University Hospital Birmingham

University Hospital Birmingham (UHB) faced a significant backlog in chest X-ray reporting driven by a workforce gap and mismatch between radiologist supply and demand across the NHS. Chest X-rays, despite being among the most common imaging procedures, were frequently deprioritised, causing delays in patient care and diagnosis.

UHB conducted the largest retrospective AI trial in the UK, evaluating Oxipit's ChestLink autonomous AI solution against a dataset of nearly 200,000 chest X-rays from a single year. ChestLink autonomously identified and reported high-confidence normal studies, with 140,000 scans within scope for processing.

ChestLink autonomously reported just under 15,000 high-confidence normal studies, representing 23.4% of all normal scans and a 10.5% reduction in total annual radiology workload. The discordance rate between ChestLink and consultant radiologist review was only 1%, meaning the AI missed only 1% of abnormal cases. The study also identified potential for significant reduction in reporting turnaround time, from days to minutes for normal studies.

• Autonomous AI can meaningfully reduce radiology workload at scale (10.5% of annual volume), freeing radiologists for complex cases.
• A 1% missed-abnormal rate after consultant review demonstrates high sensitivity, but NHS-wide deployment would require demographic generalisability validation across diverse populations.
• Retrospective studies offer a unique advantage: the full patient journey is already known, enabling modelling of AI miss scenarios against historical outcomes before live deployment.