> In his press conference today, Governor Cuomo revealed the preliminary results of a first-phase antibody test that surveyed 3,000 New Yorkers over two days in 19 counties at 40 locations that included grocery and big-box stores. The sample suggests that 13.9% of New York State residents have the antibodies, meaning they had the virus at one point and recovered. Of the regions tested–Long Island, NYC, Westchester/Rockland, and the rest of the State–New York City had the highest positive rate at 21.2%. The potential good news to come to light is that the death rate may be far lower than some estimates, at 0.5%.
This still is not a random sampling. It only samples from shoppers at grocery stores and big-box retailers. Imagine doing the same study, but of people who ordered groceries online only. Would you expect to see big differences in exposure rate? I think so.
The study itself isn't linked anywhere, nor have I seen it elsewhere. Science is all about the details. It's not hard to imagine half a dozen ways that the bottom line result of this study could have been skewed by decisions made by the study authors and ground team.
Sure all of these studies are potentially flawed but they're all generally pointing in the same direction. There are many more infections than we know about and the CFR is not anywhere close to the actual IFR.
NYC is an outlier with a 21% infection estimate but for the rest of NY (outside NYC metro/Westchester/LI) the estimate is 3.6%. Santa Clara estimate was 3%. LA County estimate was 4%. Seems like a trend is developing.
> Sure all of these studies are potentially flawed but they're all generally pointing in the same direction
That's unfortunately not the way statistics works. Combining multiple bad tests just makes the results incorrect or highly uncertain; the devil often is in the details. The Stanford/Santa Clara County study is a good example of how the details can really undermine a study.
Things do indeed seem to be converging to the idea that the CFR is not near the IFR, but none of this is new news, and the IFR remains very close to what most epidemiological predictions indicated early last month. If anything, it confirms that COVID is a difficult beast to tame.
We don't have perfect studies and perfect data but calling these tests "bad" seems slightly unfair. They give us an imperfect but useful snapshot of what's going on. But thanks for the condescension. I'd expect nothing less on HN.
The study is dramatically over-sampling exactly the people who have the most potential exposure to COVID-19, and dramatically over-sampling the people who don't.
This could easily be way off. It's not testing the people who aren't home at all, and it has a low chance of only testing the people who leave home rarely, only when strictly necessary. It's mainly finding the people who leave home a lot.
You can't assume that. I know plenty of people who haven't left their apartments in a month plus, because they're having groceries ordered in.
And you completely missed the point that people who are going outside a lot more often are both (a) more likely to have been exposed to COVID-19, and (b) more likely to be encountered by your survey. If there are e.g. 700 daily shoppers and 700 weekly shoppers going to a given store, then if you go on one day and sample every shopper you're not going to get half and half by population as you'd need to for a true random sample; instead, you're going to get 700 daily shoppers and only 100 weekly shoppers! This is a hugely biased sample!
There's more than one store in NYC. People shop at a variety of places. Daily shoppers are likely not going to the same store over and over every day. Even during this pandemic I've hit up several different grocery stores and shops, because some things are hard to find.
Are you describing the actual study or the study as you're imagining it? Because I imagine the statisticians have thought of these concerns and attempted to correct for them.
I find it interesting that this study corresponds to an IFR around 0.5%, similar to the IFR arrived at by other studies using very different methods. Perhaps the statisticians who conduct these studies actually know what they're doing better than Hacker News commenters.
And perhaps it's time to accept the politically inconvenient scientific truth these studies are revealing.
>similar to the IFR arrived at by other studies using very different methods
With the wild range of IFR from so many studies, you could trivially cherry pick supporting studies for a bunch of different numbers, especially if you throw in am error range.
I’m starting to believe that the difference in opinions on whether these studies are legitimate or not hinges heavily on personal behavior.
For my wife and I and kid, we live in SF and order food when we can from grocery. When we can’t, I put on an N99 mask and gloves and buy 2 weeks worth of groceries. Then go through a ritual to be as safe as possible entering the home and bringing groceries in.
If someone asks me at the grocery store to take the covid test, no way. How would I even do that safely???
I live in San Francisco though and am worried about the fact that the grocery is 70% instacart workers.
Everyone I know in the city (nyc, sf, Boston, Chicago etc.) is trying to use 3rd parties to get their groceries.
Everyone I know in the suburbs is going to the store.
I believe the reason people read this study differently is that their experience is different. The whole foods in downtown sf would be scary to be in without a mask. And I view it as the riskiest place I have to go.
It’s full of gig workers who are taking way more risk than the average person.
We will know more when we see the results of the study, but I think we need a more random sample before thinking this NYC study is even remotely close to showing that 20% of NYCers previously had it
I'm afraid the population who has these options (and spends its time on hackernews) is called the 1% for a reason, and probably aren't going to bias the stats significantly.
This doesn't make any sense. Grocery and food delivery is quite affordable. It's not only available to the top 1%. These options are being widely taken advantage of by a large number of people right now way down the economic spectrum. It's not a tiny minority of the wealthiest people patronizing them.
At any given point of time, a disproportionate number of shoppers in a store are super-shoppers - people that shop more often than the average person and for longer periods of time. Those people are also more likely to have caught Coronavirus before. Without even getting into online ordering, false positives, etc...this phenomenon alone may confound the test results.
How do you know how often these people are leaving their homes? Everyone has to leave their home to go to the grocery store eventually or someone in their family does. Yes there are delivery services but the vast majority of people aren't using those.
True, but if you encounter one daily shopper, two weekly shoppers and twenty monthly shoppers, it will skew the results. Same if the numbers are reversed.
But you won't, the opposite will happen; you'll proportionally encounter way more daily shoppers than infrequent shoppers, for the simple tautological reason that the daily shoppers are going more often.
This is a sampling bias, and it's the exact same problem as if you ran a survey trying to ascertain how many people have landlines ... by calling random phone numbers.
I don’t get your point. Maybe we agree. If in one city there are 1000 30-year-old daily-shopers, 1000 50-year-old weekly-shoppers and 1000 70-year-old monthly-shoppers it will not be a good idea to estimate the age of the population by sampling people who enters the store. The average age is 50, not 34.
Most people in NYC are almost certainly not shopping monthly because of storage constraints. Even bi-weekly shopping is likely to exhaust storage space for city-dwellers in small apartments.
Couldn't you get a bad study that reverses the sampling bias to get a decent picture of social distancing strategies using the two bad studies? That alone seems far more useful than knowing the IFR because its actionable information.
Or better yet, do a simple survey and ask the participants how frequently they go to the grocery store. And if you see a correlation between frequency and infection rate, then try and adjust your weights accordingly.
These bias issues are well known among survey designers. There are ways that you can try and account for them. It seems like maybe the medical community is too used to having control groups in their studies?
Most people go to the grocery store.
There seems to be this attitude on HN of immediately dismissing research. No human research is ever perfect, flaws and limitations will always be present, and in the present case, study designs are being expedited in order to be timely. Each study is a piece, and together can provide convergent evidence.
OP's whole point is it's not a good sample because people were recruited while grocery shopping...I can't think of a more tech industry-oriented view of the world; most people work from home on zoom and use instacart right??
No. I'd expect people who order deliveries online and can do their entire job on Zoom/slack all day to be a minority to the point where including them in the study would be covered in a sampling error.
You’re completely missing the point. The selection effects would be things like able-bodied / non-immunocompromised / already-recovered people being selected way higher, which would inflate the apparent infected population size estimate and underestimate the death rate among those infected.
This is not some snooty tech point of view - this kind of flaw in statistical studies matters a lot. The more important the policy implications are, the more important the validity of statistical methods.
I don't think anyone disagrees that the sample has bias. On the other hand, since the majority of people have no choice but to grocery shop in person, I do not think the sample is as tilted as you imply.
And even if it were--- it's notable that it's this easy to find a subgroup with cleared infections this far above what was anticipated to be found across NY/NYC.
> “ since the majority of people have no choice but to grocery shop in person”
This is deeply flawed.
I live in NYC and have been making donations to have food delivered to elderly people since this started.
Just the other day, de Blasio announced that taxi and rideshare drivers would be paid by the city to deliver food - and city agencies would be serving as overflow dispatchers and tech support for it.
Many, many people do not have to go to the grocery store, or they have their healthy younger family member go in their place, etc.
To boot, as Gelman has written on his blog, these studies do not agree with a wide range of other studies, and in most cases the test specificity is on par with the incidence rate itself, making incidence rate estimates from _all_ of the studies very unreliable for correlated reasons, so that pooling the studies does nothing to overcome the huge uncertainty they suffer in incidence rates.
> and in most cases the test specificity is on par with the incidence rate itself,
????? We can't exclude that specificity may be 97% (it's unlikely, based on our data, but it's at the edge of the confidence interval), which is unfortunately on par with the studies returning 2-4% positives... but can't exactly explain a return of 21%. Subtract off 4% worst-case false positives from 21%, and where are you?
Since you want to appeal to authority with Gelman, this is what he said about this on his blog: "– Those California studies estimating 2% or 4% infection rate were hard to assess because of the false-positive problem: if a test has a false positive rate of 1% and you observe 1.5% positive tests, your estimate’s gonna be super noisy. But if 20% of the tests you observe are positive, then the false-positive rate is less of a big deal." ... "– In any case, the 20% number seems reasonable. It’s hard for me to imagine it’s a lot higher, and, given the number of deaths we’ve seen already, I guess it can’t be much lower either."
> so that pooling the studies does nothing to overcome the huge uncertainty they suffer in incidence rates.
The case count multiple from the serological study for New York state (and indeed, the California counties) is right in line with what's expected with a variety of statistical measures that were made without relying upon the serological data. So if you even peeked at my source you'd not be making this argument.
https://www.medrxiv.org/content/10.1101/2020.04.18.20070821v...
> First off, 3% does not sound implausible. If they said 30%, I’d be skeptical, given how everyone’s been hiding out for awhile, but 3%, sure, maybe so.
This is a convenience sample no doubt, but convenience is important here, as timeliness is important. It is easy to naysay, but why don't you suggest a less biased convenience sample? Because unless you mandate whole population testing, samples will always be biased in some way. Those who have phones, those who have time, those who are motivated, etc etc. Given all the potential ways it could be biased, choosing a convenience sample at locations almost universally used is not bad at all, imo
> Combining multiple bad tests just makes the results incorrect or highly uncertain
But it's not just, say, taking a bunch of faulty studies and average them and saying voila. They are in agreement. They'd have to be bad in similar ways in order to converge like that.
"Preliminary test results suggest 21% of NYC residents have Covid antibodies"
Should be re-written as:
"Preliminary test results suggest 21% of of people approached in a crowded grocery store and who would agree to give blood have Covid antibodies"
All of these tests suffer the same sample bias, and that sample bias is massive.
People who are already in a grocery store (risky behavior), who are willing to give a blood sample (believes they may have been previously exposed) are not the same as people who are leaving their house only for very limited purposes and ordering food online.
If we assume that the % of people shopping in grocery stores and willing to take the test are in risk group A, and the people not shopping are in risk group B. We can run the following quick analysis.
First, assume risk group A is 5 times as likely to have had COVID than Risk group B (but plug any assumption in there necessary)
Then assume that risk group A is only 10% of the population, then rerun the numbers as follows:
For every 100 people in risk group A who tested, 21 were positive.
Risk group A is 100 people, and risk group B is 900.
Risk group A's 100 were 21 prior positive.
Risk group B's 900 could be estimated that 4.2 people per 100 were positive, so in total of the 900 people 37.8 were positive.
21+37.8 = 59 people of 1000, or 5.8%.
Plug any numbers you want in the analysis, but the assumptions drive huge variability in the % of New Yorkers infected. Without a less biased sample, we really don't know much other than that way less than 21% of New Yorkers have anti-bodies.
NYC is not suburbia. I believe a higher percentage are still working outside of their homes (and, in many cases, that also means travelling on public transport), and a smaller percentage have the financial and transport resources to significantly reduce the frequency of their shopping trips. Apartment living means a greater exposure to your neighbors whenever you do enter or exit. The sampling in NYC, while far from ideal, may therefore be better than the other regions, in that shopping behaviour may not be as dominant a risk factor as you suppose.
For the same reasons, plus the likelehood that the virus arrived in NYC early and frequently, NYC's figures cannot be extrapolated to the rest of the state or country.
I would like to see how these figures match up to the sewage-sampling method of estimation.
As a (40-something) NYC resident, I don't believe that is accurate. Half of the people I know are working from home and stocked up on 1-2 months of food at the outset of this. The other half left the city and are now staying with friends/family in other states or have rented homes elsewhere.
Those that did stay and still need food (ie produce) are having it delivered.
People who are shopping in stores are absolutely a very biased sample.
Very specifically, this bias excludes older people who are locking themselves away very very judiciously.
> Half of the people I know are working from home and stocked up on 1-2 months of food at the outset of this.
Seriously? Your personal network is evidence the this study is useless because of sampling bias?
Everyone I know voted for Hilary Clinton and yet here we are.
> People who are shopping in stores are absolutely a very biased sample.
