Daily numbers, graphs, analysis thread 11

[BigChocFrenzy]

Welcome to thread 11 of the daily updates

Resource links:

Slides & data UK govt pressers
NHS England stats including breakdown by Hospital Trust
ONS UK statistics for CV related deaths outside hospitals, released weekly each Tuesday
Financial Times Daily updates and graphs
HSJ Coronavirus updates
Worldometer UK page
Covidly.com to filter graphs using selected data filters ONS statistics for CV related deaths outside hospitals, released weekly each Tuesday
Plot COVID Graphs Our World in Data

We welcome factual, data driven, and civil discussions from all contributors 💐

I think this is positive tests through the app.

Interesting thanks. I see that Tim Spector does a weekly you tube update, which is due soon. It'll be interesting to hear what he says about the figures and Ab tests.

Has this vit d study been linked yet?

nutrition.bmj.com/content/early/2020/06/14/bmjnph-2020-000110

It was linked in this which is an easier read:

news.yahoo.com/more-evidence-lack-vitamin-d-linked-covid-19-125331345.html

Could someone give me the link to the stats from each local authority? I can't seem to find where you see them

@Firefliess

coronavirus.data.gov.uk/

Scroll down to “Download the latest cases data as CSV or JSON“. I’ve download the CSV file....it’s a massive file though - some Excel skills are very helpful to analyse it. Note local authority data is for pillar 1 tests only which appear to be only 20-25% of total positive tests (I think I slightly overestimated when I’ve posted previously), but nonetheless gives an idea of prevalence.

And 10 to 20% hospitalised for up to 190 days a bit much for the NHS to handle

The new UK dashboard is better

• all 4 nations, local authorities etc and Beta release with clearer testing and healthcare stats

Beta gives daily people tested positive, but is only the 1,006

"A week or more would not really change the r rate from 0.02 to 0.9 (around)"

larrygrylls With a low number of cases, we can expect UK R0 to oscillate a lot these coming months (but not 0.02 / 0.9)

• shooting over 1.0 with new local outbreaks, but then these coming under control.

e.g. Germany's 4-day R0 shot up from under 1.0 to 2.7 a week ago with that large meat plant outbreak,
but then was down to 0.57 by yesterday
The 7-day R0 smooths out small blips, but also oscillates after a large outbreak and has only just come down to 1.01

imo countries only need be really concerned if R0 stays above 1.0 for say a couple of weeks, which could restart exponential growth

• this can hopefully be avoided, but winter will be challenging.

Going by the fear of many particularly about schools returning,
we can expect some panic threads on MN as conditions are relaxed and we see more outbreaks / blips.

It'll take a while to accept this as the new normal, probably until a vaccination program,
or effective treatments to reduce to a tiny % those cases becoming serious enough for hospitalisation

The problem with R numbers is that they aren't real. If we test a whole infected meat plant, then it's going to tell us that the virus is going to.infect everyone in six weeks. Which is dumb and not true.

It's quite hard to produce R numbers when infection rates are fractions of a % and you essentially need to test everyone daily (ok not quite but close) to produce accurate numbers.

I commented on a past thread to some stupid clickbait local news story claiming R was above 1. This was bollocks based off of some startup feeding data into their algorithm but without further analysis , and then the journalist ignoring the disclaimers about the reliability of the data.

Infections have been going down for months and months. Continuously.

Without daily testing of everyone we don't really know what has happened, but we can be sure that there will people looking for data to 'prove'that going to beach caused loads of infections or whatever.

We need to know who was tested, where,and how many.

and as the numbers get smaller the R is even less 'real'

[quote wintertravel1980]ONS estimates are based on the numbers in the Shoots post. A shift from 11/22523 to 14/24256 doesn't say much. The numbers are simply very tiny.

If I am reading the government dashboard correctly, it looks like the Pillar 4 (surveillance) testing is being significantly expanded. It might represent ongoing testing of care home staff or it may mean the ONS testing sample (which is currently 22,000-24,000 people) may be increased to over 100,000+.

coronavirus-staging.data.gov.uk/testing[/quote]
Ooh, can I break out my party poppers yet (I'm a big fan of more testing)?! Perhaps not...
Anecdotally a friend got tested locally, within the hour, based on v mild cold symptoms, heaps of capacity.

We need to know who was tested, where,and how many.
I agree. I don't know why more data isn't localised - the pillar 1 case numbers by LA are v useful but only part of the picture.

