Coronavirus - Modelling Aspects Only

[servodude]

I think there's a new acceleration happening now... looking at 12000 deaths pd within two to three weeks time IMO.

GS

Worldwide you mean, presumably? (And your "pd" = per day?)

Moving on, something about this pandemic is puzzling me. Normally in an epidemic, huge effort is put into tracing 'Patient Zero'. Identifying Patient Zero tells us buckloads apparently, about the progress of the infection and all manner of other things.

But with SARS-CoV-2, nada. I've not seen or heard a single mention of any attempt to find and identify Patient Zero since this thing first caught my attention back around Christmas. Why is that?

Could it be that there is no need because all the guvverments know already exactly where and it came from and when, and are keeping it a collective secret for political reasons? Is that a great conspiracy theory or what?

But seriously, why does Patient Zero never get mentioned? Or is it such common knowledge that it doesn't need mentioning, in which case who were they, and where and when and how did they catch it, exactly?

I saw a few reports a while ago regarding this: e.g
https://7news.com.au/lifestyle/health-w ... o-c-766471

-sd

[swill453]

Moving on, something about this pandemic is puzzling me. Normally in an epidemic, huge effort is put into tracing 'Patient Zero'. Identifying Patient Zero tells us buckloads apparently, about the progress of the infection and all manner of other things.

Is that not just in the movies?

Actually, I seem to remember in Outbreak or Contagion or something, they spent a huge effort locating Patient Zero, but in the end it didn't actually help with anything.

Scott.

[johnhemming]

There has been quite a bit of press reporting on the question when the first infections arrived in the UK. The government however, have relied on the wrong modelling and introduced policies because they are popular in the short term rather than effective in the long term.

Matt Ridley's article is good, but where he says

And science needs to rethink its affair with models rather than data.

It isn't so much science as the government and civil service.

[GoSeigen]

I think there's a new acceleration happening now... looking at 12000 deaths pd within two to three weeks time IMO.

GS

Worldwide you mean, presumably? (And your "pd" = per day?)

Yes and yes.

But seriously, why does Patient Zero never get mentioned? Or is it such common knowledge that it doesn't need mentioning, in which case who were they, and where and when and how did they catch it, exactly?

Been watching films? Try reading this:

Patient zero: why it's such a toxic term

GS

[jfgw]

It is too early to compare the effects of different levels of lockdown. Countries that have had a more relaxed lockdown may get the whole thing over and done with a lot sooner than other countries, and it is possible that countries with a stricter lockdown may experience a second wave where we do not. It will be interesting to see what happens in the next few weeks in the UK now that the lockdown has been partly relaxed.

Julian F. G. W.

I see what you're saying but I don't entirely agree. The first wave was the crucial time when deaths increased at a very fast rate, and hospitals might be overwhelmed (as happened in Italy) or care homes might be overwhelmed (as has happened in the majority of countries).
From now on, the pace of infection will depend on lockdown measures, track'n'trace and other measures, and countries have had time to look, learn and improve their responses. There won't be any sudden increases in infection and deaths, and everyone's attention is now focused on Covid-19.

The crucial point about squashing the first wave as much as possible is that -

a) If you do it really effectively, then you stop the virus taking hold or alternatively quickly eradicate it from your country, saving lives and the economy. New Zealand did this really well, as did a few other countries.

b) If you don't do a) above but then lockdown early enough, you can reduce the first wave numbers, have far fewer deaths initially, and a vaccine may well arrive before most of the population become infected, and there may be other medical developments and treatments to reduce fatalities. Most European countries reacted in this way.

c) If you have a big first wave (like UK, Brazil, Italy,Spain), then you have a lot of people (sadly) dead, and you also need a longer lockdown to get infection numbers down to a point where you have a decent chance of being able to contain the virus whilst also opening up the economy. If you rush coming out of lockdown, you risk wasting all the economic pain of the lockdown and putting yourself back to square one.

Although I didn't make it clear, my post was, in part, in response to fca2019's BoJo-bashing post.

I accept that delaying infections can reduce the peak demand on healthcare services to a manageable level and, in that way, can save lives. It also provides more time to learn about the disease and how best to treat it. I stand by my original stance that the final numbers are not yet in, however. I pointed out weeks ago in The Snug that the numbers of infections in the UK have been sub-exponential since about March 4th, before any lockdown. (While the theoretical curve is not exponential, it is pretty close until a significant proportion of the susceptible population have been infected, and it is easy to get an idea of how close a curve is to being exponential by looking at a log-linear graph.) I have never accepted that we needed most of the population to become infected and produce antibodies for the disease to die out naturally - the notion does not fit the graphs.

