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:
“We are back to the regular weekly cycle of a detailed bulletin and data release each Friday on the ONS Infection Survey. The publication cycle was interrupted to some extent over the Christmas break. However, data collection did continue, except on 4 days (24 to 26 December, and 1 January), so that the regular flow of results can continue. (ONS say that the 4 missing days of data will have led to a greater level of statistical uncertainty than usual in the estimates in the report.)
“Given the rapidly changing state of the pandemic, and the consequent changes in Government policies, it’s unfortunate that the normal flow of reports had to be interrupted slightly, though the data collectors, ONS statisticians, and their academic collaborators needed a break as much as anyone else after their hard work on this survey for many months. The last regular bulletin (with some items missing) came out on Christmas Eve, but, to inform the media and the public about the reasons for Government decisions, there was an ad hoc data release on 5 January, just a few days ago. Much of the data in today’s bulletin was already published in that release, though today’s release adds the commentary and analysis that wasn’t included in the 5 January release, and there are additional data on some things.
“Because we’ve had most of the data already, my comments partly repeat what I wrote about the 5 January release. (Apologies if you have already waded through that once.) But I will first point out some features that weren’t there on 5 January.
“Today’s release includes data from Scotland, Wales and Northern Ireland, which were not provided on 5 January. The infection rates are smaller in all these countries than in England (and, probably, than in every English region other than the South-West). In Wales, ONS estimate that around 1 in 70 people would test positive for the virus, with a credible interval from 1 in 55 to 1 in 70 to represent the uncertainty. There’s evidence that the infection rate has started to fall in Wales. In Northern Ireland, the estimated rate is about 1 in 200 (interval 1 in 130 to 1 in 340), and though the rate had been decreasing since the peak in late October, ONS say that this decrease may have stopped. In Scotland, the estimated rate is 1 in 115 (interval 1 in 95 to 1 in 140), and is now showing signs of increase. In all these countries, the survey sample size is smaller than in England, essentially because the populations are smaller, so the level of statistical uncertainty is greater than for England taken as a whole.
“Today’s bulletin includes estimates of the proportion of people testing positive in relatively small areas of England (called ‘CIS sub-regions’ by ONS). That is accompanied by a map that indicates where the hotspots are. The map makes it clear that infection rates are high in much of London and the surrounding regions, but also high in certain parts of the North (including some relatively rural areas like Cumbria). The actual estimates of the numbers who would test positive in small areas are interesting, and scarily high for some places, but they are subject to a large degree of statistical uncertainty because the number of swabs taken within the survey in those areas can be quite small. For example, the place where I live (Milton Keynes) stands out on the map as having a darker colour (more infections) than the surrounding area, and ONS estimate that the percentage who would test positive is 3.66%, so that about 1 in 25 would test positive (rounded to the nearest 5). But the interval showing the statistical variability goes right from 1 in 15 to 1 in 45. (With rates like that, even if it’s 1 in 45, I will be being very careful.)
“These sub-regional estimates also make it very clear that the pattern of infection isn’t uniform right across regions, even geographically small regions like London. Differences in sub-regional estimates are subject to a great deal of uncertainty and shouldn’t be over-interpreted – but, for instance, in London the estimated rate of testing positive (rounded to the nearest 5) is as high as 1 in 15 in Barking and Dagenham, and 1 in 20 in Enfield, Barnet, Redbridge, Bexley and Croydon, but 1 in 75 in Harrow, 1 in 65 in Tower Hamlets and 1 in 60 in Ealing.
“Today’s release includes estimates, for England, of the rate of positive tests by age group (which were not provided on 5 January). Rates have been rising for most age groups, though not in groups of school age and younger where there’s evidence the rates are falling. Rates also appear to be falling for 35 to 49 year olds. The position in people aged from about 17 to 34 isn’t so clear – there’s quite a lot of statistical uncertainty for data on individual age groups – but the rises that were happening in those groups do look as if they might be levelling off. Overall, infection rates are still highest in people aged between about 12 and 34, but rates in older age groups, 50 to 69 and 70+, are roughly twice as high as they were three or four weeks ago. The statistical uncertainty means it’s difficult so say what the trend is in these groups, and perhaps their rates aren’t now rising a fast as they were, but the position is still worrying. It’s much too early for any effect of the new English lockdown to have shown up in these figures.
