Donate to Remove ads

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

Thanks to eyeball08,Wondergirly,bofh,johnstevens77,Bhoddhisatva, 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
zico
Lemon Quarter
Posts: 2145
Joined: November 4th, 2016, 12:12 pm
Has thanked: 1078 times
Been thanked: 1091 times

Re: Coronavirus - Modelling Aspects Only

#298927

Postby zico » April 8th, 2020, 11:39 am

This is a shocking paragraph from the blame-shifting report you've linked to.
Apparently saying "one in a hundred Britons might die" isn't spelling out clearly enough that 1% of 67 million = 670,000.
Maybe saying four-fiths of Britons would be infected is also not spelling out that 80% of 67 million = 53 million people.


If unconstrained and if the virus behaved as in China, up to four-fifths of Britons could be infected and one in a hundred might die, wrote the scientists, members of an official committee set up to model the spread of pandemic flu, on March 2. Their assessment didn’t spell it out, but that was a prediction of over 500,000 deaths in this nation of nearly 70 million.

spasmodicus
Lemon Slice
Posts: 261
Joined: November 6th, 2016, 9:35 am
Has thanked: 65 times
Been thanked: 117 times

Re: Coronavirus - Modelling Aspects Only

#298949

Postby spasmodicus » April 8th, 2020, 12:21 pm

This is a shocking paragraph from the blame-shifting report you've linked to.
Apparently saying "one in a hundred Britons might die" isn't spelling out clearly enough that 1% of 67 million = 670,000.
Maybe saying four-fiths of Britons would be infected is also not spelling out that 80% of 67 million = 53 million people.


If unconstrained and if the virus behaved as in China, up to four-fifths of Britons could be infected and one in a hundred might die, wrote the scientists, members of an official committee set up to model the spread of pandemic flu, on March 2. Their assessment didn’t spell it out, but that was a prediction of over 500,000 deaths in this nation of nearly 70 million.


Yes, quite so. Quite early on in February reports from China were suggesting a death rate of around 2%. You don't have to be a health expert to calculate the number of potential deaths if that is the case and everyone gets infected. However nobody really knew the true figure for the total number of infected people, capable or not of transmitting the virus. They only had vague figures for the number of infections, based on hospital admissions. In Italy, for example
On 26 February, Director of the Italian National Institute of Health Franco Locatelli announced that swabbing would only be performed on symptomatic patients, as 95% of the swabs previously tested were negative.[195]
(Wikipedia). The only hard data then, as now, are deaths and even these figures are unreliable as they sometimes exclude people in care homes, or include people who were going to die anyway from something else. And of course the deaths lag the corresponding infections by about 3 or 4 weeks.

Until widespread, random testing is done nobody will really know whether it is safe to lift restrictions.

S

dspp
Lemon Half
Posts: 5884
Joined: November 4th, 2016, 10:53 am
Has thanked: 5825 times
Been thanked: 2127 times

Re: Coronavirus - Modelling Aspects Only

#299019

Postby dspp » April 8th, 2020, 3:55 pm

Well CV-19 can be pretty deadly if you are in the wrong cohort

"Canada nursing home reels from death of almost half its residents [27 out of 65]"

https://www.theguardian.com/world/2020/ ... bobcaygeon

dspp

zico
Lemon Quarter
Posts: 2145
Joined: November 4th, 2016, 12:12 pm
Has thanked: 1078 times
Been thanked: 1091 times

Re: Coronavirus - Modelling Aspects Only

#299072

Postby zico » April 8th, 2020, 6:56 pm

JamesMuenchen wrote:Interesting data from ONS

https://twitter.com/NickStripe_ONS/stat ... 68837?s=19

W/e 27.3 total UK deaths are up by 1000 on the 5 year average.



Thanks for this link. I've linked below to a FT piece by Chris Giles (hopefully you can open it).
It's very bad news indeed because the true daily deaths figure is 78% higher than what's reported each day, so the 600-700 deaths reported are actually 1,050-1,250.