Now imagine a second wave far more deadly than the first (think Spanish flu). The countries that have been most successful in controlling the first wave will be the most susceptible to the second. At present, we don't know the final outcome. Until the virus is fully controlled or has died out naturally, we cannot judge for certain which approach has been most successful. We can speculate but we cannot predict with certainty.

Julian F. G. W.

[Mike4]

I think there's a new acceleration happening now... looking at 12000 deaths pd within two to three weeks time IMO.

GS

Worldwide you mean, presumably? (And your "pd" = per day?)

Yes and yes.

But seriously, why does Patient Zero never get mentioned? Or is it such common knowledge that it doesn't need mentioning, in which case who were they, and where and when and how did they catch it, exactly?

Been watching films? Try reading this:

Patient zero: why it's such a toxic term

GS

Ok thanks, so its rubbish then. And no I'm not a 'film' person! The main point of films as far as I'm concerned is they present a good opportunity for a nice sleep while everyone else is captivated.

I guess the Patient Zero thing occurred to me because its a beguilingly simple idea, a bit likw the game kids play of "what's the biggest number you can think of?" Whatever one kid says, another says with glee, "that, PLUS ONE".

Take an infected person and look to see who they caught it from. Do this again and again until you run out of anyone who could have infected them. Naïve, on reflection.

[johnhemming]

Heres a bit more hunting for patient zero (or when did this actually start)
https://www.thetimes.co.uk/edition/news ... -6bfrb0ggk

Thousands of people have emailed me with classic Covid symptoms from late December and January,” said Professor Tim Spector, a leading epidemiologist at King’s College London, who runs the Covid-19 Symptom Study app to which 3.8 million people have signed up.

“Either there was another virus behaving in a similar way which has since disappeared or these were early cases.”

If so, why was it not reflected in a spike in hospital admissions or deaths? “That’s the medical mystery,” said Spector.

It now seems likely that Walsh was not the UK’s “patient zero”. A month earlier, Susannah Ford, 53, director of the Forward Arts Foundation, which promotes poetry, had fallen ill after flying back from a skiing holiday in Austria.

The mother of two, from west London, became ill on January 6, two days after her return from a new year trip. “I felt like death,” she said.

Last week Ford paid for a test that shows whether the patient’s blood contains the antibodies that form when a person successfully fights off the disease. It came back positive, confirming that she had had Covid-19, although not when.

This is perhaps more signifcant as the person concerns has proof that they had the virus. Although they cannot prove it was early January that has to be the assumption. This is therefore reasonably reliable evidence and possibly proof (on the balance of probabilities) that the disease was infecting people in early January.

From a modelling point of view, therefore, the Oxford Gupta modelling has the right starting assumption on timing rather than the Imperial Modelling.

[Bubblesofearth]

Take an infected person and look to see who they caught it from. Do this again and again until you run out of anyone who could have infected them. Naïve, on reflection.

It is, however, incredibly important to understand how zoonotic diseases spread to humans. Just as it was important to finally understand the vectors involved in the Black Death.

Only when we understand how and why such diseases occur will we be able to attempt to prevent them in the future. It's not about finding patient zero but identifying practices that are likely to lead to him/her. This detective work should not be hindered by politics. Sadly it probably will be.

BoE

[zico]

I've been looking into conflicting views of the R-value. Basically, Cambridge study and London Tropical Medicine studies estimate R values between 0.9-1 while SAGE has decided it should be 0.7-0.9

Looks to me as though SAGE has given a lot of weight to the ONS pilot sample of 20,000 people which shows a clear downward trend over the last month. See Figure 2 in the attached link below for details.

I've dug a little further into the ONS report - which covers Covid prevalence up to 30th May.

Firstly, they estimate the percentage of England's population currently infected with Covid-19 based on a sample of 19,723 individual's swab tests. Of these, for the latest week, 21 individuals tested positive from 15 households. That's not very many people.
(Over the 4 weeks of the study, positive numbers were 61,43,31,21 - so a statistically significant fall over 4 weeks)
To estimate the percentage of England's population who have ever had Covid-19, they tested for antibodies. 885 individuals (6.78%) tested positive for antibodies.
Only 29% of people who tested positive reported any symptoms (analysis based on 88 individuals who tested positive).

