RAPID ASSISTANCE IN MODELLING THE PANDEMIC: RAMP

A call for assistance, addressed to the scientific modelling community

Coordinated by the Royal Society

Update August

RAMP has now reached the end of its initially intended duration of April-July inclusive. This means that those who have volunteered for its various initiatives should feel no continuing obligation to remain involved. However, we hope many volunteers will in fact remain involved, and indeed that the majority of RAMP's activities will continue in some form or another. Our continuity plans are described below in an update of our last progress report which is still available on our "Previous updates" page.

Now that the official volunteer phase of RAMP is formally at an end, we once again thank all RAMP volunteers for their support, including those whose offers of assistance were not ultimately called upon by RAMP.

Continuity Planning

1. UKRI funding has been secured to maintain the infrastructure of the RAMP forums until late 2021. The same funding will also support a central data, coordination and software activity, led by Graeme Ackland from Edinburgh. The RAMP Forums will therefore continue to operate largely as now, and we hope that many people who have developed an interest in COVID-19 modelling and science will actively participate by nominating and evaluating papers for community review. Papers of potential policy relevance will then continue to be screened by RAMP's Rapid Review Group (RRG) and passed by them through to Government advice channels. These channels are currently being restructured at national level, with the Joint Biosecurity Centre (JBC) taking increasing responsibility for day-to-day COVID-19 planning within England, with SAGE and SPI-M increasingly focussed on longer-term aspects of pandemic planning. Although COVID-19 planning decisions are increasingly political in character, the UK's national COVID-19 response continues to depend crucially on the latest science, which is evolving at pace. Therefore involvement of the wider RAMP community in screening and commenting on this science, via the RAMP Forums, will remain invaluable for the foreseeable future.

2. RAMP is now working on a funding bid to UKRI to secure a further 18 months funding to cover (a) a series of meetings, workshops and study groups, primarily organized by the Isaac Newton Institute, that will allow continued discussion among the many interdisciplinary teams that RAMP has established; (b) continuation of RRG's activities; and (c) continued engagement by staff of the Royal Society Science Policy Centre to ensure that scientific outputs identified as important by RAMP (including but not limited to those generated by its own teams) are communicated to the most relevant parts of Government as the latter evolve during the transition from emergency to "new normal" response modes. These all build on working patterns that have become well established in the past few months, with (a) building on the Newton Institute programme on Infectious Dynamics of Pandemics which runs through to the end of September, and (b) led from the outset by an Oxford team headed by Alain Goriely and Philip Maini.

3. Neither of the above UKRI bids will directly fund researcher time for continuation of the many projects RAMP has begun. These are so numerous and diverse that any kind of centralized funding structure for their continuation would be unmanageable, although the forums, RRG, and networking activities described above will all add value to the new ecosystem RAMP has created by maintaining strong collaboration channels between its various parts. Instead, individual RAMP Task Teams and Projects have been encouraged to apply to UKRI for their own continuation funding with official RAMP support. Six such bids have so far been made to UKRI, with an aggregate proposed budget of well over £4M. Among Task Teams already submitting bids are New Epidemic Modelling (2 bids); Urban Analytics; Environmental and Aerosol Transmission (2); and Structured Expert Judgement. The remaining Task Teams on Comorbidities; Human Dynamics in Small Spaces; and Within-Host Disease Dynamics are yet to finalize their continuity plans which may include further UKRI bids in due course.

4. Many individual volunteers were seconded by RAMP to existing teams within SPI-M. As with all RAMP volunteers, there is no obligation that these secondments continue after July, but of course they can do so by mutual agreement. RAMP volunteers have played active roles in many SPI-M projects leading to a number of reports (see item 5 below). Alongside this, several groups among the RAMP Task Teams have developed close collaborations with SPI-M member teams, such as one on Structured Expert Judgement for Primary School reopenings (using head-teacher expertise) led by vulcanologist Steve Sparks and epidemiologist Ellen Brooks-Pollock of SPI-M (both from Bristol). Several of these cross-disciplinary, group-to-group collaborations are likely to remain in place for the longer term, and may well nucleate further UKRI bids beyond those mentioned already above. In some other cases, RAMP's constituent groups, such as a large team led by Richard Reeve on behalf of Scottish Covid Response Consortium (SCRC) are already directly involved in Government advice channels and may at some point join SPI-M in their own right. The SCRC project, which lies within RAMP's New Epidemic Modelling Task, now involves over 100 scientists from 21 Universities plus partner organizations such as UKAEA, Man Group PLC, BioStatistics Scotland, and Invenia Labs.

