Hacker News Reader

Comment by sojournerc | Hacker News Reader

> For every confirmed case of COVID-19, there are likely another five to 10 people with undetected infections.

I wish this was reflected more in the numbers being presented by various authorities. There is just too much confidence that confirmed cases are representative of the spread of the virus.

This unknown denominator can change mortality rate, and severe case rate by orders of magnitude.

Because of the dearth of testing in the US, it seems highly likely that far more people are infected, and maybe even recovered.

I'm fairly certain I've just recovered from COVID-19 given my symptoms, but can't get tested here in Colorado. By the time I am able to get tested it seems likely that I won't be shedding the virus any longer, and will test negative.

This article captures my feelings on this very well: https://blog.longnow.org/02020/03/14/an-epidemic-of-false-co...

View on HN

Two things to know:

1. The current test is designed for the scenario that someone shows up to medical provider with symptoms. Test answers the question: are the symptoms caused by COVID-19?, quite reliably (negligible false positives, low false negatives).

2. The current test won't produce terribly useful data in someone without symptoms (either before, or after the illness). This is why for example certain well-known political figures saying they have been tested and cleared after potential exposure is not very meaningful.

Therefore ideas like: let's randomly test the population to see how much of this is out there, and I had what I think was COVID-19, I want to be tested to confirm, are not practical.

A test that can determine if someone was infected in the past is in development still.

I'm curious about the technical reasons for what you're saying. Can you say more? Is it a problem with figuring out the false negative rate for people without symptoms?

If you test on antigen that matches the virus, then that will return a negative result in the very beginning of infection and after recovery.

If you want accurate numbers for statistics, including people who never had symptoms, you can do antibody testing on a random sample of population, that will indicate people who had the disease - but that's kind of too late for that particular location and those particular patients, that's for epidemiological studies and helping other locations. I recall reading about one such study for Covid, but I don't remember which country did it.

I heard in an interview with Michael Osterholm, that scientists in China started testing people who had come in contact to those who were sick in the early days of the outbreak, and found people who were not showing any symptoms, but had a large amount of viral load in their throats they were actively breathing out.

What kind of testing they do in those situations?

The first one - if you have the virus, it's multiplying in you and you're infectious then that test will detect it (during that time) no matter if you have severe symptoms or light symptoms or if for you the infection doesn't have any symptoms that you would notice.

Number 2 is misleading, Trump got tested to quell opponents' fear mongering about his health. That doesn't make it more useful scientifically but no one thought it was happening for science anyway.

> I'm fairly certain I've just recovered from COVID-19 given my symptoms, but can't get tested here in Colorado.

I'm glad you're better. Please be careful going forward, though. You shouldn't assume you have an immunity unless you have a positive test result.

Even with the strict pre-test screening that focus strictly on high probability patients, some locations are reporting very high (90%) negative test results from assumed cases.

It's best to continue practicing safety cautions as if you did not have a COVID-19 infection

Yeah, I keep getting stuck on the early test positivity stats in the US. Over the last two weeks, only the most urgent cases were getting tested. But even then, something like 92 - 98% of tests were negative. There are a lot of illnesses out there that are presenting as close enough to COVID to concern health workers, even when it's not COVID.

People don't talk about it much yet, but we really need an immunity test to determine if someone has antibodies.

We are also not 100% certain that having the virus once grants immunity. People have tested positive again after appearing to make a full recovery. While these results are probably better explained by remaining infected and becoming asymptomatic for a period then getting symptoms again, we can’t yet rule out the possibility of reinfection to my knowledge. https://www.latimes.com/world-nation/story/2020-03-13/china-...

I've heard this before and it's the most troubling. Humans have adaptive immune systems. When our bodies first encounter a virus and a while blood cell successfully attaches too and breaks down a virus, it should either replicate or release the right activator proteins that will aid in breaking down the virus.

Vaccines are all about tricking our immune system into preparing those specific antibodies before an infection occurs, so it can stop it quickly. If reinfection is true, what does that say about the nature of this virus and how difficult it is to create a vaccine? Are there other viruses that reinfect like this?

I've also heard it, but I think we need more data.

If true, it would go against all priors for how humans build immunity to viruses - especially coronavirus (common cold).

