1 Recommendations

Can be used while more evidence is generated

1.1

Four artificial intelligence (AI) technologies can be used in the NHS during the evidence generation period as options to help healthcare professionals detect fractures on X‑rays in urgent care. The technologies are:

  • for people of any age:

    • Rayvolve

    • TechCare Alert

  • for people 2 years and over:

    • BoneView

    • RBfracture.

      These technologies can only be used:

  • if the evidence outlined in the evidence generation plan is being generated

  • once they have appropriate regulatory approval including NHS England's Digital Technology Assessment Criteria (DTAC) approval.

1.2

The companies must confirm that agreements are in place to generate the evidence (as outlined in NICE's evidence generation plan). They should contact NICE annually to confirm that evidence is being generated and analysed as planned. NICE may withdraw the guidance if these conditions are not met.

1.3

At the end of the evidence generation period (2 years), the companies should submit the evidence to NICE in a form that can be used for decision making. NICE will review the evidence and assess if the technologies can be routinely adopted in the NHS.

Can only be used in research

1.4

More research is needed on qMSK to help healthcare professionals detect fractures on X‑rays of adults in urgent care before it can be used in the NHS.

1.5

Access to qMSK should be through company, research or non-core NHS funding, and clinical or financial risks should be appropriately managed.

What evidence generation and research is needed

1.6

Evidence generation and more research is needed on:

  • the diagnostic accuracy of fracture detection in urgent care by healthcare professionals with and without the help of AI technologies

  • costs and clinical outcomes associated with different fracture types and missed fractures

  • fracture clinic referral rates with and without the help of AI technologies

  • any clinically significant changes in treatment decisions for fractures detected with and without the help of AI technologies

  • AI software failure rates and reasons for failure

  • detection of or failure to detect clinically significant non-fracture-related conditions by healthcare professionals with and without the help of AI technologies

  • the diagnostic accuracy of AI technologies to help healthcare professionals detect fractures in different populations

  • implementation costs of AI technologies in different urgent care centres.

    The evidence generation plan gives further information on the prioritised evidence gaps and outcomes, ongoing studies and potential real-world data sources. It includes how the evidence gaps could be resolved through real-world evidence studies.

Potential benefits of use in the NHS with evidence generation

  • Clinical benefit: Clinical evidence suggests that the AI technologies may improve fracture detection on X‑rays in urgent care without increasing the risk of incorrect diagnoses. This could help reduce the number of fractures that are missed in urgent care, which would reduce the 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.

  • System benefit: AI technologies may help reduce variation in standard care by providing a consistent baseline for X‑ray interpretation unaffected by differences in staff experience or resources between centres. AI technologies would also be unaffected by factors such as staff fatigue, distractions, or working outside normal hours. In centres or at times where definitive radiology reports are available before people are discharged (hot reporting), the benefits of AI assistance may be lower.

  • Resources: Reducing the number of fractures that are missed at initial interpretation would also reduce the number of people that reattend urgent care after discharge or are recalled to hospital after radiology review. Early results from the exploratory economic modelling show that the AI technologies could be cost effective.

  • Equality: AI technologies have the potential to reduce geographical inequalities in X‑ray interpretation and fracture detection, because they may improve fracture detection in smaller centres with fewer and less-experienced staff. It could also reduce inequalities in provision of service, because there may be improvement in service outside normal hours.

Managing the risk of use in the NHS with evidence generation

  • Clinical risk: Using AI technologies to help detect fractures on X‑rays in urgent care is considered to have a low clinical risk. This is because they are used in addition to standard care in which healthcare professionals make treatment decisions. Additionally, AI technologies do not replace the definitive radiology review. The available evidence suggests that the AI technologies may improve the accuracy of fracture detection.

  • Implementation guidance: Clear local protocols will need to be in place when using AI technologies. This is to ensure that healthcare professionals are confident about what action to take when there is disagreement between the healthcare professional and AI technology.

  • Costs: There is uncertainty around the cost of some software and the true cost of implementation and ongoing post-market surveillance. Costs were estimated at £1 per scan in the exploratory economic modelling. Centres implementing AI to help fracture detection should ensure the cost per scan is similar to the estimated cost. This guidance will be reviewed after the evidence generation period and the recommendations may change. Centres should take this into account when negotiating contract durations and licence costs.

  • Impact on workforce: If using AI technologies to help fracture detection becomes more widespread and part of the standard diagnostic pathway, there is a risk of over-reliance on the technologies. This could potentially lead to deskilling of the healthcare professionals who interpret the X‑rays. It may also reduce the level of scrutiny for non-fracture-related conditions. This risk could be mitigated if healthcare professionals interpret X‑rays before viewing the AI results.

  • Resources: There is a low risk that using AI technologies to help detect fractures on X‑rays may increase fracture clinic referrals and requests for further imaging such as CT or MRI. This should be monitored during evidence generation to inform local fracture detection protocols.

  • Limitations of AI for subgroups: The AI technologies may not be suitable for use in certain groups, for example, children and young people or people with conditions that affect bone health. Centres should ensure that the AI technologies are used within their indications and any limitations are acknowledged and clearly explained to patients.

  • Equality: There is a risk that the AI technologies may have reduced diagnostic accuracy in different populations, such as people from ethnic minority backgrounds or people with low socioeconomic status. Healthcare professionals should take this into account when interpreting X‑rays of people in these groups.