Using AI to predict severe system pressure up to 10 days in advance
For over 10 years, we’ve been delivering the UK’s flagship AI talent development programme: the Faculty Fellowship. Now in its 28th cohort, the Fellowship matches the UK’s best STEM academics with our partner organisations to solve key business challenges using data science and AI.
For over 10 years, we’ve been delivering the UK’s flagship AI talent development programme: the Faculty Fellowship. Now in its 28th cohort, the Fellowship matches the UK’s best STEM academics with our partner organisations to solve key business challenges using data science and AI.
Our Fellows receive world-class training in commercial and technical skills from Faculty staff, followed by six weeks embedded within the partner organisation scoping, validating and implementing a solution to a business problem.
To date, we have deployed seven Fellows to the NHS and trained over 50 NHS England analysts. Our clients consistently tell us that the Fellowship is a fantastic route for exploring AI opportunities within NHS organisations, with minimal outlay and producing tangible deliverables at the end. The training the Fellows receive is also available to our partner organisations, further supporting the development of their data science capabilities and upskilling of their workforce.
In the latest Fellowship cohort, Emily Thomas, a cognitive neuroscientist by trade, took up a proof of concept project with NHS Bristol, North Somerset and South Gloucestershire ICB (BNSSG) to explore the possibilities of predicting severe system pressure with readily available sitrep data.
Emily completed her PhD at Birkbeck College, University of London in 2021, where she examined how cognitive processes such as prediction and attention influence our multisensory perceptual experiences. She recently moved back to the UK to take part in the Faculty Fellowship following post doctoral research at NYU where she examined the neural mechanisms of perception across sensory modalities, using a combination of neuroscience (e.g. fMRI) and data science (e.g. machine learning) methods.
The UK’s entire healthcare system has been under unprecedented, sustained pressure since the pandemic, with both urgent and elective care recording the highest waiting times for a generation. The aim of Emily’s project with BNSSG was to use AI to develop a proof of concept early warning system for predicting Severe System Pressure (SSP) over the next 10 days, helping key decision makers to make targeted interventions to best alleviate strain on services and improve care.
SSP is an informal, emerging metric and, for the purposes of this project, was defined by two key variables: A&E wait times between the decision to admit to being admitted and ambulance response times to Category 2 999 calls. Emily’s project focused on using AI to create a single prediction of the likelihood of entering SSP over the next 10 days using an average of these variables from three different hospital sites.
The project explored trends in seasonality, identifying patterns on how these variables changed over time which can help predict similar patterns in the future. Emily incorporated a range of internal datasets into the AI models for greater accuracy. They included the number of active 999 calls, hospital admissions, and the number of patients waiting to be offloaded from an ambulance.
The outcome of Emily’s project with BNSSG was a first-of-its-kind tool for predicting SSP using AI. This tool has the potential to address many of the operational challenges faced by ICSs across the country by providing accurate predictions of SSP likelihood, with enough foresight to act on these predictions and prevent the pressure occurring.
Reflecting on the project, Emily observed:
“This proof of concept is a great step forward in showing how we can use AI tools to better prepare health systems for periods of strain. Understanding of which factors are most explanative of system pressure helps to form targeted action plans for mitigation, such as increasing on-call ambulance staff during peak periods to mitigate the impact upstream to ambulance response times and downstream to A&E wait times.
With these predictions in hand, hospitals will be able to deploy resources more efficiently, meaning that patients may no longer deteriorate waiting for care, improving patient experience and achieving better healthcare for us all.”
Richard Wood (Head of Modelling and Analytics at BNSSG ICB) provided the following insights on the project:
“The project has been valuable in helping to demonstrate the proof of concept of developing a Severe System Pressure indicator. It has helped to give us an idea of what an approach may look like, and what may be achievable in terms of the kind of accuracy we could expect.”
On the Fellowship itself and working with BNSSG, Emily said:
“It was great working with colleagues that had such in-depth technical knowledge, not just of data science and artificial intelligence techniques, but also the operational processes and challenges facing the NHS, that ultimately contributed to making this project a success.”
For more on the project watch Emily’s final presentation as part of the Fellowship.
Would like to learn more about Emily’s project, the opportunities for operational AI within the NHS, or Fellowship opportunities within your organisation? Let’s start a conversation.
