select search filters
briefings
roundups & rapid reactions
before the headlines
Fiona fox's blog
The Office for National Statistics (ONS) have released the latest date from their COVID-19 Infection Survey.
Dr Daniel Lawson, Senior Lecturer in Statistical Science, School of Mathematics, University of Bristol, said:
“The next few weeks are going to be critical and the ONS statistics are now significantly more reliable than the daily counts, since these are heavily biased by the lack of availability of tests.
“The ONS results are high quality and minimise bias. They support the clear picture that infection rates are growing in September, and this is driven by infection in people aged 17-34. Since universities have not restarted and 12-16s show no increase yet, this is not evidence that schools are part of the current infection problem.
“The ONS uses postal testing and sends their tests to the National Biosample Centre in Milton Keynes. This is one of the four main labs used for testing. At the moment it isn’t clear whether ONS data is prioritised, but we should hope so as this remains our best tool for quantifying the infection. Unfortunately the samples are not large enough to allow strong regional differentiation, and come with a delay, but they support the picture that some regions, in the North especially, were running out of control and that a local lockdown is an appropriate response.”
Dr Simon Clarke, Associate Professor of Cellular Microbiology at the University of Reading, said:
“These data show a significant increase in the number of infections, albeit possibly at a slower doubling-rate than had been reported previously. This does not yet mean that the epidemic is uncontrollable and it may still be possible to hold it at bay with sufficient social distancing and hygiene, but that does require the public to engage properly.
“There has been a lot of argument about the significance of the current uptick in infections, but this loses sight of the fact that in the spring, Covid-19 put a large numbers people into hospital, very quickly, hindering the ability of the NHS from delivering care to people in pain on waiting lists, cancer sufferers and denied pregnant women from having the support of their partners. That could easily happen again and academic hair-splitting makes no difference to people’s day-to-day lives.”
Prof Kevin McConway, Emeritus Professor of Applied Statistics, The Open University, said:
“Given all the other recent evidence pointing to increases in new COVID-19 infections, it’s not surprising that this week’s release of data from the Office for National Statistics (ONS) Infection Survey shows increases too. Data from this survey are generally easier to interpret than data from Test and Trace or from the daily UK government counts of new confirmed cases. That’s because those other data sources can be affected quite strongly by the availability of public tests and by where testing is targeted – where geographically, or which types of people might be finding it easier or harder to get a test. The ONS survey is carried out on a large representative sample of the community population of England, with the purpose of estimating the proportions and numbers of infections. So people in the survey are tested regardless of whether they are showing symptoms or have been in contact with an infected person.
“In fact the rates of increase of new cases in this latest ONS infection survey report are reasonably closely in line with the figures on new positive cases from the English Test and Trace report that came out yesterday. Using the central ONS estimates, the number of new infections in the latest week available – ending 10 September – were about 90% up on the previous week, and the previous week was about 60% up on the week before that. There’s some statistical uncertainty about those figures, as there always is on results derived from surveys. The latest week’s estimates are based on just 51 new infections in the most recent fortnight of swab tests in the survey, and that’s not enough to tie the estimates down very precisely. So the real increases in new infections out there in the population could be more than this, or less than this. The corresponding figures for yesterday’s NHS Test and Trace data were a 75% increase on the week ending 9 September compared to the previous week, and the previous week was 52% up on the week before that. These are really quite close to the ONS estimates, given the statistical margin of error in the ONS numbers and the potential biases in the Test and Trace numbers arising from possible changes in test availability and location. All of these figures are less than the doubling of new cases in a week that was being suggested in some recent media reports. But they are certainly high enough to be concerning, and the fact that the number from the two sources are comparable is another indication that this definitely isn’t just some sort of statistical fluke.
“The ONS estimate of the daily number of new infections in England, in the week ending 10 September, is 6,000, but the statistical margin of uncertainty means that the data are consistent with a daily rate between 4,200 and 8,300 daily infections. The average daily number of new confirmed cases for England reported on the UK Government coronavirus website at https://coronavirus.data.gov.uk/ is considerably lower even than the lowest plausible ONS estimate, at about 2,700. I think the main reason for this large difference is that the daily new confirmed case counts include only people whose infection as confirmed by a positive test. Most people who are tested are not in the ONS survey, or similar surveys, because those surveys are on a much smaller scale than the other ways of being tested. So most people being tested will have symptoms of some sort, even given the recent fuss about people with no symptoms requesting tests. But people tested for the ONS survey are not being tested because they have symptoms, but only to estimate infection rates across the country. The ONS survey has consistently found that the majority of people who test positive in the survey had no COVID-19 symptoms on the day of the test, or the previous week, or the following week. In the latest data1 on this, for August, only 28% of people with a positive swab test had symptoms at all on any of those days. That is, over two-thirds of people with a positive test results in the survey never reported any symptoms to the researchers. So the ONS infection survey estimate of new infections will include a much greater number of people who have no symptoms than is the case for the daily case counts.
“The ONS survey is also running in Wales and Northern Ireland, but started later in those countries than in England. There are not yet enough results to publish from Northern Ireland. The data from Wales do not so far show any evidence of an increase in numbers of positive tests, though the margin of statistical uncertainty in Wales is wider than in England because the number of people tests is smaller.”
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 member of the SMC Advisory Committee, but my quote above is in my capacity as a professional statistician.”
None others received.