People who have jobs that allow them to work from home are a very biased sample. How many grocery stockers are there in NYC compared to let’s say, software developers? I suspect you haven’t a clue, since by your own admission you don’t know a single person who needs to leave their home to work. So maybe your mental model if who makes up the population is skewed.
Sure, I agree with you, excluding them is biased as well. A more appropriate phrase to be used to describe "people who have jobs that allow them to work from home" is "a small minority".
And let's be real, biasing your sample by not including a small minority doesn't affect the outcome of a study nearly as much as doing the opposite and sampling mostly from that small minority. Does it make the results less accurate? Yes. Does it make the results nearly as inaccurate as it would be if they excluded most of people but that minority? Not at all.
> Half of the people I know are working from home and stocked up on 1-2 months of food at the outset of this.
This surprises me if for no other reason than it takes a lot of space to store 1-2 months of food. Might be true, but is certainly different than what is going on in suburban Connecticut based on my social network.
Thanks for your observations. If, however, behavior in NYC is more or less the same as elsewhere, then what might account for the difference in the results between store-customer sampling in the city and elsewhere? The results may well be biased from sampling shoppers, but if the bias is assumed to be from shopping behavior, and that behavior is essentially the same everywhere, why does it not wash out?
One explanation might be that it got an earlier start in NYC, but both the hospitalizations and fatalities seem to have started showing up on about the same dates in the city and the surrounding suburbs.
There seems to be a significant ethnic variation, but ethnicity is such a correlate to other issues that it is not likely to be informative on its own.
I’m a resident of SF downtown and previously DC and Boston. Everyone I know has stopped leaving their apartments/condos except for fresh air and are trying not to go to any stores. When possible are having food delivered. My friends in NYC have all left except for one who has a large apartment and no family without taking a flight.
I think people see these studies of places they visited once and determine that those areas are way more dense and that they have seen the news and that all those deaths are due to tons of people having it.
That is just not true. The studies are biased and likely this will go on for a long-time. We know this is way more deadly than the flu. That is obvious when you look at simple examples like the princess cruise in Japan. How often do 7+ people die on a cruise ship on a single voyage?
All three studies created their sample groups differently. Are all three methods flawed in some way? Sure. Is it more reasonable to completely ignore the picture all three are painting or to consider there might be something to the trend?
They're not just all flawed in some way -- along at least one dimension they are flawed in the same way, that is, they all have a biased sample of the population that seems more likely to be infected than a true random sample.
So, sure, they're "created differently", i.e., they're not all facebook ads, or grocery stores, but all those methods bias towards higher infection rates, don't they?
Given that, you'd expect high precision, low accuracy -- these tests all paint a trend for sure, but we don't know how close that trend is to reality.
These studies have completely different methodologies, and the only way you can plausibly call them similar is by arguing that anyone willing to get serological testing is substantially more likely to be infected than those who are not. That's little more than a fancy way of disregarding any testing result you don't happen to like.
We've been systematically biasing toward testing the old and infected for months now. Why should rates derived from PCR testing with known severity bias considered a strong prior, yet rates derived from serological surveys with hypothetical severity bias be disregarded for the same reasons?
At this point, we have many different studies from different parts of the world -- each with its own methodological flaws and biases -- and yet all are pointing in the same direction: a systematic under-count of cases, and an IFR somewhere between 0.5% and 1%, with a bias toward the lower value.
Anyone worth arguing with acknowledges the limitations of the PCR testing. You are essentially suggesting that we ignore the limitations and biases of these serological surveys. If a handful of internet denizens can quickly point out methodological flaws, maybe the researches should try to address or account for those. I fully understand the urgency of the matter, and that urgency can lead to mistakes. But these studies are going to be used in policy making and could potentially impact millions of lives. The difference between 1% and 0.5% is huge, that is over 1.5 million in the US. And powerful people are trying to use these studies to make a case for re-opening the economy. I think it is completely reasonable to push back on these flaws and biases in light of that.
"You are essentially suggesting that we ignore the limitations and biases of these serological surveys."
I am not. I am suggesting that you are preferring a rate based on an even more biased method, because the method yields a bigger number.
"If a handful of internet denizens can quickly point out methodological flaws, maybe the researches should try to address or account for those."
If a handful of internet denizens can quickly point out "methodological flaws", it usually means that the "methodological flaws" they have discovered are well-known and accounted for. In this particular case, sample bias is not a new statistical phenomenon, and we've been doing serological surveys for a long time.
> I am suggesting that you are preferring a rate based on an even more biased method, because the method yields a bigger number.
What method? I merely stated that PCR has limitations, I never said I was using it to calculate IFR.
For the record, my assumptions are that IFR is around 1%. That is largely based on the Diamond Princess data. The original paper suggested it was 0.5%[1], but at the time there were only 7 fatalities. The current number is 13. The raw IFR is now up to 1.8%. Crudely adjusting for demographics based on the ratio in the paper puts the IFR around 1%. This is the only population that was both comprehensively tested and occurred long enough ago for most of the cases to resolve (although some cases are still active/critical[2]). And I fully acknowledge that this approach has limitations. It is a small population, it happened earlier in the crisis when we knew less about treatment, etc... But I sill think it is the best data point we have.
> If a handful of internet denizens can quickly point out "methodological flaws", it usually means that the "methodological flaws" they have discovered are well-known and accounted for.
Except the Santa Clara study did not do that. They acknowledged potential bias, but did nothing to adjust for it. From the source:
"Other biases, such as bias favoring individuals in good health capable of attending
our testing sites, or bias favoring those with prior COVID-like illnesses seeking antibody confirmation are
also possible. The overall effect of such biases is hard to ascertain."
> arguing that anyone willing to get serological testing is substantially more likely to be infected than those who are not.
Yes, which is clearly true, no?
> That's little more than a fancy way of disregarding any testing result you don't happen to like.
We all have biases. I'm not sure how pointing out that adverse selection is a thing, though, is a bias.
Any stats can be quibbled with, and I have no personal skin in this game -- if anything, I would love (like almost everyone) for the IFR to be <0.1% and for hard immunity to already be a thing. But I'm also aware of (some of!) my own biases.
> and yet all are pointing in the same direction: a systematic under-count of cases
We already know we're undercounting, because we're only explicitly testing people with symptoms, and not even all people with symptoms! We didn't need these studies for that.
These studies would ideally be true random samples so that we can know what the true infection rate is. Some locales are doing that kind of testing, and I'll be much more interested in those reports than studies in which respondents can opt into participation.
No. As far as I can tell, approximately 100% of people in New York want to be tested, for many different reasons. You're assuming something and projecting it as truth.
"These studies would ideally be true random samples so that we can know what the true infection rate is."
But by your own logic, there can be no "random sample"...we can't tie people down and force them to be tested, so we have to ask them. This means we're back to the self-selection problem that apparently ruins the sample.
The truth is, the New York study did pretty much what we always do to get a random sample: asked a bunch of people at random to participate in the program.
> No. As far as I can tell, approximately 100% of people in New York want to be tested, for many different reasons. You're assuming something and projecting it as truth.
In NY, the issue was not about wanting to be tested, it was about "are you there when they're picking people to test"; quoting from the article, which is very thin on info, people were selected "at 40 locations that included grocery and big-box stores". (In Santa Clara, with the Facebook ads, people "wanting to be tested" by going to a testing center is not anywhere close to 100%.)
The assumption that people who are at grocery and big-box stores are representative, infection-wise, of the population of a whole is not obviously true to me. Many other posters in this thread point out reasons why they might be more likely to be infected.
> But by your own logic, there can be no "random sample"...we can't tie people down and force them to be tested
This approach may go against American sensibilities, but other states (SK, DE, IT) have had better luck with the approach of sampling everyone in a town/region/area. Or at least pick people truly at random, not just a random sample of people who happen to show up in a public place in the middle of a pandemic with a shelter-in-place order in effect. (Also, I'm not sure how "my own logic" makes any claims about the impossibility of a random sample, could you elaborate?)
Even in America, you can at least try to account for the ways in which your sample is likely to be different from a truly random sample. The study by Stanford in Santa Clara county I know at least tried to do this, but of course you can't account for "thinks is infected" as a demographic attribute because that's the dependent variable; I don't know if the NYC study did this type of accounting, but I assume they tried to too.
> The truth is, the New York study did pretty much what we always do to get a random sample: asked a bunch of people at random to participate in the program.
You clearly know about sample bias, so I'm not sure why you think going to a single location (or even 40 single locations) and randomly asking people to be sampled is a good way to get a random sample. Think about going to every Costco and Whole Foods in the state and asking "people at random" for their political views. You're going to get a biased sample, even though you're nominally asking people at random. You can try to account for bias, based on demographic factors you observe in your sample that correlate with political views, right?
But we don't know enough about infection rate to be able to effectively account for infection rate given that people who are showing up at grocery and big-box stores on a random day are more likely to be exposed (they're at a store!) and thus more likely to have been infected than the population as a whole? Literally the only info you have is that they're more likely to be infected -- there aren't well-established demographic correlates with infection.
"In NY, the issue was not about wanting to be tested, it was about "are you there when they're picking people to test""
It's the same critique. If you pick the people anywhere other than from their home, the argument is that they're not at home, therefore they're more likely to have it.
OK, so how do you get them in their homes? If you solicit them on the internet, then there's sample bias because you're telling them they'll be tested. If you knocking door-to-door, it's the same thing: you have to tell them what you're going to do, so you're "selecting for people who want to be tested". You can't win.
"Or at least pick people truly at random,"
You can't force people to take blood tests against their will. You have to tell them what you're doing, and why you're doing it, and they must consent to participate. The same objection always applies to any "random" selection of humans: it's biased towards the people in the place at the time of selection, who agreed to be selected.
"you can at least try to account for the ways in which your sample is likely to be different from a truly random sample."
Right, as all legitimate researchers in this area do.
As I said before: this is all just a highbrow way of rejecting studies for lowbrow reasons. You will never find a survey without some form of sample bias. You control for it and move on.
We are now seeing multiple independent serological surveys with different methods pointing in the same direction. It isn't a methodological error.
> > you can at least try to account for the ways in which your sample is likely to be different from a truly random sample.
> Right, as all legitimate researchers in this area do.
Yeah, obviously we agree here. So: how do you effectively control for them in this case? That's my point. It's hard, and you haven't offered any mechanism for this particular case, just assurances that people who know what they're doing do in fact know what they're doing.
Are you one of those people? Please fill us in on what they're doing to address sample bias. They "try to control for it" -- how in this case? So far you've offered nothing specific, just that experts control for it, and you're implying that it's illegitimate to question whether there might be a systemic bias because all the cited studies seem to select populations more likely to be infected.
> OK, so how do you get them in their homes? If you solicit them on the internet, then there's sample bias because you're telling them they'll be tested. If you knocking door-to-door, it's the same thing: you have to tell them what you're going to do, so you're "selecting for people who want to be tested". You can't win.
Am I understanding correctly that you're saying that because you can't get a perfectly random sample you shouldn't try to minimize selection bias? You can't "win", but you can get closer than these studies did. There's a clear difference between a Facebook ad "Stanford seeks people for COVID-19 tests" and "Your number was chosen at random and we'd like to test you for COVID-19 in the interest of science."
One should absolutely try to minimize selection bias, in addition to controlling for its inevitability. In Germany, as you propose, they are selecting people at random from a central database of residents, which is not correlated with whether they have symptoms, feel comfortable shopping in public, etc. That is better than showing up at a grocery store, clearly, right?
As for people who agree to be selected, yes, there's nothing we can do about that in liberal societies. I'd like to know what this rate is. In Germany, I recall that it was low.
> this is all just a highbrow way of rejecting studies for lowbrow reasons
OK, this is the second time you've implied that I'm uninformed, or am acting in bad faith or with bad motives (or whatever "lowbrow reasons" means). None of these are true. I've ignored the personal nature of your vague dismissals up to now in the interest of conversation, but I'm done doing so.
"Am I understanding correctly that you're saying that because you can't get a perfectly random sample you shouldn't try to minimize selection bias?"
No, you're not. I was explaining why your argument is a truism in disguise.
"OK, this is the second time you've implied that I'm uninformed, or am acting in bad faith or with bad motives (or whatever "lowbrow reasons" means). None of these are true. I've ignored the personal nature of your vague dismissals up to now in the interest of conversation, but I'm done doing so."
I don't know if you're doing it in bad faith or not, but you're definitely making a highbrow lowbrow dismissal. You've set up an argument that can never be refuted, for reasons I've explained.
> You've set up an argument that can never be refuted, for reasons I've explained.
I'm arguing that the selection mechanisms they're choosing to use are bad compared to what they could and should be using, namely, random sampling as was used in Germany. Yes, those still have problems as you've mentioned, but they are much better overall. I'm sure there are reasons these studies didn't use those mechanisms -- some bad, like that it's hard to get a good random sample, and it's easy to run facebook ads or camp out at a grocery store, and probably some good, like that it's easier enough that it makes results available sooner.
I'm also arguing that because of this bias, and because we are now seeing a few of these studies with similar biases despite different methodologies, and because, of the papers we could read (namely just Stanford's), the accounting for selection bias is weak, it's risky to make statements about trends indicated by these papers.
