Secret Garden or Open Landscape – Views on Trusted Research Environments
Wednesday, 25 November 2020 - Panel presentations Thursday, 10 December 2020 - Bringing the panel back together
The webinars focused on Trusted Research Environments (TREs) – how they work, why they are potentially so important and where the patient voice sits in their design, development and running. The first webinar included panellist presentations to set the scene and describe the key aims and challenges. This was followed by a separate question and answer session with the panellists.
use MY data is keen to bring a strong and direct patient voice into all conversations about patient data. These patient-led webinars brought together patients, relatives and carers, the public and professionals, were open to all and free to join.
Our previous events have highlighted the difficulties researchers face in getting the approvals to access the data that they need to undertake research. Furthermore, additional time and resource is spent in getting the data to the researcher, even once approved. Can the development of Trusted Research Environments address these issues?
We explored some fundamental points in the webinar, including what TREs are, how are they different and what problem are they meant to solve? Recognising that some TREs have been around for a long time, but probably by a different name (such as a Safe Haven); what can we learn from these? How are new TREs being established and for what purposes? What are some of the safeguards that we need and the potential pitfalls we need to avoid? And, crucially, where does the patient voice sit in this world?
The webinar panel
Health Data Research UK - Tim Hubbard, Professor of Bioinformatics at King's College London & Associate Director, Health Data Research UK London Site
SAIL Databank- Kerina Jones, Professor of Population Data Science
NHS Digital - Brian Roberts, Head of the TRE Service for England
MedConfidential- Sam Smith, Coordinator
use MY data - John Marsh, Advisory Group Member
The webinar can be watched by clicking either of the two video boxes to the right and the presentations are available here.
The first video includes the presentations from the panellists, with the second video containing the question and answer session, which was recorded when the panel reconvened on 10 December 2020.
Awareness of Trusted Research Environments (TREs)
A key challenge is that the general level of awareness and understanding of TREs is low, despite the ambitions of TREs being high.
At the start of the webinar we asked delegates to indicate their current level of knowledge of TREs and the results (in red) demonstrated that the webinar had a mix of delegates, with a range of expertise and understanding.
We followed the webinar with a survey to see whether delegates felt more informed, having attended the webinar. This showed a significant increase in awareness and understanding in delegates as a result of the webinar.
The strategic importance of Trusted Research Environments (TREs)
Tim noted that as the amount of biological data has increased and computing technology has made algorithms more portable, there has been a shift to research being carried out in TREs.
For sensitive health data, TREs are applying the Office of National Statistics (ONS) "Five Safes" to create environments where data is protected by not being distributed. These Five Safes are now being regularly used by organisations who manage health data.
General rules for the operation of TREs, including the adoption of the Five Safes, are being developed by Health Data Research UK under the UK Health Data Research Alliance. Tim highlighted a recent Green Paper about this. Details about the contents of each TRE should be recorded on the HDR-UK Gateway.
As an example of how a TRE functions, Tim described the TRE that is operated by Genomics England for the 100,000 Genomes Project, highlighting the ways that data was safeguarded and controls applied to restrict and monitor the uses of data.
“Health data research has historically been dominated by research cohorts and data distribution”
Tim noted that the public are rightly nervous about privacy and security. However, he explained that the concept of a TRE is inherently safer, as it avoids data distribution – data is always kept inside a secure research environment.
He emphasised that strong controls on access and security are still needed, but that it was easier to implement this centrally.
"The principles of the Five Safes are key steps in ensuring data is protected, managed and used ethically"
What has a long-established Trusted Research Environment (TRE) been able to do in Wales?
Kerina introduced the Secure Anonymised Information Linkage (SAIL) Databank, which was established by the Population Data Science group at Swansea University in 2007, in recognition of the immense, untapped research potential of routinely‑collected health and administrative data.
She described the challenges in enabling access to data with high utility for research whilst simultaneously safeguarding individual privacy. SAIL was set up to address these challenges.
Kerina provided a brief overview of how SAIL operates, including the types of data that are held within SAIL, and the controls which are placed on these data. She also describes the process of data linkage and how SAIL relies on a separate linkage agency to undertake this work, ensuring the data is always anonymised, providing a further level of safety.
Kerina noted that since launched, SAIL had opened up ways to answer important questions that could not otherwise be addressed or could not be addressed without prohibitive effort and cost.
SAIL has an established Consumer Panel of patients and the public. Their role was to provide public perspectives on data intensive research and on new developments, views on data protection issues and to discuss proposals and research findings.
Finally, Kerina described some challenges which SAIL is now facing, including:
How to get public views on safeguards in the use of free-text data such as from clinical letters and notes
Understanding public preferences on access models for genetic data linked to health and administrative data
Public views, knowledge, benefits and concerns on the use of personal device data for research
Working with the adoptees, adopters, birth parents and the general public on the use of data about adoptions with health and administrative data.
“Trusted Research Environments need to be trusted by those that are effectively donating their data - PATIENTS!” use MY data member
How are new Trusted Research Environments (TREs) being developed by NHS Digital for England?
Brian talked about the developments inside NHS Digital, building on their place in law as England’s Safe Haven for patient data. NHS Digital want to be the place to go for safe and timely access to the most comprehensive store of nationally collected, linkable, quality assured health and care data for secondary use.
"NHS Digital wants to be the place to go – the one stop trusted health research service – for researchers to come in and link data in a safe environment."
The mission of NHS Digital is to provide approved researchers with rapid, safe and trustworthy access to essential health and care data. Enabling timely access to data which will enable research and to do this at scale is a significant challenge. Their work so far has been driven by, and focused on, outcomes associated with COVID-19 to guide national and international decision making and recommend potential interventions to reduce the severity of COVID-19 outcomes.