There may be a time limit to immunity, but having none at all after recovery would be rare indeed.

Without significant evidence (more than one somewhat anecdotal case of re-emergence), there's little reason to assume the body doesn't "remember" the virus after recovery with antibodies.

Seems like unnecessary FUD to spread that idea around without more evidence.

> especially coronavirus (common cold).

Remember, coronavirus are ~25% of colds. There are many other families (like rhinovirues), that cause what we think of as the cold.

> If true, it would go against all priors for how humans build immunity to viruses - especially coronavirus (common cold).

Another possible implication is that there could be multiple strains in circulation, not that the first strain wasn’t immunized against.

You are totally right that we need more evidence and this is very unlikely to be happening. But, I am trying to convince OP to continue to be cautious even if they appear to have recovered, both for themselves and for others. It seems reasonable to be cautious. People have presented apparent recovery only to go on to appear infected again.

I am OP, and I'm definitely going to remain cautious. I appreciate your concern.

I will continue self-quarantine for another couple weeks, and will definitely continue social-distancing measures etc. etc. after that.

From the number of deaths you can crudely infer the number of infected, assuming the mortality on the Diamond princess and in South Korea (1%). Demographic effects are in play that don't make this quite right.

More than half that cruise ship was older than 60. All the deaths on the ship attributed to coronavirus were people above 70 years old.

https://cmmid.github.io/topics/covid19/severity/diamond_crui...

Totally agree. That’s the number I’ve been watching most to get a sense of how many total are infected for different regions e.g. 1 death today = 50-200 people were infected 2 weeks ago.

Applied to US data... 21 deaths today. 14% daily increase, doubling time of 5.2 days (2 week lag; hopefully social distancing has lengthened it since then). Assuming 1% fatality, that would mean 2100 "actual" cases two weeks ago. Assume no social distancing and doubling time of 5.2 days, or say it doubled three times in two weeks and you get 16,800 cases today. But we've already got 11k confirmed cases today, and I can't believe our confirmed count has already caught up that much.

I dunno, maybe that's an indication that the fatality rate is lower than 1%. Or that dying takes longer than 14 days.

Then once you have 2 weeks of death data you can't get the transmission rate. Finally you could go to Seattle and sample 50k people to see how many were infected with a serology test to further refine the 1% assumption

What if many of the passengers who tested negative on the Diamond Princess had it and had already recovered when they tested negative?

This is considered in the article I posted.

South Korea went over 1% fatality yesterday, and they're seeing a rise in numbers again... they might not be testing enough either, so natural fatality could still be lower. (Maybe they should have shut down for two weeks.)

Aren't cruise ships generally an older crowd?

Cruise ships have a lot in the 40-60 range.

This gives them a higher median than the general population, but a lower density of the truly old. And deaths are strongly clustered among the truly old.

Yes, they're basically floating nursing homes doubling as casinos.

It seems like you're pushing back on the CFR, possibly driven by trying to pessimize the need for action.

In the small Italian city of Nembro in Bergamo, 0.6% of their entire population has died in the last 12 days. That's 70 people in a small town of 11,000 — typically there are fewer than 70 deaths every six months. Now we're certainly _overestimating_ the denominator and still getting a number that's larger than the flu. While the numerator is also an overestimate (because it includes deaths not related to COVID-19), their healthcare system is stretched beyond capacity due to the virus leading to difficulties treating all illnesses and traumas.

https://www.washingtonpost.com/world/europe/coronavirus-obit...

You should include demographic information, since it is relevant to your argument:

  [Nembro] Age Distribution (E 2019)
  0-9 years 1,018
  10-19 years 1,191
  20-29 years 1,179
  30-39 years 1,192
  40-49 years 1,640
  50-59 years 1,850
  60-69 years 1,465
  70-79 years 1,181
  80+ years 810

> In 2010, there were 119,551 people residing in Bergamo (in which the greater area has about 500 000 inhabitants), located in the province of Bergamo, Lombardia, of whom 46.6% were male and 53.4% were female. Minors (children ages 18 and younger) totalled 16.79 percent of the population compared to pensioners who number 23.61 percent. This compares with the Italian average of 17.88 percent (minors) and 20.29 percent (pensioners).

https://www.citypopulation.de/en/italy/lombardia/bergamo/016...

https://en.wikipedia.org/wiki/Bergamo#Demographics

Cool, thanks for the data. Looks like their population is 30% over 60, whereas the US (for example) is 20%. Even if we assume _every_ death is someone over 60 (it's not), that'd adjust the percentage to 0.4% in a generic US town.

