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

#402894

Postby jfgw » April 9th, 2021, 1:15 pm

It is actually 28 days within the first positive test. If treatments improve such that patients live longer before dying, an increasing number of Covid deaths will not be included in the count.

Julian F. G. W.

swill453
Lemon Half
Posts: 7962
Joined: November 4th, 2016, 6:11 pm
Has thanked: 984 times
Been thanked: 3643 times

Re: Coronavirus - Modelling Aspects Only

#402896

Postby swill453 » April 9th, 2021, 1:19 pm

Julian wrote:Since "Within 28 days of a positive test" can be read as "within 28 days of the most recent positive test" such regular in-hospital testing would catch those sort of deaths.

It's not though. The definition is specifically "death in a person with a laboratory-confirmed positive COVID-19 test and died within (equal to or less than) 28 days of the first positive specimen date".

Scott.

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

#402908

Postby scotia » April 9th, 2021, 2:05 pm

Yes - thanks to all for bringing up the 28 day figure. This was bothering me when attempting, in the most recent data, to fit the gradient of deaths to the gradient of admissions - and even when I increased the model time slip to 22 days (and the ratio to 0.085), the death gradient was exceeding the admissions data. I felt that further increases in the time slip would make little sense - due increasingly to this 28 day figure - and also to the impact of vaccinations several weeks ago being different from the current state.

This "fit" (not to be taken seriously) is displayed below

Image

P.S. the Imperial College study reported by the BBC on the broken link between cases and deaths doesn't seem to be on public view. On attempting to link to it I get an Embargoed message and a requirement to login with a password to download it.

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

#402931

Postby Julian » April 9th, 2021, 3:45 pm

swill453 wrote:
Julian wrote:Since "Within 28 days of a positive test" can be read as "within 28 days of the most recent positive test" such regular in-hospital testing would catch those sort of deaths.

It's not though. The definition is specifically "death in a person with a laboratory-confirmed positive COVID-19 test and died within (equal to or less than) 28 days of the first positive specimen date".

Scott.

Wow. That is totally nuts. In that case if I had had a mild case back in April last year, tested positive and subsequently recovered, does that mean that I can now never officially die of Covid-19 since my first positive specimen date would have been way back on April the something-th 2020? Surely there must be some extra wording to plug that what seems to me fairly obvious loophole not to mention the potential loophole John pointed out of someone being admitted to hospital with confirmed Covid-19 and subsequently dying in hospital 29 or more days later?

- Julian

swill453
Lemon Half
Posts: 7962
Joined: November 4th, 2016, 6:11 pm
Has thanked: 984 times
Been thanked: 3643 times

Re: Coronavirus - Modelling Aspects Only

#402945

Postby swill453 » April 9th, 2021, 4:53 pm

Julian wrote:Wow. That is totally nuts. In that case if I had had a mild case back in April last year, tested positive and subsequently recovered, does that mean that I can now never officially die of Covid-19 since my first positive specimen date would have been way back on April the something-th 2020? Surely there must be some extra wording to plug that what seems to me fairly obvious loophole not to mention the potential loophole John pointed out of someone being admitted to hospital with confirmed Covid-19 and subsequently dying in hospital 29 or more days later?

You'll never appear in the "death within 28 days" daily reported figures. They have these precise rules just so they can be collated very quickly.

You would, however, appear in the "Covid-19 on the death certificate" figures which are produced at a more leisurely pace. And which in the years to come should be used for learning lessons.

Scott.

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

#403130

Postby dealtn » April 10th, 2021, 10:38 am

Julian wrote:
swill453 wrote:
Julian wrote:Since "Within 28 days of a positive test" can be read as "within 28 days of the most recent positive test" such regular in-hospital testing would catch those sort of deaths.

It's not though. The definition is specifically "death in a person with a laboratory-confirmed positive COVID-19 test and died within (equal to or less than) 28 days of the first positive specimen date".

Scott.

Wow. That is totally nuts. In that case if I had had a mild case back in April last year, tested positive and subsequently recovered, does that mean that I can now never officially die of Covid-19 since my first positive specimen date would have been way back on April the something-th 2020? Surely there must be some extra wording to plug that what seems to me fairly obvious loophole not to mention the potential loophole John pointed out of someone being admitted to hospital with confirmed Covid-19 and subsequently dying in hospital 29 or more days later?

- Julian


It's not totally nuts (but still nuts).

The measure is a crude measure put together as an attempt to quantify the number of Covid deaths. It was recognised as being inaccurate with respect to any particular death, but an improvement on the previous method.

