select search filters
briefings
roundups & rapid reactions
Fiona fox's blog
NHS Test and Trace have released the latest figures from their work for the week 10th – 16th September.
Prof Kevin McConway, Emeritus Professor of Applied Statistics at The Open University, said:
“As seems common with these weekly Test and Trace statistics, there’s bad news, good news, and important difficulties of interpreting the numbers. Overall, the news is not marvellous, though it could be much worse, and the difficulties of interpretation are potentially very important.
“The obvious bad news is that the number of positive cases for the most recent week (10-16 September) is up again compared to the previous week. And, as the bulletin says, that number is getting on for 3 times the number for the last full week in August (20-26 August) – actually 2.8 times the number for that week.
“The good news, or at least slightly better news, is that the number for the most recent week isn’t a huge increase on the previous week. There were only 508 more positive tests in the week ending 16 September, compared to the week ending 9 September. That’s an increase of only about 3% (this calculation uses the revised figure for positive tests in the week ending 9 September, which is included in this week’s data.) Also, more tests were done in the most recent week than the week before – not a whole lot more, the number of tests is up by less than 1%, but with more tests, you’d expect to get more positive test results if nothing else has changed. But, even allowing for the increased number of tests carried out, the number for this week is still something like 2% up on the week before, so this isn’t the best possible news. Numbers of positive cases are still rising. But a small increase like that isn’t in accord with the suggestion that the number of cases is doubling in around a week. It doesn’t really make sense to compare just two weeks of data, though, because statistical variation can make that picture very unclear – but looking back across the positive cases since late August, the data are consistent with a doubling time of something between two and three weeks, rather than a week. That’s hardly good news either, though. If that trend were to continue we’d still be in a pretty bad position in a few weeks from now – it would just take longer to get to really high levels of cases than if the doubling time was one week. I’m hoping very much that the recently announced measures can stop us getting to that difficult position.
“However, it’s really difficult to draw any clear conclusions from the number of positive tests reported from NHS Track and Trace. That’s because the people who are tested are not representative of the population of England. People are tested for a reason – for instance, they have symptoms that might indicate Covid-19, or they are in a high-risk area, or maybe they work in a care home. People in many of these groups would be more likely to be infected than the general level in the population. If the type of people who are tested under Track and Trace changes, on average, from one week to another, then the number of positive tests would change even if the actual number of infected people in the whole population has not changed. Or the number of positive tests under Track and Trace might change in a different way than the number infected in the population. Issues about the availability of tests in some places, or the ease of getting a test appointment at all, or the time taken to return results, might affect the number of positive cases from Track and Trace in ways that are difficult or impossible to predict. So it’s difficult to draw really clear conclusions from the NHS Track and Trace data on what’s actually going on, overall, in the English population. We should know more tomorrow, when the latest results from the ONS Infection Survey are published. That survey tests a representative sample of people, regardless of whether they have symptoms, or where they work, or whether Track and Trace tests are easily available.
“A potentially important point is that the balance between Pillar 1 tests (in hospitals and outbreak locations) and Pillar 2 tests (national swab testing) has changed in the latest week compared to the week before. For the latest week the number of Pillar 1 tests went up by a fifth, compared to the previous week, but the number of Pillar 2 tests fell (by 7%). And the percentage of Pillar 1 tests that were positive increased considerably between these weeks (from under 1% to about 1.8%), while the percentage of Pillar 2 tests that were positive fell slightly (from about 4.1% to about 4.0%). The details of exactly how this might have happened will require more examination, but it does perhaps indicate that things are getting worse in hospitals now.”
Prof Sheila Bird, former Programme Leader, MRC Biostatistics Unit, University of Cambridge, said:
“On 23 July 2020, the Royal Statistical Society (RSS) made recommendations – thus far ignored – to remedy the failures of Test & Trace (T&T) to glean intelligence. First, we need information about the symptomatic test-positive rate for T&T’s two high-risk quarantined groups:
a) Members of the household of symptomatic index cases
b) External close contacts of symptomatic index cases.