You could refute my argument by explaining how these not-very-random selection mechanisms could be effectively accounted for post-selection. I've asked you to do this several times, and you have not done so.
I am not making an argument that can never be refuted. I'm asking you to refute it, and I've given you one potential argument, and you have chosen not to do so, repeatedly.
I'm sorry, but in this exchange, the person putting forth arguments that amount to vague dismissals that can't be refuted was, um, not me.
Exactly, CFR is always expected to be higher than the IFR. Most people getting sick don't go to the doctor. The CFR is an indication of how many people feel they need to go to the doctor and get treatment due to how severe the illness is. And of those that seek treatment how many die.
In comparing the COVID-19 IFR to the flu IFR, it is important to remember that flu vaccines are widely available and limit the spread of influenza. For example, CDC retrospectives for 2018-9 estimate that 35M Americans[1] got influenza over the flu season, or less than 12%.
By contrast, the current COVID-19 infection rate in New York (from these data) is already higher than 12%. So COVID-19 has the potential to infect a larger proportion of the population than the flu usually does.
(If the CDC data is correct, the flu shot may save around a hundred thousand lives per year. Don't skip it!)
It's actually important to point out, the CDC is very explicit that they are estimating flu illnesses, in their own words "an estimated 34 million people had symptomatic influenza illness"
Many people are attempting to extrapolate asymptomatic COVID to estimate an infection fatality rate and then making comparisons to CDC flu data for symptomatic illness, which is completely wrong. Actual flu sero studies suggest a large fraction of the population gets flu every year[1] and IFR is more like 0.02->0.05%, so even taking NYC numbers at face value COVID looks 10-50x more fatal than flu.
Yes this is happening all over this thread and is absolutely not a valid comparison. I'm also seeing people compare age-stratified Covid rates with overall flu fatality rates which is also invalid for similar reasons.
>IFR is more like 0.02->0.05%, so even taking NYC numbers at face value COVID looks 10-50x more fatal than flu.
Exactly, my only point in mentioning approximate flu IFR was to demonstrate that it's significantly lower than flu CFR. And so the NY COVID-19 serological study suggests a CFR well above 1%.
Is there any actual evidence that the flu vaccines indeed work? Given that 'flu' is actually a very broad term for a disease that can be caused by =many different viruses that also mutate each year.
That point is entirely true but doesn't really change the policy implications unless you believe that we can actually contain this thing indefinitely (i.e. that a given individual can avoid being exposed to it indefinitely)
Nope, because they're all designed to skew higher. People who are out shopping, people who will go out of their way on a weekend to be tested, healthcare workers... all of them are likely skewed high. One might as well conclude (with as much evidence) that the purpose of the tests is to show higher infection rates. Still, nobody doubts that we're under testing by a factor of 3-5x to catch a majority of cases.
What's disturbing is that if you take the estimated statewide data of 2.7M infected (out of ~20M) at face value and only the current 21k dead you still get 0.7-0.8% fatality rate.
In support of my statement, it is now noted that the lead professor's wife sent out misleading emails to find people who wanted to return to work, if they tested positive. The professor says he knew nothing of it and because they are wealthy they would skew low rather than high.
Since false positives can be low single digit percentages, anything outside the NYC data is pretty meaningless, and a bunch of meaningless data is still meaningless.
>There are many more infections than we know about and the CFR is not anywhere close to the actual IFR.
That is the optimistic takeaway.
The pessimistic takeaway is that even the hardest hit areas are nowhere near herd immunity levels and that we are either going to be isolating until a vaccine is created or we can expect to see a lot more death once nonessential people are forced back to work.
I don't think this a binary outcome. Getting closer to herd immunity also makes things better. Take NYC. Now 1 out of 5 people you see can't infect you (assuming having antibodies means immunity). The closer we get, the better it is. Having 21% "immune" leads to far different outcomes than .1%.
I'm no expert here, so someone can correct me if I am wrong. But I believe as long as each infected person spreads it to more than 1 other non-infected person the disease will continue to spread until there is a herd immunity level of infections. That spread rate will decrease as more people have antibodies, but it seems unlikely to get below 1 since even all the stay at home orders haven't been able to get that number much below 1.
Sustained spread of the infection stops when percentage of people having immunity is 1-1/R0. So if R0 for Covid-19 is (hypothetically) 3, you need 67% immunity to reach herd immunity. For highly contagious diseases like smallpox that number is very close to 100%
The 1-1/R0 is the point that the virus starts to burn out but there is still a long tail of infections. The total population that is infected will be higher than 1 - 1/R0. Also if the estimates of IFR were too high than the estimates of R0 were too low.
Yes, but there's a world of difference between Rt=1.5 and Rt=2.5, both in the degree of controls necessary to bring Rt close to 1 and in how far it can get out of hand if your controls are somewhat insufficient.
> The closer we get, the better it is. Having 21% "immune" leads to far different outcomes than .1%.
It depends on how long the immunity lasts. If it's permanent, this is indeed great. If it lasts only a few months, this 21% won't make any difference for another wave of infections this fall.
We use the term "immunity" but we should remember that there is a difference between the presence of active antibodies and the presence of immune "memory cells". The latter hang around long-term, even if the former disappear. So at a minimum if one does develop infection they will recover far sooner and with better outcomes. And likely will reach a far lower peak viral load which might bring down transmission.
Of course, there's other confounds. Like the somewhat recent discovery that we end up with antigen experience / "memory" of diseases that we've never contracted. (Perhaps from viral fragments in the environment??)
At this point people advocating the position you’re advocating for are in a state of denial (this is my opinion, not a matter of fact, obviously). Your assumption is that we can effectively prevent the majority of the nation from exposure via lockdown.
Not only does evidence seem to point against that, but when you do the math on mortality due to suicide and overdose it’s not clear that containment would even save more lives in the long run.
Here’s how you can tell people’s philosophical positions: if they talk about fear of a “second wave” they are Containers, since that implies the initial “wave” will not infect the majority; ie the virus is successfully contained (EDIT: See https://news.ycombinator.com/item?id=22961927 for the caveat here).
Ironically, leaders like Fauci are verbally saying that containment is not the strategy, yet every word that he says and the IMHE model everyone is relying on are all the result of a Containment ideology.
The alternative is what I would call Pareto mitigation. The vulnerable portion of the population self isolates, while the rest of us are _allowed_ to resume working and living more or less normally (still no large gatherings presumably).
I'd like to take this moment to put out a brief PSA that the serological data coming out, while not 100% reliable, is all telling more or less the same story. Let's look at these IFRs (the second link is CFRs but for Italian healthcare workers who presumably are all getting tested so I'm treating it as a de-facto IFR):
(I'm linking to the reddit comments instead of the actual study because they're really nice tables and the links are still there for anyone who wants to double check)
As others have said, for those around age 45 or less, Covid is equally or less dangerous than Influenza. And particular for those under 30 the flu is an order of magnitude more deadly at least.
In the general population overall, Covid is undeniably more deadly than the flu, but only about 3-5x (and I think 3x personally right now).
Recall that the flu is characterized by deaths in the very young and very old, while being less harmful to those "in between", purportedly due to the "cytokine storm" which is a scorched earth reaction of the immune system. Covid is very different, it is extraordinarily deadly to the very old, extraordinarily non-deadly to the very young, and about the same as the flu to those in between.
A disease with such a "spiky" (highly variable) mortality rate based on your risk factors is precisely the kind of disease that is most effectively treated with risk-informed self quarantine rather than a national lockdown.
Unemployment is correlated with a 2-3x higher chance of suicide, of which perhaps half can be explained away by mental health confounds [1]. There's unique factors in play here - rampant social isolation and widespead fearmongering, propagated even by health experts and "trusted" news sources at times - that lead me to believe that the spike in suicide and overdoses will actually be much higher than predicted by just unemployment alone.
We're currently at 50,000 suicides per year in the US as a base rate, it is not unimaginable that we would see at least 50,000 _extra_ suicide deaths attributable to a mixture of lockdown and the general socioemotional environment.
--
I haven't even gotten to the philosophical battle of "freedom versus security". I am, ideologically, someone who drank the koolaid and really believes in freedom and civil liberties over "security" (which I view as illusory anyway), but _even just viewed through the lens of reducing mortality_, the evidence is stacking up that lockdown is going to do more harm than good.
Is the evidence fully settled? Of course not. But it's shocking to me how many people seem to be operating off of the projected CFR's we had in early February, shouting from the rooftops about "1 in 20" people dying (random recent case in point: https://news.ycombinator.com/reply?id=22952764&goto=threads%...). I don't know whether it's just that a large swath of the population already had clinical anxiety which is further magnified by social isolation and social media and news headlines, or whether something else is at play, but I'm very concerned about the state of discourse in the United States right now, and more broadly, the entire world. In fact, ironically I feel a bit luckier to be in the US than some of these other countries because in the US _every_ issue is partisan, which while entirely irrational means that roughly half the country will be in favor of ending the lockdown at any given time (the position I am advocating for, within reason, insofar as hospitals are not overwhelmed), as opposed to other places where you can get given a $1600 ticket for driving a car by yourself, based off of a superstition that _being outside_ causes Covid as opposed to exposure to infected respiratory droplets...
--
EDIT: Lastly I should mention that in a perfect world we could have voluntary variolation; I would love to be able to expose myself to a controlled dose of SARS-CoV-2 and self isolate for several weeks to ensure that I can never pass on Covid to someone else. Unfortunately that would be very hard to make a reality due to the political environment, even though I am advocating for it to be totally voluntary. I was heartened to see this recent paper toying with a variant of that approach: https://www.medrxiv.org/content/10.1101/2020.04.12.20062687v... (I don't agree with an "Immunity Card" for ideological reasons but I'm glad we have a paper attempting to model it out which does show benefit of voluntary self exposure)
Hey, I'm in broad agreement with you, but, a nit ...
> Here’s how you can tell people’s philosophical positions: if they talk about fear of a “second wave” they are Containers, since that implies the initial “wave” will not infect the majority; ie the virus is successfully contained.
Here in the SF Bay Area we pretty effectively blunted the first wave with public health controls, which meant health care was not stressed much. Despite the Stanford serosurvey saying that we have 3% with a history of infection, the real number is probably more like 1%. Daily case counts are decreasing, so the initial "wave' did not infect the majority.
It's not enough to decrease Rt noticeably at all. We need to figure out how to loosen up controls in a way that provides economic benefits, keeps a nice constantish burn towards herd immunity, but with enough safety margin to prevent catastrophe that New York got too close to.
Instead, our public health seems to be further tightening instead of experimenting with small measures to relax the controls. There's a loud contingent advocating what would effectively be permanent controls.
So, that's definitely a good point to raise. I agree and should have been more clear. I think there should not be a second wave because I think we need to resume more or less normal society and let people naturally get exposed (since voluntary self exposure will not ever be tenable in the US I fear).
But given the way that California aggressively locked down early into this, I agree that we have been sloping down and thus there is guaranteed to be subsequent waves.
Basically, the waves are real but are arbitrary and are caused by our own misguided interventions.
> We need to figure out how to loosen up controls in a way that provides economic benefits, keeps a nice constantish burn towards herd immunity, but with enough safety margin to prevent catastrophe that New York got too close to.
100% agreed. Now the argument I do somewhat agree with is that it's very hard to get that balance right when dealing with exponential transmission. Which is why on balance I'm leaning much more towards "we'll cross that bridge when we get there" (because the alternative is that we have to stay permanently locked down).
Also the portion of the population I am advocating should be allowed to return to work is precisely the portion of the population that produces very low hospitalization rates.
I strongly believe that anyone who has been reading CNN the last few weeks would be _shocked_ to learn that we get 1 hospitalization for every 500 20-29 year olds infected (and again, these numbers are not fully settled but they're at least in the right order of magnitude IMO)
> There's a loud contingent advocating what would effectively be permanent controls.
Yup, this is what has me really scared. The widespread belief being that it is actually plausible to avoid ever getting infected and therefore any infections that follow a softening of lockdown introduce deaths that never would have occurred in the hypothetical alternate universe.
--
So, thanks for raising that point, I fully agree. The TL;DR is that implicit in "we need to watch out for the second wave" is the notion that "we need to fight these waves and halt their spread" which I strongly disagree with.
EDIT: And just to be clear, if we "re-open" we'll still all be wearing masks and keeping arbitrary distance between each other so it's not like we're all running around exchanging bodily fluids willy nilly. But I really do think that the shutdown has been, in some part, effective in curtailing spread, and thus naturally I would expect a higher infection rate following a re-opening.
EDIT 2: Removed the part about the political leanings of those advocating for long-term lockdown because it's going to set off people's defense mechanisms and potentially cause them to draw the wrong impression of what I'm saying
In San Diego people under 30 are close to 20% of total confirmed and over 5% of hospitalized cases. About 7% of confirmed are hospitalized for this age group. Our positive test ratio is under 7%. Tests are constrained by policy but not dramatically so. https://www.sandiegocounty.gov/content/sdc/hhsa/programs/phs...
Another data point is the US military. I believe they test much more widely due to their living conditions. So far the cfr is about 0.4%.
Young people interact with more individuals, and engage in more behavior that exchanges microbes such as unprotected sex, protected sex, sharing pipes/vapes, sharing utensils, etc.
