Daily numbers, graphs, analysis thread 16

[BigChocFrenzy]

Welcome to thread 16 of the daily updates

Resource links:

Uk dashboard deaths, cases, hospitals, tests - 4 nations, English regions & LAs
MSAO Map of English cases
[[https://assets.publishing.service.gov.uk/government/uploads/system/uploads/attachment_data/file/909430/Contain_framework_lower_tier_local_authority__14_August_2020.pdf
Slides & data UK govt pressers
UK added daily by PHE & DHSC
R estimates UK & English regions
PHE Surveillance report infections & watchlists every Thursday
ONS England infection surveillance reports
ONS UK death stats released each Tuesday
ECDC rolling 14-day incidence EEA & UK
Daily ECDC country detail UK
WHO dashboard
Worldometer UK page
Plot FT graphs compare countries deaths, cases, raw / million pop
Covidly.com world summary & graphs
Plot COVID Graphs Our World in Data test positivity etc

We welcome factual, data driven, and civil discussions from all contributors 📈 📉 📊 👍

Epidemiological perspectives on the number of infections in the under 20 age group:

https://www.sciencemediacenter.de/en/alle-angebote/rapid-reaction/details/news/zuwachs-an-jungen-sars-cov-2-infizierten-und-tests-unter-reiserueckkehrern/

Prof. Reinhard Berner
Head of the Clinic and Polyclinic for Pediatric and Adolescent Medicine, Carl Gustav Carus University Hospital, Dresden

“The picture of the extent to which children are at risk from a SARS-CoV-2 infection
and, in particular, to what extent they also make a relevant contribution to the transmission of the virus,
gradually merges with each study, like a giant puzzle.

And the puzzle is far from finished."

Prof. Dr. Ralf Reintjes
Professor of Epidemiology and Health Reporting, Hamburg University of Applied Sciences (HAW), Hamburg

“The cases that we received reports at the beginning of the epidemic only represented the absolute tip of the iceberg.

The people with severe symptoms were primarily elderly.
Many young patients, on the other hand, showed little or no symptoms.

As a result, they were often overlooked at first,
only testing who had symptoms and direct contact with a person with proven infection.

In addition, children and young people were protected by the school closings.

It is therefore logical that if we test more now and independently of symptoms, we will also observe more young people among the infected

At the same time, there are increasing reports of increasingly changed behavior with regard to the AHA rules
< German for distance, hygiene, masks >

There are two important points to consider:

The further the virus spreads among young people, the more widely and more diversely it can spread in society.

On the one hand, young people generally have more social contacts to whom they can pass the virus on than older people.

On the other hand, they often do not notice an infection in the absence of symptoms and therefore do not isolate themselves.

If more people become infected, the number of serious illnesses also increases.
The fewer symptoms infected people show, the less overview we have of who is really infected.
As a result, measures can be used in a less targeted manner, assessments of the current situation are less realistic.

It can be assumed that the number of infections will probably continue to rise in the coming days and weeks

• and that the average age of those infected will continue to decrease.

Factors for this are, among other things, probably vacation trips and the opening of schools.

As a result, young people who have previously been intensively protected and kept out of the infection process will also be increasingly affected in the future.

They are now increasingly exposed.
The average age will therefore go down significantly.

Since school lessons are very important for children, our main focus should be on making it safe

That's a great graphic boys3, thank you.

"False positives would of course explain rising cases and falling hospital admissions, which is still what we're seeing"

It would not really explain either rising or falling, unless the standard number of magnification cycles has been changed in the UK and most other European countries.

The massive drop in deaths & hospitalisations since March-April is to be expected with the massive drop in daily cases - as discussed, only a tiny % of actual cases were confirmed during peak

The continuing gradual fall in hospital and ICU admissions would fit with the continuing drop in average age of infections
There is an increasing % in the 10-30 age group in other European countries, likely too in the UK

The number of false positives from swab tests in the UK has been repeatedly estimated from serology tests as being around 1%, vs the 30% of false negatives
So I would not place much emphasis on 1 article in the NYT unless / until there are more studies to confirm this change in estimates

The most reliable estimation of cases in the UK would be from regular surveillance reports which include serology tests as well as swabs.

A false positive would explain a single case, but it much less likely to be the case in detected clusters of even 2-3 cases
Statistically, if the chance of a false positive is N%, then the chance of that for 2 specific people is NxN and in 3 people NxNxN

whatsnext2 posted this on the study thread www.medrxiv.org/content/10.1101/2020.08.10.20171413v1

Several interesting things in that - including a number testing negative but having antibodies. That certainly suggests false negatives are more significant than false positives. The initial tests are known to only work well for a limited time period.

