Coronavirus - Modelling Aspects Only

[Nimrod103]

Here is a truly sobering argument:

https://www-spectator-co-uk.cdn.ampproj ... positives-

At a very low disease incidence (as we have at the moment), the number of positive tests is far in excess of the true prevalence of the disease. Wasn't that a bit like Trump's argument? But I jest. A staggering thought, the disease might almost have died out, even though we still get the same number of positive test results. Carl Heneghan is turning out to one of the few people in this plague who can think clearly.

[Mike4]

Here is a truly sobering argument:

https://www-spectator-co-uk.cdn.ampproj ... positives-

At a very low disease incidence (as we have at the moment), the number of positive tests is far in excess of the true prevalence of the disease. Wasn't that a bit like Trump's argument? But I jest. A staggering thought, the disease might almost have died out, even though we still get the same number of positive test results. Carl Heneghan is turning out to one of the few people in this plague who can think clearly.

A good article, thanks for posting, but I'd quibble with your last statement. This effect is clearly understood within the world of medicine and epidemiology. But is a difficult concept to grasp so it gets conveniently ignored by journalists, politicians, contributors to discussion forums etc (including me). I've encountered this effect being explained several times now, including in one of the videos by Dr John Campbell. And each time I've struggled to grasp it.

I saw the boss of a firm who make the PCR tests being interviewed and he explained the sensitivity/specificity thing much more clearly than the article. He said they face a balancing act - they cannot make a test both 100% sensitive and at the same time 100% specific, they have to draw a balance. They can easily raise the sensitivity to near 100% (i.e. detect near 100% of infections) but at the cost of specificity (i.e. more false positives). And vice versa, they can guarantee near zero false positives as the cost of missing more of the genuine positives, AIUI.

The language used (sensitivity and specificity) muddies the waters in my view, different words are needed. Until the sensitivity/specificity thing is understood, the arithmetic which follows from it makes little sense. Although Prof Heneghan explains the arithmetic nice and clearly, the preamble explaining sensitivity/specificity was as confusing as any of the explanations I've ever read.

The net result I think, is most people encountering articles like this fail to quickly understand the sensitivity/specificity thing (e.g. me), so the arithmetic explaining the low accuracy of the stats in low concentrations of infection gets lost of them. Instead, they (we, I) fall back on the intuitive trust we have in the headline figures produced by the scientific test.

On a related subject, the BBC this morning have been saying of the infection spike in Preston, that nearly 50% of new infections being detected are in people under 30. Given the absolute numbers are so small, this could be meaningless.

[Nimrod103]

Here is a truly sobering argument:

https://www-spectator-co-uk.cdn.ampproj ... positives-

At a very low disease incidence (as we have at the moment), the number of positive tests is far in excess of the true prevalence of the disease. Wasn't that a bit like Trump's argument? But I jest. A staggering thought, the disease might almost have died out, even though we still get the same number of positive test results. Carl Heneghan is turning out to one of the few people in this plague who can think clearly.

A good article, thanks for posting, but I'd quibble with your last statement. This effect is clearly understood within the world of medicine and epidemiology. But is a difficult concept to grasp so it gets conveniently ignored by journalists, politicians, contributors to discussion forums etc (including me). I've encountered this effect being explained several times now, including in one of the videos by Dr John Campbell. And each time I've struggled to grasp it.

I saw the boss of a firm who make the PCR tests being interviewed and he explained the sensitivity/specificity thing much more clearly than the article. He said they face a balancing act - they cannot make a test both 100% sensitive and at the same time 100% specific, they have to draw a balance. They can easily raise the sensitivity to near 100% (i.e. detect near 100% of infections) but at the cost of specificity (i.e. more false positives). And vice versa, they can guarantee near zero false positives as the cost of missing more of the genuine positives, AIUI.

