation, ONS has reviewed the data covered by this Agreement and confirmed that all variables listed are necessary for achieving the purposes listed. In some cases, this review has been done in conjunction with NHS England. Though the work will be driven by commissions from the Scientific Advisory Group for Emergencies (SAGE), the Chief Medical Officer (CMO), the British Government or the Department for Health and Social Care (DHSC), ONS will be solely responsible for determining what data will be processed and the manner of its processing in order to undertake the work. The work is self-funded by the ONS. Generally full datasets are required (as opposed to a subset, random sample, or aggregate data) to ensure quality statistics can be produced. More information on statistical quality can be accessed on the ONS website but the following is a short summary: 'The quality of a statistical product can be defined as the "fitness for purpose" of that product. More specifically, it is the fitness for purpose with regards to the European Statistical System dimensions of quality: • relevance - is the degree to which a statistical product meets user needs in terms of content and coverage? • accuracy and reliability - how close is the estimated value in the output is to the true result? • Timeliness and punctuality - describes the time between the date of publication and the date to which the data refers, and the time between the actual publication and the planned publication of a statistic • accessibility and clarity - is the ease with which users can access data, and the quality and sufficiency of metadata, illustrations and accompanying advice? • coherence and comparability - is the degree to which data derived from different sources or methods, but that refers to the same topic, is similar, and the degree to which data can be compared over time and domain, for example, geographic level? There are additional characteristics that should be considered when thinking about quality. These include output quality trade-offs, user needs and perceptions, performance cost and respondent burden, and confidentiality, transparency and security.' It is crucial that the chances of drawing incorrect conclusions are kept to an absolute minimum, especially when the statistics will guide decision making in a pandemic. It is also crucial that accurate statistics can be produced quickly. This could save lives. To receive less data than have been requested, or to use aggregate data, would lead to less accurate statistics with greater uncertainty. It would also take longer to produce the statistics and therefore negatively affect timeliness because methods would need to be used to account for the missing data and/or greater limitations of the data compared with ONS accessing everything that has been requested. ONS is the controller as the organisation responsible for ensuring that the Data will only be processed for the purpose described above. Though the work will be driven by commissions from the Scientific Advisory Group for Emergencies (SAGE), the Chief Medical Officer (CMO), the British Government or the Department for Health and Social Care (DHSC), ONS will be solely responsible for determining what data will be processed and the manner of its processing in order to undertake the work. [4 paragraphs unchanged] All purposes are tasks in the public interest. ONS is the sole controller for the work described above. The work is self-funded by the ONS. Crown Hosting Data Centres Limited provides IT hosting services to ONS and will store the Data as contracted by ONS. This DSA came into force mid-pandemic in response to the increased demand on ONS for rapid response analysis. Previous versions of this DSA did include specific descriptions permissions for each statistical use. Therefore, at the time of the DSA coming into force, much is already known about how the data have been linked and what statistics have been produced or will be produced. As such, in the interests of transparency, ONS details examples in the remainder of this section. Note: these are not exhaustive and the Data may be put to additional uses beyond those described. 1. Support the production of statistics on the short, medium and long-term impacts of having had COVID-19, treatment for COVID-19 or a COVID-19 Vaccine. This includes statistics on the direct impact of COVID-19 infection on mortality (acute infection) and morbidity (COVID-19 sequelea and so called long-COVID), and how these are mediated by factors such as personal characteristics (e.g. ethnicity, age, sex), vaccination status and pre-existing conditions: 1.1 COVID-19 RISK FACTORS AND MORTALITY: This work aims to determine the population-level relative risk of hospitalisation and death that COVID-19 presents to different people. Models have been published on mortality risk involving COVID-19 by ethnic group, religion and disability, controlling for comorbidities as derived from the HES and GP data. This work will continue to be updated and developed. 1.2 LONG-COVID: ONS has been tasked to continue analysis around the prevalence of, and risk factors associated with 'long-COVID' i.e. symptoms and conditions that persist/develop beyond the acute phase. This analysis is intended to inform public health messaging and interventions relating to long-COVID. 1.3 VACCINATION: ONS already receive limited, daily data on COVID-19 vaccinations from NHS England, and are now requiring monthly disseminations of the higher quality Vaccination status data from NHS England (ref: DARS-NIC-175120-W5G2X). This allows ONS to include vaccinations in the ONS COVID-19 risk models. This could be as: • the exposure of interest (i.e. is vaccination associated with reduced risk of adverse outcomes) • a confounder to be controlled for (i.e. does the relationship between socioeconomic exposure and outcome still hold after controlling for vaccination status) • an outcome (i.e. are certain socio-demographic groups more/less likely to take-up vaccination). Additional analysis focused on vaccinations data will also include estimating vaccination efficacy and investigating approaches to estimate the causal effect of vaccination on mortality. 1.4 QCOVID Analysis: ONS were required by SAGE and the Chief Medical Officer to urgently quality assure a QCOVID algorithm developed by Oxford University to identify clinical vulnerability to COVID-19. This algorithm has been used operationally by Public Health England as a replacement for the previously used shielding list. To do this, ONS ran the QCOVID algorithm against data which is independent of the dataset used to develop the model. ONS may update this work periodically. It is possible that there may be a commercial element to the QCOVID algorithm and/or its potential uses noting that 'QCovid' is a registered trademark. However, any commercial elements are peripheral to this agreement. There are no known commercial elements to ONS' involvement as an independent validator of the algorithm as requested by SAGE. 1.5 SCHOOLS INFECTIONS: ONS has been tasked with carrying out analysis looking at clustering of COVID-19 cases in schools and the impact that vaccination take-up has on infection rates of staff and pupils in each school. In addition, ONS has been asked to look at long COVID and potentially hospitalization direct from COVID-19. 2. Support production of statistics on the coronavirus pandemic and the associated social, economic and environmental impact on health and well-being. This includes statistics on the indirect impact of COVID-19 on mortality (e.g. non-COVID-19 excess mortality), morbidity (e.g. mental health), impact on vulnerable and disadvantaged groups and economic and social wellbeing.: 2.1 EXCESS MORTALITY AND MORBIDITY: To date, all work in this area has focused on excess deaths by analysing stand-alone ONS mortality data. However, it is possible that work in this area, and in particular the indirect excess mortality and morbidity that may have been caused by the pandemic (for example, missed cancer diagnoses) may require the use the data covered by this agreement. There is one example of investigating excess indirect morbidity where work has begun using the data in this agreement, and that is on mental health: 2.2 MENTAL HEALTH: ONS has been requested to look at the impact of the pandemic, and its associated lockdowns on the mental health of the population. This work initially focused on the GDPPR data working with NHS England to identify common mental health conditions in these data. ONS now plans to expand this work through the re-use Improving Access to Psychological Therapy (IAPT) data. The work includes understanding the characteristics of the population at risk (based on census data), identifying specific conditions, identifying new diagnoses, and identifying changes in existing conditions. 2.3 IMPACT OF THE PANDEMIC ON DISADVANTAGED GROUPS: ONS has been asked by DHSC to investigate the impact of the pandemic on disadvantaged groups, notably the disabled, by taking into account employment and benefits outcomes in addition to health outcomes. For example, is there evidence of increased rates of unemployment or benefits uptake in certain groups, with a consequence potentially being widening inequalities and how does this interact with health events; for example, evidence of negative economic outcomes following infection or hospitalization from COVID-19? The DHSC/DWP joint work health analysis unit are stakeholders in this area of the programme. ONS are aware of possible public concern in this area – i.e. the linkage of health and economic data. Therefore, to reiterate points made earlier, when health data are linked to benefits an income data, they will only be used to produce aggregate statistics (in this case, on impact and inequalities) by substantive ONS staff. They will never be used for operational purposes, and no DWP or DHSC officials can access the linked data and ONS retains independence and control over what analyses it conducts. A separate Data Sharing Agreement between ONS and DWP covers the use of benefits and income data for the production of official statistics. 2.4 SCHOOLS INFECTION: There is increased interest in the indirect impact of COVID-19 on the longer-term health of school children and the school workforce, with a specific focus on how this then impacts absence rates and the link to longer term education outcomes. For example, increased mental health diagnoses and also negative outcomes from other infections where the severity might be higher because of weaker immunity caused by social distancing during the pandemic. 2.5 WINTER PRESSURES: The NHS and social care were expected to face more severe pressures than normal in the winter of 2022 because of interacting risks driven by COVID-19 and compounded by the cost-of living crisis. COVID-19 has exacerbated workforce and operational pressures within the NHS and social care. The cost-of-living crisis will increase the number of vulnerable people not eating and/or heating their homes properly, worsening the impact of any surge in COVID-19 cases, making those recently infected with COVID more susceptible to cardio-vascular disease and potentially negatively impacting the health of those with long-COVID. These vulnerabilities are compounded by a surge in influenza and other viral and bacterial diseases over winter, and existing population vulnerabilities to non-communicable diseases exacerbated by poverty, indoor mixing and/or exposure to cold. ONS are producing statistics to support decision making around NHS winter pressures. This includes understanding the size and location of populations at risk, the consequences for physical and mental health including those with long COVID, and the impact that these compound pressures will have on the NHS and social care services. This work was commissioned by DHSC backed by the Chief Medical Officer. 3. Supporting numerical analysis on the quality of the underlying data itself and statistics produced in the two broad areas mentioned, so that the data can be used appropriately by others, and so that the statistics can be interpreted appropriately. All of the projects listed above under types 1 and 2 will include supporting quality information derived from the data covered by this DSA. In addition to that, ONS are pursuing a specific project investigating the quality of the NHS England ethnicity dataset (one of the sources covered by this agreement): 3.1 QUALITY OF NHS ENGLAND COVID-19 ETHNIC CATEGORY DATA ONS are working with NHS England to inform the reporting of health statistics by ethnicity. Reporting by ethnicity has become increasingly important to help understand inequalities during the pandemic. To support improvements to the reporting and quality of ethnicity health statistics and ultimately support and inform the governments' response to COVID-19 ONS will: Investigate the quality of NHS England health data sources in their reporting of ethnicity against the ONS's 2011 census data and explore methods to improve the quality of estimates using these data sources. Specifically, the first phase of this work will assess the quality of ethnicity data in NHS England’s COVID-19 Ethnic Category Data Set, as well as the Improving Access to Psychological Therapies (IAPT) data and the Birth notifications data.

Processing activities

[2 paragraphs unchanged] • General Practice Extraction Service (GPES) Data for Pandemic Planning and research (GDPPR) including the 'POSSPOSTCOVID_COD' cluster and the vaccinations and immunisations cluster obtained under DARS-NIC-388794-Z9P3J [2 paragraphs unchanged] The following Data are disseminated under a separate DSA ref: DARS-NIC-175120-W5G2X and the relevant subsets may be reused under this DSA for the purposes described above: • Hospital Episode Statistics Accident & Emergency (HES A&E) • Hospital Episode Statistics Outpatient (HES OP) • Emergency Care Dataset (ECDS) • Improving Access to Psychological Therapy (IAPT) • Birth Notifications data The following Data are disseminated under a separate DSA ref: DARS-NIC-20951- D2K6S and the relevant subsets may be reused under this DSA for the purposes described above: • Personal Demographics Service (PDS) [1 paragraph unchanged] The Data will be stored on servers at Crown Hosting Data Centres Limited. [1 paragraph unchanged] The Data will be accessed by authorised personnel via remote access. The Data will remain on the servers at Crown Hosting Data Centres Limited at all times. The Integrated Data Service (IDS) is a cloud-based Trusted Research Environment (TRE) where analysts process data for the purposes listed in this Agreement. Google UK Limited provide IT hosting services to ONS for the IDS and will store the data as contracted by ONS as a processor. The Secure Research Service (SRS) is hosted on the iTS computing Ltd Cloud platform. The servers used to store data and to host the analysis environment are located within a Pan-Government and National Cyber Security Centre (NCSC) Accredited (PGA) data centre, provided by iTS computing Ltd Data Centres and based in the United Kingdom only. This processing activity takes place in a dedicated ONS environment provided by Amazon Web Services, and based in England and Wales only. The Data will reside within Google Cloud Platform (GCP). All cloud services consumed for the storage and use of the data are scoped to securely managed GCP Projects. GCP Projects in scope for this project are not connected to ONS corporate networks. GCP Projects in scope for this project are accessible via the internet for administration/analytical work. Access is securely bound up with Google identity services, internet authentication proxies and multi-factor authentication. All platform infrastructure and storage is deployed into the Europe-west2 region (London) and in any of the 3 available zones for redundancy and high availability (where applicable). Processing of data can only be carried out on GCP infrastructure within the deployed region. Access to GCP platform is region locked to UK IP addresses only. Access to IDS and any data it holds is not permitted from outside the UK. Overseas connections are monitored, and connection attempts will lead to account suspension. The Data will be accessed by authorised personnel via remote access. [26 paragraphs unchanged] This generally revolves around taking a dataset that represents a population of interest (or at risk) and integrating both the NHS England Data covered by this DSA, and other relevant sources, with that core data to build powerful data assets. These assets are: Census 2011 based asset: ONS have linked the majority of census 2011 subjects to their NHS number as recorded in the patient registration information that ONS holds for 2011- 2013. This was successful for over 90% of Census records. This enables onward linkage to mortality data and the NHS health datasets covered by this DSA. ONS COVID Infection Study (CIS) asset: ONS have linked the majority of CIS subjects to their NHS number as recorded in the Personal Demographic Service data that ONS holds. This was successful for approximately 90% of participants. This enables onward linkage to mortality data and the NHS health datasets covered by this DSA. ONS Schools Infection data asset: The core of this data asset is the National Pupil Database (NPD) but includes the ONS Schools Infection Study (SIS) as a subset where, for participants in that study, ONS holds richer information than for the rest of the NPD population. The NPD and SIS will be linked to other sources, including health datasets through NHS number assignment and linkage (as per the first two assets), to enable production of the statistics that are requested. In addition, ONS have linked vaccination data (supplied by NHS England) and test and trace data (as supplied by DHSC) to the first two assets, and Property Attributes data from Valuation Office Agency (VOA) and Energy Performance Certificate (EPC) data on housing from MHCLG to the CIS data asset. The vaccination data will be updated with the higher quality monthly feed of vaccination from NHS England in due course. Any other sources that ONS has access to, which contain information about these populations at risk will only be linked to one or more of the assets if: • It is necessary to do so to support requests being made that fit the criteria and conditions of this DSA, and • The data sharing agreement with the relevant data supplier that other data, and supporting legal basis, provides for such linkage. As described in section 5a, the proposed purposes require linkage of records at the individual level. This is why personal identifiers such as date of birth, postcode and NHS number are required. However, ONS is only interested in producing aggregate statistics and using these to uncover trends and other useful insights based on the non-identifiable 'attribute' information. Inadvertent re-identification is still a risk but ONS will never seek to intentionally re-identify this data. ONS staff are suitably trained - for example, ONS's health analysts are experienced working with sensitive data about deaths (such as individual level data about suicides). Further, only statistical disclosure controlled aggregate outputs will be exportable from the secure data analysis environment. In other words, other than the initial transfer of the data from NHS England to ONS, the identifiable data will never be in transit and will always be protected by procedural and technical controls. [3 paragraphs unchanged] SPECIFIC PROCESSING ACTIVITIES RELATING TO QCOVID VALIDATION: ONS will run Oxford University's QCOVID algorithm against its linked data to validate the statistical performance of the QCovid algorithm in a dataset which is independent of the dataset used to develop the model. This will generate a c- statistic of between 0.5 and 1 (0.5 = no better at predicting outcome than random chance; 1 = perfect prediction), which will be provided to the researchers and clinicians that developed the algorithm. The subsequent decision on the suitability of the algorithm for clinical or operational purposes will rest with that team and not with ONS. ONS will provide the c-statistic back to the researchers and clinicians that developed the algorithm. The team at Oxford will then decide on the suitability of using the algorithm. ONS is not responsible for determining the threshold for acceptance. ONS' role is to provide validation of the algorithm's performance against an independent dataset, by an independent team of statistical experts, and not to make a clinical or operational judgement.

Expected output

[6 paragraphs unchanged] • Updated coronavirus related deaths by ethnic group and investigation of the explanatory power of hospital-based comorbidity on ethnic differences: https://www.ons.gov.uk/peoplepopulationandcommunity/birthsdeathsandmarriages/deaths/articles/updatingethniccontr astsindeathsinvolvingthecoronaviruscovid19englandandwales/deathsoccurring2marchto28july2020 https://www.ons.gov.uk/peoplepopulationandcommunity/birthsdeathsandmarriages/deaths/articles/updatingethniccontrastsindeathsinvolvingthecoronaviruscovid19englandandwales/deathsoccurring2marchto28july2020 [2 paragraphs unchanged] The output expected from the ethnicity quality work takes the form of [49 words unchanged] a technical brief on either the ONS website or another suitable website. The final publication from this project was published in May 2024: https://www.ons.gov.uk/peoplepopulationandcommunity/healthandsocialcare/healthinequalities/articles/understandingconsistencyofethnicitydatarecordedinhealthrelatedadministrativedatasetsinengland2011to2021/may2024 Update in September 2025: Since the start of 2024, the only ongoing work using GDPPR is the work that is investigating the economic impact of the pandemic on disadvantaged groups and involves linking GDPPR and other health data with benefits and income data (refer to " 2.3 IMPACT OF THE PANDEMIC ON DISADVANTAGED GROUPS" in Objective for processing. The following outputs and benefits have been generated to date: - https://blog.ons.gov.uk/2023/12/05/using-the-power-of-linked-data-to-understand-factors-preventing-people-from-working/ - https://www.ons.gov.uk/employmentandlabourmarket/peoplenotinwork/outofworkbenefits/bulletins/benefitrecipientsduringthecoronaviruscovid19pandemicengland/november2019tomarch2021 ONS has also conducted an observational, retrospective, matched cohort study of individuals admitted to hospital with Covid-19. ONS used the Hospital Episode Statistics Admitted Patient Care records for England up to 31 August 2020 and the General Practice Extraction Service Data for Pandemic Planning and Research (GDPPR) up to 30 September 2020. GDPPR is an extract of primary care records collected from surgeries by NHS Digital for pandemic research and analysis (supported by the British Medical Association and the Royal College of General Practitioners), including over 56 million individuals registered at NHS England general practice surgeries and updated fortnightly. The extract includes a subset of about 35 000 clinical codes, selected for potential use in pandemic related analysis. Death registrations from the Office for National Statistics were linked for deaths up to 30 September 2020 and registered by 7 October 2020. Link to the study can be found here: https://www.bmj.com/content/372/bmj.n693. The COVID-19 ethnicity dataset - the original project to assess the quality of NHS ethnicity information for NHS England, through a Wellcome Trust funded project – was published the final publication from this project in May 2024: https://www.ons.gov.uk/peoplepopulationandcommunity/healthandsocialcare/healthinequalities/articles/understandingconsistencyofethnicitydatarecordedinhealthrelatedadministrativedatasetsinengland2011to2021/may2024 The IMPACT OF THE PANDEMIC ON DISADVANTAGED GROUPS has led to the following outputs: - Impact of health conditions requiring hospitalisation on earnings, employment and benefits receipt, England - Benefit recipients during the coronavirus (COVID-19) pandemic, England: November 2019 to March 2021 ONS plan to publish a further breakdown of one of the conditions in the hospitalisations release, which includes benefits data. ONS plan to publish an academic paper, using results from the hospitalisations publication, including the benefits results.

Benefits reported

[4 paragraphs unchanged] The final publication was published in May 2024 for the project which assessed the quality of NHS ethnicity information for NHS England: https://www.ons.gov.uk/peoplepopulationandcommunity/healthandsocialcare/healthinequalities/articles/understandingconsistencyofethnicitydatarecordedinhealthrelatedadministrativedatasetsinengland2011to2021/may2024 [8 paragraphs unchanged] Since the start of 2024, the ongoing work is investigating the economic impact of the pandemic on disadvantaged groups and involves linking GDPPR and other health data with benefits and income data. The following outputs and benefits have been generated to date: https://blog.ons.gov.uk/2023/12/05/using-the-power-of-linked-data-to-understand-factors-preventing-people-from-working/ https://www.ons.gov.uk/employmentandlabourmarket/peoplenotinwork/outofworkbenefits/bulletins/benefitrecipientsduringthecoronaviruscovid19pandemicengland/november2019tomarch2021 ONS has also conducted an observational, retrospective, matched cohort study of individuals admitted to hospital with covid-19. ONS used the Hospital Episode Statistics Admitted Patient Care records for England up to 31 August 2020 and the General Practice Extraction Service Data for Pandemic Planning and Research (GDPPR) up to 30 September 2020. GDPPR is an extract of primary care records collected from surgeries by NHS Digital for pandemic research and analysis (supported by the British Medical Association and the Royal College of General Practitioners), including over 56 million individuals registered at NHS England general practice surgeries and updated fortnightly. The extract includes a subset of about 35 000 clinical codes, selected for potential use in pandemic related analysis. Death registrations from the Office for National Statistics were linked for deaths up to 30 September 2020 and registered by 7 October 2020.

Unchanged: Expected measurable benefits.

Text removed is struck through; text added is underlined. Unchanged paragraphs are summarised rather than repeated.

Unchanged: Objective for processing, Processing activities, Expected output, Expected measurable benefits, Benefits reported.

Text removed is struck through; text added is underlined. Unchanged paragraphs are summarised rather than repeated.