We should expect that the proportion of people with Covid should be heavily biased towards the young for that reason, as it is for any disease that does not have a biased transmission.
Most indications I've seen is that the young are not harmed much by covid, but can still get infected just as easily, thus why we see:
"X% of hospitalizations are young people" as opposed to the statement that people actually interpret it as: "X% of young people infected with covid are hospitalized".
For any test that is not given randomly or universally, there is a selection bias. A test given based on severity of symptoms bias heavily against patients with light symptoms, presumably younger people. But to get from 7% (hospitalized to confirmed) to 1 in 500 is a factor of 35. BTW the city of Canton (Guangzhou) just tested all their high school seniors in preparation for school reopenning and found zero positive case so far. And they did find cases among other groups.
I don't think anyone enjoys lockdown. At a minimum we need sufficient testing, sufficient ppes, sufficient surge capacity and training for standard supportive care. We are still quite a ways from that. Even if death is unavoidable people deserve to have reasonably good care.
> But to get from 7% (hospitalized to confirmed) to 1 in 500 is a factor of 35.
This is broadly in the range implied by a number of measures across the entire population. 20x for randomized population sampling (Iceland), 20-80x from serology studies, etc.
> I don't think anyone enjoys lockdown. At a minimum we need sufficient testing, sufficient ppes, sufficient surge capacity and training for standard supportive care.
Here in the SF Bay Area, ICU usage peaked 3 weeks ago, and controls have been tightening. We're well below seasonal norms for ICU usage at this point. Surely the original controls that caused the original trend down in disease would be good enough now, and could even be loosened slightly further to have more parity with other jurisdictions that are successful.
> In San Diego people under 30 are close to 20% of total confirmed and over 5% of hospitalized cases.
People under 30 are over half the population in San Diego, so you've just said there's 1/10th as many of them in the hospital as we'd expect if the total risk were equal.
The serology data tells us that an even greater portion of cases in the young are missed-- they are probably actually half the number of infections in San Diego, but a much smaller proportion of hospitalized cases.
The current data implies something close to what he said: about a 1 in 500 to 1 in 1000 chance of hospitalization for people under age 35.
It's not like business was going to go on as normal when people started filling up hospitals.
The lock down was a choice between what we are going through now and a much worse disaster if the spread continued through a couple more doublings. Of course the extreme stay at home orders should be continued for as short a time as possible, but what we do next also doesn't have to be exactly what we were doing before (and realistically won't be).
My impression at the moment is that containment isn't going to happen, but mostly because people don't want to bother, not because it is the more costly alternative or impractical. Given the lack of appetite to test and isolate, we had better all start doing things like wearing masks.
Note that ~80% of American's favor the stay at home orders right now. It's not a partisan issue, it's an issue where a loud minority is getting some attention from the media.
> As others have said, for those around age 45 or less, Covid is equally or less dangerous than Influenza. And particular for those under 30 the flu is an order of magnitude more deadly at least.
Can you share the specific numbers you used to reach this conclusions? The reason I ask is because I'm seeing a lot of people in this thread comparing:
1) Overall influenza fatality rates to age stratified Covid fatality rates, and
I agree that finding properly stratified data that is comparing IFR and not just symptomatic CFR is quite tricky.
At a minimum though, if we assume the chart I linked is CFR and compare to the Covid 5-20% CFRs we've seen, we still get that order of magnitude difference for the very, very young. But like I said in my previous comment here, I don't think the statement for all under 30's holds. It'd have to be <18 years old (and to be clear while https://bmcmedicine.biomedcentral.com/track/pdf/10.1186/1741... is helpful, I still haven't proven anything)
I’ll try to dig up some sources later today. But briefly, on second thought “under 30” was sloppy of me to say and so that statement is probably false. I would expect for those 25-30 years old they should fall outside the range where Influenza kills a lot. So I think the order of magnitude is almost certainly true for those under 18, but I am now doubting it holds all the way up to age 30.
> when you do the math on mortality due to suicide and overdose it’s not clear that containment would even save more lives in the long run.
I'm waiting for people to do the analysis of mass unemployment from lockdown leading to people losing their healthcare, just in time for the pandemic to widen or return this summer or fall.
You're exactly right to consider all of the externalities of our current approach.
I hadn't thought of it specifically in terms of our broken employer-based healthcare system. That's a really good point.
I quit my job Feb 7, before all this unfolded. And COBRA costs me $612.54 per month.
For the same reason I was able to quit my job, I will be totally fine. I have over 5 years of living expenses in liquid assets, so I will be fine.
But for someone who was living paycheck to paycheck and lost their job because it's not the type of job that can work from home, how the hell could they possible afford healthcare?
Along those lines, I do really think that part of the reason so many people don't see the harm in locking down for the next several months is because, like me, they work in the tech industry and really have not been affected by this as far as employment goes.
Thanks for this comment, puts into words what I have been casually thinking. Also, this is the strategy in Sweden, seems to be working out fine for them.
Thanks, I've been trying to find the words as well, which is really difficult when expressing a position that runs counter to "we need to lock down for the next 18 months" is basically characterized as wanting to kill granny.
I generally have to spend more time prefacing the ideas with "I'm not a trump supporter and I think covid is real and I think it's more deadly than the flu and..." than talking about the actual ideas themselves. Especially on Reddit...
I agree with some of what you said here, but there is one specific point I want to disagree with. All the projections for unemployment related deaths are based on society functioning as it previously did. I think many of the people who are suggesting we are in for long term isolation are also suggesting a much bigger increase in the social safety net to help people through these difficult times.
Right, but I think some of that suicide rate comes from the lack of "purpose" (it's silly that we rely on our jobs for purpose but we really do), or more broadly the desynchronization of one's internal schedule that many of us have experienced (which leads to worse sleep and therefore higher mortality).
I also think that given we know about the level of competence of our government, it will be very difficult for "real" safety nets to be put into place. That's not even getting to the partisan divide inherent to our system.
So in my eyes, we had one big problem, covid, and turned it into two big problems, covid and an economy in ruin. And these two problems affect a greater set of individuals than either one alone. Since a huge number of deaths from Covid are those who weren't working because they're 70+.
Lastly, one of the most "famous" social safety nets in America is social security, which is a farce that is known to redistribute money from the poor to the rich. (It is an eternal frustration of mine that the Left in America is so heavily in favor of social security despite it being flawed from its inception) [1]
You are missing the point in that if nothing was done in terms of trying to (at least temporarily) contain COVID, then the economy would be just as bad off.
We're currently at 50K deaths in the US, largely in a four week period. I think it's safe to say that the shutdown has cut the death rate in half, if not more. If 100K people die in 4 weeks in the US, the people will shut the economy down whether the government initiates it or not.
Either way we're screwed. And I think your data is far off in terms of IFR and CFR. Imagine the IFR being a magnitude higher for COVID versus influenza. I personally think this is a safe guess based on what we've seen in a wide variety of environments, complicated by different testing methodologies etc etc. So we'll use 1% for the IFR.
Now current thought is the R0 is between 2 and 5. Hopefully closer to 2 since R0 dramatically affects the percentage of infections required to reach herd immunity. But to be simple, we'll assume we need 50% to reach herd immunity. So we'll use 164M as the number of Americans needing to be infected before we can "return to normal."
With 164M cases, and an IFR of 1%, 1.6M Americans will die. An additional 9% will need hospitalization, but survive. That's 14.8M.
Say we're off by a lot and "it's just a flu" with an IFR of 0.1%. You'll have 164K fatalities, and we'll still have the 14.8m hospitalizations. Say the hospitalization rate is off by a magnitude, it's still 1.48M.
There's no way our country can operate "normally" with these types of numbers.
I will grant you the lack of purpose being a motivator for suicide, but none of us know how that would transpire during a global pandemic. I know there is debate over the specifics over Maslow's hierarchy of needs nowadays, but generally speaking I think a lot of us will focus less on these internal issues in times of external danger.
Also if we are going to use government incompetency as an argument here. I would throw it right back at you and say I have little faith in the government implementing a reopening plan that doesn't either kill people or damage the economy long term.
Lastly it is a false dichotomy to present a choice between protecting people from COVID-19 or protecting the economy. If we refuse to do the latter, we are going to ruin the economy anyway. Hundreds of thousands of people dying would certainly put a damper on demand. And while there are certainly people protesting about reopening sooner, I don't think think movie theaters would be selling out if they opened tomorrow.
>In the general population overall, Covid is undeniably more deadly than the flu, but only about 3-5x (and I think 3x personally right now).
I don't see how you get close to 3X more deadly than the flu. If 14% of New York state residents have been infected, 20M population, 15000 deaths + another 3000 that are infected now and will die (this disease takes a long time to kill people) you get an IFR of 0.64. If the IFR for flu is 0.05, that makes covid 12X more deadly than the flu.
Lots of people reporting an IFR of 0.5 based on the NY serological data; that is "right censoring" the deaths. It's got to be a bit higher than that. Either way you have covid with 10X the infection fatality rate of the flu. If worst flu years have IFR of 0.1 then covid is still 5X worse than the worse flu seasons. And as contagious as the worst flu seasons too.
So twice as contagious as the flu, 10X deadly. Much more likely to put you in the hospital, and thus overwhelm the hospital system, causing many ancillary deaths. This NYTimes article sums it up: https://www.nytimes.com/2020/04/06/well/live/coronavirus-doc...
At the peak in NY, hospitals stopped seeing heart attacks and strokes, because those people were too afraid to go to the hospital. Many of those people died at home, as supported by the overall death rates in NY.
The narrative put out there by those that look at the recent serology results and say "this proves that this disease was really just a bad flu all along; we can reopen the economy without fear" is just not supported by the data. An IFR of 0.64 and hospitalization rate of double that, like 1.2%, for a disease this contagious, shuts down the economy until we get a vaccine or effective treatment or a Korea-like testing/tracing regime in place.
Yes suicides go up. But this is more than compensated for by other benefits. Overall, people may well be much healthier in a depressed economy. We will certainly see a decline in car accidents.
> The pessimistic takeaway is that even the hardest hit areas are nowhere near herd immunity levels and that we are either going to be isolating until a vaccine is created or we can expect to see a lot more death once nonessential people are forced back to work.
The third option is, when you take into account that it's approximately as bad as the flu for folks under 40, we let out the young and keep the older folks and the vulnerable inside. This will boost our progress towards herd immunity without materially increasing the death counts.
You can't have it both ways. If IFR is low, then that means R0 is way higher than we thought, which also means isolating only the vulnerable will not work.
The fact is, the number of deaths is too high in NYC to be able to cherry-pick your way to an argument that supports your view. Either R0 is really high, and we need to shutdown to prevent it from infecting the entire population in a very short time span, or the IFR is way higher than the flu and we need to shutdown to prevent it from killing a lot of people.
The only sane way to open things back up at this point is to implement widespread testing and contract tracing.
> If IFR is low, then that means R0 is way higher than we thought, which also means isolating only the vulnerable will not work.
No. If you miss a constant percentage of cases, you get the same shape of exponential curve.
That is, a virus with an R0 of 2 and 1 case doubling to 2 and 2 cases doubling to 4.... looks the same when you miss 99% of infections and see 1 of 100 infections doubling to 2 of 200 doubling to 4 of 400.
This fallacy has been common in the response to this data, but it makes no sense. Large numbers of missed cases shift the curve forward and backward in time, and don't change the shape of it.
All of the current findings are still consistent with R0 in the range of 2.0 to 2.5.
We are talking about 1 serological study at a single point in time. We have no idea what percentage of cases we were missing before or after that point.
There's no reason to interpret the serological study to think a higher R0 is implied, which is what you said.
The only way that missing cases affects estimates of R0 from time series is if we're missing a much bigger proportion of cases now than we were in the past. All the evidence I've seen leans the other way-- more testing, better testing, more cases diagnosed by "presumed" criteria.
So, it actually implies the opposite of what you're saying, and argues for lower R0.
I am not interpreting it that way, I believe in the consensus estimates. The parent commenter (articbull) thinks these serological studies (SC, LA, and to a lesser degree NY) prove that the IFR is much lower than we thought because the actual infection numbers are much higher. The NY study, using reasonable assumptions, looks like the IFR might be between 0.5% - 1.0%, but the SC study was claiming 0.1-0.2%. And I am arguing, in the hypothetical case that the SC study is correct (I think it is flawed), then we have under estimated R0.
Considering we have been under lockdown for over a month (which greatly impacts effective transmission), and considering our estimates of latency hasn't changed dramatically, I think my above assumption is fair.
> And I am arguing, in the hypothetical case that the SC study is correct (I think it is flawed), then we have under estimated R0.
I don't necessarily believe the Santa Clara County study, but there are plenty of other alternatives, including a pretty likely one: gross undercounting at the start of the epidemic and/or earlier cryptic spread in communities. This is something that SCC public health officials / Dr. Sara Cody has stated in recent days is her belief-- I am not sure I agree.
At this point we're accumulating a lot of evidence IFR is 0.3-0.5%.
I have never heard of any healthy person under 40 dying from purely the flu. COVID-19 is certainly less dangerous to the young than the old, but there are plenty examples of it killing young and otherwise healthy people.
