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, estimating numbers of people currently testing positive

The Office for National Statistics (ONS) have released new data from their COVID-19 Infection Survey which estimate the number of people who have tested positive for having the disease. 


Prof Sylvia Richardson, MRC Biostatistics Unit, University of Cambridge, on behalf of the Royal Statistical Society Covid-19 Task Force, said:

“The publication of the pilot data by ONS is welcome as timely surveillance and estimation of prevalence of infection at regular points in time is a crucial component of understanding and controlling the epidemic in the UK.

“We are told that between 27th April and May 10th, out of 10,705 participants from private households tested, 33 individuals had swab-positives tests in 30 households, giving an overall positivity rate to the test for this period of 0.27% with 95% confidence limits between 0.15% and 0.41%.  The timeliness of this study was made possible by building it on previous ONS household surveys.

“While the information published will provide a useful reference point, we would like to draw a note of caution: small number of positives and the fact that response rate was not reported.  Increased sample size may be a recommendation from the pilot-phase, assuming that the participation-rate is sufficiently high.

“It would have been useful if, in addition, the fraction of the 33 swab-positives individuals who reported symptoms was given, as the extent of asymptomatic or mildly symptomatic infection is a particular dilemma in COVID-19.

“Finally, we note that in view of the self-administration of the swab test, the sensitivity is likely to be diminished and so the positivity rate could underestimate the prevalence of infected individuals in the sample.

“A future release will report the results of antibody tests which aim to tell us something different, namely the proportion of people who have been exposed to the virus since the beginning of the epidemic and developed an immune response.  Alongside these data releases, trustworthy information on the selection of the sample and the performance and quality of tests are needed in order to be able to interpret fully.”


Dr Tom Wingfield, Senior Clinical Lecturer and Honorary Consultant Physician, Liverpool School of Tropical Medicine, said:

“The preliminary findings of the ONS coronavirus (COVID-19) infection pilot survey estimate that, at any given time between 27 April and 10 May 2020, an average of 0.27% of England’s population, or 1 in 370 people, had COVID-19.  The survey also suggests that people who self-reported working in health or social care were more likely to test positive for Covid-19 infection at any given time (1.33%) than those who reported working in other sectors (0.22%).  The ONS survey does not include people with COVID-19 in hospitals, care homes or other institutional settings.

“There are a number of interesting points to note from these results.

“First, this survey tells us the numbers of participants who were positive for COVID-19 at the time of swabbing (the positivity rate).  The survey does not tell us the number of participants who have had COVID-19 (the prevalence).  Once there is an approved, effective antibody test, this could be used in similar surveys to help us to establish COVID-19 prevalence.

“Second, the ONS survey used self-swabbing and self-reporting of occupation and symptoms.  Swabbing the nose and throat can be uncomfortable and tricky, especially when self-testing, and it is not clear what the quality of these self-swabs was.  In addition, there is not yet enough data to help us to understand how confident we can be that a negative swab means that a person does not currently have COVID-19.  False-negative rates – when a swab is negative in a person who actually has COVID-19 infection – could be influenced by the quality of the swab, laboratory procedures, when the swab was taken within the course of COVID-19 illness, and other factors.

“Third, the accuracy of the ONS survey in estimating the numbers of people in England actually infected with COVID-19 at any time will improve as it is rolled out and more data accrued.  Although the ONS and research partners applied “weighting” to ensure the participants in the survey were representative of the population in terms of age, sex, region, and household ownership and size, there may still be differences that have not been accounted for, especially in such an early phase of the pilot.  It should also be noted that the estimates refer only to England and not yet to Scotland, Wales, and Northern Ireland.

“Fourth, the results of the ONS survey seem to be markedly lower than many of the estimates of numbers of the UK population infected given in various modelling studies.  However, it is worth noting that most of the modelling studies have given estimates of the cumulative numbers of people in the UK predicted to have had COVID-19 (the prevalence) whereas the ONS survey estimates those who had evidence of current COVID-19 infection at any given time (the positivity rate).  The ONS data may support modellers to further refine the parameters that they use to build their models and make predictions about COVID-19 infection rates, prevalence, and transmission.

“Fifth, in terms of groups at risk of COVID-19 infection, the survey suggests that health and social care workers are at higher risk than those in other occupations in England and, as has been shown in household contact tracing and other cohort studies from China, all age groups – from 2 years old up – seem to be similarly susceptible to catching the virus.  This is in contrast to disease severity, which shows a clear association that older people, especially those over 60, are more likely to have severe COVID-19 illness.  Although this data is still preliminary and further work is needed, the suggestion from the ONS survey that children seem similarly likely to have the virus (though not the disease) as adults is particularly important as the UK government considers when and how it may lift school closures – though this data can’t tell us anything about how likely people of different ages are to transmit the virus.”


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

“The ONS Infection Survey invited participation from households, one member of which had previously taken part in ONS-surveys and indicated their willingness to be re-contacted.  This prior willingness enabled the ONS Infection Survey to go into the field so quickly.  However, this index householder had to give consent to his/her participation in the ONS Infection Survey for the survey team to come calling.  When they did, other members of the household were invited to take part in addition to the index-member.  Surprisingly, the initial publication leaves us guessing about response rates.  The plan for the pilot had been that 10,000 households would give access to 25,000 individuals.  Results are reported for 10,705 individuals but we are not told how many households they belong to; nor are we told how many prior-willing-respondent, when approached about this survey, confirmed their willingness to take part.

“We are told that the 33 swab-positives belonged to 30 households.  Crucially and significantly, the swab-positive-rate was about 6 times higher at 1.3% (95% CI: 0.39% to 3.3%) for participants in patient-facing healthcare roles or resident-facing social care roles than for other householders at 0.22% (95% CI: 0.13% to 0.35%).  I estimate that the former were around 4.5% of the initial respondents, around 500.

“No differences were reported by the age-group of respondent but 95% confidence intervals were wide so that, to the nearest 1000, I think that respondents were around 2000, 4000, 3000 and 1000 in the four analysed age-groups (2-19 years; 20-49 years; 50-69 years; and 70+ years of age).  As the weeks go by, ONS’s Infection Survey will be increasingly informative with results on response-rates (post-pilot) and from the succinct questionnaire that accompanies each swab-test.”


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