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

expert reaction to ONS Infection survey data

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


Dr Jason Oke and Prof Carl Heneghan, Centre for Evidence Based Medicine, University of Oxford, said:

“The ONS report today that, based on their modelled estimates, the daily incidence rate has increased in recent weeks to 0.58 per 10,000 people but this appears at odds with their own non-modelled and “official reported estimates” of the rate of new COVID-19 infections.

“The modelled estimates, which are based on just 89 positive cases and restricted to the last 6 weeks of data, suggest a flat lining trend for August with an uptick going into September.  In contrast, the un-modelled data suggests a trend that has remained fairly stable since mid-June, albeit with some week-to-week variation.

“We believe that the modelled estimates should not be over-interpreted and the longer-term trend, based on the raw data, given prominence in future reports. In addition, to aid interpretation, the ONS should publish the cycle threshold for positive tests over time as this also gives an indication of the level of infectiousness as well as response rates for different age-groups.”


Prof James Naismith FRS FMedSci, Director of the Rosalind Franklin Institute, and Professor of Structural Biology, University of Oxford, said:

“The ONS survey is particularly informative but it stops at September 5th. In contrast to last week’s release, there is now evidence for the uptick in positive cases. Taken together with the real time data from yesterday’s track trace and this morning’s Imperial study, the virus is clearly spreading again. The incidence is still relatively low; we would do well to keep it there.

“There is no evidence that the virus has changed its behaviour. That the virus is more prevalent in younger people, is almost certain to always have been true. It is disheartening to see this presented as a novel observation, had we been testing in January we would have seen the same thing as now. I have seen arguments that were disproven by data in other countries rehashed. Younger people generally lead more sociable lives than older people and the majority of younger people have no symptoms so have no idea they are infectious. Brutally said, younger people are being asked to make a huge sacrifice in social life, personal life and economic prosperity to save the older population from covid19. I am very grateful at just how many of them have willingly done so, especially given that although no one is completely safe, for younger people covid19 is largely harmless.

“Many scientists have warned that covid19 was a marathon not a sprint. Winter is likely to be time of maximum danger and this is fast approaching. What happens next will be determined by a number of factors, some in our control and some not.

“We do not yet know how post-infection immunity fades (time and extent), and clearly some people are now being re-infected. The less complete the fade-out and the slower it happens, the better things will be.

“We do not yet know whether there is some T-cell cross-immunity that might protect against or lessen the disease and how many people will benefit from this, but it could be a significant positive.

“We know that medical treatment and scientific advances have improved significantly, thus even with infection rates as bad as March and April, there will be many fewer deaths.

“The disaster of the first wave resulted in many of the most vulnerable losing their lives in the most tragic circumstances. 

“The more people wash their hands and practise social distancing – especially by and around the vulnerable – the lower the number of deaths and illness we will see.

“We could benefit enormously from a system that isolates infectious people before they infect others as this will greatly slow the spread of the virus and thus save lives.  This must have rapid testing, instant follow up and effective isolation built into it: it is a three legged stool; miss one and its useless.

“Ultimately, the next few months are about buying time.  There will be at some point a vaccine and/or highly effective medicines, and keeping the virus as low as possible in the vulnerable population until then is the key to reducing the heartbreak.”


Dr Daniel Lawson, Lecturer in Statistical Science, School of Mathematics, University of Bristol said:

“The ONS survey is based on a random cross-section sample of the general public, which are less affected by sampling bias than the total test results. Sadly, these data tell a similar story: that infections are rising throughout the UK, and therefore without intervention, we will face a second outbreak.

“These data also show that the epidemic is currently growing in the younger age groups, including those who may be returning to university. Travel, moving home and reuniting friends again are likely to coincide to create opportunities for additional spread within this age group and ultimately to more vulnerable groups.

“It is worth mentioning that the survey is only unbiased if response rates are high, or unrelated to the output. The response rate is only 16% and response might be biased by age category and COVID-19 symptoms. So it is plausible the picture could evolve rapidly.”


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

“Given all the other data that we’ve heard about this week pointing in the same direction, it’s hardly surprising that ONS are saying that this week’s report on their infection survey is showing an increase in infections in England. But that’s important. The information coming on over the past week about increasing numbers of cases has been based on data from testing of people who have asked for a test, or been asked to take a test, primarily because they have symptoms. So those results could be affected by test availability, by exactly where the tests are done and on which people, and by any glitches in the testing system. Those issues don’t apply to the ONS infection survey – the people tested form a large and representative sample of the English community population, who are being tested just in order to estimate the infection rate across the country, and not because of any symptoms they have or anything of that kind. So if these results are showing increases in infections, we can be confident that they are real increases and not just some anomaly of the system.

“One important aspect is that the rise in numbers of new infections is not quite so sudden as the increase in the daily report of new UK cases last Sunday (6 September), which went up by over 1,000 – about 65% – on the previous day’s figure. But the ONS’s central estimate did rise by a fairly similar amount. The ONS estimate for the most recent week, 30 August to 5 September, is that there were 58 new infections each day for every million people in the English population. (That week does roughly correspond to the dates when tests would have been done for the UK figure for new cases for 6 September – 6 September is the date that the tests were reported, not when the swabs were actually taken.) That’s an estimated 3,200 new infections each day in England. The previous reporting week, 19-25 August, ONS estimated 36 new infections per million people per day, or 2,000 per day across England, so that’s a rise of about 60%.

“ONS are, rightly, rather guarded on reporting these results – there’s a lot of ‘suggests’, ‘may have increased’, and so on. That’s because all these figures are estimates based on a survey, so there’s an inevitable margin of error. For most of these figures, that margin is quite wide. So for the estimated new infections per day in the most recent reported week, the central estimate is 58 new infections per million people, but the data are compatible with a rate anywhere between 38 and 84 new infections per million (or 2,000 and 4,600 new infections per day across England). The rather wide margins of error arise because the overall infection rate in the population remains fairly low, despite the recent rise. So there aren’t many infections in the sample of people being tested, despite around 30,000 people being tested in the survey each week. In the latest reported fortnight (23 August to 5 September), for example, there were only 25 new infections in the sample. This problem of wide margins of error is more marked when we look at subgroups of the whole English population, by age or by region for example. So yes, the data do indicate that infection rates have been rising in the 17-24 and 25-34 age groups. But it’s not quite so clear that infection rates are stable or falling in older age groups (50 and over) – the data look that way, but there’s uncertainty. Similarly, though the estimates rates of infection in different English regions do differ quite a bit, those differences are generally within the margins of error. The survey is in the process of increasing the number of sampled households and people, and one aim of that is to reduce the margins of error on regional estimates.

“The survey now covers Wales and Northern Ireland as well as England. No results have yet been published for Northern Ireland because the survey has not yet been running long enough there. Some results for Wales are published, and, unlike the position in England, they are not showing evidence of a change in the infection rate there – but the sample size in Wales is relatively small (so far) so that the margin of error is wide.” 


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


Declared interests

None received.

in this section

filter RoundUps by year

search by tag