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
johnhemming
Lemon Quarter
Posts: 3858
Joined: November 8th, 2016, 7:13 pm
Has thanked: 9 times
Been thanked: 609 times

Re: Coronavirus - Modelling Aspects Only

#379910

Postby johnhemming » January 23rd, 2021, 11:16 am

scotia wrote:The strongest I would put my own observations is that I suggest that there could be - and if the effect I am seeing is due to an increased death rate of the variant, it looks like its in agreement with the government's tentative figure of a 30% increase .

It is, of course, possible that the health service have increased the threshold at which someone goes to hospital.

However, taking the deaths by date of death for England:
https://coronavirus.data.gov.uk/details ... me=England

They are running around the 900-1000 mark.

Hospital admissions are running around the 3600-4000 mark.

Your figure of 26.5% is just over a quarter.

So fatalities as a proportion of hospital admissions appears about the same.

It may be that the circumstances are such that increase the viral load and number of people infected, but I don't see the calculation for a 30% increase as a proportion of hospital admissions.
(which would be 1270-1300 deaths per day)

The government don't really know how many people get infected. You could take the ONS proportion of people infected and compare that to hospital admissions and/or deaths. I am not sure what happened with the ONS figures. Initially they were testing the same people every week so they could pick up when they were infected, but I think that stopped which means there is then a month's uncertainty about infection date.

funduffer
Lemon Quarter
Posts: 1328
Joined: November 4th, 2016, 12:11 pm
Has thanked: 122 times
Been thanked: 831 times

Re: Coronavirus - Modelling Aspects Only

#380010

Postby funduffer » January 23rd, 2021, 4:07 pm

The first modelling I have seen of the effect of the vaccination programme on cases, admissions, ICU and deaths, is provided by the Covid-19 Actuaries Response Group, via the Spectator:

https://www.spectator.co.uk/article/how ... e-roll-out

Key findings by mid- to late-March:

86% reduction in hospital deaths
59% reduction in hospital admissions
34% reduction in ICU admissions
15% reduction in cases

They list their assumptions, the main one being that the first 4 groups are vaccinated by mid-February.

There is also some modelling reported in Monevator, showing that results from Israel are starting to show the benefits from vaccination, using similar modelling:

https://monevator.com/the-beginning-of- ... 19-crisis/

Let's hope this all proves to be reasonably accurate!

FD

spasmodicus
Lemon Slice
Posts: 259
Joined: November 6th, 2016, 9:35 am
Has thanked: 65 times
Been thanked: 117 times

Re: Coronavirus - Modelling Aspects Only

#380277

Postby spasmodicus » January 24th, 2021, 2:34 pm

johnhemming wrote:
scotia wrote:The strongest I would put my own observations is that I suggest that there could be - and if the effect I am seeing is due to an increased death rate of the variant, it looks like its in agreement with the government's tentative figure of a 30% increase .

It is, of course, possible that the health service have increased the threshold at which someone goes to hospital.

However, taking the deaths by date of death for England:
https://coronavirus.data.gov.uk/details ... me=England

They are running around the 900-1000 mark.

Hospital admissions are running around the 3600-4000 mark.

Your figure of 26.5% is just over a quarter.

So fatalities as a proportion of hospital admissions appears about the same.

It may be that the circumstances are such that increase the viral load and number of people infected, but I don't see the calculation for a 30% increase as a proportion of hospital admissions.
(which would be 1270-1300 deaths per day)

The government don't really know how many people get infected. You could take the ONS proportion of people infected and compare that to hospital admissions and/or deaths. I am not sure what happened with the ONS figures. Initially they were testing the same people every week so they could pick up when they were infected, but I think that stopped which means there is then a month's uncertainty about infection date.


I was too was looking at deaths relative to admissions data and came across this
https://www.bbc.co.uk/news/health-55652771
which carries the headline Covid-19: Packed hospitals raised death risk by 20%

and this from the USAs first wave
https://www.cidrap.umn.edu/news-perspective/2021/01/crowded-icus-tied-higher-risk-covid-19-death

I would expect that survival chances would be poorer in an overstretched hospital. That hospital might be operating in an area where the new variant is dominant, so it's difficult to untangle cause and effect. It may be possible to draw conclusions from local data, or from the death rate for people that did not die in hospital, but imo the all England data are much too noisy.