The language used (sensitivity and specificity) muddies the waters in my view, different words are needed. Until the sensitivity/specificity thing is understood, the arithmetic which follows from it makes little sense. Although Prof Heneghan explains the arithmetic nice and clearly, the preamble explaining sensitivity/specificity was as confusing as any of the explanations I've ever read.

The net result I think, is most people encountering articles like this fail to quickly understand the sensitivity/specificity thing (e.g. me), so the arithmetic explaining the low accuracy of the stats in low concentrations of infection gets lost of them. Instead, they (we, I) fall back on the intuitive trust we have in the headline figures produced by the scientific test.

On a related subject, the BBC this morning have been saying of the infection spike in Preston, that nearly 50% of new infections being detected are in people under 30. Given the absolute numbers are so small, this could be meaningless.

I had come across this issue of sensitivity/specificity before, but it didn't have much impact on me early in the infection. Now, however, in the probable final waning stages of this infection, it becomes very important indeed. Just like Heneghan pointing out how PHE were over-counting numbers of deaths from Covid, which now has quite a significant effect on the low numbers. I think Heneghan explains things well and concisely.

[dspp]

Now, however, in the probable final waning stages of this infection, .

I'm not sure I'd be quite that certain that these are the waning stages (though it would be nice if it were). Here is the graph for Iran (leaked data via BBC, so not really copyright anyone), which kind of shows how it can go if one just takes the brakes off. It is worth recalling that Iran was the second country to get really badly hit, as it was the first to receive a goodly number of carriers from China.



(image from BCC at https://www.bbc.co.uk/news/52959756, exact copyright either BBC or Iran Health Ministry or someone else)

My personal take is that for Europe, the current amount of aggregate social distancing is keeping the infection rate within bounds, and so the overall situation under control. Wherever the aggregate distancing lapses (for whatever reason) the R number rises and a local outbreak grows quickly since cv19 is now quite widely dispersed in the population and able to seed quickly. The problem is that we are not infecting enough people to be able to work our way through an on-the-job innoculation programme and so it might take a very long time to reach herd immunity. I would be very happy if I was wrong and it was the waning stages, but I don't think the data supports that.

regards, dspp

[Nimrod103]

Now, however, in the probable final waning stages of this infection, .

I'm not sure I'd be quite that certain that these are the waning stages (though it would be nice if it were). Here is the graph for Iran (leaked data via BBC, so not really copyright anyone), which kind of shows how it can go if one just takes the brakes off. It is worth recalling that Iran was the second country to get really badly hit, as it was the first to receive a goodly number of carriers from China.

My personal take is that for Europe, the current amount of aggregate social distancing is keeping the infection rate within bounds, and so the overall situation under control. Wherever the aggregate distancing lapses (for whatever reason) the R number rises and a local outbreak grows quickly since cv19 is now quite widely dispersed in the population and able to seed quickly. The problem is that we are not infecting enough people to be able to work our way through an on-the-job innoculation programme and so it might take a very long time to reach herd immunity. I would be very happy if I was wrong and it was the waning stages, but I don't think the data supports that.

regards, dspp

I read a couple of weeks ago that the new spike is affecting a different area of Iran, not sure whether that is true or not. The true daily death numbers are still not that high compared to Europe and USA.
My own interpretation of what is happening in the UK is that the infection has burnt its way through most of those vulnerable because of genetic, health and age, or all combined. This point was reached early on in London, and as it had spread in the north of England, it makes its appearance in the form of of these sudden local outbreaks, probably caused by superspreaders. Many European countries with earlier more effective lockdowns did not reach that level of immunity, and are now fighting a second wave.
One day we may know the truth.

[dspp]

Now, however, in the probable final waning stages of this infection, .

I'm not sure I'd be quite that certain that these are the waning stages (though it would be nice if it were). Here is the graph for Iran (leaked data via BBC, so not really copyright anyone), which kind of shows how it can go if one just takes the brakes off. It is worth recalling that Iran was the second country to get really badly hit, as it was the first to receive a goodly number of carriers from China.