The same people do a self-test using swabs each week, and a blood sample is also taken. This isn't a random sample of the population, as it's a fair assumption it's a lot more likely that these people will be more cautious than others about exposing themselves to possible Covid-19 infection.
ONS don't have a random sample. 20,000 households were asked to take part, just under 50% agreed.

My conclusion is that the ONS sample is far more likely than the general population to be observing lockdown rules and taking appropriate precautions, so it isn't representative of the general population, and likely to be a big underestimate of the actual spread of the disease.
It also seems that 90% of those sampled were infected before April 26th - which seems to fit with what we know about the spread.

https://www.ons.gov.uk/peoplepopulation ... /5june2020

[tjh290633]

But seriously, why does Patient Zero never get mentioned? Or is it such common knowledge that it doesn't need mentioning, in which case who were they, and where and when and how did they catch it, exactly?

He does, a Stuart something or other who came back from the Far East to a skiing holiday in Italy and then brought it home. It's old news now.

TJH

[swill453]

But seriously, why does Patient Zero never get mentioned? Or is it such common knowledge that it doesn't need mentioning, in which case who were they, and where and when and how did they catch it, exactly?

He does, a Stuart something or other who came back from the Far East to a skiing holiday in Italy and then brought it home. It's old news now.

And he infected all of China too? Wow.

Scott.

[XFool]

Looks like Nic Lewis's model or more accurately Gabriela Gomes whose paper Nic cites has gone mainstream. Mat Ridley talks about it here :
https://www.telegraph.co.uk/news/2020/0 ... r-history/

Has the British scientific establishment made its biggest error in history?

A strange obsession with mathematical modelling has compromised the country’s covid response

MATT RIDLEY

Um... As I have not got access to the full article I will be using that as an excuse for not giving here my full and likely blunt comment on Ridley's piece.

"A strange obsession with mathematical modelling"? But is not mathematical modelling what you've got, unless you simply want to fly blind and wait and see. But then Ridley was "flying blind" at Northern Rock - so why not?

[Mike4]

But seriously, why does Patient Zero never get mentioned? Or is it such common knowledge that it doesn't need mentioning, in which case who were they, and where and when and how did they catch it, exactly?

He does, a Stuart something or other who came back from the Far East to a skiing holiday in Italy and then brought it home. It's old news now.

TJH

You're not really getting what is meant by Patient Zero, a more correct term being the index case.

As Scott points out, Stuart Wossname was long pre-dated by many thousands ill and dying in China. There is some anecdotal evidence of early cases in China back in November and prior to that even, including some infection at the Miltary Games in Wuhan back in October. Many athletes fell ill with what now look like COVID-like symptoms yet these are being studiously ignored in the MSM. I wonder if those athletes had been antibody tested and what the result was. No-one is saying.

[johnhemming]

yet these are being studiously ignored in the MSM.

The MSM in the UK are looking for early cases in the UK. The timing of the first cases in the UK is critical for the modelling systems being used.

[johnhemming]

He does, a Stuart something or other who came back from the Far East to a skiing holiday in Italy and then brought it home. It's old news now.

I think it is pretty well accepted that there were infections in the UK some time before he got infected.

[zico]

Looks like Nic Lewis's model or more accurately Gabriela Gomes whose paper Nic cites has gone mainstream. Mat Ridley talks about it here :
https://www.telegraph.co.uk/news/2020/0 ... r-history/

Has the British scientific establishment made its biggest error in history?

A strange obsession with mathematical modelling has compromised the country’s covid response

MATT RIDLEY

Um... As I have not got access to the full article I will be using that as an excuse for not giving here my full and likely blunt comment on Ridley's piece.

"A strange obsession with mathematical modelling"? But is not mathematical modelling what you've got, unless you simply want to fly blind and wait and see. But then Ridley was "flying blind" at Northern Rock - so why not?

Ridley is very critical of the UK's response, blaming scientists, civil servants, public health England - but not the government or politicians who appear completely blameless. He's cherry-picking scientific views to push his view that the pandemic is much less dangerous than everyone thinks, but completely fails to explain why New York had so many deaths (hint - it's because he can't explain New York's results).

His last sentence in this article.

And science needs to rethink its affair with models rather than data.