The latest ONS all England random sampling figures for 13th January (10558440 infected) were published yesterday and came in somewhat higher than I expected, after the previous REACT survey point for 10th January (889540 infected). As I mentioned earlier, the REACT data points seem to be consistently lower than those ONS estimates. I do not know the explanation for this, but it bodes ill for drawing precise conclusions from these very broad brush estimates.

S

dealtn
Lemon Half
Posts: 6072
Joined: November 21st, 2016, 4:26 pm
Has thanked: 441 times
Been thanked: 2324 times

Re: Coronavirus - Modelling Aspects Only

#380283

Postby dealtn » January 24th, 2021, 2:46 pm

spasmodicus wrote:
I would expect that survival chances would be poorer in an overstretched hospital. That hospital might be operating in an area where the new variant is dominant, so it's difficult to untangle cause and effect. It may be possible to draw conclusions from local data, or from the death rate for people that did not die in hospital, but imo the all England data are much too noisy.



And yet some call for less granularity in the data as it's not useful.

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

#380293

Postby Itsallaguess » January 24th, 2021, 3:12 pm

dealtn wrote:
spasmodicus wrote:
I would expect that survival chances would be poorer in an overstretched hospital. That hospital might be operating in an area where the new variant is dominant, so it's difficult to untangle cause and effect. It may be possible to draw conclusions from local data, or from the death rate for people that did not die in hospital, but imo the all England data are much too noisy.


And yet some call for less granularity in the data as it's not useful.


I think it's important not to generalise too much here though.

Granular data that might help to improve the medical side of things is a different beast entirely to granular data that might be sought to influence the social-adherence side of things.

Personally, I'd never be critical of the former, but I do continue to believe that at a certain point, the granularity of the latter begins to lose the justification for ever-deeper rabbit-holes...

Cheers,

Itsallaguess

dealtn
Lemon Half
Posts: 6072
Joined: November 21st, 2016, 4:26 pm
Has thanked: 441 times
Been thanked: 2324 times

Re: Coronavirus - Modelling Aspects Only

#380295

Postby dealtn » January 24th, 2021, 3:18 pm

Itsallaguess wrote:
dealtn wrote:
spasmodicus wrote:
I would expect that survival chances would be poorer in an overstretched hospital. That hospital might be operating in an area where the new variant is dominant, so it's difficult to untangle cause and effect. It may be possible to draw conclusions from local data, or from the death rate for people that did not die in hospital, but imo the all England data are much too noisy.


And yet some call for less granularity in the data as it's not useful.


I think it's important not to generalise too much here though.

Granular data that might help to improve the medical side of things is a different beast entirely to granular data that might be sought to influence the social-adherence side of things.

Personally, I'd never be critical of the former, but I do continue to believe that at a certain point, the granularity of the latter begins to lose the justification for ever-deeper rabbit-holes...

Cheers,

Itsallaguess


I don't disagree tremendously, there will be times granularity isn't useful.

(Odd for an adherent of granularity to be accused of generalising too much. Isn't language great?)

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

#380298

Postby Itsallaguess » January 24th, 2021, 3:24 pm

dealtn wrote:
Itsallaguess wrote:
dealtn wrote:
And yet some call for less granularity in the data as it's not useful.


I think it's important not to generalise too much here though.

Granular data that might help to improve the medical side of things is a different beast entirely to granular data that might be sought to influence the social-adherence side of things.

Personally, I'd never be critical of the former, but I do continue to believe that at a certain point, the granularity of the latter begins to lose the justification for ever-deeper rabbit-holes...


I don't disagree tremendously, there will be times granularity isn't useful


And in the past where I've questioned the usefulness of ever-more granular data in the area being discussed at the time, that was the point I was trying to make in relation to the specific discussion that was ongoing.

If I gave the impression that I was against granular data in all circumstances, then I should have been clearer in the point I was trying to make, as that wouldn't be something I'd have intended to propose.

I'm grateful for the opportunity to agree with your view above, which is that it's not always useful, and as such, just because someone might perhaps suggest that 'we should be able to find that out', that fact alone doesn't always justify the real-world usefulness in doing so...

