select search filters
roundups & rapid reactions
factsheets & briefing notes
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 there Covid-19 Infection Survey. 


Dr Julian Tang, Honorary Associate Professor in Respiratory Sciences and Clinical Virologist, University of Leicester, said:

“This ongoing increase in the number of COVID-19 cases is not surprising due to its recent exponential growth.  What this means is that if there are more cases in the population at any one time, where R>1: these will seed even more new cases into the next generation of cases – and the growth will continue.

“Without reducing the contact rate between the infecteds and susceptibles (which is still most of us), the virus will continue to spread.

“If people think the existing control measures are not working (but we actually don’t know if the number of cases might be even higher now, without them) then a decision needs to be made to impose stronger restrictions sooner rather than later (when the virus will be more difficult to control) – or that we somehow learn to live with the virus and just be prepared to deal with the impact of this.

“This is not an easy decision and unfortunately there is no decision that will be acceptable to everyone – as we have been learning so far in this pandemic.”


Prof Sheila Bird, Formerly Programme Leader, MRC Biostatistics Unit, University of Cambridge, said:

“Round 5 of REACT’s swab-test surveillance of persons aged 5+ years in England is now complete and shows a very sharp increase in prevalence between rounds 4 and 5, see Table. 

Comparison between REACT-1’s person-based and ONS Infection Survey’s household-based swab-test surveillance is intriguing.

“REACT (round 5) estimates national prevalence at 0.60% (95% confidence interval: 0.55% to 0.71%) which means that, on any one day, 450,000 (410,000 to 530,000) are infected with SARS-CoV-2 {assuming 75% sensitivity that a nose and throat swab picks up the virus}.  This prevalence translates {indirectly} into 45,000 new infections per day, by assuming that the virus remains detectable for 10 days on average.

“By contrast, ONS Infection Survey estimates half as many prevalent infections in households: that  224,400 people (95% credible interval: 203,800 to 245,700) within the community population in England had the coronavirus (COVID-19) during the most recent week, from 25 September to 1 October 2020, equating to around 1 in 240 people (95% credible interval: 1 in 270 to 1 in 220).

“The ONS infection Survey estimates incidence directly.  During its most recent week (25 September to 1 October 2020), there were around 17,200 new cases per day (95% credible interval: 13,800 to 22,900).  Again, substantially lower, here by a factor of three, than deduced by round 5 of REACT-1.

“Sampling frames differ (address file for households; NHS list of patients registered with a general practitioner) and volunteer-rates differ between the two surveillance surveys; and there is home-visiting by ONS Infection Survey’s field-force and re-imbursement of its participants.

“The ONS Infection Survey should be able to tell us more about within-household transmission than is currently reported.  More please!  On average, there are two persons per ONS surveyed household and it appears that 28% of infected households have more than one infected person (532/416 = 1.28).

“Re-weighting is necessary because REACT’s response rate, for example, is around one-third of those invited to take part. Notice that re-weighting increases the estimated prevalence so that those more likely to be positive are less likely to take part.

“Oddly, re-weighting appears to affect ONS Infection Survey’s estimated prevalence less despite their volunteer-rate being substantially lower (in part because the Address file is not current).

“Finally, what about the swab-positive rate for those whom test and Trace quarantines?  After nearly 20 weeks, Test and Trace has failed to report anything about the swab-positive rate for those whom it quarantines, namely: a) members of the household of symptomatic index case; b) external close contacts of symptomatic index case.  The Royal Statistical Society’s (RSS) recommendations on how to glean intelligence from Test & Trace were designed to help.

“But REACT could help too – by asking its participants an extra question to find out if they are in quarantine at behest of Test & Trace when they take their swab-test.

“ONS Infection Survey could help too – by deploying its field-force to conduct the RSS-recommended random home-visits to offer swab-testing to Test & Trace households.  Based on REACT, this extra stratum of visits would be expected to yield at least 10 to 20 times as many positives per 1000 visits as ONS Infection Survey currently reports, and hence be hugely cost-efficient!”


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

“The virus has not changed.