Objective for processing

The Office for National Statistics (ONS) requires access to NHS Digital data England Data for the purpose of supporting the national response to the COVID-19 pandemic. All the COVID-related analytical purposes. Processing of NHS England data requested is necessary to support the production of official national statistics. will fall under one analytical purposes outlined below: The following NHS Digital data will be accessed: • Support the production of statistics on the short, medium and long-term impacts of having had COVID-19, treatment for COVID-19 or a COVID-19 Vaccine – uses include statistics on the direct impact of COVID-19 infection on mortality (acute infection) and morbidity (COVID-19 sequelea and so called long-COVID), and how these are mediated by factors such as personal characteristics (e.g. ethnicity, age, sex), vaccination status and pre-existing conditions. • General Practice Extraction Service (GPES) Data for Pandemic Planning and research (GDPPR) including the 'POSSPOSTCOVID_COD' cluster and the vaccinations and immunisations cluster • Support production of statistics on the coronavirus pandemic and the associated social, economic and environmental impact on health and well-being - uses include statistics on the indirect impact of COVID-19 on mortality (e.g. non-COVID-19 excess mortality), morbidity (e.g. mental health), impact on vulnerable and disadvantaged groups and economic and social wellbeing. • Covid-19 Ethnic Category Data • Supporting numerical analysis on the quality of the underlying Data itself and statistics produced in the purposes mentioned, so that the Data can be used appropriately by others, and so that the statistics can be interpreted appropriately. • Hospital Episode Statistics Admitted Patient Care (HES APC) Data This DSA approves use of the Data listed for the production of statistics in the three purposes described. In doing so, the following criteria and conditions will apply: • HES A&E data • The statistics produced will meet a definitive user need or driven by commissions from the Scientific Advisory Group for Emergencies (SAGE), the Chief Medical Officer (CMO), the British Government or the Department for Health and Social Care (DHSC). These commissions are received via the National Statistician or the Director of Health and Pandemic Insight in ONS, working on his behalf; • HES Outpatient data • The statistics produced will be released as publicly available statistics on the ONS website in line with the UK code or practice for statistics; • HES Critical Care (CC) Data • The Data will never be used for operational purposes that could affect an individual – ONS is only interested in processing and linking the Data (even if this is linking at an individual level) to then enable the production of aggregate statistics that describe broad patterns in the data, therefore providing insight on what is happening in the real world; • Emergency Care Dataset (ECDS) data • Personal Demographics Service (PDS) data • Improving Access to Psychological Therapy (IAPT) data • Birth Notifications data This Agreement permits ONS to use the data listed to provide rapid response official statistics relating to the coronavirus pandemic. These statistics will inform key Government officials and their scientific advisors to make the best possible decisions as the pandemic progresses. The statistics that ONS produces under this Agreement will be driven by commissions from the Scientific Advisory Group for Emergencies (SAGE), the Chief Medical Officer (CMO), the British Government or the Department for Health and Social Care (DHSC). These commissions are received via the National Statistician or the director of Health and Pandemic Insight in ONS, working on his behalf. Experience has shown that is it often not possible to detail in advance the specifics of all the statistics and analysis needed. This includes knowing which other data sources the NHS Digital data will need to be linked to in order to produce the statistics requested. This is because of the fast moving and unpredictable nature of the pandemic and therefore being unable to predict ahead of time all of the statistics that stakeholders such as SAGE will request. However, in the broad terms, ONS can define the scope of what will be produced as: • Statistics on the short, medium and long-term impacts of having had COVID-19, treatment for COVID-19 or a COVID-19 vaccine – i.e. statistics on the direct impact of COVID-19 infection on mortality (acute infection) and morbidity (covid-19 sequelea and so called long-COVID), and how these are mediated by factors such as personal characteristics (e.g. ethnicity, age, sex), vaccination status and pre-existing conditions • Statistics on the coronavirus pandemic and the associated social, economic and environmental impact on health and well-being – i.e. statistics on the indirect impact of COVID-19 on mortality (e.g. non-COVID-19 excess mortality), morbidity (e.g. mental health), and economic and social wellbeing • Supporting numerical analysis on the quality of the underlying data itself and statistics produced in the two broad areas mentioned, so that the data can be used appropriately by others, and so that the statistics can be interpreted appropriately. The time that would be required to amend this Agreement for every specific use (once requested and therefore known) would not support the rapid response nature of this programme. Therefore, this Agreement approves use of the data listed for the production of statistics in the three broad areas described. In doing so, the following criteria and conditions will apply: • The statistics produced will meet a definitive user need or request from key Government officials and/or their scientific advisors tasked with guiding the UK’s response to the pandemic. Specifically, SAGE, the Chief Medical Officer, DHSC, and the British Government itself • The statistics produced will be released as publicly available statistics on the ONS website in line with the UK code or practice for statistics • The data will never be used for operational purposes that could affect an individual – ONS is only interested in processing and linking the data (even if this is linking at an individual level) to then enable the production of aggregate statistics that describe broad patterns in the data, therefore providing insight on what is happening in the real world [1 paragraph unchanged] • The number of ONS staff who access identifying data will be kept to the absolute minimum necessary and there will be a separation of duties between those using identifying information for linkage purposes, purposes and those doing the analysis based on integrated attribute data after the identifiers have been removed; • ONS analysts will never seek to reidentify and individual, an individual and it is worth noting that ONS analysts do not actually scrutinize individual rows of data, data; rather they run models on large scale datasets with millions of rows, rows and then scrutinize the outputs and aggregate results; • The programme already has approval from the National Statistician’s Data Ethics Advisory Committee (NSDEC), and NSDEC has reviewed and approved subsequent addendums as the work programme has expanded. NSDEC will be given opportunity to scrutinize and where necessary challenge, the ethical basis for the programme as new statistics are proposed and NSDEC’s approval will be obtained before ONS commences processing for any new use. ONS is committed to ethical use of the data, and welcomes NSDEC’s scrutiny and advice. • No other organisation will process the Data under this DSA. The Data shared with ONS will not be onwardly disseminated or shared under this DSA except as disclosure controlled aggregate statistics and/or analysis as aggregated data with small numbers suppressed, in line with the Hospital Episode Statistics Analysis Guide; • In the absence of amending this Agreement in advance with new specific uses, ONS will: • The Data will be used only for purposes which have approval from the National Statistician’s Data Ethics Advisory Committee (NSDEC). NSDEC’s role is to scrutinize and where necessary challenge, the ethical basis for the work as new statistics are proposed. i. Submit a briefing document specifying: the new uses; the expected outputs, dissemination plans and expected benefits; what datasets and what subsets thereof will be used; and any non-NHS Digital datasets with which the data will be linked. PRE-APPROVAL PROCESS FOR NEW USES: ii. Provide evidence of approval for the proposed processing by the National Statistician’s Data Ethics Advisory Committee (NSDEC). In the absence of amending this DSA ONS will complete the following steps for any new use of the Data. Any new use of the Data will be in support of one of the purposes outlined above. Before processing begins, ONS will:: iii. Not use the data for the proposed purpose before approval is confirmed in writing by NHS Digital after the above conditions are met. i. Submit a briefing document to NHS England specifying: the new uses; the expected outputs, dissemination plans and expected benefits; what datasets and what subsets thereof will be used; and any non-NHS England datasets with which the Data will be linked. If the new specific uses include either the GPES Data for Pandemic Planning and Research (GDPPR) dataset or the COVID-19 Ethnic Category Data Set, the briefing will be provided to NHS Digital’s Profession Advisory Group (PAG) and Independent Group Advising (NHS Digital) on Release of Data (IGARD) and NHS Digital approval will be subject to taking advice from IGARD and the PAG. ii. Provide evidence of approval for each new use by the NSDEC. It will not be necessary to amend this DSA before the data is used for the new purpose. iii. Not use the Data for the proposed purpose before approval is confirmed in writing by the Data Access Request Service (DARS) team in NHS England after the above conditions are met. The level of the data provided by NHS Digital will be: iv. If the new use involves either the GPES Data for Pandemic Planning and research (GDPPR) or the COVID-19 Ethnic Category Data, the briefing will be provided to Chair of NHS England’s Profession Advisory Group (PAG) and NHS England’s Advisory Group for Data (AGD), and NHS England approval will be subject to taking advice from AGD and the PAG. • Identifiable – necessary because linkage at a record level is a prerequisite to success for the proposed uses, and therefore identifiers including postcode, date of birth, sex and NHS number are required (or will be re-used) in the case of all the datasets listed. It will not be necessary to amend this DSA before the Data is used for the new use. To address the GDPR Principle of data minimisation, ONS has reviewed the data covered by this Agreement and confirmed that all variables listed are necessary for achieving the purposes listed. In some cases, this review has been done in conjunction with NHS Digital. The following NHS England Data will be accessed: • Hospital Episode Statistics (HES) Admitted Patient Care, Accident & Emergency and Outpatients – necessary because the HES data provides historical information about comorbidities relevant to negative outcomes from COVID-19, which is needed for risk factor modelling, and also provides information about the cause of secondary care contact during the pandemic (for example, being hospitalized because of acute COVID-19 infection). • Hospital Episode Statistics (HES) Critical Care – necessary because it allows ONS to stratify hospitalisations according to severity (ICU vs. non-ICU admission), which will be useful when defining pre-existing conditions, COVID-19 exposure, and outcomes following hospitalisation. ONS used this grading of severity in their long-COVID publication but it is not possible to replicate this using HES alone. These Data allow ONS to take into account the type and level of intensive care support patients were given. • Emergency Care Data Set (ECDS) – necessary because HES A&E Data has been discontinued and these data have been replaced by the more comprehensive ECDS. The ECDS Data will be used for the same reasons as HES A&E described above. • Personal Demographic Service – necessary to: o Facilitate linkage of non-NHS England data to NHS England Data by using personal identifiers to match records in the Census, CIS or NPD populations at risk with their NHS number. This enables onward linkage to other sources where NHS number is the unique identifier on those other sources; o Ensure each data asset includes the most up to date residential location for the subjects included (especially the Census based asset given the Census 2011 data is 10 years out of date); in turn, this location information can be used to assign the most up to date location-based indicators that are relevant to the statistics being produced; for example:  Determining if the current location a care home.  Determining the current Lower Layer Super Output Area of residence (and therefore deprivation score of area of residence).  Linking members of household together which is important for taking household transmission into account. • Improving Access to Psychological Therapy (IAPT) – necessary because it provides information about contact with such services and helps enable statistics on the indirect mental health impact of the pandemic. The ethnicity data included in this dataset will also help enable the quality assurance of ethnicity data on behalf of NHS England. • Birth Notifications – necessary because it includes NHS number of the mother and baby so will allow identification of the fact of a birth and allow ONS to identify babies in wider NHS England datasets. • General Practice Extraction Service (GPES) Data for Pandemic Planning and research (GDPPR) including the 'POSSPOSTCOVID_COD' cluster and the vaccinations and immunisations cluster – necessary because they provide historical information about pre-existing conditions relevant to negative outcomes from COVID-19 (acute or sequelae). The Data also provide information about diagnoses made during the pandemic relevant to indirect impacts; for example, mental health diagnoses. • COVID-19 Ethnic Category Data – necessary to assess the quality of ethnicity data in this dataset. The level of the Data will be identifiable – necessary because linkage at a record level is a prerequisite to success for the proposed uses, and therefore identifiers including postcode, date of birth, sex and NHS number are required (or will be re-used) in the case of all the datasets listed. To address the UK GDPR Principle of data minimisation, ONS has reviewed the data covered by this Agreement and confirmed that all variables listed are necessary for achieving the purposes listed. In some cases, this review has been done in conjunction with NHS England. [10 paragraphs unchanged] All purposes are tasks in the public interest. ONS is the sole data controller. ONS is the controller as the organisation responsible for ensuring that the Data will only be processed for the purpose described above. The lawful basis for processing personal data under the GDPR is: Though the work will be driven by commissions from the Scientific Advisory Group for Emergencies (SAGE), the Chief Medical Officer (CMO), the British Government or the Department for Health and Social Care (DHSC), ONS will be solely responsible for determining what data will be processed and the manner of its processing in order to undertake the work. • GDPR Article 6(1)(e) The processing is necessary for the performance of a task carried out in the public interest or in the exercise of official authority vested in the data controller. The authority for ONS to produce, promote and safeguard official statistics is found in the Statistics and Registration Service Act 2007. The lawful basis for processing personal data under the UK GDPR is: The lawful basis for processing special category data under the GDPR is: Article 6(1)(e) - processing is necessary for the performance of a task carried out in the public interest or in the exercise of official authority vested in the controller; • GDPR Article 9(2)(j) processing is necessary for archiving, statistics and research (with a basis in law). The lawful basis for processing special category data under the UK GDPR is: No other organisation will process the data under this Agreement. The data shared with ONS will not be onwardly disseminated or shared under this Agreement, except as disclosure controlled aggregate statistics and/or analysis as aggregated data with small numbers suppressed, in line with the Hospital Episode Statistics Analysis Guide. Article 9(2)(j) - processing is necessary for archiving purposes in the public interest, scientific or historical research purposes or statistical purposes in accordance with Article 89(1) based on Union or Member State law which shall be proportionate to the aim pursued, respect the essence of the right to data protection and provide for suitable and specific measures to safeguard the fundamental rights and the interests of the data subject. This Agreement came into force mid-pandemic in response to the increased demand on ONS for rapid response analysis. Previous versions of this Agreement did include specific descriptions permissions for each statistical use. Therefore, at the time of the Agreement coming into force, much is already known about how the data have been linked and what statistics have been produced or will be produced. As such, in the interests of transparency, ONS details examples in the remainder of this section. Note: these are not exhaustive and the data may be put to additional uses beyond those described. All purposes are tasks in the public interest. 1. Statistics on the short, medium and long-term impacts of having had COVID-19, treatment for COVID-19 or a COVID-19 vaccine – i.e. statistics on the direct impact of COVID-19 infection on mortality (acute infection) and morbidity (covid-19 sequelea and so called long-COVID), and how these are mediated by factors such as vaccination status and pre-existing conditions: The work is self-funded by the ONS. COVID-19 RISK FACTORS AND MORTALITY: Crown Hosting Data Centres Limited provides IT hosting services to ONS and will store the Data as contracted by ONS. This DSA came into force mid-pandemic in response to the increased demand on ONS for rapid response analysis. Previous versions of this DSA did include specific descriptions permissions for each statistical use. Therefore, at the time of the DSA coming into force, much is already known about how the data have been linked and what statistics have been produced or will be produced. As such, in the interests of transparency, ONS details examples in the remainder of this section. Note: these are not exhaustive and the Data may be put to additional uses beyond those described. 1. Support the production of statistics on the short, medium and long-term impacts of having had COVID-19, treatment for COVID-19 or a COVID-19 Vaccine. This includes statistics on the direct impact of COVID-19 infection on mortality (acute infection) and morbidity (COVID-19 sequelea and so called long-COVID), and how these are mediated by factors such as personal characteristics (e.g. ethnicity, age, sex), vaccination status and pre-existing conditions: 1.1 COVID-19 RISK FACTORS AND MORTALITY: [1 paragraph unchanged] 1.2 LONG-COVID: [1 paragraph unchanged] 1.3 VACCINATION: ONS already receive limited, daily data on COVID-19 vaccinations from NHS England, and are now requiring monthly disseminations of the higher quality Vaccination status data from NHS Digital England (ref: DARS-NIC-175120-W5G2X). [5 paragraphs unchanged] 1.4 QCOVID Analysis: [3 paragraphs unchanged] 1.5 SCHOOLS INFECTIONS: [1 paragraph unchanged] 2. Statistics Support production of statistics on the coronavirus pandemic and the associated social, economic and environmental impact on health and well-being – i.e. well-being. This includes statistics on the indirect impact of COVID-19 on mortality (e.g. non-COVID-19 excess mortality) and mortality), morbidity (e.g. mental health), impact on vulnerable and disadvantaged groups and economic and social wellbeing: wellbeing.: 2.1 EXCESS MORTALITY AND MORBIDITY: [1 paragraph unchanged] 2.2 MENTAL HEALTH: ONS has been requested to look at the impact of the pandemic, [10 words unchanged] population. This work initially focused on the GDPPR data working with NHS Digital England to identify common mental health conditions in these data. ONS now plans [29 words unchanged] identifying specific conditions, identifying new diagnoses, and identifying changes in existing conditions. 2.3 IMPACT OF THE PANDEMIC ON DISADVANTAGED GROUPS: [2 paragraphs unchanged] 2.4 SCHOOLS INFECTION: [1 paragraph unchanged] 3. Supporting numerical analysis on the quality of the underlying data itself, and statistics produced in the two broad areas mentioned, so that the data can be used appropriately by others and so that the statistics can be interpreted appropriately. 2.5 WINTER PRESSURES: All of the projects listed above under types 1 and 2 will include supporting quality information derived from the data covered by this agreement. In addition to that, ONS are pursuing a specific project investigating the quality of the NHS Digital ethnicity dataset (one of the sources covered by this agreement): The NHS and social care were expected to face more severe pressures than normal in the winter of 2022 because of interacting risks driven by COVID-19 and compounded by the cost-of living crisis. COVID-19 has exacerbated workforce and operational pressures within the NHS and social care. The cost-of-living crisis will increase the number of vulnerable people not eating and/or heating their homes properly, worsening the impact of any surge in COVID-19 cases, making those recently infected with COVID more susceptible to cardio-vascular disease and potentially negatively impacting the health of those with long-COVID. QUALITY OF NHS DIGITAL COVID-19 ETHNIC CATEGORY DATA These vulnerabilities are compounded by a surge in influenza and other viral and bacterial diseases over winter, and existing population vulnerabilities to non-communicable diseases exacerbated by poverty, indoor mixing and/or exposure to cold. ONS are working with NHS Digital to inform the reporting of health statistics by ethnicity. Reporting by ethnicity has become increasingly important to help understand inequalities during the pandemic. To support improvements to the reporting and quality of ethnicity health statistics and ultimately support and inform the governments' response to COVID-19 ONS will: ONS are producing statistics to support decision making around NHS winter pressures. This includes understanding the size and location of populations at risk, the consequences for physical and mental health including those with long COVID, and the impact that these compound pressures will have on the NHS and social care services. Investigate the quality of NHS Digital health data sources in their reporting of ethnicity against the ONS's 2011 census data and explore methods to improve the quality of estimates using these data sources. Specifically, the first phase of this work will assess the quality of ethnicity data in NHS Digital’s COVID-19 Ethnic Category Data Set, as well as the Improving Access to Psychological Therapies (IAPT) data and the Birth notifications data. This work was commissioned by DHSC backed by the Chief Medical Officer. To enable the work described requires the linkage of NHS Digital data with data sources that ONS already owns. 3. Supporting numerical analysis on the quality of the underlying data itself and statistics produced in the two broad areas mentioned, so that the data can be used appropriately by others, and so that the statistics can be interpreted appropriately. This generally revolves around taking a dataset that represents a population of interest (or at risk) and integrating both the NHS Digital data covered by this Agreement, and other relevant sources, with that core data to build powerful data assets. These assets are: All of the projects listed above under types 1 and 2 will include supporting quality information derived from the data covered by this DSA. In addition to that, ONS are pursuing a specific project investigating the quality of the NHS England ethnicity dataset (one of the sources covered by this agreement): Census 2011 based asset: ONS have linked the majority of census 2011 subjects to their NHS number as recorded in the patient registration information that ONS holds for 2011- 2013. This was successful for over 90% of Census records. This enables onward linkage to mortality data and the NHS health datasets covered by this Agreement. 3.1 QUALITY OF NHS ENGLAND COVID-19 ETHNIC CATEGORY DATA ONS COVID Infection Study (CIS) asset: ONS have linked the majority of CIS subjects to their NHS number as recorded in the Personal Demographic Service data that ONS holds. This was successful for approximately 90% of participants. This enables onward linkage to mortality data and the NHS health datasets covered by this Agreement. ONS are working with NHS England to inform the reporting of health statistics by ethnicity. Reporting by ethnicity has become increasingly important to help understand inequalities during the pandemic. To support improvements to the reporting and quality of ethnicity health statistics and ultimately support and inform the governments' response to COVID-19 ONS will: ONS Schools Infection data asset: The core of this data asset is the National Pupil Database (NPD) but includes the ONS Schools Infection Study (SIS) as a subset where, for participants in that study, ONS holds richer information than for the rest of the NPD population. The NPD and SIS will be linked to other sources, including health datasets through NHS number assignment and linkage (as per the first two assets), to enable production of the statistics that have been requested from us. Investigate the quality of NHS England health data sources in their reporting of ethnicity against the ONS's 2011 census data and explore methods to improve the quality of estimates using these data sources. Specifically, the first phase of this work will assess the quality of ethnicity data in NHS England’s COVID-19 Ethnic Category Data Set, as well as the Improving Access to Psychological Therapies (IAPT) data and the Birth notifications data. In addition, ONS have linked vaccination data (supplied by NHS England) and test and trace data (as supplied by DHSC) to the first two assets, and Property Attributes data from Valuation Office Agency (VOA) and Energy Performance Certificate (EPC) data on housing from MHCLG to the CIS data asset. The vaccination data will be updated with the higher quality monthly feed of vaccination from NHS Digital in due course. Any other sources that ONS has access to, which contain information about these populations at risk will only be linked to one or more of the assets if: • It is necessary to do so to support requests being made that fit the criteria and conditions of this agreement • The data sharing agreement with the relevant data supplier that other data, and supporting legal basis, provides for such linkage Further justification for the use of each NHS Digital dataset: Dataset 1: Dissemination of General Practice Extraction Service (GPES) Data for pandemic planning and research (GDPPR) These data are required because they provide historical information about pre-existing conditions relevant to negative outcomes from COVID-19 (acute or sequelae). The data also provide information about diagnoses made during the pandemic relevant to indirect impacts; for example, mental health diagnoses. The specification of the variables ONS receives was developed in collaboration with NHS Digital data experts and through a separate data access request which allowed three ONS researchers remote access to the data on NHS Digital systems (ref: DARS-NIC-388794-Z9P3J). Since the first supply, the specification has been updated once, to include codes relevant to long CVID and vaccinations. ONS receives an update every quarter. Dataset 2: Re-use of Hospital Episode Statistics (HES) APC, OP and A&E Primarily, the HES data provides historical information about comorbidities relevant to negative outcomes from COVID-19, which is needed for risk factor modelling, and also provides information about the cause of secondary care contact during the pandemic (for example, being hospitalized because of acute COVID-19 infection). Dataset 3: Emergency Care Dataset (ECDS) The HES data ONS already receives includes the Accident and Emergency (A&E) portion of HES. However, HES A&E data has now been discontinued by NHS Digital and these data have been replaced by the more comprehensive ECDS which started its data collection in 2017. The ECDS data will be used for the same reasons as HES A&E described above. Dataset 4: Dissemination of Critical Care Data The critical care data allow ONS to stratify hospitalisations according to severity (ICU vs. non-ICU admission), which will be useful when defining pre-existing conditions, COVID-19 exposure, and outcomes following hospitalisation. ONS used this grading of severity in their long-COVID publication, but it is not possible to replicate this using HES alone. These data allow ONS to take into account the type and level of intensive care support patients were given. The specification of the data was developed based on knowledge of the data through ONS's use of these data via a separate Agreement which allowed ONS researchers remote access to the data on NHS Digital systems (DARS- NIC-388794-Z9P3J). Dataset 5: Personal Demographic Service data (PDS) data The PDS data will be used to: • Facilitate linkage of non-NHS Digital data to NHS Digital data by using personal identifiers to match records in the Census, CIS or NPD populations at risk with their NHS number. This enables onward linkage to other sources where NHS number is the unique identifier on those other sources; • Ensure each data asset includes the most up to date residential location for the subjects included (especially the Census based asset given the Census 2011 data is 10 years out of date); in turn, this location information can be used to assign the most up to date location-based indicators that are relevant to the statistics being produced; for example: o Determining if the current location a care home o Determining the current Lower Layer Super Output Area of residence (and therefore deprivation score of area of residence) o Linking members of household together which is important for taking household transmission into account The derivation of location-based information is part of the data linkage processing phase and therefore only those who already have access to identifiers for such purposes would be involved in this re-use. Dataset 6: Re-use of Improving Access to Psychological Therapies (IAPT) The information included in the IAPT dataset will provide information about contact with such services and help enable statistics on the indirect mental health impact of the pandemic. The ethnicity data included in this dataset will also help enable the quality assurance of ethnicity data on behalf of NHS Digital. Dataset 7: Re-use Birth notifications Birth notifications include NHS number of the mother and baby so will allow identification of the fact of a birth and allow ONS to identify babies in wider NHS Digital datasets.