Cheers,

Itsallaguess

johnhemming
Lemon Quarter
Posts: 3858
Joined: November 8th, 2016, 7:13 pm
Has thanked: 9 times
Been thanked: 609 times

Re: Coronavirus - Modelling Aspects Only

#380319

Postby johnhemming » January 24th, 2021, 4:19 pm

Itsallaguess wrote:I'm grateful for the opportunity to agree with your view above, which is that it's not always useful, and as such, just because someone might perhaps suggest that 'we should be able to find that out', that fact alone doesn't always justify the real-world usefulness in doing so...

I accept that the statistical reliability reduces for smaller sample sizes. It is, however, worth keeping and analysing the data and recognising where uncertainty exists.

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

#381783

Postby scotia » January 29th, 2021, 2:05 am

Another week of (English) data. To recap:- The Blue Points are the deaths by publish date, summed over the preceding week.
As I described in previous posts there has been a strong correlation between the deaths by publish date, and the hospital admissions of some days previous. So the Red Points are the hospital admissions, summed over a week, multiplied by 0.265, and moved forward by 13 days - And these are renamed as being the Projected deaths by publish date. The size of the vertical bars are the statistical standard deviations, assuming a Poisson distribution. These factors - 13 days and 0.265 provided a good fit to data of a few months ago.

Image

It is now clear that the deaths curve has diverged significantly from the projected values based on the previous correlations with admissions. To get a better fit over the past two weeks, I reduced the time difference from 13 days to 11, and increased the multiplicative factor from 0.265 to 0.3. The result is displayed below

Image

The new (red) projected points are now fitting closely on top of the (blue) actual deaths - and some of the blue points are difficult to see since they are over-written by the red points.

And on wondering why the model parameters have changed from 13 days time slip and a 0.265 ratio to 11 days and a 0.3 ratio, the jury is still out. Is the new variant more deadly, or are the hospitals simply being over-run?

However - to more cheerful matters, the peak death number seems to have passed - it reached 7945 for the week ending 23/01/21. And although the reduction is slow, there is some encouragement from the last few admissions points where the downwards gradient seems to be increasing.

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

#381786

Postby Itsallaguess » January 29th, 2021, 6:13 am

scotia wrote:
And on wondering why the model parameters have changed from 13 days time slip and a 0.265 ratio to 11 days and a 0.3 ratio, the jury is still out.

Is the new variant more deadly, or are the hospitals simply being over-run?


I've been told by someone close to the scene regarding a local ICU that the critical care staff are really struggling to cope with it all now, at both a personal and professional level. Many are regularly in tears, simply trying to get through the working day and cope with the mental and physical issues of having to deal with this dreadful situation for so long now.

This shouldn't come as a shock, of course, and my heart goes out to any and all front-line NHS staff working in these critical-care areas, who will hopefully be the first to see the tangible benefits of the recent vaccine programme having been rolled out to the most vulnerable groups.

Whilst there may well be some medical effects being seen from the new variants, I suspect that the mental and physical exhaustion of the critical care teams will be playing some part in the quality of the care that they can sustain right now..

Cheers,

Itsallaguess

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

#383725

Postby scotia » February 4th, 2021, 7:35 pm

Another week of (English) data. To recap:- The Blue Points are the deaths by publish date, summed over the preceding week.
As I described in previous posts there has been a strong correlation between the deaths by publish date, and the hospital admissions of some days previous. So the Red Points are the hospital admissions, summed over a week, multiplied by 0.265, and moved forward by 13 days - And these are renamed as being the projected deaths by publish date. The size of the vertical bars are the statistical standard deviations, assuming a Poisson distribution. These factors - 13 days and 0.265 provided a good fit to data of a few months ago.

Image

It is now clear that the deaths curve has diverged significantly from the projected values based on the previous correlations with admissions. To get a better fit over the past few weeks, I reduced the time difference from 13 days to 11, and increased the multiplicative factor from 0.265 to 0.3. The result is displayed below

Image

The new (red) projected points are now fitting closely around the peak of the (blue) actual deaths. However - the final two blue points have both deviated downwards from the projections by more than a standard deviation. One swallow does not a summer make - nor probably does two. However if next week this deviation continues, we may be seeing the reduction in deaths due to vaccinations in care homes. This may be expected to reduce deaths more than hospital admissions - since it may have been kinder to leave old people who are ill in the care home, rather than increasing the stress of a hospital admission.
Last week I was encouraged by the apparent increase in the downward gradient of the admissions - but it looks like my optimism may not have been well founded. Perhaps I should wait for another week of data before pronouncing that the downward gradient is (unfortunately) reducing. If so, could this be due to a relaxation of strict lockdown behaviour, possibly due to a mistaken belief that all the oldies have now been vaccinated? I got my first jab (Oxford vaccine) yesterday, and my wife gets her vaccination on Saturday. Then we need to wait for 3 weeks before the immunity develops.

