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

expert reaction to new ONS data on prevalence of ongoing symptoms following COVID-19 infection in the UK: 1 April 2021

The Office for National Statistics (ONS) has released new data on the estimates of the prevalence of self-reported “long COVID”, and the duration of ongoing symptoms following confirmed coronavirus infection using COVID-19 Infection Survey data to the 6 March.


Dr Elaine Maxwell, Clinical Adviser at the National Institute for Health Research (NIHR), said:

“This is entirely consistent with the NIHR themed reviews of international literature re prevalence rates – more common in working age women and causing difficulties in activities of daily living.

“Key findings for me are not the prevalence at 12 weeks but the 70,000 at one year – a new finding internationally and the relapsing nature confounding cross sectional surveys which goes a long way to explaining the wide range of estimates at 12 weeks.  In the ONS report they say that if people reporting resolution of symptoms at one follow up only are measured, the rate at 12 weeks is 0.9%, but if they are followed up twice (so report symptoms resolved at the first follow-up but have reappeared at the 2nd follow-up) then prevalence at 12 weeks is 13.7%.

“Regarding the self reporting nature of this, given community testing was suspended for most of last year we can’t just ignore a million people so self reporting is the only option.  The fact that ONS used a control group and found an 8-fold difference in findings shows there is something happening even if the precise prevalence is unclear.

“As we at NIHR have repeatedly said, there are a number of different clusters of symptoms that may have different pathogenesis (and ONS acknowledge this in the release).  Measuring prevalence in an undifferentiated group with different phenotypes is more of a problem than self reported symptoms.”


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

“I think today’s ONS bulletin and data release about long COVID is a very important contribution to our understanding of what’s going on.

“The headline results are striking.  Over a four-week period up to 6 March, well over a million people in the UK are estimated to have had self-reported long COVID, which ONS define as symptoms persisting more than four weeks after the first suspected COVID-19 episode that aren’t explained by something else.  About three in every five of these people reported that the symptoms were adversely affecting their day-to-day activities – and of course many of the others would still be experiencing pain and distress even if their day-to-day activities weren’t affected.  The time period of four weeks is in accord with the clinical guideline from NICE for managing long-term effects of COVID-191, but what I find particularly alarming is the estimates numbers who have had long-lasting symptoms for much longer periods.  ONS estimate that 697,000 people reported that their symptoms had been persisting for at least 12 weeks, and again about three in five of them said that the symptoms were adversely affecting their day-to-day activities.  And the estimate for the number of people still experiencing symptoms a whole year after their original disease is 70,000.  That’s a pretty large number, considering that these people would have to have originally had COVID-19 on or before 6 March last year, which was before the first wave really took a firm hold.

“ONS report some quite marked differences between different population groups in the prevalence of these long-lasting symptoms.  The rates are highest in people aged 35-69, females, those living in the most deprived areas, those working in health and social care, and those with pre-existing health conditions.  But those are all groups where infection rates were relatively high, and the report can’t tell us whether the high rates of long COVID were because of differences in infection risk, differences in the chance of getting long-lasting symptoms if a person is infected, or some combination of the two.  A lot has been reported previously about the increased risk of infection, hospitalisation and death from COVID-19 in several minority ethnic groups – comparisons between ethnic groups on the prevalence of long COVID could not be made in this bulletin because the statistical uncertainty is too great.

“There are actually two different broad types of data in the report.  The first type is based on people’s replies to the questions about ongoing symptoms that can arise from COVID-19, and it includes responses from everyone, regardless of whether they ever actually had a positive test (in the survey, or elsewhere).  That makes sense, because some people will have had COVID-19 some time before they became part of the survey sample, and perhaps before testing became relatively easy to get in the UK.  A great number of people during the first wave would never have been tested during their initial infection.  Also, many people infected with the virus that causes COVID-19 don’t show any symptoms at all during their initial infection, so may well have had no reason to ask for a test – and not having symptoms during the initial infection doesn’t necessarily mean that symptoms will not occur later.

