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 Paul Hunter, Professor in Medicine, The Norwich School of Medicine, University of East Anglia, said:
“The important point to point out is that the ONS survey largely covers a period prior to the decline in cases, especially as this is a prevalence survey and people can be positive for some time after acquiring their infection. We will have to wait till next week before we can see any indication of the recent decline in cases. Generally changes in ONS data lag about 2 weeks behind daily cases data.”
Prof James Naismith, Director of the Rosalind Franklin Institute, and University of Oxford, said:
“The publication of today’s ONS data contain some hopeful news, some bad news and some troubling news.
“First the hopeful news, the rate of increase of prevalence in England is slowing. There is evidence of a small decline in prevalence in Scotland, which may also be seen in hospital admissions but it is far too early to say. However, where Scotland goes it is likely England will follow in time.
“The bad news is that the UK (all its parts) have a very high rate of prevalence. The prevalence of the virus continued to increase up to July 24 in England, that is more people were infected than got better; the very definition of increasing spread.
“The troubling news is that the daily number of test positives are currently a worthless guide to what is happening. The test positive daily data showed the beginning of a sharp drop in positive cases, we can now say this is not an accurate description of where things are.
“The ZOE app concluded the same thing, Professor Spector has made comments to this effect.
“Thus the breathless commentaries and confident predictions based on these data are misleading the general public.
“Many of these people who write these commentaries have a fixed view of what the right course of action is, data is carefully sifted and only choice bits selected to support their argument.
“This is not science; this is politics. The virus is not interested in politics.
“With so much death and disruption, the general public deserve much better. I find it depressing that we seem unable to learn lessons about over interpreting daily test numbers.
“In the ONS data, there is an estimate of the daily incidence per day; it’s in the spreadsheet. It is presented as 11.57 per ten thousand for the week ending 10th of July. With a population of 56 M in England, this is over 60 000 infections per day. The test positive data for that week are around 30, 000 for England. We are back to our old problem, around half the infected people are not being identified.
“I am not qualified to offer an explanation as to why the daily test numbers lost track of the infection at the moment (they will recover I am sure).
“I would like to make the following observations:
1 – We need to improve vaccine take up, personally I think 16 & 17 years old can give consent; there is no doubt that benefits far outweigh their risk. Vaccines are safe and effective. For children 12 to 16, I am not qualified to judge, I do think the JCVI provided a balanced and informed document.
2 – We need to improve ventilation urgently in schools before kids return, this age groups is where the virus grows most rapidly.
3 – Similarly we need to improve ventilation in indoor venues.
4 – I think Sir John Bell’s analysis of catch and release has much to recommend it. I am pleased that the government will implement it.
5 – No matter what, we are not going to rerun the deaths or illness we say in January, the vaccines have made a huge difference.
“The risks that I see are:
A – Long covid in the young, its poorly understood so there is a lot of uncertainty.
B – Spill over from high population prevalence into more vulnerable groups, delta is going to seek out those without or with only weak immunity (for whatever reason)
C – A wearing down of the NHS, I very much doubt we will burst hospitals, but medical staff need a break. Heading into winter with no respite from high demand is a recipe for difficulty.”
Dr Daniel Howdon, Senior Research Fellow, Health Economics, University of Leeds, said:
“It is important to distinguish between the number of people currently testing positive and the number of new infections. Given that some people continue to test positive for long periods of time after an initial infection, the number of people testing positive can continue to rise while the number of new infections falls – this is a particularly important distinction to make when we observe sharp recent large spikes in cases.
“It is worth noting that while the ONS provide estimates for the number of people testing positive up to the week ending 24 July, due to methodological difficulties in accurately estimating recent new infections, figures for the latter are only presented up to week ending 10 July. This was the day before the European Championship final, an event that plausibly caused the most recent large increase in cases, and the effects of which will be visible in ONS incidence estimates published next week. It should come as no surprise that estimates of new infections – mirroring the rise in cases detected until mid-July – continued to rise until this date. Given this, it is unlikely that the most useful comparison – using these estimates of new infections – with the falling new case numbers observed since mid-July can be drawn from these ONS estimates until the data release scheduled for two weeks’ time.”