My personal take is that for Europe, the current amount of aggregate social distancing is keeping the infection rate within bounds, and so the overall situation under control. Wherever the aggregate distancing lapses (for whatever reason) the R number rises and a local outbreak grows quickly since cv19 is now quite widely dispersed in the population and able to seed quickly. The problem is that we are not infecting enough people to be able to work our way through an on-the-job innoculation programme and so it might take a very long time to reach herd immunity. I would be very happy if I was wrong and it was the waning stages, but I don't think the data supports that.

regards, dspp

I read a couple of weeks ago that the new spike is affecting a different area of Iran, not sure whether that is true or not. The true daily death numbers are still not that high compared to Europe and USA.
My own interpretation of what is happening in the UK is that the infection has burnt its way through most of those vulnerable because of genetic, health and age, or all combined. This point was reached early on in London, and as it had spread in the north of England, it makes its appearance in the form of of these sudden local outbreaks, probably caused by superspreaders. Many European countries with earlier more effective lockdowns did not reach that level of immunity, and are now fighting a second wave.
One day we may know the truth.

Quite possible. And the truth shall set ye free.

[jfgw]

A good article, thanks for posting, but I'd quibble with your last statement. This effect is clearly understood within the world of medicine and epidemiology. But is a difficult concept to grasp so it gets conveniently ignored by journalists, politicians, contributors to discussion forums etc (including me). I've encountered this effect being explained several times now, including in one of the videos by Dr John Campbell. And each time I've struggled to grasp it.

I had no idea it was that difficult. If the specificity is 0.1% and you test lots of people who do not have the disease, about 0.1% of them will, nonetheless, test positive. Ignore the sensitivity bit for now if it is confusing. It does make a difference but not much if the infection rate is very low.

Since we do not know how many tests are being conducted, nor of the specificity of the test, we are left in the dark as to how many positives are false. My best guess based upon publically available information is somewhere between 10% and 99.99%.


Julian F. G. W.

[Mike4]

A good article, thanks for posting, but I'd quibble with your last statement. This effect is clearly understood within the world of medicine and epidemiology. But is a difficult concept to grasp so it gets conveniently ignored by journalists, politicians, contributors to discussion forums etc (including me). I've encountered this effect being explained several times now, including in one of the videos by Dr John Campbell. And each time I've struggled to grasp it.

I had no idea it was that difficult. If the specificity is 0.1% and you test lots of people who do not have the disease, about 0.1% of them will, nonetheless, test positive. Ignore the sensitivity bit for now if it is confusing. It does make a difference but not much if the infection rate is very low.

Since we do not know how many tests are being conducted, nor of the specificity of the test, we are left in the dark as to how many positives are false. My best guess based upon publically available information is somewhere between 10% and 99.99%.


Julian F. G. W.

Are you SURE specificity is a measure of false positives? I thought it was a measure of how few negative results you got if you tested only positive patients.

Now what was it you were saying about it not being confusing?

[jfgw]

Are you SURE specificity is a measure of false positives? I thought it was a measure of how few negative results you got if you tested only positive patients.

I see where I went wrong there. It should be "If the specificity is 99.9% and you test lots of people who do not have the disease, about 0.1% of them will, nonetheless, test positive."

Specificity is a measure of how many negative results you get if you only test negative patients. In other words, it is how specific the test is. A less specific test will detect things that are not specific to sars-cov-2.

If none of those tested have the disease:
If the specificity is 95%, 95% will test negative and 5% will test positive;
If the specificity is 99.9%, 99.9% will test negative and 0.1% will test positive.

If the specificity is 99.9% and 0.2% of those tested test positive, almost half of those will be false positives (not taking sensitivity into account).

Once that is understood, the effect of sensitivity may be considered.


Julian F. G. W.

[johnhemming]

A less specific test will detect things that are not specific to sars-cov-2.

There is also the question as to what proportion of the test population have things that will be picked up by the test. That one would assume would vary from test group to test group. Hence the specificity would not be constant.