Note that Florida is similarly 28% over 60.

If you compare the 80+ group (which has a much higher CFR), this town has closer to three times the percentage of the US population.

On top of that, this is in a region of Italy where the healthcare capacity has been completely overwhelmed. That's going to be devastating for the prognosis of an older person who needs a ventilator.

Very true. Here's some more comprehensive numbers from Italian media:

https://translate.google.com/translate?hl=en&sl=it&u=https:/...

Since Feb 25 there have been 91 deaths with positive COVID-19 diagnoses. In total there have been 330 deaths. I'd be very curious what the age breakdowns are.

Anyhow, this was just a small example that shows that we're already seeing significant mortality relative to the _total population_ and not just by confirmed positive cases.

A comment by Bradley Kuszmaul on that blog post gives a good reason why the scenario presented is not likely:

> Given the amount of testing being done in South Korea, we can rule out that there are a lot more people currently infected than we know about. And if there were a lot more people recovered than we know about, we would have seen deaths earlier in South Korea, since South Korea started their testing earlier. I suspect there's a lot more evidence that rules out the "fast and stealthy" hypothesis.

It's possible (likely imho) that there is some missing factor that we have yet to discover that explains the funky nature of the data. Something about the people rather than a property of the disease. Almost like an allergy: peanuts are quite fine for everyone to eat, except those with severe peanut allergies.

What we do know is that it can't be quite "there were zillions of people infected so when we talk about 1% IFR really it's 0.01% so everything is fine", because there are accumulating bodies in various localized places around the world. There's nothing we know to say that pattern wouldn't repeated everywhere.

The type of testing S Korea does require the Virus to be present in the body and only work in a certain window of time. Once a Serological survey is done, we’ll know the true extant.

South Korea is far from the only place COVID-19 is present, and most of the rest of the world tests barely anyone, even if they have symptoms and would like a test, either due to lack of resources or formal policies of only testing hospitalised cases. I can detect my infection, but statistics on the rate it is spreading in the UK can't.

It's also pretty well established that even in symptomatic cases where testing is widespread the disease has a multi-day incubation period.

If the people who recover after mild/minimal symptoms do so within <5days (hypothetically), but people who have severe cases take 10 days to get severe, wouldn't that account for "would have seen deaths earlier in South Korea"?

Yes, that would. But what evidence do we have for that? I believe you're engaging in special pleading: appealing to an exception because it fits your conclusion.

Given the amount of testing being done in South Korea, we can rule out that there are a lot more people currently infected than we know about

That's a very strong claim. SK has done a lot of testing relative to most places but it's still:

a) A tiny amount relative to the population

b) Not a random sample of the population

Their tests were done at testing clinics that people had to take themselves to, which in many cases had large queues and generally you'd only go there if you actually felt really sick because otherwise why would you expose yourself to lots of probably really sick and infectious people?

So far there's overwhelming evidence that either a lot of people have the disease without symptoms, or the tests have a very high FP rate. There's also a lot of people writing like the person above, who are sure they've had the disease based on symptoms but were never tested. South Korea wasn't deploying their tests on people who were asymptomatic, they are doing the same as everyone else: testing people who present for testing or who otherwise seem to need it.

China tested hundreds of thousands of people in provinces with known cases, and found very little evidence of asymptomatic circulation. See my comment from a week ago, https://news.ycombinator.com/item?id=22475853, which quotes Bruce Aylward, an assistance director general and epidemiologist for the WHO.

Our current evidence points to there not being millions of asymptomatic people.

I see your HN comment from 2 weeks ago and raise you with a point in this article lifted from a recent Science paper. Namely, ...individuals with undetected infections were so prevalent in China that they apparently were the infection source for 86 percent of confirmed cases.

This is strong evidence of asymptomatic circulation.

This conclusion is not in contradiction to the evidence behind your previous comment. In mid-February with ~60k confirmed cases that would be 300-600k asymptomatic ones. Which means that China did not have millions of asymptomatic people at the time. And so it is no surprise that a broad testing regime failed to find what did not exist at that time.

