Enabling AI and data development for health and care – a blog by Gemma Warren, Policy Manager Data and AI

As part of our work to streamline approvals for AI (artificial intelligence) and data driven research, we need to understand the challenges faced by industry when accessing health data for research and how the HRA can support.

Understanding and learning

We organised workshops in partnership with the Medical Research Council’s Regulatory Support Centre. In them we worked with applicants to understand how we can best direct them to the approvals they need as part of their study.

We met with university researchers, data scientists, research and development staff, experts in information governance and members from the research community - including those developing tools and applications using health data.

To make sure we heard from as many people as possible, an online survey asked applicants about their journeys through the approvals process when accessing health data.

The feedback collected will help shape guidance in this area. We want to make access to data less confusing while still ensuring it is done legally and ethically.

Our findings

• The people we spoke to wanted clarification on what data was accessible to them. They wanted to know at what stage they should apply for access. They also wanted help to find the right people to have conversations with to get access to these data and to help prepare necessary documents.
  
• Applicants wanted clear definitions and concepts, such as the difference between anonymisation and pseudonymisation. They asked for greater clarity on what approvals are required at each stage of developing AI and research using confidential data. They wanted to know more about the decision-making process to help them understand how decisions are made, adopt best practice and get approvals faster. Researchers reported that their peers who already had awareness and knowledge of the system were a valuable source of guidance when accessing data.
• Time taken and the complexity of gaining access were things that came up again and again. We know that the interoperability - the ability of computer systems or software to exchange and use information - between data service providers and our systems is an area that needs improvement.  However, uncertainty about the quality and completeness of the data being accessed had the potential to impact timelines and funding for developing data driven technologies.

Next steps

We’re developing new guidance that will support applicants doing data and AI-driven research.

This includes a step-by-step overview of the process, with essential considerations for those wanting to carry out trials, a directory of the key organisations in this area, and definitions of terms and concepts. This guide will explain at what stage applicants should complete documentation, what conversations to have with data service providers, and key concepts and definitions. Our guidance will be accessible and open-ended enough to provide examples and advice relevant to projects in their earliest phases, helping researchers to plan and access data proportionately.

We’re also planning a data decision tool to help researchers decide what data and approvals they might need as part of their study. The tool will enable researchers to map out and test different possibilities within trials. This will demonstrate how changes to trials may affect approvals needed by making the decision making process clear.

Thank you to everyone who has helped us with feedback so far. We’re working with partners to create guidance that is visual, dynamic, clear and easy to understand.

There will be further opportunities to feedback on the work we are doing. If you would be interested in hearing about future opportunities, please email us