On Wednesday 7 December, Mike Keoghan, Deputy National Statistician for Economic, Social and Environmental Statistics, Office for National Statistics, and Darren Morgan, Director for Economic Statistics Production and Analysis, Office for National Statistics, gave evidence to the Science and Technology Committee for their inquiry, ‘Research and Development Statistics’.
A transcript of which has been published on the UK Parliament website.
Dear Greg,
Thank you for inviting me to give evidence to the Science and Technology Committee on 8 June for your inquiry the right to privacy: digital data. I hope you and the Committee accept my sincere apologies for the technical difficulties on the day that meant I was unable to contribute to the session as much as I would have liked. Please see enclosed our answers to the outstanding questions which I hope will assist the Committee with its report.
What progress is being made in addressing weaknesses across the UK’s data ecosystem? Which areas and actions would you prioritise to accelerate further progress?
As we discussed in our written evidence for this inquiry, the Office for National Statistics (ONS) is leading in the delivery of an Integrated Data Service (IDS) in collaboration with partners across government as part of the National Data Strategy (4.2.1.). The IDS will enable faster and lower cost analysis for government through streamlined data access processes and departmental collaboration which will transform ways of working. It releases the power of data to enable government reform based on improved policy decisions. Under robust security and ethical protocols, that again we elaborated on in our written evidence, and through a Trusted Research Environment, the IDS will enable analysts to access and analyse linked data and give them the ability to disseminate a range of data.
On areas to prioritise to accelerate further progress, for the ONS to build a service that is fit for purpose, we need partners across government to feed in their requirements on data, as well as their technical and analytical needs.
How deep seated are risk-averse approaches to data sharing? What is driving this and what role can you play in addressing it?
The Committee is right to consider this issue; as we noted in our written evidence, when it comes to data sharing “we could do more still if various barriers to data sharing were eased, particularly a mindset shift from internal department boundaries and towards risk management at government level, including recognising the risk of valuable data not being shared.” We always balance the potential of data sharing for improving policy and outcomes with stringent ethical and legal safeguards to keep the public’s data safe.
There are multiple drivers to a general risk-averse mindset within Government: data governance, data quality, and a focus on risk over reward.
On data governance, while initiatives like the Central Digital and Data Office’s Playbook team have helped resolve differences in opinions on some key data shares and the IDS will bring a more comprehensive approach to cross Government data sharing, friction still remains. Ownership of data is dispersed and fragmented, with different stewardship frameworks adopted across departments.
Secondly, rapid growth of cross-government data sharing has resulted in increase in demand at a time where the administrative systems that generate data are undergoing profound transformation to improve automation and consistency of data. This, coupled with challenges in recruitment of data professionals, can create a gap between the demand for data to be shared and bottlenecks in the preparation of the data.
More broadly, data sharing between the public and private sector is hindered by an unequal risk profile: the organisation that shares the data assumes (or perceives itself to assume) legal risk, while the recipient reaps the benefit. Work is ongoing to understand how legal frameworks can be adapted to mitigate this risk.
The ONS is leading data sharing efforts across government and has over 400 agreements in place to share data for statistical and research purposes. We are the data hub for government and are working to streamline the existing landscape and build the IDS, to provide a mechanism through which to share data, while proactively engaging with privacy groups and the public about the development, to ensure the new service is supported.
What further steps, if any, are needed to ensure public trust on the sharing of their data, especially with private companies? What are the best mechanisms for engaging with the public on data sharing and how can their views be taken into account?
Secure access to data for the private sector, through the Secure Research Service, was introduced by the ONS in 2015 following a full public consultation, and extensive engagement with privacy groups, to ensure that such use was acceptable and demonstrably in the public interest. They must demonstrate the public good of their research, ethical considerations, and adhere to the accreditation conditions of any other project (including transparency).
The importance of public engagement cannot be understated, and maintaining public trust in how the ONS, Government and Accredited Researchers use data is critical to the continued use of these data for analysis. We do this in a number of ways, including:
Key to this is involving the public in decision-making. For example, through my Data Ethics Advisory Committee, which provides independent ethics advice to the ONS, across government and beyond, and is made up of data ethics experts and lay members who provide an independent public voice to the Committee. The work of the Data Ethics Committee, and various partner committees such as the Research Accreditation Panel, is transparently available to the public through published minutes, blogs from members and registers of activity.