Prof Kevin McConway, Emeritus Professor of Applied Statistics, The Open University, said:
“Sometimes the weekly bulletins on the ONS Covid-19 Infection Survey (CIS) don’t do much more than confirm that what we’ve seen in daily counts of new confirmed Covid-19 cases is reasonably accurate. That’s still important, because the CIS results aren’t subject to some biases than can affect the case counts on the dashboard at coronavirus.data.gov.uk. But I’m happy to admit that the weekly CIS bulletin isn’t always very interesting. However, this week’s CIS results aren’t like that, and the same will apply at least for the next week or two.
“The reason is that recently we’ve seen decreases in the daily numbers of new confirmed cases on the dashboard. In England, they fell day by day for roughly a week from about 15 July to about 24 July, the exact dates depending on whether they are classified by the date the result was reported or the date when the person involved was actually tested. The figures have been slowly rising since then. But many statisticians and epidemiologists, including me, were somewhat surprised by the fall, and it’s still not clear what the reasons for it might be. One possibility, though, is that, to some extent at least, it might not represent a true fall in the number of new cases. The new case counts on the dashboard come from numbers of people testing positive for the virus, so they depend on how many people turn up to be tested, and on why people are being tested. If those things change, that might cause changes in the numbers of confirmed cases that do not properly reflect what is actually going on in the country, in terms of infections. And it’s possible that they may have changed recently, because of tests of school students changing as school close for the summer, and perhaps because the concerns about people being ‘pinged’ by the NHS Covid-19 app and having to self-isolate could be deterring some people from coming forward to be tested. The CIS results shouldn’t be affected by biases like that, because the survey involves testing a representative sample of people for infection, simply to track the progress of the pandemic and not because they have symptoms or need to be tested for work or for school. So the new CIS results could throw a bit of light on what might have happened. Initially, the new CIS results don’t look very encouraging – but there are important reasons why things aren’t as bad as they look at first glance.
“The latest CIS results on the prevalence of testing positive, that is, the estimates of the numbers who would test positive for the virus that can cause Covid-19, whether they have a new infection or one that has been going on for a few days or longer, cover the week 18-24 July. That is, they cover the period when, according to the dashboard new case counts, numbers of new cases were falling in England (and, approximately, in Wales and Northern Ireland too – the pattern in Scotland is different and I’ll come to that later). New confirmed cases on the dashboard in England, by specimen date, fell by about 30% between the week 11-17 July and the week 18-24 July. However, the estimated number of people who would test positive in England rose by about 15% between those two weeks, according to the new CIS results. In England, ONS estimate that in that week ending 24 July, about 1 in every 65 people would test positive for the virus. Because this estimate comes from a survey, there’s inevitably some statistical uncertainty in it, and the margin of error runs from 1 in 60 to 1 in 70. But the previous week, the estimate was 1 in 75, the margins of error for the two weeks don’t overlap, and ONS do conclude that the number of people testing positive in England increased. They also estimate that numbers of people testing positive increased in Wales and in Northern Ireland between those weeks. On the face of it, that’s a bit discouraging. Perhaps it indicates that the reduction in dashboard figures for new cases was indeed due to some sort of bias in the dashboard counts, and wasn’t real.
However, that’s only one possibility, and there are others possibilities that are more encouraging. There’s a very important difference between what the CIS incidence counts measure and what the dashboard new cases counts measure. Those CIS figures that I’ve described represent all people who would test positive, regardless of whether their infection is brand new or whether it has been around longer. So some of the people who tested positive in the survey in the most recent week, 18-24 July, would have originally been infected the week before, or in some cases even earlier than that. New cases in England on the dashboard were increasing for most of the previous week (11-17 July) and indeed for a considerable time before that. So it’s possible that the new infections in the CIS figures for the week 18-24 July are indeed down on the new infections the previous week, but that we can’t see that in the prevalence estimates because the number of older infections from the previous week or weeks is hiding it. I absolutely can’t be sure that that is happening, but it’s a possibility. And it does fit in with the ONS statement that “there are possible signs that the rate of increase [in England] may have slowed”.