It misses many that genuinely die of Covid, that occur beyond the 28 day threshold, but includes non-Covid deaths, such as car accidents. The offsetting errors were thought to be "better" (2 wrongs do make a right!).

This has become more of an issue as treatments and care have improved through knowledge expansion, with more now surviving past 28 days, yet still eventually succumbing. One "error" group has grown.

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

#403131

Postby Julian » April 10th, 2021, 10:44 am

dealtn wrote:
Julian wrote:
swill453 wrote:It's not though. The definition is specifically "death in a person with a laboratory-confirmed positive COVID-19 test and died within (equal to or less than) 28 days of the first positive specimen date".

Scott.

Wow. That is totally nuts. In that case if I had had a mild case back in April last year, tested positive and subsequently recovered, does that mean that I can now never officially die of Covid-19 since my first positive specimen date would have been way back on April the something-th 2020? Surely there must be some extra wording to plug that what seems to me fairly obvious loophole not to mention the potential loophole John pointed out of someone being admitted to hospital with confirmed Covid-19 and subsequently dying in hospital 29 or more days later?

- Julian


It's not totally nuts (but still nuts).

The measure is a crude measure put together as an attempt to quantify the number of Covid deaths. It was recognised as being inaccurate with respect to any particular death, but an improvement on the previous method.

It misses many that genuinely die of Covid, that occur beyond the 28 day threshold, but includes non-Covid deaths, such as car accidents. The offsetting errors were thought to be "better" (2 wrongs do make a right!).

This has become more of an issue as treatments and care have improved through knowledge expansion, with more now surviving past 28 days, yet still eventually succumbing. One "error" group has grown.

Thank you. That makes sense. In light of your final paragraph on changing circumstances maybe I should change my "that is totally nuts" comment to "that has now become a little bit nuts".

- Julian

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

#403133

Postby servodude » April 10th, 2021, 11:06 am

Julian wrote:
dealtn wrote:
Julian wrote:Wow. That is totally nuts. In that case if I had had a mild case back in April last year, tested positive and subsequently recovered, does that mean that I can now never officially die of Covid-19 since my first positive specimen date would have been way back on April the something-th 2020? Surely there must be some extra wording to plug that what seems to me fairly obvious loophole not to mention the potential loophole John pointed out of someone being admitted to hospital with confirmed Covid-19 and subsequently dying in hospital 29 or more days later?

- Julian


It's not totally nuts (but still nuts).

The measure is a crude measure put together as an attempt to quantify the number of Covid deaths. It was recognised as being inaccurate with respect to any particular death, but an improvement on the previous method.

It misses many that genuinely die of Covid, that occur beyond the 28 day threshold, but includes non-Covid deaths, such as car accidents. The offsetting errors were thought to be "better" (2 wrongs do make a right!).

This has become more of an issue as treatments and care have improved through knowledge expansion, with more now surviving past 28 days, yet still eventually succumbing. One "error" group has grown.

Thank you. That makes sense. In light of your final paragraph on changing circumstances maybe I should change my "that is totally nuts" comment to "that has now become a little bit nuts".

- Julian


Yeah it's always been "just what it is"; which is fine... if somewhat misrepresented on occasion
- I hope someone's paying enough attention to the positive results to know that re-testing positive after a decent period is worth noting ;)

-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

#406390

Postby scotia » April 22nd, 2021, 11:27 pm

Further to my plot of two weeks ago, I'm displaying another "fit" - since there seems to have been a significant change in the downwards death curve. Previously I reported that the gradient of the death curve was significantly steeper than the admissions curve - even with a three-week time slip, however this pattern appears to have changed.

Image

As can be seen, the steep downward gradient of the deaths continued to the week ending 6th April, then it reversed for a week, before settling down close to the gradient of the time-slipped admissions. This oscillation could possibly have been caused by the Easter holiday weekend from 2nd to 7th April which will have caused anomalies in the deaths by publish dates, so I also had a look at the deaths by death date. These have the disadvantage of being updated retrospectively, however values of a week ago are reasonably stable. On carrying some spreadsheet calculations, it appears that there is around 3 to 4 days difference between the deaths by publish date and deaths by death date (summed over 7 days). So I checked the deaths by death date - time slipped by 20 days, with the same multiplier of 0.075. The curve was similar but with the oscillation around the Easter weekend being largely ironed out. It confirmed that the deaths and (time slipped) admissions are now following similar gradients.
So what has changed? I think the steeply falling death rate was a feature of the success of the vaccinations in the oldies (I had my second jab yesterday). There won't be such a spectacular effect due to the vaccinations of the younger cohort - but its still a downward trend.

