Osteoporosis is a common bone disease that causes weaker bones that break after a fall and lead to chronic severe pain and life-changing complications. There are treatments available that strengthen bones and lower the chance of another fracture. Many hospitals have specialist teams dedicated to finding patients who have broken a bone to prevent the next fracture. We know breaking a bone in the spine, a spine fracture, is a critical sign of osteoporosis. However, most patients don't realise they have a spine fracture as they cause temporary back pain that is often ignored. We know that over three million people have scans that include the spine, for other reasons such as lung or bowel problems. Up to 1 in 20 scans could be showing there is a spine fracture. Currently, very few fractures are recognised or acted upon. This project will use an innovative Artificial Intelligence (AI) software that automatically looks at existing CT scans to find these spine fractures and brings them directly to the specialist team's attention, to see if the patient needs bone-strengthening lifestyle advice and medicines. We call this an 'AI-enabled spine fracture pathway'. We will show that using the pathway will improve patient health and reduce costs to hospitals using multiple methods.
Background
We know that breaking a bone after a fall in adults aged 50 years and older could be the first sign of osteoporosis and all these adults should be tested for osteoporosis. There are several approved treatments that can strengthen bone and protect individuals from breaking a bone. The challenge is to find patients at risk of osteoporosis and make sure they get tested and treated for osteoporosis. While breaking an arm or a leg is obvious, many people have breaks of their spine bone that are called vertebral fractures. These vertebral fractures can cause minimal back pain that is often ignored and so most people with a vertebral fracture do not know they have one. However, breaking a bone in the back is still a very strong risk factor for osteoporosis and patients with a broken spine bone need further testing and potential treatment for osteoporosis.
We know many adults have CT body scans for other reasons. Many of these scans include the back and can detect broken spine bones if they are present. However, because the reason for the scan was not to look for broken bones, most reports do not mention the broken backbone if present. Even if it is reported it is not acted upon, so the broken spine bone remains invisible to the patient and their healthcare teams. It would take too long for hospital teams to look at the millions of CT scans in the NHS again.
Instead, an Artificial Intelligence (AI) programme is being used as part of routine NHS care that will look at old CT scans and flag ones that could have broken a spine bone. The AI programme has been developed by a company called Nanox and is certified and approved for use in the NHS. As part of routine NHS care, all the scans the AI programme identifies with a with a potential vertebral fracture will be confirmed by the local hospital team, who will then contact them to arrange further routine NHS care to improve the patient’s bone health. To measure the benefit to patients we will study the outcomes of patients who had a scan in 2017 and those who had a scan more recently in 2022 focusing on hospital admissions. The results of the comparisons between patients will be used to inform the NHS use an AI-enhanced vertebral fracture prevention pathway for patient benefit.
Aims and objectives
- To describe the performance of the AI-enabled vertebral fracture identification platform compared with NHS radiology reports and local readers.
- To describe the implementation of an AI software in NHS hospitals and develop a supporting toolkit.
- To describe how the AI-FLS pathway improves key performance indicators for patient’s identification, assessment, treatment recommendation and adherence.
- To describe the clinical and cost effectiveness of the AI-FLS pat
Study design
- Clinical audit
- Focus groups
- Interrupted time series and matched design
- Matched parallel cohort with historical controls.
TO FIND OUT MORE
Please email kassim.javaid@ndorms.ox.ac.uk.