Data, Stats Thread June 11

[PatriciaHolm]

UK govt pressers Slides & data

www.gov.uk/government/collections/slides-and-datasets-to-accompany-coronavirus-press-conferences#history

Data Dashboard coronavirus.data.gov.uk/
Covid 19 Genomics www.cogconsortium.uk/tools-analysis/public-data-analysis-2/
Covid 19 Variant Mapping Sanger Institute covid19.sanger.ac.uk/lineages/raw
NHS Vaccination data www.england.nhs.uk/statistics/statistical-work-areas/covid-19-vaccinations/
Global vaccination data ourworldindata.org/covid-vaccinations
R estimates UK & English regions www.gov.uk/guidance/the-r-number-in-the-uk
Imperial UK weekly LAs, cases / 100k, table, map, hotspots statistics imperialcollegelondon.github.io/covid19local/#map
NHS England Hospital activity www.england.nhs.uk/statistics/statistical-work-areas/covid-19-hospital-activity/
NHs England Daily deaths www.england.nhs.uk/statistics/statistical-work-areas/covid-19-daily-deaths/
Cases Tracker England Local Government lginform.local.gov.uk/reports/view/lga-research/covid-19-case-tracker
ONS MSAO Map English deaths www.england.nhs.uk/statistics/statistical-work-areas/covid-19-daily-deaths/
CovidMessenger live update by council area in England www.covidmessenger.com/
Scot gov Daily data www.gov.scot/publications/coronavirus-covid-19-daily-data-for-scotland/
Scotland TravellingTabby LAs, care homes, hospitals, tests, t&t www.travellingtabby.com/scotland-coronavirus-tracker/
PH Wales LAs, cases, tests, deaths Dashboard public.tableau.com/profile/public.health.wales.health.protection#!/vizhome/RapidCOVID-19virology-Public/Headlinesummary
ICNRC Intensive Care National Audit & Research reports www.icnarc.org/Our-Audit/Audits/Cmp/Reports
NHS t&t England & UK testing Weekly stats www.gov.uk/government/collections/nhs-test-and-trace-statistics-england-weekly-reports
PHE Surveillance reports & LA Local Watchlist Maps by LSOA (from last summer) www.gov.uk/government/collections/nhs-test-and-trace-statistics-england-weekly-reports
ONS England infection surveillance report each Friday www.ons.gov.uk/peoplepopulationandcommunity/healthandsocialcare/conditionsanddiseases/bulletins/coronaviruscovid19infectionsurveypilot/previousReleases
Datasets for ONS surveillance reports www.ons.gov.uk/peoplepopulationandcommunity/healthandsocialcare/conditionsanddiseases/datasets/coronaviruscovid19infectionsurveydata/2020
ONS Roundup deaths, infections & economic reports www.ons.gov.uk/peoplepopulationandcommunity/healthandsocialcare/conditionsanddiseases/articles/coronaviruscovid19roundup/2020-03-26
Zoe UK data covid.joinzoe.com/data#interactive-map
ECDC (European Centre for Disease Control rolling 14-day incidence EEA & UK www.ecdc.europa.eu/en/cases-2019-ncov-eueea
Worldometer UK page www.worldometers.info/coronavirus/country/uk/
Our World in Data GB test positivity etc, DIY country graphs ourworldindata.org/coronavirus/country/united-kingdom?country=~GBR
FT DIY graphs compare deaths, cases, raw / million pop ig.ft.com/coronavirus-chart/?areas=eur&areas=usa&areas=bra&areas=gbr&areas=cze&areas=hun&areasRegional=usny&areasRegional=usnj&areasRegional=usaz&areasRegional=usca&areasRegional=usnd&areasRegional=ussd&cumulative=0&logScale=0&per100K=1&startDate=2020-09-01&values=deaths
PHE local health data fingertips.phe.org.uk/profile/health-profiles
Alama Personal COVID risk assessment alama.org.uk/covid-19-medical-risk-assessment/
Local Mobility Reports for countries www.google.com/covid19/mobility/
UK Highstreet Tracker for cities & large towns Footfall, spend index, workers, visitors, economic recovery www.centreforcities.org/data/high-streets-recovery-tracker/

⏭ Our STUDIES Corner ⏮ www.mumsnet.com/Talk/coronavirus/3869571-Studies-corner?msgid=99913434

We welcome factual, data driven and analytical contributions
Please try to keep discussion focused on these

Can I check something with the experts on here?

On another thread, someone asked what the probability was of 4 LFT tests being negative (if someone was actually positive).

My calculation, taking the % of false negatives as around 50% (optimistic in real life usage for asymptomatic cases, but easy to work with), was that this was 1/16 (1 divided by 2 to the power 4).

Then the probability of 'being positive with 4 false negatives' overall was (1/16) * (1/560) (current prevalence of positive cases from ONS).