The other interesting thing is that 40% were asymptomatic even in a population that requires the care of a nursing home, so are quite likely elderly with comorbidity.

I've been wondering about this study and implications:

secure.jbs.elsevierhealth.com/action/getSharedSiteSession?rc=1&redirect=https%3A%2F%2Fwww.jpeds.com%2Farticle%2FS0022-3476%2820%2931023-4%2Ffulltext

@NeurotrashWarrior

I've been wondering about this study and implications:

secure.jbs.elsevierhealth.com/action/getSharedSiteSession?rc=1&redirect=https%3A%2F%2Fwww.jpeds.com%2Farticle%2FS0022-3476%2820%2931023-4%2Ffulltext

.... That link won't open for me Is it members only ?

@BigChocFrenzy I didn’t think the NY a times article was really about false positives though was it? It wasn’t disputing that people were carrying the virus but whether it was at such a low level that it was pretty much irrelevant? I thought false positives were where there was actually no virus present at all?

Following on from the 10:30 graphic.

This is the actual cases movement for the same two weeks.

As it shows 61 LAs saw their case numbers fall; this includes places such as Leicester. Reduction of 708 cases; although Leicester contributed around 30% of this figure

For 13 LAs case numbers in both weeks were identical.

The vast majority of LAs though - 241 in total - showed an increase in case numbers, adding between them 3584 cases.

@Bigchoc - I should have said, rising numbers of cases could be accounted for by false positives in a situation of rising testing, which is what we have had in the last 2 months. We're now testing around 180,000 a day (pillar 1 and 2 only). A false positive rate of 1% would therefore find 1,800 cases a day, even if everyone was in fact negative. So clearly the false positive rate is lower than 1%. However a false positive rate of around 0.5% would give around 900 false positives a day, up from 450 in mid June when we were testing only 90,000 a day - which would explain a large proportion of the rise we've seen in the last 2 months.

The fact that hospital admissions are falling rather than steady is more likely due to older people managing to stay away from catching it, rather than being an actual fall in the number is cases.

You're right about genuine clusters being exceedingly unlikely to be false positives - if they are actually within a family or group of friends or workmates. But we can't tell from the MSOA maps that show 3 cases in a local area whether these are actually people who've been together or not. It's the rise in the very lowest rate areas that I'm most suspicious of - as shown in @boys3's chart.

Does this work?

www.jpeds.com/article/S0022-3476(20)31023-4/fulltext

Apologies if it's already been shared.

[quote MarcelineMissouri]@BigChocFrenzy I didn’t think the NY a times article was really about false positives though was it? It wasn’t disputing that people were carrying the virus but whether it was at such a low level that it was pretty much irrelevant? I thought false positives were where there was actually no virus present at all?[/quote]
....
People with v low levels of virus, epecially the young & healthy middle-aged, are quite likely to be asymptomatic

However, they may still shed virus and infect others who are more vulnerable
Hence why they need to isolate

Same as the 11:45 Graphic but using cases per 100,000 movement as opposed to purely the absolute case number change.

As per bigchoc's posts about the age profile changing what would be useful would be to understand the demographics of the confirmed cases at an LA level. PHE had some graphs in last Friday's surveillance report at a regional level, but without the actual data - making the graphs not that easy to read - beyond London's jump in the 20-29 year group.

It might be useful to show a graph with the demographic profile of LAs as compared with to their case movement. might being the operative word though.

[quote Firefliess]@Bigchoc - I should have said, rising numbers of cases could be accounted for by false positives in a situation of rising testing, which is what we have had in the last 2 months. We're now testing around 180,000 a day (pillar 1 and 2 only). A false positive rate of 1% would therefore find 1,800 cases a day, even if everyone was in fact negative. So clearly the false positive rate is lower than 1%. However a false positive rate of around 0.5% would give around 900 false positives a day, up from 450 in mid June when we were testing only 90,000 a day - which would explain a large proportion of the rise we've seen in the last 2 months.

The fact that hospital admissions are falling rather than steady is more likely due to older people managing to stay away from catching it, rather than being an actual fall in the number is cases.