I skimmed through that so please point it out if I missed it, but the numbers don't seem to account for comorbidities. The question was whether COVID-19 is more dangerous than the flu to healthy young people not all young people.
Well even with your qualifier of "no comorbidities", the evidence is pretty unanimous that Covid and Influenza are not even the same ballpark. Worrying about a healthy young person dying of Covid is more akin to worrying about a healthy young person dying of cancer than it is to bacterial meningitis or something.
As I've said elsewhere, Influenza is defined by being deadly to the very young and very old, and so-so to those in the middle. Meaning that healthy young people do regularly die of the flu (it's still rare in absolute numbers, but it happens WAY more than covid).
Unfortunately, and I know people read this stuff and their quack heuristics start firing, the reason people are so afraid about Covid's impact on young people is because the mainstream media has intentionally promulgated a narrative that "young people are at risk too" because they fear that otherwise young people wouldn't submit to glorified house arrest for months straight.
Yes, young people get _infected_ by sars-cov-2, but they do not develop deadly cases of covid in any appreciable numbers. You really should be worrying far more about the flu as it pertains to a college-aged demographic.
EDIT: I do need to find some sources for you though. I saw some NY data broken down by comorbidites but am having trouble finding it.
> Unfortunately, and I know people read this stuff and their quack heuristics start firing, the reason people are so afraid about Covid's impact on young people is because the mainstream media has intentionally promulgated a narrative that "young people are at risk too" because they fear that otherwise young people wouldn't submit to glorified house arrest for months straight.
They've reported that because young people who very likely won't die from COVID-19 (they may still need to be put on oxygen and may have lifelong effects from it, but they likely won't die) can still spread it to those that will. Further, they'll put more strain on the health system, and the ability to care for COVID-19 patients is a big determinant in the fatality rate. Everyone needs to stay practice social distancing and shelter, because everyone can carry and spread.
Young people are also drastically less likely to get severe disease, and drastically less likely to be hospitalized, see FIGURE 1. [1] I find it very suspicious that young people who don't go to the hospital with severe disease would require oxygen or end up with life-long lung injuries.
> Everyone needs to stay practice social distancing and shelter, because everyone can carry and spread.
Sweden has demonstrated at national scale that's not the case. [2,3]
The reality is about 60% of us are going to get it one way or the other, so let's control which 60%, and in what order, and on what timeline before everyone stops listening and just walks out.
> I find it very suspicious that young people who don't go to the hospital with severe disease would require oxygen or end up with life-long lung injuries.
Or I don't know, Google ventilator effects. Lung scarring. Young people get this, and it hurts them, maybe forever. It's not common, but it's absolutely not never, and you can still spread the virus if you're asymptomatic. Seems like people should try and avoid getting it!
> Sweden has demonstrated at national scale that's not the case.
Sweden is so unlike the US in so many ways, this is a worthless anecdote. Imagine the differences in demographics, density, culture, literally everything.
> The reality is about 60% of us are going to get it one way or the other, so let's control which 60%, and in what order, and on what timeline before everyone stops listening and just walks out.
It's by no means certain 60% of us will get it. Especially if we do what we should do and build a competent testing and tracing system. The attitude that this is inevitable is lazy and puts thousands of lives in jeopardy.
Permanent restrictions, drastic social change, and large contact tracing infrastructure might keep it under the herd immunity threshold of ~60%. It might not, too.
But you can't let up at any point, because it'll still be endemic and ready to explode up to that herd immunity threshold.
I prefer approaches that get us to that 60% in an orderly fashion-- staying short of healthcare overload, and preventing as much infection in the vulnerable as possible.
Just because you've not heard about it, doesn't mean it doesn't happen. Seemingly healthy young people die from the flu each year.
Since Feb 1, 204 people from 15-34 have been confirmed to have died from COVID-19 in the US. 162 people from age 15-34 have been confirmed to have died from influenza in the same period.
I don't think you'll find either statistic broken down by comorbidities.
Not sure why you added the lower limit at 15. The two high risk age groups from influenza are 65+ and under 5. If I recall correctly, pediatric flu deaths are the only ones tracked by the CDC (for other age groups they just use statistical estimates).
There is data on comorbities with the flu from prior years. This article says half of deaths had no preexisting medical conditions.
I picked 15-34 because it corresponds with the data table I cited from the CDC, which lists confirmed deaths from each for the time period. It's already cited and linked, but here's it again: table 2, https://www.cdc.gov/nchs/nvss/vsrr/covid19/index.htm
Note that a few of the deaths are -both- COVID-19 and influenza.
You mind sharing the source on those numbers because the total deaths is somewhat meaningless without the number of infections which I imagine is much higher for the flu? Even those absolute numbers show total deaths being just over 25% higher for COVID-19 than the flu.
The statement you differed with is "The third option is, when you take into account that it's approximately as bad as the flu for folks under 40," You're ignoring data in order to make a wishy-washy statement that it must be much worse for people under 40 based on anecdotes you hear.
I don't agree flu infections were higher for those 2 months, either. Distancing has been spectacularly effective against influenza, since it has a lower Rt in the absence of controls. Under 1% of influenza surveillance tests are positive right now, which is a level usually only seen in the middle of summer.
Our best guess for overall infection fatality rate is about 0.3%, double or 3x influenza (because of the very high death rate in the elderly), but COVID-19 deaths overall have been 4x influenza in that time period, indicating that COVID-19 prevalence is higher.
First off, all those numbers are once again absolute deaths and not related to case numbers so my prior complaint still stands. If you look at the first chart, you will see that COVID-19 deaths didn't start becoming a problem until March. Distancing practices were not widespread until mid-March. Meanwhile flu deaths weren't dramatically cut until the most recent timeframe. So taking into consideration both the first chart and the second chart that breaks out the numbers by age range, there seems to be an implication that COVID-19 deaths are much higher for young people in recent weeks. Therefore the absolute numbers would look totally different if March 1st was the starting point. I really don't see anything here that demonstrates the flu is as dangerous as COVID-19 for young people.
We've said approximately as dangerous. It clearly is in a very similar ballpark. We don't have enough data right know to know if it's twice as dangerous for people under age 30, or half as dangerous.
(We do already know it's -way less dangerous- for people under age 18).
Ok, maybe this dispute is based on different definitions. I don't think 2X is "approximately", but I do agree that we don't have enough data right now to know the exact difference so I can't say for sure that COVID-19 is more than 2X as dangerous.
I also agree that COVID-19 does appear to be way less dangerous for the under 18 group. However practically none of the under 18 group is able to isolate themselves away from the 18 and over group so there is no real societal benefit to this distinction.
> I don't think 2X is "approximately", but I do agree that we don't have enough data right now to know the exact difference so I can't say for sure that COVID-19 is more than 2X as dangerous.
I think we have enough data to exclude more than 2x. The range of potential danger has influenza's risk right in the middle of it.
> I also agree that COVID-19 does appear to be way less dangerous for the under 18 group. However practically none of the under 18 group is able to isolate themselves away from the 18 and over group so there is no real societal benefit to this distinction.
There's a huge societal benefit to this distinction, because if children are at relatively low risk, and also do not seem to be index cases very often for how COVID-19 is spread to households, schools can reopen. This is different from most diseases, where schools are often a key mechanism of disease spread. e.g. see https://www.medrxiv.org/content/10.1101/2020.03.26.20044826v...
That's pretty bad considering only perhaps 5% of the population has been penetrated with covid-19 vs 100% for the flu. Assuming a 70% eventual population infection rate, using your data we should expect to see around 2900 deaths in that age cohort. Annualized flu numbers would be around 500. So covid-19 by your data is 6x more fatal.
100% of the population was not "penetrated" with the flu between February 1 and now. Evidence implies a similar proportion of the population was "penetrated" with COVID-19 and the flu in that time. This implies a similar death rate for COVID-19 and influenza in those age groups, as the poster said.
Indeed, influenza has been greatly impacted by distancing. The surveillance numbers for influenza are lower than you'll ever see for this time of year.
I can't exclude it's twice as bad as the flu (or half as bad). But even those broadly agree with the poster's statement.
I’m on mobile right now so won’t be able to look up the influenza numbers, but perhaps you could find some quality data on influenza mortality in those without comorbidities?
> This still is not a random sampling. It only samples from shoppers at grocery stores and big-box retailers. Imagine doing the same study, but of people who ordered groceries online only. Would you expect to see big differences in exposure rate? I think so.
There's a larger, related problem with this: It's way over-sampling people who go outside a lot and make unnecessary trips, and undersampling people who don't. I'm in NYC and I'm only going to one store a week (the grocery store). So I'd have much less chance of being included in this study than someone who hasn't changed their habits and is still shopping much more frequently. And there's a direct correlation between how often you're going shopping and how likely you are to have been exposed to COVID-19.
So yeah, it's very far from a random sampling, and it is specifically over-sampling people who are more at risk. So I'd take it with a grain of salt; it could actually be way off.
They should've activated the jury selection tool for this and gone to people's apartments to draw blood. That way they'd correctly sample the people who aren't leaving home at all, or only do so rarely.
> This still is not a random sampling. It only samples from shoppers at grocery stores and big-box retailers. Imagine doing the same study, but of people who ordered groceries online only. Would you expect to see big differences in exposure rate? I think so.
I'd expect people who still do their own grocery shopping and go to big box retailers to be a better sample of the population than hacker news' audience definition (ordering from instacart/whole foods, working on Zoom/slack all day).
The vast majority of the population does not have the privilege that many people in the tech industry do. Going to a grocery store is a pretty basic activity in most of the population. People who get their groceries delivered online and can maintain their livelihood being on Zoom calls all day are probably statistically insignificant enough to be covered in the sampling error.
It's not black and white. A person who goes to the store more often has a greater chance of being tested, than someone who goes infrequently.
There's also the issue that people are trying to avoid other human beings right now, including the human beings administering tests. People who are social distancing more effectively are less likely to agree to a test.
> New York City had the highest positive rate at 21.2%. The potential good news to come to light is that the death rate may be far lower than some estimates, at 0.5%.
How do they get to 0.5%? According to NYC data if you include "probable" cases, 15,400 people, or 0.18% of the population, are already dead. Wouldn't that imply 0.9% fatalities if the whole city was infected?
The data that NYC publishes daily is super interesting since it breaks down cases/deaths by race, gender, age (group), and presence of underlying health conditions [1].
The gist of it is that the majority of deaths from COVID-19 were suffered by people who had some kind of other health issue. As of the most recent data (April 22 at 6pm), 61 deaths were from people with no underlying health issues, 7,474 from people known to have underlying health issues, and 2,755 from people with unknown status [0]. The elderly in poor health are much more likely to die than the young in poor health. But when healthy, the elderly have pretty close to the same fatality rate as the young [2].
It would not be ruled as a significant factor leading to death in many of those cases, and for the rest, NYC residents might be more at risk due to higher average viral dose (subways). Also there is a lag in IgG antibody production, the test has imperfect sensitivity, and some portion of exposed people may never develop detectable antibodies. Some of these confounding variables are probably baked into the estimate.
It is very positive in that we can be more confident that our assumptions are close to reality. 0.5%*330M still equals 1.65M deaths, or (very roughly) 5M ICU beds. Spread out over 52 weeks, that's a lot of ventilators and ICU beds and PPE, but it is a target. A very big juicy target that we can and should aim directly at, and HIT.
So having killed 1.6 million in the US alone and run ICU at 100% for a year, you’re just going to hope people remain immune and the virus doesn’t mutate having passed through billions of people worldwide?
The alternative being...? We know that covid-19 will likely mutate, just like other coronaviruses, H1N1 and its mutations are still around every year; are you suggesting we must remain in quarantine until we both create a vaccine for covid-19 and then every fall quarantine until we're able to develop a vaccine for that year's bout of viruses?
It's more or less epidemiologically impossible that you will see an attack rate of 100% (as you just assumed). Once attack rate reaches >40~60%, the reproduction number drops below 1 because there just aren't enough uninfected hosts.
It will probably be significantly higher than 60%, since COVID-19's R0 without lockdown seems to be pretty high. The CDC has estimated it at 5.7 [0]. Herd immunity threshold = 1 - 1/R0, so the threshold would be 82.4%.
That R0 assumes we are taking essentially no measures to stop the virus though, right? I don't no much about it but it seems like the R0 would drop significantly if we keep doing social distancing, expand testing, make people wear masks, etc.
Right. Really this study is just confirming (with higher error bars than we'd like) what we already strongly suspected.
Note that as others are pointing out, that 0.5% isn't really extrapolated correctly. The tested case deaths are looking like they're only 70% of the total or so. Also using a current death count with contemporary infection counts is a mismatch, because it forgets that ~35% of the people currently in the hospital won't ever leave, which is about 5k more deaths.
So a better calibrated CFR might be 1% or so, which is even closer to the range we've all been assuming.
Except this data is also lagging by three weeks, depending on what antibodies are being checked. Some deaths today have come from infections in April, so the net effect could be a wash.
Per https://covidtracking.com/data#state-ny there have been 57103 hospitalizations in the state (not city, so not quite the same population), of which 15021 are still hospitalized. So that's 42082 cases that have left the hospital, of which 15740 died, or 37%.