Well, this is just brilliant. He's advocating that if we have a pandemic, let's wait to see how many deaths we get before doing anything at all. Not the optimum strategy, I'd suggest.

[Sorcery]

Looks like Nic Lewis's model or more accurately Gabriela Gomes whose paper Nic cites has gone mainstream. Mat Ridley talks about it here :
https://www.telegraph.co.uk/news/2020/0 ... r-history/

Has the British scientific establishment made its biggest error in history?

A strange obsession with mathematical modelling has compromised the country’s covid response

MATT RIDLEY

Um... As I have not got access to the full article I will be using that as an excuse for not giving here my full and likely blunt comment on Ridley's piece.

"A strange obsession with mathematical modelling"? But is not mathematical modelling what you've got, unless you simply want to fly blind and wait and see. But then Ridley was "flying blind" at Northern Rock - so why not?

Ridley is very critical of the UK's response, blaming scientists, civil servants, public health England - but not the government or politicians who appear completely blameless. He's cherry-picking scientific views to push his view that the pandemic is much less dangerous than everyone thinks, but completely fails to explain why New York had so many deaths (hint - it's because he can't explain New York's results).

His last sentence in this article.

And science needs to rethink its affair with models rather than data.

Well, this is just brilliant. He's advocating that if we have a pandemic, let's wait to see how many deaths we get before doing anything at all. Not the optimum strategy, I'd suggest.

I get the impression that Ridley is critical of Ferguson's model, but then everybody is! It's certainly a problem when Ferguson's model produces different results on separate runs, it strongly implies buggy code. Can't really trust any model that is not bug free in that way. It is absolutely basic to be able to get repeatable results.

And science needs to rethink its affair with models rather than data.

Not sure he is against all models, he seems to like Gabriela Gomes's model for example, so it's perhaps slightly strange for him to criticise them all.
I would agree with him though that we do need to look at/trust the data more, especially where the data contradicts the established models and junk a bad model if necessary. The idea that herd immunity is only reached when Ro = 1 - 1/Ro looks like a dead parrot in the light of Gomes paper (though it still works for vaccinations).

Gabriela Gomes works for the Liverpool School of Tropical Medicine and is a credible source : https://www.lstmed.ac.uk/about/people/dr-gabriela-gomes

[dspp]

I get the impression that Ridley is critical of Ferguson's model, but then everybody is! It's certainly a problem when Ferguson's model produces different results on separate runs, it strongly implies buggy code. Can't really trust any model that is not bug free in that way. It is absolutely basic to be able to get repeatable results.

I've not looked in detail at anybody's model, but I would expect all medium-complexity (or higher) models to be stochastic in nature, i.e. all runs would be different. Low-complexity deterministic models are surely only for kindergarten use ? Higher-complexity models should be both stochastic and recursive.

- dspp

[servodude]

I get the impression that Ridley is critical of Ferguson's model, but then everybody is! It's certainly a problem when Ferguson's model produces different results on separate runs, it strongly implies buggy code. Can't really trust any model that is not bug free in that way. It is absolutely basic to be able to get repeatable results.

I've not looked in detail at anybody's model, but I would expect all medium-complexity (or higher) models to be stochastic in nature, i.e. all runs would be different. Low-complexity deterministic models are surely only for kindergarten use ? Higher-complexity models should be both stochastic and recursive.

- dspp

Markov chain Monte Carlo methods seem to be popular for it; and I can see why.
I do think though that "awareness" impacts the ongoing behavior of the population though, implying that its not "really" a Markov chain in practice (as there's a history, and the path taken would modify the future behaviour). But they'll work well enough for scenario prediction.

-sd

[9873210]

You have the potential for exponential growth so the problem is ill conditioned. Small changes in initial conditions can cause large changes in the result. This is a problem with reality. Any model that does not share this property is a bad model.

Deterministic v. stochastic models is largely a matter of technique. Deterministic differential equations can be and are solved using Monte Carlo techniques. Markov chains can be solved with deterministic matrix multiplication (if you have the computing power, or are lucky and clever).

If you have limited knowledge simple models that incorporate everything you know will be better than complex models that add guess work. IMNSHO far too many stochastic models are designed by throwing far too much poor quality "data" at poorly understood tools, instead of by thinking about the problem. I don't know about medical models, but in the financial field Monte Carlo is largely synonymous with junk.