Obvious caveats are that if those 4 LFTs are taken at intervals over a period, and if you were infected at the start of the period of testing, the false negative probability would go down (because viral load would increase and thus the LFTs would become more accurate), whereas if they were all taken on the same day, the viral load would be the same on each occasion and thus the probability of false negatives would be exactly the same for each of the 4 tests.

However, I was quite loudly told that my calculations were completely wrong, because the false negatives were not independent of one another [this is the bit I don't get, mathematically] and because if it was the case that each was equally likely to give a false negative, we would be being asked to do many tests at a time to reduce false negatives [I suspect even if sensible, we are not being told to do this because it makes the false negative issue so much more visible).

Have I gone wrong, and if so, where?

@BlackeyedSusanWell the current UK numbers would then get split into very tiny groups.

It would be very possible to do that though internationally (and it would be good to look at other experiences with the delta variant).

I think so Jan.

I doubt the households that weren't keeping to lockdown closely want to test particularly. I understand the feeling of not knowing means you don't have to isolate. If you want to go out, do you want to test and risk you can't? ( Contemplating this as was poking the underside of my brain with a stick)

Too tiny to be of any use then in the UK then here, but would be interesting internationally. I suppose then you would be looking at countries with similar levels of health care too.

(It is a long time since I did stats and only at school)

I think your school needs a better system jan. The one we use looks like truely's and is one click.

I'd start a thread just as a survey, but it would be 'descended upon' by a certain subset!

@cantkeepawayforever

Can I check something with the experts on here?

On another thread, someone asked what the probability was of 4 LFT tests being negative (if someone was actually positive).

My calculation, taking the % of false negatives as around 50% (optimistic in real life usage for asymptomatic cases, but easy to work with), was that this was 1/16 (1 divided by 2 to the power 4).

Then the probability of 'being positive with 4 false negatives' overall was (1/16) * (1/560) (current prevalence of positive cases from ONS).

Obvious caveats are that if those 4 LFTs are taken at intervals over a period, and if you were infected at the start of the period of testing, the false negative probability would go down (because viral load would increase and thus the LFTs would become more accurate), whereas if they were all taken on the same day, the viral load would be the same on each occasion and thus the probability of false negatives would be exactly the same for each of the 4 tests.

However, I was quite loudly told that my calculations were completely wrong, because the false negatives were not independent of one another [this is the bit I don't get, mathematically] and because if it was the case that each was equally likely to give a false negative, we would be being asked to do many tests at a time to reduce false negatives [I suspect even if sensible, we are not being told to do this because it makes the false negative issue so much more visible).

Have I gone wrong, and if so, where?

I'd agree with you.

It depends on why they are false negatives.

1. operator error
2. All taken one after the other and low viral load
3. Faulty batch of tests with same fault
4. Something else that would effect all the tests done by that person.

Any of the above would mean that the probabilities of the tests giving 3 further false negatives , having had one, would be higher, so more likely to give 4 false negatives.

If it is simply down to test inaccuracy, then it would be (1/2)^4 (assuming 50% accuracy) as you said.

We're not going to be asked to give multiple tests because where do you stop? 4 tests? That's 1 in 16 will all be false negatives, so about 6 in every 100.
And if people have to do 4 tests that's 4 times the expense, plus you have to throw that off against doing 4 tests putting more people off doing them at all. As we were saying earlier, one test per child is a bit of a hassle, I'd imagine doing 4 tests would put off considerably more.

The problem is that we don't know more about the false negatives/false positives. It is a different matter if it's the above inaccuracies which mean if you get a false negative you're more likely to continue getting them, than if it's simply the tests make an error 50% of the time.

The odd thing was (I saw the thread you are referring to) that I felt that the person who was getting angry felt you were over exaggerating the probability of getting 4 false negatives. However their argument against your calculation would increase the probability of getting 4 false negatives not decrease it!

@Piggywaspushed

I think your school needs a better system jan. The one we use looks like truely's and is one click.

We have that, but were told that was in addition to the government site. If it's meant to be instead then I've frequently double reported, which adds a different problem!

Looking at cases per 100k the higher case numbers are localised to north west and particularly Scotland. What is going on in Sturgeon territory.

Cant - someone with a low viral load could continually get negative lateral flow tests as they are not very sensitive. An asymptomatic person with covid might never get a positive result with one.

I get that, lonelyplanet, but is that factored into the calculation anyway - so someone like that is in the 50% false negatives of the first test and the 50% false negatives in the final test too?

Or is it that the actual probability of 4 false negatives at a population level is HIGHER than 1/16, because if your viral load was low enough not to get a positive on the first, then you are more likely than 50% to get a negative on each of the subsequent ones too? Ie some people have a 1/2 probability of false negatives throughout (50% on the first making it 100% certain that the next 3 are negative even though you are infected)?

