Mihaela van der Schaar
If you’ve been reading my updates, you’ll know that my team and I have recently been thinking hard about ways to use machine learning to help hospitals predict and control their usage of thinly-stretched resources amid the COVID-19 pandemic.
Specifically, we have been adapting a system we recently developed, called Cambridge Adjutorium, to make hospital-level projections of upcoming demand for ventilators and ICU beds. A couple weeks back, I mentioned that we had successfully trained Cambridge Adjutorium using depersonalised COVID-19 patient data kindly provided by Public Health England.
For us, that was a huge and helpful breakthrough, and we managed to create a working demonstrator showing that Cambridge Adjutorium could, indeed, accurately predict hospital resource usage.
Today, I’m delighted to announce an even bigger breakthrough: a partnership between my Cambridge team, NHS Digital and Public Health England to start trialling Cambridge Adjutorium at a number of Acute Trusts in England.
The partnership
So far, our team has been adapting Cambridge Adjutorium and training it on COVID-19 patient data. We have been working with a dataset of 4,000 patients, and have demonstrated that the accuracy of the system surpasses existing state-of-the-art models. We’ve been building the system in collaboration with clinicians, since they’re our intended users.
We’ve also been in discussions with Public Health England and NHS Digital with the aim of finding ways to deploy the system at hospitals and continue to improve it based on their feedback. Obviously a rollout at scale on day 1 isn’t possible or helpful, so the number of hospitals would need to be limited before considering scaling up.
In the last couple days, we agreed with NHS Digital and Public Health England to help enhance Cambridge Adjutorium and use it to power the Capacity Planning and Analysis System (CPAS), which will initially be trialled at at a number of Acute Trusts (alpha sites) in England in order to demonstrate its viability for broader usage. Basically, it’s an alpha phase.
Looking to the future
Helping clinicians save lives is at the very core of our work, so we’re thrilled to be able to work with NHS Digital and Public Health England on this unique and pioneering project. I’m grateful to our partners for their open-mindedness and vision, and to our colleagues on the clinical side for sharing the challenges they face and guiding us as we try to solve them.
For now, our hope is that this trial phase will demonstrate that CPAS, powered by Cambridge Adjutorium, is able to provide reliable, understandable and actionable information to clinicians. We want to use this period to listen and learn, and to continue to work closely alongside Public Health England and NHS Digital on developing a system that can be rolled out more broadly across the NHS.
Going forward, I’m cautiously optimistic about the prospect of turning what we’re doing today with COVID-19 into a broader framework for helping hospitals to manage their resources. COVID-19 is a terrible disease that has torn families and friends apart, brought the UK to a virtual standstill, and upended the lives of countless healthcare workers. But if we can use projects like this to strengthen collaboration and lay the foundation for more robust digital infrastructure for healthcare, we can come out the other side of this crisis even stronger.
General info about Cambridge Adjutorium
Cambridge Adjutorium is a prediction system driven by a state-of-the art machine learning model. The system was developed by my team based in Cambridge.
Cambridge Adjutorium was initially conceived as a prognostication tool for cardiovascular disease, but from the outset was created for use with a broad range of diseases and conditions. Since then, it has been validated for cystic fibrosis and breast cancer. Training the system on a depersonalised COVID-19 patient dataset provided by Public Health England has shown that its predictive accuracy far surpasses existing state-of-the-art techniques.
Cambridge Adjutorium can provide aggregated predictions for hospitals, which could significantly help improve capacity planning for healthcare systems in response to COVID-19. The system uses its underlying predictive models to provide accurate near-term projections of the likely demand on hospital resources such as ICU beds and ventilators. These projections are shown to healthcare decision-makers in an easy-to-interpret and actionable format.
I’ve recently made an introductory video about the system in collaboration with my team members and one of our clinician partners, Dr. Ari Ercole. It’s fairly short and sweet; you can find it embedded below.
If you want to learn more about our work, leave a comment below, message me or email my team at mv472@cam.ac.uk and nm736@cam.ac.uk.
You can find the NHS Digital press release here.
Nabeel Seedat
Andreas Bedorf
Jeroen Berrevoets
[…] AutoPrognosis does not only issue predictions. It also issues interpretations in order to ensure that the predictions are not black boxes, and can be understood by clinicians in a way that’s helpful and meaningful. The result is an end-to-end automated pipeline that takes clinical data as an input and provides predictions and explanations as outputs. Since 2018, my team has applied AutoPrognosis in many clinical applications showing how autoML can outperform other methods – both statistical and machine learning approaches. We’ve made new and meaningful discoveries about lung transplant risk factors for cystic fibrosis patients (see our article in Nature here), individualized risk predictions for breast cancer (see The Alan Turing Institute story here) and cardiovascular disease (using UK Biobank), among other applications. Additionally, AutoPrognosis is at the core of our Adjutorium system, which is currently being deployed by the NHS to help hospitals with capacity planning in the face of the COVID-19 pandemic (more here). […]