Page 84 of 258

Re: COVID-19

Posted: Thu Mar 26, 2020 2:41 am
by Millennie Al
OneOffDave wrote:
Wed Mar 25, 2020 11:13 am
This has always been an interesting debate for as long as I've been an emergency planner. How many are "excess deaths" versus "early harvest". Most UK pandemic planning has always been around demand smoothing as opposed to eradication as it's perceived that total containment is effectively impossible.
There are no extra deaths. Life inevitably ends in death anyway. What matters is the effect and expense of treatment. Every one of us currently has a life-threatening condition of potential starvation, which needs the continuous treatment of frequent meals. The expense of this is so manageable that it is nothing more than a normal part of everyday life. We would not be very concerned about a disease that spread quickly and killed all who were untreated if the treatment was to drink a litre of water. Similarly, we would not be very concerned about a disease that killed those infected if it were very difficult to catch (such as rabies). The ones that are really dangerous are those which are easy to catch from another person and which have require expensive treatment - like Covid-19, which can easily overwhelm a health service by exhausting the supply of things like ventilators. (Note that I mean expensive both in money and other resources such as time - we could provide everyone with their own personal ventilator if we only had enough time to build them).

Re: COVID-19

Posted: Thu Mar 26, 2020 7:41 am
by greyspoke
dyqik wrote:
Thu Mar 26, 2020 1:06 am
greyspoke wrote:
Wed Mar 25, 2020 3:39 pm
Well that Prof Ferguson has a slightly optimistic prediction.

But is getting some flack on twitter for not open-sourcing his code - "I wrote the code (thousands of lines of undocumented C) 13+ years ago to model flu pandemics..."

I assume that his publications on the topic gave the underlying algorithms but not the actual code.
Any publications based on undocumented and unreviewed code without unit tests should probably be retracted as unrepeatable and unverifiable. You might as well use instruments with no calibration or traceability to standards.
As a matter of interest, what exactly is a "unit test"?

Re: COVID-19

Posted: Thu Mar 26, 2020 7:47 am
by basementer
greyspoke wrote:
Thu Mar 26, 2020 7:41 am
dyqik wrote:
Thu Mar 26, 2020 1:06 am
Any publications based on undocumented and unreviewed code without unit tests should probably be retracted as unrepeatable and unverifiable. You might as well use instruments with no calibration or traceability to standards.
As a matter of interest, what exactly is a "unit test"?
A test of an individual component of a system, rather than the system as a whole. Unit testing, obviously, comes before system testing.

Re: COVID-19

Posted: Thu Mar 26, 2020 8:08 am
by greyspoke
So the results of tests of the elements of the code should be published? Which means at least the structure of the elements needs to be published otherwise the test outcomes will not mean that much?

Re: COVID-19

Posted: Thu Mar 26, 2020 8:19 am
by OneOffDave
Millennie Al wrote:
Thu Mar 26, 2020 2:41 am
There are no extra deaths. Life inevitably ends in death anyway. What matters is the effect and expense of treatment. Every one of us currently has a life-threatening condition of potential starvation, which needs the continuous treatment of frequent meals. The expense of this is so manageable that it is nothing more than a normal part of everyday life. We would not be very concerned about a disease that spread quickly and killed all who were untreated if the treatment was to drink a litre of water. Similarly, we would not be very concerned about a disease that killed those infected if it were very difficult to catch (such as rabies). The ones that are really dangerous are those which are easy to catch from another person and which have require expensive treatment - like Covid-19, which can easily overwhelm a health service by exhausting the supply of things like ventilators. (Note that I mean expensive both in money and other resources such as time - we could provide everyone with their own personal ventilator if we only had enough time to build them).
On a population basis, excess deaths are those over and above what you'd expect in a given period using actuarial methods. Something out of the usual run of things produces excess deaths. Some of those will be early harvest where the individual was likely to die within a year. It's about planning for how many people you lose from the economy and how many dead people you need to manage as the capacity is based on 'normal' years

