select search filters
briefings
roundups & rapid reactions
Fiona fox's blog
The government have released regional data for the COVID-19 R value in England.
Prof Keith Neal, Emeritus Professor of the Epidemiology of Infectious Diseases, University of Nottingham, said:
“The publication of regional R values are not easy to understand particularly as the South West which has been least affected has the highest figure. What is unclear from the methods described is how many cases acquired in care home and hospitals are being included in the models used to estimate R. These cases have very little, if any, relevance to the R in the wider community which is important to the man or woman out and about. Cases in places of care need to be managed with infection control and not social distancing.”
Prof Daniela De Angelis, Professor of Statistical Science for Health and Deputy Director and Programme Leader of the MRC Biostatistics Unit, University of Cambridge, said:
What do these numbers show and what’s the significance of the ranges?
“There are a number of modelling teams, including PHE/Cambridge, LSHTM, Warwick, Imperial, Manchester and Bristol that contribute estimates for the current value for R, the reproduction number, together with a range that quantifies their uncertainty in this estimate. The values are submitted to SPI-M, a SAGE subgroup for epidemic modellers, who form a principled consensus in the form or a range of possible values. These are submitted to SAGE for approval and have, this week, been published.”
Is there anything surprising in these numbers?
“Not particularly. It is still highly likely that R is below 1 in each region. However, as the level of infection becomes smaller, R will naturally gravitate towards 1 as localised outbreaks have greater significance in estimating R and the lockdown rules are gradually relaxed and/or adherence becomes less.”
Is there anything significant in the small differences in the ranges, and what’s the reason for the differences?
“There is no real evidence for significant difference between the regions. The South West is the highest and this may be because of the localised outbreak in Western Super-Mare, or simply because this region has seen the lowest levels of infection in the country. The R value for London also appears to be increasing.”
What’s the message (if any) to the public/councils etc in the lower vs the higher areas?
“It is important to keep monitoring the situation, but as yet there is no compelling evidence for any of the regions to be considered exceptional.”
Is there any trend here; and will the next batch be more interesting when they’re released?
“The important indicator of the threat level due to the pandemic is the combined contribution of incidence (how big is the pandemic currently) and the value of R (is it growing bigger). As incidence declines, as mentioned above, R will gravitate naturally towards 1 and we may find more evidence of this in the next round of analyses. It may be that we get used to seeing these estimated intervals straddling 1 in future. This is not necessarily a cause for alarm or concern.”
Prof Rowland Kao, Professor of Veterinary Epidemiology and Data Science, University of Edinburgh, said:
“As the number of cases decline, it is only to be expected that the relative fluctuations in the R number will be greater, since even a small outbreak could cause a temporary rise. Longer term trends would be needed to indicate whether any observed change is important.
“The fact that all these numbers include ranges close to one suggest that now, as the effect of the first measures easing lockdown are being seen, some slight increases in transmission may be occurring. This emphasizes the continued importance of not moving too quickly through changes in lockdown, allowing contact tracing and testing to work in preventing small outbreaks from becoming large ones.
“Importantly, information on test and trace should be available allow epidemiologists to evaluate its efficacy. This includes not just overall numbers traced but how those tracings are distributed across positive cases (e.g. is it roughly the same number of tracings for each infected, or is there evidence that some individuals trigger many more contacts than others?) and how they are distributed across those regions.”
Dr Yuliya Kyrychko, Reader in Mathematics at the University of Sussex, said:
“Because the R number is notoriously difficult to estimate, and the ranges for all regions in England are so wide and so close to 1, this means that one has to be very careful with using these numbers to plan the strategy for easing the lockdown further. Such a close proximity of those values to 1 does not automatically mean that we are facing the next wave imminently, but it may indicate that there are some specific regions or social settings where infection is still prevalent. This highlights even further the need for a robust and efficient track-and-trace and testing programs to make sure any further infection is quickly contained to prevent a large second wave.”
Dr Konstantin Blyuss, Reader in Mathematics at the University of Sussex, said:
“The fact that R numbers look very much the same across regions may suggest that they rather represent aggregates across very large geographic areas and populations. The difficulty with this is that it may represent an average for a region but mask what is possibly a significant local variation in the level of infection, thus making it more difficult to efficiently plan local intervention strategies.
“Significant differences in social structure, age distribution, public transport infrastructure etc. result in different levels of mixing between individuals in the population, and therefore, more local estimates of R would be much more insightful from the perspective of developing and optimising strategies for disease control.”
Prof Matt Keeling, Professor of Populations and Disease at the University of Warwick, said:
“All the ranges are similar and overlapping, so we cannot say that any one region is worse than any other region.
“All the ranges are closer to the critical threshold of R=1 than we would ideally like to see – which means that the epidemic is declining relatively slowly. This also means we haven’t got much wiggle room for additional relaxation of social distancing measures.
“As the number of cases becomes smaller in many regions, these predictions will become more uncertain and more biased by small localised outbreaks.
“These are not a measure of risk, they are about the decline of the epidemic. To understand risk you need to look at both the incidence of infection and individual behaviour.
“There is still some uncertainty in these values, which are influenced by the three interacting outbreaks in hospitals, in care homes and in the community. Having input from multiple groups using a variety of methods and data streams helps to overcome these issues.”
All our previous output on this subject can be seen at this weblink:
www.sciencemediacentre.org/tag/covid-19
Declared interests
Prof Keeling: ‘I’m a member of SPI-M and therefore one of the people that generates one of the R values that make up this consensus.’
None others received.