Sorcery
Lemon Quarter
Posts: 1229
Joined: November 4th, 2016, 6:38 pm
Has thanked: 147 times
Been thanked: 366 times

Re: Coronavirus - Modelling Aspects Only

#383735

Postby Sorcery » February 4th, 2021, 8:04 pm

scotia wrote:The new (red) projected points are now fitting closely around the peak of the (blue) actual deaths. However - the final two blue points have both deviated downwards from the projections by more than a standard deviation. One swallow does not a summer make - nor probably does two. However if next week this deviation continues, we may be seeing the reduction in deaths due to vaccinations in care homes. This may be expected to reduce deaths more than hospital admissions - since it may have been kinder to leave old people who are ill in the care home, rather than increasing the stress of a hospital admission.
Last week I was encouraged by the apparent increase in the downward gradient of the admissions - but it looks like my optimism may not have been well founded. Perhaps I should wait for another week of data before pronouncing that the downward gradient is (unfortunately) reducing. If so, could this be due to a relaxation of strict lockdown behaviour, possibly due to a mistaken belief that all the oldies have now been vaccinated? I got my first jab (Oxford vaccine) yesterday, and my wife gets her vaccination on Saturday. Then we need to wait for 3 weeks before the immunity develops.


Surely deaths dropping faster than admissions is a positive? It could also be due to your refactored formula for lining up deaths and admissions, even the vaccine rollout may be changing the curves. Not something I would worry about at this stage. Looks positive to me.

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

#383748

Postby scotia » February 4th, 2021, 9:18 pm

Sorcery wrote:Surely deaths dropping faster than admissions is a positive?

Yes of course it is.
However I was also hoping for a continuing steep decline in admissions, and although the data is currently inadequate it looks like it could be slowing.

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

#383750

Postby servodude » February 4th, 2021, 9:25 pm

scotia wrote:
Sorcery wrote:Surely deaths dropping faster than admissions is a positive?

Yes of course it is.
However I was also hoping for a continuing steep decline in admissions, and although the data is currently inadequate it looks like it could be slowing.


It supports the hypothesis that supply of hospital beds was a factor in the divergence

- 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

#383755

Postby scotia » February 4th, 2021, 9:58 pm

servodude wrote:
scotia wrote:
Sorcery wrote:Surely deaths dropping faster than admissions is a positive?

Yes of course it is.
However I was also hoping for a continuing steep decline in admissions, and although the data is currently inadequate it looks like it could be slowing.


It supports the hypothesis that supply of hospital beds was a factor in the divergence

- sd

Its possible - but I think that going forward the vaccinations are going to change the dynamics significantly - and will probably put most other cause and effect prognostications into the "maybe" category.
The last (inconclusive) report which I found on the possible increased mortality rate of the new variant is from the BMJ on 26th January https://www.bmj.com/content/372/bmj.n230. That was the other candidate for the current model parameters being different from those of a few months ago. If the death rate continues to plunge, then I suspect that the statistical evidence of such an effect will remain inconclusive.

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

#383765

Postby servodude » February 4th, 2021, 10:19 pm

scotia wrote:
servodude wrote:
scotia wrote:Yes of course it is.
However I was also hoping for a continuing steep decline in admissions, and although the data is currently inadequate it looks like it could be slowing.


It supports the hypothesis that supply of hospital beds was a factor in the divergence

- sd

Its possible - but I think that going forward the vaccinations are going to change the dynamics significantly - and will probably put most other cause and effect prognostications into the "maybe" category.
The last (inconclusive) report which I found on the possible increased mortality rate of the new variant is from the BMJ on 26th January https://www.bmj.com/content/372/bmj.n230. That was the other candidate for the current model parameters being different from those of a few months ago. If the death rate continues to plunge, then I suspect that the statistical evidence of such an effect will remain inconclusive.


indeed! we're at a point in this where "the model" should very soon change dramatically... ...and for once it's for the better! :D

