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:
“This latest bulletin from the excellent ONS Infection Survey takes the information up to last week, the week ending 27 February, and the news is generally very good. The survey’s researchers report that the estimated percentages of people who would test positive for the virus that can cause Covid-19 have fallen since the previous week in all four UK countries, in most of the English regions, and in all age groups with one possible exception. Even in the regions and the age group where ONS can’t be confident that infections are falling, they aren’t concluding that they are rising, only that the trend is uncertain. Overall, the position is looking good. The virus hasn’t gone away yet, and hasn’t yet fallen to levels that we could live with in the longer term, in my view. But we’re getting there, and we’ve got effective vaccines now.
“Overall the falls in infections are pretty rapid. In England, for the most recent week, ONS estimate that 1 in 220 people in the community population would test positive. The week before that (the week ending 19 February, so the week before last), the estimate was 1 in 145. And if we go back to early January, the estimate was that 1 in 50 people would test positive. So a pretty big fall in the most recent week for which we’ve got data, and an immense fall since the post-Christmas peak. The positions in the other three UK countries are very broadly similar, though (particularly in Wales and Northern Ireland) there’s much more statistical uncertainty than in England, because the number of people included in the survey in those countries is much smaller. Allowing for the statistical margins of error, the proportions of the population testing positive in the other three countries are even smaller than in England, and not hugely different from one another – 1 in 285 in Wales, 1 in 325 in Northern Ireland, 1 in 335 in Scotland. (Because of the wide margins of error in Wales and Northern Ireland, we can’t be entirely sure that the rates there are definitely lower than in England, but they very likely are.) In three of the four countries, the estimated rate of testing positive is now at around the level of mid to late September – not yet down to the levels we saw in the middle of the summer (when the rate in England was about 1 in 2,000), but very much lower than they have been since October. I can’t say the same about Scotland on the ONS survey evidence, but that’s not any cause for concern. It’s only because the survey began later in Scotland than in the other three countries, and the Scottish data only go back to early October (when the infection rate was already quite a bit higher than it was last week).
“In England, ONS report that the rates of testing positive fell in most regions, though the current trend is uncertain in three of them – North East, East Midlands, and East of England. In each of those regions, infections have definitely fallen very substantially since January. The uncertainty is about the trend in the last week or two, where the decrease may have levelled off. But there’s quite a lot of statistical uncertainty about regional trends, because the number of people swabbed for the survey is obviously a lot smaller than the number swabbed in the whole country, so the margins of error are wider. The same applies to the estimates for different age groups. In England, ONS estimate that the rates of testing positive fell, comparing the most recent week with the week before, in all but one of the age groups they look at. The exception is for children of secondary school age, from about 11 to about 16 – and there, ONS are again not saying that infections are rising, just that the previous decrease may have levelled off in the last few weeks. The upshot of all these trends is that infection rates across all the age groups are more similar than they were during the January peak. That doesn’t apply to the oldest group in the report, those aged 70+, where the infection rate is relatively low. In that group, around 1 in 480 would test positive, compared to (for example) 1 in 225 for those aged between 25 and 49. It would be tempting to put this difference down to vaccination, but we can’t really say that at all. Infection rates in older people have generally been lower than other groups, even though the effects of infection, if someone is infected, can be much worse on average in older people. And, particularly during lockdowns and similar restrictions, it’s generally younger people who are more at risk of infection, because it’s more likely that they would have to go out of the home to work, possibly in front-line jobs.
Further information
“The big advantage of the ONS Infection Survey, compared to the numbers of new confirmed cases reported daily on the dashboard at coronavirus.data.gov.uk, is that the survey gets its estimates by testing a reasonably representative sample of people simply to estimate trends in infection. New case counts can be affected by changes in which people choose to come forward for testing, which people are actually allowed to book a test (because that may be impossible if one has no clear symptoms), what jobs people are in, and so on. That won’t apply to the ONS survey. Also, new case counts are obviously estimating people newly becoming infected (so-called incidence), while the ONS survey estimates the number of people infected on a given day (prevalence), which would include some whose infection started some time ago as well as those newly infected. Therefore you wouldn’t expect the two sets of estimates to track one another exactly anyway. In fact, they do match up pretty well. There has been some concern recently that the downward trend in confirmed cases was slowing up, or even levelling off, but that no longer seems to be the position. Since somewhere around 20 February, the confirmed cases have started decreasing faster again. There is a hint in the ONS survey data of a slight slowing up in the rate of decrease for the most recent week (though not for Scotland), but that is far from clear, and if the survey figures do continue to track the confirmed case counts, I’d expect that the slowing up might no longer be there in next week’s ONS data anyway.
