select search filters
briefings
roundups & rapid reactions
before the headlines
Fiona fox's blog
The Office for National Statistics (ONS) have released the latest data from their COVID-19 Infection Survey.
Prof Kevin McConway, Emeritus Professor of Applied Statistics, The Open University, said:
“The latest weekly data release from the ONS Infection Survey takes the data up to last week (the week from 28 March to Easter Saturday, 3 April). Very broadly speaking, across the UK the trend in the rate of testing positive for the virus that can cause Covid-19 is similar to what it has been for a few weeks now – in all of the four UK countries, infection rates (by this measure) are very, very much less than they were in the peak in early January, but over the past three or four weeks the fall in rates has roughly levelled off and the rate of testing positive hasn’t changed much in either direction. There have been small changes over that period in the four UK countries, but these changes have been in different directions in different countries and in different weeks. This roughly static position doesn’t concern me much. I’d be happier if rates were continuing to fall. But there’s no clear sign that they are going up generally, as they could have done after the reopening of schools and the loosening of some of the lockdown restrictions. In three of the UK countries, the infection level seems now to be roughly where it was around the middle of September last year, and it’s a bit lower than that in Wales. Those levels are still quite a lot higher than they were in July and early August last year – they are now about six times as high in England as they were in July, so we still do have to be careful. But one reason why I’m not very concerned is that vaccination has much reduced the chance that someone will get very seriously ill or die, if they become infected. Vaccination doesn’t take the risk away completely, though, and anyway not all the population is vaccinated, and those are more reasons for continuing to be careful.
“For the most recent week, ONS estimate that the rate of positive tests has been roughly level in England over the past two weeks, with between 1 in 300 and 1 in 380 people testing positive in the latest week. (That’s a range that allows for the inevitable statistical variability in estimating a rate from a sample of swabs.) In Wales, ONS estimate that the rate has fallen over those two weeks, with between 1 in 470 and 1 in 1,630 testing positive in the most recent week – a wide range of uncertainty because the number of people swabbed in Wales (and indeed in Northern Ireland and in Scotland) is considerably smaller than in England because of the smaller populations. In Northern Ireland, ONS estimate that the rate has been level, but the statistical uncertainty is large, with the range going from 1 in 180 to 1 in 580. In Scotland, ONS report that the rate of testing positive continues to decrease, as it did the previous week, and between 1 in 280 and 1 in 640 people would test positive in the most recent week. Because the ranges of uncertainty overlap quite a lot, it’s difficult to be certain from all this how the countries compare. The infection rate in Wales does pretty definitely seem to be lower than in the other three countries, and the Scottish rate might well be lower than in England and Northern Ireland, but there isn’t enough statistical precision to be sure of that.
“In the English regions, ONS begin by pointing out that there are wide margins of error for the estimates of the positivity rate, because the number of people swabbed in a single region is much less than for the whole country. As has been the position for some time now, rates are higher in the northern and midland regions than in the south of the country. Despite this, every week ONS report that a few regions have ‘early signs of’ a particular trend, though very often such early signs don’t continue in subsequent weeks. This week, ONS are more definite that the rate of positive tests is increasing in the North West, which, with Yorkshire and the Humber, is estimated to have the highest positivity rate of all the English regions. ONS also say that infections could perhaps be beginning to increase in the South East, and to decrease in the South West and the North East, but I don’t think there is enough certainty about these trends to pay much attention to them, unless they continue into future weeks.