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

#406466

Postby Julian » April 23rd, 2021, 1:11 pm

scotia wrote:...
So what has changed? I think the steeply falling death rate was a feature of the success of the vaccinations in the oldies (I had my second jab yesterday). There won't be such a spectacular effect due to the vaccinations of the younger cohort - but its still a downward trend.

Thanks. Love this stuff. Much appreciated.

On that last point I do note that with April vaccine shortages the rate of increase of first-dose-only vaccinations has dramatically reduced as the majority of daily doses are allocated to second jabs so to some extend that vaccination of the younger cohorts has, for now at least, slowed to a pretty slow crawl. That probably makes quite a difference since even after the first dose efficacy against symptomatic disease looks extremely respectable at 72% (https://www.theguardian.com/society/202 ... y-65-study) and even a quite respectable 57% for asymptomatic infections. One would assume that efficacy against hospitalisation and death, which is what you care about for your model, is even stronger after just one dose.

So, I'm absolutely NOT saying that the UK is doing the wrong thing because extending dosing intervals even further might be a risk too far in terms of compromising overall double-dose efficacy so 2nd doses at 10 - 12 weeks must be adhered to but in the short term the relatively modest levels of increase in efficacy in going from single dose to double dose for someone already vaccinated vs a significant level of protection conferred within a few weeks for an extra person had that jab gone into their arm as a first dose means that, strictly from the point of view of your modelling, a lot of doses are being at least somewhat "wasted" by giving a modest bump in protection to someone who already had quite a lot of protection vs giving a quite high level of protection to one extra member of the population.

As I say, I'm not saying those second jabs should not be happening, they absolutely should be because one cannot extend and extend the dosing interval thus risking peoples overall medium/long term protection, I'm just commenting on how it might be affecting your model.

- Julian

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

#409929

Postby scotia » May 6th, 2021, 9:57 pm

Further to my plot of two weeks ago, I'm displaying another "fit" - with the same parameters, but with data to 6/5/21. This seems to confirm that the death curve gradient has significantly reduced from the position of some months ago, and it now seems to be back in step with the time-slipped admissions - albeit with a bit of oscillation (once again a holiday weekend is included).

Image

Its a fortnight since I had my second jab, and the first jab is now being rolled out to my children (in their forties). So the huge reduction in the deaths of the oldies has now worked through the system - and further reductions are likely to occur at a slower pace. The recent, and continuing, removal of many restrictions will likely increase the infection rate - probably chiefly in the (unvaccinated) younger age groups. This may start to show through in the admissions, but hopefully not in the deaths. No doubt government statisticians will be watching this carefully.

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

#409945

Postby servodude » May 6th, 2021, 11:05 pm

scotia wrote:This may start to show through in the admissions, but hopefully not in the deaths. No doubt government statisticians will be watching this carefully


No doubt indeed; and I hope they are given the freedom to do so with an open mind.
If they can get enough confidence in how it should be running that should leave them in good stead for picking up any digression, e.g. if a VOC escapes.
As it stands now though it's looking really good.

-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

#413806

Postby scotia » May 21st, 2021, 10:36 am

Another fortnight of data - and another "fit" - with the same parameters. With the much reduced numbers, the statistical accuracy is poor, but an optimist may suggest that the death curve gradient is once again slipping below the the time slipped admissions.

Image

So currently it continues to look like good news. It will probably be a few weeks before the reduction in restrictions shows any effect.

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

#413833

Postby servodude » May 21st, 2021, 11:10 am

scotia wrote:So currently it continues to look like good news. It will probably be a few weeks before the reduction in restrictions shows any effect.

I'm not one for putting arbitrary targets on data. .
..but it sounds like a good excuse for a trip to the pub! :D ;)

Thanks for such a sterling and committed effort to keeping track of this stuff. I'm sure I'm not the only one that appreciates your work

Stay well
-sd

GrahamPlatt
Lemon Quarter
Posts: 2059
Joined: November 4th, 2016, 9:40 am
Has thanked: 1032 times
Been thanked: 824 times

Re: Coronavirus - Modelling Aspects Only

#414636

Postby GrahamPlatt » May 24th, 2021, 2:48 pm

B1.617.2 "India" variant progressing rapidly. AZ relatively ineffective:
https://mobile.twitter.com/chrischirp/s ... 7349872644

(Sorry, unsure whether to post on "no statistics" or here. Plenty of figures though, so I thought here.).