The outreach campaign involved in the 2021 Census was an example of engaging with a wide range of community and representative groups, as well as citizens directly, to build understanding of why the numbers matter, how we keep personal information safe, and build confidence in supplying their data. We are now taking this successful approach and applying it to all of our public engagement work.
We take an insight-led approach and are confident in our understanding of the public’s attitudes, barriers and motivations. We will continue to regularly engage with audiences, through quarterly surveys and 6-monthly focus groups, to ensure we keep up to date on any shifts.
This understanding, for example, includes that the public have higher levels of acceptance for the use of data for the ‘public good’ than for commercial gain, and that both COVID-19 and GDPR seem to have had a positive impact on attitudes and behaviours in relation to data sharing. Sharing personal information is thought to be an inevitable part of modern life, with perceptions that people have little choice in sharing their data, but there is a growing need for reassurance that (linked) admin data will be anonymised. There is a very low awareness about both admin data collection, linkage and use, but this leads to broad acceptance once the concept is introduced depending on why and how it is being used. Finally, the high level of awareness and trust in the ONS, as evidenced in the most recent Public Confidence in Official Statistics report where 89% of respondents able to express a view trusted the ONS and 75% of respondents were aware of the ONS, means we are well positioned as trusted data guardians.
Furthermore, close relationships and collaboration with business, charities, public sector and many other organisations are useful for taking a wider range of views into account, and we recently established the National Statistician’s Expert User Advisory Committee (NSEUAC) to bring us even closer to our users and further build their diverse needs into our approach.
Professor Goldacre told us of the risk of using pseudonymised health data and advocated its eradication. How much of a problem is pseudonymisation across the rest of the UK’s data ecosystem and how can any risks be dealt with?
There are trade-offs when it comes to pseudonymisation. The ONS is a world leader in trying to mitigate both risks: the risk to privacy that Professor Goldacre highlights, and the risk of lower quality research. We have done so by investing in the Five Safes Framework. The wider controls within a TRE and the Five Safes Framework combine to reduce the risk of reidentification and enable the safe use of detailed microdata.
Should Trusted Research Environments be networked and if so, how can this be done safely and effectively?
While in principle Trusted Research Environments (TREs) can be networked, we would not recommend this as this would increase security risk without improving access to, or use of, data.
However, TREs must work closely together to ensure that, where there is a need to combine data held in different environments (for example to create a UK-wide Integrated Data Asset, from data held separately in the four nations, or combine Health and Administrative records), data can be securely transferred, linked, and made available for analysis. This requires that the organisations responsible for the TREs develop the technical mechanisms, and related Information Governance, to enable such collaboration and to so safely, securely, and ethically, in line with established technical, security and capability standards for processing of data for research purposes.
The value of such an approach is recognised by all major TREs in the UK, with good examples of success during the COVID-19 pandemic, and we continue to work closely together. For the ONS, this is an essential component of the IDS.
How should science research be defined for the purposes of data sharing?
The Digital Economy Act 2017 (DEA) Code of Practice does not explicitly define this but does say that the research powers support: “helping researchers and policy-makers build a better understanding of how people live their lives, their patterns of need and use of different services and the resultant outcomes, to support the design and delivery of more effective and efficient public services.”
It explains that data can be used if the purpose “serves the public good”, meaning the use must do one or more of the following:
We would highlight that the focus of this framework is on the use and the public benefits deriving from the research, regardless of professional background or academic discipline.
Do you have any concerns regarding the Government’s proposals regarding the use of AI?
While giving evidence to the Committee on 8 June, we briefly discussed AI and its challenges for transparency. I noted then that I would like to see the various groups thinking about the ethics of AI convened by government to create “a set of clear ethical procedures.” I also highlighted the need for good regulation of AI by regulators with an understanding of both the models and the data being used.
What progress is being made on digital skills in government and what role are you playing in improving them?
We have recognised the need to increase data skills and awareness in government, and through our Data Science Campus, we aim to build data science capability across the public sector. This work is aligned to the National Data Strategy and is achieved through a range of programmes and community engagement. In 2021/22 the Data Science Campus has provided data science learning to over 6,000 people across a wide variety of organisations and experience levels.