Re: COVID-19

Posted: Thu Mar 26, 2020 8:25 am
by OneOffDave
Seeing more calls for the "highest level possible PPE" for NHS staff. Did a few quick thumbnail calculations (no of hospitals x itu staff x three shifts) and it works out at around 160,000 suits a day. A Respirex PRPS suit costs about £3k so we'd be up to about £47m a day. This assumes that we can find enough trained staff. Three months means over 14 million suits which would be difficult to manufacture

Re: COVID-19

Posted: Thu Mar 26, 2020 9:26 am
by jimbob
jimbob wrote:
Wed Mar 25, 2020 8:03 pm
lpm wrote:
Wed Mar 25, 2020 7:05 pm
I still don't get why anyone is taking the 50% thing remotely seriously.

50% implies about 32 million now. At a couple of doublings a week, that's

8 million 18 March
2 million 11 March
500,000 4 March
125,000 26 Feb

These vast numbers a few weeks ago never showed up as positive in testing, implying testing was a dud. On 26 Feb there were 12 confirmed cases from 125,000 actual infections? 0.01%? Absurd. By 4 March we had tested 16,000 people but only managed to find 87 cases. Out of 500,000 infected people?
Yup
HMG Website wrote:As of 9am on 24 March 2020, a total of 90,436 people have been tested, of which 82,359 were confirmed negative and 8,077 were confirmed positive. 422 patients in the UK who tested positive for coronavirus (COVID-19) have died.
That suggests that of those suspected of being most likely to have COVID-19 about 90% have been negative. About 12% of the positive tests were today.

The Oxford figures seem very difficult to fit to any data.
And now
HMG wrote:Number of cases
As of 9am on 25 March 2020, a total of 97,019 people have been tested, of which 87,490 were confirmed negative and 9,529 were confirmed positive. 463 patients in the UK who tested positive for coronavirus (COVID-19) have died
So by my maths, I make it 6583 tests yesterday and 1452 positive or 22% of those most likely to be infected

Re: COVID-19

Posted: Thu Mar 26, 2020 9:30 am
by greyspoke
Woodchopper wrote:
Wed Mar 25, 2020 12:56 pm
lpm wrote:
Wed Mar 25, 2020 12:30 pm
...

For example, if half the population on average is infected, you wouldn't expect one county to be at 75% and the neighbour at 25%. 50% would surely result in averaging out of peaks and troughs.

For example, Kent, Surrey and Hampshire. Why would they diverge from each other in a 50% scenario? Why would known cases be different?

Here's the current cases per million population:

Kent 45
Surrey 110
Hampshire 207
I agree in general, but there would be hot spots for mortality and people needing hospital treatment. But those hotspots would be correlated with the presence of at risk populations. For example, towns or villages with relatively large populations of retirees. ETA which isn't what we are seeing - eg a hotspot in New York which has a comparatively young population.
And indeed there is a hotspot in Gwent, though neighbouring areas share most of its demographic features.

Re: COVID-19

Posted: Thu Mar 26, 2020 10:25 am
by El Pollo Diablo
Death numbers properly low again yesterday. Not a full 24 hour period, but it does look like, deathwise, it's slowing down here. The next few days will confirm. We've gone from doubling every ~1.9 days to doubling every ~2.7 days. Last week we looked like we'd hit 1,000 deaths at this point, and we have less than half of that. Two weeks ago yesterday Italy had 827 deaths, we reported 463 last night. On Saturday we were tracking identically to them, 14 days later. 463 deaths is what Italy had 17 days ago.

Re: COVID-19

Posted: Thu Mar 26, 2020 10:28 am
by snoozeofreason
jimbob wrote:
Wed Mar 25, 2020 8:03 pm
HMG Website wrote:As of 9am on 24 March 2020, a total of 90,436 people have been tested, of which 82,359 were confirmed negative and 8,077 were confirmed positive. 422 patients in the UK who tested positive for coronavirus (COVID-19) have died.
That's interesting. Do you have a link to the original source (on Public Health England?). Googling it just throws up lots of secondary sources.

