Source: United Kingdom – Executive Government & Departments
The Office for National Statistics (ONS) have released their monthly report on the Coronavirus Infection Survey looking at characteristics of people testing positive for COVID-19 in England, covering the month of October.
Prof Kevin McConway, Emeritus Professor of Applied Statistics, The Open University, said:
“These monthly reports on the ONS Infection Survey don’t really add to the main information about how infection rates are changing across the UK countries and in the English regions. Those aspects are covered by the weekly reports. So there’s not really anything here that tells us more about how fast the rate of infection might be increasing, or whether the recent increases are showing any signs of levelling off. We’ll (possibly) know more about that when the weekly results come out later this week. Instead the monthly reports provide useful information about some characteristics of the people who test positive, within the survey, that aren’t available in the weekly reports. The data have the same advantages as other results from the ONS survey. They are based on swab tests done on a representative sample of the community population (who do not live in communal establishments such as care homes or some University halls of residence). Those tests are done only to provide information on the state of the pandemic, and unlike most other data sources, the results are therefore not affected by changes in the availability of tests, or in the type of people that are being tested. For these monthly reports, results are given only for England.
“What the monthly results mostly do not provide, however, is a clear explanation of the reasons for the patterns that are reported. Generally that’s because the survey results themselves cannot provide clear information on causes – to find out more about what is driving the infection patterns would usually require other data beyond what is collected in the survey. But many of the patterns reported this month are interesting and do raise questions.
“It has been a feature of the results from the survey, since it began, that a large proportion of the people who test positive for the virus did not report symptoms at the time of their positive test, or indeed (often) at testing times the week before or the week after. In the latest data (covering the two weeks from 28 September to 11 October), about two-third (66%) of those who tested positive did not report any symptoms at the time of their positive test, and over half (55%) did not report any symptoms around the time of their positive test (including a week before it and a week after). That indicates that a large proportion of infected people seem not to show any symptoms. That’s not new in general – but the proportion reporting no symptoms was much higher in late June and early July than it is now (and has been since August). The ONS statisticians do not comment on reasons for this change. Is it something to do with the fact that the survey participants report the symptoms themselves, rather than having them diagnosed by health professionals? Maybe the true position on symptoms hasn’t changed, but people are reporting differently for some reason. Or does it have anything to do with the kind of people who were testing positive in June and July, compared to those testing positive now? But I’m speculating here – the available data don’t provide information on these points.
“The new report also gives data on three other characteristics of the people who tested positive – their ages, whether they live in urban or rural places, and whether they had travelled abroad in the 30 days before their positive test. For the first two of these (age, urban/rural), it’s not surprising, given previous data from the survey and data from other sources, that infection rates are higher in younger age groups, particularly 17 to 24 years of age, and that infection rates are generally higher in urban places than in rural places (even within the same region). But ONS indicate that these differences between ages and between urban and rural places appear to be more marked in the regions where the overall infection rate has been highest. That’s interesting, and potentially important in making decisions on measures to reduce infections, but again ONS don’t speculate on the reasons for these differences, and neither will I.
“The position on travelling abroad is possibly more interesting, but also harder to interpret. The pattern that ONS report is that the rate of testing positive was higher in people that had travelled abroad in the previous 30 days, than in people who hadn’t. That was particularly marked for tests carried out between mid-August and mid-September – for the second half of August, people who had travelled abroad in the previous 30 days were almost 7 times as likely to test positive as people who had not travelled abroad in that period. But since then, the difference in the rate of testing positive has got smaller and smaller (alongside the overall rate of positive tests increasing), and in the most recent period in this particular data (25 September to 8 October) there wasn’t clear statistical evidence for any difference in positivity between the overseas travellers and the others. It’s very tempting to speculate that, earlier in the year when infection rates were much lower than they are now, people were bringing back infections from their overseas trips. Indeed that could be the explanation, in part at least, and one could go on to speculate that the reason there’s no clear difference now is that infections being brought back from overseas simply don’t show up statistically on top of the much greater number of infections being transmitted within the country. But that’s not the only possible explanation. The trouble is that this is observational data, and it brings with it all the difficulties of interpretation that apply to any observational research. There will be many differences between the people who travelled abroad, and the people who didn’t, apart from their travel. Maybe they have different ages, or live in different kinds of place, or are better or worse off, or work in different types of job. Any of these differences, or combinations of them, might be the cause of differences in the positivity rate between travellers and non-travellers. The overseas travel itself might not be the cause of the difference in positivity, or only part of the cause. We just can’t tell.
“This issue is more important for the data on overseas travel than the data on age or on urban or rural residence. That’s because the statisticians made adjustments to their estimates of positivity rates, for different age groups and urban or rural residence, to allow for several differences in other characteristics of the people involved. These adjustments don’t entirely remove the difficulty of knowing what causes what, but they make it rather less acute. However, these adjustments were not done for the data on overseas travel, and that makes those figures particularly hard to interpret. So we must be really careful not to jump to conclusions about it.”
All our previous output on this subject can be seen at this weblink:
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.”