Obviously there's some apples/oranges here, both due to the population being different from the city and the fact that some of those deaths were outside of hospitals. But it's close.
It's been covered in the past that NYC's total fatality rate has been good but its ICU survival rate has been anomalously bad.
Yeah, still 150,000 death toll if everybody gets infected. And this is not considering the permanent lung and other tissue damage that's getting reported from "recovered" patients.
This was my thought exactly. Some of my friends in the Bay area got COVID-19 (confirmed positive tests) after quarantining for over two weeks at home and taking all of the precautions. The only place they had been was the grocery store, so they suspect that they got it there.
The incubation period runs out to around 27 days, though above 2 weeks is uncommon. It seems possible one of them already had it before quarantine, and just happened to be one of those who had a long incubation period.
I obviously don’t know whether that’s the case here but its also possible that some of these anecdotal stories (I’ve heard similar from colleagues) involve someone in the same household having had an asymptomatic case acquired prior to quarantining and transmitting it to others in the same household who then get symptoms after a week or two.
I'm in the East Bay and was quarantining for almost a month before my symptoms showed. I also assume I got it at the grocery store since that was my only outing.
Are they not taking public transit? Your friends must be some of the most unlucky people ever. The Bay Area has had under 3k cases since the start of April outside group living settings (nursing homes, homeless shelters, etc.). That's ~1/2000 chance of getting covid from anywhere outside.
Remember that’s not a count of cases. That’s a count of the under 3000 people that came up positive after permitted to take a limited availability test.
Take total case deaths, divide by CFR, that’s how many likely total infected there were 3 weeks ago, and then compound daily by the daily spread rate to find how many likely today.
I literally made the same comment a few days ago but here goes: every thread about a study such as this one is has criticisms concerning the scientific accuracy even though all these studies point to the same conclusion and not one of them provide evidence to the contrary.
This isn't exact science, we're in the middle of a pandemic. But all these antibody studies are pointing towards the same conclusion.
Do they all use the same testing mechanism? If they do, a flaw in that mechanism could result in them all being incorrect or misleading in the same way(s).
I can't believe people are still screaming false results at this point every single random sampling has been the same or point to more infections. At this point based on everyone's method of random sampling that should merit the badge of random sampling, but people argue it on every post.
The biggest problem looking at these tests is that they are completely on the opposite side compared to the mortality estimated using South Korea numbers.
We know for a fact that South Koreans tracked pretty much every single person with covid given that their new cases are increasing by a single digit per day. We know that as of today they have 10708 cases and 240 deaths, with still 1967 active cases.
This give us a lower bound for the mortality at 2.2%.
I think this is the best way to estimate mortality and I think that a lot of studies are giving numbers for the infected that are just out of touch with reality.
I remember an Oxford study that was saying that half the UK was infected at the beginning of April. Those numbers were simply insane.
I still don’t have a perfect explanation why the mortality numbers from the antibody tests seem so low, but I think that the accuracy of these tests or the way they select the people to be tested may be involved.
Do you see any flaw in using South Korea numbers to calculate the mortality?
South Korea apparently has conducted 11500 tests/ 1m population. It's ridiculous to claim that they detected every covid case and to suggest that their cfr is a lower bound for the ifr.
They tested everyone that was in contact with a reported case, I never claimed that they tested the whole population. How do you explain that they are showing only a single digit increase of new cases per day if, as you assert, there is a huge number of unaccounted cases?
Probably by implementing strict quarantine measures at the start of the pandemic. They likely had many undetected cases in the beginning as well, the quarantine reduced the pool of infected people a lot. Given that the number of infected people is relatively low right now, their testing capacity probably allows for fairly good tracing of cases, which is one of the primary reasons why their numbers remain low, in addition to factors such as social distancing and high hygiene standards. But the large number of undetected cases in the beginning would would reduce the ifr significantly.
Yes more people will live if they have access to care because the system isn’t overwhelmed. People in the later stages will also benefit from the institutional knowledge hospitals gain from treating patients. There will also be more PPE and essential medication that will make treatment more effective. Time is our ally.
This is like the third or fourth spot I've seen on this very site where people are postulating that covid deaths would have really died in "months" anyway. Why is this so weirdly consistent?
To be clear: there is absolutely no evidence for a number like that, none at all. Typical at-risk patients with comorbidity like advanced age or obesity are routinely expected to live for decades. Actual conditions with a death-in-months prognosis are extremely rare, and tend to require hospital care already. Think advanced heart failure or stage 5 cancer.
In Sweden, which only has a voluntary lockdown, over 30% of covid19 dead come from elderly homes[0]. Swedish elderly care is highly focused on keeping the elderly independent (at home) as long as possible, so the people in elderly homes are the ones in the worst state - 20% don't even survive one month from arrival[1] (I saw another statistic that 40% don't survive 6 months, but I can't find that source, google search results are now overwhelmed with covid19 news)
That may point to an upper bound of 30% on the people dying from this that would have died "within months/a year anyway"
There are some bayesian tripwires in that analysis, though I think the data is good and it fits roughly with my intuition that bulk of covid deaths are in the "basically healthy and will live for many years" population.
The fact that a large-ish fraction of the entrants to the care homes die soon means that the people who don't die soon are going to be over-represented in the population. So if you pick a bunch of people from these homes at random, you probably (depending on the total population and total number of such deaths, which isn't cited) don't find people at the edge of death.
So I buy the upper bound, but think the actual number is likely to be much lower.
Stated more intuitively: there just aren't that many people "within months of death" at any time, and there are a lot of people dying to covid.
If you look at the median age of covid fatalities in Italy at age 81.5, this question is entirely valid. Some of the numbers I read pointed to a 16-20% all cause mortality rate for that group irrespective of covid. So by definition the circles overlap to a meaningful extent.
The vast majority of deaths in Canada (for example since I live here) are those who are 80+ and around half lived in nursing homes. Given average life expectancy in our country is only 82, the vast majority of coronavirus victims' probability of death any given year was quite high.
There's a metric called something like "estimated years of life lost" based on mortality rate by age and life expectancy, but you'd also want to factor in individual life expectancy—a lot of the people dying from this already had other health conditions.
Except your hypothetical "never leave house" population is probably <5% of the population - every single person I know is still going out to public places at some time during the week.
This study doesn't appear to suffer from the same methodology problems as the Stanford study of a few days ago. In that study, they recruited people through Facebook, and reported an infection rate that was low enough that it could have been caused by false positives.
Here, it's closer to a random sample, but more importantly it shows really high rates. Those rates overwhelm any error due to false positives.
Also, it shows numbers that are in line with our intuition. It shows higher infections in NYC, and higher infections among blacks. That reflects what the hospitals are seeing.
Sure, but a person who goes to the store 7 times a week is ~7 times as likely to have been sampled compared to someone who goes only once a week. They are also ~7 times as likely to have contracted covid-19 at the grocery store. It's likely that they are also, in general, isolating less than the other.
This sampling would have a (possibly slight, but still unknown) bias towards people who are isolating less (who you would expect to then have a higher than average percent positive for antibodies, assuming isolating is helpful).
I don't think anyone knows whether or not this is a more representative sample than the stanford study.
Even with worst-case scenario of bias, this is still extremely good news. My worry was always with super spreaders who refuse to isolate but the data is suggesting that those spreaders will still soon achieve herd immunity among themselves. As long as the rest of the society behaves we will still hit zero cases relatively quickly.
(which wasn't always clear to me before since I initially predicted that this will take years to work out)
I don't think you are looking at this right. It will still take about a year (+/- a few months) for this to run through the population of the country. NYC spreads faster and was hit harder and earlier. Still only about 1/5 have had it. The remaining 4/5 (maybe 3/5 if there is some fraction of people that just are naturally immune/resistant) are going to get it. It's going to be a while to get there.
All of these numbers are inline with what experts have been saying and modeling for more than a month.
We will be dealing with this for at least another year. We will cycle through policy to loosen up and close back down a little. In the best case for NYC they are 1/4th of the way to herd immunity in numbers. With the drop in transmissions, this might be 1/8th of the way in terms of time. There will be more deaths in the future than have been recorded so far.
Look for the bright side in things, but zero cases is a pipe dream.
Herd immunity doesn't mean 100% exposure. It means a high enough incidence of antibodies such that the effective R0 goes lower than one, meaning that new outbreaks tend to shrink over time and not grow.
With most endemic viruses, antibody incidence is somewhere around 30-50% I believe, but I haven't seen any modelling for what covid is expected to do specifically.
If there are a high number of asymptomatic/mild cases that are to infectious to others this means it will have to be more.
i've seen numbers between 70% and 80%. Thats why i took 1/4th to mean 80%
Bear in mind that the CDC estimate of 45 million influenza cases[1] is the number of symptomatic cases, and therefore it doesn't really make sense to directly compare that with Covid-19 IFR rates calculated from antibody studies which include both symptomatic and asymptomatic cases.
It's not 25%. 0.5% versus 0.13% is not the only issue here in terms of how much worse it is. It's the long time in the ICU. The flu kills you fairly quickly or you get better fairly quickly, so you don't take up hospital capacity so long. Herd "immunity" does not require 100%, but that's a decent approximation. Sure, I'll grant that it "starts" to show an effect around 65%, but the effect is not so strong. 70%, much stronger. 80% very strong. Heck, you could probably do containment by then without waiting to get to 100%. Because inadvertent spreading would be so low.
The non-spreaders are the people who are self isolating.
I think they're saying that people who go out and spread the disease will quickly catch it, recover, and be unable to catch it and spread it. That relatively small group of people who are refusing to self isolate, will gain herd immunity, causing the virus to die out in that group of people, preventing them from giving it to the non-spreaders that have been at home the whole time.
Whether or not it will work out that way, I don't know.
While this dynamic may exist in some form, I don’t think it’s a powerful enough affect to stop the spread. There is not a firm dichotomy of spreaders vs. isolators and the composition probably changes over time, such that the virus still has many opportunities to spread to previously-isolated groups.
otoh, it gives about the same ifr as the recent study in geneva. at the end of the day, it looks like the first wave will have an ifr of about 0.6% - more than ten times greater than the flu, but not the bubonic plague
In the other direction, people who are sick or have recently been sick may well isolate and avoid going to the grocery store, dragging the detection rate back down.
Or truly random sampling that includes appropriate proportions of sub groups, e.g, socioeconomic status etc. Otherwise you have to know on all of the doors. But it's still a lot: for a 95% confidence level and 2% margin of error, for NYC you would need to test about 2,400 people. Assuming a response rate of 10%, you need to knock on 24,000 doors.
Wouldn't the ability to refuse taking the test introduce bias? i.e. the 10% that do take it still may not be representative of the entire population. Likely better than the grocery store sample, but still not ideal
Yeah, that is still an issue. You can offer compensation to incentivize participation. And you can pre-select your sample (instead of random door-to-door) and require multiple follow ups with the selected persons to reduce nonresponse bias.
You can also do a separate phone survey in addition to surveying the test participants. Questions like "Do you think you have had COVID?" and "How many times per week do you leave the house?". If the responses for the test participants vary significantly from the phone survey participants, you can try and weight your data accordingly.
Sure, phone surveys for political purposes (presidential approval ratings etc.) have to deal with that all of the time. There are methods for estimating non-response impact. [0] One method of mitigating it that I've seen it to reach out again to non-responders. You then analyze their results to see how they differ from the baseline responders to estimate the non-responder population. If there's little/no difference, you can be fairly confident the risk of bias is low. It's called non-response follow up, and is a pretty common method.
There's also literature that suggests that you don't discard outlier values in the actual responders as they may help approximate the non-responder population, i.e., the non-outliers represent typical responders while outliers are more likely to represent non-responders [1]
Wow, downvoted for providing factual information on how researchers mitigate non-response bias. Didn't think there was anything controversial there. I'm willing to accept the additional downvoted for this comment though.
Not downvoted by me! Appreciate you sharing this information, it is helping me better understand how researchers mitigate non-response bias. Interesting stuff.
Even with door to door knocking, you're going to introduce a bias -- toward people who are around to answer the door, which would tend to undersample essential workers.
Well, sort of. The increase in likelihood of being sampled due to going to the store N times more than the average person is a function of N and P (the probability of being sampled in any given trip) such that
f(N, P): (1 - ((1-P)^N)) / P
For a large P you're right, but as P gets very small (let's say that in any given trip you have a 0.01% chance of being sampled) then the increase in likelihood of being sampled approaches N times as much (in this example, it would be 6.999 times as likely). Of course, this assumes they're taking measures to ensure no one gets sampled twice.
I couldn't find any detailed write up about their selection methods for participants, only the basics of where they found them. Without that, it is very hard to determine whether or not there's some flavor of selection bias: If it is voluntary, and they do not screen out people who report having had cold/flu symptoms, then they run the risk of attracting a disproportionate number of people who volunteer because they're curious if their prior symptoms were actually coronavirus. That would make it far from a random sample. The fact that they sampled only people actually leaving their house is also a form of selection bias: these are the people more likely to be exposed and may represent a disproportionately high infection rate as well.
So I await further information.