However what it also means is that as the number of confirmed cases skyrockets, it WILL be true (and possibly already is) that the number of asymptomatic cases will be in the millions.

The study is on the initial outbreak in Wuhan when testing was in short supply.

And has the testing story improved relative to the need?

Also from the article and the same Science paper, For every confirmed case of COVID-19, there are likely another five to 10 people with undetected infections.

They were talking about abysmal testing rate:

https://science.sciencemag.org/content/early/2020/03/13/scie...

”We estimate 86% of all infections were undocumented (95% CI: [82%–90%]) prior to 23 January 2020 travel restrictions.”

“I’ll see … and raise” is known as a string raise and is seen as bad form or even cheating. Actions in poker must be atomic.

https://upswingpoker.com/angle-shooting-string-bet-raise/

I will keep that in mind if we ever play poker.

But the phrase remains a common idiom. And in an internet discussion like this, the rationale that makes it bad form in poker does not apply. I cannot see anyone's reactions until after my action is complete and I hit "reply". And therefore I cannot gain a read and final decision based on my incomplete first stage of the action.

Sorry, I don't know anything about poker, but didn't Riker do this all the time on TNG?

He did, and the poker on the show was generally abysmal. But partly for the ‘I’ll see you... and raise you’ prevalence in drama is that it creates a more dramatic situation. The dramatic effect is the very thing that can be taken advantage of in poker, and is why there’s a rule at most professionally run tables that what you claim, on your turn, is what you must do. Once you call, you can no longer raise no matter how much you want to.

Though the bigger reason for not allowing it, is that it can really add confusion and slow down the game when a person says call and the next person now has to wait until they’re sure the current asshole isn’t going to say ‘and raise’ before they can act.

Oh, and keep in mind that poker is full of assholes who will do all kinds of anti social behavior to gain an edge. So the fact that this is a standard rule really gives some idea as to just how annoying it is to the poker community.

I haven't been able to make heads nor tales of his logic. In Guangdong province, at the peak of the outbreak, 0.47% of people tested were positive. Guangdong province has a population of about 110 million people. If that was a representative random sample, that would mean a lot of infections, so he's clearly relying on the fact that it's a sample of people who saw a doctor because they had symptoms. Except that I don't see how you can use that to say anything about asymptomatic cases. It just doesn't make sense.

It's 0.5% of people who were tested for the flu, not the general public.

That still does not rule out mild cases which are regarded as a regular viral illness.

What is the detection limit threshold on that test?

SK is the exception.

There is a blood serum test already available that can determine if you had coronavirus (by testing for COVID-specific antibodies produced by your body). I'd say wait until the pandemic dies down and then go get that if you're curious.

There's going to be some very interesting population studies done post-pandemic to determine exactly what percentage of people had asymptomatic cases (and thus could have unknowingly been contributing to the spread).

Hopefully serological tests will be widely available which would tell us if we’ve had it or not.

Agreed.

That's the necessary data to know when we can start rolling back the extreme measures being taken to slow the spread.

We need a measure to detect herd immunity

It sounds like there are antibody tests in the offing:

https://www.technologyreview.com/s/615379/antibody-test-how-...

>There is just too much confidence that confirmed cases are representative of the spread of the virus.

Is that really the case? I'm finding it to be pretty common knowledge that "confirmed cases" is simply just that.

Another factor is that “confirmed cases” is very very frequently reported as just “cases”.

Unfortunately, that is not how a lot of people think. Wish it was.

Furthermore, when you see that (larger) number it has a psychological effect of "this might be serious".

If conversations with my family are any indication, yes. People see low-looking numbers and place names of faraway cities in headlines and think it isn't serious.

After making them think about incubation times that starts changing, but it takes a conversation.

What is common knowledge depends very much on who you're common with.

People are mostly looking at the death count, which is still very low for a pandemic.

It's hard to convince people to sacrifice so much when they look outside their windows and see nothing special, plus read the news and find that some bomb/earth quake/accident somewhere killed more people in a day than the corona virus in a month.

Not to mention it is still extremely difficult to be tested in the US. Yes, doctors can test you, but you basically can’t get a test unless you have all of the symptoms and are sick enough to need to visit the ER.

