Donate to Remove ads

Got a credit card? use our Credit Card & Finance Calculators

Thanks to brightncheerful,Dormouse,Silverstar64,dave559,OLTB, for Donating to support the site

Coronavirus - Modelling Aspects Only

The home for all non-political Coronavirus (Covid-19) discussions on The Lemon Fool
Forum rules
This is the home for all non-political Coronavirus (Covid-19) discussions on The Lemon Fool
dspp
Lemon Half
Posts: 6803
Joined: November 4th, 2016, 10:53 am
Has thanked: 5678 times
Been thanked: 2037 times

Coronavirus - Modelling Aspects Only

#297303

Postby dspp » April 3rd, 2020, 11:20 am

Coronavirus - modelling aspects only please

Midsmartin
2 Lemon pips
Posts: 228
Joined: November 4th, 2016, 7:18 am
Has thanked: 29 times
Been thanked: 117 times

Re: Coronavirus - Modelling Aspects Only

#297311

Postby Midsmartin » April 3rd, 2020, 11:35 am

It's hard to keep all these threads separate! Modelling I think throws a lot of light on the likely economic and investment outcomes too.

Anyway, to start with, here's a freely available article in Nature:
"Special report: The simulations driving the world’s response to COVID-19"
https://www.nature.com/articles/d41586-020-01003-6


And this referenced in the above:
https://www.medrxiv.org/content/10.1101 ... 20042291v1

For the future, finding out how many are asymptomatic but infected is vital.

zico
Lemon Quarter
Posts: 2115
Joined: November 4th, 2016, 12:12 pm
Has thanked: 581 times
Been thanked: 589 times

Re: Coronavirus - Modelling Aspects Only

#297374

Postby zico » April 3rd, 2020, 1:44 pm

Thanks for opening this up.
I'd posted on Health & Well-being,
viewtopic.php?f=86&t=22298&p=297368#p297368

I'll put future posts on here, and below is pretty much what I posted on the Health board, to save people looking at the link.

Daily hospital death figures from the Dept of Health are quite lumpy. By now the numbers are large enough that there should be a smooth curve, but government is getting permission from next of kin before adding the numbers to the statistics. Office for National Statistics published a weekly total of all deaths from Covid-19, showing there's an extra 20% non-hospital death figures.

Assuming there's an average 10 day lag between new cases & deaths (which looks reasonable for the figures) then there would be 20,000 deaths in 10 days time, even if there were no further new cases starting tomorrow (obviously an extremely optimistic assumption). I hope I'm wrong but it looks pretty bad.

servodude wrote:

Is that assuming that around half of those that test positive (with the current testing/sampling methodology) die?

- sd



My initial thought was "no, not at all. I'm simply assuming the current 10-day lagged ratio of new cases to deaths remains constant."
But I went back and checked my model, and it's showing a horrifying death rate for new cases of 54%.

Sense-checking this, there were 6,700 total confirmed cases on 22nd March, and we've had 2,900 deaths in total, which would be a 43% death rate even if you assume that no deaths at all have occured from confirmed cases since 22nd March.
(It seems that reported "new cases" are actually hospitalised cases, and the above death rate refers to infected people who are in the 2nd stage of the virus, where it becomes much more serious and hospitalisation is required.)

Reuters are reporting a Govt source says the UK government has a worst case death toll of 50,000 if social distancing was only 50% adhered to, but they are saying a "good" outcome would be under 20,000. (Obviously "good" is a relative term for a disaster like this.) Source also expects Easter Sunday (April 12th) to be peak day for deaths.

My assumption of a 10-day lag is based on taking a weekly average of new cases % increase (16th-22nd March) and weekly average of death % increase (26th March-1st April). These figures are 24% and 26.5% respectively, showing a reasonable fit. To estimate deaths in future days, I've plugged in %new cases increase from 10 days earlier.

Results -
Adding in 20% from care homes, it seems inevitable that 17,000 deaths will occur based on existing 2nd stage infection cases. Assuming the latest reported %increase in new cases (14.6%) is replicated tomorrow would give 20,000 deaths by Easter Sunday.
To make that the peak day for deaths, I've assumed the %increase in new cases goes down tomorrow to 10%, then 8% the next day, 6% the day after, then 4%, 2% which would give 27,000 deaths by 17th April.