The ONS figures include all reported Covid-19 deaths (including nursing homes which only account for 7% of deaths - ONS expect this %age to increase).
However, the daily reported deaths figures don't have all deaths registered, so lag by about 3 days.
28th March - Govt announced 928 Covid-19 deaths in hospitals in England.
Deaths registered up to 27th March - 1,568 had Covid-19 mentioned on death certificate
England hospitals reconciled their figures with actual date of death, which gives 1,649 Covid-19 deaths by 27th March.

Not all deaths will yet be registered so ONS expect the 78% figure to go even higher when all deaths are in.

https://www.ft.com/content/c07e267b-7bc ... 37841d178a

Nimrod103
Lemon Half
Posts: 6600
Joined: November 4th, 2016, 6:10 pm
Has thanked: 969 times
Been thanked: 2315 times

Re: Coronavirus - Modelling Aspects Only

#299189

Postby Nimrod103 » April 9th, 2020, 9:11 am

zico wrote:
JamesMuenchen wrote:Interesting data from ONS

https://twitter.com/NickStripe_ONS/stat ... 68837?s=19

W/e 27.3 total UK deaths are up by 1000 on the 5 year average.



Thanks for this link. I've linked below to a FT piece by Chris Giles (hopefully you can open it).
It's very bad news indeed because the true daily deaths figure is 78% higher than what's reported each day, so the 600-700 deaths reported are actually 1,050-1,250.

The ONS figures include all reported Covid-19 deaths (including nursing homes which only account for 7% of deaths - ONS expect this %age to increase).
However, the daily reported deaths figures don't have all deaths registered, so lag by about 3 days.
28th March - Govt announced 928 Covid-19 deaths in hospitals in England.
Deaths registered up to 27th March - 1,568 had Covid-19 mentioned on death certificate
England hospitals reconciled their figures with actual date of death, which gives 1,649 Covid-19 deaths by 27th March.

Not all deaths will yet be registered so ONS expect the 78% figure to go even higher when all deaths are in.

https://www.ft.com/content/c07e267b-7bc ... 37841d178a


Agreed, this is due to a backlog in reporting. The article is based on only March figures up to the 27th, so I am not clear how much of a delay there still is. I note that already in the death data, there is a clear depression over weekends due to reporting problems, but this seems to be caught up with by the following Tuesday.

The effects of this backlog will be balanced when the rate of deaths declines, due to the effectiveness of the lockdown. The actual rate of decline will be steeper than the announced figures would suggest.

SalvorHardin
Lemon Quarter
Posts: 2062
Joined: November 4th, 2016, 10:32 am
Has thanked: 5358 times
Been thanked: 2485 times

Re: Coronavirus - Modelling Aspects Only

#299225

Postby SalvorHardin » April 9th, 2020, 10:17 am

The Economist has just posted an article referring to a new American study which estimates that the mortality rate of the Wuhan Coronavirus is about 0.1%. That's because rate at which people have been showing the symptoms of an "Influenze-Like Illness" (ILI) has rocketed for people who do not have the flu (i.e. they have the Coronavirus).

"They find that non-flu ILI has surged. Its rise has the same geographic pattern as covid-19 cases: modest in states with few positive tests, like Kentucky, and steep in ones with big outbreaks, such as New Jersey. In total, estimated non-flu ILI from March 8th to 28th exceeded a historical baseline by 23m cases—200 times the number of positive covid-19 tests in that period. This may overstate the spread of covid-19, since non-flu ILI has other causes. It could also be too low, because people with asymptomatic or mild covid-19 would not report non-flu ILI."

https://www.economist.com/graphic-detail/2020/04/11/why-a-study-showing-that-covid-19-is-everywhere-is-good-news?fsrc=scn/tw/te/bl/ed/footprintsoftheinvisibleenemywhyastudyshowingthatcovid19iseverywhereisgoodnewsgraphicdetail

No surprise to me. The official mortality rates have been consistently overestimated because the recorded cases are dominated by tests made upon those who have been admitted to hospital. The group which has been admitted to hospital is not representative of the general population.

