Goldacre recommendations for making the most of health data for research
The Goldacre review, 'Better, broader, safer: using health data for research and analysis' is a hard-hitting report with recommendations for enabling the proper use of NHS health and care data for research , to drive improvements in health and care
8 April 2022
A hard-hitting new report with recommendations for actions to enable the proper use of NHS health and care data for research – and hence, to drive improvements in health and care – was released today.
The Goldacre review, Better, broader, safer: using health data for research and analysis, was led by Prof Ben Goldacre and commissioned by the Health Secretary to inform and work with the wider NHS Data Strategy. It was specifically focused on identifying the barriers to making better use of NHS data, and making recommendations on how these should be overcome.
Leading the world means changing the system
This long-anticipated report does not disappoint. It is highly technical – because it proposes solutions to in many cases highly technical problems – but also outlines broader policy actions. Emphasising the unique and rich resource of health data held within the NHS, which is widely recognised, and the ‘world class researchers’ who are ideally placed to make use of it, the report also explains the need for data to be curated, managed and made readily available research via well designed platforms with appropriate security and transparency to earn public trust. At present, it notes, data management systems and projects are fragmented and siloed, duplicating costs, fostering monopolies over data, using inadequate security and relying on slow, risk-averse governance systems.
Goldacre concludes that such a system ‘can never scale to the kind of access needed for a world leader in data science’ and emphasises the compelling need for action to meet this ‘generational opportunity’ by advocating for platforms, people and organisations to facilitate data analysis and meet current and potential future needs for research driven improvements in health services and solutions to health challenges. The report concludes:
‘For less than the cost of digitising one hospital the system can have the secure data platforms and workforce needed to realise the full value of NHS data, driving research, health service improvement, and innovation’.
Creating and using Trusted Research Environments
To create this coherent, connected system at scale, the report makes specific recommendations. A move to the use of a small number of shared platforms that create Trusted Research Environments or TREs as the default approach for analysis of NHS patient records by researchers is advocated, for all patient information with any privacy risks and without consent for wider forms of data sharing.
In contrast, the report recommends urgent recognition of the limitations of pseudonymisation of patient data as a safety measure, and of public engagement and communications alone (ie. without robust systems to ensure security) as a means to build trust.
Public trust is unquestionably a big issue when it comes to health data, with an ongoing House of Commons Science and Technology Committee inquiry, The right to privacy: digital data, currently examining issues of concern arising from government ambitions to share personal health data across the health system, as set out in the draft Data Saves Lives strategy.
A consistent and collaborative approach
When it comes to data storage, management, curation and analysis, the report recommends funding the development of shared access systems, coding, standards and practices for use across all NHS TREs in an open and collaborative fashion, with clear, accessible technical documentation. This, it is suggested, should be achieved through driven by open, competitive funding to develop such methods and tools, and to curate health data, with clear disclosure of public, academic or commercial funding sources (as well as that from UKRI and NIHR), objectives, work in progress and linked outcomes.
It also calls for the creation of a new NHS Analyst profession to underpin a commitment to Reproducible Analytical Pipelines as a core working practice for all NHS platforms and teams. This new NHS Analyst specialty should, it is suggested, include a formal service modelled on the Government Economic Service and Statistical Service, an Open College to oversee development of curricula for roles, training and continuing professional development, and ‘realistic salaries’ ie. sufficient to attract people with the necessary skills and talents required to fulfil the roles.
More widely, the report calls for development of Research Software Engineering as a new academic discipline (especially in health); for training of academic researchers and NHS analysts in computational data science; and for training of software developers and data scientists in health services research and epidemiology.
When it comes to managing these new and improved systems for health data, the report makes ambitious calls for the development of common approval processes including for ethics, data access and information governance and recognition of ‘the risk of data controllers asserting access monopolies to obstruct competitors’. To address this issue – noting that at present 160 NHS Trusts and 6,5000 GP practices act as distinct data controllers, which nicely illustrates to scale of the problem - it suggests that either one national organisation acting as Data Controller for a copy of all NHS patients’ records in a TRE; or the ability of separate Trusts and GPs to nominate a single entity to review and approve data access requests on their behalf. Rules about the use of patient records for NHS performance management are also advocated with a view to improving services but not distracting providers with ‘unhelpful’ measures.
The review also recommends ‘high quality PPIE’ (patient and public involvement and engagement) and ‘a frank public conversation about commercial use of NHS data for innovation, but only after privacy issues have been addressed through adoption of TREs’ – suggesting that current efforts to address public concerns around data sharing and health research might be better delayed until what are clearly seen as the most important solutions are in place.
Making it happen: leadership
Finally, the Goldacre reviews sets out clear recommendations for steps to enable all the other proposals to progress, calling for the NHS to build ‘impatiently, but incrementally’ by demonstrating the value of new data management and use and gradually replacing old practices with new. It suggests identifying ‘data pioneer’ teams, recruited for the best technical skills and suitably equipped, for integrated care systems, national quality improvement registry or audit, academic birth cohort or electronic health record analysis and national NHS analysis. It also emphasises the importance of putting people who understand complex technical issues in ‘very senior leadership roles’, as well as training other senior leaders in the basics of data analysis, software and informatics as far as possible.
These recommendations are practical, and bold; they pull no punches about the genuine barriers to progress, and as such are likely (if implemented) to cause some discomfort, especially for the more radical changes. However, the review clearly emphasises the driver for change: ‘If this job is done well, then the system can finally unleash the full power of all NHS data ever collected, in one fell swoop’.
The work has already begun; the government announced £200 million to support the development of TREs. If the forthcoming final version of Data Saves Lives is in line with the Goldacre recommendations, and if the government can put the funding and impetus required behind not only technical development but also cultural change, training, infrastructure and other support – this could indeed be the turning point in making data deliver for research and for health.
Many of the findings in this report echo those identified in previous PHG Foundation research, including in relation to the pandemic experience of patient data sharing for genomic and health research under the COPI notices. Further legal and policy analysis of these recommendations will follow.