Science, Statistics and Studies [SARS-CoV-2]

Thanks for posting, great little science lesson over coffee.

https://amp.ft.com/content/67e6a4ee-3d05-43bc-ba03-e239799fa6ab#click=https://t.co/YPPhk6f4ms

Death registrations and extrapolations from that right now aren't that helpful.

They will be though. When it's over.
Looking at excess deaths is a good crude measure of "how many more people died than usual".

This bit:

The ONS data also showed that the vast majority of all excess deaths
were people aged over 75 years old. This age bracket accounted for 70
per cent of the total, the same proportion as those with Covid-19 on
their death certificates.

Will be worth looking at later. As will demographics and inequality.
Looking at death registrations, pre-processed, is a bit of nightmare. Free text innit.

Does anyone know why the graph data published by John Burn-Murdoch in the FT every day doesn't contain data for Sweden?

I think they might be reporting deaths in a different way but seems strange to leave out the only control data available in this global lockdown experiment.

Chris Giles again of the FT

Calling the registered deaths 'unprecedented' but looks within the standard deviation
https://twitter.com/ChrisGiles_

Its in there as one of the little grey lines and also the individual nation inset graphs below... or at least it was last I checked. I think they are just guessing which counties will have the broadest interest.

1 Attachment

HSJ have had researchers looking at all the reported healthcare /social care staff deaths

https://www.hsj.co.uk/exclusive-deaths-of-nhs-staff-from-covid-19-analysed/7027471.article

but looks within the standard deviation

Maybe you’re eyeballing something different to me, but definitely doesn’t look like we’re within +/-1SD (or even several SD) of the mean number of weekly deaths either in general, or for this time of year.

It is, it is the grey line below Turkey/small grey dot under Canada, you have to cross reference with the small multiples to identify all the countries

I think the use of the word unprecedented is hyperbole when there is clearly precedent in winter months, it is however unusual for the time of year

The little blue dots suggest a wide variation of excess deaths week by week for the whole season over the last 50 years.

Its certainly 'unprecedented' for week 15 to have this many deaths, but its not for the season as a whole to swing by the odd 40,000 eitherway.

I don’t know, even from eyeballing I’d guess that the gradient of the line (size of the spike) IS unprecedented in other time period 1970-2020. Sure there may have been a small number of weeks when the number of recorded deaths was similar ballpark to now, but I suspect that the magnitude of those spikes above the recent running average would be comparatively small and attributable to background factors such as (possibly) an interaction between unusually cold weather and economic hardship.

Either way I think the ft is using the Johns Hopkins database so it would be pretty easy to download the data to check... but I’ve got to get back to work.

https://www.hug-ge.ch/medias/communique-presse/seroprevalence-covid-19-premiere-estimation

Its ONS data not John Hopkins.

A better measure of the impact of Corona as it covers all deaths, not just those thought related to the COVID-19 virus .

https://www.ons.gov.uk/peoplepopulationandcommunity/healthandsocialcare/causesofdeath/articles/comparisonofweeklydeathoccurrencesinenglandandwales/uptoweekending10april2020

I couldn't find the full 50 years of death data online but I did find 2001 to 2017
*Couldn't be bothered to workout the Sigma by hand but Excel tells me it the dataset has a standard deviation of 34439

So you are right, 40000 deaths is outside of a single deviation
It might even be 'unprecedented' but with the level of varation, well within expectation.

Found the Corba charts sh0wn in the daily briefings are antithetical to those shown in the FT

1 Attachment

NYT drilling down into total deaths vs corona deaths for different countries.
https://www.nytimes.com/interactive/2020/04/21/world/coronavirus-missing-deaths.html

Again the control (sweden) showing negative correlation in government mandated/influenced lockdowns as an effective tool in the management of the disease.

1 Attachment

Better chart showing UK excess deaths in 2020 compared to the last five previous years.
Would like to see this done for the last 50 for it to be relative.

Should open the debate to why previous excess death rates were accepted (or even noticed)

Panel chart that and I might look at it.

Neil Ferguson finally open sources his model...

https://github.com/mrc-ide/covid-sim

*Warning you need visual studio and 20gb of Ram

Visual studio. Huh.

Latest release by ONS to 17th April https://www.ons.gov.uk/peoplepopulationandcommunity/birthsdeathsandmarriages/deaths/datasets/weeklyprovisionalfiguresondeathsregisteredinenglandandwales shows 1,776 total deaths from all respiratory diseases in England and Wales as underlying cause of death, this is a reduction in total respiratory deaths from the week two winter high of 2,477 and less than half the excess Winter death toll in 2017/18 which averaged 3,854 over a 13 week 'winter' period totalling 50,100 EWD.
The highest EWDs recorded was in 75/76 with 58,100 or an average of 4,469 weekly deaths (not deaths recorded simply because a respiratory disease is 'mentioned' on death certificate as).

The comparison in actual deaths by underlying cause is stark a death WITH and a death FROM are two totally different things, previously only a death as underlying cause is mentioned with regards to respiratory disease deaths on the register of deaths, a death WITH or mentioned on a death certificate is not recorded.
See screen shot from 2018 register with the respiratory deaths as underlying cause, no deaths counted because a death certificate mentioned a respiratory disease, same as 2019 and the same up until C.19 was a thing.

Also there is no mention how many of the 1,776 deaths are solely attributed to pneumonia, influenza, Covid19 or any of the other respiratory ailments which is in line with previous years, how many of the 1,776 are solely FROM Covid19?

Of the deaths mentioned on the weekly register, ONS state that where influenza, pneumonia or C.19 are included they are all included in the C.19 section only, see the screen grab from the ONS page. The number of deaths with a mention of C.19, influenza or pneumonia will of course with greater testing show a massively high number due to huge testing for such in the deceased.

This slants the figures because as mentioned by many doctors, scientists and experts, testing for influenza in people dying with significant underlying conditions would not be done in anywhere near the number as with the COV2 test, so of course this massively increases positive tests in deceased persons, that's before you even get to the fact that as per ONS statement all pneumonia and influenza deaths are being counted in the C.19 section only.

2 Attachments

See my post with regards actual deaths by underlying cause in the respiratory section, the increases in deaths are not FROM COVID19, or in fact any respiratory death, the ONS chart showing the register of deaths proves this conclusively.

Is the effects of lockdown having a direct influence on people dying, fewer staff, people suffering more strokes in the home/residential/care homes and not being seen to and dying, more vulnerable persons avoiding hospital.
The deaths are because of something and that something is NOT C.19, in fact the respiratory deaths (as underlying cause) as of 17th April are lower than they were for Week 2 of January by almost a third.

Weekly deaths England & Wales over the last 50 years *per capita for population growth

To spell it out, 2020 Excess deaths are not significant, its excess deaths for week 15 to 20 in 2020 that show significance.

Edit - either you edited to clarify or I stopped reading before finishing your post.

Assuming a gaussian distribution a standard Alpha of 0.05

Even if you wanted to remove the outlier in 1970, we haven't deviated from the mean for the year as a whole.