A study published May 21, 2024, in The Lancet Digital Health, led by researchers from Imperial College London including Dr. Susan Toh and Prof. Peter Dean, found that integrating Artificial Intelligence (AI) into the UK's National Health Service (NHS) Breast Screening Programme could significantly improve cost-effectiveness and reduce radiologist workload. The research aimed to address the shortage of specialist radiologists and enhance screening efficiency and accuracy for breast cancer detection in the UK.
The study utilized a decision-analytic model to simulate health outcomes and costs over a 20-year period, comparing the current two-human-reader screening method with various AI-supported models. These models included AI for triage, as a concurrent reader, and as an independent reader, using retrospective data from 124,775 women screened between 2009 and 2014 across NHS trusts in England, including Norfolk and Norwich University Hospital NHS Foundation Trust, Imperial College Healthcare NHS Trust, and Manchester University NHS Foundation Trust. The collaborative effort involved clinicians, health economists, and computational scientists.
The 'AI-first' model, where AI acts as an independent reader with a human arbiter, emerged as the most cost-effective, potentially saving approximately £10,000 per quality-adjusted life-year (QALY) gained compared to the existing human-only standard. This model could reduce the number of human readers needed by 25% to 50%, maintain or improve breast cancer detection rates, and lead to an estimated annual saving of around £40 million across the UK, while preventing more breast cancer deaths. This 'AI-first' model, found to be 98% more cost-effective, is currently being trialled in a prospective clinical study called MAMMOTH.