In countries where their testing for the Wuhan Coronavirus has tried to sample the general population, notably Germany and Iceland, mortality rates have consistently been dramatically lower than those in countries which haven't been randomly sampling the population.

zico
Lemon Quarter
Posts: 2145
Joined: November 4th, 2016, 12:12 pm
Has thanked: 1078 times
Been thanked: 1091 times

Re: Coronavirus - Modelling Aspects Only

#299300

Postby zico » April 9th, 2020, 1:30 pm

SalvorHardin wrote:The Economist has just posted an article referring to a new American study which estimates that the mortality rate of the Wuhan Coronavirus is about 0.1%. That's because rate at which people have been showing the symptoms of an "Influenze-Like Illness" (ILI) has rocketed for people who do not have the flu (i.e. they have the Coronavirus).

"They find that non-flu ILI has surged. Its rise has the same geographic pattern as covid-19 cases: modest in states with few positive tests, like Kentucky, and steep in ones with big outbreaks, such as New Jersey. In total, estimated non-flu ILI from March 8th to 28th exceeded a historical baseline by 23m cases—200 times the number of positive covid-19 tests in that period. This may overstate the spread of covid-19, since non-flu ILI has other causes. It could also be too low, because people with asymptomatic or mild covid-19 would not report non-flu ILI."

https://www.economist.com/graphic-detail/2020/04/11/why-a-study-showing-that-covid-19-is-everywhere-is-good-news?fsrc=scn/tw/te/bl/ed/footprintsoftheinvisibleenemywhyastudyshowingthatcovid19iseverywhereisgoodnewsgraphicdetail

No surprise to me. The official mortality rates have been consistently overestimated because the recorded cases are dominated by tests made upon those who have been admitted to hospital. The group which has been admitted to hospital is not representative of the general population.

In countries where their testing for the Wuhan Coronavirus has tried to sample the general population, notably Germany and Iceland, mortality rates have consistently been dramatically lower than those in countries which haven't been randomly sampling the population.



Yes, official mortality rates are pretty pointless at the moment for the reasons you give, and also because there isn't yet a cohort of people going through all the stages of the disease, including full recovery.
The other point about "mortality rates" is that probably no country is picking up every new infections. UK "new cases" are less than 10% of actual new infections (and hopefully much lower). Germany is testing far more people and so will pick up a higher percentage of actual new infections, but there's no reason to assume they will therefore have a lower death rate - it's the rate of spread that's the key factor.

Couldn't read the article (paywall) but looking at your quote from the article, 23m people with flu-like symptoms would translate into 46m people with Covid-19 (using the estimate of 50% asymptomatic which seems to be the generally accepted figure).
Working back from 46m people, and assuming the cases double every 2 cases, then the first case would have arrived in USA 2 1/2 months ago, which seems plausible. A US state (e.g. Montana) that had its first case 10 days later than this would only have about 3% of the cases of the worst affected state, so again this is plausible with what we actually do know about the rate of spread.

If the 0.1% estimate is plausible, that would imply 67,000 UK deaths in total, even with no lockdown, which is vastly below the government's estimate of 500,000 deaths with no action. If the 0.1% is actually 0.15% (but rounded down to 0.1%) then that's still 100,00 - just a fifth of the UK government's estimate. I would like it to be true, but it doesn't seem to be in the right ball park.

