University of Birmingham

Developing an automated evaluation pipeline to identify the effectiveness of digital interventions in acute care: A pilot study assessing inequalities in the effectiveness of DERM

Artificial Intelligence (AI) has the potential to help doctors detect skin cancer earlier. For example, DERM, an AI tool, can analyse photos of skin lumps and compare them to millions of images of skin cancer. It has been used successfully at UHB since 2020.
However, there is a concern that AI tools like DERM might not work equally well for all patients. For example, the AI might be less accurate for people with darker skin tones because it may not have been trained with enough examples of diverse skin types. A similar issue was found in pulse oximeters, which worked less well on people with darker skin.
The project has three aims: check if DERM works well in practice by comparing skin cancer diagnosis rates before and after the AI tool was introduced; identify if any patient groups see worse outcomes, including ethnic minorities; and create a system to quickly test other AI tools.
The researchers will use data on patients referred for a skin cancer check, whether they were diagnosed, how long it took, and information about ethnicity.
This project will benefit patients by making sure that AI tools like DERM work equally well for everyone. The results will be shared with the developers of DERM so they can improve the tool for use in other hospitals.

This project will benefit patients by identifying potential harms to patient subgroups and supporting the development of mechanisms to monitor and mitigate future risks. The findings will be shared with UHB and the developers of DERM to guide improvements and reduce any identified harms. This will benefit patients within UHB and other trusts using the tool by making sure that AI tools like DERM work equally well for everyone.
The creation of an automated analysis pipeline will support future evaluations of AI tools within UHB, helping ensure patients across different disease pathways receive the best possible care.
The project will also benefit the NHS by providing a low-cost, largely automated framework for evaluating AI tools. The resulting evidence will support informed funding decisions, accelerate future evaluations, reduce costs, and help identify and address inequalities in AI-supported care.

16/06/2026

Live - Contracts Signed

SDE096

Cancer and neoplasms

WMSDE trusted research environment

No

Not yet published.