“As well as the data on swab tests that are compatible with the new virus variant that was first detected in England (see below; those data were in the 5 January release), data are provided in this new release on the Ct (cycle threshold) values recorded during the PCR testing of swabs in the whole UK, in each of the four UK countries, and in each English region. Some of these data were published in an ad hoc release by ONS before Christmas, but the new release updates those numbers. The Ct values are related to the viral load in the tested swab, which has some association with how infectious the person might be. However, I’m no virologist and cannot comment in detail on these values, and the text of the ONS bulletin does not interpret them. I hope someone with the appropriate expertise will comment. What they do illustrate, though, is that in fact the result of a PCR test goes further than a simple positive or negative result, and decisions have had to be made about which combinations of findings should be counted as positive.
“Now a slightly edited version of my comments from 5 January. These data on the numbers of infections in the English community are alarming, though not surprising given the bad news we’ve all heard recently about rising rates of infection. The data come from the ONS Infection Survey, in which a reasonably representative sample of the English community population is tested for the virus, not because the people tested have symptoms or are in a certain job or live in a place where mass testing is being done, but simply to monitor how infection levels are changing. Most of the information on the rapid rise in infection rates that we’ve seen in recent days comes from counts of confirmed cases (as shown on coronavirus.data.gov.uk, for example), and those numbers can depend on which people are coming forward for testing. The pattern of who is being tested generally could well have changed over the Christmas period, when most aspects of life aren’t the same as usual. So it’s particularly important to check the changes in confirmed cases against survey results which, I’d expect, would be much less affected by changes in people’s behaviour in deciding to be tested.
“The ONS survey numbers, like the counts of confirmed cases, do show a rapid increase since the November lockdown in England ended at the start of December. The percentage of people who would test positive for the virus, across England as a whole, more than doubled in the four weeks between the week ending 5 December and the week ending 2 January. Very roughly speaking, the percentage was increasing, over that period, at a rate that corresponds to a doubling in infections every three weeks, ending up with an estimate that 1 in 50 people in the English community population were infected by 2 January. That number comes from a survey, and all survey data are subject to some statistical uncertainty, but it is a large survey, and ONS state that the interval representing the uncertainty is pretty narrow – from 1 in 45 to 1 in 50 (rounded to the nearest 5).
“Over about the same period, the number of daily confirmed cases in England rose even faster, ending up at the end of December at three and a half times the number at the beginning of the month. Now it’s not really meaningful to compare these two rates of increase, because the daily confirmed cases figure is essentially a count of new cases, while the numbers from the infection survey are for people who have an infection at the time they were tested, which will include some people who have just newly been infected together with others whose infection happened some days or even a couple of weeks earlier. Because of this difference, I’d expect the number of people who would test positive to increase more slowly than the number of new confirmed cases, as indeed seems to be happening – but I think the ONS data might be compatible with a slightly lower rate of increase in the pandemic than the new confirmed cases. There isn’t really enough data to be at all sure of that, though – and in any case, an increase this fast is alarming even if it does turn out to be slightly slower than the increase in confirmed cases might indicate. Some hospitals in certain parts of the country are already pretty full. The lockdown announced this week will reduce new infections, and vaccination might possibly start to reduce new infections. But those effects won’t show up in hospital admissions for another two or three weeks, because it generally takes time after they are infected for people to become ill enough to need hospital treatment. People going into hospital now could well have been infected two or three weeks ago, when the proportion of people infected was more like 1 in 80 than the current 1 in 50. The numbers needing hospital treatment are going to keep rising. If they do double in the next three weeks, as more people who are already infected need hospitalisation, things are going to get very difficult in hospitals across a lot of the country.
