We are taking part in a collaboration with NHSx, alongside the Big Data Institute at the University of Oxford (led by epidemiologist Christophe Fraser's group) and IBM.
Together, we are modelling how Covid-19 spreads in communities and which policy interventions are most effective at halting its spread. For example, this might be modelling a contact-tracing approach, or quarantining sections of the population, and comparing this to how transmission is affected within the community.
Recently, our group was specifically tasked with modelling nosocomial transmissions (the spread within hospitals) between healthcare workers and patients. We have also modelled how bed availability within the hospital wards changes with the peak of the pandemic, and how a lack of beds can change the progression of the disease.
This involved working closely with epidemiologists at Oxford in Christophe Fraser's group, who first developed the model, and interviewing healthcare workers to fully understand how best to describe this intricate system. IBM were tasked with optimising the codebase and took charge of day-to-day project management and our ‘scrums’, where we gave them feedback on our progress.
As data protection is imperative at the NHS, we worked without any access to actual patient data. We focused on designing and programming the infrastructure code of the hospital feature in the model, which is now with another team at NHSx. This team have patient-data access clearance, so they can calibrate the parameters we created to the real data.
Our work is interdisciplinary – most of us are computational modellers, several with an interest in software engineering, but we’re integrated with a small in vitro wet lab who also perform in vivo imaging in collaborations. Up until the pandemic started, our focus was modelling both individual and collective cell migration and decisions.
For another project, we’re using 3D imaging of vessels in mouse retinas to understand abnormal vessel growth in eye disease. The team was also in the process of starting up experiments growing vascular endothelial cells on micro-patterned substrates to validate model predictions (and then the labs shut down!)
Fortunately, a lot of us computational researchers can work quite well at home. On this project, three of us in the lab switched our activities almost entirely in order to develop the model required by NHSx as quickly as possible.
An interesting change to our workday, that I mentioned earlier, was our daily (virtual) ‘scrums’ when we met with the IBM team and others at NHSx and Oxford. This involved chatting with the teams over video calls, summarising what we had achieved and planning our work for the day using shared Trello boards, which is a standard (maybe obvious) way of keeping software development focused and efficient. These scrums certainly helped us learn a lot more about industry standard software engineering and project management across teams of developers.
The team at NHSx are now calibrating the feature we developed to the patient data they can access. We are keeping in touch and will continue to support and interact with the NHSx teams on this codebase, but we’ve mostly moved back to our Crick lab to work on our normal projects.
One interesting new direction has arisen from this – we are now developing an extension of the open-source hospital feature model we developed to help simulate COVID spread and effects among cancer patients. It’s part of a project led by Crick group leader Samra Turajlic, integrating with her longitudinal clinical CAPTURE study based at the Royal Marsden – watch this space!