“This is the third week in which ONS have published estimates of incidence – that is, of the rate of new infections – using their new estimation method. The other figures, that I’ve described up to now in this comment, are for prevalence – that is, the rate of testing positive for the virus regardless of whether the infection is new or the person has already been infected for some time. Again the picture is mixed. The incidence estimates go up only to the week ending 27 March (because the relevant data take longer to collect and process than for the prevalence data). ONS report that the rate of new infections increased in England over the two weeks ending on 27 March, and that the rate probably fell in Scotland over that period, though that’s not certain because the statistical margin of error is wide. The trends aren’t certain in Wales or Northern Ireland. Overall, allowing for the considerable margins of error, the rates of new infections are pretty similar in England, Wales, and Northern Ireland, with between 27 and 29 new infections per 100,000 people per day, but the rate of new infections in Wales is estimated to be less than half that, about 11 new infections per 100,000 people per day (but the margin of error on that Welsh estimate goes from about 2 to about 22 per 100,000 people per day – incidence is not easy to estimate).
Further information
“These aren’t the only data on the progress of the pandemic to have come out this week. Yesterday the REACT-1 study from Imperial College published its own estimates of the rate of testing positive in England. These estimates, like the ONS estimates, are based on a survey in which a reasonably representative sample of the community population (not including people living in communal institutions like care homes or prisons) provides swab samples which are tested for the virus, and the fact that they are swabbed only to estimate infection rates should mean that they aren’t affected by some of the biases that can occur in other data sources on cases and infections. The REACT-1 and ONS surveys are, however, carried out and processed in rather different ways. Their estimates therefore don’t usually coincide exactly. Though they usually at least agree on general trends, there can be differences in what they say about shorter term trends or on some of the details. This week’s REACT-1 preprint covers round 10 of their survey, which ran from 11 to 30 March. The headline estimate from REACT-1 was that 0.20% of the English community population would have tested positive, that is, 1 in 500 people, but the statistical margin of error means that the number could plausibly be between about 1 in 430 and about 1 in 590. It’s not really appropriate to compare that with the very latest ONS estimate, which is for the week 28 March to 3 April and doesn’t overlap much with the REACT-1 round. So instead I’ve compared it with the average of the ONS daily modelled estimates for the period of the REACT-1 round, 11-30 March. That ONS figure is a 0.29% positivity rate, or about 1 in 350, with a margin of error from about 1 in 315 to about 1 in 390. That’s clearly a positivity rate that’s quite a lot higher than the REACT-1 estimate for the same period, and it would take considerably more analysis that I have time for to tie down the reason for that difference. However, the two surveys do generally agree on the trend in the positivity rate over that period – both say that it was roughly static.
“Where REACT-1 and the ONS survey do differ a lot, though, is on the estimates of positivity rates for the English regions. Given that the overall REACT-1 positivity estimate is about two-thirds of the ONS positivity estimate for the same period of time, you might expect that the REACT-1 estimates for the regions would also be roughly two-thirds of the ONS estimates. But in fact the REACT-1 estimates differ a lot relative to the ONS ones. In the South East, the REACT-1 estimate is just a little over one third of the ONS estimate for the same period, and in the North East, the REACT-1 estimate is 10% higher than the ONS estimate. I believe these differences stem from the detailed way in which the two surveys draw their samples and process their results. But they do draw attention to the fact that, despite the very large number of people swabbed for both surveys, it’s really not easy to get precise estimates for regions (and particularly so, I think, for the North East region, which has the smallest population of all the UK regions).
“Another regular source of data on the incidence of infection is the ZOE Covid Symptom Study, which bases its results on reports from volunteers who have chosen to use the ZOE app to report their state of health, as well as on results of tests on some of those people. This week ZOE are reporting quite large falls in prevalence, to about 1 in 1,400 for the UK as a whole, which is a lot less than the latest ONS estimates. That’s the ZOE figure for this week – for the previous week, which corresponds to the most recent ONS data, the ZOE prevalence estimate for the UK as a whole was about 1 in 1,000, which is still considerably lower than the ONS estimate. The difference is partly because ONS are estimating the prevalence of all infections, while the ZOE study is estimating the prevalence of symptomatic infections. But some of the difference is bound also to be caused by the very different data sources and estimation methods that they use, compared to the survey methods of the ONS (and REACT-1). The ZOE method needs to take account of what kinds of people have volunteered to use the app, and if that changes over time, biases may be introduced.
