Donate to Remove ads

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

Thanks to johnstevens77,Bhoddhisatva,scotia,Anonymous,Cornytiv34, 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
jfgw
Lemon Quarter
Posts: 2539
Joined: November 4th, 2016, 3:36 pm
Has thanked: 1097 times
Been thanked: 1146 times

Re: Coronavirus - Modelling Aspects Only

#426428

Postby jfgw » July 10th, 2021, 1:31 pm

Since cases are high in the North West, I thought I would look at the regional case data. Blackburn with Darwen, and Bolton look interesting.

Image

Here is the same but going back to the end of October 2020,

Image

Both graphs:
My graph. Contains public sector information licensed under the Open Government Licence v3.0. : https://www.nationalarchives.gov.uk/doc/open-government-licence/version/3/: https://coronavirus.data.gov.uk/details/cases?areaType=overview&areaName=United%20Kingdom
Population data are mid-2020 estimates from the ONS: https://www.ons.gov.uk/peoplepopulationandcommunity/populationandmigration/populationestimates/bulletins/annualmidyearpopulationestimates/mid2020


Julian F. G. W.

Hallucigenia
Lemon Quarter
Posts: 2618
Joined: November 5th, 2016, 3:03 am
Has thanked: 166 times
Been thanked: 1718 times

Re: Coronavirus - Modelling Aspects Only

#429400

Postby Hallucigenia » July 21st, 2021, 5:35 pm

1nvest wrote:Nowadays I'm more inclined to just look at the death rates and in seeing relatively low numbers it seems more like that Covid is just another cold. 1% of the population will die on average each year, so 700,000 type number, near 2000/day.


The number that is currently most worrying is the number on ventilation. If you translate that into the equivalent last autumn, 618 on ventilators is more than 17 October, a few days into the tiers. We're 0.62 doublings away from the 1,003 on ventilation on 5 November when lockdown2 was announced, and 2.7 doublings away from the all-time high of 24 January when intensive care was in meltdown, way over capacity. Recently cases have been doubling every 10-14 days, so you might expect ventilation to be similar, with a time lag.

If you're worried about non-Covid deaths then you should be terrified by the prospect of intensive care being saturated with Covid patients. Aside from the fact that the infection risk alone can close down chunks of an ICU to non-Covid patients, if it's at capacity then that means that serious operations like heart transplants and cancer surgery have to be postponed. The real problem is that they just stay in ICU for so long, 20 days on average whereas recovery from surgery might need a day or two in intensive care.

So once the ICU is at capacity, one Covid patient can lead to major operations being postponed for 10-20 other patients who are seriously ill, and there's a fair chance that will not end well for at least one of the non-Covid patients.

1nvest wrote:I suspect there are nowadays more suicides/day perhaps as a function of lockdowns/controls than Covid related deaths.


Citation needed.

https://gh.bmj.com/content/6/8/e006653
"Careful analysis of excess mortality suggests that lockdowns are not associated with large numbers of deaths in places that avoided large COVID-19 epidemics (eg, Australia, New Zealand)

This evidence must be weighed against the very severe harms caused by COVID-19 itself, as seen for example in Brazil and India

It is unlikely that government interventions have been worse than the pandemic itself in most situations using data collected to date
"

scotia
Lemon Quarter
Posts: 3561
Joined: November 4th, 2016, 8:43 pm
Has thanked: 2371 times
Been thanked: 1943 times

Re: Coronavirus - Modelling Aspects Only

#429768

Postby scotia » July 22nd, 2021, 9:15 pm

Two weeks ago I noted that the newest data on hospital admissions was now showing a steep climb. The current situation is

Image

At first sight it appears that the climb in admissions is continuing, however on fitting curves to the data, its gradient is reducing. An optimist would be looking for a turning point in the not too distant future - I'll wait for more data. Within statistical accuracy, the deaths curve seems to be following reasonably the admissions, time slipped by 8 days. I'll guess that the importation of the Delta variant has played a significant part in this new outbreak.

I have also been keeping a lookout on the Scottish data - but since the population of England is approximately 10 times that of Scotland, the statistical accuracy is much poorer. However there are some important features. It looks like the rise in cases has been more rapid in Scotland, and most have been centred on the cities. Going back two weeks, the 7-day admissions and deaths were around one fifth of those in England - whereas one tenth would have been expected on a population pro-rata basis. However about 10 days ago the (7-day) Admissions peaked, and over the most recent three days the (7-day) deaths seem to have levelled off to around 50 - albeit with a poor statistical accuracy.