“The ONS survey is grim reading and in conjunction with the REACT-5 study alarming.  The two surveys report rather different absolute numbers, there are methodological differences between them, but the message is crystal clear the virus has taken off.  Case numbers are rapidly increasing in the North of England, really only in the South East and East have numbers so far remained stable.  The infections in the teen to 24 year group is spiralling.  It is depressing and somewhat shaming to see how disadvantage and deprivation are such markers for infection.

“Scientists look at data, these data show as a country that we have failed to use the summer breathing space to defeat covid19.  The numbers in Europe and USA are showing similar trends, we are not alone in this failure.  Covid19 is an awful problem.  I cannot forget the pain suffered by so many families in March and worry about its repetition.

“I realise many will want to spend time on identifying the who, why and what of this failure; but the data indicate we do not have time for this.

“There are dire warnings that hospitals in the North of England will hit bed capacity.  It is hard to over state the risk that brings.  There has been a revolution in science and medicine, provided people can be looked after in hospital, the death toll will be much lower.  If however, the hospital system overloads, then triage will operate, meaning choices will be who gets treated and who does not.  Untreated severe Covid-19 is a significant killer.  Overloading hospitals will have spill over into many other illnesses where a drop in care will also lead in time to deaths and disability.  This overloading is the thing that I most fear.

“We have to remember that young people are mostly asymptomatic and even as the infection builds in this population, we wont see big rises in disease provided it stays in this population.  It can never be stressed enough that young people are not immune, a small number of them will die and others will have life changing illness.

“However, the ONS survey which samples the community, shows that infections in the over 70’s are increasing, thankfully much more slowly than the teen to 24 year old.

“On a more hopeful note, ONS indicates that the over 70s also showed a slower increase than groups aged 25 to 69.  This REACT study would lead to a more pessimistic conclusion about the rate of difference in the over 70s relative to 25 to 69.  Once again, irrespective of the absolute numbers, both studies agree that the virus is spreading in the over 70s again.

“What does science tell us about the various options ahead?  I stress that only politicians can chose between the options.

“If the virus continues to spread as rapidly as it has in September, we will see thousands of deaths, maybe tens of thousands with many more people suffering lasting illness.

“In the absence of vaccine, the only way to stop covid19 spread is to disrupt transmission.  This means stopping infectious people infecting uninfected people.

“Hand washing, masks and social distances do reduce transmission, but on their own have not halted covid19.

“We therefore need to prevent infectious people meeting uninfected people.

“Track trace and isolate as means to prevent infectious people spreading the disease has failed.  The Scottish system which claims a better performance than the English one, has likewise not prevented sharp increase in cases.  It is therefore unlikely that we can increase the performance enough in time to make any real difference now.

“We are really only left with brute force restrictions on people’s lives.  These bring serious damage to the economy, social life and personal life of the entire population.

“In theory, an effective shield the “vulnerable” policy could reduce harm while allowing the less vulnerable to resume normal life.  However, such a policy needs urgent answers to several key questions including who defines the “vulnerable”, how long will shielding last, what is the quality of life for the shielded, how effectively can such shielding be implemented, how many of the less vulnerable will die if covid19 let rip?  There are also very serious concerns as to whether such a policy is even feasible, even if desirable.  The data on deaths show that neither Europe or the US have managed as of today to shield the vulnerable.  Carrying out a shield the vulnerable experiment with no more than a hope we can pull it off in the face of evidence of failure,  risks many thousands of deaths.  For the families that would be affected were such the experiment to fail, I doubt there will be much consolation in the victims mainly being over 70.

“Local ‘lockdowns’ in the North of England have not obviously worked so far.  Greater Glasgow NHS has continued to see rises in cases, despite stricter measures.

“We could wait another week, more data will clarify the picture, the newer restrictions may have begun so slow the spread.  This is not unreasonable but we have to be aware it is human nature to postpone difficult decisions.

“If current measures have failed, then an extra week will have made things much worse due to the exponential growth of the virus.

“Implementing stricter control measures today essentially banning movement except for work, will as before, bring the viral spread under control for a period.  The cost of these measures will be high financially, socially, in physical and mental health terms.  This needs experts in social science, economics, psychology and mental health to quantity these.

