Hacker News Reader

Comment by nabla9 | Hacker News Reader

So far the best tests are from Iceland and Italian town Vo.

In Iceland over 11% of population is tested, 4% positive. Results indicate that over 50% of infected are asymptomatic.

Italian town Vo tested everyone in their village with similar results. Over 50% of those who test positive are completely asymptomatic.

View on HN

> Results indicate that over 50% of infected are asymptomatic.

If infections double every five days, and it takes 5 days for symptoms to show, then a huge number of samples will be presymptomatic.

The same thing happened with the Diamond Princess, where initial testing showed there were many “asymptomatic” people, but they then got symptoms and the true “asymptomatic “ number dropped to a much lower number.

I have been reading this recent paper which suggests that c19 spreads most in the days before symptoms: https://www.nature.com/articles/s41591-020-0869-5

I believe this is true of most such viruses. The flu is also contagious the day before you show symptoms, then most contagious 5-7 days after [1].

[EDIT] Actually, it says as much in the graph in your article.

[1] https://www.cdc.gov/flu/about/disease/spread.htm

... at the time of the test. Or there was a follow up?

And what were the false positive rates?

If I could upvote this a million times I would. If 50% of those tested positive are asymptomatic, then a good first-order approximation is that the false positive rate is 2x the actual prevalence, and the results say very little about prevalence that actual hospitalizations don't say.

People, and to a large degree the media, are putting a huge emphasis on testing, not always in a way sensitive to the essential difficulty of evaluating test accuracy, especially in epidemiological settings.

Would it be disingenuous to say that false positives in the current case fatality rate would bring the rate down because hospitals are using symptoms combined with PCR for deaths?

From what I've heard, at least in some areas, a Covid test is often performed only after a (usually very fast) flu test is performed. So a death resulting from a person negative for flu but also false-positive for Covid would seem very unlikely. As would any death of this sort from a person who came up positive. Much more likely would be that false positives cluster among those with mild or no symptoms (with the mild symptoms instead coming from a more mundane infection)

If false positives were pushing the published death rate up, it's important to remember there are also deaths among people infected but never tested that would be pushing the rate down. I don't know if they would cancel each other out though. I would however really like to see published data from the CDC about reported pneumonia deaths just prior to the "official" outbreak to see if there was an otherwise unexplained increase.

Agreed. It will be very nice to have more data. There are too many ifs right now.

Are false positives more prevalent in this type of testing than false negatives? Edit: Due to the sars-cov-1?

I don't know but the problem is more with low infection rates. Suppose you have 3% false positives and 3% false negatives, and 0.1% of people are infected.

The false negatives will be 3% of 0.1%. The false positives will be 3% of 99.9%. You exaggerate your infection rate by about 30X even though the test is equally inaccurate in both directions.

Fantastic explanation!

In Vo they first tested everyone, then there was 14 day lockdown for those infected, then they tested them again. 6 of them were still positive.

Well maybe positive but not necessarily able to infect. If they used PCR tests it could give positive on broken up inactivated viruses for awhile after you cleared the infection

> Results indicate that over 50% of infected are asymptomatic.

Please stop spreading this meme. Pre-symptomatic, not asymptomatic

https://www.icelandreview.com/sci-tech/is-icelands-coronavir...

Have these tests been antibodies though? Normal tests otherwise only test for active infections afaik.

What criteria was used for testing? Completely random? If it was volunteer, their method of obtaining and screening the volunteers is very important as well.

Same with the Diamond Princess.

Diamond Princess was a very old population, and started off this whole scare campaign that this was going to be Spanish Flu Part 2. It's like doing study within just a nursing home. It won't be representative.

I very much agree.

No, Wuhan's medical system getting overloaded started the scare campaign, and then Italy's medical system getting overloaded confirmed that this is, in fact, the real McCoy.

It did not confirm anything was the real McCoy. What's confirming things is the latest serological studies that show the data from Italy was far from representative, and so was the data from Wuhan. The data from the rest of China was much more in line (CFR 0.6%) with the serological studies we're seeing of late (IFR 0.25-1%).

The data from Italy isn't representative of anything other than Lombardy's overwhelmingly old and sick population, Italy's unfortunate response to the situation and the fact the disease hits older folks hundreds of times harder than younger folks. Italy's CFR is about 2 orders of magnitude higher than the IFR we should be basing our broader response on.

Reply on news.ycombinator.com