SalvorHardin
Lemon Quarter
Posts: 2062
Joined: November 4th, 2016, 10:32 am
Has thanked: 5358 times
Been thanked: 2485 times

Re: Coronavirus - Modelling Aspects Only

#299305

Postby SalvorHardin » April 9th, 2020, 1:45 pm

zico wrote:Couldn't read the article (paywall) but looking at your quote from the article, 23m people with flu-like symptoms would translate into 46m people with Covid-19 (using the estimate of 50% asymptomatic which seems to be the generally accepted figure).
Working back from 46m people, and assuming the cases double every 2 cases, then the first case would have arrived in USA 2 1/2 months ago, which seems plausible. A US state (e.g. Montana) that had its first case 10 days later than this would only have about 3% of the cases of the worst affected state, so again this is plausible with what we actually do know about the rate of spread.

If the 0.1% estimate is plausible, that would imply 67,000 UK deaths in total, even with no lockdown, which is vastly below the government's estimate of 500,000 deaths with no action. If the 0.1% is actually 0.15% (but rounded down to 0.1%) then that's still 100,00 - just a fifth of the UK government's estimate. I would like it to be true, but it doesn't seem to be in the right ball park.

Yes, 0.1% is probably a bit optimistic. But the lockdown will buy us time and there is a lot of effort currently being put into antiviral drugs. If one of these proves effective on late-stage patients (my money is on Remdesivir) then that will drastically drop the mortality rate. Then we wait for the vaccine(s). Moderna is seriously talking about September for full scale trials for its vaccines.

Sorry about the paywall. I forgot that the Economist operates one for non-subscriber. I believe that you can register as a non-subscriber and get a few articles per month. The original paper can be found here:

https://www.medrxiv.org/content/10.1101/2020.04.01.20050542v1

zico
Lemon Quarter
Posts: 2145
Joined: November 4th, 2016, 12:12 pm
Has thanked: 1078 times
Been thanked: 1091 times

Re: Coronavirus - Modelling Aspects Only

#299310

Postby zico » April 9th, 2020, 1:54 pm

I'm just adding a couple of links that were posted in Beerpig's Snug about modelling, because I think they'll be more useful to inform discussion here.

sq31 posted the following.

Coronavirus: low antibody levels raise questions about reinfection risk

https://www.scmp.com/news/china/science ... ions-about

Researchers in Shanghai hope to determine whether some recovered coronavirus patients
have a higher risk of reinfection after finding surprisingly low levels of Covid-19 antibodies in a number of people discharged from hospital.

A team from Fudan University analysed blood samples from 175 patients discharged from the Shanghai Public Health Clinical Centre and found that nearly a third had unexpectedly low levels of antibodies.

In some cases, antibodies could not be detected at all.

“Whether these patients were at high risk of rebound or reinfection should be explored in further studies,” the team wrote in preliminary research released on Monday on Medrxiv.org, an online platform for preprint papers.

Although the study was preliminary and not peer-reviewed, it was the world’s first systematic examination of antibody levels in patients who had recovered from Covid-19, the disease caused by the coronavirus, the researchers said.

Simoan posted the following.

A friend of mine sent a link to a very interesting article by Prof. Julia Gog of Cambridge University. She is currently working full-time for SAGE. The article covers the basic maths of the pandemic and possible future scenarios based on a series of lockdowns: https://plus.maths.org/content/how-can- ... a-pandemic

Warning: it's not a particularly positive conclusion.
All the best, Si

(I've looked at this paper and can't understand the graphs in it at all)

GoSeigen
Lemon Quarter
Posts: 4406
Joined: November 8th, 2016, 11:14 pm
Has thanked: 1603 times
Been thanked: 1593 times

Re: Coronavirus - Modelling Aspects Only

#299325

Postby GoSeigen » April 9th, 2020, 2:30 pm

SalvorHardin wrote:In countries where their testing for the Wuhan Coronavirus has tried to sample the general population, notably Germany and Iceland, mortality rates have consistently been dramatically lower than those in countries which haven't been randomly sampling the population.


The Iceland thing has been quoted widely, including by people who should know better. Any figure derived from that study should IMO be treated with huge skepticism, for the simple reason that the virus had barely to started spread through the population at the time of the study. When the infection rate gets to around 5-10% of the population we can start to take note, but by then it will may be too late of course.

