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.
Dr Simon Clarke, Associate Professor of Cellular Microbiology at the University of Reading, said:
“The latest ONS infection data show that we’ve reached a point where the numbers of people in the population infected with the coronavirus is levelling out. It seems that we are currently at the shoulder of the peak and can expect to see the numbers stabilise over the next week or so, before decreasing; this stabilisation is also reflected in the decreased R number announced today. The numbers of hospitalisations and deaths can be expected to stabilise in the coming weeks. Two weeks into lockdown, this is where we should expect to be, but it must be remembered that the burden of the virus in the community remains relatively high. We are not yet in a place where the restrictions on our lives can be lifted without the virus ‘taking off’ again.”
Prof Sheila Bird, Formerly Programme Leader, MRC Biostatistics Unit, University of Cambridge, said:
“My reading of the confidence intervals about ONS Infection Survey’s weekly reported estimates of SARS-CoV-2 incidence for England is, broadly, of increased incidence after the first and second fortnights. We do not yet have convincing evidence that England’s incidence is back down to around 5 per 10,000.
“REACT-1 versus ONS infection Survey in England: Unlike REACT-1, the ONS Infection Survey now has low volunteer-rates despite reimbursement and is shy – even in data-files – to cite number of participants by region, age-group, deprivation quintile, etc. so that readers can appreciate the before/after impact of the re-weighting that underlies much of the output we read. Let’s have better transparency, please.
“Re-weighting to each nation’s demography: England’s sample size is around 24 times greater than Scotland’s in the most recent two fortnights and, consequently, England’s 95% confidence interval would be expected to be about a fifth the width of Scotland’s. Comparison of re-weighting percentages testing positive for COVID-19 is illustrated below for England, Wales, Norther Ireland and Scotland. In the most recent fortnight, re-weighted prevalence is significantly lower in Northern Ireland and Scotland than in England.
“England’s re-weighting: Comparison can be made only indirectly between “Official reported estimates of the percentage testing positive, England” (see Table 1a) which are derived by use of modelling and the re-weighted estimates for the above non-overlapping 14-day periods. Modelling gives 1.04% [0.98% to 1.10%] for 17-23 October and 1.13% [1.07% to 1.20%] for 25-31 October, both systematically lower than the above re-weighted estimate.
“Why? Does this mean that the modelled Official Reported Estimates are not re-weighted?
“See ONS Infection Survey’s Table 1i for corresponding re-weighted information for all English regions, as shown below for the North East and the North West. The population of the North West is about 2.7 times greater than in the North East.
“In the three most recent fortnights, prevalence was clearly higher in the North West than in North East – albeit increasing in both. And, in each fortnight, prevalence in the North East was markedly higher than in Scotland.”
Prof Kevin McConway, Emeritus Professor of Applied Statistics, The Open University, said:
“Today’s weekly update on the ONS infection survey takes the estimates up to the week ending 14 November. As always, the survey should give more reliable estimates of the number of infected people that can be obtained from Test and Trace data and the daily counts of new confirmed cases on the gov.uk dashboard. That’s because it tests a representative sample of the community population of England, and tests them only to estimate infection rates, not because they have symptoms and ask for a test, or work in certain jobs, or live in certain places.
“Overall, for England, the good news nationally is that the estimated number of people who would test positive for the virus seems to be increasing much more slowly than a few weeks ago, and indeed this week’s numbers are little different from last week’s. For the most recent week, the estimate is that 1.22% of the English community population would test positive. Last week’s estimate was 1.20%. But these numbers, as with any estimate based on a survey, have margins of possible statistical error around them – from 1.15% to 1.29% for the latest week. So this week’s estimate is too close to last week’s to be sure what direction the rate of infection has actually moved. In numbers, ONS estimate for the latest week that between 1 in 80 and 1 in 85 of the English community population (of about 54.5 million people aged 2 and over) would test positive. The range for the previous week was 1 in 80 to 1 in 90. This isn’t the best news we could have – there’s no clear indication that the number of infected people is falling, and it needs to fall if we are to get out of the restrictions we are under. The latest data are for the week from 8 to 14 November, so entirely within the period of the current national restrictions in England – but the estimated numbers of infected people include some people who would have been infected before those restrictions came in on 5 November and I wouldn’t have expected a big fall yet, however effective those restrictions turn out to be. The numbers I have quoted so far are estimates of the number of people in the population who would test positive if a swab was taken. At these levels of infection, false negatives will exceed false positives on this test. ONS have again produced some estimates for the numbers that are infected, allowing for the possibilities of false negatives and false positives, and for the most recent fortnight (1 to 14 November) that would give a range for the number of people infected from 1 in 60 to 1 in 70 of the population. That’s still a lot of infected people.
“The ONS survey also produces estimates of the number of new infections per day. It can do that separately from the total number of infections, because people are tested more than once. The latest week’s estimate is 714 new infections per day, per million people, with a range representing the likely statistical error from 629 to 809 per day, per million people. In numbers across the English population every day, that’s a range from 34,300 to 44,100 new infections each day. There’s relatively more uncertainty about these estimates than about the estimates of all people who would test positive, because there are fewer positive tests on which to base the estimates. This is on the whole better news than for the total infections – the central estimate has fallen by about a fifth since the previous week. We can’t be sure that there is a real fall, because of the relatively high level of statistical uncertainty, but it’s certainly a rough indication that, nationally, things really have started moving in the right direction.