Processing activities

Under this DSA, NHS England will disseminate the following identifiable data: No data will flow to NHS England for the purposes of this Data Sharing Agreement (DSA). Under this DSA, NHS England will disseminate the following Data: [4 paragraphs unchanged] • Hospital Episode Statistics Admitted Patient Care Accident & Emergency (HES APC) A&E) • HES Accident & Emergency data • Hospital Episode Statistics Outpatient (HES OP) • HES Outpatient [1 paragraph unchanged] • Improving Access to Psychological Therapy (IAPT) data [2 paragraphs unchanged] • Personal Demographics Service (PDS) data The Data will contain directly identifying data items including NHS Number, Date of Birth and Postcode which are required to link the Data at record level with data already held by ONS. The Data will be stored on servers at Crown Hosting Data Centres Limited. The Data will not be transferred to any other location. The Data will be accessed by authorised personnel via remote access. The Data will remain on the servers at Crown Hosting Data Centres Limited at all times. Remote processing will only be through a secure electronic network and organisational controls prohibit personnel from downloading or copying data to local devices. Remote processing will be subject to the following being in place: • Multifactor authentication (MFA); • Access controls granting users the minimum level of access required; • Secure connections (e.g., VPNs or secure protocols) to protect data during remote access; • Device security, including up-to-date software and operating systems, antivirus software, and enabled firewalls. All remote access is undertaken within the scope of the relevant organisations’ DSPT (or other security arrangements as per this Data Sharing Agreement (DSA)). [15 paragraphs unchanged] ONS will keep a record of any processing of the Data and will provide a copy of such record to NHS England on request. ONS will not transfer or permit the transfer of the data to any territory outside England/Wales. The Data will not leave England & Wales at any time. All personnel accessing the Data have been appropriately trained in data protection and confidentiality. To enable the work described requires the linkage of NHS England Data with data sources that ONS already owns. This generally revolves around taking a dataset that represents a population of interest (or at risk) and integrating both the NHS England Data covered by this DSA, and other relevant sources, with that core data to build powerful data assets. These assets are: Census 2011 based asset: ONS have linked the majority of census 2011 subjects to their NHS number as recorded in the patient registration information that ONS holds for 2011- 2013. This was successful for over 90% of Census records. This enables onward linkage to mortality data and the NHS health datasets covered by this DSA. ONS COVID Infection Study (CIS) asset: ONS have linked the majority of CIS subjects to their NHS number as recorded in the Personal Demographic Service data that ONS holds. This was successful for approximately 90% of participants. This enables onward linkage to mortality data and the NHS health datasets covered by this DSA. ONS Schools Infection data asset: The core of this data asset is the National Pupil Database (NPD) but includes the ONS Schools Infection Study (SIS) as a subset where, for participants in that study, ONS holds richer information than for the rest of the NPD population. The NPD and SIS will be linked to other sources, including health datasets through NHS number assignment and linkage (as per the first two assets), to enable production of the statistics that are requested. In addition, ONS have linked vaccination data (supplied by NHS England) and test and trace data (as supplied by DHSC) to the first two assets, and Property Attributes data from Valuation Office Agency (VOA) and Energy Performance Certificate (EPC) data on housing from MHCLG to the CIS data asset. The vaccination data will be updated with the higher quality monthly feed of vaccination from NHS England in due course. Any other sources that ONS has access to, which contain information about these populations at risk will only be linked to one or more of the assets if: • It is necessary to do so to support requests being made that fit the criteria and conditions of this DSA, and • The data sharing agreement with the relevant data supplier that other data, and supporting legal basis, provides for such linkage. [1 paragraph unchanged] Inadvertent re-identification is still a risk but ONS will never seek to [67 words unchanged] be in transit and will always be protected by procedural and technical controls: controls. [1 paragraph unchanged] The data Data shared with ONS under this Agreement DSA will not be onwardly disseminated or shared, except as disclosure controlled aggregate [7 words unchanged] small numbers suppressed, in line with the Hospital Episode Statistics Analysis Guide. Analysts from the ONS will process/analyse the Data for the purposes described above. [3 paragraphs unchanged]

Expected measurable benefits

[1 paragraph unchanged] This These statistics are of national public health importance and have been requested by [40 words unchanged] refine its policy response to the pandemic using the best evidence available. [5 paragraphs unchanged]

Benefits reported

[3 paragraphs unchanged] ONS has also published initial findings on the ethnicity data quality work described above. In order to better understand the quality of ethnicity coding in electronic health records widely used to estimate COVID-19 mortality rates, ONS has published person-level comparisons of ethnicity recorded in three key health data sources (HES, GDPPR) compared with ethnicity information recorded in Census data (which is widely regarded as the most robust source of ethnicity data). This analysis showed the consistency of ethnicity recording compared with Census varied substantially for different ethnic groups – and was generally lower for those in the Mixed and ‘other’ ethnic groups. This analysis lays the foundations for future work which will look at solutions and methods analysts can use to produce more reliable estimates despite the differences between sources. [4 paragraphs unchanged] ONS has also published experimental statistics relating to the risk of hospital admission for coronavirus (COVID-19) and death involving COVID-19 by vaccination status by age group. This work is also currently undergoing peer-review for publication in the International Journal of Epidemiology. [3 paragraphs unchanged]

Unchanged: Expected output.

Text removed is struck through; text added is underlined. Unchanged paragraphs are summarised rather than repeated.

Objective for processing

[52 paragraphs unchanged] This Agreement came into force mid-pandemic in response to the increased demand [8 words unchanged] of this Agreement did include specific descriptions permissions for each statistical use. Therefore, at the time of the Agreement coming into force, much is already known about how the data have been linked and what statistics have been produced or will be produced. As such, in the interests of transparency, ONS details examples in the remainder of this section. Note: these are not exhaustive and the data may be put to additional uses beyond those described. Therefore, at time of the Agreement coming into force, much is already known about how the data have been linked and what statistics have been produced or will be produced. As such, in the interests of transparency, ONS details examples in the remainder of this section. Note: these are not exhaustive and the data may be put to additional uses beyond those described. [37 paragraphs unchanged] ONS Schools Infection data asset: The core of this data asset is the National Pupil Database (NPD) but includes the ONS Schools Infection Study (SIS) as a subset where where, for participants in that study, ONS holds richer information than for the [30 words unchanged] to enable production of the statistics that have been requested from us. [27 paragraphs unchanged]

Processing activities

No data will flow from ONS to NHS Digital under this Data Sharing Agreement (DSA). Under this DSA, NHS England will disseminate the following identifiable data: Under this DSA, NHS Digital will disseminate the following identifiable data: [2 paragraphs unchanged] • HES Critical Care (CC) Data [5 paragraphs unchanged] • HES Critical Care (CC) Data [18 paragraphs unchanged] With reasonable notice, periodic written/verbal checks may be conducted by an authorised employee of NHS Digital England to confirm compliance with this Agreement. ONS will keep a record of any processing of the Data and will provide a copy of such record to NHS Digital England on request. ONS will not transfer or permit the transfer of the data to any territory outside England/Wales. [1 paragraph unchanged] Inadvertent re-identification is still a risk but ONS will never seek to [47 words unchanged] other words, other than the initial transfer of the data from NHS Digital England to ONS, the identifiable data will never be in transit and will always be protected by procedural and technical controls: [5 paragraphs unchanged]

Expected output

[9 paragraphs unchanged] The output expected from the ethnicity quality work takes the form of [9 words unchanged] different sources. This would initially be an internal report for ONS, NHS Digital England and appropriate partners to use to inform their work and ensure appropriate [15 words unchanged] a technical brief on either the ONS website or another suitable website.

Changed only in punctuation, spacing or capitalisation: Benefits reported.

Unchanged: Expected measurable benefits.

Text removed is struck through; text added is underlined. Unchanged paragraphs are summarised rather than repeated.

Datasets: + Birth Notification Data; + Improving Access to Psychological Therapies Data Set_v1.5; + Personal Demographic Service

