Kettering General Hospital (KGH) recently completed a pilot evaluation to assess the AI fracture detection tool, RBfracture™. The aim was to reduce missed fracture rates in the hospital’s Accident and Emergency (A&E) department, especially in the out-of-hours, where acute musculoskeletal exams are not immediately reported by radiologists or reporting radiographers.
The pilot aimed to measure the impact of RBfracture in an NHS setting. The assessment of RBfracture involved a retrospective audit of the algorithm’s standalone performance and a prospective analysis of missed fracture rates.
A total number of 14,791 patient exams were processed during the pilot period and a total of 319 patient exams were included in the audit of the RBfracture standalone performance.
RBfracture’s performance was assessed by calculating accuracy, sensitivity (true positive rate) and specificity (true negative rate) on a dataset of 319 cases selected by KGH.
RBfracture showed an outstanding performance, with sensitivity of 94%, specificity of 94%, and accuracy of 94% in the audit.
Missed fracture incidents are flagged retrospectively when the initial interpretation made by the A&E staff is compared to the official radiology report.
Prior to the introduction of RBfracture, the missed fracture rate ranged between 1.9 to 4.5 per 1000 examined patients. Following its introduction in August 2023 and through January 2024, 15 cases of missed fractures were reported, equivalent to a missed fracture rate of 1.0 per 1000 patients. This represents a 47%-62% reduction in missed fracture rate when compared to the second half of 2021 and 2022, respectively. Of the recorded 15 missed cases, 8 were indicated up by RBfracture, but still missed by the A&E.
The deployment of RBfracture at Kettering General Hospital has improved the initial assessment of X-rays. It has enhanced medical professionals’ performance, resulting in quantifiable benefits such as a significant decrease in missed fracture rates. This improvement translates to better patient outcomes, reducing the likelihood of incorrect discharge and subsequent readmission, thus alleviating both cost and care burdens.