That said, even if it's overall 0.6%, that is still 6x higher than flu, and higher than H1N1 which had a CFR around 0.1 for areas with adequate health care. (much higher when there was not adequate care, but that is not dissimilar to Covid.)
So no matter what, no one should be walking away from this study saying "See it's no big deal! Just the Flu/Cold etc!"
Flu is generally well under 0.1% of those infected on average. If you’re comparing them you want to either include or exclude asymptotic people from both populations. “Symptomatic Illnesses” https://www.cdc.gov/flu/about/burden/index.html
As to NYC deaths, many people currently infected will die in the future. You can make various estimates to account for this but a reasonable first approximation is to double current deaths based on NYC’s infection curve vs South Korea’s.
Flu rates are based on all known knowledge, population testing, etc. We don't have that for covid-19. If you want a good apple-to-apples comparison of lethality of another pandemic, you need to find CFR numbers that were available during the pandemic. That is what I provided in my post. Here is the source I got them from [0] which gives the CFR at the 10-week mark for H1N1, somewhat similar to where we are now. Here's the relevant quote:
>"The overall case fatality rate as of 16 July 2009 (10 weeks after the first international alert) with pandemic H1N1 influenza varied from 0.1% to 5.1% depending on the country."
There are near instant tests for flue in most every urgent care center I've been to (easier and faster than getting a dr appt.)
Whenever I have gone or brought my kids when sick, they perform the test to rule out any potential need for antibiotics. When I asked, they said it's systematically reported to the the CDC as part of their flu surveillance system, whether positive or negative. That surveillance network is significantly more robust than a survey.
0.6% this is the estimated ifr. comparing it to the cfr of flu is not right, which is about ten times less than the flu ifr - or about 60 times less than this figure. your conclusion is upheld, but even moreso
According to the CDC, the estimated prevalence of symptomatic flu last year was 45 million illnesses and 61,000 deaths.
For NY, assuming it’s evenly distributed, that would translate into 13.8% of 20 million = 2.8m symptomatic flu cases and 3,753 deaths.
COVID appears to be significantly more prevalent (roughly twice as prevalent in just 3 months as flu gets all year despite incredible efforts), and roughly 2-3x as fatal.
I think you need to rethink essentially all of your assumptions/math. COVID is not more prevalent than the flu and certainly more than slightly more fatal.
20% infection rates already puts SARS-CoV-2 at higher prevalence than the flu. And we're barely a couple months into the time period of significant spread.
Why would you think that SARS-CoV-2 would not be more prevalent than the flu? The fact that we have no natural immunity, combined with how virulent it is, indicates the endpoint infection rate will be significantly higher than an average flu season.
"In the U.S., for example, in recent years about 8.3% of the total population get sick from flu each season, a CDC study found; including people who carry the flu virus but show no symptoms, that estimate ranges to up to 20%."
The endpoint percentage of fatalities is a multiple of the infection rate times the fatality rate. Both numbers are equally important.
We've been hearing the "10% hospitalization" and low-single digit CFR numbers for months. At 20% prevalence, the NYC actually had 1.7 million cases, 36,723 hospitalizations, for a hospitalization rate of ~2.1%.
According to the CDC web site, those numbers are 35.5M flu cases and 34,200 deaths (https://www.cdc.gov/flu/about/burden/2018-2019.html), giving an IFR of more like 0.1%. Which would put COVID-19 at more like 5-6x more fatal than the flu. Combine that with the virtually complete lack of natural and vaccinated resistance compared to the flu, and it's clear that the potential fatality totals for COVID-19, absent drastic action, would be much, much higher than for the seasonal flu.
I agree 100% it will be significantly more prevalent, due to the nature of the virus combined with the total lack of natural or vaccinated immunity.
However, it's not clear at all what our "drastic action" is achieving, aside from 20 million unemployed and trillions of dollars spent trying to hold everything together. 20% prevalence doesn't exactly speak volumes towards the efficacy of social distancing.
That 45 million in your first link is Symptomatic Illnesses which represent ~2/3 of total cases as Approximately 33% of people with influenza are asymptomatic. https://en.wikipedia.org/wiki/Influenza So, total infections would be closer to 67.5 million.
But, it’s important to keep things in context, another year lists 9.3 million symptomatic infections and 12,000 deaths. The average year is well below peak years.
"... it's not clear at all what our "drastic action" is achieving..."
It's no coincidence that the city that shows a much higher prevalence also experienced a much more acute infection, straining their medical infrastructure. That isn't evidence that social distancing doesn't work, it's evidence that the infection was already well on its way before social distancing was implemented there. NYC was the canary in the coal mine that triggered social distancing across the US, and as a result the experience in NYC has not been repeated elsewhere. Aside from the results, which should speak for themselves, the epidemiological connection between R0 and the prevailing rate of social interaction is clear, which is why it's accounted for in the models.
I actually agree with everything you said, but I still haven’t found the basis for the tremendously costly actions taken.
As we can see, NYC did not particularly flatten their curve, and they did not see excess deaths due to lack of available care. They already hit the endpoint prevalence that you would see with the flu in a very bad year.
It’s not clear how much further SARS-CoV-2 would go naturally — the S curve doesn’t ever go to 100%. But more importantly, without a vaccine, which we won’t see for 12-18 months, we are going to find out exactly how high the S curve goes.
We have a small degree of control over exactly when that happens by burning extremely large amounts of money. So, when do you want it to happen? Now, while flu is at its nadir, or after spending $10 trillion dollars perhaps we can see the final surge during the peak of the next flu season?
The fact is that there isn’t actually a known treatment for COVID, and even with treating the symptoms (invasive ventilation) 80% of people in the ICU will die. This isn’t a disease for which “flattening the curve” makes particular sense. That would require there actually be an effective treatment.
NYC did not particularly flatten their curve, and they did not see excess deaths due to lack of available care.
I'm pretty sure all the family members of the people who died would regard their loved ones as "excess deaths". Your technocratic language ascribes no value to the lives lost, in which case of course the actions seem excessive. For people who don't value the lives of others, it would be hard to understand all the fuss.
This isn’t a disease for which “flattening the curve” makes particular sense. That would require there actually be an effective treatment.
This is a self-contradictory statement. The whole point of flattening the curve is to buy time to develop exactly those effective treatments. It is also possible to flatten the curve to an extent that there are fewer deaths under it, and based on the continual downward revision of the IHME model fatality estimates, that appears to be exactly what is happening.
So the most charitable interpretation of my comment is to convince yourself I’m a sociopath?
Excess deaths is a term of art to describe a number of deaths over a baseline. It can be used in general, or specifically, as I was using it, to describe the repercussions of an overloaded health care system. Here, for example, is a Sky News story from today using the term in its headline. [1]
I think you owe me an apology, but I won’t hold my breathe.
“Flattening the curve” is absolutely not intended to buy time to develop a cure or vaccine. The timeline over which the economic devastation of flattening the curve can be sustained is measured in weeks not 12-18 months.
The purpose of flattening the curve is to be sure there are the necessary number of hospital beds and ICU beds to provide effective treatment. If effective treatment can not be provided it leads to “excess deaths”.
Bad projections based on a 10% hospitalization rate and based on assumptions that ICU ventilators wouldn’t kill 80% of patients, have led to disastrously bad public policy which is causing untold suffering throughout the country. This suffering, which many people seem to be blind to, is notably self-imposed rather than natural, and through logical and reasonable interpretation of factual data we can stop this suffering.
This is very much unlike the suffering caused by people who are dying from COVID, for which unfortunately we have no effective cure or treatment, and for which it is not reasonable to assume we will develop one in the timeline of this current pandemic.
This is the key takeaway from “flattening the curve”. It cannot and will not decrease the total number of people who are ultimately exposed at this stage of the pandemic. Once the health system capacity is high enough to handle the number of cases coming in with effective care, continuing social distancing causes extraordinary damage without any benefit.
Since effective care at this point has proven to be both simple and extremely scalable (non-invasive ventilation, antibiotics, and frequent repositioning) the resource curves for effective care are extremely larger than projected, back when they were based on availability of ventilators.
Simply put, the data has changed, and it’s time to update the policy. It’s extremely incompassionate to do otherwise.
However you choose to define the term "excess deaths", the fact remains that there are tens of thousands of people alive today who would be dead if we all took your advice and ignored the notion of social distancing. Does that make you a sociopath? No, it simply makes you wrong, policy-wise, if you value human life. Unless you really are a sociopath, and advocate the policy precisely because you don't value those lives. I can either give you the benefit of the doubt morally, or intellectually. Which do you prefer?
You are correct that social distancing and flattening the curve protects the health care system from overloading. That doesn't mean that it is the only purpose, or effect. If we took a poll among health care policymakers about the benefits of flattening the curve, I'd be happy to bet any amount you'd like to wager that "buying time to develop effective treatments or a vaccine" would be on their list.
This is the key takeaway from “flattening the curve”. It cannot and will not decrease the total number of people who are ultimately exposed at this stage of the pandemic.
This is just mathematically untrue. There is clearly a point where the curve is so flattened that the number of deaths under it is less than a baseline "no action" scenario. If we could somehow get everyone in the US to truly isolate for three weeks, the virus would die out entirely. So clearly there is a spectrum of actions and their corresponding outcomes. The number of infected (and dead) is not a fixed number with only the duration of the outbreak changing.
If your premise were simply that the current measures are unsustainable, I couldn't agree more. Fortunately, those measures have bought us the time to take more focused, informed action based on the latest data. In a crisis this fast-moving, every day is time to update the policy.
Those are the people who constitute the denominator for the CFR. It's a completely different metric. You would never try to impute population exposure from such cases. This study is trying to to impute population exposure. There isn't any comparison between the two in terms of which has more/less selection bias, they are completely different things.
That's because all of the statistical reporting is whacked up across the globe. The only thing that can reasonably be measured now is CFR. Without extensive, reliable serological testing with high specificity, IFR can only be guessed at.
Is there a write-up for this study? Or anywhere with data/methodology released?
Also, is this particularly good news? Using the raw numbers from the headline and the deaths from here https://www1.nyc.gov/site/doh/covid/covid-19-data.page we can make coarse fatality rate estimates. I assume population of NYC is 8,400,000.
10290 / (.212 * 8.4e6) = 0.6%
Including probable deaths (which makes the death count line up more closely with excess deaths)
15411 / (.212 * 8.4e6) = 0.9%.
So it seems consistent with a IFR of 0.5%-1%. However, this doesn't account for the lag between antibody presence and death. Also, we don't know what bias sampling from grocery stores introduces, it could affect the results in either direction. Hopefully, New York releases a paper so we can get more accurate estimates.
It's good to see a prevalence study with presumably less flaws.
Yes, the study is really good news for NYC. My thinking is that the state of emergency, currently in place until 5/15, will be extended at most once to 6/15. After that we'll be somewhat open for business. I imagine everyone will be advised to wear mask/gloves, to keep the social distance and to keep washing hands. Lots of people will continue to work from home, others will bike to work. Being summer, the virus will not be that contagious. When the November rebound is forecast to happen, the herd immunity will be much higher, hospitals better prepared, the medical world will have learned the most effective ways to treat this disease, so the fatality rate will go down to levels seen in Germany, or lower. Bottom line, I don't think we'll have another shelter-in-place in November-December.
That said, science is enamored with significance levels, p-values, etc. Most people just do that because that's what they've learned, and think that's how it should be done. They don't think it comes from Bayesian inference, where you put a very low prior probability of a drug being useful (because the very vast majority of chemicals we could put in our bodies are either harmful, or have harmful side-effects). In this case, the Bayesian prior is that common colds in general become less prevalent during the Summer, and Covid19 is caused by a coronavirus, which is related to viruses causing the common cold. The burden of proof should be lowered many, many times, but I nobody who did these statistical surveys did that, because this is a total no-no in the field.
So, I'll state my conclusion, after reading your link: the infectiousness of Covid19 is more likely than not to decrease in the Summer, based on the studies that were performed. It will not decrease to zero, so that does not contradict your statement that it is transmissible in hot countries.
> the Bayesian prior is that common colds in general become less prevalent during the Summer, and Covid19 is caused by a coronavirus, which is related to viruses causing the common cold
MERS is also a coronavirus and does just fine in the warm climate of the middle east. I hope that warmer temps slow the spread, but it's far from certain.
It's not just far from certain. It is simply unlikely. The "hypotheses" for explaining seasonal variations in transmissibility are nothing but a wild-guess/hope.
Kinshasa, the capital of DR Congo, is a city with 11 million people. About the same size as Wuhan or New York City. According to [1] there were only 25 deaths due to Covid19 in the whole country. Why so few? 21 million people live in Lagos, Nigeria, yet the whole of Nigeria has seen only 31 deaths. On the other hand, I have to admit, there were quite a bit more deaths in Sao Paolo, Brazil, but it's still an order of magnitude less than in the New York State.
Maybe Africa and Brazil are faking the numbers, but Australia most likely not. Only 76 deaths so far. Argentina has recorded only 159 deaths.