Objective for processing

The Office for National Statistics (ONS) requires NHS Digital data for three purposes, all relating the purpose of supporting the national response to the COVID-19 pandemic, and which are described here as Purpose 1, Purpose 2 and Purpose 3. pandemic. All the data requested is necessary to support the production of official national statistics. PURPOSE 1: Emerging questions relating to Covid-19 The following NHS Digital data will be accessed: PURPSOE 2: QCOVID Analysis PURPOSE 3: Investigating the quality of ethnicity data All purposes are tasks in the public interest. ONS is the sole data controller and will process personal data for this purpose under Articles 6(1)(e) and 9(2)(j) of the General Data Protection Regulation. • GDPR Article 6(1)(e) The processing is necessary for the performance of a task carried out in the public interest or in the exercise of official authority vested in the data controller. The authority for ONS to produce, promote and safeguard official statistics is found in the Statistics and Registration Service Act 2007. • GDPR Article 9(2)(j) processing is necessary for archiving, statistics and research (with a basis in law). No other organisation will process the data under this Agreement. The data shared with ONS will not be onwardly disseminated or shared under this Agreement, except as disclosure controlled aggregate statistics and/or analysis as aggregated data with small numbers suppressed, in line with the Hospital Episode Statistics Analysis Guide. The present Data Sharing Agreement (DSA) permits the dissemination or re-use of the following identifiable data: • Hospital Episode Statistics Admitted Patient Care (HES APC) Data disseminated under DARS-NIC-175120-W5G2X • HES A&E data disseminated under DARS-NIC-175120-W5G2X • HES Outpatient data disseminated under DARS-NIC-175120-W5G2X • Emergency Care Dataset (ECDS) data disseminated under DARS-NIC-175120-W5G2X • HES Critical Care (CC) Data [2 paragraphs unchanged] ONS currently hold the following non-NHS Digital datasets and have permissions, or are seeking permission, to link these to NHS Digital data: • Hospital Episode Statistics Admitted Patient Care (HES APC) Data • COVID-19 Vaccinations data from NHS England • HES A&E data • 2011 Census • HES Outpatient data • COVID-19 Testing data from UK Health Security Agency • HES Critical Care (CC) Data • Coronavirus Infection Survey (CIS) from ONS & University of Oxford • Emergency Care Dataset (ECDS) data • Property Attributes data from Valuation Office Agency (VOA) • Personal Demographics Service (PDS) data • Energy Performance Certificate (EPC) data on housing from MHCLG • Improving Access to Psychological Therapy (IAPT) data • National Pupil Database, which includes the English School Census • Birth Notifications data • School Workforce Census This Agreement permits ONS to use the data listed to provide rapid response official statistics relating to the coronavirus pandemic. These statistics will inform key Government officials and their scientific advisors to make the best possible decisions as the pandemic progresses. • Annual Population Survey (APS) The statistics that ONS produces under this Agreement will be driven by commissions from the Scientific Advisory Group for Emergencies (SAGE), the Chief Medical Officer (CMO), the British Government or the Department for Health and Social Care (DHSC). These commissions are received via the National Statistician or the director of Health and Pandemic Insight in ONS, working on his behalf. • Additional Census data including the 2021 Census Experience has shown that is it often not possible to detail in advance the specifics of all the statistics and analysis needed. This includes knowing which other data sources the NHS Digital data will need to be linked to in order to produce the statistics requested. • Genomics testing data This is because of the fast moving and unpredictable nature of the pandemic and therefore being unable to predict ahead of time all of the statistics that stakeholders such as SAGE will request. However, in the broad terms, ONS can define the scope of what will be produced as: • Birth registrations data • Statistics on the short, medium and long-term impacts of having had COVID-19, treatment for COVID-19 or a COVID-19 vaccine – i.e. statistics on the direct impact of COVID-19 infection on mortality (acute infection) and morbidity (covid-19 sequelea and so called long-COVID), and how these are mediated by factors such as personal characteristics (e.g. ethnicity, age, sex), vaccination status and pre-existing conditions Linkage at a record level is a prerequisite to success for the proposed uses, and therefore identifiers including postcode, date of birth, sex and NHS number are required. ONS does not require identifying details for any other reason than data linkage or to derive location-based variables for analysis. • Statistics on the coronavirus pandemic and the associated social, economic and environmental impact on health and well-being – i.e. statistics on the indirect impact of COVID-19 on mortality (e.g. non-COVID-19 excess mortality), morbidity (e.g. mental health), and economic and social wellbeing ONS require patient-level data because ONS are interested in patient-level socio-demographics, clinical profiles and outcomes. Use of more aggregated (e.g. regional level) data would result in a lack of statistical precision and risks the analysis being subjected to the so-called ecological fallacy. • Supporting numerical analysis on the quality of the underlying data itself and statistics produced in the two broad areas mentioned, so that the data can be used appropriately by others, and so that the statistics can be interpreted appropriately. To address the GDPR Principle of Data Minimisation ONS have reviewed the data disseminated and re-used under this Agreement and can confirm that all variables listed within this Agreement have been deemed to be necessary for achieving the purposes listed within this Agreement. In some cases, this review has been done in conjunction with NHS Digital. The time that would be required to amend this Agreement for every specific use (once requested and therefore known) would not support the rapid response nature of this programme. Therefore, this Agreement approves use of the data listed for the production of statistics in the three broad areas described. In doing so, the following criteria and conditions will apply: The necessity for the processing of the data for the are largely to do with ensuring the quality and therefore the value of the statistics that can be produced using such complete and record level data compared with less than this (for example, a subset, random sample, or aggregate data). There is more on statistical quality on the ONS website including the following: • The statistics produced will meet a definitive user need or request from key Government officials and/or their scientific advisors tasked with guiding the UK’s response to the pandemic. Specifically, SAGE, the Chief Medical Officer, DHSC, and the British Government itself • The statistics produced will be released as publicly available statistics on the ONS website in line with the UK code or practice for statistics • The data will never be used for operational purposes that could affect an individual – ONS is only interested in processing and linking the data (even if this is linking at an individual level) to then enable the production of aggregate statistics that describe broad patterns in the data, therefore providing insight on what is happening in the real world • These aggregate statistics will be disclosure controlled and will never identify an individual; • The number of ONS staff who access identifying data will be kept to the absolute minimum necessary and there will be a separation of duties between those using identifying information for linkage purposes, and those doing the analysis based on integrated attribute data after the identifiers have been removed; • ONS analysts will never seek to reidentify and individual, and it is worth noting that ONS analysts do not actually scrutinize individual rows of data, rather they run models on large scale datasets with millions of rows, and then scrutinize the outputs and aggregate results; • The programme already has approval from the National Statistician’s Data Ethics Advisory Committee (NSDEC), and NSDEC has reviewed and approved subsequent addendums as the work programme has expanded. NSDEC will be given opportunity to scrutinize and where necessary challenge, the ethical basis for the programme as new statistics are proposed and NSDEC’s approval will be obtained before ONS commences processing for any new use. ONS is committed to ethical use of the data, and welcomes NSDEC’s scrutiny and advice. • In the absence of amending this Agreement in advance with new specific uses, ONS will: i. Submit a briefing document specifying: the new uses; the expected outputs, dissemination plans and expected benefits; what datasets and what subsets thereof will be used; and any non-NHS Digital datasets with which the data will be linked. ii. Provide evidence of approval for the proposed processing by the National Statistician’s Data Ethics Advisory Committee (NSDEC). iii. Not use the data for the proposed purpose before approval is confirmed in writing by NHS Digital after the above conditions are met. If the new specific uses include either the GPES Data for Pandemic Planning and Research (GDPPR) dataset or the COVID-19 Ethnic Category Data Set, the briefing will be provided to NHS Digital’s Profession Advisory Group (PAG) and Independent Group Advising (NHS Digital) on Release of Data (IGARD) and NHS Digital approval will be subject to taking advice from IGARD and the PAG. It will not be necessary to amend this DSA before the data is used for the new purpose. The level of the data provided by NHS Digital will be: • Identifiable – necessary because linkage at a record level is a prerequisite to success for the proposed uses, and therefore identifiers including postcode, date of birth, sex and NHS number are required (or will be re-used) in the case of all the datasets listed. To address the GDPR Principle of data minimisation, ONS has reviewed the data covered by this Agreement and confirmed that all variables listed are necessary for achieving the purposes listed. In some cases, this review has been done in conjunction with NHS Digital. Generally full datasets are required (as opposed to a subset, random sample, or aggregate data) to ensure quality statistics can be produced. More information on statistical quality can be accessed on the ONS website but the following is a short summary: [7 paragraphs unchanged] PURPOSE 1: Emerging questions relating to COVID-19 It is crucial that the chances of drawing incorrect conclusions are kept to an absolute minimum, especially when the statistics will guide decision making in a pandemic. It is also crucial that accurate statistics can be produced quickly. This could save lives. The Office for National Statistics (ONS) has been asked to provide rapid responses to the coronavirus pandemic on emerging research questions which are arising as the pandemic continues, and as understanding of COVID-19 and its impact increases. This work is to support the ongoing government response, and has been requested the Scientific Advisory Group for Emergencies (SAGE) and the Government, via the National Statistician (a member of SAGE). To receive less data than have been requested, or to use aggregate data, would lead to less accurate statistics with greater uncertainty. It would also take longer to produce the statistics and therefore negatively affect timeliness because methods would need to be used to account for the missing data and/or greater limitations of the data compared with ONS accessing everything that has been requested. The work aims to improve understanding of, and support the development of, statistics on: All purposes are tasks in the public interest. ONS is the sole data controller. • the short, medium and long-term impacts of having had COVID-19, treatment for COVID-19 or a COVID-19 vaccine The lawful basis for processing personal data under the GDPR is: • the coronavirus pandemic and its associated social, economic and environmental impacts has had on health and well-being. • GDPR Article 6(1)(e) The processing is necessary for the performance of a task carried out in the public interest or in the exercise of official authority vested in the data controller. The authority for ONS to produce, promote and safeguard official statistics is found in the Statistics and Registration Service Act 2007. ONS will use the data under this Agreement to produce rapid responses to emerging questions driven by clinical issues relating to the COVID-19 pandemic as requested by either the Scientific Advisory Group for Emergencies (SAGE) or the Chief Medical Officer (CMO) via the National Statistician. The lawful basis for processing special category data under the GDPR is: As new questions emerge which require investigation using the data under this Agreement, ONS will submit a briefing paper to NHS Digital outlining the question(s) to be answered and identify the source of the request as either SAGE or the CMO. • GDPR Article 9(2)(j) processing is necessary for archiving, statistics and research (with a basis in law). Some specific examples of the work ONS has been requested to undertake are as follows: No other organisation will process the data under this Agreement. The data shared with ONS will not be onwardly disseminated or shared under this Agreement, except as disclosure controlled aggregate statistics and/or analysis as aggregated data with small numbers suppressed, in line with the Hospital Episode Statistics Analysis Guide. COVID-19 RISK FACTORS This Agreement came into force mid-pandemic in response to the increased demand on ONS for rapid response analysis. Previous versions of this Agreement did include specific descriptions permissions for each statistical use. This work aims to determine the population-level relative risk of hospitalisation or death that COVID-19 presents to different people. Data has been published and subsequently updated with the latest deaths data involving COVID-19, including by ethnic group using the power of hospital-based comorbidity data. Variation by religion and disability have also been published. Current work continues to use the linked data and will now incorporate primary care data on comorbidities and other risk factors. For example, ONS will shortly be publishing an article on the relationship between COVID-19 mortality and disability using data from GDPPR (to define comorbidities as well as learning disability as an exposure of interest). Therefore, at time of the Agreement coming into force, much is already known about how the data have been linked and what statistics have been produced or will be produced. As such, in the interests of transparency, ONS details examples in the remainder of this section. Note: these are not exhaustive and the data may be put to additional uses beyond those described. LONG-COVID 1. Statistics on the short, medium and long-term impacts of having had COVID-19, treatment for COVID-19 or a COVID-19 vaccine – i.e. statistics on the direct impact of COVID-19 infection on mortality (acute infection) and morbidity (covid-19 sequelea and so called long-COVID), and how these are mediated by factors such as vaccination status and pre-existing conditions: ONS has been tasked to continue analysis around the prevalence of, and risk factors associated with ‘long-COVID’ i.e. symptoms and conditions that persist/develop beyond the acute phase. This analysis is intended to inform public health messaging and interventions relating to long-COVID. Initial analysis was been carried out using data obtained through ONS’s remote access (ref: DARS-NIC-388794-Z9P3J) and this learning and analysis will be transferred to ONS systems and widened to enable further analysis using linked census data to explore socio-demographic variation in outcomes. COVID-19 RISK FACTORS AND MORTALITY: MENTAL HEALTH This work aims to determine the population-level relative risk of hospitalisation and death that COVID-19 presents to different people. Models have been published on mortality risk involving COVID-19 by ethnic group, religion and disability, controlling for comorbidities as derived from the HES and GP data. This work will continue to be updated and developed. ONS has been requested to look at the impact of the pandemic, and its associated lockdowns. on mental health of the population. This work initially focused on the GDPPR data working with NHS Digital to identify common mental health conditions in these new data. LONG-COVID: VACCINATION ONS has been tasked to continue analysis around the prevalence of, and risk factors associated with 'long-COVID' i.e. symptoms and conditions that persist/develop beyond the acute phase. This analysis is intended to inform public health messaging and interventions relating to long-COVID. As the coronavirus vaccination rolls out new research questions are emerging, and statistics being requested. VACCINATION: ONS will be receiving limited data on COVID-19 vaccinations from NHS England and linking these to ONS mortality, census data and Coronavirus Infection Survey (CIS) data. ONS request permission here to additionally link these data to the NHS Digital data outlined in this agreement to be used in a range of its COVID-19 analyses. ONS already receive limited, daily data on COVID-19 vaccinations from NHS England, and are now requiring monthly disseminations of the higher quality Vaccination status data from NHS Digital (ref: DARS-NIC-175120-W5G2X). Linking these data will allow ONS to include vaccinations in the ONS COVID-19 risk models. This could be as (i) the exposure of interest (i.e. is vaccination associated with reduced risk of adverse outcomes); (ii) a confounder to be controlled for (i.e. does the relationship between socioeconomic exposure and outcome still hold after controlling for vaccination status); or (iii) an outcome (i.e. are certain socio-demographic groups more/less likely to take-up vaccination). For example, initial work includes estimating the difference in efficacy between the different vaccine types and investigating approaches to estimate the causal effect of vaccination on mortality. This allows ONS to include vaccinations in the ONS COVID-19 risk models. This could be as: • the exposure of interest (i.e. is vaccination associated with reduced risk of adverse outcomes) • a confounder to be controlled for (i.e. does the relationship between socioeconomic exposure and outcome still hold after controlling for vaccination status) • an outcome (i.e. are certain socio-demographic groups more/less likely to take-up vaccination). [1 paragraph unchanged] The datasets ONS hold or are seeking to acquire (listed above) are required to be linked to each other in order to answer a wide range of emerging questions about the impact of the pandemic. For example: QCOVID Analysis: • Mortality data identifies information on the outcome of death including COVID-19 ONS were required by SAGE and the Chief Medical Officer to urgently quality assure a QCOVID algorithm developed by Oxford University to identify clinical vulnerability to COVID-19. This algorithm has been used operationally by Public Health England as a replacement for the previously used shielding list. • Information on socio-demographic characteristics such as ethnicity comes from Census data; Census 2021 data will provide more up to data population characteristics, for example on occupation To do this, ONS ran the QCOVID algorithm against data which is independent of the dataset used to develop the model. ONS may update this work periodically. • HES & ECDS data can identify hospitalisations (severe illness) as well as more severe existing/subsequent conditions or comorbidities record It is possible that there may be a commercial element to the QCOVID algorithm and/or its potential uses noting that 'QCovid' is a registered trademark. However, any commercial elements are peripheral to this agreement. There are no known commercial elements to ONS' involvement as an independent validator of the algorithm as requested by SAGE. • HES Critical care data can allow grading of morbidity according to severity, e.g. to stratify analysis of COVID-19 patients into those hospitalised with and without ICU admission and into type and level of intensive care support patients were given SCHOOLS INFECTIONS: • GDPPR data provide up to date and a more complete history of conditions, comorbidities and risk factors and so the underlying health of the population, as well as diagnoses of post-COVID syndrome ("long-COVID") as an outcome following infection ONS has been tasked with carrying out analysis looking at clustering of COVID-19 cases in schools and the impact that vaccination take-up has on infection rates of staff and pupils in each school. In addition, ONS has been asked to look at long COVID and potentially hospitalization direct from COVID-19. • COVID-19 testing data identifies people testing positive for COVID-19 including those who may not have been in contact with primary or secondary care services 2. Statistics on the coronavirus pandemic and the associated social, economic and environmental impact on health and well-being – i.e. statistics on the indirect impact of COVID-19 on mortality (e.g. non-COVID-19 excess mortality) and morbidity (e.g. mental health), and economic and social wellbeing: • COVID-19 vaccinations data provided by NHS England EXCESS MORTALITY AND MORBIDITY: • ODS Codes will allow us to link Vaccine data via SITE_CODE in order to identify the specific sites where vaccinations take place, for example to be used as a flag activity in specific regions. To date, all work in this area has focused on excess deaths by analysing stand-alone ONS mortality data. However, it is possible that work in this area, and in particular the indirect excess mortality and morbidity that may have been caused by the pandemic (for example, missed cancer diagnoses) may require the use the data covered by this agreement. There is one example of investigating excess indirect morbidity where work has begun using the data in this agreement, and that is on mental health: • Property attributes and housing data provides information on property type and size which could impact or help explain infection spread MENTAL HEALTH: •National Pupil Database, which includes the English School Census, and School Workforce Census will allow ONS to monitor ONS to monitor the infection rates among pupils and staff in schools, with linkage to HES, ECDS and GP data allowing analysists to identify cases of long-COVID, severe Covid requiring hospitalisation and other respiratory infections. This will also contribute to analysis of effectiveness of the vaccination programme amongst school staff and, should the programme be extended to under-18s, school children. ONS has been requested to look at the impact of the pandemic, and its associated lockdowns on the mental health of the population. This work initially focused on the GDPPR data working with NHS Digital to identify common mental health conditions in these data. ONS now plans to expand this work through the re-use Improving Access to Psychological Therapy (IAPT) data. The work includes understanding the characteristics of the population at risk (based on census data), identifying specific conditions, identifying new diagnoses, and identifying changes in existing conditions. • COVID-19 Ethnic Category Data Set to provide information on ethnicity according to NHSD sources. IMPACT OF THE PANDEMIC ON DISADVANTAGED GROUPS: • APS data to improve characteristics data availability, detail and timeliness compared to that of the 2011 census data (for a sample of the study population), for example data on self-reported health, impairments and activity limitations to inform disability analysis or other characteristics such as sexual orientation and more up to date occupation and employment status data. ONS has been asked by DHSC to investigate the impact of the pandemic on disadvantaged groups, notably the disabled, by taking into account employment and benefits outcomes in addition to health outcomes. For example, is there evidence of increased rates of unemployment or benefits uptake in certain groups, with a consequence potentially being widening inequalities and how does this interact with health events; for example, evidence of negative economic outcomes following infection or hospitalization from COVID-19? • DWP and HMRC data to understand employment outcomes, and when linked to health data, health outcomes due to COVID-19. The DHSC/DWP joint work health analysis unit are stakeholders in this area of the programme. ONS are aware of possible public concern in this area – i.e. the linkage of health and economic data. Therefore, to reiterate points made earlier, when health data are linked to benefits an income data, they will only be used to produce aggregate statistics (in this case, on impact and inequalities) by substantive ONS staff. They will never be used for operational purposes, and no DWP or DHSC officials can access the linked data and ONS retains independence and control over what analyses it conducts. A separate Data Sharing Agreement between ONS and DWP covers the use of benefits and income data for the production of official statistics. • Genomics testing data to investigate different COVID-19 variants and so associated outcomes. Note, these data refer to the genomics of the virus, not the infected individual. SCHOOLS INFECTION: JUSTIFICATION FOR DATA USAGE UNDER PURPOSE 1: There is increased interest in the indirect impact of COVID-19 on the longer-term health of school children and the school workforce, with a specific focus on how this then impacts absence rates and the link to longer term education outcomes. For example, increased mental health diagnoses and also negative outcomes from other infections where the severity might be higher because of weaker immunity caused by social distancing during the pandemic. Dataset 1: Hospital Episode Statistics (HES) APC, OP and A&E 3. Supporting numerical analysis on the quality of the underlying data itself, and statistics produced in the two broad areas mentioned, so that the data can be used appropriately by others and so that the statistics can be interpreted appropriately. ONS already hold HES Admitted Patient Care, Outpatient and Accident & Emergency data under DARS-NIC-175120-W5G2X, and will receive monthly updates on an ongoing basis. The purpose section of that Agreement has been updated to be compatible with the reuse of that data for the uses described here. All of the projects listed above under types 1 and 2 will include supporting quality information derived from the data covered by this agreement. In addition to that, ONS are pursuing a specific project investigating the quality of the NHS Digital ethnicity dataset (one of the sources covered by this agreement): Dataset 2: Emergency Care Dataset (ECDS) QUALITY OF NHS DIGITAL COVID-19 ETHNIC CATEGORY DATA The HES data ONS already receives includes the Accident and Emergency (A&E) portion of HES. However, HES A&E data has now been discontinued by NHS Digital and these data have been replaced by the more comprehensive ECDS which started its data collection in 2017. ONS holds HES A&E data covering April 2009 through to March 2020. ONS are working with NHS Digital to inform the reporting of health statistics by ethnicity. Reporting by ethnicity has become increasingly important to help understand inequalities during the pandemic. To support improvements to the reporting and quality of ethnicity health statistics and ultimately support and inform the governments' response to COVID-19 ONS will: ONS received ECDS monthly, that includes: Investigate the quality of NHS Digital health data sources in their reporting of ethnicity against the ONS's 2011 census data and explore methods to improve the quality of estimates using these data sources. Specifically, the first phase of this work will assess the quality of ethnicity data in NHS Digital’s COVID-19 Ethnic Category Data Set, as well as the Improving Access to Psychological Therapies (IAPT) data and the Birth notifications data. • Variables equivalent to those previously received from the HES A&E data To enable the work described requires the linkage of NHS Digital data with data sources that ONS already owns. • Additional variables that were not available in HES A&E that ONS needs for the purposes of this project, specifically to support understanding of comorbidities and outcomes. This generally revolves around taking a dataset that represents a population of interest (or at risk) and integrating both the NHS Digital data covered by this Agreement, and other relevant sources, with that core data to build powerful data assets. These assets are: The specification being requested was developed in collaboration with NHS Digital data experts to ensure the data being shared are of sufficient quality (e.g. coverage, accuracy, relevance) to be likely to support the statistical purpose intended. This is described fully in the separate Agreement (ref: DARS-NIC-175120-W5G2X). Census 2011 based asset: ONS have linked the majority of census 2011 subjects to their NHS number as recorded in the patient registration information that ONS holds for 2011- 2013. This was successful for over 90% of Census records. This enables onward linkage to mortality data and the NHS health datasets covered by this Agreement. Dataset 3: General Practice Extraction Service (GPES) Data for pandemic planning and research (GDPPR) ONS COVID Infection Study (CIS) asset: ONS have linked the majority of CIS subjects to their NHS number as recorded in the Personal Demographic Service data that ONS holds. This was successful for approximately 90% of participants. This enables onward linkage to mortality data and the NHS health datasets covered by this Agreement. These data are still new and sensitive. Therefore, the specification of the variables ONS are receiving was developed in collaboration with NHS Digital data experts and through a separate data access request which allowed three ONS researchers remote access to the data on NHS Digital systems (ref: DARS-NIC-388794-Z9P3J) ONS Schools Infection data asset: The core of this data asset is the National Pupil Database (NPD) but includes the ONS Schools Infection Study (SIS) as a subset where for participants in that study, ONS holds richer information than for the rest of the NPD population. The NPD and SIS will be linked to other sources, including health datasets through NHS number assignment and linkage (as per the first two assets), to enable production of the statistics that have been requested from us. The ONS researchers used this access, and worked with NHS Digital analysts, to learn about the GP data and to run some analyses by combining the data with mortality and HES data which are also present on NHS Digital systems. This has ensured the data being shared under this Agreement are of sufficient quality (e.g. coverage, accuracy, relevance) to be likely to support the statistical purpose intended in line with ONS requirements for producing official statistics under the Code of Practice (particularly transparency, quality and improvement). In addition, ONS have linked vaccination data (supplied by NHS England) and test and trace data (as supplied by DHSC) to the first two assets, and Property Attributes data from Valuation Office Agency (VOA) and Energy Performance Certificate (EPC) data on housing from MHCLG to the CIS data asset. The vaccination data will be updated with the higher quality monthly feed of vaccination from NHS Digital in due course. This work helped ONS come to a decision about if and what parts of the GP data needed to be transferred to ONS systems to enable ONS to meet the proposed purposes. ONS is therefore confident that the data specification being requested has been minimised to that absolutely necessary, in line with GDPR principles. Any other sources that ONS has access to, which contain information about these populations at risk will only be linked to one or more of the assets if: Since this first flow of GDPPR data there have been several developments which have highlighted the importance of a full and robust set of primary care data. The concept of long-covid has emerged, NICE has agreed a set of SNOMED codes for use in primary care and NHS Digital has requested primary care IT suppliers to implement these in GP systems for future supplies. Additionally, COVID-19 vaccinations are being rolled out. ONS request that the specification for the GDPPR data flows be updated, when the data are made available, to include these aspects to be used for the purposes described in this Agreement, this includes codes in the: • It is necessary to do so to support requests being made that fit the criteria and conditions of this agreement • ‘POSSPOSTCOVID_COD’ cluster • The data sharing agreement with the relevant data supplier that other data, and supporting legal basis, provides for such linkage • COVID-19 Vaccination under the vaccinations and immunisations cluster Further justification for the use of each NHS Digital dataset: Dataset 4: Critical Care Data Dataset 1: Dissemination of General Practice Extraction Service (GPES) Data for pandemic planning and research (GDPPR) The specification of the data ONS is requesting was developed based on knowledge of the data through ONS’s use of these data via a separate Agreement which allowed ONS researchers remote access to the data on NHS Digital systems (DARS-NIC-388794-Z9P3J). The ONS researchers used this access, to learn about the Critical care data and to run some analyses by combining the data with mortality, HES and GDPPR data which are also present on NHS Digital systems. Initial analysis using these data on long-COVID has been carried out and published (on the ONS website and as an academic article). These data are required because they provide historical information about pre-existing conditions relevant to negative outcomes from COVID-19 (acute or sequelae). The data also provide information about diagnoses made during the pandemic relevant to indirect impacts; for example, mental health diagnoses. The critical care data will allow ONS to stratify hospitalisations according to severity (ICU vs. non-ICU admission), which will be useful when defining pre-existing conditions, COVID-19 exposure, and outcomes following hospitalisation. ONS used this grading of severity in their long-COVID publication, but it is not possible to replicate this using the HES data to which ONS currently have access in ONS. The data requested will also allow ONS to take into account the type and level of intensive care support patients were given, e.g., for how many days did they receive advanced respiratory support, did the patient require level 2 or level 3 care, etc. These data will help ONS grade the severity of exposures and outcomes. The specification of the variables ONS receives was developed in collaboration with NHS Digital data experts and through a separate data access request which allowed three ONS researchers remote access to the data on NHS Digital systems (ref: DARS-NIC-388794-Z9P3J). Since the first supply, the specification has been updated once, to include codes relevant to long CVID and vaccinations. ONS receives an update every quarter. This work has ensured the data being shared under this Agreement are of sufficient quality (e.g. coverage, accuracy, relevance) and relevance to be likely to support the statistical purpose intended in line with ONS requirements for producing official statistics under the Code of Practice (particularly transparency, quality and improvement). Dataset 2: Re-use of Hospital Episode Statistics (HES) APC, OP and A&E ONS are now requesting to receive these data to be transferred to ONS systems so they can widen this initial use of the critical care data to enable further analysis particularly using the linked census data to explore socio-demographic variation in exposures and outcomes. Primarily, the HES data provides historical information about comorbidities relevant to negative outcomes from COVID-19, which is needed for risk factor modelling, and also provides information about the cause of secondary care contact during the pandemic (for example, being hospitalized because of acute COVID-19 infection). ONS requests Critical Care to be updated monthly. The years of data for this request are in line with the current supply of HES data ONS receive and will allow for comparisons over time to assess any COVID specific impact. Dataset 3: Emergency Care Dataset (ECDS) PURPOSE 2: QCOVID analysis The HES data ONS already receives includes the Accident and Emergency (A&E) portion of HES. However, HES A&E data has now been discontinued by NHS Digital and these data have been replaced by the more comprehensive ECDS which started its data collection in 2017. The ECDS data will be used for the same reasons as HES A&E described above. ONS request permission to re-use HES, ECDS and GDPPR data to support QCOVID Analysis . Dataset 4: Dissemination of Critical Care Data The Office for National Statistics (ONS) has been required by the Chief Medical Officer to urgently quality assure a QCOVID algorithm developed by Oxford University to identify clinical vulnerability to COVID-19. This algorithm will be used operationally by Public Health England as a replacement for the previously used shielding list. ONS is confident the information being requested for their part in assuring this algorithm are needed for statistical purposes as detailed below. The critical care data allow ONS to stratify hospitalisations according to severity (ICU vs. non-ICU admission), which will be useful when defining pre-existing conditions, COVID-19 exposure, and outcomes following hospitalisation. ONS used this grading of severity in their long-COVID publication, but it is not possible to replicate this using HES alone. These data allow ONS to take into account the type and level of intensive care support patients were given. ONS will run QCOVID against a range of health data assets to validate the statistical performance of the QCovid algorithm in a dataset which is independent of the dataset used to develop the model. The validation will also be undertaken by an independent team of experts, providing the highest level of assurance regarding the performance of the algorithm in line with TRIPOD (Transparent reporting of a multivariable prediction model for individual prognosis or diagnosis) guidance. The specification of the data was developed based on knowledge of the data through ONS's use of these data via a separate Agreement which allowed ONS researchers remote access to the data on NHS Digital systems (DARS- NIC-388794-Z9P3J). The data being used for this work are either owned by ONS, have been acquired through its statutory powers, or have been provided to ONS under COPI Notice for this purpose. The required dataset will be produced by linking information on outcomes with information on characteristics and underlying health conditions at a record level. Dataset 5: Personal Demographic Service data (PDS) data The information being linked and used in this dataset by ONS are as follows: The PDS data will be used to: • Information on the outcome of death comes from the death registration data ONS already processes and regularly publishes statistics on. • Facilitate linkage of non-NHS Digital data to NHS Digital data by using personal identifiers to match records in the Census, CIS or NPD populations at risk with their NHS number. This enables onward linkage to other sources where NHS number is the unique identifier on those other sources; • Information on socio-demographic characteristics such as ethnicity comes from the Census 2011. • Ensure each data asset includes the most up to date residential location for the subjects included (especially the Census based asset given the Census 2011 data is 10 years out of date); in turn, this location information can be used to assign the most up to date location-based indicators that are relevant to the statistics being produced; for example: • Information on hospitalisation (serious illness) from COVID-19 as an outcome, and information on underlying conditions that result in hospital contact, come from Hospital Episodes Statistics (HES) data. o Determining if the current location a care home • General Practice Extraction Service (GPES) Data for pandemic planning and research (GDPPR). o Determining the current Lower Layer Super Output Area of residence (and therefore deprivation score of area of residence) • Information on all patients, recorded in the National Radiotherapy Dataset or Systemic Anti-Cancer Therapy Dataset who received treatment (all types) with Radiotherapy between 01/06/2018 to 31/08/2020 or treatment with SACT between 01/06/2018 to 30/06/2020 o Linking members of household together which is important for taking household transmission into account • Information on persons who received a test result (positive only) for Covid-19 up to and including 27th October 2020 from the PHE Second Generation Surveillance System (SGSS) The derivation of location-based information is part of the data linkage processing phase and therefore only those who already have access to identifiers for such purposes would be involved in this re-use. The subsets of National Radiotherapy Dataset, Systemic Anti-Cancer Therapy Dataset and PHE Second Generation Surveillance System data are being shared directly with ONS by Public Health England under Regulation 3(4) of the Health Service Control of Patient Information Regulations 2002 for the explicit purpose of identifying and understanding information about patients or potential patients with or at risk of Covid-19 through the validation of the statistical performance of the QCovid algorithm, and will be held for no longer than is required for this purpose, currently agreed to be a six month period. Dataset 6: Re-use of Improving Access to Psychological Therapies (IAPT) The GDPPR data to be used in this work will be limited to: The information included in the IAPT dataset will provide information about contact with such services and help enable statistics on the indirect mental health impact of the pandemic. The ethnicity data included in this dataset will also help enable the quality assurance of ethnicity data on behalf of NHS Digital. • Patient identifiers Dataset 7: Re-use Birth notifications • patient demographics Birth notifications include NHS number of the mother and baby so will allow identification of the fact of a birth and allow ONS to identify babies in wider NHS Digital datasets. • diagnoses and findings • medications and other prescribed items • investigations, tests and results • treatments and outcomes • vaccinations and immunisations This work is of critical priority across Government as part of the UK’s response to the COVID-19 pandemic. It will contribute to the wider understanding of the virus, helping to inform a range of policy decisions taken by central government, health services and others. Optimising such decision making could ultimately save lives. It is possible that there may be a commercial element to the QCOVID algorithm and/or its potential uses noting that ‘QCovid’ is a registered trademark. However, any commercial elements are peripheral to this application. There are no known commercial elements to ONS’ intention to perform an independent validation of the algorithm as requested by SAGE. ONS have been commissioned to validate the revised QCovid model, meaning that this analysis is still ongoing. PURPOSE 3: Investigating quality of ethnicity data ONS are working with NHS Digital to inform the reporting of health statistics by ethnicity. Reporting by ethnicity has become increasingly apparent as a key requirement to help understand inequalities during the coronavirus pandemic. This has again been demonstrated in the work ONS recently published looking at ethnic differences in mortality outcomes in waves 1 and 2 of the pandemic which provides valuable insights on which groups are most at risk. This work will support improvements to the reporting and quality of ethnicity health statistics and ultimately support and inform the governments’ response to COVID-19. ONS will investigate the quality of NHS Digital health data sources in their reporting of ethnicity against the ONS’s 2011 census data, and explore methods to improve the quality of estimates using these data sources. ONS request additional permission to reuse and link datasets listed within this agreement to ONS deaths and Census data and to the non-NHS Digital data sources listed above. Specifically, the first phase of this work will include the additional following sources for this purpose: • COVID-19 Ethnic Category Data Set (NHSD) • additional years of past censuses including the 2021 Census once available and appropriate permissions are in place from data owners Depending on the outcome of the first phase of quality assurance additional datasets as listed in Purpose 1 and specifically required for COVID-19 analysis will be used to further assess quality of ethnicity data recorded in the linked data. These data sources add to the range of data on ethnicity available to investigate the quality of NHS Digital health data sources. Linking the sources listed above together will allow ONS to explore the consistency of ethnic group assignment between data sources, and assess reliability. In addition to the ethnicity data personal identifiers, including postcode, date of birth, sex and NHS number, from each of these datasets will be required to link the sources together to investigate the quality. ONS does not require identifying details for any other reason than data linkage or to derive location-based variables for analysis.

Processing activities

The present Data Sharing Agreement (DSA) permits the dissemination or re-use of the following identifiable data: No data will flow from ONS to NHS Digital under this Data Sharing Agreement (DSA). • Hospital Episode Statistics Admitted Patient Care (HES APC) Data disseminated under DARS-NIC-175120-W5G2X Under this DSA, NHS Digital will disseminate the following identifiable data: • HES A&E data disseminated under DARS-NIC-175120-W5G2X • General Practice Extraction Service (GPES) Data for Pandemic Planning and research (GDPPR) including the 'POSSPOSTCOVID_COD' cluster and the vaccinations and immunisations cluster obtained under DARS-NIC-388794-Z9P3J • HES Outpatient data disseminated under DARS-NIC-175120-W5G2X • Covid-19 Ethnic Category Data • Emergency Care Dataset (ECDS) data disseminated under DARS-NIC-175120-W5G2X The following data are disseminated under a separate DSA ref: DARS-NIC-175120-W5G2X and the relevant subsets may be reused under this DSA for the purposes described above: • Hospital Episode Statistics Admitted Patient Care (HES APC) • HES Accident & Emergency data • HES Outpatient • Emergency Care Dataset (ECDS) [1 paragraph unchanged] • General Practice Extraction Service (GPES) Data for Pandemic Planning and research (GDPPR) including the ‘POSSPOSTCOVID_COD’ cluster and the vaccinations and immunisations cluster • Improving Access to Psychological Therapy (IAPT) data • Covid-19 Ethnic Category Data • Birth Notifications data The following data are disseminated under a separate DSA ref: DARS-NIC-20951- D2K6S and the relevant subsets may be reused under this DSA for the purposes described above: • Personal Demographics Service (PDS) data [1 paragraph unchanged] • Need To Access applied through user account access and management . Access - access to the data is restricted to individuals granted access on the basis of a justified need to access the data data; • Controlled ingest and export of data into/out from the DAP environment environment; • Controlled account access using unique credentials based on job role role; • Logged and monitored access of user activity within the DAP environment environment; • Secure build configuration for infrastructure infrastructure; • Vulnerability tested infrastructure with appropriate remediation and patching patching; • Compliance checks against security enforcing controls controls; • Architectural review against standards and best practice practice; • Staff security cleared to the appropriate level based on their supervised and/or unsupervised access to sensitive data in accordance with ONS clearance policies and data access processes processes; • Education and awareness of environment users covering security policies and secure working practices practices; • Operational support processes to securely manage the environment environment; [3 paragraphs unchanged] ONS will keep a record of any processing of Personal the Data and will provide a copy of such record to NHS Digital on request. ONS will not transfer or permit the transfer of the Data data to any territory outside the UK without the prior written consent of NHS Digital. England/Wales. As described in section 5a, the proposed purposes require linkage of records at the individual level. This is why personal identifiers such as date of birth, postcode and NHS number are required. However, ONS is only interested in producing aggregate statistics and using these to uncover trends and other useful insights based on the non-identifiable ‘attribute’ 'attribute' information. [1 paragraph unchanged] Access to data held within the Data Access Platform (DAP) (DAP), which includes HES, ECDS and GDPPR data, is granted to users on a need-to-know basis depending on their role, [53 words unchanged] is higher than the standard basic clearance required for all ONS staff. Only authorised ONS staff with appropriate security clearance will have access to identifiable data with regular audit and monitoring in place to ensure compliance SPECIFIC PROCESSING ACTIVITIES RELATING TO PURPOSE 2: QCOVID analysis The data shared with ONS under this Agreement will not be onwardly disseminated or shared, except as disclosure controlled aggregate statistics and/or analysis as aggregated data with small numbers suppressed, in line with the Hospital Episode Statistics Analysis Guide. For the purpose of the quality assurance of the QCOVID algorithm, ONS will link the HES, ECDS and GDPPR data with the additional datasets described in the section above. SPECIFIC PROCESSING ACTIVITIES RELATING TO QCOVID VALIDATION: ONS will run the QCOVID algorithm against the linked data sets to validate the statistical performance of the algorithm. ONS will run Oxford University's QCOVID algorithm against its linked data to validate the statistical performance of the QCovid algorithm in a dataset which is independent of the dataset used to develop the model. This will generate a c- statistic of between 0.5 and 1 (0.5 = no better at predicting outcome than random chance; 1 = perfect prediction), which will be provided to the researchers and clinicians that developed the algorithm. The subsequent decision on the suitability of the algorithm for clinical or operational purposes will rest with that team and not with ONS. ONS will run Oxford University’s QCOVID algorithm against their data asset to validate the statistical performance of the QCovid algorithm in a dataset which is independent of the dataset used to develop the model. This will generate a c-statistic of between 0.5 and 1 (0.5 = no better at predicting outcome than random chance; 1 = perfect prediction), which will be provided to the researchers and clinicians that developed the algorithm. The subsequent decision on the suitability of the algorithm for clinical or operational purposes will rest with that team and not with ONS. [1 paragraph unchanged] This statistical quality assurance role will operate in parallel to any operational decisions on the roll out of the QCovid model for clinical purposes, which rest fully with Public Health England. Such decisions are outside the scope of this application. The team conducting the validation study is led by a Principal Statistician at ONS and a member of the Government Statistician Group, who is also working as a Research Fellow at the London School of Hygiene and Tropical Medicine and has years of experience using complex administrative records to conduct research and analysis. The team includes experienced statisticians and data scientists, some of whom have years of experience using administrative records and in particular primary care and hospital data. The ONS team will produce metrics as agreed with the Principal Researcher at Oxford University. The data shared with ONS under this Agreement will not be onwardly disseminated or shared, except as disclosure controlled aggregate statistics and/or analysis as aggregated data with small numbers suppressed, in line with the Hospital Episode Statistics Analysis Guide. SPECFIC PROCESSING ACTIVITIES RELATING TO PURPOSE 3: Investigating quality of ethnicity data The key data required for this purpose is limited to the use of ethnicity as recorded in the datasets listed in purpose 3 which will be compared to the ONS’s 2011 census self-reported ethnicity data. In addition to the ethnicity data socio-demographic variables such as age, sex and location-based indicators created under purpose 3 will be required to quality assure the ethnicity data and explore the distribution of ethnic populations reported in different sources. As well as these analytical variables personal identifiers, including postcode, date of birth, sex and NHS number, from each of these datasets will be required to link the sources together to investigate the quality. ONS does not require identifying details for any other reason than data linkage or to derive location-based variables for analysis.