This assumes the social distancing measures start to really kick in from tomorrow (in affecting new cases) and show a dramatic improvement.
It's reasonable to assume social distancing measures should kick in any day now giving an immediate reduction in % increase of new cases.
I've been optimistic in my assumption of how quickly and how sharply social distancing will affect infection growth rates, but more pessimistic assumptions don't change the total deaths by more than about 7,000.
The key driver in the number of deaths is the exponential growth rate. Compared to the actual lockdown date, a lockdown 3 days earlier would have halved the deaths, lockdown 3 days later would have doubled deaths.
I've also assumed the NHS doesn't get overwhelmed by the new cases, and death rate remains unchanged. If this isn't the case, there could be in the region of 5,000 extra deaths.

However, we need to remember this is just the initial stage of the virus. After this first wave, the virus will still be present, and loosening of the social distancing measures will see further waves of new infections and deaths.

I'm afraid this is a very bleak picture, but I'm simply trying to make sense of the numbers, and give some estimate of what may happen, in a dispassionate way.
Social distancing needs to start working very soon, be very effective and also to be very sustained if we are to have any hope of keeping the death total below 50,000 in this first phase.
Today's 2pm figures will be absolutely crucial - we desperately need to see the % increase in new cases reduce from yesterday's 14.6% to below 10% either today or tomorrow.

Please note, there is absolutely no praise/criticism of the government, NHS or public behaviour in my post, and I wouldn't want this thread to descend into any sort of political point-scoring (there's also Polite Discussions for that).
I'd be very interested to see anyone's alternative methodologies and thoughts on this.


Just to clarify, my 54% death rate estimate is for Govt-registered new cases, which appear to be people in the more serious 2nd stage of the virus.
The death rate for the general population will be massively lower.
Having said that, if we assume that 25% of the general population (17 million) have already contracted the disease, then even 20,000 deaths would be a 0.13% death rate amongst the general population.

Assuming 5 million people already have the disease gives a 0.4% overall fatality rate.
Assuming 1 million gives a 2% fatality rate.
This is pure estimation/guesswork, but it seems likely to me that between 1 to 5 million people have now already contracted the virus in the UK.

dspp
Lemon Half
Posts: 6803
Joined: November 4th, 2016, 10:53 am
Has thanked: 5678 times
Been thanked: 2037 times

Re: Coronavirus - Modelling Aspects Only

#297404

Postby dspp » April 3rd, 2020, 2:56 pm


SUGGESTION
***************
If you use imgur to load a jpg you can then insert the jpg url into an image file on here. That way we can all quickly see each other's COVID-19 spreadsheets and graphs.


Here is an example I did for TSLA that I monitor, so you can see how easy it makes communicating.

regards, dspp

Image

Sorcery
Lemon Quarter
Posts: 1105
Joined: November 4th, 2016, 6:38 pm
Has thanked: 35 times
Been thanked: 46 times

Re: Coronavirus - Modelling Aspects Only

#297559

Postby Sorcery » April 3rd, 2020, 10:11 pm

I must be lost if I find myself here. Is there a reason it's buried?
Yes Coronavirus is serious even if it just makes us over-react.
Zico I agree with you about the likely low death rate in the global population as a %. Just hope that those affected have had a good life.

Nimrod103
Lemon Quarter
Posts: 3388
Joined: November 4th, 2016, 6:10 pm
Has thanked: 22 times
Been thanked: 180 times

Re: Coronavirus - Modelling Aspects Only

#297668

Postby Nimrod103 » April 4th, 2020, 11:58 am

A couple of newspaper articles today were interesting:

https://www.theguardian.com/world/2020/ ... l-is-wrong

So the death toll is rising strongly, but the lumpiness is probably due to undercounting in the past and maybe even overcounting now(?). The histogram in this article is interesting because quite clearly the low death reports on 28/29th March were because it was a weekend.

ttps://www.telegraph.co.uk/news/2020/04 ... val-covid/

Oxford and Imperial say they respect each others work, but either one of them is wrong, or maybe both are wrong.

zico
Lemon Quarter
Posts: 2115
Joined: November 4th, 2016, 12:12 pm
Has thanked: 581 times
Been thanked: 589 times

Re: Coronavirus - Modelling Aspects Only

#297675

Postby zico » April 4th, 2020, 12:13 pm

Nimrod103 wrote:A couple of newspaper articles today were interesting:

https://www.theguardian.com/world/2020/ ... l-is-wrong

So the death toll is rising strongly, but the lumpiness is probably due to undercounting in the past and maybe even overcounting now(?). The histogram in this article is interesting because quite clearly the low death reports on 28/29th March were because it was a weekend.

ttps://www.telegraph.co.uk/news/2020/04 ... val-covid/

Oxford and Imperial say they respect each others work, but either one of them is wrong, or maybe both are wrong.


