March 1, 2021

expert reaction to results from the COVID-19 Schools Infection Survey Round 2

The Office for National Statistics (ONS), London School of Hygiene & Tropical Medicine (LSHTM), and Public Health England (PHE) have released the second round of results for their COVID-19 Schools Infection Survey.

This Roundup accompanied an SMC Briefing.

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

“With the reopening of schools in England planned for a week today (8 March), any data on how infections might occur in schools is bound to be welcome. So it’s good that this new report on a long-term project run by ONS, LSHTM and PHE has appeared. These data have some important strengths – in particular the results on current infection levels were collected from pupils (and teachers) who were sampled regardless of whether they currently had symptoms. That’s important, because other sources indicate that school-age children who are infected are less likely to show symptoms than are adults. However, probably the most important feature that it would be good to know about the virus in schools is how and when transmission occurs, and the design of this study doesn’t allow any clear conclusions to be drawn about that. Also, there are many limitations and caveats that apply to the study – I’ll describe some of them below, but there are more of them discussed in the bulletin.

“Many of the possible comparisons that might be made from the results are not very clear, in statistical terms. That’s because the percentages of people (pupils and staff) who would test positive for current infection, or who would test positive for antibodies, are estimated from survey samples, and for various reasons the numbers of people tested were not very large. So there’s rather a lot of statistical uncertainty, and many of the differences that were observed are within the range that might happen just because of this statistical uncertainty – so they might or might not be true differences. The rates of current infection were lower (in both survey rounds) in primary pupils and staff than in secondary pupils and staff – but the difference was not large, and could possibly have been explained just by statistical uncertainty from the sampling. Infection rates in the December round were rather higher than in the November round, but again this difference was small enough to be explainable by sampling variability. There were small differences in the levels of antibody positivity in school staff, between the two rounds and between primary and secondary schools, but again these could be explained by sampling variability (Antibody positivity measures for pupils in round 2 are not (yet) available.) Given that infection rates in the country as a whole were quite high in the period between the two survey rounds, I’d expect that school staff would have been infected  over that time, and so that antibody positivity rates would have gone up in school staff. The survey data are compatible with that, but they can’t confirm it, because of the wide statistical uncertainty.

“One set of comparisons that I do find interesting is that the researchers compared current infection levels, and the staff antibody levels, found in this survey with the levels in the general population of the local authorities concerned, as measured in the ONS Infection Survey. In every case, infection and antibody levels as measured in this schools survey were lower than the general population figures from the ONS Infection Survey. In some cases the difference was rather small – that’s true of antibody levels in teachers. But in both pupils and teacher on current infection levels they were larger, though still not really large enough to be statistically meaningful. The comparisons were done on a pretty reasonable basis – comparing the pupils in this survey with children of the same age group in the ONS infection survey and so on – though adjustments could not be made to allow for systematic differences between the age and gender of school staff and the general population. But the fact that they all went in the same direction may be saying something about the existence of biases in the estimates from this survey, from the ONS infection survey, or both. One can’t tell directly from these results, but given the many limitations in this schools survey, my suspicion is that this is another warning to be careful in interpreting the results from the schools survey.

“My general feeling is that the survey results on the measures taken in schools to prevent Covid transmission are more important and probably more reliable than the results on current infections and antibody levels, though the researchers are not in a position to link the measures to actual infection levels.

“The study was designed to follow up patterns of current infection and antibody levels over time in a sample of schools. That’s a useful thing to do. However, only two rounds of sampling have so far occurred (in early to mid November, and in early December, last year), and that’s not really enough so far to see trends over time. There was to be a third round in January but it was cancelled because schools are closed. We are promised more results and deeper analyses after future rounds.

“So what are the limitations? Here are some.