Why not accept that mortality rates are unknown and just look at death figures which tell their own story.


GS

dealtn
Lemon Half
Posts: 6091
Joined: November 21st, 2016, 4:26 pm
Has thanked: 442 times
Been thanked: 2338 times

Re: Coronavirus - Modelling Aspects Only

#299327

Postby dealtn » April 9th, 2020, 2:41 pm

GoSeigen wrote:
Why not accept that mortality rates are unknown and just look at death figures which tell their own story.



Because they are, very much, a lagging indicator, so have very little predictive power on how the pandemic is developing, nor on how any measures introduced to deal with it are working (or not).

Nimrod103
Lemon Half
Posts: 6600
Joined: November 4th, 2016, 6:10 pm
Has thanked: 969 times
Been thanked: 2315 times

Re: Coronavirus - Modelling Aspects Only

#299343

Postby Nimrod103 » April 9th, 2020, 3:12 pm

dealtn wrote:
GoSeigen wrote:
Why not accept that mortality rates are unknown and just look at death figures which tell their own story.



Because they are, very much, a lagging indicator, so have very little predictive power on how the pandemic is developing, nor on how any measures introduced to deal with it are working (or not).


It is true that deaths are a lagging indicator. From the Telegraph today:

Of the 765 new hospital deaths announced today by NHS England, 140 occurred on April 8 while 568 took place between April 1 and April 7.
The remaining 57 deaths occurred in March, including two on March 19 and one on March 16


It is the trend that matters, and I imagine the experts have access to much more up to date data and predictions.

alphab1
Lemon Pip
Posts: 93
Joined: November 6th, 2016, 5:32 pm
Has thanked: 8 times
Been thanked: 5 times

Re: Coronavirus - Modelling Aspects Only

#299362

Postby alphab1 » April 9th, 2020, 4:10 pm

Just a query:
Unless the daily number of patients tested for the virus is roughly constant, will it not be more appropriate to use the percentage of positive cases rather than the actual number of positive cases? From what I have found, all the charts from reputable organizations provide the actual number of positive cases but not the sample size.

Since only a small percentage of the population is being tested on a day the total number of cases can change quite erratically if the daily sample size changes - which is quite likely.

It is the curve of daily percentage of positive cases that should be monitored to see if it has been flattened sufficiently.

Regarding the criterion for easing the lockdown can we not monitor how the spare capacity of the hospitals increases as the 'curve' is flattened, and set a target percentage when a week's notice for lifting the lockdown is given. If the situation deteriorates, the decision can be withdrawn.

Regards.

zico
Lemon Quarter
Posts: 2145
Joined: November 4th, 2016, 12:12 pm
Has thanked: 1078 times
Been thanked: 1091 times

Re: Coronavirus - Modelling Aspects Only

#299460

Postby zico » April 9th, 2020, 11:23 pm

alphab1 wrote:Just a query:
Unless the daily number of patients tested for the virus is roughly constant, will it not be more appropriate to use the percentage of positive cases rather than the actual number of positive cases? From what I have found, all the charts from reputable organizations provide the actual number of positive cases but not the sample size.

Since only a small percentage of the population is being tested on a day the total number of cases can change quite erratically if the daily sample size changes - which is quite likely.


For many days, the total numbers tested varied little from the 9,000 mark, so it was a good proxy for total cases, and indeed, my model has predicted reasonably well. Howeve, in the last few days, the numbers tested have increased to 12,000-13,000 so you'd expect the positive cases to have increased, simply because more tests are being carried out.

alphab1 wrote:It is the curve of daily percentage of positive cases that should be monitored to see if it has been flattened sufficiently.