At the top of government, the Data Science Masterclass for Senior Leaders is being rolled out to senior leaders, predominantly permanent secretaries and Senior Civil Servants with plans to expand across the public sector. Over 4,500 learners have registered across 48 departments and agencies, with further cohorts planned. We have also developed a core data science foundational curriculum and learning pathway that is delivered across the public sector, including the Civil Service Fast Stream. This is an optional part of the Fast Stream curriculum and mandatory for 300 generalists, thereby increasing the data skills of future leaders.
The Campus also support more technical training such as the specialist Data Science Graduate Programme, a two-year programme covering core data science skills available to existing civil servants and new joiners. We are also working with academic partners to deliver a government designed master’s programme (MDataGov) and to provide placements and secondments for students at all levels.
Finally, we run accelerator and mentoring programmes. In 2021 the accelerator programmes in data science and data visualisation had a cohort of 57 from across a variety of public sector organisations. Our Community Programme supports the data science specialist community across government and recently delivered a hugely successful Data Science Festival which had over 90 speakers, presentations, a hackathon, training sessions, community and social events delivered virtually to over 3,000 participants.
Through my role as Head of the Analysis Function, we aim to integrate analysts in all facets of government. The proliferation of data has expanded the scale of the task but also means analysis can have a greater impact than ever before.
To take advantage of the opportunities presented by analysis (and avoid pitfalls), it is necessary for analytical skills to extend beyond traditional professional boundaries. By operating as a function, professional analysts will act as the catalyst for other professions contributing to analytical insight. This will be built on multi-disciplinary partnerships and developing capability.
To that end, we recently undertook an audit of the analysis skills within the Policy Profession, which highlighted where we need to focus our attention on continuing to mature these skills. We will use the recommendations in the report to continue work together to drive enhanced analytical skills across government.
I hope this is helpful to the Committee, and please do let me know if I can assist further.
Yours sincerely,
Professor Sir Ian Diamond
On 8 June 2022 Professor Sir Ian Diamond, the National Statistician at the Office for National Statistics, gave evidence to the Science and Technology Committee’s inquiry on the right to privacy: digital data.
A transcript of which has been published on the UK Parliament website.
Dear Mr Clark,
On 2 March 2022, I gave oral evidence to the Science and Technology Committee. In my evidence in response to a question about the approach to measuring excess deaths I referred to the Office for National Statistics (ONS) approach, stating that: “For the five-year rolling average excess deaths, ONS has decided for 2022 to drop 2020 from the five-year calculation… It is 2016, 2017, 2018, 2019 and 21, and it has dropped 2020 because 2020 had such an unusual peak of deaths.”
I wanted to clarify that there are a number of ways to measure excess deaths, that is, the difference between the expected number of deaths and the actual number of deaths in a given period of time. The ONS headline approach, shown in the weekly deaths publication is as I described above and has been agreed across the devolved administrations. ONS publications also refer to excess deaths based on the five-year average from 2015 to 2019. ONS has published a blog explaining more about its choice of five-year average.
As the ONS blog notes, more complex methods can also be used to calculate expected deaths. The Office for Health Improvement and Disparities (OHID) uses a more sophisticated modelling approach, taking into consideration the ageing population, differing mortality trends in subgroups of the population and variation in registrations around bank holidays. OHID developed a different method to that of ONS because it needed more accurate data for operational decision making during the pandemic. OHID continues to use the years prior to the pandemic (2015-2019) as the five-year baseline for its measure of excess deaths.
I would be happy to discuss this further if you would find it helpful.
Yours sincerely,
Ed Humpherson
Director General for Regulation
On Wednesday 2 March 2022 Ed Humpherson, Head of Regulation at the Office for Statistics Regulation, gave oral evidence to the Science and Technology Committee’s inquiry on UK science, research and technology capability and influence in global disease outbreaks; specifically on statistics and modelling.
A transcript of which has been published on the UK Parliament website.
Dear Mr Clark,
I write in response to the Science and Technology Committee’s call for evidence for its inquiry ‘the right to privacy: digital data’. Our submission focuses on the questions raised in the call for evidence, including:
This inquiry is particularly relevant to the Office for National Statistics (and, in turn, the UK Statistics Authority) as we progress in two areas: data sharing and data ethics. As detailed in the annexed evidence, the benefits of data sharing are extensive; our accredited Secure Research Service (SRS) currently ensures researchers can access data safely, in line with the five safes framework, and we hope to harness the potential of data sharing further through the Integrated Data Service (IDS), a cross-government initiative we are leading.
