Artificial intelligence technologies to help detect fractures on X-rays in urgent care: early value assessment

Fracture assessment and diagnosis in urgent care typically involves triage in which a nurse, advanced clinical practitioner or doctor will do an initial assessment before requesting imaging. X‑rays taken by a diagnostic radiographer are usually the first-line imaging approach for non-complex fractures. X‑rays are initially interpreted by a healthcare professional in the urgent care centre, who may be assisted by a preliminary clinical evaluation from the diagnostic radiographer. Multiple surgical and non-surgical treatment options are available depending on the type of fracture.

NICE's guideline on non-complex fractures recommends that a radiologist, radiographer or other trained reporter should review X‑rays and provide a definitive report before the injured person is discharged (hot reporting). Clinical experts explained that in practice this is not always possible and reporting delays can occur ranging from days to weeks.

Missed fractures are reported to be the most common diagnostic error in the emergency department. Hussain et al. (2019) found that 44% of diagnostic errors in fractures resulted from inappropriate response to imaging. Missed or delayed diagnosis of fractures on radiographs is reported to occur in around 3% to 10% of cases (Kuo et al. 2022).

Artificial intelligence (AI) technologies that can help healthcare professionals detect fractures on X‑ray images could improve the accuracy of fracture diagnoses in urgent care. This could help reduce the:

• number of fractures that are missed before a radiologist or reporting radiographer reviews the X‑rays

• number of people being recalled to hospital after radiology review

• risk of further injury or harm to people during the time between the initial interpretation and treatment decision in urgent care and the definitive radiology report

• number of unnecessary referrals to fracture clinics.
  
  AI technologies may also improve consistency of X‑ray interpretation when the ability of healthcare professionals to interpret X‑rays is reduced, for example, when they are tired, distracted or working outside normal hours.

The technologies included in this early value assessment are standalone software that use AI-derived algorithms to analyse X‑ray images to detect fractures. They are intended to be used as decision aids for healthcare professionals interpreting the X‑ray image. Some companies provide the software directly, whereas others provide it through multivendor platforms. The technologies use X‑ray radiographs in digital imaging and communications in medicine (DICOM) format, which are stored on the hospital's picture archiving and communications system (PACS). Images are then interpreted using proprietary AI-derived algorithms. The technologies included in this assessment are shown in table 1.

The comparator is standard care for fracture assessment in which the urgent care healthcare professional interprets the X‑ray radiograph without AI assistance.

The reference standard is based on the consultant radiologist or reporting radiographer interpretation and report.