If - and I should say that it is a very big IF - there has been a turnaround in Scotland, it might suggest that the effect of the Delta variant has passed its worst. And hopefully England will follow.

tjh290633
Lemon Half
Posts: 8209
Joined: November 4th, 2016, 11:20 am
Has thanked: 913 times
Been thanked: 4097 times

Re: Coronavirus - Modelling Aspects Only

#429780

Postby tjh290633 » July 22nd, 2021, 10:57 pm

It always strikes me that the function being measured should be the second derivative, i.e the change in slope of the curve.

If that goes negative then we are on a winning streak.

TJH

servodude
Lemon Half
Posts: 8271
Joined: November 8th, 2016, 5:56 am
Has thanked: 4435 times
Been thanked: 3564 times

Re: Coronavirus - Modelling Aspects Only

#429782

Postby servodude » July 22nd, 2021, 11:08 pm

tjh290633 wrote:It always strikes me that the function being measured should be the second derivative, i.e the change in slope of the curve.

If that goes negative then we are on a winning streak.

TJH


Agreed!
Isn't that why the epidemiology chap(esse)s bang on about the R number a lot?
- and it seems intuitively relevant given the virus spreads geometrically (albeit with a constantly varying gain given various factors)

From Scotia's graphs it looks like it would be possible to get something useable from the admissions
- the deaths figures though would look to be giving a few false dawns (perhaps widening the averaging window would help this at the cost of a bit of resolution)

- sd

9873210
Lemon Slice
Posts: 986
Joined: December 9th, 2016, 6:44 am
Has thanked: 226 times
Been thanked: 296 times

Re: Coronavirus - Modelling Aspects Only

#429808

Postby 9873210 » July 23rd, 2021, 3:07 am

tjh290633 wrote:It always strikes me that the function being measured should be the second derivative, i.e the change in slope of the curve.

If that goes negative then we are on a winning streak.

TJH


First, higher derivatives are hard to measure. Small amounts of noise in the data cause huge differences in estimates of the second derivative.

Second, d^2 Ax+ln(x) /dx^2 = -1/x^2. So you can have a negative second derivative and still be in for a world of hurt if A is large.

scotia
Lemon Quarter
Posts: 3561
Joined: November 4th, 2016, 8:43 pm
Has thanked: 2371 times
Been thanked: 1943 times

Re: Coronavirus - Modelling Aspects Only

#429883

Postby scotia » July 23rd, 2021, 11:16 am

9873210 wrote:First, higher derivatives are hard to measure. Small amounts of noise in the data cause huge differences in estimates of the second derivative.

Second, d^2 Ax+ln(x) /dx^2 = -1/x^2. So you can have a negative second derivative and still be in for a world of hurt if A is large.


Agreed! With small numbers and consequently significant standard deviations its a near impossible task.

scotia
Lemon Quarter
Posts: 3561
Joined: November 4th, 2016, 8:43 pm
Has thanked: 2371 times
Been thanked: 1943 times

Re: Coronavirus - Modelling Aspects Only

#430041

Postby scotia » July 24th, 2021, 12:14 am

servodude wrote:From Scotia's graphs it looks like it would be possible to get something useable from the admissions

I have tried analysing the admissions data over the past month in several different ways. The best I could say is that the rate of increase (i.e. the first derivative) appears to be falling. I'll wait for more data. I see that The UK new infection numbers appears to have peaked - but this may change with the lifting of restrictions on July 19th.

GrahamPlatt
Lemon Quarter
Posts: 2059
Joined: November 4th, 2016, 9:40 am
Has thanked: 1032 times
Been thanked: 823 times

Re: Coronavirus - Modelling Aspects Only

#430057

Postby GrahamPlatt » July 24th, 2021, 7:01 am


jfgw
Lemon Quarter
Posts: 2539
Joined: November 4th, 2016, 3:36 pm
Has thanked: 1097 times
Been thanked: 1146 times

Re: Coronavirus - Modelling Aspects Only

#430283

Postby jfgw » July 25th, 2021, 2:54 pm

Hospital admissions in England still rising in all areas,
Image

Scotland going down,
Image

15 to 24 age group cases are highest but have started to fall,
Image

First vaccine: 18–24s have caught up with 25–29s (England),
Image

Second vaccine (England),
Image

All graphs are mine. Contains public sector information licensed under the Open Government Licence v3.0. : https://www.nationalarchives.gov.uk/doc/open-government-licence/version/3/


Julian F. G. W.