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

#414650

Postby Itsallaguess » May 24th, 2021, 3:09 pm

GrahamPlatt wrote:
B1.617.2 "India" variant progressing rapidly.

AZ relatively ineffective:

https://mobile.twitter.com/chrischirp/s ... 7349872644


Did you read the whole series of tweets?

She doesn't say that it's 'ineffective' at all - she's just commenting on the relatively low initial effectivity of a single AZ dose, alongside the current rise and domination of the Indian variant at a time where the much-improved effectiveness of a double-dose has yet to be delivered to all adults, partly due to our initial longer drum-beat between doses, which is now being reduced.....

It's a timing issue that she's really concerned with - and not primarily a 'vaccine effectiveness' issue...

Cheers,

Itsallaguess

GrahamPlatt
Lemon Quarter
Posts: 2059
Joined: November 4th, 2016, 9:40 am
Has thanked: 1032 times
Been thanked: 824 times

Re: Coronavirus - Modelling Aspects Only

#414672

Postby GrahamPlatt » May 24th, 2021, 4:32 pm

Itsallaguess wrote:
GrahamPlatt wrote:
AZ relatively ineffective:

https://mobile.twitter.com/chrischirp/s ... 7349872644


Did you read the whole series of tweets?

Itsallaguess



I don't wish to bicker, but did you read "relatively"?

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

#414688

Postby Itsallaguess » May 24th, 2021, 5:32 pm

GrahamPlatt wrote:
Itsallaguess wrote:
GrahamPlatt wrote:
AZ relatively ineffective:

https://mobile.twitter.com/chrischirp/s ... 7349872644


Did you read the whole series of tweets?


I don't wish to bicker, but did you read "relatively"?


I don't either, and if you'd said 'relatively less effective' then I think that would have perhaps properly reflected the current position in relation to the AZ effectiveness compared to the Pfizer effectiveness, but I think to say 'relatively ineffective' puts a completely different and unwarranted slant on the piece, as that's simply not what's being said in it....

Cheers,

Itsallaguess

GrahamPlatt
Lemon Quarter
Posts: 2059
Joined: November 4th, 2016, 9:40 am
Has thanked: 1032 times
Been thanked: 824 times

Re: Coronavirus - Modelling Aspects Only

#414704

Postby GrahamPlatt » May 24th, 2021, 6:00 pm

Itsallaguess wrote:I don't either, and if you'd said 'relatively less effective' then I think that would have perhaps properly reflected the current position in relation to the AZ effectiveness compared to the Pfizer effectiveness, but I think to say 'relatively ineffective' puts a completely different and unwarranted slant on the piece, as that's simply not what's being said in it....

Cheers,

Itsallaguess


OK, I’ll accept that, yes. It’s a nuance, because to me they’re almost equivalent, but they are indeed not the same.

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

#414709

Postby Itsallaguess » May 24th, 2021, 6:15 pm

GrahamPlatt wrote:
Itsallaguess wrote:
I don't either, and if you'd said 'relatively less effective' then I think that would have perhaps properly reflected the current position in relation to the AZ effectiveness compared to the Pfizer effectiveness, but I think to say 'relatively ineffective' puts a completely different and unwarranted slant on the piece, as that's simply not what's being said in it....


OK, I’ll accept that, yes.

It’s a nuance, because to me they’re almost equivalent, but they are indeed not the same.


I think in some non-pandemic contexts, the difference becomes much less of an issue, but if during a global pandemic that's already killed 3.5 million people, someone who'd had the AZ vaccine read someone saying that they'd had a version of the vaccine that was 'relatively ineffective', it's likely to cause some real alarm and concern, and especially if they'd perhaps just read a headline on the BBC site that said -

The Pfizer and AstraZeneca coronavirus vaccines are highly effective against the variant identified in India after two doses, a study has found.

https://www.bbc.co.uk/news/uk-57214596

The study you linked to was primarily discussing the 'effectiveness gap' between the initial relatively low protection straight after a single jab, and the end-result of the higher two-dose protection, and the author was concerned that the Indian variant might be becoming both dominant and prevalent in the UK at a time where it might well be able to take advantage of that 'effectiveness gap', either in relation to those that have yet to have any initial dose, or those that are between doses and not yet at 'higher effectiveness'....

They were saying that steps could be taken to counter that 'effectiveness gap', one of which is already being taken by the now-shortened delivery times between 1st and 2nd jabs for those in the most vulnerable groups.

Cheers,

Itsallaguess


Return to “Coronavirus Discussions”

Who is online

Users browsing this forum: No registered users and 4 guests