You're right about genuine clusters being exceedingly unlikely to be false positives - if they are actually within a family or group of friends or workmates. But we can't tell from the MSOA maps that show 3 cases in a local area whether these are actually people who've been together or not. It's the rise in the very lowest rate areas that I'm most suspicious of - as shown in @boys3's chart. [/quote]
....
Many things are possible, but we should not say they are probable when there are other explanations just as likely
The key word is "could"

it is very unwise to seize on an article / study that confirms what one wishes, or believes to be true

The most reliable evidence comes from serology tests and surveillance reports, which are not currently indicating that the low % of swab positives is a significant over-estimate

IgM antibodies indicate the presence of infection

(As the infection clears, IgM antibodies decay and can no longer be detected.
Once that has happened, "IgG" antibodies form and these are what provide detectable immunity, whereas T cells reduce the severity of an infection)

The Covid Messenger service is MNr @littleowl1's baby!

Of course, with maybe 3% of the infections now compared to peak, very small errors in % positives become important
which is why it is sensible to require at least 2 confirmed cases before shutting down a whole class or business

• that masively reduces the chances:

There is no peer-reviewed study or afaik serious evidence that false positives are higher than false negatives
It was when the cases were much higher - and hence less sensitive to small errors - that the relative percentages were calculated and checked vs serology tests

A 1% chance of false positive is dwarfed by the 30% chance of false negatives, so it is probable that fewer people isolate when they shouldn't vs those who don't when they are positive - which is not a problem with the current much lower number of cases

@bigchoc. Whether a 1% false positive rate is smaller than a 30% false negative depends on the proportion of all cases tested that that are positive. Eg if 1000 positive people were tested then 700 would be positive and 300 false negative. But if you also tested 200,000 negative people you would find 2,000 false positive results from them (as well as 198,000 negatives) - dwarfing the 300 false negatives.

So you need to know the proportion of times that false positives and negatives occur, and also the positivity rate to work out which is larger. Nobody seems very sure what the actual false positive and negative proportions are, which is a problem.

Thanks to @littleowl1. Its a good service, I hope it's a success for you.

So much effort is going into getting this data and it is used so widely yet is unreliable. Do we know whether there is any research into the rates of false results whether positive or negative?

My understanding from what I've seen about the New York Times article (apologies if it's a different one as I don't have an account) is that the concern is the test picks up people as postive sometimes several weeks after their infection at which point they are no longer infectious. So it's not a case of overestimating cases (as they are true cases) but more about picking up cases too late and telling people to isolate after the horse has bolted so to speak. The author is questioning whether this is useful and whether our attention/money would be better spent on less sensitive but faster tests (like the rapid saliva ones that have been talked about it) that could be rolled out quickly in hotspots to try to quickly catch cases earlier.

To expand, the PCR test amplifies viral material by putting it through cycles. In the states in some labs, at least, the number of cycles allowed is so many that it can pick up very small levels of material. The person has/had covid-19 but is no longer infectious. So we are seeing cases too late

@Littlebelina

My understanding from what I've seen about the New York Times article (apologies if it's a different one as I don't have an account) is that the concern is the test picks up people as postive sometimes several weeks after their infection at which point they are no longer infectious. So it's not a case of overestimating cases (as they are true cases) but more about picking up cases too late and telling people to isolate after the horse has bolted so to speak. The author is questioning whether this is useful and whether our attention/money would be better spent on less sensitive but faster tests (like the rapid saliva ones that have been talked about it) that could be rolled out quickly in hotspots to try to quickly catch cases earlier.

This is my understanding too.

Thanks @Littlebelina. So that's not a problem that would affect our overall count of cases (as we only count new cases) It would be a problem for keyworkers who needed a negative test to return to work.

"Nobody seems very sure what the actual false positive and negative proportions are, which is a problem."

The proportions were estimated when there were a large number of cases and deaths
They were also checked vs serology tests for antibodies

If there is a significantly higher % of false positives, then this would mean there were fewer actual cases for the deaths

i.e. ==> mortality rate would be higher than currently estimated - and there is no real evidence for this

I can't access the NYT article - I may have exceeded my monthly free articles,
but I read that it was stating that many people were being told to isolate unnecessarily

However, someone being tested because of symptoms - which in itself increases the likelihood of COVID compared to a random person -
has a 30% chance of false negative according to e.g. the NHS vs a 1% positive according to available evidence

I repeat:
anything is possible , but we cannot throw away almost all previous evidence based on hundreds of studies on the basis of 1 NYT article, just because it says what people want to hear
We must have more studies from completely different & independent scientists

In other population studies, serology tests often found more people had been infected than expected
e.g. in the detailed study of Bad Feilbach in Bavaria, there were 2.6 x the number previously known
https://www.rki.de/DE/Content/Gesundheitsmonitoring/Studien/cml-studie/FactsheetBadd_Feilnbach.html