This initial work on COVID-19 is now being extended into other areas, including heart disease and cancer.
"Public and patient engagement is a watchword for the service...the PPI voice is really important to all of this."
As with the other panellists, Brian described how the Five Safes were also being implemented in the design of the NHS Digital TRE solution. He highlighted the areas of development for their TRE service, which would mark a significant shift from data dissemination to (safe) access.
There remained a significant need for wider engagement. Reflecting that it was difficult to engage with everyone individually, he encouraged people to work with groups such as use MY data.
"We actively encourage people to go through groups like use MY data to have their say on things like, how we design and build the service and how we run it day-to-day"
In conclusion, Brian highlighted that their TRE development was very much at the start, and that the enduring TRE service will enable safe and timely research at scale for:
Benchmarking (such as for a clinical intervention, a service etc.)
Health and care surveillance/monitoring over time
Evaluations of clinical and financial impact
Improved care commissioning, assessment and planning.
What about privacy, governance, transparency and trust?
"I suggest the technology is now the easy bit to do"
Sam started by reflecting that everything we had been heard so far suggests that TREs are wonderful. He agreed that dissemination of data always carries risk, and that in safety terms, TREs are a better solution than disseminating data.
But he also raised some questions about the way that TREs will be run. For instance, how will we find out if rules are broken? And how can we find out if things go wrong?
He asked who is best placed to run a TRE, what might the governance look like, and what should be avoided.
Sam noted that the ONS has good principles and governance established over a period of time, but that a small organisation running their own TRE probably doesn’t have this.
Whilst the TRE model (where access is granted to the data for a researcher, rather than the data being sent out to the researcher) is a step forward, how do we assure that this is not an open door and that all the uses of data are ethical, monitored and reported?
On a wider level, Sam thought that the success of TREs will not depend on technology, but on the politics around their use. Who is going to 'certify' a TRE?
"The success of Trusted Research Environments will not depend on technology, but on the politics around their use"
And how does a patient feel about this?
John gave his views about the potential benefits and the potential risks of Trusted Research Environments (TREs) from a patient’s perspective. He brought a perspective from both the patient side and from the side of being a supporter of TREs, contrasting the benefits, risks and challenges.
"What controls are there over access to the data and over who can use it and for what purposes?"
John agreed that TREs have a number of benefits. They bring together data from diverse sources to enable medical research that has not been possible before. He also noted that data which is obtained from diverse sources will have different terms and conditions around it and will have a different quality.
Patients need to know that their data is being used legitimately. Patients need to know who is accessing their data and how the access has improved outcomes. John closed with the key point that the success of TREs would not depend on technology, but on the wider engagement and support of patients.
“It’s the single biggest prize in moving forward UK data-driven health research” use MY data member
Questions to the speakers
Chris thanked the speakers for their contributions, and opened up the webinar to delegate questions, noting the limited time available for questions.
Chris put two initial questions to the panel:
Do commercial users have the same rights of access to a TRE as public sector or academic users?
How easy is it for someone to be able to find out what uses had been made of a TRE?
Brian responded that currently there were no active commercial users of the TRE, but that this would be open to commercial organisations. He stressed that the route of access would always be through the existing Data Access Request Service (DARS) and that all approved requests would still be subject to a Data Sharing Agreement.
Kerina said that SAIL follow the same type of process, with the focus on ensuring there was public benefit. Of around 460 projects which have used SAIL, around 75% of these were by universities, 13% by the NHS, 4% by government and the rest were classified as “other”, a very small number of which would be commercial organisations.
In terms of the details of these uses being accessible, there is an internal SAIL dashboard. Chris highlighted the importance of transparency of use and suggested that this be checked by SAIL after the webinar.
Tim noted that the Genomics England TRE has an access committee which reviewed applications.
Sam and John both echoed the need for TREs to be transparent about who is using them, including the need to highlight the publications which were produced as a result of the data access.
Summary and close
Chris highlighted key points which had arisen during the webinar. As the amount of data has increased and computing technology has made algorithms more portable, there has been a shift to research being carried out in research environments. He also noted:
Awareness of these Trusted Research Environments is low
Positively, we heard a consistent commitment to abide by the Five Safes from all
The Green Paper about TRE development needs to be publicised better
Transparency around access to (and data releases from) TREs is essential and currently is not comprehensive
The SAIL Databank would benefit from publishing their access details in full
NHS Digital will need to describe data access and data releases for their TRE.
Chris brought the webinar to a close, summarising:
"It's clear that if they deliver what they intend, this could see significant changes, improvements and efficiencies in the ability for bona-fide applicants to use patient data in a secure environment. If they work, then the traditional data release could begin to disappear, being replaced by data access. It's also clear that if all this is going to happen, we need much greater understanding and involvement to build trust in the wider patient and public population. And that’s a joint challenge to us all."
Bringing the panel back together - Thursday, 10 December 2020
The following questions were covered during the second webinar, with the same panellists.
Clarity is needed about costs of access to TREs. Is there a standard model, or are TREs going to become competitive? Lowest cost might not always be in the patient interest
We seem to have a number of TREs run by different groups. What is the strategy to link or coordinate between them so we have a national service rather than separate TREs?
Does research suffer by being limited to databases within the home nations rather than being truly national - i.e. whole UK?
Will TREs do away with the need for all data releases?
What are the different models for patient involvement in existing and developing TREs - contrast GEL, SAIL and NHSD
How is HDR-UK bringing to bear what we know about public attitudes towards third parties accessing health data (even if de-personalised) that is routinely collected?
Most data in TREs will be routinely collected, and not consented. GEL is the exception. How will TREs deal with the problems of using non-consented data?
The key closing point from the panel: while there was awareness of the risks of using TRES, overall there was optimism for their role in driving forward health research more rapidly and widely than other data access methods.