> By the time I am able to get tested it seems likely that I won't be shedding the virus any longer, and will test negative.

Shouldn't the test detect antibodies that your immune system has developed?

There are different kinds of testing.

The current testing that's reported in the media is an antigen tests, that directly tests for the presence of the virus. Once your viral load comes down, you will test negative for the virus. Hence the reason people are declared recovered after having two negative tests in a row.

A test for the presence of antibodies is a serology test. The last time I looked into it (~one week ago), serology tests were under active development, but had not been deployed anywhere in a widespread way.

So, right now the testing will only tell you if you currently have the virus. Coming relatively soon are tests that will tell you if you previously had the virus.

Yes, but I don't think the PCR test is testing for that, currently. It's looking for pieces of viral RNA that are present in your blood. Antibody serology tests aren't readily available anywhere yet. In fact, the CDC is still developing theirs.

Can't edit this, but to be completely correct, it's looking for viral RNA present in samples, not necessarily blood. More commonly sputum or respiratory samples when it comes to coronavirus.

That sort of test is in development but isn't reliable enough for use yet. The tests currently in use detect virus DNA, not antibodies.

It doesn’t get called out, but when I read about the Spanish Flu I can’t help but feel that certain genotypes got very strong selective pressure.

Conversely, there are probably genotypes running around today that are completely asymptomatic to Covid. You could have whole families shedding virus wherever they go and never even know what they’ve done.

And one of the ways we learn about immunology is to study people who are immune or asymptomatic, so it’s not just about quarantining people that we need to identify infected individuals.

This is absolutely correct.

Quick recombination so that a beneficial gene can spread without losing genetic diversity of the population is is one of the benefits of sexual reproduction. And immune resistance to new threats is one of the top reasons why we need this capability.

In fact we broadcast signatures of our immune systems in pheromones and women are attracted to the smell of men whose immune systems are different from their own. This improves the probability that the children will be resistant to a wider range of possible diseases.

https://en.wikipedia.org/wiki/Body_odour_and_sexual_attracti... is one of many places that you can start learning more about this topic.

> Because of the dearth of testing in the US, it seems highly likely that far more people are infected, and maybe even recovered.

It's not just highly likely, it's absolutely certain. I recommend the recent Khan academy video on this if you want to understand the math and you want a healthy dose of panic:

https://www.youtube.com/watch?v=mCa0JXEwDEk

> This unknown denominator can change mortality rate, and severe case rate by orders of magnitude.

The two best data points we have would be South Korea and the Diamond Princess cruise ship. Both of these groups have high rates of testing and both have had enough time that we can know how the most of the cases played out, and in both cases we see the death rate at about 1%. This doesn't take into account possibly different age group distributions. The cruise ship probably has less children and less 80+ people than on typical distributions.

Additionally both the Diamond Princess and South Korea cases assume ready access to medical treatment. If the medical system is overwhelmed we will see much higher CFR due to critical cases not receiving adequate care.

>I wish this was reflected more in the numbers being presented by various authorities.

UK authorities presented this exact number at their press conference last week.

you don't need to get tested the vast majority of people (even ederly) are going to have mild flu like symptoms. stay home and get better.

Is there a test, or, what are the chances that there will be a test, to check if somebody already had covid19?

I had been wondering this too

I know of one that’s in the works https://www.sciencemag.org/news/2020/02/singapore-claims-fir...

I have seen other talk about such tests on Twitter but it’s too difficult to find it now.

Such tests seem to be called antibody tests or serological tests/assays.

EDIT: this twitter thread discusses such tests https://twitter.com/nachristakis/status/1240689935557865472?...

The German virologist Christian Drosten mentioned in his daily coronavirus update that some specialized labs at least in Germany are able to do an antibody blood test.

There are also some commercial antibody tests, which have some error rate and as all antibody tests can identify the virus after 7-10 days after onset of symptoms.

https://www.ndr.de/nachrichten/info/15-Coronavirus-Update-In...

I think the term "asymptomatic" gives a wrong picture of the severity of the disease, when various authors use it to mean "testing positive before disease onset" or even outright untested due to testing criteria and availability.

Reply on news.ycombinator.com