Expected output

The outputs expected from this work are statistics, statistical reports and analyses [13 words unchanged] outputs contain any data, this data will be aggregated with small numbers supressed suppressed in line with HES Analysis Guidance. EXPECTED AND YIELDED OUTPUTS RELATING TO PURPOSE 1: Emerging questions relating to COVID-19 The outputs associated will typically be official statistics, reports and/or briefings for technical/expert audiences such as SAGE. Any official statistics produced will be published so that they are publicly available, for example on the Office for National Statistics website. The outputs associated with this purpose will typically be official statistics, reports and/or briefings for technical/expert audiences such as SAGE. Any official statistics produced will be shared with MPs, members of SAGE and other government officials to inform the Government’s response to the COVID-19 pandemic. Any official statistics produced will be published so that they are publicly available, for example on the Office for National Statistics website. [3 paragraphs unchanged] • A report assessing Coronavirus related deaths by ethnic group: https://www.ons.gov.uk/peoplepopulationandcommunity/birthsdeathsandmarriages/deaths/articles/coronaviruscovid19relateddeathsbyethnicgroupenglandandwales/2march2020to15may2020 https://www.ons.gov.uk/peoplepopulationandcommunity/birthsdeathsandmarriages/deaths/articles/coronaviruscovid19 relateddeathsbyethnicgroupenglandandwales/2march2020to15may2020 • Updated coronavirus related deaths by ethnic group and investigation of the explanatory power of hospital-based comorbidity on ethnic differences: https://www.ons.gov.uk/peoplepopulationandcommunity/birthsdeathsandmarriages/deaths/articles/updatingethniccontrastsindeathsinvolvingthecoronaviruscovid19englandandwales/deathsoccurring2marchto28july2020 https://www.ons.gov.uk/peoplepopulationandcommunity/birthsdeathsandmarriages/deaths/articles/updatingethniccontr astsindeathsinvolvingthecoronaviruscovid19englandandwales/deathsoccurring2marchto28july2020 EXPECTED AND YIELDED OUTPUTS RELATING TO PURPOSE 2: QCOVID analysis [1 paragraph unchanged] The result of the exercise will be shared with the Scientific Advisory Group for Emergencies (SAGE) and the University of Oxford. The bulk of the analysis has now been complete, but ONS still require continued access to these data to carry out any required revision and to quality assure any revisions to the algorithm. Outside of the work carried out by ONS, it is expected that there will be several journal articles published in relation to the QCOVID Algorithm. The bulk of the analysis has now been complete, but ONS still require continued access to these data to carry out any required revision and to quality assure any revisions to the algorithm. The output expected from the ethnicity quality work takes the form of a quality report providing aggregate data comparisons of the different sources. This would initially be an internal report for ONS, NHS Digital and appropriate partners to use to inform their work and ensure appropriate reporting and interpretation of health statistics by ethnicity. If appropriate it could be published as a technical brief on either the ONS website or another suitable website. Outside of the work carried out by ONS, it is expected that there will be several journal articles published in relation to the QCOVID Algorithm. EXPECTED OUTPUTS RELATING TO PURPOSE 3: Investigating quality of ethnicity data The output expected from this purpose takes the form of a quality report providing aggregate data comparisons of the different sources. This could include for example ethnic groups presented by age and sex to understand if reporting of ethnicity varied across socio-demographic characteristics. This would initially be an internal report for ONS, NHS Digital and appropriate partners to use to inform their work and ensure appropriate reporting and interpretation of health statistics by and COVID-19 analysis ethnicity. If appropriate it could be published as a technical brief on either the ONS website or another suitable website. The ONS will then explore the consistency of ethnic group assignment within these data sources over time as well as assess the consistency between data sources. ONS would aim to identify what individual factors influence the propensity for ethnic group ‘change’ (e.g., age and sex). Again, this analysis would first be an internal report, and ONS would seek to publish as appropriate. The ultimate aim of this research would be to produce methodological guidance to account for observed bias and missingness when analysing the health datasets included in this research.

Expected measurable benefits

PURPOSE 1: Emerging questions relating to COVID-19; [1 paragraph unchanged] This analysis is statistics are of national public health importance and has have been requested by central government leaders and advisors such as SAGE and the government, SAGE, via the National Statistician. The results of the analysis will be used to inform members of SAGE, Members of Parliament (MPs) and other key government officials of the differing COVID-19 risk profiles experienced by UK citizens. officials. These statistics will enable the government to refine its policy response to the pandemic using the best evidence available. [2 paragraphs unchanged] PURPOSE 2: QCOVID analysis The benefit of the exercise to validate the QCOVID algorithm is independent validation of an algorithm which will inform whether the model should be used or continue to be used to support government decision-making. This work is of critical priority across Government as part of the UK's response to the COVID-19 pandemic. It will contribute to the wider understanding of the virus, helping to inform a range of policy decisions taken by central government, health services and others. The benefit of the exercise to validate the QCOVID algorithm is independent validation of an algorithm which will inform whether the model should be used or continue to be used to support government decision-making. The benefit of the ethnicity data quality work will be: This work is of critical priority across Government as part of the UK’s response to the COVID-19 pandemic. It will contribute to the wider understanding of the virus, helping to inform a range of policy decisions taken by central government, health services and others. Optimising such decision-making could ultimately save lives. PURPOSE 3: Investigating quality of ethnicity data [1 paragraph unchanged]

Benefits reported

ONS has worked on a range out of topics and outputs in line with this agreement. Agreement. These statistics and analysis have been used to inform government policy and health campaign decisions, impacts include: Wide media coverage, for example highlighting inequalities in COVID-19 outcomes: Ethnic differences in COVID-19 mortality: • “Coronavirus: Black Britons face ‘twice the risk of death’ say ONS” was the top news story on the BBC website on the day of release The ONS has used the PHDA to highlight Ethnic differences in COVID-19 mortality and help understand the drivers. Numerous ONS bulletin and articles were published in peer-reviewed journals. The results have been presented multiple times at Scientific Advisory Group for Emergencies (SAGE), including in a very recent meeting. ONS has actively contributed to the SAGE Ethnicity subgroup, and data from the PHDA were used to inform policy papers. The initial analysis, which was subsequently published in the International Journal of Epidemiology, has been widely cited in academic research, including regularly in the British Medical Journal (BMJ). Professor Peter Goldblatt, University College London Institute for Health Equity, described the analysis as “a milestone in ethnicity analysis”. It has provided the baseline for further research into the social, clinical and biological reasons for the differences in COVID-19 mortality. It was awarded the RSS Campion award for excellence in official statistics. All subsequent analyses generated high media attention and were published in internationally renowned journals. Policy impact Qcovid validation: • UK Government ordered a review of the disproportionate impact of COVID-19 on Black, Asian and Minority Ethnic (BAME) communities ONS was commissioned by the CMO to conduct the validation of the QCovid risk model - an algorithm predicting the risk of COVID-19 death based on clinical records. The validation was conducted at pace and the results published in the Lancet Digital Health. As as a result, the algorithm was successfully used to update the shielding list and inform the prioritisation of the vaccination campaign. The overall project won the RSS Nightingale award • Community-led interventions to reduce inequalities and raise awareness COVID-19 mortality and vaccination: • Regular updates and further analyses provided to SAGE ethnicity subgroup ONS used the PHDA to analyse the links between COVID-19 vaccination and COVID-19 mortality. First, ONS conducted a study on vaccine effectiveness (VE) against mortality using a regression discontinuity design. The study was praised by Sir David Spiegelhalter as a ‘clever application of an innovative method’ and was used to assess the potential bias in the regular VE estimates produced by the UKHSA. ONS also used the PHDA to monitor COVID-19 vaccination rates by vaccination status. The findings were widely reported in the media and used by the Government and the NHS to promote the vaccination campaign. COVID-19 risk factors Media coverage: ONS examined inequalities in COVID-19 mortality by ethnicity, religion and disability. Using the data from the ONS Public Health Data Asset (Linked 2011 census, primary care and hospitalisation records, death registration data), ONS estimated mortality rates and examined whether these differences were driven by other socio-demographic factors. ONS also investigated how COVID-19 mortality varied by occupation, and whether these differences were driven by non-workplace factors. This recent piece with the National Statistician gives an overview of ONS’s role in the pandemic, although the specifics do not include much on linked data analysis: This work has furthered understanding of the interaction between COVID-19 in the following ways: https://www.newstatesman.com/encounter/2022/02/counting-the-cost-of-covid-with-ian-diamond • Ethnic minority groups were at an elevated risk of COVID-19 mortality, however lockdowns were associated with reductions in excess mortality risk in these groups. • Disabled people have a marked risk of mortality involving COVID-19 compared to non-disabled people, and as a result were prioritised within the pandemic response. • The risk of COVID-19 mortality was explained by controlling for sociodemographic and geographic determinants, however, those of Jewish affiliation remained at higher risk of death compared to all other groups. • Elderly people living with younger people are at increased risk of COVID-19 mortality, and that this is a contributing factor to the excess risks experienced by older South Asian women compared to White Women. • Working conditions are likely to play a role in COVID-19 mortality, particular where workers are coming into contact with COVID-19 patients or the public. The above findings have been used, and will continue to be used to inform the governments on-going response to the pandemic. Validation of the QCovid risk model Commissioned by the Chief Medical Officer for England, ONS validated the novel clinical risk prediction model QCovid to identify risks of short-term severe outcomes due to COVID-19. Following success of this validation ONS have been commissioned to validate the revised QCovid model. This model has been shown to perform well, and showed high levels of discrimination for COVID-19 deaths in men and women. Inequality in vaccination coverage Linking the ONS Public Health Data Asset to National Immunisation Management System (NIMS) data ONS investigated inequality in the coverage of vaccination against COVID-19, focusing on a range of sociodemographic characteristics, such as ethnicity, religion, disability and deprivation. One ONS study showed that populations that are most likely to be disproportionately affected by COVID-19 seemed most hesitant to vaccination. The government was able to adapt policies to improve these disparities.

Text removed is struck through; text added is underlined. Unchanged paragraphs are summarised rather than repeated.

Datasets: + COVID-19 Ethnic Category Data Set · − HES-ID to MPS-ID HES Accident and Emergency; − HES-ID to MPS-ID HES Admitted Patient Care; − HES-ID to MPS-ID HES Outpatients

Objective for processing

The Office for National Statistics (ONS) requires data for three purposes purposes, all relating to the COVID-19 pandemic pandemic, and which are described here as PURPOSE Purpose 1, PURPOSE Purpose 2 and PURPOSE Purpose 3. All the data requested is necessary for to support the production of official national statistics. All 3 purposes are considered to be tasks in the public interest. ONS is the sole data controller and will process personal data for this purpose under Articles 6(1)(e) and 9(2)(j) of the General Data Protection Regulation. PURPOSE 1: Emerging questions relating to Covid-19 PURPSOE 2: QCOVID Analysis PURPOSE 3: Investigating the quality of ethnicity data All purposes are tasks in the public interest. ONS is the sole data controller and will process personal data for this purpose under Articles 6(1)(e) and 9(2)(j) of the General Data Protection Regulation. • GDPR Article 6(1)(e) The processing is necessary for the performance of a task carried out in the public interest or in the exercise of official authority vested in the data controller. The authority for ONS to produce, promote and safeguard official statistics is found in the Statistics and Registration Service Act 2007. • GDPR Article 9(2)(j) processing is necessary for archiving, statistics and research (with a basis in law). [1 paragraph unchanged] PURPOSE 1: Emerging questions relating to COVID-19 The present Data Sharing Agreement (DSA) permits the dissemination or re-use of the following identifiable data: The Office for National Statistics (ONS) has been asked to provide rapid responses to the coronavirus pandemic on emerging research questions which are arising as the pandemic continues and as understanding of COVID-19 and its impact increases. This work is to support the ongoing government response and has been requested by central government leaders and advisors such as SAGE and the government, via the National Statistician. The work will improve understanding of and support the development of statistics on what the data can be used to learn about: • Hospital Episode Statistics Admitted Patient Care (HES APC) Data disseminated under DARS-NIC-175120-W5G2X • the short, medium and long-term impacts of having had COVID-19, treatment for COVID-19 or a COVID-19 vaccine • HES A&E data disseminated under DARS-NIC-175120-W5G2X • the coronavirus pandemic and its associated social, economic and environmental impacts has had on health and well-being. • HES Outpatient data disseminated under DARS-NIC-175120-W5G2X Some examples of the work ONS has been asked to carry out follows. • Emergency Care Dataset (ECDS) data disseminated under DARS-NIC-175120-W5G2X Initial work described in a previous version of this Agreement on analysis to determine the population-level relative risk of hospitalisation or death that COVID-19 presents to different people continues. Data have been published and subsequently updated with the latest deaths data on mortality involving COVID-19 by ethnic group using the power of hospital-based comorbidity data. Variation by religion and disability have also been published. Current work continues to use the linked data and will next incorporate primary care data on comorbidities and risk factors. For example, ONS will shortly be publishing an article on the relationship between COVID-19 mortality and disability using data from GDPPR (to define comorbidities as well as learning disability as an exposure of interest). • HES Critical Care (CC) Data As a second example, ONS has been tasked to continue analysis around the prevalence and risk factors for ‘long-COVID’ i.e. symptoms and conditions that persist/develop beyond the acute phase. This analysis is intended to inform public health messaging and interventions relating to long-COVID. Initial analysis has been carried out and published (on the ONS website and as an academic article) through ONS’s remote access (DARS-NIC-388794-Z9P3J-v2.3) and this learning and analysis will be transferred to ONS systems and widened to enable further analysis using linked census data to explore socio-demographic variation in outcomes. • General Practice Extraction Service (GPES) Data for Pandemic Planning and research (GDPPR) including the ‘POSSPOSTCOVID_COD’ cluster and the vaccinations and immunisations cluster As another example, ONS has been requested as a priority to look at the impact of the pandemic and the lockdown on mental health for the whole population. This work will initially focus on the GDPPR data working with NHS Digital to identify common mental health conditions in these new data and has started through ONSs remote access. The work includes understanding the characteristics of the population at risk, identifying specific conditions, identifying new diagnoses and identifying changes in existing conditions. The analysis will be transferred to ONSs systems and widened to enable further analysis using linked census data to explore socio-demographic variation in outcomes. • Covid-19 Ethnic Category Data As the coronavirus vaccination rolls out new research questions are emerging and statistics being requested. ONS will be receiving limited data on COVID-19 vaccinations from NHS England and linking these to ONS mortality, census data and Coronavirus Infection Survey (CIS) data. ONS request permission here to additionally link these data to the NHS Digital data outlined in this agreement to be used in a range of its COVID-19 analyses. Linking these data will allow ONS to include vaccinations in the ONS COVID-19 risk models. This could be as (i) the exposure of interest (i.e. is vaccination associated with reduced risk of adverse outcomes); (ii) a confounder to be controlled for (i.e. does the relationship between socio-economic exposure and outcome still hold after controlling for vaccination status); or (iii) an outcome (i.e. are certain socio-demographic groups more/less likely to take-up vaccination). For example, initial work includes estimating the difference in efficacy between the different vaccines types and investigating approaches to estimate the causal effect of vaccination on mortality. ONS currently hold the following non-NHS Digital datasets and have permissions, or are seeking permission, to link these to NHS Digital data: The data being used so far are being processed and analysed on ONS’s secure data platform. These data are either owned by ONS or have been acquired under previous versions of this Data Sharing Agreement or under separate Data Sharing Agreements. ONS have already been able to fill some important gaps in the data primarily, this is to do with comorbidities. Without a complete picture of comorbidities, it is not possible to explain all of the differences being found in COVID-19 related mortality and morbidity between different groups/characteristics (such as people of different ethnicity). This gap has been filled by the acquisition and linkage of HES, ECDS and GDPPR data to the ONS-owned census and deaths data. Based on ONS’s experience and evidence of the utility of the data already analysed and linked in the NHS Digital remote access area (DARS-NIC-388794-Z9P3J) and the evolving data landscape, ONS are requesting additional data and linkage permissions to support their ongoing response to the COVID-19 pandemic. Under the previous version of this Agreement, ONS was approved to link the following NHS Digital data with deaths and demographics (2011 Census): • Hospital Episode Statistics (HES) Admitted Patient Care and Outpatient subsets 2017/18 to 2020/21 • HES Accident & Emergency 2017/18-2018/19 • Emergency Care DataSet (ECDS) 2019/20-2020/21 • General Practice Extraction Service (GPES) Data for Pandemic Planning and research (GDPPR) This Agreement will additionally permit ONS to receive or reuse and link: • Hospital Episode Statistics (HES) Admitted Patient Care and Outpatient subsets 2009/10 to 2016/17 and provisional monthly data for 2021/22 • HES Accident & Emergency 2009/10 to 2016/17 and 2019/20 • Emergency Care DataSet (ECDS) provisional monthly data for 2021/22 • HES critical care data 2009/10 to 2019/20 annual data and provisional monthly data for 2021/22 • Additional HES variables (Operation and Diagnosis Codes) and the historical time series updates as received under DARS-NIC-175120-W5G2X • New long-COVID SNOMED codes developed and collected as part of the GDPPR refresh • GDPPR data which identifies COVID-19 vaccination ONS also hold or are looking to acquire further non-NHS Digital data to support the COVID-19 analyses and the purposes described in this agreement and request permission to link these external data to the NHS Digital data held or reused as part of this agreement along with the link to ONS deaths and census data. Specifically, ONS seek permission to link to the following sources once appropriate permissions are in place from their respective data owners: [1 paragraph unchanged] • Test and Trace data from PHE • 2011 Census • COVID-19 Testing data from UK Health Security Agency [3 paragraphs unchanged] The datasets ONS hold or are seeking to acquire are required to be linked to each other in order to answer a wide range of emerging questions about the impact of the pandemic. For example: • National Pupil Database, which includes the English School Census • School Workforce Census • Annual Population Survey (APS) • Additional Census data including the 2021 Census • Genomics testing data • Birth registrations data Linkage at a record level is a prerequisite to success for the proposed uses, and therefore identifiers including postcode, date of birth, sex and NHS number are required. ONS does not require identifying details for any other reason than data linkage or to derive location-based variables for analysis. ONS require patient-level data because ONS are interested in patient-level socio-demographics, clinical profiles and outcomes. Use of more aggregated (e.g. regional level) data would result in a lack of statistical precision and risks the analysis being subjected to the so-called ecological fallacy. To address the GDPR Principle of Data Minimisation ONS have reviewed the data disseminated and re-used under this Agreement and can confirm that all variables listed within this Agreement have been deemed to be necessary for achieving the purposes listed within this Agreement. In some cases, this review has been done in conjunction with NHS Digital. The necessity for the processing of the data for the are largely to do with ensuring the quality and therefore the value of the statistics that can be produced using such complete and record level data compared with less than this (for example, a subset, random sample, or aggregate data). There is more on statistical quality on the ONS website including the following: ‘The quality of a statistical product can be defined as the “fitness for purpose” of that product. More specifically, it is the fitness for purpose with regards to the European Statistical System dimensions of quality: • relevance – is the degree to which a statistical product meets user needs in terms of content and coverage? • accuracy and reliability – how close is the estimated value in the output is to the true result? • timeliness and punctuality – describes the time between the date of publication and the date to which the data refers, and the time between the actual publication and the planned publication of a statistic • accessibility and clarity – is the ease with which users can access data, and the quality and sufficiency of metadata, illustrations and accompanying advice? • coherence and comparability – is the degree to which data derived from different sources or methods, but that refers to the same topic, is similar, and the degree to which data can be compared over time and domain, for example, geographic level? There are additional characteristics that should be considered when thinking about quality. These include output quality trade-offs, user needs and perceptions, performance cost and respondent burden, and confidentiality, transparency and security.’ PURPOSE 1: Emerging questions relating to COVID-19 The Office for National Statistics (ONS) has been asked to provide rapid responses to the coronavirus pandemic on emerging research questions which are arising as the pandemic continues, and as understanding of COVID-19 and its impact increases. This work is to support the ongoing government response, and has been requested the Scientific Advisory Group for Emergencies (SAGE) and the Government, via the National Statistician (a member of SAGE). The work aims to improve understanding of, and support the development of, statistics on: • the short, medium and long-term impacts of having had COVID-19, treatment for COVID-19 or a COVID-19 vaccine • the coronavirus pandemic and its associated social, economic and environmental impacts has had on health and well-being. ONS will use the data under this Agreement to produce rapid responses to emerging questions driven by clinical issues relating to the COVID-19 pandemic as requested by either the Scientific Advisory Group for Emergencies (SAGE) or the Chief Medical Officer (CMO) via the National Statistician. As new questions emerge which require investigation using the data under this Agreement, ONS will submit a briefing paper to NHS Digital outlining the question(s) to be answered and identify the source of the request as either SAGE or the CMO. Some specific examples of the work ONS has been requested to undertake are as follows: COVID-19 RISK FACTORS This work aims to determine the population-level relative risk of hospitalisation or death that COVID-19 presents to different people. Data has been published and subsequently updated with the latest deaths data involving COVID-19, including by ethnic group using the power of hospital-based comorbidity data. Variation by religion and disability have also been published. Current work continues to use the linked data and will now incorporate primary care data on comorbidities and other risk factors. For example, ONS will shortly be publishing an article on the relationship between COVID-19 mortality and disability using data from GDPPR (to define comorbidities as well as learning disability as an exposure of interest). LONG-COVID ONS has been tasked to continue analysis around the prevalence of, and risk factors associated with ‘long-COVID’ i.e. symptoms and conditions that persist/develop beyond the acute phase. This analysis is intended to inform public health messaging and interventions relating to long-COVID. Initial analysis was been carried out using data obtained through ONS’s remote access (ref: DARS-NIC-388794-Z9P3J) and this learning and analysis will be transferred to ONS systems and widened to enable further analysis using linked census data to explore socio-demographic variation in outcomes. MENTAL HEALTH ONS has been requested to look at the impact of the pandemic, and its associated lockdowns. on mental health of the population. This work initially focused on the GDPPR data working with NHS Digital to identify common mental health conditions in these new data. VACCINATION As the coronavirus vaccination rolls out new research questions are emerging, and statistics being requested. ONS will be receiving limited data on COVID-19 vaccinations from NHS England and linking these to ONS mortality, census data and Coronavirus Infection Survey (CIS) data. ONS request permission here to additionally link these data to the NHS Digital data outlined in this agreement to be used in a range of its COVID-19 analyses. Linking these data will allow ONS to include vaccinations in the ONS COVID-19 risk models. This could be as (i) the exposure of interest (i.e. is vaccination associated with reduced risk of adverse outcomes); (ii) a confounder to be controlled for (i.e. does the relationship between socioeconomic exposure and outcome still hold after controlling for vaccination status); or (iii) an outcome (i.e. are certain socio-demographic groups more/less likely to take-up vaccination). For example, initial work includes estimating the difference in efficacy between the different vaccine types and investigating approaches to estimate the causal effect of vaccination on mortality. Additional analysis focused on vaccinations data will also include estimating vaccination efficacy and investigating approaches to estimate the causal effect of vaccination on mortality. The datasets ONS hold or are seeking to acquire (listed above) are required to be linked to each other in order to answer a wide range of emerging questions about the impact of the pandemic. For example: [1 paragraph unchanged] • Information on socio-demographic characteristics such as ethnicity comes from the Census 2011 data; Census 2021 data will provide more up to data population characteristics, for example on occupation [4 paragraphs unchanged] • COVID-19 vaccinations data identifies people who have received a coronavirus vaccine and could be used as a control variable when assessing outcomes as well as to explore vaccinations take up and efficacy • COVID-19 vaccinations data provided by NHS England • ODS Codes will allow us to link Vaccine data via SITE_CODE in order to identify the specific sites where vaccinations take place, for example to be used as a flag activity in specific regions. [1 paragraph unchanged] Taken together these data will allow ONS to build a rich picture of the health state and experiences of the population due to COVID-19 and answer emerging questions on the impact of the pandemic. As the pandemic continues and understanding of coronavirus increases and changes, ONS need this ability to respond to emerging questions quickly. ONS will use the data under this Agreement to produce rapid responses to emerging questions driven by clinical issues relating to the COVID-19 pandemic as requested by either the Scientific Advisory Group for Emergencies (SAGE) or the Chief Medical Officer (CMO) via the National Statistician. •National Pupil Database, which includes the English School Census, and School Workforce Census will allow ONS to monitor ONS to monitor the infection rates among pupils and staff in schools, with linkage to HES, ECDS and GP data allowing analysists to identify cases of long-COVID, severe Covid requiring hospitalisation and other respiratory infections. This will also contribute to analysis of effectiveness of the vaccination programme amongst school staff and, should the programme be extended to under-18s, school children. As new questions emerge which require investigation using the data under this Agreement, ONS will submit a briefing paper to NHS Digital outlining the question(s) to be answered and identify the source of the request as either SAGE or the CMO. • COVID-19 Ethnic Category Data Set to provide information on ethnicity according to NHSD sources. ONS already receives the HES data from NHS Digital under a separate Data Sharing Agreement. ONS compelled NHS Digital to share this HES data with ONS for the purposes of official statistics under Section 45C of the Statistics and Registration Services Act (2008), as amended by the Digital Economy Act (2017), in January 2019. • APS data to improve characteristics data availability, detail and timeliness compared to that of the 2011 census data (for a sample of the study population), for example data on self-reported health, impairments and activity limitations to inform disability analysis or other characteristics such as sexual orientation and more up to date occupation and employment status data. The Government’s Scientific Advisory Group for Emergencies (SAGE) and the National Statistician are all clear that they want this improvement to the breadth of ONSs COVID-19 analyses to be enabled by NHS Digital. The insight gained from this improved linked data asset will inform decision making that could ultimately save lives. • DWP and HMRC data to understand employment outcomes, and when linked to health data, health outcomes due to COVID-19. This Agreement authorises ONS to continue to receive extracts of General Practice Extraction Service (GPES) Data for Pandemic Planning and research (GDPPR) data (quarterly), and Emergency Care Data Set (ECDS) (monthly) and to receive extracts of the Critical Care Data (monthly) in line with HES data already flowing. These data will be used solely for the purposes described in this Agreement. • Genomics testing data to investigate different COVID-19 variants and so associated outcomes. Note, these data refer to the genomics of the virus, not the infected individual. The Agreement also authorises reuse a subset of the HES data already supplied under a separate Agreement (ref: DARS-NIC-175120-W5G2X) for the purpose described in this Agreement. JUSTIFICATION FOR DATA USAGE UNDER PURPOSE 1: Additional HES variables and HES Critical Care data will be supplied under this Agreement. Dataset 1: Hospital Episode Statistics (HES) APC, OP and A&E All NHS Digital data will be linked to data on Deaths and demographics (2011 Census) and subsequently to the named external datasets held by ONS to produce statistics on emerging research questions which are arising as the pandemic continues and as understanding of COVID-19 and its impact increases. ONS already hold HES Admitted Patient Care, Outpatient and Accident & Emergency data under DARS-NIC-175120-W5G2X, and will receive monthly updates on an ongoing basis. The purpose section of that Agreement has been updated to be compatible with the reuse of that data for the uses described here. Linkage at a record level is a prerequisite to success for the proposed uses, and therefore identifiers including postcode, date of birth, sex and NHS number are required. ONS does not require identifying details for any other reason than data linkage. ONS require patient-level data because ONS are interested in patient-level socio-demographics, clinical profiles and outcomes. Use of more aggregated (e.g. regional level) data would result in a lack of statistical precision and risks the analysis being subjected to the so-called ecological fallacy. Dataset 1: Hospital Episode Statistics (HES) At the start of the pandemic, ONS already held HES Admitted Patient Care, Outpatient and Accident & Emergency data covering from April 2009 to March 2019, and was to receive annual updates on an ongoing basis. This is covered under a separate Agreement, with the data being used for other statistical purposes. That Agreement has since been updated such that ONS now has HES data from April 2019 to the most recent HES data available, and will continue to receive further HES data on a monthly basis. The purpose section of that Agreement was also updated to be compatible with the reuse of that data for the uses described here. ONS has already used and published analysis using these HES data for several purposes described in this Agreement. Based on learning from this work ONS are requesting further additional variables and historical times series updates where some variables were only requested for a subset of years to the current HES subset they receive. This will be described fully in the separate Agreement and all these data will be reused under this agreement. Therefore, from a HES perspective, the only change under this Agreement is that the HES data ONS holds will be linked to the other datasets described in this Agreement. [1 paragraph unchanged] The HES data ONS already receives includes the Accident and Emergency (A&E) [27 words unchanged] its data collection in 2017. ONS holds HES A&E data covering April 2010 2009 through to March 2020. ONS received an ECDS extract for the financial year 2019/20 and will receive data covering April 2020 onwards, to be updated monthly, that includes: [2 paragraphs unchanged] The specification being requested was developed in collaboration with NHS Digital data [12 words unchanged] coverage, accuracy, relevance) to be likely to support the statistical purpose intended. This is described fully in the separate Agreement (ref: DARS-NIC-175120-W5G2X). [1 paragraph unchanged] These data are still new and clearly very sensitive. Therefore, the specification of the variables ONS are receiving was developed [15 words unchanged] allowed three ONS researchers remote access to the data on NHS Digital systems. systems (ref: DARS-NIC-388794-Z9P3J) [7 paragraphs unchanged] The critical care data will allow ONS to stratify hospitalisations according to [16 words unchanged] outcomes following hospitalisation. ONS used this grading of severity in their long-COVID publication publication, but it is not possible to replicate this using the HES data [18 words unchanged] account the type and level of intensive care support patients were given, eg e.g., for how many days did they receive advanced respiratory support, did the [9 words unchanged] These data will help ONS grade the severity of exposures and outcomes. [2 paragraphs unchanged] ONS requests Critical Care data for the financial years 2009/10 to 2019/20 and data covering April 2020 onwards, to be updated monthly. The years of data for this request are [12 words unchanged] will allow for comparisons over time to assess any COVID specific impact. [1 paragraph unchanged] Version 1.3 of this Agreement amended the original purpose to authorise the reuse of the HES, ECDS and GDPPR data for the following additional purpose. ONS request permission to re-use HES, ECDS and GDPPR data to support QCOVID Analysis . [2 paragraphs unchanged] The data being used for this work are either owned by ONS, [30 words unchanged] with information on characteristics and underlying health conditions at a record level. The data being used so far are being processed and analysed on ONS’s secure data platform. [16 paragraphs unchanged] Linkage at a record level is a prerequisite to success for the proposed use, and therefore identifiers including postcode, date of birth, sex and NHS number are required. ONS does not require identifying details for any other reason than data linkage. The diagnostic and treatment information are included so patients’ comorbidity profiles can be fully explored and controlled for in the work. [2 paragraphs unchanged] ONS have been commissioned to validate the revised QCovid model, meaning that this analysis is still ongoing. [1 paragraph unchanged] ONS are working with NHS Digital to inform the reporting of health [37 words unchanged] in mortality outcomes in waves 1 and 2 of the pandemic which provides. provides valuable insights on which groups are most at risk. This work will support improvements to the reporting and quality of ethnicity health statistics and ultimately support and inform the governments’ response to COVID-19. ONS will investigate the quality of NHS Digital health data sources in their reporting of ethnicity against the ONS’s 2011 census data. The data required for this purpose is limited to the use of ethnicity as recorded in GDPPR and HES data collections which will be compared to the ONS’s 2011 census self-reported ethnicity data. In addition to the ethnicity data personal identifiers, including postcode, date of birth, sex and NHS number, from each of these datasets will be required to link the sources together to investigate the quality. ONS does not require identifying details for any other reason than data linkage. No additional data are requested to that already held by ONS to carry out this work. ONS will investigate the quality of NHS Digital health data sources in their reporting of ethnicity against the ONS’s 2011 census data, and explore methods to improve the quality of estimates using these data sources. ONS request additional permission to reuse and link datasets listed within this agreement to ONS deaths and Census data and to the non-NHS Digital data sources listed above. Specifically, the first phase of this work will include the additional following sources for this purpose: • COVID-19 Ethnic Category Data Set (NHSD) • additional years of past censuses including the 2021 Census once available and appropriate permissions are in place from data owners Depending on the outcome of the first phase of quality assurance additional datasets as listed in Purpose 1 and specifically required for COVID-19 analysis will be used to further assess quality of ethnicity data recorded in the linked data. These data sources add to the range of data on ethnicity available to investigate the quality of NHS Digital health data sources. Linking the sources listed above together will allow ONS to explore the consistency of ethnic group assignment between data sources, and assess reliability. In addition to the ethnicity data personal identifiers, including postcode, date of birth, sex and NHS number, from each of these datasets will be required to link the sources together to investigate the quality. ONS does not require identifying details for any other reason than data linkage or to derive location-based variables for analysis.