Itsallaguess
Lemon Half
Posts: 9129
Joined: November 4th, 2016, 1:16 pm
Has thanked: 4140 times
Been thanked: 10023 times

Re: Coronavirus - Modelling Aspects Only

#432102

Postby Itsallaguess » August 2nd, 2021, 6:39 pm

Some welcoming news on hospital admissions from the BBC -

Clear signs of fall in England's hospital admissions -

There are now clear signs that the number of people being admitted into hospital with Covid has started falling in England following the drop in detected cases.

Infections have been falling for well over a week. But there is always a lag between a change in cases and hospitalisations as it tends to take a week to 10 days for someone who is infected to become ill enough to require hospital treatment.

On Saturday there were 593 hospitalisations - that followed 650 on Friday and 758 on Thursday. On the same days the previous week there were 734, 725 and 827 admissions respectively.


Image

Source - https://www.bbc.co.uk/news/live/uk-58053529

Cheers,

Itsallaguess

jfgw
Lemon Quarter
Posts: 2539
Joined: November 4th, 2016, 3:36 pm
Has thanked: 1097 times
Been thanked: 1146 times

Re: Coronavirus - Modelling Aspects Only

#432122

Postby jfgw » August 2nd, 2021, 9:06 pm

Similar graphs for admissions: peaked or seemingly about to,
Image
My graph. Contains public sector information licensed under the Open Government Licence v3.0. : https://www.nationalarchives.gov.uk/doc/open-government-licence/version/3/

Cases going up for the 15 to 24s, with 25 to 29s levelled off,
Image
My graph. Contains public sector information licensed under the Open Government Licence v3.0. : https://www.nationalarchives.gov.uk/doc/open-government-licence/version/3/


Julian F. G. W.

88V8
Lemon Half
Posts: 5769
Joined: November 4th, 2016, 11:22 am
Has thanked: 4098 times
Been thanked: 2560 times

Re: Coronavirus - Modelling Aspects Only

#432312

Postby 88V8 » August 3rd, 2021, 6:58 pm

Study modelling the potential for the rise of new variants:

Our model suggests three specific risk factors that favour the emergence and establishment of a vaccine-resistant strain that are intuitively obvious: high probability of initial emergence of the resistant strain, high number of infected individuals and low rate of vaccination. By contrast, a counterintuitive result of our analysis is that the highest risk of resistant strain establishment occurs when a large fraction of the population has already been vaccinated but the transmission is not controlled.

That would pretty well describe the situation here unless social distancing/mask wearing is maintained, and superspreader events are either not permitted or restricted to the double-jabbed.

https://www.nature.com/articles/s41598-021-95025-3

It is regrettable that the political capital rightly garnered by the vaccine rollout is now being squandered by the shambolic handling of travel restrictions. Inevitably this will distract the govt, which is still shilly-shallying about mandatory vacc of are staff and NHS staff, and keeps threatening vaccine passports without actually imposing them.

V8

servodude
Lemon Half
Posts: 8271
Joined: November 8th, 2016, 5:56 am
Has thanked: 4435 times
Been thanked: 3564 times

Re: Coronavirus - Modelling Aspects Only

#432370

Postby servodude » August 4th, 2021, 2:36 am

88V8 wrote:Study modelling the potential for the rise of new variants:

Our model suggests three specific risk factors that favour the emergence and establishment of a vaccine-resistant strain that are intuitively obvious: high probability of initial emergence of the resistant strain, high number of infected individuals and low rate of vaccination. By contrast, a counterintuitive result of our analysis is that the highest risk of resistant strain establishment occurs when a large fraction of the population has already been vaccinated but the transmission is not controlled.

That would pretty well describe the situation here unless social distancing/mask wearing is maintained, and superspreader events are either not permitted or restricted to the double-jabbed.

https://www.nature.com/articles/s41598-021-95025-3

It is regrettable that the political capital rightly garnered by the vaccine rollout is now being squandered by the shambolic handling of travel restrictions. Inevitably this will distract the govt, which is still shilly-shallying about mandatory vacc of are staff and NHS staff, and keeps threatening vaccine passports without actually imposing them.

V8


It's a big bold experiment alright

I'm not sure restricting events to double-jabbed would make much difference on the emergence of resistant strains; they're the cohort exerting the evolutionary pressure on the virus

If a resistant strain is going to arise it will probably do so in a vaccinated patient who becomes ill for a long time
It's that time in the patient that gives the population of virus (every 10hrs or so giving rise to a new generation) within them the opportunity to "solve the puzzle" of their immune defences
Seeing as this is the modelling thread in the parlance of Genetic Algorithms those patients provide a better (more generally applicable) cost function and more generations/cycles for the algorithm to run (1)

