select search filters
briefings
roundups & rapid reactions
Fiona fox's blog
The Office for National Statistics (ONS), London School of Hygiene & Tropical Medicine (LSHTM), and Public Health England (PHE) have released the sixth round of results for their COVID-19 Schools Infection Survey, providing initial estimates of staff and pupils testing positive for COVID-19.
Prof Kevin McConway, Emeritus Professor of Applied Statistics, The Open University, said:
“These results from the Covid-19 Schools Infection Survey are, as always with this survey, really quite hard to interpret – or perhaps I should say that they could be very easy to over-interpret. This survey, run jointly by ONS, the London School of Hygiene and Tropical Medicine, and Public Health England, shouldn’t be confused with the ONS Covid-19 Infection Survey (CIS). The CIS covers the whole of the UK with a sample of participants that is representative of the whole community population (aged 2 and over, or 16 and over for the CIS antibody results). The schools survey runs in a limited set of local authorities in England – originally 15, but many of the results are based on data from only 11 of them. These authorities were never supposed to form a representative sample for the whole of England. Instead areas with high infection levels back at the beginning of the 2020-21 school year, last September, had more than their representative share of the number of participants. Since then, infection levels have varied quite a bit between local areas. So, as the ONS bulletin clearly warns, the results from the schools survey simply can’t be used to provide estimates of the position across the whole of England.
“There are other limitations and considerations too. Only pupils and staff who were present in the school building on the day when testing took place have contributed data. People absent for any reason, including absences because they were ill with Covid-19 or self-isolating, aren’t taken into account. Rules to prevent the disclosure of data that could identify individuals mean that some results for some groups cannot be provided at all. The antibody tests used in the schools survey are different from those used in several other studies and surveys, in that they detect only antibodies from natural infections and not antibodies arising from vaccination (not that many of the pupils would have been vaccinated). There are reasons for all these aspects, but they do mean that the detailed results could potentially mislead, if one isn’t careful.
“As an example of why we’ve got to be careful with the results, the data file that comes with the ONS release presents two different sets of figures for the percentage of pupils in primary schools in Rounds 5 and 6 of the survey, which took place May (round 5) and in the second half of June and the first week of July (round 6) this year. One set covers data from 14 of the original 15 local authorities, and on the face of it appears to show a fall in infection rates in primary pupils between round 5 and round 6. In Round 5, the central estimate is that about 7 in every thousand primary pupils was infected, while in round 6 it was about 3 in every thousand. However, there is a great deal of statistical uncertainty around those numbers, partly because the sample size is not immense and partly because the prevalence of infection is really very low. The Round 5 data are compatible with between about 3 per thousand and about 13 per thousand being infected, and the Round 6 data with between less than 1 per thousand and 8 per thousand being infected. So we can’t even be sure that the proportion infected did really fall between round 5 and round 6 – for example, it could plausibly have been about 5 per thousand in both rounds. Anyway, these average infection rates apply only to the 14 local authorities involved, and not by any means the whole of England. That’s made clear by looking at the other set of figures, which covers only 11 of the original 15 local authorities, which are those that provided data for all the rounds so far (rounds 1, 2, 4, 5 and 6 – round 3 did not take place because schools were closed in the January to March lockdown this year). From those figures for 11 authorities, the central estimate for round 5 is that about 4 in every thousand primary pupils (more precisely, 3.6 in every thousand) were infected, and for round 6 it is about 3 in every thousand (more precisely, 3.4 in every thousand). Again the margins of statistical error are really wide – from about 1 in every thousand to 10 in every thousand for both rounds. It is these figures that are behind the statement in the ONS bulletin that “There was little change in the percentage of primary school pupils testing positive in Round 6 compared with Round 5.” That’s a valid thing to say, but it applies only to pupils from the 11 local authorities involved. The fact that the central estimates for the 11 local authorities are quite a lot different from those for the 14 local authorities in the other set of figures – in round 5, 7 in every thousand pupils for the 14 authorities and 4 in every thousand for the 11 authorities – does show that it really makes little sense to try to apply these results to the whole country. The trends over time for the individual local authorities are reasonably firmly based, but the level of statistical uncertainty for each of them is really pretty large.
“I could make similar points about most of the other results. It’s good, though not surprising, to see that school staff were changing from negative to positive for antibodies at a slower rate between rounds 5 (May) and 6 (June/July) compared to between rounds 4 (March) and 5. Not surprising, because the antibody test detects only antibodies from previous infections, not from vaccination, and many more of the staff were vaccinated at the later dates and so would be less likely to become newly infected. Also, in statistical terms we can’t even really be sure that there was a decrease in the rate of change at all, even in the 11 local authorities that provided the data, because the statistical margins of error around the estimates of the rate of change on antibody levels are very wide and are compatible with there having been no change at all.
“It’s interesting to see again that parents of both primary and secondary pupils would mostly be willing at least to consider having their children vaccinated, if they were offered a Covid-19 vaccine. The numbers saying yes to these questions have changed very little between the latest release (covering late May to late June this year) and the previous one (covering mid-April to late May), and the change that did occur is well within the margins of statistical error. But we’ve got to remember again that these figures are from just 14 local authorities (in this case), and that vaccines are still not available for most ages of pupils, even now. Would parents think differently if a vaccine were actually available for their children? We can’t say from these data.
“As well as the main release on round 6 of the schools survey, ONS have also just released data on antibodies in pupils from round 4, using tests carried out in the second half of March. That was quite a long time ago now, and the interpretation issues that I’ve already described apply to these results too. The huge range of different antibody prevalence estimates across the included local authorities, from about 1 in 20 secondary pupils being positive for antibodies in some local authorities to about 1 in 4 in others, indicates that it probably wouldn’t make sense to look just at averages for the whole of England even if we did have representative data for the whole country, which we don’t. And, as ONS rightly point out, the margins of statistical variability on these estimates for individual local authorities are very wide. So I’ll say no more about these pupil antibody results.”
Prof Mark Woolhouse, Professor of Infectious Disease Epidemiology, University of Edinburgh, said:
“There is now a wealth of evidence from around the world that schools are not the main driver of COVID-19 epidemics.
“Although outbreaks do occur in schools these tend to reflect levels of infection in the wider community. Outbreaks in schools in the UK involve staff at least as often as students.
“However, the epidemiology of COVID-19 in the UK has changed in recent months. First, the great majority of adults have now been vaccinated, so the relative contribution of school-aged children to transmission will have increased. Second, the delta variant is significantly more transmissible than its predecessors in all age groups, meaning that the absolute contribution of school-aged children to transmission will have increased.
“For these reasons close monitoring of COVID-19 in schools continues to be important, so the publication of the ONS’s latest Schools Infection Survey is extremely welcome.
“One reassuring result from the survey is that – despite the advent of the delta variant – infection rates during the summer term of 2021 were actually lower than during the autumn term of 2020. It is possible that this is at least partly due to mitigation measures implemented in schools, including routine testing.
“The report does not raise any immediate concerns about the re-opening of schools after the summer holidays.
“A key question for the planned inquiry into COVID-19 in the UK is whether there was ever any need to close schools at all. The epidemiological evidence to date suggests the answer may be no.”
Round 6: June 2021 https://www.ons.gov.uk/releases/covid19schoolsinfectionsurveyround6englandjune2021
Round 4: March 2021 https://www.ons.gov.uk/releases/covid19schoolsinfectionsurveyround4pupilantibodiesenglandmarch2021
All our previous output on this subject can be seen at this weblink:
www.sciencemediacentre.org/tag/covid-19
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.