University of Exeter leads “rapid response” data analysis to aid NHS during COVID-19 pandemic

The research team, led by Professor Gavin Shaddick from the University of Exeter, has developed a new data modelling tool to help forecast demand on crucial NHS resources in the region.

The forecasts have been used to help predict the demand for in-patient beds, intensive care, PPE, ventilators, oxygen and testing kits. Crucially, it has also incorporated not only predicted hospital staffing requirements, but also the levels of staff that may be off-duty due to illness and self-isolation.

The team of data scientists and health professionals have been collaborating to produce high quality models and data to aid quick decision making, within a rapidly evolving environment, to influence key healthcare decisions.

Nic Harrison, Principal Analyst and NHS Lead for Northern and Eastern Devon collaborative COVID analysis and modelling said: “This is work that the NHS is having to deliver at pace, so we are delighted that the University is helping to support us with this at such a challenging time so that key decisions are based on the strongest possible evidence base.”

Angela Hibbard, Director of Finance and Performance at NDHT and also representing the RD&E, has been leading the project for the local NHS. She described the tool as “being hugely important in informing the  decision-making within both Trusts during the COVID-19  crisis. It has helped us make well informed and evidence-based decisions across a range of key issues such as PPE and ventilators being available to frontline staff when they need them. More importantly it is helping us model scenarios as we start to come out of lockdown and what this may mean to our hospital capacity requirements going forward.”

Professor Shaddick, Chair of Data Science and Statistics at the University of Exeter said: “This is a truly collaborative effort, and it has been an honour to work with our NHS colleagues over the past weeks. The immediate response from colleagues across the University who offered to contribute was fantastic and by working closely with our NHS colleagues we were able to produce prototype models and initial results in under a week.”

The forecasting model integrates epidemiological modelling with statistical forecasting, using both national and local data.

The analysis works by comparing local patterns of the spread of COVID-19 with other areas both nationally, and abroad. The researchers are able to use daily ‘live’ data to adjust their forecasts as the spread of the disease evolved in local populations.

The latest aspect of this project has been to consider the impact of a wide range of scenarios which may occur when lockdown is released.

The forecasting model developed by the University team has been made available online to a group of senior clinicians, including microbiologists and disease infection and prevention specialists, and executives, as well as public health and commissioning experts.

This reference group will be giving feedback to the team so the model can be iterated to best reflect the latest research and understanding in the field.

Professor Shaddick continued: “Each day, new data becomes available that helps our understanding of the spread of COVID-19 in the region. We are now using this data to produce daily forecasts to provide the most up-to-date information to the NHS”.

The vital research was conducted by the University of Exeter and the NIHR Applied Research Collaboration South West Peninsula (PenARC), in conjunction with the Northern Devon Healthcare Trust (NDHT) and the Royal Devon and Exeter Foundation Trust (RD&E).

The Data Scientist research team includes Turing Research Fellows James Salter and Fiona Spooner, Institute of Data Science and Artificial Intelligence Research Fellow Oliver Stoner, and UKRI CDT in Environmental Intelligence postgraduate students Chris Kerry, Josh Redmond and Arthur Vandervoort.

They have been working with clinician and PhD researcher Rob Challen, and Martin Pitt, Stuart Logan and colleagues in the NIHR PenARC team.