Which is fine if the patient then eventually beats it or dies trying; the danger though is if it escapes from that patient in to the wild (where it now has the advantages it has won over vaccination defences)
- and if "the wild" at that time looks like verdant pastures without any hindrances everything repeats again and we eventually run out of Greek letters

At this stage it would be wise to treated vaccinated patients who require hospitalisation with extra caution (but the stuff around reducing the spread as much as you can still holds)

The political stuff?....it has always been a distraction
- both in terms of taking effort from the work that matters
- and in keeping people convinced that "we're doing a good job" irrespective

Thanks for the link

-sd

1: Back in the dawn of the internet this was the GA library I used in my PhD work http://lancet.mit.edu/galib-2.4/ amazing what's still lying around out there - and props to the writers and maintainers - it saved me a lot of work

scotia
Lemon Quarter
Posts: 3561
Joined: November 4th, 2016, 8:43 pm
Has thanked: 2371 times
Been thanked: 1943 times

Re: Coronavirus - Modelling Aspects Only

#432797

Postby scotia » August 5th, 2021, 11:35 pm

Two weeks ago I noted the continuing rise in the (7 day) English hospital admissions, but suggested that the gradient of the climb was reducing. The results today shows a definite peak

Image

And the prediction is that the (7-day) deaths have now probably reached a peak - but more data is required.
Note that I have tweaked the parameters slightly to fit the more recent data.

The Scottish data has a much poorer statistical accuracy, due to the smaller numbers, and the hospital admissions are reported in a different fashion - usually two days behind the English data. However as previously reported, the Scottish 7-day admissions peaked around 13th July, and I can now report that the 7 day deaths peaked around 25th July. However after falling for a week, the deaths are now showing a slight upward trend, but this is not mirrored in the admissions data. It could simply be an effect of the poor statistical accuracy - I'll wait for more data before ringing alarm bells.

servodude
Lemon Half
Posts: 8271
Joined: November 8th, 2016, 5:56 am
Has thanked: 4435 times
Been thanked: 3564 times

Re: Coronavirus - Modelling Aspects Only

#432802

Postby servodude » August 6th, 2021, 12:28 am

scotia wrote:The Scottish data has a much poorer statistical accuracy, due to the smaller numbers, and the hospital admissions are reported in a different fashion - usually two days behind the English data. However as previously reported, the Scottish 7-day admissions peaked around 13th July, and I can now report that the 7 day deaths peaked around 25th July. However after falling for a week, the deaths are now showing a slight upward trend, but this is not mirrored in the admissions data. It could simply be an effect of the poor statistical accuracy - I'll wait for more data before ringing alarm bells.


Ah... that'll be bumps in reporting due to the Glasgow Fair!? ;)
- could never get anything done last couple of weeks in July (and it'll be the Paisley Fair at the moment)

It'll all settle down shortly - fingers crossed the data continues to improve

All the best
- sd

scotia
Lemon Quarter
Posts: 3561
Joined: November 4th, 2016, 8:43 pm
Has thanked: 2371 times
Been thanked: 1943 times

Re: Coronavirus - Modelling Aspects Only

#432890

Postby scotia » August 6th, 2021, 12:38 pm

servodude wrote:
scotia wrote:The Scottish data has a much poorer statistical accuracy, due to the smaller numbers, and the hospital admissions are reported in a different fashion - usually two days behind the English data. However as previously reported, the Scottish 7-day admissions peaked around 13th July, and I can now report that the 7 day deaths peaked around 25th July. However after falling for a week, the deaths are now showing a slight upward trend, but this is not mirrored in the admissions data. It could simply be an effect of the poor statistical accuracy - I'll wait for more data before ringing alarm bells.


Ah... that'll be bumps in reporting due to the Glasgow Fair!? ;)
- could never get anything done last couple of weeks in July (and it'll be the Paisley Fair at the moment)

It'll all settle down shortly - fingers crossed the data continues to improve

All the best
- sd


Yes - holidays skew the data. Also Scotland has a longer time to register a death (5 days in England, 8 days in Scotland). So these combinations may have affected the returns. Fingers crossed as advised.

Itsallaguess
Lemon Half
Posts: 9129
Joined: November 4th, 2016, 1:16 pm
Has thanked: 4140 times
Been thanked: 10023 times

Re: Coronavirus - Modelling Aspects Only

#432894

Postby Itsallaguess » August 6th, 2021, 1:02 pm

It looks like one of the perplexing disconnects between higher ONS estimates and lower recent case-data might be starting to re-align in a positive way, given data released from the ONS this morning -

First fall in number of positive Covid infections in UK since May -

Latest estimates from the Office of National Statistics (ONS) show the first weekly fall in positive coronavirus infections across the UK since May.

