Scrutable

According to this Sky News story https://news.sky.com/story/coronavirus- ... d-11980194 the R rate has increased from 0.7 to 0.96 since they eased restrictions.

Brightonian wrote: Wed Apr 29, 2020 7:47 am Oxford vaccine trial in South London area:
https://www.covid19vaccinetrial.co.uk/p ... -georges-0

Thank you! Have registered and spread the word.

Blackcountryboy wrote: Wed Apr 29, 2020 4:11 pm According to this Sky News story https://news.sky.com/story/coronavirus- ... d-11980194 the R rate has increased from 0.7 to 0.96 since they eased restrictions.

How could they possibly know? Even with fantastic testing, there must be undiscovered pockets of infection?

lpm wrote: Wed Apr 29, 2020 5:41 pm

Blackcountryboy wrote: Wed Apr 29, 2020 4:11 pm According to this Sky News story https://news.sky.com/story/coronavirus- ... d-11980194 the R rate has increased from 0.7 to 0.96 since they eased restrictions.

How could they possibly know? Even with fantastic testing, there must be undiscovered pockets of infection?

You could probably proxy it with the rate of hospitalizations. At a very basic level assume that a constant percentage of those infected need hospital treatment. It would be possible to make it more complex by adjusting for age etc

The problem with hospitalizations is the lag between infection and needing treatment.

The daily situation report from the Robert Koch Institute states that R is 0.75 https://www.rki.de/DE/Content/InfAZ/N/N ... cationFile page 6
https://www.rki.de/DE/Content/InfAZ/N/N ... esamt.html

I'm not sure where Sky got their estimate from.

Methodology can be found here: https://www.rki.de/DE/Content/Infekt/Ep ... rt_02.html (in German)

Blackcountryboy wrote: Wed Apr 29, 2020 4:11 pm According to this Sky News story https://news.sky.com/story/coronavirus- ... d-11980194 the R rate has increased from 0.7 to 0.96 since they eased restrictions.

It looks like they corrected it, and it should be 0.75.
https://www.merkur.de/welt/corona-deuts ... 30761.html

Woodchopper wrote: Wed Apr 29, 2020 8:51 pm

Blackcountryboy wrote: Wed Apr 29, 2020 4:11 pm According to this Sky News story https://news.sky.com/story/coronavirus- ... d-11980194 the R rate has increased from 0.7 to 0.96 since they eased restrictions.

It looks like they corrected it, and it should be 0.75.
https://www.merkur.de/welt/corona-deuts ... 30761.html

Unfortunately, many people completely fail to understand exponentials. With R=0.7 for each person currently infected you will get 2.333... more infected until the disease dies out. But with R=0.75 (which seems like a very small change) you get 3 more infected. So that small change in R of 0.05 causes a 29% increase in infections (and, presumably, deaths).

From the article:

Lothar Wieler, head of Germany's disease control centre the Robert Koch Institute, stressed the importance of the reproduction rate.

"The number should stay below one, that is the big goal," he told a news conference.

That is a very bad mistake to make. As R gets closer to 1, smaller changes have bigger effects. For example, if it went from 0.95 to 0.96, that would mean an extra 5 people would get infected per initially infected person (24 instead of 19). And note also how that's hugely greater than the number infected at R=0.75. The big goal is not to keep R less than one, it's to get R as low as practical and keep it there until the disease is eradicated. This can be by lockdown or other means. Keeping R less than one is no more than keeping the barest minimum of control over the disease. Any value of R >= 1 means the same number of people get infected - the value only changes how fast this happens.

Millennie Al wrote: Thu Apr 30, 2020 3:31 am

Woodchopper wrote: Wed Apr 29, 2020 8:51 pm

Blackcountryboy wrote: Wed Apr 29, 2020 4:11 pm According to this Sky News story https://news.sky.com/story/coronavirus- ... d-11980194 the R rate has increased from 0.7 to 0.96 since they eased restrictions.

It looks like they corrected it, and it should be 0.75.
https://www.merkur.de/welt/corona-deuts ... 30761.html

Unfortunately, many people completely fail to understand exponentials. With R=0.7 for each person currently infected you will get 2.333... more infected until the disease dies out. But with R=0.75 (which seems like a very small change) you get 3 more infected. So that small change in R of 0.05 causes a 29% increase in infections (and, presumably, deaths).

What you're characterising there is not an exponential. It's the sum of an infinite geometric series, and it blows up as R nears 1 because there's a 1-R in the denominator - it's approaching division by zero. An even more acute problem than exponential growth.