Task Teams: Progress to date

5. As mentioned already, seconded RAMP volunteers have actively participated in many SPI-M teams covering all aspects of pandemic modelling. However, because of the urgency with which SPI-M modellers have had to deliver results to Government, the ability of these teams to write up work for mainstream scientific publication has been severely impaired in recent months. (This has led to a peculiar situation where the scientific productivity of these experts may temporarily have appeared to be less than new entrants to the field.) We are currently gathering data on reports where coauthorship by RAMP volunteers has occurred or is intended. Plans are afoot to have a special issue of Phil Trans Roy Soc B in which some of these papers will appear. The possibility of a second such issue, allowing coverage of papers from RAMP Task Teams as well as SPI-M, is also under discussion. Meanwhile the Royal Society's Open Science journal actively encourages COVID-19 submissions from RAMP participants among others.

6. The RAMP forums have discussed over 250 papers and reports with about 10% of these being passed through to the RRG for Rapid Review. Since April the RRG has scrutinized about 60 nominations for Rapid Review, not only from the forums and from SPI-M but in some cases direct from Government. The RRG has commissioned rapid reviews (usually two per paper) on over 40 of the reports nominated.

7. New Epidemic Modelling

A number of groups contributed to RAMP by setting up modelling efforts from scratch. We identified a small number of lead groups, and assigned volunteers to them. We saw value in this in terms of avoiding potential "groupthink", while trialling and testing new concepts, software engineering and model comparison. The guiding principle was to do "high-risk, high-reward" work that might lead to methodological breakthroughs, at a time when the well-established teams of SPI-M were generally too busy running their models for forecasting purposes to innovate much on the model-building or coding side. RAMP's code development work has been successful on many fronts, although so far its deployment to directly inform policy has been frustrated by the difficulties (often shared by SPI-M) in accessing UK disease data more detailed than openly available. This situation is gradually changing in part by the authorities getting to grips with the widely identified 'data problem' and in part by the RAMP teams joining forces with SPI-M groups.

Highlights have so far included the wide range of interlinked models with a shared 'data pipeline' structure developed by SCRC (see item 5 above), and novel compartment models 'PyRoss' and 'PyRossGeo' developed in Cambridge. These modelling platforms allow consideration of interactions from an individual level, building upwards to make predictions at a national level. Assumptions about what processes are important for disease dynamics of course involve a degree of expert judgement, but numerical parameter estimates can be learned reliably from data, once enough of this is available. Both these platforms are already fully functional and producing results based on the publicly available data from the UK and elsewhere.

These and other New Modelling projects can best be appreciated via the following github repositories:

A second thread in the New Epidemic Modelling Task Team has been to study existing codes, to do code comparisons, as well as sensitivity and validity testing. Significant work has been done on the Imperial COVIDSim code, including Red-Team testing at Edinburgh and parameter sentivity tests at UCL, and with Microsoft Research new work has been done on the development of an open framework for pandemic modelling which facilitates model comparison and unifies data formats.

8. Urban Analytics

The Urban Analytics Task Team comprises over 30 individuals. It has developed a fully functional model for the County of Devon and has developed tools to compare intervention scenarios for this demonstrator region. A national model is being developed in engagement with both the Met Office and computer games company Improbable as technology partners. The team is cooperating with the Connected Places Catapult in code verification and in access to high-resolution mobile phone data, and is also partnering with a number of small companies to allow integration of mobility and economic infrastructure data into models that combine human behaviour with disease dynamics. A continuation funding bid has been submitted to UKRI, led by the Alan Turing Institute, with University partners from Cambridge, Exeter, Leeds and UCL, various non-academic partners and support from the JBC. Overall, substantial progress has been made towards the hugely challenging goal of creating genuinely integrated urban pandemic models during RAMP's voluntary period, and there is every possibility of genuine breakthroughs if the project is continued, as we hope, with UKRI support.

9. Human Dynamics in Small Spaces

From the open call, 11 groups volunteered their services, most of whom had a professional interest in how the Pandemic was affecting movement in small spaces such as buildings, larger complexes such as stations and malls, and streets as well as within vehicles and spaces within buildings. Unlike most of the volunteers, in this area, non-academic organisations were dominant and the work they were already doing with respect to their own normal activities was the focus of their interest in linking to other features of modelling the Pandemic which they were not expert in. The groups in question are BAESystems, CASA-UCL/Tesco, Centre for Numerical Modelling & Fire Safety Group–University of Greenwich, City Modelling Lab – Arup, Connected Places Catapult, Martin Centre –Cambridge, Network Rail, Ordnance Survey–Rapid Prototyping Team, Pamela Lab–CEGE–UCL, PWC–Artificial Intelligence Team, Technology & Investment Group, and Sainsbury’s. Each group has produced substantial demonstrators and about half have produced papers for the academic/professional audiences. All groups were represented in presentations to the Newton Institute IDP Programme. Continuing work is being developed with these groups in touch with one another. There have been two outputs thus far: Ordnance Survey's "Predicting the Geospatial Spread of Disease using Spatial Interaction Modelling with Gridded Data" and Sainsbury’s "Simulating human interactions in supermarkets to measure the risk of COVID-19 contagion at scale".