The daily percentage of positive cases has varied quite a lot from day to day. We don't know the criteria for doing the tests, but it seems to change from day to day. I expect that some days, there are far more NHS staff in ICUs being routinely tested, or something similar which means you can't rely on the percentage of positive tests.

zico
Lemon Quarter
Posts: 2145
Joined: November 4th, 2016, 12:12 pm
Has thanked: 1078 times
Been thanked: 1091 times

Re: Coronavirus - Modelling Aspects Only

#299464

Postby zico » April 9th, 2020, 11:37 pm

Official figures were released unusually late today, around 8pm.
I couldn't reconcile the number of total tests with the 16,095 tests given in the figures.
New positive cases 4,344
Daily Deaths - 881

More odd things in today's press conference. 3 graphs were shown -
1st was for new cases, where it was claimed peak new cases have already happened - the graph didn't support that.
2nd was for hospitalisations, where it was claimed the rate of increase was slowing - though not that obvious to the naked eye.
3rd was for critical cases in hospitals, where it was claimed these have now flatlined.
The problem with the above is that with the progress of the diseases (identification, then hospitalisation, then critical care) you'd expect to see the first flatline in new cases, followed a few days later by flatlining hospital admissions, then a few days later by flatlining critical cases. But the critical cases seem to have flatlined first.

Vallance told us the peak of daily deaths is expected in 2 weeks time, but at the start of April, govt sources said the peak would be on Easter Sunday. That's 10 days later, which is a big difference, and no explanation why there should have been such a change, or why it should be so different from the pattern seen in Italy and France. Obviously Vallance has access to much better data quality and consistency than what's published, so I'm sure he's right, but I don't know why the peak should have been delayed so much. Possibly, UK modellers expected to replicate Italy results, but UK lockdown is less rigorous? Just speculation, but it's odd all the same.

Whitty was asked about the weekly ONS Covid-19 death stats, and said they would be more complete and accurate, also adding that the ONS stats would include deaths where a doctor thought it was Covid-19, but didn't have a positive test carried out.

Looks like the UK is going to have significantly more deaths in this first wave than were originally expected.

Itsallaguess
Lemon Half
Posts: 9129
Joined: November 4th, 2016, 1:16 pm
Has thanked: 4140 times
Been thanked: 10025 times

Re: Coronavirus - Modelling Aspects Only

#299487

Postby Itsallaguess » April 10th, 2020, 6:28 am

zico wrote:
More odd things in today's press conference. 3 graphs were shown -

1st was for new cases, where it was claimed peak new cases have already happened - the graph didn't support that.

2nd was for hospitalisations, where it was claimed the rate of increase was slowing - though not that obvious to the naked eye.

3rd was for critical cases in hospitals, where it was claimed these have now flat-lined.

The problem with the above is that with the progress of the diseases (identification, then hospitalisation, then critical care) you'd expect to see the first flat-line in new cases, followed a few days later by flat-lining hospital admissions, then a few days later by flat-lining critical cases.

But the critical cases seem to have flat-lined first.


I saw the press conference and initially thought the same as you, regarding the 'critical-cases in hospitals' figures looking like flatlining earlier, and then I thought that this would make complete sense if the section of the community that might end up in that particular section of the statistics were self-isolating to a much more stringent degree than those that might still become poorly, but not necessarily then go on to need hospital treatment...

On the flip side of that, you'd perhaps then expect more and more of the actual 'daily new cases' figures to be more strongly represented over time by people who are clearly still catching it, but perhaps then tend to be less and less reliant on hospital treatment, given that they might not be 'taking that extra level of care' simply because they don't consider themselves to be in that 'extra-vulnerable group' that might then end up requiring hospital treatment, or because many of them are perhaps part of the relatively younger section of society that largely forms our current 'critical services' workers, who by default will be exposed to more risk of catching the virus on a daily basis, simply by having to carry out those 'critical services'.

If we were to perhaps compare those two 'less-susceptible-to-needing-hospital-treatment' groups of people with the 'older and much more susceptible group' that are being told to carry out much more stringent self-isolation processes, then we might expect those 'critical-cases in hospitals' figures to start to drop off quite early on, as a higher and higher proportion of the less-susceptible (to needing hospital treatment..) group continue feed into the 'daily new cases' figures..