• “The survey results cannot be used to estimate infection levels nationally. That’s because the sampling of schools was deliberately not done in a uniform way. Instead, when the local authorities where the survey was to be done were chosen in the early autumn, more local authorities where infection rates were high at the time were chosen, presumably to allow for more accurate surveillance there. This makes it pretty well impossible to use the results to provide national estimates – and also, the areas that had high or low prevalence back in September won’t necessarily have high or low prevalence as the study goes on. Ten of the local authorities chosen were amongst the top 20% of prevalence in September; by the time of round 2, none of them were still in the top 20%. The other five local authorities were in the bottom 8% of prevalence in September; by the time of round 2, four of them still had low prevalence, but one (Barking and Dagenham) has the sixth highest prevalence of all the English local authorities.
• “The survey couldn’t include pupils or students who were absent from school on the days the tests were done, either because of Covid-19 or for any other reason. This could well bias the estimates of infection levels in school downwards.
• “Response rates amongst pupils were low (14% in round 1, 16% in round 2). Perhaps the pupils whose parents agreed for them to be tested were systematically different from the others. (Response rates amongst staff were much higher – over 40%.)
• “It could have been interesting to compare antibody positivity levels in pupils and teachers (though the antibody levels for pupils in round 2 are not available yet). But this could be misleading, because the tests used were different (blood test for staff, saliva test for pupils) and would have different error rates.”

Prof Russell Viner, Professor of Adolescent Health, UCL, said:

“These are very useful data from an overall high quality study. There were low response rates amongst students, which limits validity, however this is common for large surveys during COVID-19.

“In essence the study suggests that school levels of infection in late 2020 were similar to community levels for both adults and children. This suggests that schools are not amplifiers of infection, as some have feared, but that school infections largely reflect community virus levels. Unfortunately the study as set up cannot be directly informative about transmission between children and from children to staff.

“The study also found that the great majority of schools studied are doing well with coronavirus control measures.

“Both these main findings are reassuring as we prepare to reopen schools in the UK.”

Dr Julian Tang, Honorary Associate Professor/Clinical Virologist, University of Leicester, said:

“It is important to get children back to school but there is a risk of enhanced viral transmission with this.

“Fundamentally, from a virological viewpoint, all the other seasonal respiratory viruses demonstrate higher viral loads that shed for longer in young children – more so with younger children – and this is the same for SARS-COV-2:

https://www.jpeds.com/article/S0022-3476(20)31023-4/fulltext

https://jamanetwork.com/journals/jamapediatrics/fullarticle/2770150

https://www.ncbi.nlm.nih.gov/pmc/articles/PMC7510743/

and there are reports showing transmission amongst teachers/children in schools:

https://www.cdc.gov/mmwr/volumes/70/wr/mm7008e4.htm

https://www.eurosurveillance.org/content/10.2807/1560-7917.ES.2020.25.29.2001352

where masks may help to control this:

https://www.eurosurveillance.org/content/10.2807/1560-7917.ES.2020.25.36.2001587

“The fact that these are not higher than the background community rate of transmission does not mean that it is not happening in schools (it is always hard to tell exactly where anyone caught the virus) – rather that the transmission rates in/out of schools in the community and homes have all reached an equilibrium – this is quite normal in compartments where movement of infected hosts/viruses is free and unimpeded between these compartments.

“This does not mean that virus transmission in schools is not happening – but it has reached similar levels in schools, community and homes.

“One way of highlighting the differential transmission rates between the school-community-home environments may be to look at the emergence of the various new virus variants in these compartments – where the variant has not yet reached a uniform distribution across these different populations – to see if there is any enhanced spread of the virus in these different compartments – as was done in the UK and South Africa to examine the enhanced transmissibility of the Kent (B.1.1.7) and South African (B.1.351) variants, respectively, against the background rate of spread of the original Wuhan virus.

“I’m sure some of the modelling teams are looking at this – and I very much look forward to what they find as all schools reopen after 8 March. With almost 10 million children and staff mixing in schools again, it will be revealing and instructive.”

https://www.ons.gov.uk/releases/covid19schoolsinfectionsurveyround2englanddecember2020 

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

www.sciencemediacentre.org/tag/covid-19

Declared interests

Prof 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.”

Prof Viner: “Member of SAGE and President of RCPCH.”

None others received.