When the NHS is approached with a new tool which can improve diabetic eye screening, how can it be sure that the tool will live up to its creator’s claims? The Health Economics Unit was asked to provide independent peer review of an evaluation, helping to give a clear and credible picture of the tool’s potential to have an impact on treatment.
OptosAI is a machine learning tool developed by Optos to support screening that looks for an eye condition called diabetic retinopathy, which can be caused by diabetes. The tool analyses fundus images of the retina and recommends referrals based on severity of diabetic retinopathy and diabetic macular oedema.
To strengthen the evidence base behind the technology, Optos undertook a retrospective evaluation commissioned by NHS England’s Accelerated Access Collaborative (AAC).
The Accelerated Access Collaborative (AAC) is a national NHS England programme that supports the evaluation, adoption, and spread of innovative technologies and models of care across the NHS.
The Health Economics Unit (HEU), with the Strategy Unit, is an evaluation partner for the AAC. The team partnered with the clinical trials unit at Keele University to provide independent peer review of the evaluation – ensuring the evaluation met high scientific and ethical standards and that the findings would stand up to scrutiny.
A transparent approach to evaluation
From the outset, maintaining independence, rigour, and practical support would be essential to helping the evaluation succeed.
The HEU set up a Data Monitoring Committee (DMC) made up of specialists in clinical trials, health economics, machine learning, ophthalmology, and statistics. This group brought the right blend of expertise to constructively challenge, support, and guide the evaluation team throughout the process.
The DMC met at three key stages of the project – early on, midway, and at the end – to review progress, evaluate methods, and assess data quality.
At each meeting, Optos presented their study design, interim findings, and next steps. The HEU provided structured feedback through open discussion and private deliberation sessions. Following each DMC, the team wrote to the AAC summarising their findings and any recommendations.
Adding value through improvements
The HEU’s work with the AAC was important to ensure independent peer review of the evaluation, however the team wanted to ensure added value and further strengthening for the evaluation.
For example, the HEU recommended more sophisticated statistical analyses beyond basic agreement percentages, helped sharpen the sampling methodology to reduce bias risk, and advised Optos to consider how artefacts and multiple images should be handled to more accurately reflect real-world grading practices.
The project team also made sure the evaluation considered practical realities that could affect future implementation. For instance, the HEU questioned how different imaging equipment might impact results and flagged the need for clear guidance on camera compatibility for wider roll-out.
Optos said they found the HEU’s input helpful and used it to refine their study methods and improve their documentation between meetings.
Throughout, the HEU team made sure the peer review remained responsive, practical, and rooted in real-world needs – providing the kind of scrutiny that supports innovation without stifling it.
Delivering a clear study free from bias
The HEU’s independent peer review helped ensure that the OptosAI evaluation was conducted with rigour, transparency, and credibility. The team’s involvement strengthened the methodology, challenged assumptions where needed, and ultimately helped Optos deliver a clearer, more robust study.
The findings are now well placed to inform future research and contribute to decisions about wider adoption within the NHS. The HEU are proud to have played a part in supporting innovation that could improve diabetic eye screening – and to have demonstrated how expert peer review can add real value at every stage of technology development