“Second, robust intelligence is needed – requiring home-visits on randomly sampled days – about asymptomatic swab-positive rates in the first five days of quarantine and in the next nine days for T&T’s two high-risk groups.
“Third, those random home-visits allow adherence to “stay at home” to be properly audited.
“Without ever having measured robustly adherence to “stay at home”, as recommended by RSS, we appear now to leap into punishing those who fail to adhere. Punishment means that the door may remain closed to any T&T survey team who knocks on 2 random days [in first five days next nine days] of a) index-case household’s or b) external contact’s quarantine to offer swab-testing. Amnesty from punishment, please, for household members subject to these random visits in order that co-operation is not destroyed before we ever start! Scottish Universities’ student amnesty is well-done.
“After 16 weeks of T&T’s reporting on operations and 9 weeks after the RSS’s recommendations, we are still failing to learn about asymptomatic positivity, especially how soon those quarantined – either a) in the household of symptomatic index cases or b) as external close contact – become swab-test positive which could have economic implications if shorter quarantine were justifiable.
“To their credit, both Test & Trace in England and Test & Protect in Scotland are trying to use record-linkage to find out at least how many of those who have developed symptoms and tested positive had been in quarantined. The “quarantined symptomatic positives” are the “successes” of T&T/T&P because taken out of circulation while incubating and infectious. Both systems’ efforts have not got answers yet. In case they never succeed, a new approach is needed now which takes account of UK’s new contact tracing APPs.
“New approach: when a positive swab-test-result is reported back, ask (and record answers centrally) the person who receives the positive notification, what was their status when swab-tested:
1 = in quarantine or just-out of quarantine as member of household with a symptomatic index case
2 = in quarantine or just-out of quarantine as an external contact of a symptomatic positive case
3 = in quarantine or just-out of quarantine as having returned to UK from COVID-high-risk country
4 = alerted by app to take a swab-test because of recent at-risk contact
5 = tested as part of routine active surveillance as patient-facing social care or healthcare worker
6 = none of the above but I developed at least one of three key COVID symptoms and got tested
7 = other reason (please specify).
“The positives that take us by surprise (those coded 6 and 7) are the ones we need to count.”
Prof James Naismith FRS FMedSci, Director of the Rosalind Franklin Institute, and Professor of Structural Biology, University of Oxford, said:
“At face value, these data are better news than might have been expected. The number of people entering the tracing system in fact exceeded the number of people testing positive. This would indicate that the system is not yet being overwhelmed. Less than 80 % of those transferred were reached to provide their contacts, 75 % of those contacts were reached. This means around only 60 % of people who had contact with a positive case were contacted. The report is silent as to whether this enough or not. The report does not disclose the compliance with isolation, if this is low, the system will fail no matter how well we reach contacts. Another point to note is 60 % of the contacts reached reside in the same household, it needs to established whether this is reflection of the actual reality or self-reporting bias. This can be done by cross correlating patterns of virus spread, without it we could well be looking in the wrong places for contacts. We should have clearer statements as to what levels of contacts need to reached and isolate for it to have a meaningful impact on viral spread. There is a danger we simply obsess about weekly scores rather asking the key question about what the system is supposed to deliver.
“Even at face value, there are reasons to be uneasy. The report looks back a week ago to 10 – 16 September, and the number of positive tests in that week (19,278) was a small increase upon the previous week of 3 – 9 September (18,371). I would be concerned that the small weekly increase (900) reflects oddities in the reporting testing system, rather than a sudden plateau in viral cases. The numbers are clearly rising with a doubling trend of around 2 weeks. Such problems in reporting with sudden dips and spikes bedevilled clear public messaging last time, it would be disappointing if this were happening again. The report mentions issues with Pillar 2 and in this context Figure 3 showed a drop in positive cases whereas pillar 1 shows an increase.”
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 member of the SMC Advisory Committee, but my quote above is in my capacity as a professional statistician.”
None others received.