Processing activities

Except where specified, the following processing activities are relevant to all three purposes listed above. The present Data Sharing Agreement (DSA) permits the dissemination or re-use of the following identifiable data: Dataset 1: HES data (Admitted Patient Care, Outpatient and Accident & Emergency) • Hospital Episode Statistics Admitted Patient Care (HES APC) Data disseminated under DARS-NIC-175120-W5G2X ONS receives HES data under a separate Data Sharing Agreement (DARS-NIC-175120-W5G2X) which will be reused under this Agreement. • HES A&E data disseminated under DARS-NIC-175120-W5G2X Additional variables from the Admitted Patient Care, Outpatient and Accident & Emergency datasets which are not supplied under Agreement DARS-NIC-175120-W5G2X will be supplied under this Agreement. • HES Outpatient data disseminated under DARS-NIC-175120-W5G2X Datasets 2,3 and 4: Emergency Care Dataset (ECDS), General Practice Extraction Service (GPES) Data for pandemic planning and research (GDPPR) and Critical Care data. • Emergency Care Dataset (ECDS) data disseminated under DARS-NIC-175120-W5G2X As with the HES data which is transferred from NHS Digital to ONS on a monthly basis, these data will be transferred by Secure Electronic File Transfer. • HES Critical Care (CC) Data Once received by ONS, Data security for storage and linkage of the data will be provided within an assured ONS data analysis environment that includes the following elements of security control: • General Practice Extraction Service (GPES) Data for Pandemic Planning and research (GDPPR) including the ‘POSSPOSTCOVID_COD’ cluster and the vaccinations and immunisations cluster • Covid-19 Ethnic Category Data Once this data is received by ONS, data security for storage and linkage of the data will be provided within an assured ONS data analysis environment that includes the following elements of security control: [12 paragraphs unchanged] Following policy specified by the ONS Chief Security Officer, ONS user access to the data environment is only after approval of an application by the Information Asset Owner (IAO) including ethical assessment of proposed data use. A list of approved users is available on request. ONS will keep the number of staff permitted to process identifiers to an absolute minimum and these staff will have a higher level of clearance. All other staff will only be permitted to access non-identifying data. [2 paragraphs unchanged] As described in section 5a, the proposed purposes require linkage of records at the individual level. This is why personal identifiers such as date of birth, postcode and NHS number are required. However, ONS is only interested in producing aggregate statistics and using these to uncover trends and other useful insights based on the non-identifiable ‘attribute’ information. ONS will keep the number of staff permitted to process identifiers to an absolute minimum and these staff will have a higher level of clearance. All other staff will only be permitted to access non-identifying data. Inadvertent re-identification is still a risk but ONS will never seek to intentionally re-identify this data. ONS staff are suitably trained - for example, ONS’s health analysts are experienced working with sensitive data about deaths (such as individual level data about suicides). Further, only statistical disclosure controlled aggregate outputs will be exportable from the secure data analysis environment. In other words, other than the initial transfer of the data from NHS Digital to ONS, the identifiable data will never be in transit and will always be protected by procedural and technical controls: Inadvertent re-identification is still a risk but ONS will never seek to intentionally re-identify this data. ONS staff are suitably trained - for example, ONS’s health analysts in particular are experienced working with sensitive data about deaths (such as individual level data about suicides). Further, only statistical disclosure controlled aggregate outputs will be exportable from the secure data analysis environment. In other words, other than the initial transfer of the data from NHS Digital to ONS, the identifiable data will never be in transit and will always be protected by procedural and technical controls: Access to data held within the Data Access Platform (DAP) is granted to users on a need-to-know basis depending on their role, through a request process which provides a business justification. Access is authorised on a case-by-case basis by the ONS Information Asset Owner (IAO) responsible for the data, with advice from Security and Information Management. Staff requesting access to sensitive data such as these must be cleared to the appropriate National Vetting level, which is higher than the standard basic clearance required for all ONS staff. Only authorised ONS staff with appropriate security clearance will have access to identifiable data with regular audit and monitoring in place to ensure compliance Access to data held within the Data Access Platform (DAP), which includes HES, ECDS and GDPPR data, is granted to users on a need-to-know basis depending on their role, through a request process which provides a business justification. Access is authorised on a case-by-case basis by the ONS Information Asset Owner (IAO) responsible for the data, with advice from Security and Information Management. Staff requesting access to sensitive data such as these must be cleared to the appropriate National Vetting level, which is higher than the standard basic clearance required for all ONS staff. Only authorised ONS staff with appropriate security clearance will have access to identifiable HES, ECDS and GDPPR data, with regular audit and monitoring in place to ensure compliance SPECIFIC PROCESSING ACTIVITIES RELATING TO PURPOSE 2: QCOVID analysis The necessity for the processing of the data for the purposes described in section 5a are largely to do with ensuring the quality and therefore the value of the statistics that can be produced using such complete and record level data compared with less than this (for example, a subset, random sample, or aggregate data). There is more on statistical quality on the ONS website including the following: ‘The quality of a statistical product can be defined as the “fitness for purpose” of that product. More specifically, it is the fitness for purpose with regards to the European Statistical System dimensions of quality: • relevance – is the degree to which a statistical product meets user needs in terms of content and coverage? • accuracy and reliability – how close is the estimated value in the output is to the true result? • timeliness and punctuality – describes the time between the date of publication and the date to which the data refers, and the time between the actual publication and the planned publication of a statistic • accessibility and clarity – is the ease with which users can access data, and the quality and sufficiency of metadata, illustrations and accompanying advice? • coherence and comparability – is the degree to which data derived from different sources or methods, but that refers to the same topic, is similar, and the degree to which data can be compared over time and domain, for example, geographic level? There are additional characteristics that should be considered when thinking about quality. These include output quality trade-offs, user needs and perceptions, performance cost and respondent burden, and confidentiality, transparency and security.’ In the case of this project, it is crucial that the chances of drawing incorrect conclusions are kept to an absolute minimum, and that accurate statistics can be produced quickly. This could save lives. To receive less data than are being requested, or to use aggregate data will lead to less accurate statistics with greater uncertainty. It will also take longer to produce the statistics as methods will need to be used to account for the missing data and/or greater limitations of the data compared with if ONS have access to everything that has been requested. PURPOSE 2: QCOVID analysis [5 paragraphs unchanged] The team conducting the validation study is led by a Principal Statistician [14 words unchanged] as a Research Fellow at the London School of Hygiene and Tropical Medicine, Medicine and has years of experience using complex administrative records to conduct research [16 words unchanged] experience using administrative records and in particular primary care and hospital data. The ONS team will produce metrics as agreed with the Principal Researcher at Oxford University, who is a Professor of Clinical Epidemiology and General Practice, and an active GP with over two decades of experience. She is also the Founder and Director of the QResearch and QSurveillance platforms, one of the world’s largest clinical datasets and a real time infectious disease surveillance system, respectively. The ONS team will produce metrics as agreed with the Principal Researcher at Oxford University. [1 paragraph unchanged] SPECFIC PROCESSING ACTIVITIES RELATING TO PURPOSE 3: Investigating quality of ethnicity data The key data required for this purpose is limited to the use of ethnicity as recorded in GDPPR and HES data collections the datasets listed in purpose 3 which will be compared to the ONS’s 2011 census self-reported ethnicity data. In addition to the ethnicity data socio-demographic variables such as age, sex and location-based indicators created under purpose 3 will be required to quality assure the ethnicity data and explore the distribution of ethnic populations reported in different sources. As well as these analytical variables personal identifiers, including postcode, date of birth, sex and NHS number, from [16 words unchanged] ONS does not require identifying details for any other reason than data linkage. No additional data are requested linkage or to that already held by ONS to carry out this work. derive location-based variables for analysis.

Expected output

PURPOSE 1: Emerging questions relating to COVID-19 The outputs expected from this work are statistics, statistical reports and analyses that can be used to support the on-going response to the pandemic. Should outputs contain any data, this data will be aggregated with small numbers supressed in line with HES Analysis Guidance. The outputs will typically be official statistics, reports and/or briefings for technical/expert audiences such as SAGE. EXPECTED AND YIELDED OUTPUTS RELATING TO PURPOSE 1: Emerging questions relating to COVID-19 Any official statistics produced will be shared with MPs, members of SAGE and other government officials to inform the Government’s response to the COVID-19 pandemic. Any official statistics produced will be published, for example on the Office for National Statistics website. Briefings for technical/expert audiences such as SAGE will not include any new statistics/figures that are not in the publicly available release (written in such a way as to be accessible to all including the general public). ONS will share any briefings produced for technical/expert audiences with the British Medical Association (BMA) and Royal College of General Practitioners. All data outputs will be subject to any required disclosure control practices. The outputs associated with this purpose will typically be official statistics, reports and/or briefings for technical/expert audiences such as SAGE. Any official statistics produced will be shared with MPs, members of SAGE and other government officials to inform the Government’s response to the COVID-19 pandemic. Any official statistics produced will be published so that they are publicly available, for example on the Office for National Statistics website. Briefings for technical/expert audiences such as SAGE will not include any new statistics/figures that are not in the publicly available release (written in such a way as to be accessible to all including the general public). ONS will share any briefings produced for technical/expert audiences with the British Medical Association (BMA) and Royal College of General Practitioners. [1 paragraph unchanged] • Prevalence A report assessing the prevalence of long-COVID symptoms and COVID-19 complications: https://www.ons.gov.uk/news/statementsandletters/theprevalenceoflongcovidsymptomsandcovid19complications • A report assessing Coronavirus related deaths by ethnic group: https://www.ons.gov.uk/peoplepopulationandcommunity/birthsdeathsandmarriages/deaths/articles/coronaviruscovid19relateddeathsbyethnicgroupenglandandwales/2march2020to15may2020 • Updating Updated coronavirus related deaths by ethnic group and investigation of the explanatory power of hospital-based comorbidity on ethnic differences: https://www.ons.gov.uk/peoplepopulationandcommunity/birthsdeathsandmarriages/deaths/articles/updatingethniccontrastsindeathsinvolvingthecoronaviruscovid19englandandwales/deathsoccurring2marchto28july2020 EXPECTED AND YIELDED OUTPUTS RELATING TO PURPOSE 2: QCOVID analysis The output of the exercise to validate the QCOVID algorithm will be [30 words unchanged] will be used (or continue to be used) based on a threshold. ONS have provided initial concordance statistics to the University of Oxford, but they will now be providing concordance statistics for revised versions of the QCOVID Algorithm. The result of the exercise will be shared with the Scientific Advisory Group for Emergencies (SAGE) and with PHE and the University of Oxford. The bulk of the analysis has now been complete complete, but ONS still require continued access to these data to carry out any required revisions. revision and to quality assure any revisions to the algorithm. PURPOSE 3: Investigating quality of ethnicity data Outside of the work carried out by ONS, it is expected that there will be several journal articles published in relation to the QCOVID Algorithm. The output would be a quality report providing aggregate data comparisons of the different sources. This could include for example ethnic groups presented by age and sex (from the census data) to understand if reporting of ethnicity varied across socio-demographic characteristics. This would initially be an internal report for ONS and NHS Digital to use to inform their work and ensure appropriate reporting and interpretation of health statistics by ethnicity. If appropriate it could be published as a technical brief on either the ONS or NHS Digital website. EXPECTED OUTPUTS RELATING TO PURPOSE 3: Investigating quality of ethnicity data The output expected from this purpose takes the form of a quality report providing aggregate data comparisons of the different sources. This could include for example ethnic groups presented by age and sex to understand if reporting of ethnicity varied across socio-demographic characteristics. This would initially be an internal report for ONS, NHS Digital and appropriate partners to use to inform their work and ensure appropriate reporting and interpretation of health statistics by and COVID-19 analysis ethnicity. If appropriate it could be published as a technical brief on either the ONS website or another suitable website. The ONS will then explore the consistency of ethnic group assignment within these data sources over time as well as assess the consistency between data sources. ONS would aim to identify what individual factors influence the propensity for ethnic group ‘change’ (e.g., age and sex). Again, this analysis would first be an internal report, and ONS would seek to publish as appropriate. The ultimate aim of this research would be to produce methodological guidance to account for observed bias and missingness when analysing the health datasets included in this research.

Expected measurable benefits

PURPOSE 1: Emerging questions relating to COVID-19 COVID-19; [10 paragraphs unchanged]

Benefits reported

The data processed under this Agreement has already allowed work to extend initial analysis carried out by ONS on their linked Census and Mortality data looking at coronavirus related deaths by ethnic group, by linking in HES data specifically to investigate the explanatory power of hospital-based comorbidity on ethnic differences. This has been published on the ONS website and has been used to inform government policy and health campaign decisions: https://www.ons.gov.uk/peoplepopulationandcommunity/birthsdeathsandmarriages/deaths/articles/updatingethniccontrastsindeathsinvolvingthecoronaviruscovid19englandandwales/deathsoccurring2marchto28july2020 ONS has worked on a range out topics and outputs in line with this agreement. These statistics and analysis have been used to inform government policy and health campaign decisions, impacts include: This analysis has been further extended to take advantage of the acquisition of the GDPPR data in controlling for health status and studying ethnic variations in the mortality risk in the first and second wave of the pandemic. This has been published as a pre-print article and can be used to inform government policy and health campaign decisions: https://www.medrxiv.org/content/10.1101/2021.02.03.21251004v1.full.pdf Wide media coverage, for example highlighting inequalities in COVID-19 outcomes: Though revisions may yet be required, initial validation of the QCOVID algorithm has been completed with results shared as agreed with the Scientific Advisory Group for Emergencies (SAGE) and with PHE and the University of Oxford. • “Coronavirus: Black Britons face ‘twice the risk of death’ say ONS” was the top news story on the BBC website on the day of release Policy impact • UK Government ordered a review of the disproportionate impact of COVID-19 on Black, Asian and Minority Ethnic (BAME) communities • Community-led interventions to reduce inequalities and raise awareness • Regular updates and further analyses provided to SAGE ethnicity subgroup COVID-19 risk factors ONS examined inequalities in COVID-19 mortality by ethnicity, religion and disability. Using the data from the ONS Public Health Data Asset (Linked 2011 census, primary care and hospitalisation records, death registration data), ONS estimated mortality rates and examined whether these differences were driven by other socio-demographic factors. ONS also investigated how COVID-19 mortality varied by occupation, and whether these differences were driven by non-workplace factors. This work has furthered understanding of the interaction between COVID-19 in the following ways: • Ethnic minority groups were at an elevated risk of COVID-19 mortality, however lockdowns were associated with reductions in excess mortality risk in these groups. • Disabled people have a marked risk of mortality involving COVID-19 compared to non-disabled people, and as a result were prioritised within the pandemic response. • The risk of COVID-19 mortality was explained by controlling for sociodemographic and geographic determinants, however, those of Jewish affiliation remained at higher risk of death compared to all other groups. • Elderly people living with younger people are at increased risk of COVID-19 mortality, and that this is a contributing factor to the excess risks experienced by older South Asian women compared to White Women. • Working conditions are likely to play a role in COVID-19 mortality, particular where workers are coming into contact with COVID-19 patients or the public. The above findings have been used, and will continue to be used to inform the governments on-going response to the pandemic. Validation of the QCovid risk model Commissioned by the Chief Medical Officer for England, ONS validated the novel clinical risk prediction model QCovid to identify risks of short-term severe outcomes due to COVID-19. Following success of this validation ONS have been commissioned to validate the revised QCovid model. This model has been shown to perform well, and showed high levels of discrimination for COVID-19 deaths in men and women. Inequality in vaccination coverage Linking the ONS Public Health Data Asset to National Immunisation Management System (NIMS) data ONS investigated inequality in the coverage of vaccination against COVID-19, focusing on a range of sociodemographic characteristics, such as ethnicity, religion, disability and deprivation. One ONS study showed that populations that are most likely to be disproportionately affected by COVID-19 seemed most hesitant to vaccination. The government was able to adapt policies to improve these disparities.