What also puzzles me is that the vaccines seem to decrease hospitalisation proportionately more than death in the latest figures

This could easily be explained if the deaths are not caused by covid, the rates of hospitalisation and death could be the same if covid+vaccination could not cause any illness at all (it would just be the background rate of +ve test in the sick population), which means the reduction in the two would look in the stats as proportionately higher in protecting against hospitalisation compared to death.

Yes, we have school and gov site for reporting too. Luckily, DS does do his own. Both sites together take me maybe 2 mins some of which is looking for my glasses!

@Angrymum22

Looking at cases per 100k the higher case numbers are localised to north west and particularly Scotland. What is going on in Sturgeon territory.

Im concerned about the people that will be allowed in Glasgow for the match with no tests at all , what can go wrong

I never understand probabilities... but have read a couple of (credible-sounding) Account’s of people who had several times tested positive on LF but negative on PCR. One had a HCP confirm that some people just do that. They weren’t anti-test /Covid denier rants.

piggy, how is my school devising a better system going to help my reporting on the gov website?

It isn't jan, just meant the school bit. But the gov bit takes me about a minute too, honest!

Vaccines don't set up a force-field that stop the virus getting into your nose and throat. The idea is that once it does the vaccine has primed your immune system to deal with it much quicker and better than it would if you were unvaccinated.

Over 1300 people die each day on average in the UK at this time of year. It wouldn't be unreasonable to expect that the majority of them would be over 65. It would also be likely that with health declining, many would be in close contact with a larger number of people than usual so would have an increased likelihood of enountering Covid and having it enter their nose and/or throat where it would be picked up on testing. Then, when they died of whatever ailed them, it would be within 28 days of a positive test, regardless of whether or not the presence of the virus had any discernable affect on their health.

As information about the state of health of those hospitalised or dying post vaccination is not provided, we have little idea of whether or not the vaccination is providing any, some or a lot of protection against Covid in elderly people who die. We do know, however, that people die as they age, as they always have, with or without Covid or vaccination.

And the older people who are more likely to die are also more likely to have been double vaccinated.

@amicissimma

And the older people who are more likely to die are also more likely to have been double vaccinated.

The thing is that the population who were double vaccinated were even older when the initial reports were coming out that vaccines gave very good protection against death.

The numbers though are pretty small all the way through.

So it could be

• a function of a new variant that behaves differently

• vaccine effectiveness waning

• complexities of diagnosis (mind you these older people might have been more likely to die of other things in winter and early spring, I'd guess)

• related to the small numbers of the cases we have

The thing is that the population who were double vaccinated were even older when the initial reports were coming out that vaccines gave very good protection against death.

This isn't true the initial data on which the JVCR based the 12 week interval was trial data, the majority in the trials were young and fit, very, very few will have been over 80.

Risk of death should be 95% in that double jabbed group and they make up only 29% of deaths instead. If 5% of deaths has become 61% then the risk of death has reduced dramatically.

At a certain point, the majority of deaths will be vaccinated because the majority of people will be vaccinated and:

1. someone has to be in the 5% who don't develop antibodies
2. someone has to be very frail or immune suppressed where any bug will carry them off but they've been exposed to covid so that's the one
3. someone has to die of other causes within 28 days of a positive test.

I forgot to also note than two jabs may need more than 14 days for immunity. I've no idea why scientists consistently keep using 2 weeks for efficacy when they know it takes longer (for AZ at least).

People much more better at maths etc than me have looked at the hospitalisation, deaths and vaccinations figures

They worked out that from the higher risk groups. We would have had 570 people die not 12 if they were not fully vaccinated. That works out at around 95% protection apparently

Probably not explained that correctly but either way 12 is a lot less than 570. So deaths after hospitalisation in the priority groups is in fact low. Something like that anyway🤷‍♀️

@MargaretThursday

The odd thing was (I saw the thread you are referring to) that I felt that the person who was getting angry felt you were over exaggerating the probability of getting 4 false negatives. However their argument against your calculation would increase the probability of getting 4 false negatives not decrease it!

Huh - the argument I was making was that you cannot remove the error by taking multiple tests!

You stated the reasons why an LF test may give a false -ve - the main one is likely to be lack of sufficient virus in the sample taken which could be user error, or simply low viral load in the patient.

While multiple attempts may reduce an error caused by user fault, you CANNOT change the patient's viral load, so it doesn't matter HOW many tests you take!

What made me angry though was the fact you have a numpty, who clearly doesn't understand how the tests work, issuing advice to others that totally contravenes the official guidance. It's not only irresponsible, it's a perfect example of how stupidity and arrogance go hand in hand so often.

@strangeshapedpotato - it's great to have a good debate and I see merit in what you say. On this particular very long-running thread though, many efforts have been made to avoid argument so by mutual consent posters don't target words like numpty and stupidity at other posters on the thread. Fine to use them elsewhere as is normal, but it would be better here to say "I disagree because", please.