Regarding data sharing barriers, negotiations to secure agreement to share data, and the associated work to put in place the legal, security and other arrangements required, can be protracted and complex, including with willing partners. We believe there is potential for elevating the consideration of risk for data sharing beyond silo departmental perspectives to better reflect the value of that data use elsewhere.
When it comes to data, the importance of individual privacy and security cannot be understated, and we elaborate on the many safeguards we put in place to protect this. Finally, following the Public Administration and Constitutional Affairs Committee recommendations in 2019, the Authority has taken a leading role in data ethics, including through the National Statistician’s Data Ethics Committee (NSDEC), and we explain this further in the annexed note.
I hope this is useful, and please do let me know if we can provide further evidence or discuss directly with the Committee.
Yours sincerely,
Alison Pritchard
Deputy National Statistician and Director General for Data Capability, Office for National Statistics
The Office for National Statistics (ONS) has demonstrated the value of data sharing and access to data over the course of the pandemic and will go further through the Integrated Data Service (IDS), all the while within the parameters set out in the Digital Economy Act 2017 (DEA). However, we could do more still if various barriers to data sharing were eased, particularly a mindset shift from internal department boundaries and towards risk management at government level, including recognising the risk of valuable data not being shared.
We are ambitious and radical in our pursuit of greater data access for better analysis (and therefore, better policy); with this in mind, our ethics framework is extensive and carefully considered. In addition to our ethical principles, which we enact through both a self-assessment tool for researchers and the National Statistician’s Data Ethics Committee (NSDEC), we follow the required legal safeguards, and together these ensure the ONS is well-equipped to use, link and share data for the public good.
Effective data use and sharing has many benefits to government and society. An integrated, collaborative approach to data use enables in-depth and efficient analytical and research practices across organisational and sector boundaries. This directly informs government and society, supports the formulation of cost-effective, transparent public policy, and ultimately, delivers better outcomes which can be evaluated.
It is important to note, as we will do in more detail in other sections of this submission, that we only use and share data for research purposes and ensure that data is anonymised as early as possible. Building and ensuring public trust in data sharing is the most important factor when considering the benefits and barriers, and we currently do this through the application of ethical principles and legal safeguards. We are absolutely committed to increasing levels of public engagement and involvement to improve awareness, understanding and trust. We are developing an engagement approach with other key partners so the public not only understand the framework that is in place to safeguard their data but are also alive to the huge benefits that effective data use and sharing can bring.
The ONS is leading a cross government initiative to develop the IDS, which will bring together ready-to-use data to enable faster and wider collaborative analysis for the public good. It will provide the facility to fully exploit the opportunities for safe and secure access to data provided for in the DEA. There is a clear need for such a service: the COVID-19 pandemic illustrated the potential for government and public services to use and share data to help and protect people. When data are shared effectively, the speed at which analysis can be done means time-critical policy issues can be understood and addressed quickly; the ONS demonstrated this on a regular basis throughout the pandemic in partnership across government and with other organisations, using ONS’s trusted research environment (TRE), the Secure Research Service (SRS). Recognising the importance of TREs to research, we have agreed to support and participate in HDR UK’s pan-UK Data Governance Steering group to help streamline data governance approaches for better data linkage for health data research.
The IDS will build on the success of the SRS, which has successfully and securely hosted de-identified data for over 15 years but is reaching its capacity while becoming increasingly costly to run. The IDS will reshape the way that data users share assets and offer controlled access to integrated de-identified data assets, a broader range of data to address policy analysis and evaluation and will work with the Central Digital & Data Office (CDDO) to ensure common data standards, governance, and quality measures. The IDS is a fully cloud-based platform, which will enable connectivity with data where they are held, reducing the friction caused by sharing data multiple times and the complexity of multiple data sharing agreements. This will organically increase the appetite to share data, as it will not need to be moved across the governments data estate and will therefore reduce cost burden on the supplier.
The IDS programme has just concluded its first year, with its most recent success the conclusion of the cloud procurement activity, which led to Google Cloud Platform (GCP) being awarded the contract to take forward the next stage of its development. Migration from the existing provider will conclude in spring and allow for a userbase of 40 across government and the devolved administrations. Throughout 2022, capability, data available and projects will significantly increase, and its userbase will be broadened and increased to around 1000 users by autumn, to include users outside of government.
