Hacker News Reader

Comment by jupp0r | Hacker News Reader

> The sensitivity is 97.90 %, the specificity is 91.77%.

This seems like it could provide a false sense of security to many if used widely by the general population.

It also means that, if that number is correct, 8.23% of not-yet-infected people will test positive. At a time when only 0.01% of people are infected, that means that 8.2301% of people will test positive, and of those people, on average one person† of every 82301 people will have the infection; the other 82300 are all false positives. They'll need to improve that for this to be useful, since at the time 1% of the population is infected, you're already suffering the worst catastrophe in a century.

† (well, precisely 0.979 out of every 82300.979 people, but the difference is not significant in this context)

Yeah, that's the Base Rate Fallacy https://en.wikipedia.org/wiki/Base_rate_fallacy.

I'm a little confused at your calculations.

With a population size of 1000 and with 0.01% of the population infected (i.e. 1 individual) with this test we'd get 82 false positives and 0.979 true positives.

For every true positive result in your population, you'll have 8 false positive results and 0,021 false negatives.

0.01% of 1000 is 0.1 people, not 1 people, and 82÷1 is 82 and not 8, but you're right that my calculations weren't correct either. You'd get 82 false positives and 0.1 (or 0.0979) true positives, so for every true positive result, you have 820 false positives, not 82300 as I said.

RT-PCR tests can be way worse than this.

https://i1.wp.com/emcrit.org/wp-content/uploads/2020/03/perf...

https://emcrit.org/ibcc/COVID19/

And also this.

>Positive results may be due to past or present infection with non-SARS-CoV-2 coronavirus strains, such as coronavirus HKU1, NL63, OC43, or 229E.

I feel like false positives are less damaging than false negatives at this point. Unnecessarily putting people into self quarantine is not the worst, as opposed to giving false security to potentially infectious people.

Not so great if people are using to test to see if they're "immune"

Testing should perhaps not be done at home, at all.

True positives could be sent for PCR testing to confirm whether they are infected or already cleared.

Negatives that show symptoms could likewise be sent for PCR testing.

This way, you'd massively reduce the strain on PCR testing resources.

But this is an antibody test, for people who have already recovered from the virus. False positives in this case are precisely what would give people a false sense of security.

Antibodies are present as soon as there are symptoms or soon after, so these tests are designed to be taken by sick people.

So other people don't have to google like I did: this means this test should correctly diagnose 97.9% of people with the disease, and should correctly diagnose (i.e. give a negative result to) 91.77% of people without the disease.

Nothing is life is certain

Some smart ass always has to comment how this is not perfect and thus not useful. Look at the alternatives. Should we just keep on waiting otherwise?

It is probably useful for getting population statistics and directing higher level actions, or even at some point overquarantining people based on positives and false positives, but people that get a false sense of security are a risk. There was an article about a village leader who tested negative and had 200 villagers come to congratulate him, then he tested positive on a subsequent test.

But there are other tests, the US just didn't use them. South Korea has had tremendous success with testing.

> On Feb. 4, two weeks after South Korea reported its first coronavirus case, the government gave “emergency approval” for Seoul-based Kogene Biotech to move ahead with the test kits.

> Kogene currently exports test kits for the novel virus to 35 countries in Asia, Europe and the Middle East.

https://www.washingtonpost.com/world/asia_pacific/coronaviru...

> Researchers say a viral test is relatively easy to develop. Rather, scientists say, the chasm between the testing haves and have-nots reflects politics, public health strategies and, in some cases, blunders.

> South Korea opened nearly 600 testing clinics, including dozens of drive-through stations. More than 250,000 people have been tested

https://www.nytimes.com/2020/03/20/world/europe/coronavirus-...

Not true. Go deeper. There was ~10% false negatives in South Korea too. But it is still much better than nothing. But of course no on here will say anything about it since everything America does is likely bad and foreign countries are always a step ahead.

jupp0r is correct; this is a real problem for a disease that is so contagious.

Lets say the covid-19 (real) prevalence is 0.1% in the US now. That means 10k infected people for every 10 Million people. If 10M take the test, 97.90 % sensitivity means, of the 10k (true) infected people, 210 will show up as negative. If these people go out and "mingle", there is potential for the infection to pick up again.

Worse, if 10M take the test, 97.90% sensitivity means, 210 000 will show up as positive. (Only 10 000 actually are.)

Reply on news.ycombinator.com