“There are also other sources of data on new infections or new cases, that are also published regularly. One is the daily counts of new confirmed cases on the dashboard at coronavirus.data.gov.uk. The patterns of new cases on the dashboard do differ somewhat between the UK countries, but broadly they did all show a flattening, in early March, of the decline from the peak of cases early in the year, followed by an increase in the rate of decline in the last week or two of March. The ONS incidence data do show that flattening off, but not yet very clear signs of the reduction speeding up again, except in Scotland. That may be because the ONS data go up only to 27 March and don’t cover the most recent weeks, and it may also be because that is picked up as a confirmed case in the dashboard data does depend on who is being tested, and patterns of testing for those data do change over time. The ZOE symptom study also estimates incidence, with its latest estimate for the UK being about 1,900 new cases per day. The ONS estimate for its latest week is of a total of about 17,300 new infections a day. Again, ZOE are estimating only symptomatic infections, while ONS are estimating all infections whether or not the person has symptoms – and the ONS figures are for two weeks ago whereas the ZOE figures are for this week. However, ZOE’s estimate for new symptomatic cases across the UK two weeks ago was about 4,800 a day, and the difference between that and the ONS figure is just too large to be explained by the fact that ZOE don’t include cases where there are no symptoms, I’d say. A problem, though, is that the ZOE estimates depend on who is using the app to report their symptoms, and if that changes fast for some reason, it could affect the ZOE estimates in a rather unpredictable way.”
Prof James Naismith, Director of the Rosalind Franklin Institute, and University of Oxford, said:
“Today’s numbers shows that as of last week, in England around 3 people in every 1000 are infected with covid19. In Scotland, a previous upward blip to 4 in 1000, has fallen to just over 2. Wales has just over 1 person in a 1000. The number in England has held roughly steady and that the upward blip in Scotland has fallen back is very encouraging.
“There is some evidence that opening of schools saw a small rise in children but this appears to levelled off at around 5 in 1000. Most hopefully, the prevalence in the most vulnerable age groups is 2 in 1000 for 50 to 69, with 1 in 1000 for 70+. It would seem likely that the vaccine campaign is responsible the significantly lower prevalence in these age groups relative to younger cohorts. As more people complete their vaccinations and more ages are vaccinated, incidence will fall further.
“Whilst this is very encouraging and suggests that the re-opening can continue as planned, there are grounds to remain careful for a bit longer.
“The virus is still present, for example 4 in 1000 people aged 35 to 49 are currently infected. Without social distancing this would increase and although we would not see anything like the same proportion of deaths per infection, a surge in this age group (because of its size) would place severe stress on the NHS. A surge of infection in one age group also gives more chance of new mutants emerging. The less infections the smaller chance for mutants.
“Vaccines are working exactly as predicted, this is science at work. I hope the public are convinced of the value in investing in basic science even if sometimes it seems “blue skies”.
“The UK lockdown has saved many lives. The new variants are particularly hard to control and the decision to space out the first and second jab to maximise coverage has been vindicated. The purchasing and roll out of the UK vaccines has saved lives.
“What is true about circulating virus in the UK is true globally. We should do everything we can to support vaccination in developing countries. “Stopping travel” only slows the disease spread for a period, as we can see with the Kent strain it is getting everywhere. Allowing the virus to rage unchecked in a human population is to gamble on the emergence of new strains. No matter what we do such new strains would reach the UK. We are not safe until everyone is safe.
“I believe we should continue our effort on developing medicines against coronaviruses. There are a number of promising leads, stopping this research now because of the success of the vaccination campaign is unwise. Vaccines take time to develop and administer, having medicines which are active against all coronaviruses, even if not perfect, would give the world the time between a new virus emerging and the completion of a vaccination campaign.”
All our previous output on this subject can be seen at this weblink:
www.sciencemediacentre.org/tag/covid-19
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.