[Mike4]

Are you SURE specificity is a measure of false positives? I thought it was a measure of how few negative results you got if you tested only positive patients.

I see where I went wrong there. It should be "If the specificity is 99.9% and you test lots of people who do not have the disease, about 0.1% of them will, nonetheless, test positive."

Specificity is a measure of how many negative results you get if you only test negative patients. In other words, it is how specific the test is. A less specific test will detect things that are not specific to sars-cov-2.

If none of those tested have the disease:
If the specificity is 95%, 95% will test negative and 5% will test positive;
If the specificity is 99.9%, 99.9% will test negative and 0.1% will test positive.

If the specificity is 99.9% and 0.2% of those tested test positive, almost half of those will be false positives (not taking sensitivity into account).

Once that is understood, the effect of sensitivity may be considered.


Julian F. G. W.

Do you see now why I say yer average politician/journalist/football fan avoids even thinking about it?

[Mike4]

A less specific test will detect things that are not specific to sars-cov-2.

There is also the question as to what proportion of the test population have things that will be picked up by the test. That one would assume would vary from test group to test group. Hence the specificity would not be constant.

Yes this is another point made by the bod from the company making the tests. He said with the lower specificity that comes if very high sensitivity is specified, closely related virus e.g. other coronaviruses can get detected.

There is however, completely new thinking on this subject. If a not-particularly-accurate test can be sold for a dollar apiece, the value in controlling a pandemic is sky high compared to a 99.99% accurate PCR test costing $100 each as the masses can test themselves every day with the cheap test, then make epidemiologically valuable decisions each day as to whether to send the kids to school or go to work themselves, or not. Even a test 50% accurate is better than no test at all. Explained concisely in this five-minute short video by Dr Michael Mina here:

https://www.youtube.com/watch?v=AZWuyvBAWWQ

[Nimrod103]

A less specific test will detect things that are not specific to sars-cov-2.

There is also the question as to what proportion of the test population have things that will be picked up by the test. That one would assume would vary from test group to test group. Hence the specificity would not be constant.

Yes this is another point made by the bod from the company making the tests. He said with the lower specificity that comes if very high sensitivity is specified, closely related virus e.g. other coronaviruses can get detected.

There is however, completely new thinking on this subject. If a not-particularly-accurate test can be sold for a dollar apiece, the value in controlling a pandemic is sky high compared to a 99.99% accurate PCR test costing $100 each as the masses can test themselves every day with the cheap test, then make epidemiologically valuable decisions each day as to whether to send the kids to school or go to work themselves, or not. Even a test 50% accurate is better than no test at all. Explained concisely in this five-minute short video by Dr Michael Mina here:

https://www.youtube.com/watch?v=AZWuyvBAWWQ

What is the point of testing so often, unless people abide by the results? If you test positive on a particular day, 80% chance you will have no symptoms. Are you going to stay at home, or go out and earn a dollar?

[Mike4]

What is the point of testing so often, unless people abide by the results? If you test positive on a particular day, 80% chance you will have no symptoms. Are you going to stay at home, or go out and earn a dollar?

This is true.

As I've been saying since the dawn of time, people need to be paid to self-isolate or they simply won't do it, and lootman keeps arguing with me about it.

I give it three months before Matt Hancock comes up with the bright idea of compensating people for self-isolating, to encourage them to do it. Politicians seem to take three months longer than the rest of us to realise the bleedin' obvious, to paraphrase them Pythons.

[jfgw]

Even a test 50% accurate is better than no test at all.

There is an online test here that is 50% accurate, https://justflipacoin.com/

For a real sars-cov-2 test, the accuracy will be different for positive cases and negative cases (hence values for both sensitivity and specificity). A test with a sensitivity of 50% will detect 50% of infections (which I assume is what you meant).

Julian F. G. W.