The percentage of people testing positive for the virus decreased in England, Wales and Scotland - but increased in Northern Ireland.

Estimates from the ONS survey suggest just over 810,000 people in the UK would test positive for coronavirus in the week to 31 July, down from 950,000 people the previous week.

This equates to 1.3% of the population – or one in 80 people. The week before it was 1.5% of the population, or one in 70 people.

Breaking down the data by nations, it means around one in 75 people in private households in England had Covid-19 in the week to 31 July - down from one in 65 in the previous week.

In Scotland it was one in 120 and in Wales, one in 230; In Northern Ireland, the percentage of people continued to increase, to one in 55.

Sarah Crofts, The ONS head of analytical outputs for the Covid-19 Infection Survey, called the latest data "encouraging".


Image

Source - https://www.bbc.co.uk/news/live/uk-58112093

Cheers,

Itsallaguess

Hallucigenia
Lemon Quarter
Posts: 2618
Joined: November 5th, 2016, 3:03 am
Has thanked: 166 times
Been thanked: 1718 times

Re: Coronavirus - Modelling Aspects Only

#434531

Postby Hallucigenia » August 13th, 2021, 11:42 am

Hallucigenia wrote:The number that is currently most worrying is the number on ventilation. If you translate that into the equivalent last autumn, 618 on ventilators is more than 17 October, a few days into the tiers. We're 0.62 doublings away from the 1,003 on ventilation on 5 November when lockdown2 was announced, and 2.7 doublings away from the all-time high of 24 January when intensive care was in meltdown, way over capacity. Recently cases have been doubling every 10-14 days, so you might expect ventilation to be similar, with a time lag.

If you're worried about non-Covid deaths then you should be terrified by the prospect of intensive care being saturated with Covid patients. Aside from the fact that the infection risk alone can close down chunks of an ICU to non-Covid patients, if it's at capacity then that means that serious operations like heart transplants and cancer surgery have to be postponed. The real problem is that they just stay in ICU for so long, 20 days on average whereas recovery from surgery might need a day or two in intensive care.

So once the ICU is at capacity, one Covid patient can lead to major operations being postponed for 10-20 other patients who are seriously ill, and there's a fair chance that will not end well for at least one of the non-Covid patients.


Ventilation seems to have peaked at 895 on 2 August, equivalent to 27 October but is continuing at just under that level. I can't imagine it's much fun working on an intensive care ward at the moment. But given that Covid patients are typically in intensive care for 3 weeks, and admissions peaked on 22 July but are continuing at a fairly high level, we should hopefully see numbers on ventilation drift down from now on.

Still, we're currently generating ~5k new cases of long Covid per day....

redsturgeon
Lemon Half
Posts: 8912
Joined: November 4th, 2016, 9:06 am
Has thanked: 1309 times
Been thanked: 3667 times

Re: Coronavirus - Modelling Aspects Only

#434560

Postby redsturgeon » August 13th, 2021, 12:44 pm

Hallucigenia wrote:
Hallucigenia wrote:The number that is currently most worrying is the number on ventilation. If you translate that into the equivalent last autumn, 618 on ventilators is more than 17 October, a few days into the tiers. We're 0.62 doublings away from the 1,003 on ventilation on 5 November when lockdown2 was announced, and 2.7 doublings away from the all-time high of 24 January when intensive care was in meltdown, way over capacity. Recently cases have been doubling every 10-14 days, so you might expect ventilation to be similar, with a time lag.

If you're worried about non-Covid deaths then you should be terrified by the prospect of intensive care being saturated with Covid patients. Aside from the fact that the infection risk alone can close down chunks of an ICU to non-Covid patients, if it's at capacity then that means that serious operations like heart transplants and cancer surgery have to be postponed. The real problem is that they just stay in ICU for so long, 20 days on average whereas recovery from surgery might need a day or two in intensive care.

So once the ICU is at capacity, one Covid patient can lead to major operations being postponed for 10-20 other patients who are seriously ill, and there's a fair chance that will not end well for at least one of the non-Covid patients.


Ventilation seems to have peaked at 895 on 2 August, equivalent to 27 October but is continuing at just under that level. I can't imagine it's much fun working on an intensive care ward at the moment. But given that Covid patients are typically in intensive care for 3 weeks, and admissions peaked on 22 July but are continuing at a fairly high level, we should hopefully see numbers on ventilation drift down from now on.

Still, we're currently generating ~5k new cases of long Covid per day....



Plus diabetes seems to be a long term complication.

John


Return to “Coronavirus Discussions”

Who is online

Users browsing this forum: No registered users and 5 guests