“The ONS survey is not the only survey tracking infections in England. Imperial College’s REACT-1 survey does something similar, and also uses a reasonably representative sample of the English community population to examine trends in infection. The latest REACT-1 report1 came out earlier this week, and for its most recent period (13-23 February, their Round 9b), the REACT-1 researchers estimated that 0.47% of the English community population would test positive (which is about 1 in 210). That’s not a lot different from the ONS survey estimate for England for the most recent week (21-27 February), of 0.45% or 1 in 220 – though the time periods are different, and the average estimate from the ONS modelled estimates for the same period as the REACT-1 figure is about 0.6% or 1 in 165. However, I would not expect the two sets of survey results to match up exactly anyway. They use different lists from which to draw the people they test, they use rather different testing procedures, and they cover slightly different ages. (ONS includes people aged 2 or over, REACT-1 5 and over.) It would be more concerning if the general trends were different in the two surveys, and they are not. REACT-1 reported that there was some evidence from their latest part-round that the rate of decrease of infections was slowing up, compared to the very fast decrease between their January round and the first part of their February round. Their evidence for that slowing was based on data from a very short time span, so it is not completely convincing statistically. It does however match the pattern in the dashboard confirmed case numbers, and as I’ve mentioned, there is some hint of a slowing down of the decrease in the ONS results too. But that slowing up in the rate of decrease in the confirmed cases on the dashboard seems to have stopped at some time around 20 February, right at the end of the REACT-1 timescale, and any speeding up in those last two or three days of the REACT-1 round wouldn’t have shown up in their data yet. So, at the broad national level for England, the two surveys and the dashboard case numbers are showing similar trends. It’s great that we have these different sources of information for how infections are changing in England, and can check them against one another.
“Where REACT-1 and the ONS survey don’t entirely agree, however, is in their assessment of the regional trends. The REACT-1 report said that, comparing their two part-rounds during February, infections were falling in four of the English regions (North East, North West, East of England and South West), but showed no apparent change in either direction in Yorkshire and the Humber, and apparently rising in the remaining four regions (London, South East, East Midlands and West Midlands). The ONS survey reports that infections are falling in six of the English regions, but that the trends are unclear in the other three (North East, East Midlands and East of England). The time periods, and statistical estimation methods, are not the same, so I wouldn’t expect things to match up perfectly, and we’ve got to remember that both surveys show very big falls in infection in all regions since the January peak. But it’s a little strange that REACT-1 report a clear falling pattern in the North East and the East of England during February, while ONS report that the trend there is unclear, and that ONS report a falling trend in London, South East and West Midlands but REACT-1 report an increasing trend there. Regional estimates are much less precise, statistically, than national estimates, and that’s particularly true of the North East (which has the smallest population of all the regions, and hence the smallest numbers of people tested in the surveys and the least precise estimates). But it does illustrate that we should be careful not to over-interpret every small move in these regional estimates. The REACT-1 researchers fitted statistical models to some data for smaller areas, and reported some evidence that there were patterns of differences between different places that did not match clearly onto the standard regions. Many of the regions included areas where infections seemed to be increasing as well as other areas where they seemed to be decreasing. ONS do report figures for smaller areas too, but (so far at least) have not published an analysis that puts these together in the way that the REACT-1 researcher did. That might, or might not, throw some light in the differences between the two surveys in regional trends. In any case, I don’t think there’s great cause for concern now, but we’ve got to continue tracking infections regularly.”
1 https://spiral.imperial.ac.uk/handle/10044/1/86343
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.”