In a way, the above scenario is how I would expect any future 'herd immunity' approach to have to play out - where the ultra-susceptible (in terms of potentially requiring hospital treatment...) are told to stay in a different 'situation' than those that are not as likely to need hospital treatment if and when they were to go on to be affected by the virus, and in that scenario we might see exactly what you're scratching your head about here, where 'new cases' continues to rise, but with 'critical hospital cases' continuing to flat-line or even drop-off, so long as that ultra-susceptible (to requiring hospital treatment) cohort were to continue with their very stringent self-isolation routines...

Given that in the absence of any type of reliable vaccine, the above scenario is also likely to be one that might get the economy going again using what we might consider to be the largely more 'economically active' and also large sections of the 'working population', where the younger/fitter/still-working might largely be considered to be that section of society that are much less likely to need hospital treatment, I would expect the above 'two-tiered' approach to be something they might go on to look at when they ever begin to walk-back on the current lock-down arrangements...

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

#299507

Postby johnhemming » April 10th, 2020, 8:52 am

The figure on cases depends primarily on testing. Hence it really does not provide much useful information.

The figure on admissions is the best predictor remembering that it needs a seasonable adjustment to work out what is actually happening.

Most useful are indicators as to what proportion of the population have had the disease
https://www.telegraph.co.uk/news/2020/0 ... man-study/

The above German study indicates 15%. (for a particular town).

The only useful information we really have on this is a mori poll that said that 20% of people thought they have probably had the virus.

Nimrod103
Lemon Half
Posts: 6600
Joined: November 4th, 2016, 6:10 pm
Has thanked: 969 times
Been thanked: 2315 times

Re: Coronavirus - Modelling Aspects Only

#299508

Postby Nimrod103 » April 10th, 2020, 8:57 am

johnhemming wrote:The figure on cases depends primarily on testing. Hence it really does not provide much useful information.

The figure on admissions is the best predictor remembering that it needs a seasonable adjustment to work out what is actually happening.

Most useful are indicators as to what proportion of the population have had the disease
https://www.telegraph.co.uk/news/2020/0 ... man-study/

The above German study indicates 15%. (for a particular town).

The only useful information we really have on this is a mori poll that said that 20% of people thought they have probably had the virus.


I am unclear how the Germans have done this study, given that we have been told the antibody test is unreliable (at the moment).

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

#299510

Postby johnhemming » April 10th, 2020, 9:29 am

a) It depends which antibody test
b) It is worth spending some serious cash to get some figures
c) Information in this area has statistical validity even if it is not 100% reliable on an individual basis.

Nimrod103
Lemon Half
Posts: 6600
Joined: November 4th, 2016, 6:10 pm
Has thanked: 969 times
Been thanked: 2315 times

Re: Coronavirus - Modelling Aspects Only

#299516

Postby Nimrod103 » April 10th, 2020, 9:42 am

johnhemming wrote:a) It depends which antibody test
b) It is worth spending some serious cash to get some figures
c) Information in this area has statistical validity even if it is not 100% reliable on an individual basis.


Yes, but. AIUI the fingerprick test which every Govt wants to rely on in order to lift the lockdown, reportedly has a 30% reliability at present. And Van Tam's comment a week ago was that it takes up to a month for people to develop measurable antibodies. And there was also a study out of China (which may or may not be a lie) that they had identified a significant number who had recovered from the infection, who had no or only very low antibody levels.

With all medical studies I have ever read, I never get a good idea of the range of uncertainties. For instance in drug trials I have occasionally read about in newspapers over the years, reserchers publish about trials on as few as a dozen people, when I would have expected a sample size of thousands.


Return to “Coronavirus Discussions”

Who is online

Users browsing this forum: No registered users and 23 guests