I just thought it was worth mentioning that what your plots show is what you would have expected if there was a "saturation block" on the admissions data; a limit that would not have been present on the deaths or "cases" data

After a bit of initial scepticism, I have become convinced of the increased infectiousness of the "new variant"
- teasing anything concrete about whether it is also more deadly will be very difficult in the absence of any controlled investigations (I suppose though that being more "infectious" you might expect it to spread more readily in the patient also?)
- hopefully we won't have to muse too much about it for much longer

- sd

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

#383768

Postby jfgw » February 4th, 2021, 10:28 pm

servodude wrote:It supports the hypothesis that supply of hospital beds was a factor in the divergence


Hospital bed occupancy has certainly come down (still high, though); it peaked mid to late January so, if this is the main factor, we should see a reconvergence.

Mechanical ventilator bed occupancy is still high, passed the peak though.


Julian F. G. W.

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

#383774

Postby servodude » February 4th, 2021, 11:18 pm

jfgw wrote:
servodude wrote:It supports the hypothesis that supply of hospital beds was a factor in the divergence


Hospital bed occupancy has certainly come down (still high, though); it peaked mid to late January so, if this is the main factor, we should see a reconvergence.

Mechanical ventilator bed occupancy is still high, passed the peak though.

Julian F. G. W.


I really hope that vaccination becomes the dominant factor quickly - ideally before we saw any reconvergence

There was some interesting discussion about patients on mechanical ventilator in the More or Less Podcast recently (https://www.bbc.co.uk/programmes/m000rccm) at about 5 minutes in
- we might expect the occupancy to fall more slowly than the other figures due to how triaged it is; only 25% of ICU admissions are over 70 years old (25% are under 52)
- this also means that we can expect the vaccine programme to have little effect on these numbers for some time

- 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

#385870

Postby scotia » February 12th, 2021, 12:00 am

Another week of (English) data. To recap:- The Blue Points are the deaths by publish date, summed over the preceding week.
As I described in previous posts there has been a strong correlation between the deaths by publish date, and the hospital admissions of some days previous. The Red Points are the hospital admissions, summed over a week, multiplied by 0.3, and moved forward by 11 days - And these are renamed as being the projected deaths by publish date. The size of the vertical bars are the statistical standard deviations, assuming a Poisson distribution. These parameters provided a reasonable fit a few weeks ago, however going back a few months, a 13 day slip and a factor of 0.265 provided an optimal fit. There was speculation that the increase in the mortality ratio between those times could possibly have been due to over-stretched hospitals, or the new Kent virus variant. However - back to the present and some good news

Image

The recent actual deaths are now diverging from the projected deaths - in the downward direction. So the 3715 projected deaths in the week ending 20/2/21 seems over-pessimistic. It looks likely that the actual deaths will be less than 3000. We are possibly starting to see the benefits of the early vaccinations in Care Homes - which should significantly affect the deaths, but have a smaller effect on admissions. I would expect to see further vaccination-related reductions in the ratio of deaths to admissions in the coming weeks.

Julian
Lemon Quarter
Posts: 1385
Joined: November 4th, 2016, 9:58 am
Has thanked: 532 times
Been thanked: 676 times

Re: Coronavirus - Modelling Aspects Only

#385947

Postby Julian » February 12th, 2021, 10:54 am

scotia wrote:...
The recent actual deaths are now diverging from the projected deaths - in the downward direction. So the 3715 projected deaths in the week ending 20/2/21 seems over-pessimistic. It looks likely that the actual deaths will be less than 3000. We are possibly starting to see the benefits of the early vaccinations in Care Homes - which should significantly affect the deaths, but have a smaller effect on admissions. I would expect to see further vaccination-related reductions in the ratio of deaths to admissions in the coming weeks.

Good news indeed. I also note that the news yesterday was talking about the RECOVERY project (their capitalisation!) showing a tocilizumab + dexamethasone treatment giving ~33% reduction in mortality for those requiring oxygen therapy and almost a 50% reduction in mortality for patients on mechanical ventilation [ Source: https://www.nihr.ac.uk/news/recovery-tr ... d-19/26844 ]. Presumably that will be another factor effecting your model at some point if not already, I assume by exerting a downward pressure on your current 0.3 multiplier.

- Julian


Return to “Coronavirus Discussions”

Who is online

Users browsing this forum: No registered users and 6 guests