Geometric growth and exponential growth are the same thing, just one is discrete and the other is continuous.

In other words, one is the sum of an infinite series and the other is the integral of the function. An integral is just a sum with infinitessimal intervals. You're right that even if R_0 < 1 the length of the decay tail makes a difference to the total number of cases, but it's still good if it's going down, and 0.75 is a lot better than 0.96.

But neither can it actually go to infinity in the real world of course; for R_0 > 1 the fraction of the population which gets infected* is something like 1-1/R_0, so even there there's a difference between R_0 slightly more than 1 and R_0 a lot bigger than 1.

* - something like this anyway; I might be thinking of the maximum fraction which gets infected at once, can't be bothered to look it up.

... and it's more complicated when R_0 changes because of various social measures etc.

... but we're still in the exponential growth phase in just about every country and/or region I think, except maybe in a very few localised hot spots.

Millennie Al wrote: Thu Apr 30, 2020 3:31 amThe big goal is not to keep R less than one, it's to get R as low as practical and keep it there until the disease is eradicated. This can be by lockdown or other means. Keeping R less than one is no more than keeping the barest minimum of control over the disease. Any value of R >= 1 means the same number of people get infected - the value only changes how fast this happens.

It appears that across Europe the policy is that the lowest practical value for R is just below 1. Which is why we are now seeing phased unlockdowns in countries that are still experiencing hundreds of deaths per day.

All of the patients who die of covid die therefore the CFR is 100%

Well ok that's not exactly what it says but you can't really say Covid-19 is as bad as Ebola either.

These results are still better than the roughly 50:50 rate we previously had for intensive care, right? I think the numbers will continue to improve a bit, because recovery takes time and intensivists are learning about the best ways to support patients. For example, intubation may not be so strongly indicated.

I read that this morning and thought WTF?

They’re comparing deaths in hospitalised cases for COVID-19 with the death rate in the general population for Ebola. Bonkers!

The death rate for all cases for Ebola is 25-90%, but for COVID-19 it’s 0.5-1%.

What the f.ck are they talking about?

Sure, if you are serious enough to be hospitalised (thus probably with co-morbidities etc), your chances of survival are on par with anyone getting Ebola (regardless of health before). But that’s apples and oranges.

headshot wrote: Thu Apr 30, 2020 10:50 am I read that this morning and thought WTF?

They’re comparing deaths in hospitalised cases for COVID-19 with the death rate in the general population for Ebola. Bonkers!

The death rate for all cases for Ebola is 25-90%, but for COVID-19 it’s 0.5-1%.

What the f.ck are they talking about?

Sure, if you are serious enough to be hospitalised (thus probably with co-morbidities etc), your chances of survival are on par with anyone getting Ebola (regardless of health before). But that’s apples and oranges.

It's atrocious journalism but it looks like it was Calum Semple, the chief investigator on the trial, who came up with comparison...

Coronavirus 'as deadly as Ebola in hospital'
https://www.bbc.co.uk/news/health-52473524

What the 33% mortality rate for hospitalised patients does suggest - is that in the UK the threshold for hospitalisation is quite high. Or our definition of "hospital" is more restrictive than other countries.

China reported a lower mortality rate in "hospitalised" patients - but they may have been taking in a much broader range of cases and I'm not sure whether some of the quarantine centres may have counted a "hospitals".

This is off a Facebook post so take it how you will but here's a graph of estimated daily excess deaths in Lombardy (red) compared to officially registered covid deaths (blue).
https://www.facebook.com/DatiAnalisiCoronavirus/

(the baseline death rate is about 300 per day, to put those values into perspective: the peak is over 1000 per day; the population of Lombardy is about 10 million)

This is the daily death rate in Bergamo (the whole province has a population of the order of a million) in 2020 and 2016.

Looking at those graphs reminds me...

Are there graphs of cases or deaths by local region - say by county - for the UK anywhere? I can see the current numbers on the government dashboard but not previous numbers, and nothing came up on a very quick google search. Surely someone is graphing them somewhere.

Also read this article in the Grauniad yesterday about 1918 influenza pandemic, which has cool handrawn graphs from the time, talks about how very similar social distancing measures were used and how even back then there were numpties protesting against them.

Even more interesting was a link BTL to this National Geographic article from back in 2014, post-avian flu and SARS, that postulated the 1918 'flu may have originated in China. I don't know if that's still current thinking, or if it was ever an opinion widely held though.