Thanks for posting this. My model adds 20% (Source : ONS weekly stats) to allow for daily deaths only being hospital deaths.
I haven't allowed for the other point from the Guardian in my modelling about effect of delaying reporting (which is likely to have the effect of making the death rate look better than it actually is). As they rightly state, delaying the reporting of deaths has the effect of making the trend look better than it actually is. It's been clear for a week or two now that the daily deaths figure isn't purely hospital deaths in last 24 hours, because there has been a clearly increasing trend in new cases which mathematically must lead to a rise in daily deaths.

Could you please repost the telegraph link - I can't access it.

PinkDalek
Lemon Half
Posts: 6037
Joined: November 4th, 2016, 1:12 pm
Has thanked: 1586 times
Been thanked: 1735 times

Re: Coronavirus - Modelling Aspects Only

#297678

Postby PinkDalek » April 4th, 2020, 12:22 pm

Adding the leading 'h':

Science clash: Imperial vs Oxford, and the sex smear that created rival Covid-19 studies
https://www.telegraph.co.uk/news/2020/04/04/science-clash-imperial-vs-oxford-sex-smear-created-rival-covid/

You’ve reached your free article limit...

Nimrod103
Lemon Quarter
Posts: 3388
Joined: November 4th, 2016, 6:10 pm
Has thanked: 22 times
Been thanked: 180 times

Re: Coronavirus - Modelling Aspects Only

#297692

Postby Nimrod103 » April 4th, 2020, 12:58 pm

zico wrote:Could you please repost the telegraph link - I can't access it.


I'm not sure how to link to articles behind a paywall, but here is the actual headline and sub:

Science clash: Imperial vs Oxford, and the sex smear that created rival Covid-19 studies
Rivalry began two decades ago when leading lights of Imperial College and Oxford University worked together

It is more of a university politics article with no explanation of the science and maths, but explains how the former head of the Imperial team (Prof Sir Roy Anderson) was at Oxford and fell out with Sunetra Gupta over her appointment 20 years ago. He was then forced to leave, and moved to Imperial. Now he is doing other things, while his former understudy at Imperial, Ferguson, now gets on OK with Gupta - at least professionally.

PS sorry I left the h off of https

zico
Lemon Quarter
Posts: 2115
Joined: November 4th, 2016, 12:12 pm
Has thanked: 581 times
Been thanked: 589 times

Re: Coronavirus - Modelling Aspects Only

#297702

Postby zico » April 4th, 2020, 1:29 pm

I'm trying to download a couple of charts I uploaded to Imgur - let's see if it works.

https://imgur.com/a/RQe7Jpp

I've modelled 2 scenarios. Scenario 1 assumes we've reached peak daily new cases and there will be a gradual decline. Scenario 2 assumes a much faster decline in new cases.

Note : both scenarios as for hospitalised deaths only (total deaths will be around 20% higher).

In the first, I've assumed future % increases in daily new cases are 10% (today), 9% (tomorrow), 8%, 7%, etc down to 1%, then assumed 0.1% thereafter.
In the second, I've assumed future % increases in daily new cases are 10% (today), 5% (tomorrow), 3%, 1% and 0.1% thereafter.

Data up to 2nd April are actuals (Note - I'm treating each day's figures as referring to previous day - so yesterday's data released on 3rd April refers to 2nd April). I'm reasonably confident about the cumulative deaths curve up until 12th April, but after that date, it's based on assumptions about new cases.

I think Scenario 1 is reasonably optimistic but possible - it implies we've already reached our peak daily cases, and future daily cases should decrease to less than 3,000 within the next 7 days.

For Scenario 2 to work, Sunday's increase in new cases would need to drop to 2,100 then Monday's increase would be 1,323 and Tuesday's would be 454, and after that, 46 new cases per day.

