Daily numbers, graphs, analysis thread 9

[Barracker]

Welcome to thread 9 of the daily updates.

Resource links:
Worldometer UK page
Financial Times Daily updates and graphs
HSJ Coronavirus updates
Johns Hopkins Coronavirus Resource Centre
NHS England stats, including breakdown by Hospital Trust
Covidly.com to filter graphs using selected data filters
ONS statistics for CV related deaths outside hospitals, released weekly each Tuesday

Thank you to all contributors for their factual, data driven, and civil discussions.

However, I doubt “all” Swedes are obeying the rules! Swedes are still human after all.

How much do we trust all of the data ?

Every country has different systems
Many countries have no systems at all

Unless the data collection and descriptors are the same between countries, comparisons lose validity.

Less than 3% of births are "registered" in Tanzania
but over 20% of "deaths" are

Some countries have not had an accurate census for 20 years so
deaths per 100,000 population are finger in the air guess work
not statistics

Might Sweden have been ramping up tests like we have done, Derbygerbil?

Hi Folks long time lurker on here and the ONS website.
The NHS trust I work for has offered antibody tests to all 11000 staff. That will be interesting. They also did random voluntary testing of asymptomatic staff one weekend and just over 5% came back positive

Head of UK Statistics Authority "questions" Matt Hancock on his presentation of testing figures:
https://www.statisticsauthority.gov.uk/wp-content/uploads/2020/06/02.06.2020SDNNMattHancockkMP.pdf

Why are the UK daily figures always so late being released now? It used to be 2pm but it rarely seems to be before the briefing now?

It seems there has been a serious, possibly criminal failure to protect care home residents in certain parts of the country.

For example, in West Berkshire, 41 deaths took place in care homes in weeks 2-11. As of Census 2011, there were 696 care home residents.

This increased to 136 in weeks 12-21 (19.5% of 2011 residents) i.e. deaths were 3.3x normal levels. In the wider community, deaths increased by only 4%, though covid deaths did total 31% of w2-11 deaths.

Similar occurred in Merthyr Tydfil, where 47 deaths occurred in W12-21, which was up from 12 deaths in W2-11.

Unfortunately the C2011 data are quite old. So it may be that comparing the W2-W11 deaths to W12-21 is a more reasonable approach than estimating mortality from 2011 population counts.

The largest care home rises in deaths occurred in

Tower Hamlets (4.2x)
Enfield (4.0x)
Merthyr (3.9x)
Hammersmith (3.8x)
Islington (3.7x)
Newham (3.7x)
Hillingdon (3.5x)
Cardiff (3.5x)
Croydon (3.4x)
West Berks (3.3x)
Slough (3.3x)
Kingston London (3.2x)
Camden (3.2x)
Greenwich (3.2x)
Reading (3.2x)
Waltham F (3.2x)
Walsall (3.2x)
Merton (3.1x)
Hackney (3.1x)
Liverpool (3.0x)
Ealing (3.0x)
Haringey (3.0x)

Certain areas avoided 'harvesting'. Most notably, Cornwall which had 4737 care home residents in 2011 experienced a small rise from 327 to 387. The Isle of Wight, had a rise of only 18, out of 1835 residents in 2011, as did most of Wales:

Anglesey: fall from 37 to 25 deaths
Pembrokeshire: rise from 74 to 79
Denbighshire: rise from 55 to 60
Ceredigion: rise from 37 to 41
Gwynedd: rise from 60 to 78
Conwy: rise of 31%
Carmarthenshire: 41% rise
Vale of Glamorgan: 43% rise
Blaenau Gwent: 43% rise
Swansea: 43% rise
Neath Port Talbot: 44% rise
Wrexham: 52% rise
Torfaen 58% rise
Flintshire 67% rise
Carephilly 89% rise
Rhondda 108% rise
Newport 112%
Monmouth 118%
Powys 124%
Bridgend 129%
Cardiff 247%
Merthyr 292%

Considering the 50 worst local authorities, deaths increased from 2623 to 7846, an increase of 200%, or 5223 people.

The best 50 went from 3358 to 4183.