Text removed is struck through; text added is underlined. Unchanged paragraphs are summarised rather than repeated.

Datasets: + HES-ID to MPS-ID HES Accident and Emergency; + HES-ID to MPS-ID HES Admitted Patient Care; + HES-ID to MPS-ID HES Outpatients; + Hospital Episode Statistics Critical Care (HES Critical Care)

Objective for processing

The Office for National Statistics (ONS) is working on urgent analysis to determine the population-level relative risk of hospitalisation or death that COVID-19 presents to different people. This is being achieved by linking information on outcomes with information on characteristics and underlying health conditions at a record level. The data being used so far are being processed and analysed on ONS’s secure data platform. These data are either owned by ONS or have been acquired through its statutory powers. The Office for National Statistics (ONS) requires data for three purposes all relating to the COVID-19 pandemic and which are described here as PURPOSE 1, PURPOSE 2 and PURPOSE 3. All the data requested is necessary for the production of official statistics. However, there are some important information gaps in the data that ONS has linked and has analysed so far. Primarily, this is to do with comorbidities. Without a complete picture of comorbidities, it is not possible to explain all of the differences being found in COVID-19 related mortality and morbidity between different groups/characteristics (such as people of different ethnicity). All 3 purposes are considered to be tasks in the public interest. ONS is the sole data controller and will process personal data for this purpose under Articles 6(1)(e) and 9(2)(j) of the General Data Protection Regulation. The information that has already been linked and is being analysed by ONS are as follows: No other organisation will process the data under this Agreement. The data shared with ONS will not be onwardly disseminated or shared under this Agreement, except as disclosure controlled aggregate statistics and/or analysis as aggregated data with small numbers suppressed, in line with the Hospital Episode Statistics Analysis Guide. • Information on the outcome of death comes from the death registration data ONS already processes and regularly publishes statistics on. PURPOSE 1: Emerging questions relating to COVID-19 • Information on socio-demographic characteristics such as ethnicity comes from the Census 2011. The Office for National Statistics (ONS) has been asked to provide rapid responses to the coronavirus pandemic on emerging research questions which are arising as the pandemic continues and as understanding of COVID-19 and its impact increases. This work is to support the ongoing government response and has been requested by central government leaders and advisors such as SAGE and the government, via the National Statistician. The work will improve understanding of and support the development of statistics on what the data can be used to learn about: • Information on hospitalisation (serious illness) from COVID-19 since February 2020 as an outcome, and information on underlying conditions that resulted in hospital contact since 2017/18, come from Hospital Episodes Statistics (HES) data. • the short, medium and long-term impacts of having had COVID-19, treatment for COVID-19 or a COVID-19 vaccine ONS already receives this HES data from NHS Digital under a separate Data Sharing Agreement. ONS compelled NHS Digital to share this HES data with ONS for the purposes of official statistics under Section 45C of the Statistics and Registration Services Act (2008), as amended by the Digital Economy Act (2017), in January 2019. • the coronavirus pandemic and its associated social, economic and environmental impacts has had on health and well-being. ONS is currently unable to control for all comorbidities in statistical models using only the HES data. This is because hospital attendances will only represent the most serious cases, with minor illnesses or managed chronic conditions being handled solely in primary care. These missing comorbidities will be mediating some of the differences that have been found between different groups/characteristics. Some examples of the work ONS has been asked to carry out follows. To overcome this limitation, ONS is seeking to include primary care data for all or most of the population in its models This will involve ONS acquiring and transferring the data onto its secure systems where it can be linked to the other data sources described at a record level. Initial work described in a previous version of this Agreement on analysis to determine the population-level relative risk of hospitalisation or death that COVID-19 presents to different people continues. Data have been published and subsequently updated with the latest deaths data on mortality involving COVID-19 by ethnic group using the power of hospital-based comorbidity data. Variation by religion and disability have also been published. Current work continues to use the linked data and will next incorporate primary care data on comorbidities and risk factors. For example, ONS will shortly be publishing an article on the relationship between COVID-19 mortality and disability using data from GDPPR (to define comorbidities as well as learning disability as an exposure of interest). In addition, a change in reporting and data collection by NHS Digital means that their A&E dataset within HES has been discontinued and the more detailed Emergency Care Dataset (ECDS) has been stood up to replace it. Therefore, ONS are also seeking access to an extract of the ECDS data, both to replace the A&E information on a like for like basis, and take advantage of some additional useful information that is new to ECDS compared with HES A&E. As a second example, ONS has been tasked to continue analysis around the prevalence and risk factors for ‘long-COVID’ i.e. symptoms and conditions that persist/develop beyond the acute phase. This analysis is intended to inform public health messaging and interventions relating to long-COVID. Initial analysis has been carried out and published (on the ONS website and as an academic article) through ONS’s remote access (DARS-NIC-388794-Z9P3J-v2.3) and this learning and analysis will be transferred to ONS systems and widened to enable further analysis using linked census data to explore socio-demographic variation in outcomes. The Government’s Scientific Advisory Group for Emergencies (SAGE) and the National Statistician are all clear that they want this improvement to the project to be enabled by NHS Digital. As a result, both ONS and NHS Digital are coming under significant pressure share relevant data quickly. More importantly, the insight gained from an improved study will inform decision making that could ultimately save lives. As another example, ONS has been requested as a priority to look at the impact of the pandemic and the lockdown on mental health for the whole population. This work will initially focus on the GDPPR data working with NHS Digital to identify common mental health conditions in these new data and has started through ONSs remote access. The work includes understanding the characteristics of the population at risk, identifying specific conditions, identifying new diagnoses and identifying changes in existing conditions. The analysis will be transferred to ONSs systems and widened to enable further analysis using linked census data to explore socio-demographic variation in outcomes. This is a task in the public interest. ONS is the sole data controller and will process personal data for this purpose under Articles 6(1)(e) and 9(2)(j) of the General Data Protection Regulation. As the coronavirus vaccination rolls out new research questions are emerging and statistics being requested. ONS will be receiving limited data on COVID-19 vaccinations from NHS England and linking these to ONS mortality, census data and Coronavirus Infection Survey (CIS) data. ONS request permission here to additionally link these data to the NHS Digital data outlined in this agreement to be used in a range of its COVID-19 analyses. Linking these data will allow ONS to include vaccinations in the ONS COVID-19 risk models. This could be as (i) the exposure of interest (i.e. is vaccination associated with reduced risk of adverse outcomes); (ii) a confounder to be controlled for (i.e. does the relationship between socio-economic exposure and outcome still hold after controlling for vaccination status); or (iii) an outcome (i.e. are certain socio-demographic groups more/less likely to take-up vaccination). For example, initial work includes estimating the difference in efficacy between the different vaccines types and investigating approaches to estimate the causal effect of vaccination on mortality. Aside from ONS, no other organisation will process the data under this Agreement. The data being used so far are being processed and analysed on ONS’s secure data platform. These data are either owned by ONS or have been acquired under previous versions of this Data Sharing Agreement or under separate Data Sharing Agreements. This Agreement authorises ONS to receive extracts of General Practice Extraction Service (GPES) Data for Pandemic Planning and research (GDPPR) data, and Emergency Care Data Set (ECDS). The ECDS and GDPPR data will be used solely for the purposes described in this Agreement. ONS have already been able to fill some important gaps in the data primarily, this is to do with comorbidities. Without a complete picture of comorbidities, it is not possible to explain all of the differences being found in COVID-19 related mortality and morbidity between different groups/characteristics (such as people of different ethnicity). This gap has been filled by the acquisition and linkage of HES, ECDS and GDPPR data to the ONS-owned census and deaths data. Based on ONS’s experience and evidence of the utility of the data already analysed and linked in the NHS Digital remote access area (DARS-NIC-388794-Z9P3J) and the evolving data landscape, ONS are requesting additional data and linkage permissions to support their ongoing response to the COVID-19 pandemic. Under the previous version of this Agreement, ONS was approved to link the following NHS Digital data with deaths and demographics (2011 Census): • Hospital Episode Statistics (HES) Admitted Patient Care and Outpatient subsets 2017/18 to 2020/21 • HES Accident & Emergency 2017/18-2018/19 • Emergency Care DataSet (ECDS) 2019/20-2020/21 • General Practice Extraction Service (GPES) Data for Pandemic Planning and research (GDPPR) This Agreement will additionally permit ONS to receive or reuse and link: • Hospital Episode Statistics (HES) Admitted Patient Care and Outpatient subsets 2009/10 to 2016/17 and provisional monthly data for 2021/22 • HES Accident & Emergency 2009/10 to 2016/17 and 2019/20 • Emergency Care DataSet (ECDS) provisional monthly data for 2021/22 • HES critical care data 2009/10 to 2019/20 annual data and provisional monthly data for 2021/22 • Additional HES variables (Operation and Diagnosis Codes) and the historical time series updates as received under DARS-NIC-175120-W5G2X • New long-COVID SNOMED codes developed and collected as part of the GDPPR refresh • GDPPR data which identifies COVID-19 vaccination ONS also hold or are looking to acquire further non-NHS Digital data to support the COVID-19 analyses and the purposes described in this agreement and request permission to link these external data to the NHS Digital data held or reused as part of this agreement along with the link to ONS deaths and census data. Specifically, ONS seek permission to link to the following sources once appropriate permissions are in place from their respective data owners: • COVID-19 Vaccinations data from NHS England • Test and Trace data from PHE • Coronavirus Infection Survey (CIS) from ONS & University of Oxford • Property Attributes data from Valuation Office Agency (VOA) • Energy Performance Certificate (EPC) data on housing from MHCLG The datasets ONS hold or are seeking to acquire are required to be linked to each other in order to answer a wide range of emerging questions about the impact of the pandemic. For example: • Mortality data identifies information on the outcome of death including COVID-19 • Information on socio-demographic characteristics such as ethnicity comes from the Census 2011 • HES & ECDS data can identify hospitalisations (severe illness) as well as more severe existing/subsequent conditions or comorbidities record • HES Critical care data can allow grading of morbidity according to severity, e.g. to stratify analysis of COVID-19 patients into those hospitalised with and without ICU admission and into type and level of intensive care support patients were given • GDPPR data provide up to date and a more complete history of conditions, comorbidities and risk factors and so the underlying health of the population, as well as diagnoses of post-COVID syndrome ("long-COVID") as an outcome following infection • COVID-19 testing data identifies people testing positive for COVID-19 including those who may not have been in contact with primary or secondary care services • COVID-19 vaccinations data identifies people who have received a coronavirus vaccine and could be used as a control variable when assessing outcomes as well as to explore vaccinations take up and efficacy • Property attributes and housing data provides information on property type and size which could impact or help explain infection spread Taken together these data will allow ONS to build a rich picture of the health state and experiences of the population due to COVID-19 and answer emerging questions on the impact of the pandemic. As the pandemic continues and understanding of coronavirus increases and changes, ONS need this ability to respond to emerging questions quickly. ONS will use the data under this Agreement to produce rapid responses to emerging questions driven by clinical issues relating to the COVID-19 pandemic as requested by either the Scientific Advisory Group for Emergencies (SAGE) or the Chief Medical Officer (CMO) via the National Statistician. As new questions emerge which require investigation using the data under this Agreement, ONS will submit a briefing paper to NHS Digital outlining the question(s) to be answered and identify the source of the request as either SAGE or the CMO. ONS already receives the HES data from NHS Digital under a separate Data Sharing Agreement. ONS compelled NHS Digital to share this HES data with ONS for the purposes of official statistics under Section 45C of the Statistics and Registration Services Act (2008), as amended by the Digital Economy Act (2017), in January 2019. The Government’s Scientific Advisory Group for Emergencies (SAGE) and the National Statistician are all clear that they want this improvement to the breadth of ONSs COVID-19 analyses to be enabled by NHS Digital. The insight gained from this improved linked data asset will inform decision making that could ultimately save lives. This Agreement authorises ONS to continue to receive extracts of General Practice Extraction Service (GPES) Data for Pandemic Planning and research (GDPPR) data (quarterly), and Emergency Care Data Set (ECDS) (monthly) and to receive extracts of the Critical Care Data (monthly) in line with HES data already flowing. These data will be used solely for the purposes described in this Agreement. [1 paragraph unchanged] More detail on purpose broken down by dataset: Additional HES variables and HES Critical Care data will be supplied under this Agreement. The HES, ECDS and GDPPR All NHS Digital data will be linked to data on Deaths and demographics (2011 Census) and subsequently to the named external datasets held by ONS to produce statistics on emerging research questions which are arising as the risk factors, including comorbidities, associated with COVID-19. pandemic continues and as understanding of COVID-19 and its impact increases. Linkage at a record level is a prerequisite to success for the proposed use, uses, and therefore identifiers including postcode, date of birth, sex and NHS number are required. ONS does not require identifying details for any other reason than data linkage. [1 paragraph unchanged] At present the work ONS has done is only able to control for decade-old socio-demographic factors in its COVID-19 risk models. Up-to-date primary care data on clinical diagnoses, treatments and histories would allow ONS to substantially enhance the risk models, as comorbidities are likely to explain relatively large proportion of the variability in COVID-19 mortality risk. Whilst certain conditions are known to be risk factors for COVID-19 mortality/morbidity (e.g. patients with severe lung conditions or those on immunosuppressants), it is necessary to have access to the full range of diagnostic and treatment codes (ie both HES and GDPPR data) so patients’ comorbidity profiles can be fully explored and controlled for in the models. [1 paragraph unchanged] Hospital episodes data serves two purposes in the project: At the start of the pandemic, ONS already held HES Admitted Patient Care, Outpatient and Accident & Emergency data covering from April 2009 to March 2019, and was to receive annual updates on an ongoing basis. This is covered under a separate Agreement, with the data being used for other statistical purposes. That Agreement has since been updated such that ONS now has HES data from April 2019 to the most recent HES data available, and will continue to receive further HES data on a monthly basis. The purpose section of that Agreement was also updated to be compatible with the reuse of that data for the uses described here. ONS has already used and published analysis using these HES data for several purposes described in this Agreement. Based on learning from this work ONS are requesting further additional variables and historical times series updates where some variables were only requested for a subset of years to the current HES subset they receive. This will be described fully in the separate Agreement and all these data will be reused under this agreement. Firstly, once linked to Census and mortality data at a record level, HES data (since 2017/18) provides insight into some of the pre-existing conditions (i.e. those that require hospital contact) for those whose death was caused by or involved COVID-19. ONS analysts can then produce statistics on differences by comorbidity, and control for comorbidity when modelling differences between socioeconomic groups. Therefore, from a HES perspective, the only change under this Agreement is that the HES data ONS holds will be linked to the other datasets described in this Agreement. Secondly, more recent HES data (since the March 2020) allows ONS to identify incidences where people are hospitalised because of COVID-19 but then recover. ONS analysts can then look at ‘serious illness from COVID-19’ as an outcome within the population in addition to simply modelling the binary outcome of death (died / did not die). At the start of the pandemic, ONS already held HES data covering from April 2010 to March 2019, and was to receive annual updates on an ongoing basis. This is covered under a separate Agreement, with the data being used for other statistical purposes. That Agreement has since been updated such that ONS now has HES data from April 2019 to the most recent HES data available, and will continue to receive further HES data on a monthly basis. The purpose section of that Agreement was also updated to reflect the uses described here. Therefore, from a HES perspective, the only change permitted under this Agreement is that the HES data ONS holds will be linked to the ECDS and GDPPR data should ONS be granted access to the extracts being sought as per below. [1 paragraph unchanged] These data are being requested for the same purpose and reasons and described for the HES data above. [1 paragraph unchanged] ONS requests access to received an ECDS extract for the financial year 2019/20 and will receive data covering April 2020 onwards, to be updated monthly, that includes: [4 paragraphs unchanged] These data are still new and clearly very sensitive. Therefore, the specification of the variables being requested has been ONS are receiving was developed in collaboration with NHS Digital data experts and through a separate [5 words unchanged] three ONS researchers remote access to the data on NHS Digital systems. [1 paragraph unchanged] This work has helped ONS come to a decision about if and what parts of the GP data need needed to be transferred to ONS systems to enable ONS’ COVID-19 risk modelling project (see ONS to meet the proposed use below). purposes. ONS is therefore confident that the data specification being requested has been minimised to that absolutely necessary, in line with GDPR principles. More information on how the remote access to GDPPR was used to help confirm the need for GP data and minimise the request are detailed below. The GDPPR data will be linked to data on Deaths, demographics (2011 Census) and hospital data (see dataset 2 above) to establish and assess comorbidities and risk factors associated with COVID-19. Since this first flow of GDPPR data there have been several developments which have highlighted the importance of a full and robust set of primary care data. The concept of long-covid has emerged, NICE has agreed a set of SNOMED codes for use in primary care and NHS Digital has requested primary care IT suppliers to implement these in GP systems for future supplies. Additionally, COVID-19 vaccinations are being rolled out. ONS request that the specification for the GDPPR data flows be updated, when the data are made available, to include these aspects to be used for the purposes described in this Agreement, this includes codes in the: The data are needed to understand the full range of comorbidities, patient history and risk factors which could influence COVID-19 outcomes. For example, diabetes and asthma sufferers may well be managing their condition in consultation with their GP, and have had no recent hospital contact (and therefore do not appear in HES). These missing comorbidities will mediate some of the variations in mortality that the project has found so far. • ‘POSSPOSTCOVID_COD’ cluster For example, ONS has found significant differences in mortality risk between different ethnic groups, and it is not yet fully understood what is causing this. Differences in prevalence of different comorbidities between ethnic groups is likely to be one reason, and ONS can only assess its contribution with complete comorbidity data. • COVID-19 Vaccination under the vaccinations and immunisations cluster This work is of critical priority across Government as part of the UK’s response to the COVID-19 pandemic. It will contribute to the wider understanding of the virus, helping to inform a range of policy decisions taken by central government, health services and others. Optimising such decision making could ultimately save lives. Dataset 4: Critical Care Data Data Minimisation The specification of the data ONS is requesting was developed based on knowledge of the data through ONS’s use of these data via a separate Agreement which allowed ONS researchers remote access to the data on NHS Digital systems (DARS-NIC-388794-Z9P3J). The ONS researchers used this access, to learn about the Critical care data and to run some analyses by combining the data with mortality, HES and GDPPR data which are also present on NHS Digital systems. Initial analysis using these data on long-COVID has been carried out and published (on the ONS website and as an academic article). Prior to ONS requesting extracts of the ECDS and GDPPR datasets, ONS and NHS Digital agreed that some groundwork was needed. The critical care data will allow ONS to stratify hospitalisations according to severity (ICU vs. non-ICU admission), which will be useful when defining pre-existing conditions, COVID-19 exposure, and outcomes following hospitalisation. ONS used this grading of severity in their long-COVID publication but it is not possible to replicate this using the HES data to which ONS currently have access in ONS. The data requested will also allow ONS to take into account the type and level of intensive care support patients were given, eg for how many days did they receive advanced respiratory support, did the patient require level 2 or level 3 care, etc. These data will help ONS grade the severity of exposures and outcomes. This involved ONS and NHS Digital collaborating to analyse the GPES Data for Pandemic Planning and Research (GDPPR) data on NHS Digital systems, where it was linked to the HES and mortality data (i.e. the only project data missing was the new ECDS and ONS Census data). This work has ensured the data being shared under this Agreement are of sufficient quality (e.g. coverage, accuracy, relevance) and relevance to be likely to support the statistical purpose intended in line with ONS requirements for producing official statistics under the Code of Practice (particularly transparency, quality and improvement). This allowed ONS and NHS Digital analysts to explore and analyse the data in situ, with a view to: ONS are now requesting to receive these data to be transferred to ONS systems so they can widen this initial use of the critical care data to enable further analysis particularly using the linked census data to explore socio-demographic variation in exposures and outcomes. a) confirming there was a strong enough ‘public good’ case for the data being transferred onto ONS systems, and ONS requests Critical Care data for the financial years 2009/10 to 2019/20 and data covering April 2020 onwards, to be updated monthly. The years of data for this request are in line with the current supply of HES data ONS receive and will allow for comparisons over time to assess any COVID specific impact. b) refining and minimising any GDPPR extract/specification which ONS would subsequently want NHS Digital to share via this application PURPOSE 2: QCOVID analysis ONS have now confirmed that the data will be needed for the project. In reaching this conclusion, ONS: Version 1.3 of this Agreement amended the original purpose to authorise the reuse of the HES, ECDS and GDPPR data for the following additional purpose. • Documented how the GP data will improve the risk factor modelling for COVID-19 outcomes • Considered which analyses can only be done through linkage to Census data (and therefore require data to be transferred to ONS), and how crucial these are • Considered whether the Census data could be transferred to NHS Digital as a way to get all the data in the same data platform without the GP data needing to leave NHS Digital • Investigated the general quality and coverage of the GDPPR data • Carried out analysis to confirm if and how the GP data will improve understanding of comorbidities compared with HES data alone • Carried out analysis to confirm if and how the data could be minimised, for example, in respect of years of data, patient groups, cluster codes and identifiable information i) How the GP data will improve the risk factor modelling for COVID-19 outcomes • Without the GP data, ONS analysts are constrained to using hospital based comorbidity and pre-existing conditions in their modelling of risk of COVID-19 deaths, which means they were failing to assign conditions such as diabetes and hypertension to those with such a condition who have either not had a hospital admission in the past three years or where such a condition may not be recorded on the hospital record for those with hospital contact. • GP data and longer-term patient histories of comorbidities and conditions provide a more valid association between a given disease and risk of covid-19 mortality than reliance on only hospital-based comorbidity. See section (v) below. • GP data will also encompass mental health conditions which are obscured when relying on hospital only data. This will provide the opportunity to investigate links between mental illness and COVID-19 mortality. • GP data will give ONS a more accurate measure of duration of pre-existing conditions and potentially in combination with hospital data, a measure of severity of diseases as a means by which vulnerable groups can be identified. • GP data also benefits from risk factors for disease identified on the patient primary care record such as smoking status and obesity which will complement understanding of causal pathways and vulnerability once infected. • As well as improving the way risk factors for COVID-19 outcomes are captured, the GP data may also be used to define the COVID-19 outcomes themselves. In the future there is likely to be policy interest in the impact of COVID-19 on the population’s physical and mental health. For example, is a COVID-19 diagnosis associated with elevated risk of subsequent respiratory illness, and are individuals in certain socio-economic groups more likely to experience post-lockdown deterioration in mental health. ii) Which analyses can only be done only through linkage to census data, and how crucial these are • The Census provides a population of interest at a point in time to follow-up people from – a list of everyone who was in the country on a particular day that can then be followed up through time. ONS cannot get a complete, point-in-time index population from administrative sources such as GDPPR; certain parts of the population are less likely to have frequent contact with a GP, and these individuals may also be at systematically greater/lesser risk of COVID-19 mortality, so their omission from a GP-based study population would bias estimates. • Linkage to Census allows investigation of the risk of a hospital episode for COVID-19 treatment given a/set of pre-existing condition(s) and how this varies across Census characteristics such as ethnicity, socio-economic position, household type. Given a specific or combination of conditions and population density locale, are there differences in likelihood of hospital admission for COVID-19 by ethnic group? If population density is a good indicator of infection risk, and pre-existing conditions a good indicator of prognosis then this will inform whether ethnic group differences persist given someone’s location and clinical history. • How much further do population level measures of pre-existing conditions explain existing ethnic contrasts in COVID-19 mortality? Is there an interaction between comorbidity and ethnicity in risk of COVID-19 mortality? • How much does Census based ethnic background assignment match results of previous studies using health record-based measures of ethnic background that adjusted for comorbidity? Ethnicity is well populated on the 2011 Census (the non-response rate was just 3%), but ONS has found the rate of missingness in the GP data to be much greater , reducing the utility of the ethnicity information available in the GDPPR dataset and increasing the need to link to Census. • Establish whether ethnic background has an independent effect on risk of COVID-19 death when set against measures of comorbidity, hospital contact and socio-demographic characteristics available from the Census. Compared to a baseline model including Census data on only ethnicity, age, sex, region and IMD decile (the socio-demographic characteristics available in the GDPPR dataset), the excess risk of COVID-19 mortality amongst ethnic minority groups is attenuated by up to 30% when additional socio-economic, household and occupation Census variables are included in the model. This result indicates the importance of combining clinical variables from HES and GDPPR with the full range of socio-demographic characteristics collected by the Census. • ONS is also interested in exploring the phenomenon of delayed access to hospital care given presence of a pre-existing conditions and what this means for future risk of COVID-19 and all-cause mortality. iii) ONS and NHS D agreed that transfer of census data to NHS Digital was not appropriate. Firstly, ONS has never shared record level Census data for analysis and is committed to keeping all personal information collected in the Census safe and confidential and follow a strict security regime to protect subjects’ data. The public are assured of this in Census publicity materials, and to transfer Census data to NHSD would go against these commitments. This could be detrimental to response rates and the success of Census. In turn, that would be detrimental to decision making based on the Census statistics, with the associated risk of significant human and financial costs. Secondly, given the extent of the work already carried out within ONS systems to enable linkage of health data to census data there would be significant delay if Census data were transferred to NHS Digital systems and this work had to be repeated. This would be detrimental to pandemic response and decision-making for officials who have been calling for further analysis from ONS. Details of the linkage methods have been published in a technical document: https://www.ons.gov.uk/peoplepopulationandcommunity/birthsdeathsandmarriages/deaths/methodologies/coronavirusrelateddeathsbyethnicgroupenglandandwalesmethodology Thirdly, ONS already securely processes and analyses identifiable Hospital Episode Statistics data, and has been doing so since March 2019. iv) Investigated the general quality and coverage of the GDPPR data Through the remote access previously granted ONS analysts have reviewed the data items, coverage, quality and completeness of GDPPR data in line with ONS requirements for producing official statistics under the Code of Practice (particularly transparency, quality and improvement). ONS have carried out basic quality assurance on the GDPPR data and are content that the data are of sufficient quality to support production of official statistics. Analyses included a check on missingness for key variables, such as date of birth and NHS number, check on population coverage and distributions for key variables, and a review of possible duplicate records. For example, based on a large sample of records there was only one instance of a single NHS Number being associated with multiple dates of birth, and on clerical examination these records appeared to belong to the same individual. It is important to note that assessing the quality of the data is a key requirement to produce official statistics so that the strengths and limitations of the different data items can be understood and applied or mitigated as required. ONS has to undertake preliminary work to assess the appropriateness of a datasets/data sources for use in the production of official statistics. v) Carried out analysis to confirm if and how the GP data will improve our understanding of comorbidities compared with HES data alone ONS analysts aimed to carry out analysis to assess how underlying conditions or comorbidities are associated with outcomes. For example, investigating the association of a history of cardiovascular (CV) conditions on COVID-19 death controlling for age and sex. Based on a sample of patients with recent contact with the primary care system (to manage processing time) ONS compared a 5, 10, 15, 20 and 25-year patient history of CV conditions (based on diagnoses and medication codes) and found that longer (20-year) patient history data added more value to the analysis and strength to any associations than shorter (<20-year) patient histories. ONS’s current access to 3 years of HES data which includes diagnosis codes cannot provide the same vital patient history data which will be used to help predict outcomes. vi) Carried out analysis to confirm if and how we can minimise the data, for example, in respect of years of data, patient groups, cluster codes and identifiable information. ONS considered how it could minimise the data requested as follows: Years of data - Ideally full histories for patients would be required to ensure we capture the full history of comorbidities, but these could be minimised to a 20-year time period prior to the COVID-19 period (i.e. back to 1 January 2000) to improve the predictive value of these data on outcomes (see above) and to allow for a good history of pre-existing conditions for our study population (i.e. people alive on Census day 2011). To note this would be for a 20-year history of date of activity, based on the DATE field in the GDPPR dataset i.e. when the condition / event happened (not when the patient record was updated as per the RECORD_DATE field). Patient groups - The whole population is required because the data are used to predict outcomes associated with COVID-19, which can be serious but remain relatively rare in the general population, and as such ONS needs to ensure it has a big enough population to be able to perform the analysis in a statistically robust and reliable manner. The risk model aims to identify how those experiencing adverse outcomes associated with COVID-19 are characteristically different to the rest of the population, so it requires data on risk factors across different population groups that are representative of the general population, rather than on any specific patient group. The linked study dataset ONS has produced includes approximately 50 million subjects for whom a full Census record is available and for whom it has been possible accurately assign an NHS number to (i.e. those who can then be more easily linked to other datasets that include NHS number such as HES and mortality data). It is currently difficult to specify any further minimisation within the cluster groups requested until full analysis of conditions and comorbidities and any association with COVID-19 outcomes are made. A key part of this project is to understand and assess the quality of the data in the context of producing official statistics, and any further minimisation may hamper such an assessment. Furthermore, the pressing policy questions of today (primarily around risk factors for adverse COVID-19 outcomes) are likely to be different to those of the near future (such as how the prevalence of COVID-19 sequalae might vary between different population groups), and further minimisation at this stage may prevent ONS from being able to answer such questions in a timely and effective manner. Personal identifiable information – ONS have considered how to minimise the requirement for identifiable data through a review of the data linkage and methodology required (i.e. linkage to the data currently held by ONS - HES, deaths and Census data). ONS receive NHS number, date of birth and postcode for the current HES data supply and request the same identifiers for ECDS and GDPPR (with the addition of date of death for GDPPR). Identifiable data will only be used in for the specific purpose of data linkage and quality assurance of that linkage. Further identifiable data such as full name and address would potentially allow ONS to refine the current linkage approach and develop a bespoke method, for example, to link people from the study population (Census base) who have could not be identified in the patient register and given an NHS number. This could also allow for replenishment of the study population with post-2011 arrivals. However, at this time and in consideration of the specific purpose, urgency and needs of this project, and the sensitivity of these data, ONS feel it is sufficient not to include these more detailed identifiers in this request. ONS analysts have worked on the NHSD remote access platform to understand the GDPPR data in more detail, and to refine and minimise any GDPPR extract/specification. The extent of this work has been balanced against the need for timely access to the GDPPR data in response to the urgency of this work to inform the public good. ONS have carried out a full and thorough minimisation exercise. This Agreement permits the reuse of the HES, ECDS and GDPPR data for the following additional purpose. [21 paragraphs unchanged] The data shared with ONS under this Agreement will not be onwardly disseminated or shared, except as disclosure controlled aggregate statistics and/or analysis as aggregated data with small numbers suppressed, in line with the Hospital Episode Statistics Analysis Guide. [1 paragraph unchanged] PURPOSE 3: Investigating quality of ethnicity data ONS are working with NHS Digital to inform the reporting of health statistics by ethnicity. Reporting by ethnicity has become increasingly apparent as a key requirement to help understand inequalities during the coronavirus pandemic. This has again been demonstrated in the work ONS recently published looking at ethnic differences in mortality outcomes in waves 1 and 2 of the pandemic which provides. This work will support improvements to the reporting and quality of ethnicity health statistics and ultimately support and inform the governments’ response to COVID-19. ONS will investigate the quality of NHS Digital health data sources in their reporting of ethnicity against the ONS’s 2011 census data. The data required for this purpose is limited to the use of ethnicity as recorded in GDPPR and HES data collections which will be compared to the ONS’s 2011 census self-reported ethnicity data. In addition to the ethnicity data personal identifiers, including postcode, date of birth, sex and NHS number, from each of these datasets will be required to link the sources together to investigate the quality. ONS does not require identifying details for any other reason than data linkage. No additional data are requested to that already held by ONS to carry out this work.