Re: COVID-19

Posted: Thu Mar 26, 2020 10:29 am
by El Pollo Diablo
snoozeofreason wrote:
Thu Mar 26, 2020 10:28 am
jimbob wrote:
Wed Mar 25, 2020 8:03 pm
HMG Website wrote:As of 9am on 24 March 2020, a total of 90,436 people have been tested, of which 82,359 were confirmed negative and 8,077 were confirmed positive. 422 patients in the UK who tested positive for coronavirus (COVID-19) have died.
That's interesting. Do you have a link to the original source (on Public Health England?). Googling it just throws up lots of secondary sources.
DHSC twitter. I've been following their reports and uploading the data to a google doc (as has every other bored analyst in the country):

https://docs.google.com/spreadsheets/d/ ... sp=sharing

Re: COVID-19

Posted: Thu Mar 26, 2020 10:34 am
by El Pollo Diablo
But wait. I mean. Actually amazed.

https://twitter.com/ToryFibs/status/1243095749686067202

Apparently they need the permission of relatives to release the figures for people who have died of Covid-19. Why, for f.cks sake? This is a public health emergency.

Re: COVID-19

Posted: Thu Mar 26, 2020 10:40 am
by snoozeofreason
El Pollo Diablo wrote:
Thu Mar 26, 2020 10:29 am
DHSC twitter. I've been following their reports and uploading the data to a google doc (as has every other bored analyst in the country):
https://docs.google.com/spreadsheets/d/ ... sp=sharing
Thanks muchly. I am looking for reasons to be calm at the moment, and that statistic suggests that, of people with symptoms of Covid, only a minority actually have Covid (at least I hope it does). That sounds like a reason to be a bit calmer.

Re: COVID-19

Posted: Thu Mar 26, 2020 10:53 am
by lpm
El Pollo Diablo wrote:
Thu Mar 26, 2020 10:34 am
But wait. I mean. Actually amazed.

https://twitter.com/ToryFibs/status/1243095749686067202

Apparently they need the permission of relatives to release the figures for people who have died of Covid-19. Why, for f.cks sake? This is a public health emergency.
That can't be true. There's absolutely no need for consent for release of any health statistics.

The published figures will be real, apart from a bit of lag and mismatched timings. There has been a genuine slowdown over the last 3-4 days. We were on track for 1,000 deaths at the end of today, that now looks 2-3 days delayed, making us 16-17 days behind Italy.

Slowdown potentially due to quicker processes at hotspot hospitals - they have got good at identifying and rushing critical cases into ICUs? Learning how to respond? There's been no overload that can't be handled by nearby hospitals, whereas Italy had far worse geography.

Re: COVID-19

Posted: Thu Mar 26, 2020 11:01 am
by OneOffDave
Daily figures can be found at GOV.UK

Re: COVID-19

Posted: Thu Mar 26, 2020 11:02 am
by OneOffDave
El Pollo Diablo wrote:
Thu Mar 26, 2020 10:34 am
But wait. I mean. Actually amazed.

https://twitter.com/ToryFibs/status/1243095749686067202

Apparently they need the permission of relatives to release the figures for people who have died of Covid-19. Why, for f.cks sake? This is a public health emergency.
I have to admit I'm a bit sceptical about this claim. I've not heard anything about that

Re: COVID-19

Posted: Thu Mar 26, 2020 11:05 am
by El Pollo Diablo
OneOffDave wrote:
Thu Mar 26, 2020 11:02 am
El Pollo Diablo wrote:
Thu Mar 26, 2020 10:34 am
But wait. I mean. Actually amazed.

https://twitter.com/ToryFibs/status/1243095749686067202

Apparently they need the permission of relatives to release the figures for people who have died of Covid-19. Why, for f.cks sake? This is a public health emergency.
I have to admit I'm a bit sceptical about this claim. I've not heard anything about that
Yeah, it makes so little sense that it's making me trust the Government, which is no mean feat.