So with total E&W care home deaths increasing from 24,574 to 46,682, or 22,108 excess deaths, if everywhere had been as bad as the worst places, we'd have now 50,000 excess care home deaths alone. And if things had been like the best places, we've have only 6,037.

It's also worth noting that some of the places with the FEWEST care home deaths are also those who've had most relative care home deaths. E.g, in Arun (Littlehampton), care home deaths went from 131 to 208, but only 30 non-care home covid-19 deaths were declared, relative to 333 non-care home deaths overall.

Equally, Cornwall, Devon, etc. care homes have all relatively escaped, BUT this seems more like luck than judgement, as for example care home deaths in N Devon went from 57 to 91, while there was just 14 covid deaths in the wider community. However we need some caution - some of those 34 excess deaths might be 'noise', as only 10 were certified with covid-19. The true total will certainly be 20+ but probably not as high as the full excess...

Also in somewhere like Lincoln, which as part of the Lincolnshire region which has had very low covid-19 deaths, with just 15 non-care-home deaths compared to 125 deaths in w 2-11, the care home deaths went from 41 to 84 with just 7 certified covid-19.

If we look at the weekly data, then Lincoln averaged 4 weekly care home deaths W2-11, and I've attached a chart.

The chart shows NO obvious excess mortality for hospital deaths, whereas we should note covid-19 was only 400 people for W12 nationally, 1858 for W13, 5133 for W14, 8151 for W15, etc.

So the W12 care home death figure of 7 for Lincoln is almost certainly just noise - we don't expect care home residents to be dying yet, especially not in Lincoln. W13 is down to 5, and then W14-W19 show a consistent spike in deaths. These would have been workers in specific care homes infecting care homes. Note that the total population was 824 in 2011, and if we consider 30-40 excess deaths, and a death rate of up to 15%, and an average home size of 20 beds, then it implies that perhaps a third of Lincoln care homes were infected. This compares with somewhere like Enfield, which has both a large number of care home residents (1715 as of 2011), and a massive rise in deaths (68 to 274).

So we'd assume that every care home in Enfield was infected with covid-19, which isn't such a surprise given the MUCH higher level of community deaths in Enfield

Also, while we can find many places where the wider community has not suffered many deaths but care homes have, the reverse is not true. Essentially everywhere with large totals of non-care-home covid-19 deaths has also had large numbers of care home deaths.

Head of UK Statistics Authority "questions" Matt Hancock on his presentation of testing figure

He's correct of course. But when I was at school you could get an A Level in maths without taking any statistics at all. Now I think that's no longer the case, but the problem with the testing strategy is:

• the 100,000 test target is not in itself useful, meaningful or relevant, and there is no significance particularly to this number as opposed to 60,000 or 72,574 or whatever.
• however most people don't understand statistics at all, even a little bit, and rather than saying a 'sufficient tests for a 99% confidence interval for infection rates for the population' , we got this 3-year-old's '100,000 daily test' figure
• having made this target, the government was attacked for not delivering it, even though it had no inherent value.
• no, or few MPs, are capable of understanding the issues in Sir David's letter, even though they are obvious from the data without Sir David pointing them out, to anyone with a basic understanding of statistics.

With regard to the data presented it has been pisspoor since March. At that time I was calculating the difference between each day's results to try and work out the number testing positive, because it was not directly published from the government. It turns out this was not even possible because it was not made clear at that time that some people were tested twice.

It strikes me that all of the issues raised are not only obvious to me, to Sir David Norgrove, but should also be blindingly fucking obvious to whoever is being paid to work on this shit.

I mean if I was collecting a salary to put out these numbers I would be complaining to my boss that the job was shit. If I was doing this in the private sector as a consultant I'd be getting paid hundreds of pounds a day and I would consider it my responsibility to point this out. If my boss said 'don't care', then I'd make sure it was in writing in email, and continue sending out my invoices.

I don't know WHO is doing this job, I'm assuming it's not outsourced to Indian workers who culturally FAIL to address deficiencies in what has been asked for 'by the boss', and simply do what has been told. I'm assuming that this is being done by British workers. If this is being done via a fat short-term contract by some shitty consultancy firm who is billing a flat, fat rate, and will ask for a huge fee to make any changes, I can perhaps understand the outcome. But if it's being done by a salaried staff, the question is WHERE the failure occurs. Because someone in this chain of command should have a fucking clue what they are doing. And by that I don't mean Matt Hancock. Why is there no-one competent working on this?

