The impact of Artificial Intelligence (AI)

Adapted from 'Healthcare Digital Transformation'¹¹

The digital health future has the potential to:

• enable and support online patient access and experiences
• improve the experiences of caregivers
• enable use of digital administrative functions
• enhance health and wellbeing in communities

Osteoporosis is a highly prevalent skeletal disease that causes bones to become weaker and more prone to fractures. In the UK, approximately 33% of women and 20% of men aged over 50 will suffer an osteoporotic fracture. Annual NHS costs of treating these fractures are predicted to rise to over £5.5 billion by 2025. Fractures in the vertebrae, the bone in the spine, are often the earliest clinical manifestation of the disease, so identifying them is key to providing appropriate management to reduce the risk of future, more compromising fractures. Inexpensive drug treatments are available that can halve fracture risk. However, osteoporosis is under-diagnosed in clinical practice. In particular, vertebral fragility fractures (VFFs) are visualised incidentally by approximately 20% of CR procedures in the over-50s acquired for other clinical indications, but only one quarter of them are accurately reported by radiologists and only 2.6% of these patients are referred for appropriate management. The University of Manchester, in collaboration with the Manchester University NHS Foundation Trust and Optasia Medical Ltd., has developed computer-aided diagnostic software to automatically identify VFFs in Computerised Tomography (CT) images. It uses machine learning technology to identify vertebrae visualised in CT images and diagnose any VFFs present.

The software forms the basis of ASPIRE^TM, a tele-radiology service that allows hospital radiology departments to out-source the secondary reporting of VFFs on CT images. ASPIRE^TM combines high levels of automation with oversight from radiologists to improve the efficiency and accuracy of VFF identification compared to in-hospital reporting, whilst reducing clinical workload within radiology departments. 

The number of patients undergoing chemotherapy is estimated to increase by 5-7% per passing year. This will increase the workload of pharmacists who validate the regimens. Also, in due time more pharmacists will be needed to cover the extra capacity to deal with the increased number of validations.

The process of validating a regimen, called screening, is based on well defined protocols with set criteria and baselines. 90% of the screenings are proved to be correct with no further actions needed. And only 1-2% of the rest of the screenings lead to a need for change in the regimen prescribed. Working with The Christie NHS Foundation Trust we explored whether the use of AI can reduce the time needed to screen a regimen, and therefore free up the time of the pharmacists while supporting a bigger capacity of screenings for chemotherapy patients.

The study aims to assess the feasibility of devising a new AI tool to be embedded in existing workflows to support the decision process for regimen selection within clinical systems, as well as to formulate the framework in which Machine Learning (ML) work-flows can facilitate improved patient outcomes and optimise operation clinical pathways.

• Provide a sound basis on which to evaluate the completeness and accuracy of data for the purposes of initially validating prescribed regimens leading to decision support for regimen selection within clinical systems.
• Provide efficiencies for the pharmacy department through freeing up staff from a manual process.

We are following an Agile-inspired methodology based on an incremental and iterative approach to collect information needed and develop data journey models representing the data landscapes. We work closely with the stakeholders involved in the workflows and patient pathway, establishing ongoing communication channels, and taking into consideration their feedback.

An initial analysis based on a focus group and observation sessions with a lead pharmacist on Trastuzumab SC EBC regimen, showed that the screening workflow can be supported by a decision support tool to flag up regimens that require extra verifications to be made by a pharmacist. The majority of pharmacists’ time lies in tracking the right documents in time-order to identify the latest values. This leads to an opportunity for a tool/solution to support the data provenance of information integrated in the data warehouse.