Re: COVID-19

Posted: Thu Mar 26, 2020 11:09 am
by Woodchopper
El Pollo Diablo wrote:
Thu Mar 26, 2020 10:34 am
But wait. I mean. Actually amazed.

https://twitter.com/ToryFibs/status/1243095749686067202

Apparently they need the permission of relatives to release the figures for people who have died of Covid-19. Why, for f.cks sake? This is a public health emergency.
Anyone can apply to obtain a copy of a death certificate for a specific individual. So the cause of death isn’t inherently confidential.

It’s possible that they are trying to give family members some control over when information is made public. In some areas there are still very few confirmed cases to it might be possible to guess who had died.

Re: COVID-19

Posted: Thu Mar 26, 2020 11:39 am
by dccarm
OneOffDave wrote:
Thu Mar 26, 2020 8:25 am
Seeing more calls for the "highest level possible PPE" for NHS staff. Did a few quick thumbnail calculations (no of hospitals x itu staff x three shifts) and it works out at around 160,000 suits a day. A Respirex PRPS suit costs about £3k so we'd be up to about £47m a day. This assumes that we can find enough trained staff. Three months means over 14 million suits which would be difficult to manufacture
£47m a day is just short of £350m a week. We can spend our brexit dividend on PPE and still have a few quid to spare

Re: COVID-19

Posted: Thu Mar 26, 2020 12:27 pm
by bjn
greyspoke wrote:
Thu Mar 26, 2020 8:08 am
So the results of tests of the elements of the code should be published? Which means at least the structure of the elements needs to be published otherwise the test outcomes will not mean that much?
Unittests typically work by directly exercising the code in question and seeing that the results are as expected. So publishing the unit tests in isolation and the results of the test runs would be useful-ish. You can at least see that 1+1 really does result in 2. Some folk may be reluctant to do that as they may not have unit tests; the tests are too integrated into their code so they can't publish the tests separately; the unit tests will show how their software works and they may want to keep that secret; they can't be arsed.

Ideally both tests and tested code should be published.

Re: COVID-19

Posted: Thu Mar 26, 2020 12:42 pm
by lpm
WTF?

They did change the way the did numbers yesterday. Something to do with reporting time.

Didn't tell anyone, unable to say what the change meant.
The spokesman confirmed that the way UK coronavirus deaths are recorded and made public is changing (see 11.07am), but he was unable to give details of how. He said Public Health England is moving to a different reporting time. Yesterday was “a cross-over day” in the way they were recording the numbers, he said. But he was unable to explain what would change.
If you want to change reporting times, fine. Just go back and adjust the historicals in the same way.

Re: COVID-19

Posted: Thu Mar 26, 2020 1:24 pm
by JellyandJackson
Something on time of recording of deaths from Chris Smyth of the Times.
https://twitter.com/Smyth_Chris/status/ ... 4106462208

Re: COVID-19

Posted: Thu Mar 26, 2020 1:27 pm
by EACLucifer
El Pollo Diablo wrote:
Thu Mar 26, 2020 10:34 am
But wait. I mean. Actually amazed.

https://twitter.com/ToryFibs/status/1243095749686067202

Apparently they need the permission of relatives to release the figures for people who have died of Covid-19. Why, for f.cks sake? This is a public health emergency.
We are in a public health emergency. Johnson's government still seem to think it's a PR issue.

Re: COVID-19

Posted: Thu Mar 26, 2020 1:37 pm
by jimbob
lpm wrote:
Thu Mar 26, 2020 12:42 pm
WTF?

They did change the way the did numbers yesterday. Something to do with reporting time.

Didn't tell anyone, unable to say what the change meant.
The spokesman confirmed that the way UK coronavirus deaths are recorded and made public is changing (see 11.07am), but he was unable to give details of how. He said Public Health England is moving to a different reporting time. Yesterday was “a cross-over day” in the way they were recording the numbers, he said. But he was unable to explain what would change.
If you want to change reporting times, fine. Just go back and adjust the historicals in the same way.
They really are making it difficult to easily check.

The websute from the NHS link only gives the larest numbers in detail for example.

Re: COVID-19

Posted: Thu Mar 26, 2020 1:43 pm
by greyspoke
I thought at first it couldn't be. But it looks like it really is the old "keep changing the accounting basis and nobody will be able to see how you are cooking the books" trick.