Is it because:

(a) the relevant department at the DoH is shit, and the people working on this have no insight
(b) the relevant department at the DoH is shit, and the bosses crush the insights of the people working on this who understand the proble far better than they do
(c) both of the above
(d) because this was hurriedly outsourced to some cancerous consultancy firm

Either way there is a cultural problem.

Remind me not to work for the fucking government.

I don't quite understand why his concerns are in a public letter as opposed to an email. Did he contact the DfH. It seems a slightly odd way of working unless there is prior context.

Listening We're not looking at Tanzania when it comes to analysing UK data,

Even when comparing countries, it's just modern European countries, plus seeing what worked in HK, SKorea, Taiwan, Japan.

The ONS is scupulous about trying to produce the best quality data available for the UK

imo, the above countries are giving as accurate figures as they can
However, the whistleblower resigning in iirc Florida because of being asked to falsely reduce COVID deaths indicates political pressure may make data unreliable in certain USA states.

"why his concerns are in a public letter"

He already tried the quiet official route and / or wanted to express his general concern for the record ?

"the testing statistics still fall well short of its expectations. It is not surprising that given their inadequacy data on testing are so widely criticised and often mistrusted."

I'm glad to read that - I was beginning to think I was being too hard on the test stats
They seem deliberately murky, unclear about how many actual people tested and don't have the data on demographics etc that we need

The NHS trust I work for has offered antibody tests to all 11000 staff. That will be interesting. They also did random voluntary testing of asymptomatic staff one weekend and just over 5% came back positive

gingercat02 that's interesting (to me as someone who wants to know as much as possible about testing asymptomatic people!).
What do BigChocFrenzy and shoots make of this 5% figure compared with what we already know? Obviously it's likely to be higher in this population than in general but that's still potentially loads of people still carrying it around with no other attempt (unless there is more planned?) to identify asymptomatic cases?

I was also wondering what people thought about the statement I hear often which is 'we'll all get it at some point, there's no escape unless there is a vaccine'. Do the numbers indicate this, if we assume any vaccine is more than 2 years away for example? What about vulnerable people?

"We'll all get it at some point"

Not necessarily
and there may not be a 2nd wave

Cases & deaths have fallen massively across most of Europe and continue to fall despite relaxation of lockdown
We now know more about which social distancing measures are the most important; doctors know more about treating patients - this should improve further over time.

We must see if the strategy of local lockdown for new outbreaks can keep COVID under reasonable control

However, the whistleblower resigning in iirc Florida because of being asked to falsely reduce COVID deaths indicates political pressure may make data unreliable in certain USA states.

Nothing so dramatic as that.

She provided 'too much data'. Also she was asked to 'manipulate' data for rural counties for infection rates.

www.heraldtribune.com/opinion/20200602/editorial-investigate-claims-of-covid-19-data-manipulation

This might make sense if there are small samples/populations. But regardless, essentially she was providing A-Z whereas her bosses only wanted her to give A.

This looks a lot like what Sir David is complaining about here. In other words it might be that the shit statistics published on gov.uk are deliberately shit for political reasons.

So it's not obvious that we are any better than Florida inasmuch as they had an employee actually produce good data till they stopped her whereas in the UK such data has never been published at all.

Depends if you see it here as malicious or incompetent.

Interesting paper

Greater risk of severe COVID-19 in non-White ethnicities is not explained by cardiometabolic, socioeconomic, or behavioural factors, or by 25(OH)-vitamin D status: study of 1,326 cases from the UK Biobank

www.medrxiv.org/content/10.1101/2020.06.01.20118943v1

PHE: Disparities in the risk and outcomes of COVID-19

Confirms what we have seen before, but the 70 x death rate when infected, for age 80+ vs 40 still has the power to shock.
We need to keep repeating this to the young & middle-aged worried about being in the "vulnerable" 17 million

https://assets.publishing.service.gov.uk/government/uploads/system/uploads/attachmentdata/file/889195/disparitiess_review.pdf

Working age males diagnosed with COVID-19 were twice as likely to die as females.

