In the second in our blog series showcasing the DARE UK Sprint Exemplar Projects, Professor Elizabeth Sapey and Suzy Gallier from University Hospitals Birmingham discuss how the FED-NET project explored federated analytics as a secure solution to combining datasets for research.

Society faces ever evolving and complex challenges in healthcare. Solving these problems requires health data to be collected for different purposes, from different sources across varying geographical location, at scale.

However, combining this data is technically challenging: routine data can be collected in different data languages, and building secure computer networks to house this data requires infrastructure and specialist knowledge. There is also the need to address public concerns about data access and build trust through robust governance of health data.

The FED-NET project’s remit was to test whether federated analytics could offer a scalable solution to the technical and governance challenges of analysing these datasets. Federated analytics is where different datasets are stored within separate trusted research environments (TREs), and instead of the data moving to one location for analysis, the analytical code is sent to each TRE, and the results of the analysis pooled.

Engaging the public

FED-NET had public involvement and engagement embedded throughout the project lifespan. This was essential to inform our planning and delivery and ensure FED-NET delivered a truly secure solution to combining sensitive health data that met the needs of researchers, and was also trusted by patients.

We worked closely with our established group of patients and the public to understand their views, support conversations about our planned programme of work, the benefits and challenges of federated analytics, and what this approach might mean for patients and their data.

In order to facilitate these conversations, we held a series of workshops with members of the public to co-design a brochure that explains federated analytics and the aims of FED-NET in lay language that is easy to interpret and understand.

Involving key decision-makers

As well as involving patients and the public, FED-NET wished to understand the views of key decision-makers in the health data research space regarding data discovery, access and federated analytics. The FED-NET team held workshops for strategic data providers and end-users to explore their views on the health data research landscape, their current awareness of TREs, the perceived benefits and challenges of running a TRE, and the expectations and uncertainties of federated analytics and future data analytics tools.

A FED-NET workshop was held virtually, with a range of attendees, including Chief Information Officers (CIOs), Chief Clinical Information Officers (CCIOs), TRE leads, Informatics Programme Managers, Data Scientists and Data Architects, representing NHS and academic organisations and the charitable sector. The attendees from NHS Trusts included representation from across the four nations of the UK. There was also representation from the university sector, national charities, and health research regulators.

After brief presentations on the current landscape of health data research and federated analytics, there was an open discussion and several themes emerged through the group discussion. These were expertise, resources and sustainability; equity of provision and access; public trust; harmonisation of standards and approach; scale; and interoperability.

Following the workshops, a report was published to capture the discussions and was subsequently circulated to the attendees for any further feedback. It was agreed that the group would form a FED-NET working group to continue the conversation of how to support Trusts and bring together a network of experts and key decision-makers to provide guidance and share knowledge around TREs. This will help bring about greater equity of provision, access and benefits of shared data.

Key successes

We set ourselves six ambitious work packages, and were able to successfully deliver on them all:

  1. We built and tested the technical architecture to perform federated analytics between TREs with both the same and different underpinning structures.
  2. We curated a real-world dataset including electronic health record, meteorological, air quality and inflammatory biomarker data for patients with asthma. We built a synthetic dataset of the same parameters and tested if standardised data models/languages were useful with such a complex dataset.
  3. We built a metadata catalogue of the data to enable researchers to understand the type of data we held.
  4. We performed federated analytics across multiple TREs, testing the accuracy of results when using different federated approaches, versus pooling the data.
  5. We worked with healthcare organisations from across the UK to understand their requirements for federated analytics. This was co-developed into a report and a working group has been formed for future discussions.
  6. We engaged with diverse patients and members of the public to understand what they wanted to know about federated analytics. We co-produced a public brochure to explain federated analytics, which is now being trialled in wider public groups.