DARE-FX: Delivering a federated network of TREs to enable safe analytics

Trusted Research Environments (TREs) are secure locations in which data are placed
for researchers to analyse. They host administrative data, hospital data or any other
data that must be securely isolated and only accessible for approved queries by
approved researchers. However, it is hard for a researcher to perform analysis across
multiple TREs, for example when data is to be analysed across geographical or
governance boundaries, such as the devolved nature of healthcare in the United
Kingdom. Yet this ability is urgently needed. Analysis across a federation of TREs
would enable timely analysis of UK wide scattered data to answer urgent questions,
as we needed in the COVID-19 pandemic.

The technologies and standards we need to be able to do this are available now. They
do not need to be invented. DARE-FX is assembling leading technology providers from
ELIXIR-UK and HDR-UK, with three TRE providers and two leading analysis platforms
to show through a real reference implementation how we can use secure Research
Objects to move between TREs while still supporting the Five Safes principles that
govern and protect patient data; all overseen by patient representatives.

The impact will be a step change for how researchers can safely combine data from
many sources, and for how data providers from any sector can safely implement this
using technology and standards we already have today.

Background
The main concept of a Trusted Research Environment is that data remains within the
secure boundaries of a technical and governance wrapper, to enable analysis in
accordance with the Five Safes principles (Safe People, Project, Data, Settings,
Outputs). Yet researchers need to perform analysis across multiple TREs that may
also span traditional funder silos: across multiple geographical locations; when some
data exists in one TRE and other data in another. When analytics are federated,
queries are pulled into the TREs containing the data, then processed and the results
returned subject to disclosure checking, without the data leaving the TREs.

Currently TREs do not have an effective approach to standardise the transfer of
information in and out of their environments or a common approach to controlling how
analyses are deployed and run. A framework is needed to ensure that federated
analytic queries run on TREs across the UKRI and ADR-UK networks are
interoperable, reusable and auditable to the Five Safes. Such a potential framework
has been developed by ELIXIR-UK, the national node for the European Research
Infrastructure for Life Science Data, and pre-piloted by the FED-NET TRE.

By using existing tooling and open standards within a Service-API driven architecture,
DARE-FX aims to demonstrate an exemplar framework and reference implementation
for federated analytic queries that enables Five Safes analytics over a federated
network of UK TREs. The framework and reference implementation will be developed
with PPIE as full team members.

Methods
We have gathered expertise from across the UK, including three different kinds of
TREs from HDR-UK/ADR-UK (UKSerp/SAIL, TREEHOOSE, PIONEER), pre-existing
federated analytic services (DataShield, BitFount), and ELIXIR’s interoperability
technology (workflow execution, RO-Crate Research Objects). The project includes
partners from previous DARE UK projects: FAIR TREATMENT, TREEHOOSE and
FED-NET. This expert team will assemble a core set of federated services that
demonstrate how existing tooling and open standards can be used to create a Five
Safes RO-Crate framework that adheres to the HDR-UK Five Safes model and will
thus ensure public trust.

To prove cross council utility across research domains and sectors we will collaborate
with the TRE providers SAIL, FED-NET and HIC to deliver a open-source reference
implementation that interfaces with the existing federated analytic services
DataSHIELD (to allow querying of ONS administrative data) and a commercial partner
Bitfount (to enable querying of healthcare data). An ELIXIR sponsored workshop has
already been conducted to explore the proposed concepts and implementation with
all partners.

To ensure our implementation is open for public scrutiny we will deliver a whitepaper
based on our framework, publicly available Bitfount and DataSHIELD federated
analytic workflows, an integration with the HDR data use registry and a publication
detailing the implementation. All results will be open source.