select search filters
roundups & rapid reactions
before the headlines
Fiona fox's blog

expert reaction to latest data from the ONS COVID-19 Infection Survey

The Office for National Statistics (ONS) have released the latest data from their COVID-19 Infection Survey.


Prof Paul Hunter, Professor in Medicine, The Norwich School of Medicine, University of East Anglia, said:

“The latest report from ONS provides further confirmation that infection rates really are falling in the UK.  In Northern Ireland however cases are still rising.  The fall has been most notable in younger age groups.  The percentage fall was not as great as we saw in case numbers but this was largely due to the fact that the ONS study reports on PCR positives and people who become positive can remain so for more than two weeks so there will always be a lag in decline identified by ONS and by clinical sampling.

“Whether this decline will continue or reach an equilibrium before schools return in September is not yet clear.  We may see numbers remaining around the level they are at now.”


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

“The latest ONS weekly release on their Covid-19 Infection Survey (CIS) is good news, for most of the UK at any rate.  Things don’t look good in Northern Ireland, however.  This week’s release takes the estimates on prevalence – the numbers of people in the community who would test positive on a swab test for a current infections with the virus that can cause Covid-19 – up to the week 25-31 July.  ONS estimate that the number testing positive decreased in England, Wales and Scotland during that week.  The CIS positivity rates were increasing in England and in Wales the previous week, though they were decreasing in Scotland.  So this is a turnround in the trends of infection in England and in Wales.

“You might ask why this is news.  The daily counts of confirmed new cases, as reported on the dashboard at and elsewhere, have been falling in England and in Wales since the middle of July, so for about three weeks now.  But there must always be some doubts about trends in those confirmed case figures.  That’s because they depend on the number of people turning up for routine Covid testing, and on the reasons why people choose to be tested or not.  If the numbers of tests change, or if people are making different decisions on whether to be tested, that could cause the numbers of confirmed cases to fall (or rise) without there being a real underlying change in the number of infected people out there.  The figures from the CIS come from a survey of a representative sample of the population, who are tested (regardless of symptoms) only for the purpose of tracking the pandemic.  So they aren’t subject to those biases that can affect the daily case figures.  One important reason for keeping an eye on the CIS estimates is that they are a vital check on the accuracy of the trends in new confirmed cases.

“However, a snag is that the CIS prevalence estimates count people who are positive for the virus, whether they are newly infected or have already been infected for a few days or even a couple of weeks.  This means that, when new infections start to fall, it takes a bit longer for prevalence estimates to fall, because they still include some of the people who first became infected when new infections were coming up to their peak.  In simple terms, what this means is that you wouldn’t expect the CIS prevalence estimates to start falling until two or three weeks after numbers of new infections began to fall – and that’s exactly what we’ve seen in England and Wales.  Total infections (measured by CIS prevalence) were still rising there in the week 18-24 July, though the rate of increase had slowed.  That’s what you’d expect if the numbers of new infections began to fall about the middle of the month – that decrease wouldn’t have led to a decrease in total infections so soon afterwards.  But in the latest week, 25-31 July, total numbers of infections have started to fall according to CIS, as you’d expect if the fall in confirmed cases is real.  The same pattern was seen earlier in Scotland – new confirmed cases started falling there a couple of weeks earlier than in England and Wales, at about the end of June, and the turnround in the trend in total infections in the CIS also happened a couple of weeks earlier than in England and in Wales.  What I take from this is that the falling trends in new confirmed cases on the dashboard, first in Scotland and later in England and Wales, aren’t just happening because of changes in who is turning up to be tested, and do reflect what’s actually going on in terms of infections.

“We shouldn’t get too excited about this, though, despite it being good news for three of the UK’s countries.  The numbers of people who would test positive in England and in Wales for the latest week are only just a little below the levels from two weeks before that (the week 11-17 July) and are still a very long way above the infections levels we saw in April this year.  ONS still estimate that about 1 in every 75 people in England would test positive in the latest week – in late April and early May it was fewer than 1 in 1,000.  In Wales the estimate is about 1 in 230 for the latest week, compared to fewer than 1 in 4,000 in early May.  Because the falling trend in Scotland has been going on longer, the current level of 1 in 120 is about the same as at the end of June, compared to fewer than 1 in 1,000 in early May.  And we don’t know how long these falls might continue.  Given that confirmed cases on the dashboard are still generally falling in England, Wales and Scotland, I’d expect the falling trends in the CID prevalence figures to continue there for a little longer, at least.  But, for instance, any increase in infections that happens as a consequence of the removal of restrictions in England on 19 July might not have shown up yet in the data, if it’s taking more than a few days for people to change their behaviour.  So we’ve still got to be careful and keep watching the data.