As part of IDS’s private beta, which launched in September 2021, the service launched 3 collaborative projects with partners across government to further demonstrate the benefits of data use and sharing, as well as having an environment to collaboratively run analysis. These projects, on key topics such as climate change, remain ongoing, and the programme is currently working across government to develop a portfolio of projects with a view to significantly scaling up use of the service in 2022, aligned with the aforementioned numbers.
In the meantime, government can and are already linking administrative data to inform policymaking, using powers from the DEA which facilitates the linking and sharing of de-identified data by public authorities for research purposes. For example, Longitudinal Educational Outcomes (LEO) is a de-identified, person level administrative dataset that brings together education data with employment, benefits, and earnings data from DfE, HESA, DWP and HMRC. We used this data to publish initial findings understanding earnings outcomes for free school meals students, and Ofsted is using it to look at the impact of quality of schooling on long-term labour market outcomes.
Other examples of linked datasets include the Data First programme, which links administrative datasets from across the justice system and beyond for research such as investigating racial bias in court case outcomes in England and Wales. Finally, Growing up in England (GUiE), which links 2011 Census data from the ONS with educational attainment data from DfE will be used by the London School of Economics to build up new quantitative evidence on Gypsy, Roma and Traveller young people. These projects give an idea of the real potential and benefits that can be gained when we effectively use and share data across government and beyond, and the IDS should only make these analyses easier in the future.
While the legal right of access to data is an important enabler, many prospective data suppliers face difficulties associated with the provision of their data, especially when datasets are to be shared for the first time. There are several challenges to data sharing most data suppliers experience: risk appetite; capacity; legal clarity (including GDPR) and governance.
In terms of risk appetite, when we know the environment is safe and secure, there is still too much weight on the risk of data sharing, as opposed to the very real risk of policy harm where valuable data is not being actively used and shared. There needs to be a cultural shift in approach, which the IDS should contribute to, and the Committee’s support will also be useful in this regard.
When we talk about capacity, one issue is personnel turnover, particularly when new decision-makers have differential understanding of legal frameworks or approaches to risk management in the context of data sharing, which can introduce further delays and blockages.
Lack of legal clarity can also play a role. For example, the current approach in data handling is to favour individual departmental boundaries, which is not suited to multi-departmental use of data. Particularly where infrastructure and governance are compliant with data protection legislation, it should be possible to treat such organisations as non-third party. Onward sharing of linked datasets is often another challenge a data supplier may face. Data are often brought together from separate departments to undertake topic-based analysis in areas such as education, labour market or health. If these linked data are then shared, the data governance models of the supplier departments may differ, adding further complexity due to the lack of standard data governance across the government.
These obstacles are not uncommon to most government departments, public bodies, and commercial organisations when it comes to data sharing. The ONS actively supports data suppliers and works with them to explore all options that enables the sharing of data and reduces the burden on the supplier, but there can still be tensions.
The IDS should help begin the shift from this mindset, but we need buy-in and recognition of the opportunity data sharing can present to do this with great success. Happily, the IDS should also mean that other traditional complexities such as the lack of a common gateway for sharing and lack of incentive to share data will soon be rectified. For example, the IDS will look to incentivise departmental sharing through wider access to a broader range of government data from partnering departments. The IDS data stewardship model considers the importance of data security and has the appropriate ethical controls to ensure privacy are at the heart of its design.
The importance of data ethics in the use and sharing of individual’s health data cannot be understated as we enable this data to be used in ever more radical, ambitious, inclusive and sustainable ways, as set out in our strategy. It is also crucial that the UK Statistics Authority (the Authority) guarantees public trust and acceptability and reduces potential harm to individuals involved in research; both of which can only be ensured through the application of ethical principles.
The Authority has developed the following ethical principles, which all ONS research projects must comply with, including uses of health and care data for statistical purposes:
The SRSA (as amended by the DEA) provides the ONS with statutory powers to acquire data for the production of public good statistics. These powers are governed by numerous safeguards to ensure the safety and confidentiality of personal information. For example, the ONS must ensure the data it receives is both necessary and proportionate for the intended purposes and all acquisitions must be GDPR compliant. This is set out within a published Code of Practice and statement of principles. The principles require that the ONS must work collaboratively with suppliers to reach a point of mutual agreement.