[johnhemming]

Not one of the usual suspects:
https://www.thetimes.co.uk/edition/news ... -ppdxjm0hb

Some hospitals didn’t have a single Covid-19 patient on their wards last week, the Sunday Times reports, with one top doctor suggesting that Britain is “almost reaching herd immunity”.

It's not yet known for sure if herd immunity can stop or even slow Covid-19, with some scientists expressing concern that even with coronavirus antibodies it may be possible to catch the virus a second time.Ron Daniels, an intensive care consultant in Birmingham, one of the worst-hit areas, told The Sunday Times there had been a big fall in admissions.

Last Thursday, across three hospitals that serve more than 50 per cent of Birmingham’s population, there were three critically ill Covid-19 patients.

“Compare that to where we were a couple of months ago, when we had almost 200 patients ventilated at any one given time, and this is a huge downturn,” Dr Daniels said.

He said the figures showed there was cause to be optimistic, even with the recent rise in cases in some areas such as Aberdeen and Preston.

He added that he didn’t expect an increase in hospital admissions. “I think that’s highly unlikely, because the pubs have been open for over a month [and] people have been socially interacting heavily during that time and the natural history of this disease is that if you contract the virus and you’re going to end up in hospital, you’re pretty much in hospital within 15 days,” he said.

[Mike4]

Not one of the usual suspects:
https://www.thetimes.co.uk/edition/news ... -ppdxjm0hb

Some hospitals didn’t have a single Covid-19 patient on their wards last week, the Sunday Times reports, with one top doctor suggesting that Britain is “almost reaching herd immunity”.

It's not yet known for sure if herd immunity can stop or even slow Covid-19, with some scientists expressing concern that even with coronavirus antibodies it may be possible to catch the virus a second time.Ron Daniels, an intensive care consultant in Birmingham, one of the worst-hit areas, told The Sunday Times there had been a big fall in admissions.

Last Thursday, across three hospitals that serve more than 50 per cent of Birmingham’s population, there were three critically ill Covid-19 patients.

“Compare that to where we were a couple of months ago, when we had almost 200 patients ventilated at any one given time, and this is a huge downturn,” Dr Daniels said.

He said the figures showed there was cause to be optimistic, even with the recent rise in cases in some areas such as Aberdeen and Preston.

He added that he didn’t expect an increase in hospital admissions. “I think that’s highly unlikely, because the pubs have been open for over a month [and] people have been socially interacting heavily during that time and the natural history of this disease is that if you contract the virus and you’re going to end up in hospital, you’re pretty much in hospital within 15 days,” he said.

Dr Malcolm Kendrick was reporting something similar from Sweden on his blog on 7th August 2020. In fact read back through some of his earlier articles too for a different perspective on this whole thing.

https://drmalcolmkendrick.org/

[johnhemming]

I think the mistake in this is to ignore the fact that a proportion of people already have resistance to Covid-19 and are not hence susceptible. That brings the IFR up to more around 0.2-3 rather than 0.12.

[tjh290633]

I think the mistake in this is to ignore the fact that a proportion of people already have resistance to Covid-19 and are not hence susceptible. That brings the IFR up to more around 0.2-3 rather than 0.12.

Why should that change the IFR? If they don't get it because they are resistant to it, surely that is a contributor to the lower figure.

When they are testing for antibodies, do they assume that anyone with T-cells has had the virus?

TJH

[Mike4]

I think the mistake in this is to ignore the fact that a proportion of people already have resistance to Covid-19 and are not hence susceptible. That brings the IFR up to more around 0.2-3 rather than 0.12.

Why should that change the IFR? If they don't get it because they are resistant to it, surely that is a contributor to the lower figure.

When they are testing for antibodies, do they assume that anyone with T-cells has had the virus?

TJH

Eh? Everyone has T cells AIUI, but T cells which recognise SARS-CoV-2 are not detected in any antibody test. Again AIUI.

Agree with your point about the IFR not being affected by any innate resistance to SARS-CoV-2 though.