select search filters
briefings
roundups & rapid reactions
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 figures from the ONS Infection survey move the position on to the week from 29 November to 5 December. The headline results on numbers of people who would test positive in England, and indeed in the other UK countries, may look as if they don’t add much to what we’ve already heard from other data sources. Overall, in the English community population, the number that would test positive (if everyone could be tested) has fallen in the most recent week compared to the week before, and indeed is 28% down in the most recent week compared to the peak estimated rate of infection in the week from 8-14 November. In that week, ONS estimated that 1 in 80 people would test positive – for the 29 November-5 December week they are estimating 1 in 115 (though with an interval, allowing for statistical uncertainty in making an estimate from a sample survey, from 1 in 105 to 1 in 120). There are some strong signs that infections are increasing in London, and statistically weaker evidence of a rise in the East of England too. Nationally, infection rates are highest in the age group that contains secondary school age children.
“That does all match up with evidence from numbers of confirmed cases, and some of that evidence is newer than the latest survey results. But that certainly does not mean that the ONS survey does not matter. Numbers of confirmed cases depend on who comes forward to be tested, and that in turn depends on a lot of other matters – which people feel that their symptoms are clear enough to ask for a test, where people live (because some areas have started mass testing of people who do not have symptoms), and so on. They can also be affected by glitches and technical problems in the Test and Trace system. The ONS survey tests a reasonably representative sample right across the population, regardless of whether they have symptoms and of other factors that might affect whether they would be eligible for a Test and Trace test. So these ONS results are a very important check on whether the numbers of confirmed cases are actually presenting an accurate picture. But there are other advantages too, that go well beyond the top-line results. For example, for some weeks now, ONS have been providing that they call a “sub-regional analysis” that gives estimates of the percentage that would test positive in reasonably small areas within England. Data are also available on confirmed cases in small areas, but they are likely to be affected by changes in local testing availability, and statistical glitches, to a considerable extent in some cases, probably much more so than estimates at a national or regional level, and it’s hard to get any kind of a measure of the size of the resulting uncertainty. The ONS infection survey figures for these regions also cannot always be very precise for these smaller areas, but they do give an interval that shows the uncertainty about the percentages.
“This week, one of the other usual advantages of the ONS infection survey results isn’t available. Because the ONS survey swabs people more than once at intervals of time, it has been able to estimate roughly when someone becomes infected, not just that they are infected at one point in time. A survey that swabs people just once can estimate who is infected at a point in time, but can’t tell whether they are just newly infected then, or have been infected for a week or two. Because the ONS survey can roughly measure, in some cases, when an infection actually begins, it has been able to give estimates of the incidence rate of infection – that is, the number of people who are newly infected in a week. However, these estimates have not been published for the latest week. That’s because ONS need to work further on how to do that estimation. They explain that a major reason for this reconsideration is because the intervals between swabs of the same person have changed, on average, because of the way the survey is conducted. In the survey, participants are swabbed weekly for five weeks after they are first recruited, and after that they are swabbed monthly. While the survey was getting under way and while numbers of participants were increasing fast, most results came from people being swabbed weekly. But now a greater proportion of participants have moved to monthly swabbing, simply because more of them have been in the survey for a relatively long time. Obviously, you can get a reasonably precise estimate of when a new infection starts if a person tested negative a week ago, and test positive this week. You don’t know the exact day, but you do know the week. If a person tested negative a month ago, and positive now, then you don’t know anywhere near as precisely when the new infection began. ONS have decided to take more time to work out the best way of dealing with the increased proportion of participants who are tested only once a month. I think that’s wise.
“The other figures that regularly appear each Friday are the revised ranges for the R number and the growth rate of infection, from SAGE and the Government Office for Science (GOS)1. This week, for the UK, the R range and the range for the growth rate are the same as they were two weeks ago, 0.9 to 1.0 for R and -2% to 0% per day for the growth rate. Both of those ranges are a little higher than they were last week (0.8 to 1.0 for R, -3% to -1% for the growth rate). This does not mean that either of the quantities has definitely increased compared to the week before – for example, R could have been 0.9 both weeks, and the growth rate -2% both weeks, and would still have been inside all the ranges. But the change does indicate that R might be closer to 1 than it was, and the number of infections may be decreasing more slowly, than was the case. Again that’s in line with evidence from other sources, including the ONS survey, that the reduction in infections may be slowing nationally.”
1 https://www.gov.uk/guidance/the-r-number-in-the-uk
Dr Simon Clarke, Associate Professor of Cellular Microbiology at the University of Reading, said:
“These latest ONS infection data show a mixed picture across the country. We’ve reached a point where there are clear decreases in the numbers of people infected with the coronavirus in areas that are largely under Tier 3, but nationally the rate of decrease may be slowing down or even levelling out. New infections in London are increasing and there are signs of a deteriorating situation in the East of England. It seems that we may be at the shoulder of the trough which might see the decline slow further and stop over the next week or so, before the Covid restrictions are loosened for Christmas. If this happens, the numbers of hospitalisations and deaths can also be expected to cease their decline in the coming weeks. Two weeks after lockdown was lifted, it is disappointing that infection numbers are not decreasing everywhere, but it is perhaps unsurprising given the time of year. Accordingly, decision makers and members of the public may wish to reflect on how they handle the festive period.”
Dr Daniel Howdon, Senior Research Fellow, Health Economics, University of Leeds, said:
“This week’s ONS Infection Survey features no new modelled incidence estimates. This week’s estimate of new such figures is replaced by a section explaining the absence of these figures pending a review, and offering guidance on how to interpret incidence estimates offered by the ONS each week. Following – to some extent understandably – confused media coverage in the last few days, it is welcome that the ONS are taking time to consider the best way to present these figures, both in terms of ensuring their accuracy and in terms of offering guidance on their interpretation. This will help to maintain the ONS’s well-deserved credibility in offering outstanding and comprehensive work on the provision of COVID-19 data.
“The trend in infection incidence numbers offered by the ONS as their best estimate – peaking around late October – looks to be in line with other publicly available COVID-19 data series for England.”
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 the Advisory Committee, but my quote above is in my capacity as a professional statistician.”
None received.