The separate but complimentary Code of Practice for Statistics is based on three pillars: trustworthiness, quality and value. Data governance is a key principle of the Code, requiring the ONS to follow their statutory obligations around data collection, confidentiality, and data sharing.
The primary function of the ONS is the production and publication of official statistics for the public good. All data collected and acquired by the ONS may only ever be used for its statutory functions and never for decisions about an individual or business. Data that is collected is anonymised/de-identified at the earliest possible opportunity as the ONS’ focus is never on specific individuals but on the broader application of data to produce statistics that reflect society as a whole.
Before obtaining any data from another organisation, an agreement is put in place between the relevant parties that establishes how the sharing will take place. These agreements ensure the confidentiality of the data is maintained by prescribing secure transfer methods, storage requirements, secure access control and retention periods. Such agreements work to ensure adherence to the safeguards set out in both legislation and relevant Codes of Practice referred to above.
The SRSA contains safeguards to ensure that we do not disclose personal information. Namely, information that identifies a particular person (whether living or deceased) or a body corporate must not be disclosed by the ONS unless there is a statutory exemption that applies.
In accordance with the DEA, the ONS may make certain data it and other organisations hold available to others for research purposes. This is achieved through the accredited platform the SRS. The SRS provides a safe setting, as part of the Five Safes Framework, to protect data confidentiality. The framework is a set of principles adopted by a range of secure labs, including the ONS, which provides complete assurance for data owners. The Five Safes are:
All data processing by the ONS is guided by our Data Strategy ‘to mobilise the power of data to help Britain make better decisions’ and is fully supported by a range of data policies, principles and standards.
Before undertaking projects that require the large-scale processing of personal data or the processing of special category personal data, we undertake data protection impact assessments (DPIAs). This ensures that privacy is built into the design of a project from the outset and allows us to identify and mitigate any privacy concerns as early as possible.
Prior to the publication of statistical information, thorough disclosure control methodology is applied to ensure that we do not reveal any personal information. This requires a careful balance of risk and utility of the data, i.e. ensuring that the data is still functional for users, but that sufficient safeguards and measures have been applied to eliminate the risk of disclosing personal or sensitive data.
The ONS is committed to full transparency in relation to its processing of identifiable data. Extensive privacy information is made available to data subjects, whether through dedicated privacy information for survey respondents or more generally. We have a published log of data acquisitions on the ONS website. Every effort is made to ensure data subjects are aware of how to contact us for further information or to exercise their data subject rights.
Our ONS Security framework both governs and operates a security strategy, security principles and a security policy framework, which incorporates and references appropriate recognised security standards and guidance from within UK Government (Cabinet Office, National Cyber Security Centre [NCSC], Centre for Protection of National Infrastructure [CPNI]) and international standards.
We have a dedicated data protection officer who advises the organisation in line with legislation and the organisation’s data protection policy, supported by an experienced and qualified team that manage information rights requests and staff training in security and data protection.
We also lead several committees that focus on scrutinising how data is used across the organisation including the Data Governance Committee and the National Statistician’s Data Ethics Committee (NSDEC).
There has been much interest in data ethics over the last couple of years. Much of this work has focused on the ‘theory’ of data ethics and has fallen short of making applied decisions about the ethics of the use of data for analysis.
The Authority’s work on data ethics, through the UK Statistics Authority’s Centre for Applied Data Ethics, has focused on supporting analysts to apply high level ethical principles/frameworks to their work and providing them with the support to be able to do this efficiently, to enable timely research that is ethically appropriate. This has involved creating and user supporting an ethics self-assessment tool, which empowers researchers to themselves apply the Authority’s data ethics principles to their research projects. This enables researchers to identify and mitigate against ethical risks in their projects with support from the UK Statistics Authority’s Centre for Applied Data Ethics’ expert user support services, published guidance and training. This tool has had a significant impact and has been widely used since its inception, with 258 projects last year using the tool, including those from academia, the ONS, other government departments, and devolved administrations.
