COVID-19 Modelling Consortium: quantitative epidemiological predictions in response to an evolving pandemic

Lead Research Organisation: University of Warwick
Department Name: Mathematics

Abstract

Since the beginning of the COVID-19 pandemic in early 2020, mathematical and statistical modelling have been used to provide estimates of the epidemic in the UK, and to make short- and long-term predictions about the impact of interventions. The teams of epidemiological modellers and statisticians in our JUNIPER (Joint UNIversity Pandemic Epidemiological Research) consortium represent a core of committed and experienced university research groups that have dedicated themselves since February 2020 to generating predictions, forecasts and insights. These findings feed directly into the Scientific Pandemic Influenza Group on Modelling (SPI-M) and the Scientific Advisory Group for Emergencies (SAGE), both of whom advise the UK government on scientific matters relating to the UK's response to the pandemic. As part of SPI-M this group has brought together a range of analyses to underpin diverse policy decisions including early estimates of the scale of an uncontrolled epidemic, reasonable worst-case scenarios and the impact of reopening schools.
Moving forward, critical research gaps remain unaddressed, and further translational work must be conducted to generate the necessary insights. The requested funding will ensure these key groups, with their extensive experience of delivering science for policy and deep understanding of this outbreak, will be able to continue and expand their activities. The Juniper consortium members will continue to respond to rapid requests from the UK government via SPI-M and SAGE, including providing weekly forecasts of the reproductive number R and growth rate in the UK and predictions of the likely impact of policy decisions and interventions. The research teams will be flexible and adaptive to the changing phases of the epidemic, and will proactively consider novel methodology, analysis or modelling that is required, as well as horizon scan the impact of new scientific findings and how this will impact on current and future modelling.
The programme of work will address a core set of eight overarching questions that the consortium has identified as being important over the next 12-18 months: 1. How to best address issues around the storage, curation, and processing of the growing number of COVID-related data streams 2. Improving statistical and computational fundamentals for outbreaks 3. Refining methodology for the detection of hotspots or regions in need of greater control 4. Developing bespoke methods to analyse and model Surveillance, Test and Trace 5. Refining methodologies to determine risks posed by structured environments such as workplaces, care homes, hospitals, schools, universities 6. Producing realistic individual-scale modelling of contemporary social interactions 7. Implication of finer-scale individual-level characteristics and impacts of short- and long-term immunity in models. 8. Detailed retrospective analysis of the first wave.
Our consortium will embed these scientific activities within an open and collaborative framework, including considerable public outreach so that scientific assumptions and findings are effectively communicated. Our consortium will be outward-facing and inclusive, helping to add value to a range of existing and new COVID-19 activities. We aim to build national capacity and the proposed programme will also contribute to training the next generation of applied epidemiological modellers.

Technical Summary

Mathematical and statistical modelling has been hugely influential providing rigorous estimates of the COVID-19 epidemic in the UK and making short-term and long-term predictions for decisions on interventions.
We are leaving the first phase of this epidemic, cases are slowly declining but there are local outbreaks and variation between regions is of increasing importance. Although standard epidemiological modelling tools have worked well so far, a suite of new tools are now needed that can deal with spatially- and socially-structured stochastic dynamics and population heterogeneities.
The teams of epidemiological modellers and statisticians in this consortium represent a core of committed and experienced research groups that have dedicated the last six months to generating predictions, forecasts and insights feeding into SPI-M and SAGE. To tackle the challenges of the next 18 months, these teams require investment in staff time and personnel. The proposed consortium will support these established and collaborating research teams, build national capacity and help train the next generation of applied epidemiological modellers.
We have developed a core set of eight overarching questions that we feel underpin the future challenges that will need to be addressed by SPI-M, SAGE & JBC:
1. Data collation, processing and analysis
2. Statistical and computational fundamentals for outbreaks
3. Detection of hotspots or regions in need of greater control
4. Surveillance, Test and Trace
5. Structured environments (workplaces, care homes, hospitals, schools, universities) 6. Realistic individual-scale modelling of contemporary social interactions
7. Implications of finer-scale individual-level characteristics
8. Detailed retrospective analysis of the first wave
 
Description Calculation of R and Medium Term Predictions
Geographic Reach National 
Policy Influence Type Participation in a advisory committee
Impact The reproduction number (R) is a commonly used metric for gauging the spread of the SARS-CoV-2 pandemic in the UK and the impact of control measures. It is generated by combining around 15 estimates (the precise number of estimates varies week-by-week), with the JUNIPER consortium providing 11 out of these 15 estimates. Medium Term Projections are a useful planning tool across government, translating recent behaviour into projections for the next 4-6 weeks. JUNIPER consortium members regularly provide over half of all medium term projections.
 
Description Predictions for Relaxation RoadMap
Geographic Reach National 
Policy Influence Type Participation in a advisory committee
Impact In January / February of 2021, modelling work within the JUNIPER consortium together with work from Imperial College was used to provide a predictive underpinning of the government road-map for releasing restrictions. This was extensive work commissioned through SPI-MO (of which 12 members of the JUNIPER consortium are part), with the aim of forecasting the potential impact of relaxing control measures over different time scales. The work was presented to SPI-MO, SAGE by members of JUNIPER and was then communicated to members of government.
 
Description Vaccine priority Phase 1 and 2.
Geographic Reach National 
Policy Influence Type Participation in a advisory committee
Impact Work from the JUNIPER consortium was pivotal in providing quantitative information on the likely impact of vaccination, and the ordering of priority groups for both Phase 1 and Phase 2 of the UK vaccine role out. This was directly presented to the Joint Committee on Vaccination and Immunisation (of which Keeling is a member). Our initial work clearly showed that an age-prioritised ordering had considerable advantages over other schemes in terms of reducing mortality and morbidity over short time scales. Our second piece of work considering the optimal timing of vaccinating the Phase 1 group, and helped to support the recommendation that there should be a 12-week interval between first and second doses. (This work is currently being written-up for publication). Our final piece of work considered the optimisation of Phase 2, and showed relatively little difference between any ordering of groups - speed and hence simplicity was key; this helped to support the roll-out of Phase 2 in age-order.