“The CIS does also provide so-called ‘incidence’ estimates of the numbers of new infections, as opposed to the ‘prevalence’ estimates of total numbers of currently infected people. However, these estimates take longer to produce, partly because they depend particularly strongly on the latest survey data. This week, the latest incidence estimates go up only to the week ending 10 July. In that week, ONS estimate that new infections were increasing in England, Wales and Northern Ireland, but look as if they decreased in Scotland compared to the previous week (ending 3 July). That does broadly match the patterns in the dashboard new case counts over that period – they were still increasing in three of the UK countries but started to decrease in Scotland at about the end of June.
“The prevalence (all infections) count from the CIS in Scotland for the most recent week (18-24 July) also fell, according to ONS, compared to the previous week. But that fall became clear only for the most recent week. Last week, ONS said that the trend for Scotland was ‘uncertain’, though in fact the estimated number testing positive was up a little compared to the week before, though not up by enough to be statistically certain. So it took until the most recent week for the decline in new infections to make itself clear as a decline in all infections, new or old. That’s a delay of about three weeks after the time when new infections began falling. That delay could well be because the initial falls in new infections were obscured by continuing numbers of people infected in previous days or weeks, going back to the time when new infections in Scotland were increasing. That does tend to increase my confidence that the same thing could be happening in England, Wales and Northern Ireland as well – that new infections are indeed falling but we can’t see that in the prevalence (all infections) data yet because it is obscured by older infections. But there’s no way I can be certain, yet, that that is happening.
“A further complication is that the relaxation of restrictions in England on July 19 might change the trend in new infections anyway. We wouldn’t see that yet in the latest CIS data, because that change occurred in the latest week of data, and it takes a few days for virus to build up enough to be detectable by the test (and to cause illness, if that happens).
“Because of all these different aspects, we won’t get a clear picture from CIS of exactly what’s happening in terms of new infections, or indeed of any effects of the 19 July changes, for a couple of weeks yet.”
Dr Simon Clarke, Associate Professor in Cellular Microbiology at the University of Reading, said:
“These ONS data are for the week ending 24 July, a period we were led to believe of declining infections. We’ve seen infections down, up, and with these ONS data, up again. Overall, we are seeing chaotic datasets that reflect a lot of different things happening at the same time. It’s important to remember that different data sets report different things. Daily infection numbers rely largely on self-reporting, but these ONS data use large scale sampling, across the country and in an unbiased way, giving us a more reliable indication of the considerable numbers of infections
“The ONS recently reported that more than 1 in 5 people who are known to be infected with the coronavirus do not self-isolate, which is likely to be one of the key drivers in increasing numbers. The virus only transmits when people interact with each other and things like school summer holidays, warm weather, and the lifting of restrictions are all factors which will determine how many people get infected. Recent hopes were pinned on the overall trend heading downwards, with some inaccurate commentary that ‘it’s all over’, but a few days infection numbers here or there are not a reliable way to predict what the future holds.
“While vaccinations are clearly having a very significant impact on reducing the worst effects of Covid, their efficacy is not 100% and we still have one fifth of the population unvaccinated. On top of that, the possibilities of future variants trying their best to get round our defences are something we should not ignore. If we’ve learnt anything in the last 18 months, it’s that infection trends can turn ‘on a dime’ and we should not be tempted into making attention-grabbing predictions. That the future remains uncertain is our only current certainty.”
Prof Rowland Kao who is Chair of Veterinary Epidemiology and Data Science, University of Edinburgh, said:
“The recent downturn in cases observed both in England and in Scotland has led to justifiable speculation about whether or not we are firmly heading towards lower cases numbers after the peak in mid July. ONS Surveillance data provide a useful, relatively unbiased snapshot of the level of infection in the population, however, because people can test positive with a substantial probability that a person will test positive even two weeks after becoming infected, the percentage of positive tests does not reflect current trends. Thus, a continued rise in the percentage testing positive reflects this fact – i.e. that this datastream will typically respond more slowly to downturns in cases. As the downturn we observed in the middle of July was particularly pronounced, the difference is also more marked. For similar reasons, the recent upturn in recorded cases, is also therefore only having a small contribution to the continued increase in positive tests – its effects will only be observable later, should the rise be sustained – unfortunately we do not yet know if this is the case.”
All our previous output on this subject can be seen at this weblink:
www.sciencemediacentre.org/tag/covid-19
Declared interests
None received.