10. Environmental and Aerosol Transmission

This Task Team has engaged approximately 190 researchers led by Paul Linden (Cambridge) and Christopher Pain (Imperial). The Team is split into six subgroups, as follows: (a) Exhalation and Ventilation: This group plans to produce a final document on wintertime ventilation. (b) Aerosols: Work is ongoing to agree baseline specification to characterise the initial conditions for the production of droplet clouds resulting from various types of respiration activity. A sub-group has formed which will look for funding to investigate brass, woodwind and singing. (c) People Movement: A paper on modelling wake mixing has been accepted in the Journal of Fluid Mechanics and another paper has been submitted showing dispersal of CO2 (as a proxy for virus-laden aerosol) over 15m from source in a hospital setting. The group has developed models of dispersal and work is ongoing on these. (d) Deposition: This group is developing a continuous time-based model for looking at risk from a variety of different routes in an office environment. (e) Inhalation: This group is developing a systems-based model for infectious bio-aerosols and will apply for funding to take this forward. (f) Case-studies: This group has considered the above topics through the lens of real-life situations working with industry. One proposal looking at transportation has already been funded, and one focused on schools has been submitted to UKRI. There will be an online conference for the Task findings on September 8th.

11. Within-Host Dynamics

The task team was divided into two sub-groups – one focussed on investigating a viral load model, and the other focussed on various aspects of a multiscale spatio-temporal model. Each group met weekly via Zoom. Developing a meaningful viral-load model at this time requires good data and efforts so far have therefore focussed on finding data and datasets to analyse. Currently the group is studying longitudinal data from ICU COVID patients at Columbia Medical Center (NYC), and also lymphocyte counts in blood from 20 patients from collaborators at St George's Hospital (London). A mathematical framework for evaluating the role of nucleoside analogue drugs against coronavirus using genome sequence data has been developed and the analysis of data examining the consequences of remdesivir upon the within-host evolution of the virus has started.

For the spatio-temporal modelling, progress has been made through an interntional collaboration developing the following SARS-CoV-2 tissue simulator:

A framework paper describing a new approach to modelling pandemics (including within-host dynamics) has been accepted for publication.

Work on an agent-based model of how the virus affects lung tissue is also being developed through a 6-month grant funded by the Chief Scientist Office (CSO) Call for Rapid Research in COVID-19 programme.

Both sub-groups will continue the work and efforts reported above.

12. Comorbidities

This group undertook an investigation of the role of individual comorbidities in COVID-19 outcomes. Data on patients admitted to hospital with COVID-19 were available from the International Severe Acute Respiratory and emerging Infections Consortium WHO Clinical Characterisation Protocol UK (ISARIC WHO CCP-UK), which has collected clinical care data from 260 hospitals in England, Scotland, and Wales on patients admitted to hospital since January 2020. Comorbidities investigated included obesity, diabetes, chronic cardiac disease, smoking and 14 others, and the role of sex and age were also investigated. By incorporating Health Survey for England data on population prevalence of comorbidities we were able to estimate the risk of hospitalisation with COVID-19 by sex, age and comorbidities. Using the ISARIC data on over 40,000 patients admitted to hospital since Match 10th 2020 we estimated risk of death and ICU admission in patients hospitalised due to COVID-19 by sex, age and comorbidities, alongside estimates of rates of transitions between different levels of care and the expected length of stay in different states. We developed techniques for estimation of conditional length of stay, given the eventual patient pathway through levels of care in the hospital. Two reports based on the work of this group have been presented: “Relative risks of COVID-19 hospitalisation and mortality-in-hospital by long-term health conditions in the UK population” and “Risk of death and ICU admission, state transitions and length of stay in people hospitalised with COVID-19 in the UK: associations with sex, age, and comorbidities”. A further report and accompanying github repository are in preparation, alongside a paper describing the statistical methodological developments made.

13. Structured Expert Judgement

The Structured Expert Judgement Task Team conducted a structured expert elicitation of Primary School Headteachers to quantify contact patterns within schools in pre-COVID-19 times and how these patterns were expected to change upon re-opening. Additionally, we surveyed school Headteachers about risk mitigation strategies and their anticipated effectiveness. Our findings suggest that while DfE guidelines form the basis for risk mitigation generically, individual schools have adopted their own bespoke strategies, often going beyond the guidelines. A follow-up elicitation, prior to schools fully re-opening in September, could beneficially inform COVID-19 transmission modelling. The Task Team also constructed a COVID-19 infection hazard model for the return of pupils that takes into account uncertainties in model input parameters, derived from the elicitation. The model estimated likely number of primary schools with one or more infected persons under three different return-to-school scenarios: 1st June when schools re-opened with about one-third of pupils; mid-June full return of children in the same age cohorts; and return of all primary age children in September. The infection hazard model has been evaluated with different levels of community prevalence and inventories of children, teachers and support staff per school. The Team have prepared two full-length manuscripts reporting the work just outlined, which will soon appear on MedrXiv.

Future updates

We expect to give our next update on RAMP activities at the end of the summer.

Original call | Previous updates | Privacy Statement for the RAMP initiative | Accessibility Statement | Contact