In circumstances where research teams require additional independent ethical advice and support to mitigate against significant ethical risks identified via the ethics self-assessment process, we also provide expert ethical advice through NSDEC which provides transparent and timely ethical advice and assurance to the National Statistician. Over the last year, 17 projects including the COVID-19 Schools Infection Survey (ONS), a study to assess the algorithmic feasibility of COVID-19 specific vocal biomarker detection (JBC) and research to determine the population-level relative risk of hospitalisation or death that COVID-19 presents to people with different socio-demographic characteristics and co-morbidities (ONS), have been escalated to NSDEC to seek expert independent advice and assurance on those most pressing ethical issues that the ONS and the wider research community has faced. Our experience has taught us that providing this applied data ethics support plays a vital role in empowering analysts to use data in innovative ways that are ethically appropriate and efficient.
Office for National Statistics
January 2022
Dear Mr Hunt,
While providing evidence to the Health and Social Care Committee and the Science and Technology Committee for the inquiry, “Coronavirus: Lessons Learnt” on 1 December, I promised to provide further information. Specifically, I am responding to queries from the Committees on international comparisons of mortality rates, specifically for groups who are ethnic minorities in the UK. I will address each of those questions in turn.
At what stage in the pandemic did the ONS become aware of the higher propensity for people from minority ethnic groups to die from COVID-19? (Q620)
Clinical studies identified apparent ethnic differences in COVID-19 mortality from April onwards, although the early evidence was inconclusive. The Office for National Statistics (ONS) began a research programme at the beginning of April to investigate population factors affecting COVID-19 mortality using linked data, and our first results were published on 7 May covering deaths that occurred between 2 March and 10 April. Subsequent analyses have included deaths occurring up to 28 July.
How has the UK done in comparison to other countries when it comes to ethnic minorities? How many Bangladeshis have died here in the UK [in comparison to Bangladesh]? (Q655)
Figures published by the World Health Organisation (WHO) show that the cumulative mortality rate (as of 29 November) from COVID-19 in Bangladesh is 9.9 per 100,000. Using WHO’s methods, the equivalent rate for the UK is 85.5 per 100,000. These figures cannot be directly compared to the ONS age standardised mortality rates, which for people of Bangladeshi ethnic background in England and Wales are 270.5 deaths per 100,000 in males and 110.0 per 100,000 in females, because of differences in methods. The figures from WHO on COVID-19 mortality are based on a mixture of reporting systems and estimates, and in developing countries such as Bangladesh are likely to be based on incomplete coverage; they are not age standardised, which is important as Bangladesh has a much younger population than the UK; and are not calculated based on the same exact time periods.
ONS has worked with partners including WHO, Eurostat, the French Institute for Demographic Studies (INED) and the Max Planck Institute for Demographic Research in Germany to help make comparative international data on COVID-19 mortality available. We published an analysis of excess all-cause mortality comparing the UK and most European countries on 30 July – all-cause mortality can be compared more reliably across countries than deaths from COVID-19 specifically. However, comparisons outside the small group of countries like the UK that produce comprehensive, detailed and timely cause of death data are very difficult due to different data collection systems and definitions, delays in availability, and in some cases substantial incompleteness. Even in countries with quite promptly available high-level data, breakdowns by ethnicity are not necessarily available. This is partly because ethnicity as it is measured in the UK is only given importance in certain countries, while in some (for example France and Germany) ethnicity is not officially recorded because of privacy concerns.
The published literature on COVID-19 and ethnicity so far focusses on the UK and the USA. The concept of ‘race’ used in official statistics in the USA is somewhat similar to the grouping of ethnicity in the UK. According to the US Centers for Disease Control (CDC) people of Black or African American race have 2.8 times the risk of death from COVID-19 compared to White people; see table 1 (below). These comparisons are based on age-standardised rates but not adjusted for socioeconomic factors. They show disparities of a similar magnitude to our findings for England and Wales, which were for example that males of Black African ethnic background have a rate of death involving COVID-19 2.7 times higher than White males, and females of Black Caribbean ethnic background have a rate 2.0 times higher than White females.
Table 1: Risk of death from COVID-19 for racial or ethnic groups in the USA, compared to White people (as of 30 November 2020)
| Racial or ethnic group | Increased risk relative to White group |
|---|---|
| Native American or Alaskan Native | 2.6 |
| Hispanic or Latino | 2.8 |
| Black or African American | 2.8 |
| Asian | 1.1 |
Source: Centers for Disease Control, November 2020
Why is there a discrepancy between the mortality rates for COVID-19 for minority ethnic groups in this country, making a direct comparison with countries where they are the ethnic majority? (Q656)
Comparison between the experience of COVID-19 among specific ethnic groups in the UK and in other countries associated with the ‘place of origin’ of those groups is fraught with difficulty for three main reasons.