I read something recently that "Spanish" Flu actually originated in Kansas.

https://www.ncbi.nlm.nih.gov/pmc/articles/PMC340389/

raven wrote: Thu Apr 30, 2020 1:04 pm Looking at those graphs reminds me...

Are there graphs of cases or deaths by local region - say by county - for the UK anywhere? I can see the current numbers on the government dashboard but not previous numbers, and nothing came up on a very quick google search. Surely someone is graphing them somewhere.

There's data in this github https://github.com/tomwhite/covid-19-uk-data and there are also various python scripts for getting it out of the csv files.

Armed nuts protesting lockdown at the Michigan capitol

Jesus

https://twitter.com/shannonrwatts/statu ... 05792?s=21

https://twitter.com/vote4robgill/status ... 46947?s=21

shpalman wrote: Thu Apr 30, 2020 8:43 am But neither can it actually go to infinity in the real world of course; for R_0 > 1 the fraction of the population which gets infected* is something like 1-1/R_0, so even there there's a difference between R_0 slightly more than 1 and R_0 a lot bigger than 1.

Indeed. The exponential is only a very simple model. It's actually a logistic, which is very similar when the growth has not reached much of the population. Except, it's not even that, because people die and more get born, so the population is gradually renewed with newly vulnerable individuals. Except, it's not even that, because natural selection means the population's susceptibility gradually changes over time. And then there's the question of how long immunity lasts after you have had it.

Which is why I think it's best to have models which are so simple their defects are obvious than to have ones so complex the defects are hidden.

That last sentence is a zinger, Mal

Millennie Al wrote: Fri May 01, 2020 12:56 am

shpalman wrote: Thu Apr 30, 2020 8:43 am But neither can it actually go to infinity in the real world of course; for R_0 > 1 the fraction of the population which gets infected* is something like 1-1/R_0, so even there there's a difference between R_0 slightly more than 1 and R_0 a lot bigger than 1.

Indeed. The exponential is only a very simple model. It's actually a logistic, which is very similar when the growth has not reached much of the population. Except, it's not even that, because people die and more get born, so the population is gradually renewed with newly vulnerable individuals. Except, it's not even that, because natural selection means the population's susceptibility gradually changes over time. And then there's the question of how long immunity lasts after you have had it.

Those latter effects aren't strongly relevant in the case of a pandemic which has gone on for less than half a year so far. It's not even particularly logistic yet, since a substantial fraction of the population remains susceptible in all but the hottest of hot spots, unless you let the maximum value be a free parameter. Lots of people did try fitting logistic curves when the first few days of slowed infection rates came through after the lockdown, but that was the effect of the lockdown lengthening the exponential doubling time, and it only gave the misleading impression that it would saturate on its own. A few days later it was always clear that it was still an exponential growth.

Which is why I think it's best to have models which are so simple their defects are obvious than to have ones so complex the defects are hidden.

Yes, this.

To get R0 from the exponential rate requires an assumption of how many days a case is infection for (because the rate is new cases per day but R0 is new cases per case) but you can just look at the number of new cases per day and see for yourself if the trend is up or down.

The UK's figure of 21678 deaths a couple of days ago was similar to Italy's number on the 15th of April but it turns out that number should have been 25302 which is somewhere between Italy's numbers for the 22nd and 23rd of April.

The most recent tweet about it from the Department of Health and Social Care says that "as of 5pm on 29 April, of those tested positive for coronavirus across all settings, 26,771 have sadly died"

i.e. the number I have in my spreadsheet for yesterday is anyway from the evening of the day before, and it corresponds to the number released in Italy at 5pm on the 27 of April.

So the UK has gone from "four weeks" to two weeks to two days behind Italy.

At least yesterday a decent number of swabs were processed (more than 80,000 while Italy has been averaging about 57,000 for the past couple of weeks) but it predictably led to the highest number of new positives being recorded so far. Italy hasn't had that many new cases per day for over a month and it really makes you wonder how many cases the UK really had at the beginning of April when only about 10,000 swabs were being done per day but there were still between 3000 and 6000 new cases per day.

EuroMOMO Bulletin, Week 17 is here...

https://www.euromomo.eu/bulletins/2020-17/

What is particularly worrying is that England seems to have by far the largest increase in all-cause mortality in the 15-64 year age group...

https://www.euromomo.eu/graphs-and-maps/

What could be the explanation?

- Overall COVID-19 prevalence being way higher in England than other European countries?
- The NHS being less effective at treating COVID-19 than other health systems?
- 15-64 year olds with other treatable conditions avoiding medical care?

Any ideas?