Processing activities

Dataset 1: HES data Except where specified, the following processing activities are relevant to all three purposes listed above. ONS receives HES data under a separate Data Sharing Agreement (DARS-NIC-175120-W5G2X) which will be reused for this purpose. Dataset 1: HES data (Admitted Patient Care, Outpatient and Accident & Emergency) Datasets 2 and 3: Emergency Care Dataset (ECDS) and General Practice Extraction Service (GPES) Data for pandemic planning and research (GDPPR) ONS receives HES data under a separate Data Sharing Agreement (DARS-NIC-175120-W5G2X) which will be reused under this Agreement. Additional variables from the Admitted Patient Care, Outpatient and Accident & Emergency datasets which are not supplied under Agreement DARS-NIC-175120-W5G2X will be supplied under this Agreement. Datasets 2,3 and 4: Emergency Care Dataset (ECDS), General Practice Extraction Service (GPES) Data for pandemic planning and research (GDPPR) and Critical Care data. [18 paragraphs unchanged] ONS will keep the number of staff permitted to process identifiers to an absolute minimum, minimum and these staff will have a higher level of clearance. All other staff will only be permitted to access non-identifying data. Inadvertent re-identification is still a risk but ONS will never seek to intentionally re-identify this data. ONS staff are suitably trained; trained - for example example, ONS’s health analysts in particular are experienced working with sensitive data about [46 words unchanged] in transit and will always be protected by procedural and technical controls: [10 paragraphs unchanged] PURPOSE 2: QCOVID analysis [2 paragraphs unchanged] ONS will run Oxford University’s QCOVID algorithm against thier their data asset to validate the statistical performance of the QCovid algorithm in [61 words unchanged] or operational purposes will rest with that team and not with ONS. [5 paragraphs unchanged] PURPOSE 3: Investigating quality of ethnicity data The data required for this purpose is limited to the use of ethnicity as recorded in GDPPR and HES data collections which will be compared to the ONS’s 2011 census self-reported ethnicity data. In addition to the ethnicity data personal identifiers, including postcode, date of birth, sex and NHS number, from each of these datasets will be required to link the sources together to investigate the quality. ONS does not require identifying details for any other reason than data linkage. No additional data are requested to that already held by ONS to carry out this work.

Expected output

Statistics produced by the project will be published in the form of reports and aggregate data on the ONS website. PURPOSE 1: Emerging questions relating to COVID-19 These will include analysis of COVID-19 outcomes by socio-demographics, comorbidities and risk factors associated with COVID-19. The outputs will typically be official statistics, reports and/or briefings for technical/expert audiences such as SAGE. This will extend the work already published by the analysis team such as: Any official statistics produced will be shared with MPs, members of SAGE and other government officials to inform the Government’s response to the COVID-19 pandemic. Any official statistics produced will be published, for example on the Office for National Statistics website. Briefings for technical/expert audiences such as SAGE will not include any new statistics/figures that are not in the publicly available release (written in such a way as to be accessible to all including the general public). ONS will share any briefings produced for technical/expert audiences with the British Medical Association (BMA) and Royal College of General Practitioners. All data outputs will be subject to any required disclosure control practices. https://www.ons.gov.uk/peoplepopulationandcommunity/birthsdeathsandmarriages/deaths/articles/coronaviruscovid19relateddeathsbyethnicgroupenglandandwales/2march2020to15may2020 Outputs will extend the work already published by the analysis team such as: Any official statistics produced will be shared with MPs, members of SAGE and other government officials to inform the government’s response to the COVID-19 pandemic. Briefing specific to those users will be produced to accompany the published reports and statistics themselves. • Prevalence of long-COVID symptoms and COVID-19 complications: https://www.ons.gov.uk/news/statementsandletters/theprevalenceoflongcovidsymptomsandcovid19complications • Coronavirus related deaths by ethnic group: https://www.ons.gov.uk/peoplepopulationandcommunity/birthsdeathsandmarriages/deaths/articles/coronaviruscovid19relateddeathsbyethnicgroupenglandandwales/2march2020to15may2020 • Updating coronavirus related deaths by ethnic group and investigation of the explanatory power of hospital-based comorbidity on ethnic differences: https://www.ons.gov.uk/peoplepopulationandcommunity/birthsdeathsandmarriages/deaths/articles/updatingethniccontrastsindeathsinvolvingthecoronaviruscovid19englandandwales/deathsoccurring2marchto28july2020 PURPOSE 2: QCOVID analysis [2 paragraphs unchanged] The bulk of the analysis has now been complete but ONS still require continued access to these data to carry out any required revisions. PURPOSE 3: Investigating quality of ethnicity data The output would be a quality report providing aggregate data comparisons of the different sources. This could include for example ethnic groups presented by age and sex (from the census data) to understand if reporting of ethnicity varied across socio-demographic characteristics. This would initially be an internal report for ONS and NHS Digital to use to inform their work and ensure appropriate reporting and interpretation of health statistics by ethnicity. If appropriate it could be published as a technical brief on either the ONS or NHS Digital website.

Expected measurable benefits

This analysis is of national public health importance and has been requested by the National Statistician, NHS Digital’s Chief Statistician and members of the Scientific Advisory Group for Emergencies (SAGE). The results of the analysis will be used to inform members of SAGE, Members of Parliament (MPs) and other government officials of the differing COVID-19 risk profiles experienced by UK citizens. These statistics will enable the government to refine its policy response to the pandemic using the best evidence available. PURPOSE 1: Emerging questions relating to COVID-19 Analysis of the impact of having had COVID-19 and the impact of the pandemic on society, the economy and the environment will enable the government to better respond to the ongoing public health crisis, for example through tailored public health interventions. This analysis is of national public health importance and has been requested by central government leaders and advisors such as SAGE and the government, via the National Statistician. The results of the analysis will be used to inform members of SAGE, Members of Parliament (MPs) and other government officials of the differing COVID-19 risk profiles experienced by UK citizens. These statistics will enable the government to refine its policy response to the pandemic using the best evidence available. [2 paragraphs unchanged] PURPOSE 2: QCOVID analysis [3 paragraphs unchanged] PURPOSE 3: Investigating quality of ethnicity data Given the public and government interest in addressing health inequalities, and the now well-known raised risk of death from COVID-19 for people of Black and South Asian ethnic background, it is urgent to improve the evidence base around ethnicity and health in the UK. This project will advance understanding of how ethnicity is recorded in key health-related datasets and propose methods to improve comparability of analyses and reliability of statistics based on inconsistent and imperfect ethnic classification. It will therefore lead to better understanding of ethnic disparities in health and their determinants. In the short term, it will also have the immediate benefit of validating data used across government to monitor equality of service delivery for ethnic minority groups.

Benefits reported

Not stated in the previous version; added here.

The data processed under this Agreement has already allowed work to extend initial analysis carried out by ONS on their linked Census and Mortality data looking at coronavirus related deaths by ethnic group, by linking in HES data specifically to investigate the explanatory power of hospital-based comorbidity on ethnic differences. This has been published on the ONS website and has been used to inform government policy and health campaign decisions: https://www.ons.gov.uk/peoplepopulationandcommunity/birthsdeathsandmarriages/deaths/articles/updatingethniccontrastsindeathsinvolvingthecoronaviruscovid19englandandwales/deathsoccurring2marchto28july2020

This analysis has been further extended to take advantage of the acquisition of the GDPPR data in controlling for health status and studying ethnic variations in the mortality risk in the first and second wave of the pandemic. This has been published as a pre-print article and can be used to inform government policy and health campaign decisions: https://www.medrxiv.org/content/10.1101/2021.02.03.21251004v1.full.pdf

Though revisions may yet be required, initial validation of the QCOVID algorithm has been completed with results shared as agreed with the Scientific Advisory Group for Emergencies (SAGE) and with PHE and the University of Oxford.

Text removed is struck through; text added is underlined. Unchanged paragraphs are summarised rather than repeated.

Objective for processing

[85 paragraphs unchanged] Cluster codes – ONS analysts have made an assessment of the high-level cluster code groups and identified 3 that are NOT required for our analysis: • SNOMED codes clusters - Declines, contraindications and other exceptions • SNOMED codes clusters - Review and monitoring • SNOMED codes clusters - Vaccinations and immunisations [4 paragraphs unchanged] This Agreement permits the reuse of the HES, ECDS and GDPPR data for the following additional purpose. The Office for National Statistics (ONS) has been required by the Chief Medical Officer to urgently quality assure a QCOVID algorithm developed by Oxford University to identify clinical vulnerability to COVID-19. This algorithm will be used operationally by Public Health England as a replacement for the previously used shielding list. ONS is confident the information being requested for their part in assuring this algorithm are needed for statistical purposes as detailed below. ONS will run QCOVID against a range of health data assets to validate the statistical performance of the QCovid algorithm in a dataset which is independent of the dataset used to develop the model. The validation will also be undertaken by an independent team of experts, providing the highest level of assurance regarding the performance of the algorithm in line with TRIPOD (Transparent reporting of a multivariable prediction model for individual prognosis or diagnosis) guidance. The data being used for this work are either owned by ONS, have been acquired through its statutory powers, or have been provided to ONS under COPI Notice for this purpose. The required dataset will be produced by linking information on outcomes with information on characteristics and underlying health conditions at a record level. The data being used so far are being processed and analysed on ONS’s secure data platform. The information being linked and used in this dataset by ONS are as follows: • Information on the outcome of death comes from the death registration data ONS already processes and regularly publishes statistics on. • Information on socio-demographic characteristics such as ethnicity comes from the Census 2011. • Information on hospitalisation (serious illness) from COVID-19 as an outcome, and information on underlying conditions that result in hospital contact, come from Hospital Episodes Statistics (HES) data. • General Practice Extraction Service (GPES) Data for pandemic planning and research (GDPPR). • Information on all patients, recorded in the National Radiotherapy Dataset or Systemic Anti-Cancer Therapy Dataset who received treatment (all types) with Radiotherapy between 01/06/2018 to 31/08/2020 or treatment with SACT between 01/06/2018 to 30/06/2020 • Information on persons who received a test result (positive only) for Covid-19 up to and including 27th October 2020 from the PHE Second Generation Surveillance System (SGSS) The subsets of National Radiotherapy Dataset, Systemic Anti-Cancer Therapy Dataset and PHE Second Generation Surveillance System data are being shared directly with ONS by Public Health England under Regulation 3(4) of the Health Service Control of Patient Information Regulations 2002 for the explicit purpose of identifying and understanding information about patients or potential patients with or at risk of Covid-19 through the validation of the statistical performance of the QCovid algorithm, and will be held for no longer than is required for this purpose, currently agreed to be a six month period. The GDPPR data to be used in this work will be limited to: • Patient identifiers • patient demographics • diagnoses and findings • medications and other prescribed items • investigations, tests and results • treatments and outcomes • vaccinations and immunisations Linkage at a record level is a prerequisite to success for the proposed use, and therefore identifiers including postcode, date of birth, sex and NHS number are required. ONS does not require identifying details for any other reason than data linkage. The diagnostic and treatment information are included so patients’ comorbidity profiles can be fully explored and controlled for in the work. This work is of critical priority across Government as part of the UK’s response to the COVID-19 pandemic. It will contribute to the wider understanding of the virus, helping to inform a range of policy decisions taken by central government, health services and others. Optimising such decision making could ultimately save lives. [1 paragraph unchanged] It is possible that there may be a commercial element to the QCOVID algorithm and/or its potential uses noting that ‘QCovid’ is a registered trademark. However, any commercial elements are peripheral to this application. There are no known commercial elements to ONS’ intention to perform an independent validation of the algorithm as requested by SAGE.

Processing activities

[33 paragraphs unchanged] For the purpose of the quality assurance of the QCOVID algorithm, ONS will link the HES, ECDS and GDPPR data with the additional datasets described in the section above. ONS will run the QCOVID algorithm against the linked data sets to validate the statistical performance of the algorithm. ONS will run Oxford University’s QCOVID algorithm against thier data asset to validate the statistical performance of the QCovid algorithm in a dataset which is independent of the dataset used to develop the model. This will generate a c-statistic of between 0.5 and 1 (0.5 = no better at predicting outcome than random chance; 1 = perfect prediction), which will be provided to the researchers and clinicians that developed the algorithm. The subsequent decision on the suitability of the algorithm for clinical or operational purposes will rest with that team and not with ONS. ONS will provide the c-statistic back to the researchers and clinicians that developed the algorithm. The team at Oxford will then decide on the suitability of using the algorithm. ONS is not responsible for determining the threshold for acceptance. ONS' role is to provide validation of the algorithm’s performance against an independent dataset, by an independent team of statistical experts, and not to make a clinical or operational judgement. This statistical quality assurance role will operate in parallel to any operational decisions on the roll out of the QCovid model for clinical purposes, which rest fully with Public Health England. Such decisions are outside the scope of this application. The team conducting the validation study is led by a Principal Statistician at ONS and a member of the Government Statistician Group, who is also working as a Research Fellow at the London School of Hygiene and Tropical Medicine, and has years of experience using complex administrative records to conduct research and analysis. The team includes experienced statisticians and data scientists, some of whom have years of experience using administrative records and in particular primary care and hospital data. The ONS team will produce metrics as agreed with the Principal Researcher at Oxford University, who is a Professor of Clinical Epidemiology and General Practice, and an active GP with over two decades of experience. She is also the Founder and Director of the QResearch and QSurveillance platforms, one of the world’s largest clinical datasets and a real time infectious disease surveillance system, respectively. The data shared with ONS under this Agreement will not be onwardly disseminated or shared, except as disclosure controlled aggregate statistics and/or analysis as aggregated data with small numbers suppressed, in line with the Hospital Episode Statistics Analysis Guide.

Expected output

[5 paragraphs unchanged] The output of the exercise to validate the QCOVID algorithm will be a concordance statistic (c-statistic) of between 0.5 and 1 (0.5 = no better at predicting outcome than random chance; 1 = perfect prediction) with the outcome determining whether the model will be used (or continue to be used) based on a threshold. The result of the exercise will be shared with the Scientific Advisory Group for Emergencies (SAGE) and with PHE and the University of Oxford.

Expected measurable benefits

[3 paragraphs unchanged] The benefit of the exercise to validate the QCOVID algorithm is independent validation of an algorithm which will inform whether the model should be used or continue to be used to support government decision-making. This work is of critical priority across Government as part of the UK’s response to the COVID-19 pandemic. It will contribute to the wider understanding of the virus, helping to inform a range of policy decisions taken by central government, health services and others. Optimising such decision-making could ultimately save lives.

Benefits reported

Stated in the previous version and removed here.

Yielded Benefits is not a requirement for new applications.