Data & analysis thread, started 1 December

[NoGoodPunsLeft]

New thread!

Link to previous:

Data and analysis thread, started 12 November www.mumsnet.com/Talk/coronavirus/4077794-data-and-analysis-thread-started-12-november

@littleowl1 that bar chart shown as a time series just for the past few weeks would be interesting.

As you say in a really short time frame we have moved - England at least - from a very high percentage of councils with a downward direction to a real reversal.

That said there are still very clear regional differences, and for the current week their is little difference between the overall rate in Tier 2 as compared with Tier 3, and possibly if Kent were taken out of the calc Tier 3 might even be lower.

At this late stage I don’t think they will change the Christmas relaxation of restrictions.

The London tier decision this week will be interesting.

@BigWoollyJumpers DH has done a lot of work on longevity and deprevation. He has some uk maps he showed me recently of correlations by regional areas. Overlay the maps of deprevation, high co morbidities, low life span, and latterly Covid, and they match exactly. I don't understand that, surely the areas that have high levels of Covid aren't static. I know some areas have been stubbornly high but in other areas they rise and fall.

Is he looking at covid at a particular time or anywhere that has had high rates at any time during the pandemic? I'm thinking of places like Cornwall where there are very definitely areas of high deprivation but very low covid and in London life expectancy is high but covid is very variable. Of course Cornwall and London both have wealthy populations and very poor populations so they might be different to other areas.

Or is he looking at much smaller areas?

@Augustbreeze
Somewhere I did write if I was MIx :)

Matt Hancock statement in parliament this afternoon. Possible earlier Tier announcement for London. / ?

twitter.com/bbchughpym/status/1338441826097836032?s=21
From Hugh Pym’s twitter
“Health ministers and officials are understood to be “deeply concerned” about a sharp rise in virus cases across London. MPs from the capital were briefed this morning on the latest data. It’s understood that a move of London to Tier 3 is highly likely possibly today.“

@TheSunIsStillShining - Yes, that's exactly it! We mint even be able to skip the algorithm if we can pump enough data into the AI for it to start chain spotting, just say it gives enough into to target enhanced testing.

Oh yes sun you did, sorry!

@Jenasaurus - ok, my 14th Jan State-Of-Covid prediction for you, but to play fair everyone needs to do the same including you:

1. Cases 32k/day holding fairly static after some bumpy ups and downs,
2. Death rates just about stable at 500/day and everyone wondering when they will fall because 80+ vaccination is at 65% plus all care homes completing by end Jan; lots of reminders that it's deaths within 28 days of diagnosis and lots of concern that Pfizer vaccine transfer to care homes wasn't stable, think government are trying to kill people age obviously should have originally sent the Oxford vaccine instead.
3. South West and East of England having cases rising very fast while they're falling elsewhere hence the "static", plus a resurgence in just one of the badly hit northern cities (probably Manchester) and Kent (having had one peak and down, then rising again)
4. Major flu outbreak affecting NHS capacity and everyone wondering how it's transmitting.
5. New different viral strain detected in London causing slight difference i(n transmission / timeframe / severity / ...?) And covered by vaccine immunity from only some of the vaccines
6. New viral strain coming through from Africa that has major differences, nobody knows if this will affect vaccines or not and borders closing quickly
7. Vaccine postcode lottery is a big debate as some areas are already vaccinating 70+ and inviting CEV, while others have only done a small percentage of 80+.

netherlands have announced a full lockdown from wednesday until mid january.
schools will be closed though primary will stay open as childcare for key worker children.
secondary will move to online lessons.

Looks like London is going into Tier 3 this afternoon then. Must be what the 3.30 briefing is, right?

[quote MRex]@TheSunIsStillShining - Yes, that's exactly it! We mint even be able to skip the algorithm if we can pump enough data into the AI for it to start chain spotting, just say it gives enough into to target enhanced testing.[/quote]
The algorithm is the AI :)
Have a think about the following:
How could you substitute sensitive data?

I've been thinking on the data side.
Normally in AI training we use anonimized data, so there are many solutions that make personal data still usable but not identifiable.

