[pilowsky]

 Zelandakh, on 2020-September-01, 10:00, said:

Which is why I mentioned not only Fleming but also his team. There are many who believe that La Touche should get more credit than Fleming for the discovery as that is probably where the culture came from but it was Fleming's paper and so, according to scientific precedent, he takes the primary credit.


Did you actually read my post? Here's the very first sentence:


My personal opinion is that what Florey and his team did is much more significant for the development of penicillin than Fleming and his team. My point was not that Florey's achievements should be downplayed but rather celebrated for what they are and not embellished. He did not invent penicillin. Acknowledge that you used the wrong word here (privately, if not publicly), learn from it and move on.

It does not matter what the others did. They did not invent penicillin as a cure for disease. Neither did the other people that Florey and Chain employed. You don't know anything about Science or Medicine or much of else by the sound of it.
If you pay me I will give you lessons.
Your 'personal opinion' is not worth anything because it is completely ill-informed. I was wrong, you don't really know what is going on.
It's like having a discussion with a Trump supporter. Black is not White.

[y66]

In the news today:

Quote

The World Health Organization strongly endorsed steroids as a covid-19 treatment after reviewing studies that found cheap and widely available varieties of the drug could reduce deaths by 20 percent for critically ill patients, among other benefits.

The development was “electrifying,” a critical care physician at Vanderbilt University Medical Center told The Post. It “sets steroids as the standard and the expectation that patients [who] are critically ill will get treated with this.

I understand stuff takes time and the need for caution but when you're dealing with something new that's killing a lot of people, you have to take some risks as the doctors on this expert panel decided to do when treating their critically ill covid-19 patients in January and February for whom the judicious use of steroids seemed obvious.

[y66]

Quote

Screwtape, CS Lewis’s unforgettable devil, has this advice for crushing people who are facing a test of endurance. “Feed him with false hopes . . . Exaggerate the weariness by making him think it will soon be over.”

Thanks to the coronavirus pandemic, we are starting to learn all about weariness and false hopes. It seems endless. And since a highly effective vaccine remains an uncertain prospect, is there any way we might get back to normality without one?

I think there is. The image I can’t shake off is that of the Ready Brek advertisements that have run since I was a child in the 1970s. They show children walking to school in wet and gloomy British winters shielded by a warm orange glow because they ate their porridge-adjacent breakfast.

So indulge me for a moment of science fiction. What if everyone who was infectious glowed bright orange? The virus would be extinct in humans within a month.

Coronavirus is an information problem. A few people are infected, but we don’t know who. We are forced to assume that anybody might be — and, as the economist Joshua Gans observes in his forthcoming book The Pandemic Information Gap, this is extraordinarily costly.

It would be worth a lot to know who is and who is not infectious, and the obvious way to approach the orange glow scenario is to produce tests so cheap, so plentiful and so easy to administer that everyone could test themselves frequently, by spitting on to a strip of special paper.

Such an idea has been discussed for months in certain circles, including in a campaign by Paul Romer, a Nobel laureate economist. But it came to the foreground in the UK this week when — against the backdrop of a faltering testing system — Prime Minister Boris Johnson promised millions of tests a day.

A promise from Mr Johnson might well have negative value. He and his ministers have repeatedly set and missed targets for the scale of testing and the speed at which tests are processed. And while industry press releases have suggested that cheap, accurate tests are just around the corner, independent experts such as Jon Deeks, a professor at Birmingham university, point to a history of over-promising and to a lack of trustworthy evidence that new tests are effective.

Even if a cheap-as-a-cup-of-coffee test could be produced and distributed in sufficient quantities, problems remain. The most obvious is the issue of false positives. The Office for National Statistics estimated that 1 in 2,000 people were infected with coronavirus in England and Wales in late August. With a test that has a false-positive rate of just 1 per cent, you would wrongly flag up 20 uninfected people for every genuine case. The rarer the virus, the worse this problem becomes.

No wonder the experts are sceptical — not just that a cheap test will be produced, but that it could be useful for mass screening if it were. Yet while false positives are a disaster if we react to them in the wrong way, even a shoddy test used well can nudge the odds in our favour.

Imagine a quick-and-dirty test which takes five minutes to conduct, but produces a false positive rate of 10 per cent. Two hundred false positives might be flagged for every genuine case. That is disastrously high if we apply the current UK rules, in which the tested person and everyone in their household has to self-isolate for more than a week.

But even the bad test produces some information: the person with a positive test is 10 times more likely to be infected than a randomly selected person.

So what about the following rule? If you take the test at the school gate and test positive, you must go home and try again tomorrow. If you take a positive test at the theatre entrance, you will need to leave and your ticket will be refunded. If you take it on arrival at Heathrow airport, you’ll have to do a more accurate swab test and isolate until results arrive.

These are all irritating scenarios for the 200 out of 201 who do not actually have the virus. But they are not nearly as irritating as no school, no theatre, no flights and everyone back in lockdown by Christmas.

Fast, cheap tests don’t need to be perfect to help contain the virus. They don’t even need to be nearly perfect. Cheap and quick is enough — provided we use the information wisely. We can’t shut down a school or an office block because one person tests positive on a ropey test: the risk of false positives is too great. But we can ask them to stay at home instead and book a more accurate test.

I have long believed that we undervalue two things. First, when it comes to technology, we undervalue quick-and-dirty relative to expensive-and-perfect. Vaccines get vastly more attention than the prospect of spitting on to a paper strip to produce a result that is probably wrong.

Second, we undervalue data. We spin it, make league tables out of it, turn it into targets, lie about it and disbelieve it. But data, even noisy data, about who is infectious is information that could save both livelihoods, and lives.

https://www.ft.com/c...5f-d7283c7407b3

[y66]

The fall surge is here.

[Winstonm]

Quote

Fast, cheap tests don’t need to be perfect to help contain the virus. They don’t even need to be nearly perfect. Cheap and quick is enough — provided we use the information wisely. We can’t shut down a school or an office block because one person tests positive on a ropey test: the risk of false positives is too great. But we can ask them to stay at home instead and book a more accurate test.

But wouldn't an imperfect test that produced false negatives cause more problems than false positives?

[barmar]

 Winstonm, on 2020-September-22, 08:36, said:

But wouldn't an imperfect test that produced false negatives cause more problems than false positives?

Maybe the idea is that if the test is fast and cheap people can get tested more often, and the false results of both types will be reduced.

[Winstonm]

 barmar, on 2020-September-22, 09:20, said:

Maybe the idea is that if the test is fast and cheap people can get tested more often, and the false results of both types will be reduced.

A false positive makes sense as the next step would be a retest followed by a more sensitive test; on the other hand, a false negative simply means the contagion is quite likely to be spread by an unknowing person who has been reassured of his safety. I don't think those two balance.

[awm]

 Winstonm, on 2020-September-22, 08:36, said:

But wouldn't an imperfect test that produced false negatives cause more problems than false positives?

Not necessarily. If the test has a 1/20 rate of false negatives, we can reduce the transmission rate to something like R/20 at which point the virus dies out quickly.

[cherdano]

 awm, on 2020-September-22, 10:15, said:

Not necessarily. If the test has a 1/20 rate of false negatives, we can reduce the transmission rate to something like R/20 at which point the virus dies out quickly.

Depends on whom you test. If you test a symptomatic person that would self-isolate without a test, they get a false negative test, and go out and infect others, you are creating a problem.

cherdano's rule of testing:

• You shall test asymptomatic people with cheap quick tests with high specificity (no false positives) even if they aren't very sensitive (some false negatives). The more tests the better, the quicker the returns, the better.
  
• You shall test symptomatic people with tests that are highly sensitive (no false negatives), even if they are expensive.
  
• You shall never test asymptomatic people who have just had long close exposure to a known infected person - no test is sensitive enough to justify letting them go to the pub again tomorrow even if the test is negative.Smallprint: testing them after 4-5 days of isolation, or 4-5 days after exposure, is a good idea though - by then PCR tests are sensitive enough.

The good news is that PCR tests for Sars-Cov-2 are really amazingly accurate - at least 99.96% specificity (i.e., at least 99.96% of non-infected people get a negative test), and likely much higher, according to the UK's ONS estimates; the highest sensitivity; and while they are expensive, they are not prohibitively expensive.

The other good news is that many of the cheap tests with lower sensitivity are still almost as sensitive as PCR when people are infectious (logical, as higher virus load = more infectious, and easier to detect the virus).

[pilowsky]

 cherdano, on 2020-September-22, 14:36, said:

Depends on whom you test. If you test a symptomatic person that would self-isolate without a test, they get a false negative test, and go out and infect others, you are creating a problem.

cherdano's rule of testing:

• You shall test asymptomatic people with cheap quick tests with high specificity (no false positives) even if they aren't very sensitive (some false negatives). The more tests the better, the quicker the returns, the better.
• You shall test symptomatic people with tests that are highly sensitive (no false negatives), even if they are expensive.
• You shall never test asymptomatic people who have just had long close exposure to a known infected person - no test is sensitive enough to justify letting them go to the pub again tomorrow even if the test is negative.Smallprint: testing them after 4-5 days of isolation, or 4-5 days after exposure, is a good idea though - by then PCR tests are sensitive enough.

The good news is that PCR tests for Sars-Cov-2 are really amazingly accurate - at least 99.96% specificity (i.e., at least 99.96% of non-infected people get a negative test), and likely much higher, according to the UK's ONS estimates; the highest sensitivity; and while they are expensive, they are not prohibitively expensive.

The other good news is that many of the cheap tests with lower sensitivity are still almost as sensitive as PCR when people are infectious (logical, as higher virus load = more infectious, and easier to detect the virus).

Provide the PVP Mr expert.

[cherdano]

 pilowsky, on 2020-September-22, 14:40, said:

Provide the PVP Mr expert.

To be honest, I think it would be more productive if you are direct and explicit about the issues you have either with me or my posts...

[pilowsky]

 cherdano, on 2020-September-22, 15:29, said:

To be honest, I think it would be more productive if you are direct and explicit about the issues you have either with me or my posts...

I have no issues - The PVP is the predictive value positive. It's the only important metric in a diagnostic test. I am suggesting that you incorporate it into your thinking.
It will improve and better inform your investigations.

[cherdano]

Well, I know how to get from the specificity, the sensitivity and the prevalence to that number. But of course it is wrong to say that this is the only important number - see example of a symptomatic person going to the pub after a (false) negative test...

[helene_t]

 pilowsky, on 2020-September-22, 15:34, said:

I have no issues - The PVP is the predictive value positive. It's the only important metric in a diagnostic test. I am suggesting that you incorporate it into your thinking.
It will improve and better inform your investigations.

Nah. PPV is relevant once you have tested someone and he was positive and you want to know what to do. But since it depends on the prior it's different for everyone so the average ppv is not useful. And at the stage when we're discussing whom to test the average ppv doesn't even exist since we haven't yet decided which population to average across.

[pilowsky]

I see where you are both coming from.
One way we are running into difficulties is that COVID19 is an incredibly infectious disease.
Let's say we have 1 million people and a test that has a sensitivity of 99.9 %. That's pretty good - right.
But it still means that 100 sick people will not be identified. Even if the test is 99.99% sensitive (very unlikely) 10 people will not be detected.
With a disease as infectious as COVID19 this is a big problem. That's why things like smallpox and so had to be eradicated and why people were so annoyed when labs were holding onto samples for 'study' "just in case".
With cancer, it's bad if a small number don't get picked up, but not as bad - they can't spread it about. Unless, it's something like HPV of course, in which case they can which is why circumcision is a good thing.
I could go on but what I am getting at is that behind all the numbers there are real people and real problems. We just need to be careful about how we express things.

When I was an undergraduate student I was part of a group that published a paper using Bayes theorem (the maths, amongst other things, was my contribution) to combine two diagnostic tests to improve the accuracy of the diagnosis of a disorder.
I am aware that your credentials in mathematics and statistics are a great deal more sophisticated than mine. I think that it might be useful if we we could contribute something together that would jointly be of benefit to the Bridge community.

[cherdano]

So on the one hand, PVP is the only thing that matters. On the other hand, high sensitivity is really the most important thing. Got it.

[pilowsky]

 cherdano, on 2020-September-23, 02:35, said:

So on the one hand, PVP is the only thing that matters. On the other hand, high sensitivity is really the most important thing. Got it.

Was ist los? There appears to be a little misunderstanding here.

[cherdano]

 pilowsky, on 2020-September-23, 03:00, said:

Was ist los? There appears to be a little misunderstanding here.

I agree, and the pillowsky of post #874 should sort it out with the pillowsky who wrote post #877.

[pilowsky]

That's too many l's for any Pilowsky to cope with Herr B.

[Zelandakh]

 pilowsky, on 2020-September-23, 03:00, said:

Was ist los? There appears to be a little misunderstanding here.

You appear to misunderstand pretty much everything in every discussion here. If BBFers want reliable Covid information, they look to Arend, Helene and pescetom, who have all been extremely helpful during this period. The others calling themselves experts have, like the US Expert-in-Chief, proven themselves to be extremely unreliable.