“But all COVID-19 symptoms can also be caused by other things that the virus, and there has sometimes been speculation that some people reporting that they have long COVID actually have symptoms with some entirely different cause.  The second type of data in the ONS report gives a measure of this possibility.  The researchers took a group of about 20,000 people who were known to have tested positive for the virus, within the ONS survey itself.  They compared what this group reported about long-lasting symptoms with what was reported by a control group of 20,000 survey participants, matched to the people who had tested positive for age, sex, and when they entered the survey, but who were unlikely to have been infected.  (They hadn’t tested positive for an active infection or antibodies to a previous infection either in the survey itself or outside it, and they reported no direct contact with an infected person.)  Not everyone in the survey would have fitted into one or other of these two groups – because neither of them included people who had tested positive for the virus outside the ONS survey, or who had never tested positive but had reported direct contact with an infected person.  But using the two matched groups allows a comparison on long-term symptoms that’s not affected by including people whose infection status could be unclear.

“The differences in the rates of reporting long-lasting symptoms in the group known to be infected and the group unlikely to have been infected are huge.  Yes, some people in the probably uninfected control group reported long-lasting symptoms, but only about 1 in 35 of them at 5 weeks after the equivalent date to the date their matched infected person had been infected, and 1 in 60 at 12 weeks after.  But the rates were very, very much higher in people who were known to have been infected – 1 in 5 reporting ongoing symptoms 5 weeks after when they were infected, and nearly 1 in 7 with ongoing symptoms after 12 weeks.

“That’s a rather strong indication that an actual infection with the virus that causes COVID-19 is behind most of the cases where symptoms persist.  The ONS researchers are very careful to point out that this is observational data from a survey, and does not on its own prove that the virus is the cause.  That’s largely because there are other differences between the infected group and the control group than the age and sex which were used for matching.  But the results certainly do point in the direction of an infection with the virus playing a major role in these long-lasting symptoms.

“ONS have reported on long COVID before, but this new release takes things much further.  The previous reports used data from the form of the infection survey that didn’t specifically ask about long COVID (because it was designed before long COVID emerged as a major issue).  Now ONS have added questions on long COVID to the survey questionnaire, and they have also improved and extended their methods for analysing the data.  They do point out that this new release contains statistics that they label as ‘experimental’, and doubtless they will make refinements and improvements in future releases.  But in my view, what they have done already is appropriate and impressive.

Further information

“It’s worth pointing out that the analysis, comparing people known to be infected with controls who probably weren’t infected, is the only part of the ONS report that looked at individual symptoms, and it also used a different definition of having symptoms to the other analysis, of self-reported long-lasting COVID.  That other analysis, of self-reported long COVID, isn’t based on which specific symptoms people said they had had.  But the analysis comparing infected people with controls counted people as still symptomatic only if they reported having one of a list of 12 symptoms (abdominal pain, cough, diarrhoea, fatigue, fever, headache, loss of taste, loss of smell, muscle pain, nausea/vomiting, shortness of breath, sore throat).  This doesn’t include all the symptoms that are asked about on the infection survey questionnaire.  There are 21 symptoms on that list, which also includes loss of appetite, chest pain, palpitations, worry/anxiety, vertigo/dizziness, low mood, trouble sleeping, memory loss or confusion, difficulty concentrating.  Those are all on the list of commonly reported symptoms in the NICE guidance (along with several others).  Rather little is said about individual symptoms in the ONS bulletin, and that would appear to be because the data on individual symptoms isn’t good, because many people don’t report exactly what symptoms they had.  But that means that the results can’t, for example, tell us anything about the ‘brain fog’ that is often mentioned in relation to long COVID.

“The data comes from the ONS Infection Survey, and the way that survey is carried out gives it major advantages over some other sources of information about the condition.  The people who are included in the survey are a pretty representative, random, sample of the community population of the UK.  So there will be fewer biases compared to other ways of finding people to ask about their symptoms – being in this survey sample doesn’t depend on whether you have symptoms, or where you live, or what job you do, for instance.  (People living in communal establishments like care homes or prisons aren’t included, but they make up only a small percentage of the total population.)  And in this survey, unlike other large surveys, people are tested for the virus and asked about their symptoms on repeated occasions (initially weekly, then monthly).  That allows ONS to put together symptom reports from the same person over time, which improves the precision of the estimates.

“This isn’t the only publication today from these ONS researchers.  With academic researchers, they have today published in the BMJ an important report2 on the post-discharge experience of people hospitalised with COVID-19.”





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