MIL-OSI United Kingdom: expert reaction to preprint from round 12 of the REACT-1 study on COVID-19 prevalence in England


Source: United Kingdom – Executive Government & Departments

A preprint, an unpublished non-peer reviewed study, reports on the latest data from the REACT-1 study on COVID-19 spread across England.

This Roundup accompanied an SMC Briefing.

Prof James Naismith, Director of the Rosalind Franklin Institute, and University of Oxford, said:

“The React survey shows a 50 % increase in prevalence between roughly the last two weeks of April and a month later (roughly last two weeks of May).  The study estimates a doubling time of 11 days with a credible interval with as short of 7 and as long as 23 days.  As of today, we are most likely a further cycle of doubling further down the road, although the summer will help blunt its growth.  This study therefore matches other evidence that we are in a third period of increasing infection.  The delta variant virus has largely replaced the alpha variant in the UK, a remarkable display of the power of evolution.  It is the delta variant that has led to the UK having one of the highest number of cases per million and substantially higher than the US and EU and our case numbers are increasing whilst US and EU states are mostly falling.  Had the delta variant not arrived in the UK as early as it has done, we would have been able to end restrictions with almost no risk as originally planned.  It must be understood that without vaccines, the delta variant would have been a disaster for the UK.  This is because rapid increase cases will in the absence of vaccine fill hospitals requiring shut society down or health care collapse.  The other measures that the UK has put in place (track and trace, isolation, etc.) have once again proven ineffective at preventing exponential growth.  Thus far lockdown and vaccines are the only things that work at stopping exponential growth.

“The survey shows that the delta variant is spreading most rapidly in the younger unvaccinated cohort.  The second vaccine dose is once again shown to be highly effective at reducing infection and hospitalisation, this is most clearly seen in the elder (mainly vaccinated) cohort.  The case for increasing the coverage of the double vaccinated as rapidly as possible is clear.  I am confident that when and if we reach 85 %  coverage of double vaccination of the population capable of onward spread, then the UK will indeed be able to live with the virus.

“It is troubling to see that Britons identifying as having Black and Asian ethnicity appear to more likely to be infected.  This is a persistent disparity that has not been solved.  As the report notes, those from the most deprived section of society, are more likely to be infected.  I strongly support efforts to reach out to those communities to enable them to be vaccinated.”

Prof Kevin McConway, Emeritus Professor of Applied Statistics, The Open University, said:

“This latest REACT-1 preprint covers their Round 12, which involved swabs taken between 20 May and 7 June.  Several of the findings on this round do match what we’ve seen before in recent days and weeks – that’s particularly true about the overall levels of infections.  But there are other important data too, including some that give us increased hope about the effectiveness of the vaccines against hospitalisations and deaths.  And that’s good news.

“In broad terms, the results on the latest round look like what we’ve already seen from other sources.  Both the ONS Covid-19 Infection Survey (CIS) and the daily data on new confirmed cases on the dashboard at also show increases in infection rates in England, both comparing the time of REACT-1 Round 11 (15 April to 3 May) with the time of Round 12, and during the period of Round 12.  All three sources also show, broadly, infections falling in England from the huge peak at the start of this year, roughly until early to mid May (with a few wobbles) by which time the level was very low, but then increasing after that.  Some of the details are different.  That’s particularly true in comparing the daily dashboard figures with REACT-1 and the CIS.  But that’s maybe not surprising – the two surveys (REACT-1 and CIS) estimate the percentage of people who would test positive for the virus, regardless of whether they have just been infected or got their infection some days before, while the dashboard figures count only new cases for whom the infection has just been detected.  Also, the dashboard figures come from routine virus testing, and can potentially be biased if the types and numbers of people turning up for testing change over time.  REACT-1 and CIS test people only for the purpose of estimating infection levels, so aren’t affected by any biases of that sort.

“What’s remarkable is how close the REACT-1 results for this round (and the round before) are to the CIS results for England.  REACT-1 estimates that, on average during their latest round, 0.15% (or 15 in every 10,000) of the English community population would test positive for the virus.  That’s about 1 in 670.  Because the estimate comes from a survey, there’s some statistical uncertainty, and the number could plausibly be between about 1 in 560 and 1 in 830.  The latest official estimate from CIS for England, for the week from 30 May to 5 June, was that about 1 in 560 would test positive, with a margin of error from 1 in 480 to 1 in 680.  That’s a rather higher positivity rate than the REACT-1 estimate for the whole of the latest round, but it covers a slightly different time period.  The CIS also give daily ‘modelled’ estimates, and if I average them for 20 May to 5 June (roughly the same period as this REACT-1 round; it can’t be identical because CIS has not yet given estimates for 6 and 7 June), the average comes out again at about 1 in 670, with roughly the same margin of error as REACT-1.