“As usual, there are differences in infection levels between different parts of England and between different age groups. On these, ONS provides estimates only for the numbers who would test positive, and not for the rate of new infections. I believe that that is because the estimates for regions and for age groups are bound to be subject to greater statistical uncertainty, because they are based on smaller numbers of samples than are the national figures, and because estimates of new infections (as opposed to total positive tests) are also subject to more inaccuracy. There are indications that infection rates are falling in at least two of the regions where they have been high: the North West and the East Midlands. In the East Midlands, the rate could well now be below the England average. The directions of change are less clear now in some of the other regions where they have been high: Yorkshire & the Humber, West Midlands, and North East. The regions in the South of England (London, South East, South West, East of England) have infection rates below the England average, but there are reasonably clear indications that rates in all of those regions (except the South West) are continuing to increase. What all these regional averages hide, however, is that there can be large variations in infection rates between different parts of the same region. This week, ONS have added to their report estimates of the rate of positive tests in smaller areas of the UK. I won’t describe the details of these, and care needs to be taken because the numbers tested in the survey in these areas are rather small so that the statistical uncertainty is relatively large (and is shown in the interactive graphic within the report). But they do make it clear how variable the swab positivity rate can be between different parts of a single region. Definitely worth a look.
“On age groups, ONS report indications of a continuing increase in infection rates in primary school age children, though rates there are still close to the average across all age groups. In the slightly older age groups (secondary school age up to age 34), the infection rates continue to be higher than primary school ages and also than older age groups, but the pattern of change isn’t consistent. Rates seem to be continuing to fall the group aged from about 17 (school year 12) to 24 – that group has for many weeks had the highest estimated rate of positive tests of all the age groups considered, but the rate there is now much closer to the groups just older and just younger, and indeed it’s quite likely that the infection rate for younger secondary school age students is now higher. For the 25-34 group, the rate looks as if it has levelled off. In the older age groups, there were some worrying indications in weeks before the latest one that their infection rates were increasing, but those increases now appear to have levelled off.
“The ONS survey has now been running for some time in the other UK countries. The trends there are generally less clear, because the survey there hasn’t been running so long and because the number of swabs tested is lower (mainly because the populations of the other countries are a lot smaller than the English population). ONS comment that the positivity rates in Wales and in Northern Ireland appear to be decreasing, from peaks in late October in Wales and mid-October in Northern Ireland. In Scotland, ONS say that positivity rates increased up to the end of October but now appear to have levelled off.
“In very broad terms, these results are compatible with the new estimates of R and the growth rate from SAGE and the Government Office for Science. This week the range for R is from 1.0 to 1.1. That indicates that it may well still be above 1, so that infections would continue to grow across the UK, with every 10 infected people infecting between 10 and 11 others. Last week, the range was 1.0 to 1.2, so R has probably fallen a bit. The range for the daily growth rate in infections is 0% (no change) to +2%. Since these estimates of R and the growth rate are based partly on data that change in line with new infections two or three of weeks ago, like hospital admissions and deaths, they do tend to lag maybe a couple of weeks behind current estimates of new infections. So it’s not yet really concerning that the current national restrictions in England have not brought the R range below 1 or brought negative growth rates (showing decreasing infections). Any changes from those restrictions will not yet have completely fed through into the SAGE estimates. The good news is that the R range for the North West region is now 0.8 to 1.0, so probably below 1. The infection rate has for some weeks been highest in that region, so it’s particularly good to see this indication that it is falling there. That indication matches what was found in the ONS infection survey, which has shown a fairly clear decline in infection rates in the North West for a couple of weeks. That’s probably long enough so that the time lags that are inevitable in the SAGE R estimates would have worked through. Let’s hope the same thing happens, in the next week or two, with the R ranges in other parts of the country.”
Prof James Naismith FRS FMedSci, Director of the Rosalind Franklin Institute, and University of Oxford, said:
“Yesterday saw just over 500 deaths announced. Everyone should reflect on the sadness this has meant for loved ones.
“The ONS data continue a run of data suggesting that the number of new infections is now beginning to fall (39,000, 95% CI 34,000 to 44,000).
“These numbers would be the first where we might hope to see the national lockdown beginning to impact.
“We know that social restrictions are the most effective way to bring down the number of new infections.
“The lower the number of infections, the less pressure on hospitals and the fewer number of deaths.
“Some people will live to see Christmas, who if these numbers had not fallen, would have died.
“Obviously as the restrictions continue, we all hope and I expect the number of new infections to further fall.
“The true success of the track and trace system is whether it prevents social restrictions. It has failed to achieve this.
“A successful system was always going to be very hard.
“We can see why it has failed, by observing that a large portion of community infections are not identified by testing (it varies but around half).
“Neither these infected people nor their contacts are isolating, this is the elephant in the room.
“Until this problem is fixed, the system will not work. The detailed numerical analysis week to week generates heat but little light.”
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.