1. We need identifiable data for tracking purposes. But after tracking and followup is done that data could be transferred to person's NHS record. Meaning that in 3 years time in case of an illness it could be relevant to know if tested for covid, infected. If someone doesn't need treatment or tested negative it's not recorded on NHS profile (I think, don't know for sure). So this could be an additional benefit.

And as usual I am screaming in my head that most ppl allow google to track them, but if that was to be used for outbreak control everyone would be up in arms about personal privacy. It's just idiotic.
I am not an altruistic person by nature, but there is a level of "greater good" where the benefits to all outweigh the benefits to me.

Anonymise by keeping it all at MSOA level and numbering by case identification, so case # 1849403 in MSOA X.
I don't think you need their actual contacts, places of work etc. Just the pattern that a chain-style clump has arisen on neighbouring MSOAs X and Y, so send the testers into that area and have them be guided by local public health about the streets and workplaces to start in.

London and parts off Essex and Herts to move into tier 3. Timing not yet stated. Hancock half hour at 15:30

I highlighted parts as that would be a departure from the initial approach, if that is what is going to be announced.

That would let the genie out of the wider bottle

@TheSunIsStillShining negative COVID tests are currently stored on patient records - mine show on my online patient access to my GP records

Timing from 00:01 this Wednesday

BBC news says it will be from 00.01 Wednesday

Sorry cross post

News today is saying antibody tests show 1 in 11 have had virus.
A quick calculation using number of deaths vs population show the overall death rate as 0.1% or 1 in 1000

@MRex

Anonymise by keeping it all at MSOA level and numbering by case identification, so case # 1849403 in MSOA X. I don't think you need their actual contacts, places of work etc. Just the pattern that a chain-style clump has arisen on neighbouring MSOAs X and Y, so send the testers into that area and have them be guided by local public health about the streets and workplaces to start in.

But if it is to have practical value - which I think is paramount- than at one point there has to be the opportunity to translate UID18473 into John Smith along the chain. This is what I referred to that health professionals need to identify how deep is it feasible to trace and contact.

Okay, read again and I can see how your solution would work.
Identify chain with persons 1-10. 1 and 5 are in school B which is blinking red: so knowing person 1 and 5 age we know yeargroups so PHE can go into school B without knowledge of actual identified person(s)

Okay, but what if :)
the blinking red by the AI is a place liek a bus or a restaurant where people are only in for and interim?

1. you'd need to do person tracking
2. just go with the fact that eg bus 77 at 4.34 at X, going to T was an infection point, but only trace the blinking reds that come after it.
3. It could be a combination dependent on potentiality of infection. Eg. above 80% of being infected the greater good comes into play and a separate entity who are already entitled to deal with personal level data (serco unfortunately) comes into play, extracts those specific data points ....

I am wondering if there is anything similar out there for ebola prediction?

[quote lurker101]@TheSunIsStillShining negative COVID tests are currently stored on patient records - mine show on my online patient access to my GP records[/quote]
Are you a regular NHS user or work for the NHS?
Either way, thanks, good to know! :) Just wondering if they treat records differently.

@Ninbuscl

News today is saying antibody tests show 1 in 11 have had virus. A quick calculation using number of deaths vs population show the overall death rate as 0.1% or 1 in 1000

I made that mistake at the beginning. You can't use whole population x covid deaths for covid death rate.

You should be using all infected (=67m/11?) = 6,090,909

so covid death rate is 63,000/6,090,909*100=1.03 %

Which is approximately what has been said for many months. between 1-3%

someone check my math pls.

Not seen a listing yet but if I heard Hancock correctly the only parts of Essex not being put into tier 3 are Colchester , Uttlesford and Tendring?

The antibody test infection number of 1/11 is for England and doesn't, crucially, include care home residents or the under 16s, which is likely to up the number infected significantly.

1/11 of England population of 55m = 5m. Assuming the under 16s have been infected in at least the same proportion as adults, that adds another approx 1.2m = 6.2m. Plus say a third of care home residents? 100k. Gives around 6.3m in England, vs deaths of 56k = 0.89%.

Which matches the REACT studies overall 0.9% done a few months ago pretty well.

www.medrxiv.org/content/10.1101/2020.08.12.20173690v2.full.pdf

Key to remember, though, that IFR will differ widely amongst the population, driven largely by age. REACT suggests over 10% for the over 75s, but 0.03% for the 15-44s.