Hope you can see the graphs and my explanations make some kind of sense! Happy to explain further if they don't.

Edited once - to correct spelling mistake.

johnhemming
Lemon Quarter
Posts: 4665
Joined: November 8th, 2016, 7:13 pm
Has thanked: 10 times
Been thanked: 606 times

Re: Coronavirus - Modelling Aspects Only

#297707

Postby johnhemming » April 4th, 2020, 1:47 pm

Another information problem is that there is a varying delay between date of death and date of reporting of death. This varies in a number of ways so that it is relatively difficult to track the daily rate of deaths with precision for perhaps 5 days. In the end one would assume the government have figures as to admissions with breathing problems (not all of which have Covid 19, not all of those will die), that is a useful statistic from which to see how things are going.

zico
Lemon Quarter
Posts: 2115
Joined: November 4th, 2016, 12:12 pm
Has thanked: 581 times
Been thanked: 589 times

Re: Coronavirus - Modelling Aspects Only

#297710

Postby zico » April 4th, 2020, 1:56 pm

johnhemming wrote:Another information problem is that there is a varying delay between date of death and date of reporting of death. This varies in a number of ways so that it is relatively difficult to track the daily rate of deaths with precision for perhaps 5 days. In the end one would assume the government have figures as to admissions with breathing problems (not all of which have Covid 19, not all of those will die), that is a useful statistic from which to see how things are going.


You're quite right. To mitigate this issue, my model uses % increase of cumulative new cases & deaths, so as to remove the fluctuations in daily reporting. The %increase in new cases is driving the model, and these should be more accurate, although there have still been days when the daily new case figure was lower than the previous day, which is statistically unexpected at a time when the numbers are doubling every 3 days.

zico
Lemon Quarter
Posts: 2115
Joined: November 4th, 2016, 12:12 pm
Has thanked: 581 times
Been thanked: 589 times

Re: Coronavirus - Modelling Aspects Only

#297716

Postby zico » April 4th, 2020, 2:07 pm

Figures from Intensive Care units below. Based on China figures, Intensive Care cases account for about 1/3rd of hospital admissions, so hopefully my estimated death rate is higher than the true figure.

The death rate of those admitted to intensive care with Covid-19 has topped 50%, according to the latest figures.

The figures come from data compiled by the Intensive Care National Audit and Research Centre (ICNARC) based on a sample of 2,249 coronavirus patients.

The data showed that 50.1% (346) of the 690 patients in the sample whose care outcomes were known, had died. The other 344 had been discharged.

The remaining patients, 1,559, were reported still to be in critical care.

zico
Lemon Quarter
Posts: 2115
Joined: November 4th, 2016, 12:12 pm
Has thanked: 581 times
Been thanked: 589 times

Re: Coronavirus - Modelling Aspects Only

#297721

Postby zico » April 4th, 2020, 2:17 pm

Updated figures for today are
41,903 positive cases and 4,313 deaths.
Daily figures are 3,735 new cases and 708 deaths.

The important figure is the new cases, which is down 17% on the previous few days total.
It's a daily increase of 9.8% (I used 10% in my model, so no need to update it).

This is welcome news, and hopefully tomorrow will bring another large percentage fall in new cases.

fca2019
2 Lemon pips
Posts: 222
Joined: July 18th, 2019, 8:37 am
Has thanked: 180 times
Been thanked: 64 times

Re: Coronavirus - Modelling Aspects Only

#297726

Postby fca2019 » April 4th, 2020, 2:23 pm

I think 5 million as the real number, is more realistic than 1 million. I've had symptoms and self isolated, and finding out neighbours, extended family and work colleagues who've had symptoms and self isolated too. All with varying degrees of illness and none of whom have been tested. So suspect were on the curve to herd immunity regardless of govt policy.

dspp
Lemon Half
Posts: 6803
Joined: November 4th, 2016, 10:53 am
Has thanked: 5678 times
Been thanked: 2037 times

Re: Coronavirus - Modelling Aspects Only

#297730

Postby dspp » April 4th, 2020, 2:35 pm

zico wrote:I'm trying to download a couple of charts I uploaded to Imgur - let's see if it works.

https://imgur.com/a/RQe7Jpp

.


Zico,
You have not quite done it correctly. Use the Img button in the text editor to insert the img start and end codes. Between the codes paste your URL from imgur.