“That’s much more true in Northern Ireland, where infections are continuing to increase in the CIS prevalence estimates.  For the week 25-31 July, ONS estimate that 1 in 55 people would test positive, the highest rate of any of the UK countries, though there’s quite a lot of uncertainty about that (because of the small population size and the relatively small number of people tested for the survey) – the margin of error runs from 1 in 40 to 1 in 75.  That estimated prevalence is the highest that it’s been in CIS results since the survey began in Northern Ireland last September, apart from just one week in January when the estimate was slightly worse at 1 in 50.  New confirmed cases on the dashboard for Northern Ireland are showing some signs of falling, but the fall started later than in England and Wales, is proceeding much more slowly, and hasn’t shown up at all in the CIS estimates yet.  Let’s hope that those case rates do correspond to a real fall in infection there – but it can’t yet be confirmed by the survey results.  A plausible reason for the infection position being worse in Northern Ireland than in the rest of the UK is that the delta variant of the virus is now dominant there (as in the rest of the UK), but that dominance is coming much later than it did in the rest of the UK, so that Northern Ireland is only recently experiencing the serious effects of that variant that the rest of the UK saw earlier.

“Back in England, the CIS positivity estimates are also pretty encouraging for individual regions and age groups.  ONS estimate that the numbers who would test positive were decreasing, in the week ending 31 July, in five of the English regions (North West, East Midlands, West Midlands, London, and the South East), and while they say that trends are uncertain in the other four regions (North East, Yorkshire and The Humber, South West, East of England), no region has evidence of increasing rates.  On age groups, ONS consider that rates are decreasing in some of the age groups where concern about infections was highest – that is, in younger adults (school year 12, aged 16 or 17, to 49), as well as in the oldest age group that they use (70+).  They regard the trends as uncertain in the remaining age groups (age 2 up to school year 12, and 50-69), but again there’s no clear sign of an increasing trend in any of those groups.

“I should mention that ONS do also publish estimates of numbers of new infections – so-called incidence estimates – based on the CIS.  Since these are new infections, and don’t include people that have already been infected for some days or even a couple of weeks, they should be a better match (in terms of trends) to the daily confirmed case numbers, without being subject to the biases that could affect the daily case numbers.  The CIS incidence estimates would also count newly infected people whether or not they have symptoms.  But a snag in using those incidence figures to check trends in the confirmed cases is that the CIS incidence figures are not so up to date as the prevalence estimates, for technical reasons.  The latest incidence figures only go up to the week ending 10 July, so they can’t be used to check the declines in new confirmed cases in England and Wales (or in Northern Ireland), because those declines started after 10 July.  The CIS incidence estimate for Scotland for the week ending 10 July is, however, about 6% lower than the estimate for the previous week (ending 3 July), so it does broadly correspond to the decrease in confirmed cases in Scotland that began about the end of June.”


Prof Rowland Kao, the Sir Timothy O’Shea Professor of Veterinary Epidemiology and Data Science, University of Edinburgh, said:

“The broad, continued decline of positive tests in the ONS survey in every nation but Northern Ireland is consistent with the trends in reported cases which arise from the voluntary testing recorded in pillars one and two.  As the ONS survey is intended to be a random and therefore unbiased sample of the population, it is a more robust source of information than the case data, and so this decline in numbers (with appropriate statistical uncertainty) is reassuring.  However, such random surveys do not well reflect the current trends in infections as individuals can test positive for a considerable period after becoming infected.  This capture of lingering historical infections is particularly important in periods when cases are declining, as this is when older cases contribute a greater proportion of the survey than newer cases (the opposite is true when infections are increasing).  Thus trends are detected much more slowly by the ONS survey.

“There are some indications that case numbers reported from Pillar one and two testing are either stabilising, or possibly may be about to increase again.  Whether or not they do again, will depend on a set of transient, complex factors that have been influencing the various data streams (some up, some down).  These include school holidays, weather impacts, changes in restrictions, the end of the Euro 2020 tournament, the ‘pingdemic’ and possible shifts in human behaviour.  If those trends result in rising infections we shall detect them first in the case data, but only after in ONS survey results as they come through in the coming weeks.”



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.



in this section

filter RoundUps by year

search by tag