“The ONS data are less precise on the infection rates in different regions of England, because the number of people tested in each region is smaller than for the whole country. Given the inevitable statistical uncertainty, there’s really no completely clear evidence of a difference in infection rates between regions, except for two. The rate of positive tests in London is estimated at about 1 in 30, and London is a big region so that there isn’t much statistical uncertainty (from about 1 in 25 to 1 in 30). That is very high indeed. There are some indications that the rate in London may have levelled off or even fallen in recent days, but that conclusion is far from being statistically convincing – I can do no more than hope it’s true. There are also some indications that the increases could be levelling off in the other regions near London (South East, and East of England) where the rates have been high, but again there’s considerable uncertainty about that.
“The only region where the rate is clearly lower than the English average is the South West, where around 1 person in every 135 would test positive (but with more uncertainty than in London – a range from 1 in 105 to 1 in 180). But you’ve got to remember that 1 in 135 is quite a high rate by the standards of just a couple of months ago. Indeed the rate in London was about 1 in 135 just about a month ago, at a time when the rate in the South West was only about 1 in 235. Infections have been increasing pretty fast in the South West, and there’s no place for complacency there or anywhere else.
“The new data also give information on the prevalence of the new variant of the virus, though there’s quite a lot of statistical uncertainty about many of those figures. The good news is that there’s some evidence that the prevalence of the new variant is levelling off, or possibly even falling, in some of the regions where it has been dominant (London and the South East, but not really in the East). But that evidence is statistically quite weak. The bad news is that the prevalence of the new variant is increasing in all the other English regions, and in Scotland and Northern Ireland (but not, currently, Wales). It may already have outstripped the older variants in the East Midlands, and isn’t far behind the older variants in some other regions.”
Dr Alexander Edwards, Associate Professor in Biomedical Technology, Reading School of Pharmacy, University of Reading, said:
“As we are only too sadly aware, the number of infections in the UK is now very high (around 1 in 50), placing massive pressures on our health service and also on public health monitoring (test, trace and isolate). This in turn makes it ever harder to contain spread. It’s stating the obvious – but the more cases out there, the harder it is to contain and prevent avoidable deaths, alongside the massive additional damage caused by NHS overload and the economic cost of restrictions that become inevitable with such high rates of infection.
“The exact number of infections has however been harder to follow in recent weeks, because of the understandable disruptions caused by holidays, and because of unknown impact of school and university holiday, Christmas visits, and the rapid frequent changes to restrictions. We applaud those continuing to work hard to track infections – without this we have no hope of tackling this deadly infection.
“Today’s release by the ONS infection survey starts to provide clarity about the current situation and will help to follow trends as the lockdown hopefully starts to cut the rate of spread.
“The ONS infection survey is different from the daily COVID-19 infection reports, and tells us more accurately about rates of increase or decrease. The daily reports reflect the vital symptomatic testing – essential for contact tracing to try to cut transmission. But there are several reasons symptomatic testing can’t tell us exactly how many people are infected. We have known for weeks that too many symptomatic tests are positive for this daily reported case number to be accurate – when more than 10% of tests conducted are positive, we know many cases are missing. Furthermore, over the break, maybe fewer symptomatic people will come forward for testing, and it may be harder to access testing.
“This is why the ONS infection survey is so vital – it randomly, systematically, and regularly tests volunteers and is therefore able to estimate accurately the overall pattern of infections. Unfortunately, this makes grim reading, because overall the virus has spread – yet again – to so many people across the UK. And we know now exactly how overall cases drive COVID-19 deaths and severe disease, and can expect high excess deaths for many weeks to come. It will take some time for us to know how spread is affected by the new lockdown – the next few weeks of ONS data will be even more important than today’s release.”
https://www.ons.gov.uk/releases/coronaviruscovid19infectionsurveyuk
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 others received.