Submitted by James Cantwell on Thu, 27/08/2026 - 09:05
Risk prediction tools are designed to help GPs identify the patients who may be at higher risk of disease. Some of these tools use AI, and could help with earlier diagnosis, more personalised care, and preventing illness, which is why they feature in several national plans in England, including the 10 Year Health Plan and the Cancer Plan.
While many risk prediction tools are being developed, most of them are never used in everyday GP practices, despite considerable time, money, and effort has gone into developing and testing them. This is a problem: it wastes investment in research and means patients miss out on the potential benefits.
New research led by the University of Cambridge identifies that wider system changes are needed to help risk prediction tools move from research projects into everyday clinical practice.
Why aren’t risk predictions tools used by GP surgeries?
Most GP surgeries are small, independent businesses scattered across different areas. Unlike hospitals, they don’t always have direct control over their own technology. Practices often rely on IT systems provided by other NHS organisations, often leaving GPs with disjointed software, basic computers, and very little time to use new tools during short patient appointments.
Previous reports on digital and AI tools in the NHS have rarely looked at primary care. Where they have, they’ve looked mainly at population-level data and not at the risks or advantages of using tools to flag health risks for individual patients during their appointments.
Our research explains what needs to happen for risk prediction tools to move from research projects into everyday GP care. The work was funded by the NIHR Policy Research Unit on Cancer Awareness, Screening and Early Diagnosis.
How we carried out the research
We started out by reviewing recent UK policy reports (published between 2022 and 2025) alongside a broad review of academic research on technology barriers, examining the technical, infrastructure, system, and usability challenges that prevent digital and AI tools from being successfully adopted in everyday primary care.
We then held an in-person workshop with 22 experts, including family doctors, clinical informaticians (specialists who manage patient and clinical data systems to help hospitals, doctors, and nurses improve patient care), researchers, funders, regulators (NHS England and the Medicines and Healthcare products Regulatory Agency), and medical software developers. This was followed by three online meetings and a feedback survey.
As part of this process, the experts reviewed five different technical routes for connecting these tools to GP computer systems, ranging from individual custom-built software to a shared system of standard, reusable tools.
What we found
To work effectively in GP settings, risk prediction tools need to be built to solve a real clinical problem and fit into existing GP systems without creating extra work. It should also be clear how they produce their results, so that clinicians and patients can trust them, and they need to be checked and updated regularly.
Four system changes needed for adoption
In the future, risk prediction tools could help GPs spot people who may be at higher risk of illness earlier, supporting earlier tests and diagnosis in some cases. However our research found that whether GPs can actually use these tools depends much less on the tools themselves, and much more the wider healthcare, academic, regulatory, and funding systems working together to support their use.
We identified four changes that could help:
- Early joint working - different groups need to work together from the start.
- Shared systems and infrastructure - a common, shared way of using and supporting risk prediction tools across the health system.
- Clear and practical rules - clear and workable guidance for how tools are approved and used in the NHS.
- Shared responsibility and long-term funding - a clear, shared agreement on responsibility if something goes wrong when the tools are used.
These changes will need to happen together, because if only one or two are addressed, it is unlikely that the tools will become part of routine care.
Further information is available in the published paper, lay summary and policy briefing below.
Links
-
Read the full paper: https://doi.org/10.1016/j.lanprc.2026.100199
-
Citation: Usher-Smith JA, Taylor LC, Blackburn L, Babb de Villiers C, Krishnan B, Brunsdon E, Lyratzopoulos G, Ercole A, Peek N. Overcoming technical barriers to
enable real-time risk prediction in primary care: recommendations from expert workshops. Lancet Prim Care 2026. https://doi.org/10.1016/j.lanprc.2026.100199 -
Download the lay summary:
-
Download the policy briefing:
