Tony_Tarantula on 18/4/2020 at 16:51
Iowenz why are you trying to push Trump's bullshit?
You pushing Chloroquine treatments is only going to hurt people and I encourage you to think about what you're doing. It's bad and you should feel bad.
Quote:
Obviously, the best time to get rid of your parachute is right after it has slowed your descent.
Of course and unlike Trump voters you actually have a functioning brain to understand that.
Teabaggers and Trumptards keep pointing to Sweden. Suddenly socialism is wonderful.
To the non brain-dead it's obvious:
Sweden doesn't need to flatten the curve as much because they actually care about people, and have enough beds and ventilators to handle a crisis. That's not the case in the US where we only care about the profits for Republican donors.
zombe on 18/4/2020 at 17:49
Quote Posted by heywood
Whether or not it's any good for getting people back to work depends on the specificity of the test. The false positive rate has to be significantly less than the actual percentage of the population who have antibodies, otherwise positive results give people a false sense of security.
Let's say the tests improve and achieve a specificity of 99%. Sounds pretty good, but if the true number of people with antibodies in the population is 1%, then there's likely to be as many false positives as true positives, and half the people you're green lighting to go back to work don't actually have antibodies. If the true percentage is 2%, then 1/3 of the people you're clearing for work don't have antibodies. And so on.
Your math is broken, but that is besides the point ...
Now, that would be some interesting data to have ... to ascertain the spread characteristics of the virus. It is hard to plan any future actions without having that data. Sending thous, presumably, few who have had it is a drop in the ocean - that is not a reliable plan to begin with (unless there are a lot more of thous than expected). Not having infectious people run around by mistake is impossible to avoid - no plan can rely on even pretending to be doing the impossible. Since it has become a pandemic - there is no getting rid of the virus. Any plan for the future must find a way to just live with it - ie. keep it under control. Understanding its spread characteristics etc in any region of concern is vital for that. There currently are still too many unknowns about it.
edit: what is the expected specificity for antibody testing? I would expect false negatives to greatly overshadow false positives ... but, heywood has made me wonder if that is actually true :/
Pyrian on 18/4/2020 at 18:29
Typically we test a LOT more people that don't have something than people who do, so false positive *raw numbers* can be much higher than false negative *raw numbers* if they have the same rate. To combat that effect, assays are usually designed to have very low or even negligible false positive rates, even at the cost of increasing the false negative rate. But corona viruses in general are dirt common, so my understanding is that it's been challenging to design really good (or even adequate) tests.
lowenz on 18/4/2020 at 19:05
Quote Posted by Tony_Tarantula
Iowenz why are you trying to push Trump's bullshit?
You pushing Chloroquine treatments is only going to hurt people and I encourage you to think about what you're doing. It's bad and you should feel bad.
I push what?
heywood on 18/4/2020 at 19:11
zombe, where did I go wrong in my math?
I've been looking for information on the specificity of these tests, but it's hard to find anything concrete at this point. There's at least 100 different companies rushing them onto the market. I've looked at a few of their websites, like Premier Biotech who supplied the tests for the Stanford study, and most don't say anything. The first antibody test to be authorized by our FDA is made by Cellex, who claim 96.4% according to this article:
<a href="https://www.cidrap.umn.edu/news-perspective/2020/04/antibody-tests-may-hold-clues-covid-19-exposure-immunity-its-complicated">https://www.cidrap.umn.edu/news-perspective/2020/04/antibody-tests-may-hold-clues-covid-19-exposure-immunity-its-complicated</a>
At the low end, here's one at 91%:
(
https://www.mobihealthnews.com/news/los-angeles-company-falsely-announces-fda-authorization-two-minute-covid-19-diagnostic-test)
But there are Chinese-made tests claiming 100%. Perhaps the 100% is true, but out of a sample size of what? If you look at a summary of various tests here, can see that some of the numbers are based on sample sizes of just 30 individuals:
(
https://www.centerforhealthsecurity.org/resources/COVID-19/serology/Serology-based-tests-for-COVID-19.html)
Here in the US, antibody tests and hydroxychloroquine are being pushed by the right. The great right hope was that antibody tests will reveal that a large percentage of the population has already been exposed and developed antibodies - meaning we've already achieved some degree of herd immunity. This idea circled around conservative media about 10 days ago after a military historian wrote about the Stanford study (then in progress) in a San Francisco newspaper. The article was taken down otherwise I'd link it. And now that the study results are out, they're not talking about it.
The other great hope for antibody tests is that they will clear people to go back to work. Maybe there is some promise there, but if the false positive rate is on the same order of magnitude as the true positive rate among the population, it's not going to be of any help.
Quote Posted by lowenz
I push what?
Tony is being disingenuous. Ignore.
zombe on 18/4/2020 at 19:11
English is hard man.
I have been trying to understand the meanings of "specificity" (how good at avoiding false positives) and "sensitivity" (how good at finding the antibody) for a hour now ... using some real numbers from some developed tests ... and i have got nowhere. I just cannot find a way to apply them and have the numbers line up. I give up. Guess it is prudent to ignore my comments about math - but that was besides the point anyway.
That said, if anyone can figure out an example math to show what the actual F' 0.956 specificity and 0.938 sensitivity means - i would appreciate it.
zombe on 18/4/2020 at 19:20
Quote Posted by heywood
zombe, where did I go wrong in my math?
At this point, you ninja poster, i assume you did not ... as i cannot wrap my head around the exact mathematical meaning of the words as notes in my previous post. The only thing i can say for sure is that your math is definitely correct if error chance is the same in both ways. I assumed that to be very unlikely - but now i do not know nor can figure out whether it matters as noted in previous post.
Phatose on 18/4/2020 at 20:19
I think it's like this (all made up numbers).
Imagine you have a population of 1,000 people you tested. Of those, 100 were actually sick, and 900 were actually healthy.
Of those 100 sick people, 90 test positive correctly, and 10 incorrectly test negatively.
Of the 900 healthy people tested, 870 are reported negatively, while 30 get false positive.
Sensitivity = number of times the test correctly reported sick people sick / total number of actually sick people. For us, it would be 90 / 100, or .9
Specificity = number of times the test correctly reported healthy people healthy / total number of healthy people. For us, 870 / 900, or .966
I'm getting the idea the important bits are going to be 1- Specificity and 1 - Sensitivity, which should be the percentage of the tests being incorrect. So, .956 specificity means a 4.4% chance a negative result is wrong, it misses the disease, and .938 sensitivity means 6.2% chance of a positive result being wrong, healthy people being told they're sick.