Among people with a positive test, when compared with those under 40, those who were 80 or older were seventy times more likely to die.
....
The regional pattern in diagnoses rates and death rates in confirmed cases among males were similar.
London had the highest rates followed by the North West, the North East and the West Midlands.
The South West had the lowest.

For females the North East and the North West had higher diagnosis rates than London, while London had the highest death rate.

Local authorities with the highest diagnoses and death rates are mostly urban.
Death rates in London from COVID-19 were more than three times higher than in the region with the lowest rates, the South West.
This level of inequality between regions is much greater than the inequalities in all cause mortality rates in previous years.
....
ONS reported that men working as security guards, taxi drivers and chauffeurs, bus and coach drivers, chefs, sales and retail assistants, lower skilled workers in construction and processing plants,
and men and women working in social care had significantly high rates of death from COVID-19.

Our analysis expands on this and shows that nursing auxiliaries and assistants have seen an increase in all cause deaths since 2014 to 2018

"malicious or incompetent"

I suspect both, but am pretty sure of the incompetence at least

My Trust is also offering antibody blood tests to all staff within the hospital of all different roles.

I would love it if they did a mass testing of staff to see how many of us wandering around the hospital have it but are asymptomatic...

Sorry, skipped an important word: "age 80+ vs under 40"

"They also did random voluntary testing of asymptomatic staff one weekend and just over 5% came back positive"

Just to confirm: that's antibody testing ?
So they have had it sometime in the past and are now very probably immune

Frontline staff without PPE might be more likely to be infected than the general public, but overall death rates seem similar

BigChocFrenzy no the 5% was positive Covid 19 tests in asymptomatic staff sorry

OK, thx

Adam Kucharski@AdamJKucharski (math / epid LSHTM)

I'm getting asked more about
the 'k' parameter that describes variation in the reproduction number, R
(i.e. describes superspreading)

But what does this parameter actually mean?
A short statistical thread... 1/

R measures average transmission per case,
but in reality some cases may generate more infection than others,
e.g. because of events/places they visit while infectious.

So we need a way to estimate variation in R at the individual-level... 2/

We can do this by fitting a curve to the distribution of secondary infections, and see how much variation there is.

A commonly used tool is the negative binomial distribution,
which has mean=R and variation captured by a dispersion parameter 'k'

Superspreading and the effect of individual variation on disease emerg
From Typhoid Mary to SARS, it has long been known that some people spread disease more than others. But for diseases transmitted via casual contact, contagiousness arises from a plethora of social an…
www.nature.com/articles/nature04153
3/

If k is very large, every case generates transmission randomly at constant rate with mean=R
(i.e. equivalent to a Poisson process as k->infinity).

Here's what the distribution of transmission looks like when R=3, k=1000 (dashed line shows R). 4/

If k is smaller, then there is more variability - some cases generate a lot of new infections, while most generate very few.

Here's the negative binomial distribution when R=3, k=0.2 (plausible for COVID, SARS).
Note x-scale is cropped at 20, but can obviously go higher 5/

So how do we calculate k?
One way is to estimate directly from transmission chains reconstructed from contact tracing data, e.g.

Clustering and superspreading potential of severe acute respiratory syndrome coronavirus 2 (SARS-CoV-2) infections in Hong Kong
Superspreading events have characterised previous epidemics of severe acute respiratory syndrome coronavirus (SARS-CoV) and Middle East respiratory syndrome coronavirus (MERS-CoV) infections. Using c…
www.researchsquare.com/article/rs-29548/v1
6/

Alternatively, we can estimate k from the distribution of outbreak sizes after infections are introduced to a new location,
using a simple transmission model:

wellcomeopenresearch.org/articles/5-67 7/

As 'k' is a bit hard to interpret directly,
we can also use it to calculate what proportion of infections generate a given amount of transmission,
e.g. does transmission follow something like a '20/80 rule'?

Here's a conversion table for R=3... 8/8

324 today, lower than previous post-weekend figures, so looks like a steady but slow decline is continuing.