Also you have probably used the wrong imgur link. Use the second one down, the "Direct Link".

If you do it right you will the actual image to display here.

(I don't have mod rights on these boards so cannot edit your posts to fix this for you.)

Regarding your model output I think it depends greatly on a) how much unsymptomatic transmission is going on; and b) the extent to which weather may become a factor. To get at a) they need to be doing smallscale / wide extent random population testing, but I suspect they do not have sufficient spare testing capacity to do that in most locations yet.

regards,
dspp

dealtn
Lemon Quarter
Posts: 3152
Joined: November 21st, 2016, 4:26 pm
Has thanked: 95 times
Been thanked: 1033 times

Re: Coronavirus - Modelling Aspects Only

#297736

Postby dealtn » April 4th, 2020, 2:56 pm

zico wrote:
The important figure is the new cases, which is down 17% on the previous few days total.
It's a daily increase of 9.8% (I used 10% in my model, so no need to update it).



Confused. Which is it? Or are you referring to 2 things here?

What is down 17%, and what is an increase of 9.8%?

regards

zico
Lemon Quarter
Posts: 2115
Joined: November 4th, 2016, 12:12 pm
Has thanked: 581 times
Been thanked: 589 times

Re: Coronavirus - Modelling Aspects Only

#297738

Postby zico » April 4th, 2020, 3:00 pm

dealtn wrote:
zico wrote:
The important figure is the new cases, which is down 17% on the previous few days total.
It's a daily increase of 9.8% (I used 10% in my model, so no need to update it).



Confused. Which is it? Or are you referring to 2 things here?

What is down 17%, and what is an increase of 9.8%?

regards


Sorry, it is confusing. The daily new cases figure is down 17% (a very good thing) and the cumulative new cases figure has increased by 9.8%.

zico
Lemon Quarter
Posts: 2115
Joined: November 4th, 2016, 12:12 pm
Has thanked: 581 times
Been thanked: 589 times

Re: Coronavirus - Modelling Aspects Only

#297754

Postby zico » April 4th, 2020, 3:26 pm

Death rate is the key to this, and the available data are very confusing.

The European Centre for Disease Prevention and Control ECDPC), Stockholm, 2020 gives a death rate for hospitalised cases of 12% which should be a whole lot more accurate than what comes out of my simple model.
However, the UK stats show 5,000 confirmed cases by 20th March and 1,000 deaths 7 days later (27th March) which gives a death rate of at least 20%.
Applying the ECDPC death rate to the highest daily new cases figure (4,300) would give daily deaths of 630, but UK daily deaths are already higher than that, and we only reached the 4,300 figure 4 days ago, so these figures shouldn't yet be translating into new deaths.
One possible explanation is that UK new cases are actually more severe hospitalisation cases than were previously seen in Europe.
It's puzzling.

https://www.ecdc.europa.eu/sites/defaul ... VID-19.pdf

Extract

Among all cases:Hospitalisation occurred in 30% (13 122of 43438) ofcasesreported from 17countries (median country-specific estimate,interquartile range(IQR):24%,11-41%)Severe illness(requiring ICU and/or respiratory support) accounted for 2 179 of 49 282 (4%) cases from 16 countries(median, IQR:3%, 2-8%).Among hospitalised cases:Severe illness was reported in 15% (1 894of 12 961) of hospitalised cases from 15countries (median, IQR: 16%, 10-24%).Death o

dealtn
Lemon Quarter
Posts: 3152
Joined: November 21st, 2016, 4:26 pm
Has thanked: 95 times
Been thanked: 1033 times

Re: Coronavirus - Modelling Aspects Only

#297757

Postby dealtn » April 4th, 2020, 3:31 pm

zico wrote:
dealtn wrote:
zico wrote:
The important figure is the new cases, which is down 17% on the previous few days total.
It's a daily increase of 9.8% (I used 10% in my model, so no need to update it).



Confused. Which is it? Or are you referring to 2 things here?

What is down 17%, and what is an increase of 9.8%?

regards


Sorry, it is confusing. The daily new cases figure is down 17% (a very good thing) and the cumulative new cases figure has increased by 9.8%.


OK, great I can understand what you are saying now, and agree the drop in the daily new cases is what is important here (and a very good thing).


Return to “Coronavirus Discussions”

Who is online

Users browsing this forum: gryffron and 2 guests