“These rates are still very low compared to the peaks at in January, or in Spring of 2020.  The current REACT-1 estimate is roughly the same as it was for their round 4, back in late August 2020.  The latest CIS estimate is roughly the same as for early September.  But the concern now, which was also the concern at the start of September last year, is that the rates of infection are increasing – and if they increase exponentially, they could soon reach very high levels.  The REACT-1 researchers estimate that positivity increased with a doubling time of about 11 days, during their round 12.  The CIS results show a slightly longer doubling time than that, though their official rate for England did roughly double in the last two weeks of May.

“You might wonder why there are two different survey-based estimated for England, if they are so similar.  One reason is that they can be checked against one another.  But another reason is that, as well as the overall estimates for the whole country, they provide extra information on different things.  Both provide estimates for different subgroups of the population, for example by age and by the region where people live.  These don’t coincide exactly, but you wouldn’t expect them to.  With infection rates at the current low levels, the survey results don’t provide high levels of accuracy on subgroups of the population.  (For example, in REACT-1 round 12, there were only 135 positive swabs for the whole country out of nearly 109,000 swabs taken.  That’s challenging enough, but in estimating the positivity for the North East region, for instance, their calculations have to be based on only 6 positive swabs out of a total number of just over 5,000.)  So there are some differences in details between the two surveys, but again there is a lot of similarity, with (for instance) the highest regional rate of positivity in the North West and the lowest in the South West, and with the highest positivity rates in children and young people under 24.  REACT-1 does carry out some more detailed geographical analyses, and reports pockets of infection in some areas of the North West, as we’ve seen in the dashboard data for the past few weeks.

“But, perhaps most interesting and important, the REACT-1 questionnaire, that participants complete, asks people whether they have been vaccinated, and this, together with some other analyses, is used to give a broad picture of how vaccination might be changing things.  Those results are pretty encouraging.  The rate of testing positive increased between round 11 and round 12 in most age groups, but the increase was (on average) considerably greater in those aged up to 49 than in older people, who (at these times) would be much more likely to have been vaccinated, often with two doses.  And the REACT-1 researchers compared their estimates of the numbers testing positive with figures from the dashboard on daily hospitalisations and deaths, over all rounds of the survey going back to May last year.  Comparing deaths with the infection levels about four weeks earlier, and using appropriate scales, the two sets of numbers tracked one another fairly closely last year.  But since the peak at the start of this year, infections have reduced hugely but deaths have reduced even more quickly.  That’s true of hospitalisations too (with a time lag of about three weeks rather than four), though the difference is quite a lot smaller.  Then the researchers did the same, but separately for those aged up to 64 and those aged 65+.  On deaths, during the decline from the January peak, for both age groups the numbers of deaths fell faster than the infections, compared to last year, but the difference was considerably greater and clearer for the 65+ group than for the younger people.  The same age difference is true of hospitalisations – the rate of decrease in the under 65 group, compared to the decline in infections, is pretty much the same as last year, but hospitalisations have gone down rather faster in the 65+ group.

“Of course, there’s no way of proving from these data that that difference is definitely caused by the fact that many more of the older group would have received vaccines than of the younger group.  There are other differences too, including improvements in treatment.  And I can’t claim (and nor do the researchers) that the patterns are always really clear.  But this does at least give me considerable hope that the vaccines are working, to some extent anyway, to break the link between infections on one hand, and serious illness and death on the other.  That again fits in with other detailed studies of the effectiveness of vaccines.  There’s still quite a lot of uncertainty about the exact level of protection, but there’s certainly protection.  If that’s the case, we don’t have to be as concerned with increasing cases as we were the last time they were at this kind of level, last September.  We do still have to be careful and keep looking at the data, because there’s still a lot of uncertainty about some of the details.  But there’s more hope.

Further information

“As usual, the REACT-1 results come from swabs taken from a representative sample of the community population of England aged 5+, which are tested for the virus.  REACT-1 have changed the details of the way they choose the sample this time, but they explain how their reweighting procedure (that adjusts for imbalances in the sample compared to the population) means that the results are still comparable with those from previous rounds.  CIS also uses a representative sample of the community population in England (and in the other three UK countries, though I have not discussed data for the other countries in this comment).  But it samples from children aged as young as 2, not 5, and chooses the samples in a different way using a different list of people.  The fact that we have two different sample surveys to estimate SARS-CoV-2 infections in England is admirable, because any survey can on occasion come up with ‘rogue’ results, and with two surveys we can check them against each other.  Most countries don’t even have one such survey.  (But I can’t deny that two surveys cost a lot more than one.)”

Preprint (not a paper): ‘REACT-1 round 12 report: resurgence of SARS-CoV-2 infections in England associated with increased frequency of the Delta variant’ by Steven Riley et al. was posted online at 00:01 UK time on Thursday 17 June 2021. This work is not peer-reviewed.

All our previous output on this subject can be seen at this weblink:

Declared interests

Prof Kevin McConway: “I am a Trustee of the SMC and a member of its Advisory Committee.  I am also a member of the Public Data Advisory Group, which provides expert advice to the Cabinet Office on aspects of public understanding of data during the pandemic.  My quote above is in my capacity as an independent professional statistician.”

None others received.

MIL OSI United Kingdom