“Bringing in them a week early, if they were found to be unnecessary, represents a real burden for no gain.

“It is only human nature to seek a different lower risk / lower cost choice.  I believe that no such “simple” option exists.  What lies before our politicians is the need to reduce overall harm, by balancing the harm brought about by restrictions and by covid19.

“A vaccine will come, so will much better medicines, we are not doomed to endless cycles of lockdown.”


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

There isn’t much in the way of positive news, for England at least, in the latest results from the ONS Infection Survey.  They cover the week from 25 September to 1 October.  In last week’s ONS infection survey data release, though numbers of infections were still estimated to be increasing, there were some indications that perhaps the rate of increasing was slowing.  But those indications aren’t there any more.  The ONS survey produces estimates of two different rates – the prevalence rate, which is the percentage of people in the English community population that would test positive if they were tested for a current infection with the virus, and the incidence rate, which is the number of people in the population who would be newly infected each day.  And, for the most recent week, both of those rates show substantial increases in the previous week.

“Like the REACT-1 survey that also published a new report today, the ONS survey is based on swab tests on a large representative sample of the English population.  Both the ONS survey and REACT-1 have large sample sizes – REACT-1 tested almost 175,000 people in its latest round that ran from 18 September to 5 October, and the ONS survey tested over 127,000 people in its most recent fortnight (18 September to 1 October).  It’s a strength of the ONS survey, in comparison with the REACT-1 survey, that ONS can directly estimate the incidence rate of new cases.  That’s because people are tested more than once in the ONS survey.  But the main strength of both studies is that they test people simply to estimate the prevalence of infections.  They are not affected by changes in the availability of tests to the general public, or extra testing to help control outbreaks.

“The increases in the prevalence and incidence rates in the latest week for the ONS survey, compared to the previous week, are substantial.  The prevalence rate almost doubled, from 0.21% in the week ending 14 September to 0.41% in the week ending 1 October.  Putting it another way, in the latest week 1 in 240 people in the population would test positive, but the week before it was 1 in 470.  There’s statistical uncertainty about these results, of course, since they are based on a survey, and ONS estimate that the latest week’s data is compatible with anywhere between 1 in 220 and 1 in 270 people testing positive, but that’s a substantial increase however you look at it. On incidence (new infections), the estimate for the week ending 1 October is 316 new infections each day for every million people in the population (with a credible interval, showing the uncertainty, from 253 to 419).  That’s slightly more than double the figure for the week ending 24 October, when the central estimate of the rate was 154 new infections per million people per day.  The figure for the latest week corresponds to 17,200 new infections in the English community population each day (with an interval, showing the uncertainty, from 13,800 to 22,900).

“As other data sources are showing, the differences between regions in the rates of testing positive are pretty clear in the new ONS data, with much higher rates in the Northern regions (North East, North West, Yorkshire and the Humber) than in the rest of England.  The rates of infection have been rising much faster in those regions that elsewhere.  Infection rates across the country are highest in the age group from school year 12 (age about 17) to age 24, and again that is the group in which infections have been growing fastest.  That’s, of course, the age group that includes most university students, and we’ve been hearing a lot about high infections rates amongst university students, particularly in the North – but there’s nothing in the ONS data to indicate how infection rates in students compare with those of other people of the same age.

“We now have two different new sets of survey estimates, REACT-1 and the ONS survey, covering overlapping periods.  How do the results compare?  The main quantity that’s estimated in both is the prevalence of infection, and the figures from the two surveys do differ.  The ONS survey estimates that 0.41% of the English community population would test positive, in the week from 25 September to 1 October.  That’s 1 in 240 of the population, or 224,400 people, testing positive on any one day. REACT-1 estimates that 0.60% of the English population would test positive, based on their data collected between 18 September and 5 October, and they report that that corresponds to 450,000 people testing positive on any one day . So the REACT-1 figures are higher.  There are several possible reasons for this, as follows.

“First, the time periods covered are different, and in particular the REACT-1 results are based on a period that ends several days after the latest ONS results.  When infections are increasing, as both studies indicate, that could be a reason for a higher estimate in REACT-1.