DM's care home opening up next week to visitors. They were one of a group of homes that shut down early, and have had no reported deaths from Covid, so interesting that they are loosening their controls. One person can now visit their relative, for 30 mins, in the garden, with a mask, pre-booked.

I've just seen an email from the Zoe app people to explain the change in the method of calculating their figures. Seems quite a major cock-up on their part:

"You will notice that this week's incidence figures are lower than last week, yet in our latest update, we suggest that new daily cases have stopped falling in the UK this week. This is because our team took the decision this week to update our approach to calculating daily new cases. Our previous methodology incorrectly included some antibody test results, however, these results provide information about past COVID infections rather than current ones.

By removing reported antibody tests from our analysis, we are able to provide a more accurate reflection of the number of new COVID cases in the population. We have applied this correction to all of the incidence charts, which app users can view through the app."

Thanks @Derbygerbil. I've had some fun playing around with that data. Cases in my area are very low. So I will relax for now about kids seeing partners, etc, and keep an eye on the rates. Makes sense to be more cautious when there's more risk of catching anything. In a population of around 150k we've had no cases at all in the 10 days :)

@fadingfast "took the decision....." It's not a decision that needs careful consideration FFS! They clearly cocked it up badly, and have now started correcting it. They really ought to admit this was an error though. There's no way they could previously have thought it was sensible to include some (just some!) of the antibody tests in with the others.

I'm a bit confused

Yesterday's new cases were 1006 (according to government website). However, they say the cumulative total is 374 higher than if you added yesterday's total to the cumulative total due to historical cases.

So, does this mean that 374 cases occured/were identifed on other days (not yesterday) but hadn't been added to the relevant daily total because of reporting delays or whatever?

New cases today back below 1000 (890).

7 day rolling average also below 1000 for the first time since March

Just seen that too Doordine positive news about case numbers and, maybe more importantly, the 7 day rolling average.

Good news that cases continue to fall... Many areas haven’t seen a case (pillar one at least) for over a week now. Outbreaks appearing to be more localised.

I read this article

www.theguardian.com/uk-news/2020/jun/27/people-were-abandoned-injustices-of-pandemic-laid-bare-in-brent

It's somewhat misleading in that it's not quite clear that Brent did not have the most deaths per capita, only when 'age-standardised'

Also it focuses on a MSOA that turns out to be much larger than average, at 13,026 people. One in every 362 people died of covid-19. This is not even the highest in Brent, let alone London. It's only the third-worst hit area in Brent, behind Barnhill The Avenue (Brent 006 MSOA), and Stonebridge (Brent 027)

Church End turns out to be number 110 in the country, not number 2, when one adjusts for the fact it's almost twice the average MSOA size.

The highest death rate (per population) was in Sheffield 020 Crabtree & Fir Vale. More than double the death rate of Church End. This would still be higher on an age-adjusted basis, as Church End was 2.9% 80+, versus 3.6% 80+ for C&FV.

This latter figure is not large at all - East Devon 020 is 20% 80+, and East Devon as a whole 9% 80+. New Forest, Arun, Sefton, Christchurch, Poole and other areas also have a number of MSOAs with 14%+ 80+

It has been suggested that the high death rate in C&FV is due to the hospital, however as I understand it deaths are coded by place of residence, not death.

Colton, Austhorpe & Whitkirk, Leeds 073 was not far behind, and it has no hospitals

@PumpkinPie2016 Yes, exactly.

Case data is all over the place, as more than 2/3rds comes from pillar 2 - commercial partners testing in the community. So sometimes a backlog of data appears for old dates, which makes no sense to lump into today's numbers (which are themselves not great, as they are not relating to the last 24 hours anyway, they are just "cases we received results of yesterday")

It's like they sometimes find a whole bunch of test results down the back of a sofa (could be a data entry error, or a set of results wasn't sent in, an data audit throws something up or they realise they haven't been recording results from a test centre, who knows-they rarely tell us). Adding them to the daily figure wouldn't be representative so they do a historic correction. Ideally the data would be fully audited and finalised before we see it but the demands of the situation would make this difficult.

Thanks Patricia it gets so muddling! Same with 'daily deaths' because they didn't all occur in the past 24 hours. Ultimately, it obviously matter when someone died,it just matters that they did but it makes the data rather puzzling at times.