Firstly, this is because of the limitations of ethnic group as a measure for analysis of mortality:
Secondly, mortality data which are timely, accurate and complete are hard to come by outside a relatively small number of countries which have well-developed civil registration and statistical systems. Some African and South Asian countries have less than 20% completeness in medical certification of causes of death. While this does not mean that the death rates from COVID-19 reported for those countries are completely unreliable, it should be recognised that they are estimates based on limited information. Limited access to virus testing and other diagnostic facilities in some countries could also play a role.
Thirdly, the demographic and geographical circumstances of many of the relevant countries are very different from the UK in ways which are likely to affect the outcomes of the pandemic:
I hope this information is helpful and look forward to an opportunity to update you on our further analysis on this important subject.
Yours sincerely,
Iain Bell
Deputy National Statistician and Director General for Population & Public Policy
On 21 October 2020 Professor Sir Ian Diamond, National Statistician, gave evidence to the Science and Technology Committee and the Health and Social Care Committee’s joint inquiry on Coronavirus: lessons learnt, specifically considering the use of statistics and modelling.
A transcript of which has been published on the UK Parliament Website.
Dear Mr Clark,
While providing evidence to the Committee on 7 May, for the inquiry ‘UK Science, Research and Technology Capability and Influence in Global Disease Outbreaks’, I promised to provide further information to the Committee on excess deaths, and to clarify the release schedule of the results of the COVID-19 Infection Survey. I have also been informed the Committee would be keen for ore
detail on the potential of timely electronic recording of deaths, and whether we have responsibility for publishing a value of ‘R’.
The Office for National Statistics (ONS) publish provisional weekly deaths registrations, which are currently published for deaths registered up to 1 May 2020. National Records Scotland (NRS) and the Northern Ireland Statistics and Research Agency (NISRA) are responsible for publishing the number of deaths registered in Scotland and Northern Ireland respectively.
Figures from the weekly ONS deaths bulletin show that the recent overall increase in deaths compared to the five-year average is not solely due to deaths involving COVID-19, known as excess deaths.
The ONS is publishing a report on the increase of non-COVID-19 deaths observed in weekly deaths statistics later this month, which I will send to the Committee when released. The report will analyse how the number of non-COVID-19 deaths occurring in different places of death, for different age groups and for different causes of death differ from previous years’ data and will suggest how these findings correspond with possible reasons for the increase.
I stated that electronic recording of deaths would be useful to increase the timeliness comprehensiveness of our mortality statistics, which while among some of the most timely recorded, have an 11 day lag. The timing of those registrations would still have to be balanced with the need to provide cause of death, and ensure our death registration remains informative and valuable, as opposed to a simple count. However, electronic registration would be a vast improvement, particularly when noting the ONS still receives some paper registrations, and the average registration takes etween 4-5 days.
On the COVID-19 Infection Survey, we have now produced and published the first estimates. We will produce estimates on a weekly basis to begin with and publish these each Thursday. Over time, as the Infection Survey develops, we aim to produce and publish estimates twice a week. We will do this once we are sure that publishing twice weekly maintains quality, relevance and coherence of the data to users.
Moreover, the Infection Survey is vital for the Scientific Advisory Group for Emergencies (SAGE) to calculate estimates of R. The ONS will assist SAGE, but will not provide, produce or publish any
alternative estimates of R. I hope this is helpful, and please do not hesitate to contact me further with any additional questions. I am copying this letter to the Chair of the Public Administration and Constitutional Affairs Committee.
Yours sincerely,
Professor Sir Ian Diamond
Related Links:
Professor Sir Ian Diamond’s oral evidence (May 2020)
On Tuesday 7 May 2020 Professor Sir Ian Diamond, National Statistician gave evidence to the Science and Technology Select Committee as part of their inquiry: UK Science, Research and
Technology Capability and Influence in Global Disease Outbreaks.
A transcript of which has been published on the UK Parliament’s website.
Related Links:
Professor Sir Ian Diamond’s written evidence (May 2020)