“Second, all these estimates come from surveys, and are subject to inevitable statistical uncertainty.  The credible interval for the ONS rate of positive tests runs from 0.37% to 0.45%, and the credible interval for the REACT-1 estimate runs from 0.55% to 0.71%.  If the true rate for the ONS time period is actually towards the higher end of its interval, and the REACT-1 one towards the lower end, the difference isn’t so great as it looked at first (but it’s still there).

“Third, the two surveys are considering slightly different populations, and that’s reflected in the lists from which they draw their samples of participants.  The ONS survey covers the English ‘community’ population, so it excludes people who live in communal premises, such as care homes, prisons, or university halls of residence.  It draws its participants from a comprehensive list of residential addresses. REACT-1 intends to cover the whole English population, and draws its participants from lists of people registered with GP practices.  The proportion of the total population that lives in communal premises is not large – well under 2% at the most recent Census – but maybe this could lead to higher estimates in REACT-1 than in the ONS survey, and maybe that could be particularly true when of there are a lot of infections amongst university students, though here I’m speculating because neither report gives specific data on this point.

“Fourth, and importantly, the difference between the estimates of the numbers of infected people in the population appear to differ more than the estimates of the percentages.  For ONS, the central estimate is 224,400, with an interval from 203,800 to 245,700.  For REACT-1 it is 450,000, with an interval from 410,000 to 530,000 – so roughly twice as big.  One important reason for this has nothing to do with differences in the percentages – it’s that the two surveys are estimating different things with these numbers.  ONS estimate the number of people that would test positive, if everyone on the English community population were tested.  REACT-1 estimate the number of people that are infected in the English population.  When the rate of positive swab tests gets up to the sorts of levels we are now seeing, that difference in definitions begins to matter.  The reason is that there can be substantial numbers of false negatives – that is, people who are actually infected but have a negative test result.  The nature of swab tests is that the rate of false negatives, amongst people who are actually infected, can be reasonably high, and is difficult to estimate.  REACT-1 estimate it at 25%, though other research has tended to put it lower than that.  The ONS view is that, since there aren’t particularly good estimates of the false negative rate, it’s better not to make this kind of adjustment.  In fact, when the rate of positive swab tests is rather lower than it appears to be now, work done by ONS and by their academic partners on the infection survey indicates that these false negatives might, very roughly, cancel out with false positives (people who are really not infected but have a positive test result anyway), so that the number of people who test positive might be reasonably close to the number of infected people.  I think that’s not so likely to be true when positive test rates are running at the levels they have now reached.  This does all add a bit more uncertainty.

“I wouldn’t, though, want to make too much of the differences in detail between the results of the two surveys.  They both continue to provide valuable information about the path of the pandemic.  There will be another report next week from the ONS survey, taking the data up to and beyond the end of the latest REACT-1 round.  We can then see whether the slowing up in the rate of increase, that REACT-1 report, also appears in the ONS data.

“The ONS survey also presents data on infection rates in Wales and in Northern Ireland.  (The survey is also running in Scotland but there isn’t yet enough data from there to publish.)  In both Wales and Northern Ireland, the sample sizes are smaller, so that changes are more difficult to detect.  In Wales, ONS report that there’s some evidence that a recent increase in rates of positive tests is levelling off.  There aren’t enough results to be certain about this, but there’s nothing that looks like the rise in infections seen in England.  It’s really too early to say much about the position in Northern Ireland, because the statistical uncertainty is great, but the ONS survey is again not showing a huge rise in infections.  Comparisons with REACT-1 cannot be made in these other nations, because REACT-1 is not running there.”


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


Declared interests

Prof Sheila Bird: “SMB is a member of the Royal Statistical Society’s COVID-19 Taskforce which, on 23 July 2020, made recommendations on statistical methods for gleaning intelligence about the effectiveness of Test & Trace.”

Prof Kevin McConway: “I am a Trustee of the SMC and a member of the Advisory Committee, but my quote above is in my capacity as a professional statistician.”

None others received.

in this section

filter RoundUps by year

search by tag