You could say this is only circumstantial evidence. My point is that we are not trying to prove things "beyond any reasonable doubt". We are trying to form the most informed opinion. And there are quite a few factors that point in the direction of reduced infectiousness during the Summer. Schools closed? Check. People away for vacations? Check. More UV light which kills airborne viruses? Check. People having more vitamin D in their bodies? Check. Air more humid, so the tissue inside your nose is better protected? Sure thing. Better immune systems due to less stress due to the ambient light? Sure thing. Etc, etc.
You can question any of these things. But in a Bayesian framework they tilt the scales little by little. But that's just me.
Jakarta and Istanbul have a huge rise in deaths, similar to European countries, but they have suspiciously low recorded numbers of deaths officially due to Covid. http://archive.vn/vOoRp
It looks to me like you are cherry picking data, and wrapping it up to appear objective. Looking at “case” numbers from countries with poor heath systems is especially misleading. Australia could be an outlier. Jakarta and Istanbul are better proof that temperature is not that important.
My prior is that I have seen lots of people use the “temperature” argument because they want to believe it, which trumps your prior, wink.
Edit: Note that science is mostly about finding counter-examples to a hypothesis. Jakarta is a good counter-examples to the hypothesis “Covid is not much of a problem in hot temperatures”. Istanbul was not (it is actually tepid there in March). And this brings up some other counter-examples: https://thehill.com/homenews/coronavirus-report/494428-trump...
Which infectiousness rate are you talking about? How could your prediction account for the impact of increased immunity and other interventions?
My point is that the infectiousness will decrease in the summer because of these factors. There isn't much reason to think there is anything special about it being Summer.
Finally, it's a very tenuous link between common cold prevalence and coronavirus.
It will increase, because vastly more people will have already been exposed. The seasonal variation in the cold exists for many reasons, including SCHOOL is open, and people tend to congregate in enclosed spaces more. None of these things are relevant to COVID
It's really difficult to compare the currently reported fatality rates between different countries and regions since the amount of testing varies so greatly.
Compared to many other countries, Germany has likely confirmed through testing a higher proportion of its total number of cases in the country, lowering the reported fatality rate.
In countries where testing is less widely available only those already showing severe symptoms get tested disproportionately, so confirmed cases are more likely to be fatal even if they get good hospital care.
I agree with everything you are saying. This does look like good news.
However, we should remember that this is probably a mild overestimate, as the study population was assembled from those who were out and about (shopping at the grocery store, etc). People who are more strictly staying home -- and thus less likely to have been infected -- wouldn't be included in the survey.
Another way to look at it is that 20% of people most likely to contribute to a high transmission rate already have had it, in turn flattening the curve by a considerable amount
This one's from a different company (BioMedomics) and it was a random test of people pulled straight off the streets. Happened at an entirely different geographical area (Chelsea, Massachusetts) as well. I don't want to be too optimistic but there are some signs that we are heavily under-counting the actual number of cases (at least in the US).
What I learned from an article of a major German newspaper is that in order to really estimate the quality of testing, two criteria are important: sensitivity and specifity of the applied test.
If the real infection rate is still low throughout the population, a random sampling will not lead to reliable results (you could also toss a coin).
Although the article is in German, it is worth to have a look on the graphic in the middle of the page, it should be understandable.
> Here, it's closer to a random sample,
but more importantly it shows really
high rates. Those rates overwhelm any error
due to false positives.
I'm not seeing that, at least not from what I've seen. Whether false positives skew results significantly is highly dependent on how accurate the antibody test they used is (in addition to how large a subset of the population is positive).
This guy has an interesting visual tool that helps you to see how much a study could be affected. [1] Also, he says "Here's an interesting relationship. When a test with 95% sensitivity and 95% specificity is applied to a population with <5% prevalence of disease, MOST of the patients with positive tests are FALSE POSITIVEs." I.e., positive rate shows up as greater than 10%, when it's actually less than 5%.
Do we know the sensitivity and specificity accuracies for the antibody tests used in NY?
You'd like to think they know what they're doing. But this Bayesian stuff can be tricky, especially if they're rushing something through (esp. regarding testing of accuracy of the antibody tests themselves). And the California studies, although they seem to have some competent people behind them, seem to inflate/exaggerate the lower bound of uncertainty in their projection. [2]
Sorry, can someone spell out how this might be good news? 21% is still a long way from herd immunity, and NYC's hospital system has been severely strained getting to this point. On top of that, the hospitalization rate still seems disproportionate to someone getting the flu - maybe it's five times as fatal, but it's > 5x the flu hospitalization rate. It's not like the virus has become less dangerous, we're just realizing how dangerous it has been, what with the impact we've already experienced.
Besides, generally if a virus is less fatal than previously expected, it means it's more contagious, meaning that much harder to get to herd immunity.
Good news would be things like: evidence the virus has mutated into something less severe; evidence of an anti-viral treatment that improves outcomes for everyone so it's not as big a deal to catch it; evidence that community spread has halted in an area and the boundaries are controlled so people in that area can feel secure they won't catch it; evidence of an impending vaccine.
Is this good news just because we're finally establishing that people have caught it once can't re-catch it for now? I guess I can see that as good news but that is so expected that it's more like it would be horrendously bad news if we found evidence that recovered people didn't have antibodies. But generally I don't really see what policy impact this has, other than identifying a pool of workers that can go work in meat-packing factories without fear of catching it again.
The only good news that I am getting out of this, is that NYC seems to have avoided the dreaded 3-5% fatality rate that was the presumed worse case scenario (massive community spread, overloaded healthcare system). But the lockdown likely helped a lot, and it seems like they are seeing around 1% IFR, so I am not sure this really changes anything. Our understanding of the disease and how to treat it may also be getting better, but it still seems like if we let this thing run wild through the population (as some people on here are proposing) local health care systems will collapse and we will have a IFR orders of magnitude higher than the flu.
And as you pointed out, if the study came back with an infection rate of 50%, I am not sure I would consider that good news either. That means it would be nearly impossible to isolate vulnerable populations. So while a 50% infection rate would mean the IFR is lower, it also means opening things back up and only isolating the vulnerable would not work to protect them.
If the infection rate were 50% we would be close to herd immunity (which I've read would require about 70% for this virus), so that would be better news in a sense.
Unfortunately 21% is a long way from 70%, and it's taken a massive amount of death to get to that point.
The R0 value impacts herd immunity %. So if NYC is already at 50%, then it means R0 is much higher than we thought, which means herd immunity would probably be as high as +90%. With numbers like that, not shutting down would result in the entire population getting infected in the span of a month or two.
The point I was trying to make, was that for the 'open back up' crowd, they are arguing that the IFR is similar to the flu, and only vulnerable populations are really impacted. So they say we should open up and just keep vulnerable population in lockdown. But they are ignoring the implications of the R0 value in their argument. i.e. if the IFR is really as low as they think (and consequently, the infected population is as high as they think), then nothing short of a total lockdown (or very aggressive testing and contract tracing) would stop vulnerable populations from getting infected.
The herd gets substantial benefits long before you get to 70% or whatever rate for full herd immunity.
The spread starts to slow before that point. If you're walking around infected and 20% of the people you come in contact with cannot catch it from you, 20% less people are going to catch it, no matter how much you cough on them.
> Sorry, can someone spell out how this might be good news?
It's good news because it strongly suggests that mortality is much less than previously suspected. There were numbers floating around from anywhere between 10% to 3% a few weeks back. A mortality rate < 1% is very good news because it means fewer people will die in the long run.
> mortality is much less than previously suspected
Mortality remains pretty much as suspected already two months ago:
“Based on these available analyses, current IFR estimates10,11,12 range from 0.3% to 1%. Without population-based serologic studies, it is not yet possible to know what proportion of the population has been infected with COVID-19.”
The princess cruise ship study also gave an IFR (for China) of 0.5%, and an early epidemiological modeling study put the symptomatic CFR at 1.4% which would imply 0.7% IFR assuming it's 50% symptomatic.
Yes, and then almost the entire US press spun the 3.4% figure as the real WHO-confirmed fatality rate and sub-1% numbers as a Trumpian lie as part of a stupid, cynical, partisan attempt to get Trump. There's been a lot of that. (The UK press, meanwhile, happily quoted the 1% figure - if I remember rightly, some outlets like the Guardian with both UK-facing and US-facing sides pushed both narratives to different audiences at the same time.)
In reality, the WHO number was just confirmed deaths divided by confirmed cases, which was of course almost completely meaningless.
This comment is entirely wrong. The mortality rates are right in line with what was expected with an IFR of > .5% unlike the Stanford study which was claiming something much lower.
Mid single digit percentage mortality rates were the numbers for case fatality rate, not infection fatality rates. Infection fatality rates have consistently been around 1%.
I don't know if the hospitalisation rate is so out of whack vs the flu. It should be higher. But does 20% of the population catches the flu in a matter of weeks?
Every person who is infected that we don't know about lowers the estimated hospitalization and mortality rates and reduces the ability of the virus to spread (for as long as immunity lasts in the individual).
And the specificity of the test might be worse than advertised but it wouldn't be credible that it could be making a big different with the 20% positive rate here.
Why does recruiting through Facebook invalidate the results? If it is a representative sample it is a representative sample regardless of how people were recruited.
> Why does recruiting through Facebook invalidate the results? If
Because “people who both use Facebook and don't automatically discount every ad or other solicitation on Facebook not from someone they personally know, especially if it invokes a major news story, because of the risk of it being a scam looking to steal personal information or do something similarly nefarious” are not representative of “people”.
If a representative portion of the population didn't click on Facebook ads then Facebook wouldn't generate revenue from these ads. Testing doesn't rely on the person being intelligent either. Because it sounds like you're just saying "results are bad because only idiots would click on a Facebook ad".
> If a representative portion of the population didn't click on Facebook ads then Facebook wouldn't generate revenue from these ads.
No, if valuable advertising demographics didn't do that, Facebook wouldn't make money. Valuable advertising demographics and representative samples of the general population are very different things.
> If a representative portion of the population didn't click on Facebook ads then Facebook wouldn't generate revenue from these ads.
That doesn't make any sense. Ads can make plenty of money even if they're only seen by women, or people between the ages of 30 and 70, or people in zip codes divisible by 3.
One critique was that the segment of the population who would respond to the ad and actually test may be more likely to have experienced COVID-like symptoms.
Also people reported sharing the Santa Clara link with others who might want the test (due to having had symptoms). In principle you could have the same problem with this survey: call your buddy and say “come on down to Costco—they’re doing free antibody tests”.
The obvious fix is to not tell people the results of their own tests. Not sure of the ethics/consequences of that approach.
The rates being this high actually casts doubt on the study. It’s very very implausible, and suggests almost surely a selection bias in the sample towards a population much more likely to have contracted it.
Actually in reading closer and seeing the data is collected at big box & grocery stores, it’s almost surely very biased.
This cohort would skew younger and wealthier, which correlates with better preexisting health and fewer risk factors, and would exclude populations who systematically left the city, or who have known heightened risk factors.
Comparing the death rate overall with the infection rate of this skewed sample would be likely to greatly underestimate the actual death rate of the virus.
This weekend San Francisco plans to do a hard test of about 6000 people in a couple of square blocks of the Mission district. They are going door to door to encourage people to get tested.
Still biased but probably better than anything so far.
I wonder about that. They are trying to test a neighborhood they think is most effected. One wonders if it's possible to follow up and retest positives to exclude false one.
Suspect the Bay Area is moving towards what Korea is doing. Intense testing and contact tracing.
Friend who works at a biomedical corp says scuttlebutt is they're separating their workforce into groups based on critical need and how well they can work remotely. Workers like him not expected to return to work onsite until next year.
Retesting positives won't work because we don't know whether it's independent. If there's some rare common cold coronavirus antibodies that some of the population has, that set off the test, they'll still have them later.
Not sure how well having the Bay Area do one thing, and large chunks of the rest of the country dash to herd immunity is going to work.
A little, but not anything close to what we consider normalcy. And you get to maintain them forever, or until a vaccine arrives.
Of course, our case count is so high right now. If you get Rt down to a miraculous 0.5, it's like 180 days before you can get to the point where contact tracing is effective. And you probably don't achieve an Rt of 0.5.
What's weird is that the Stanford study was suggested to have exaggerated the results, since those who suspected themselves as having the virus at one time were more likely to volunteer (allegedly). However, those results were nowhere near as big as this (I think like 2%-3% versus the 20%!)
Roughly 1/1000 New Yorkers have died of coronavirus. At a 1% IFR, that suggests at least 10% of the state has gotten the virus. Probably more, because deaths are undercounted and many people who will die of the virus already have it but haven't died yet.
In the Bay Area, almost no one has died of coronavirus so the infection rate should be next to nil.
that's crazy to me, what are the demographics of NYC? Is the population more susceptible to dying of Covid19 somehow? I still can't get my head around why NY has suffered so disproportionately.
There are many close-knit communities in NYC and the surrounding area where people live much more densely (i.e. large family in a single house) and congregate more often with one another than do West Coast types. Jewish communities in New York seem to be especially hit by this; not sure if this is because they did not follow social distancing guidelines, or if they're not possible to follow given unavoidable physical constraints.
Yes, if you look at Israel and Montreal, Ultra-Orthodox communities were hit very hard. Based on infections numbers, not deaths(since it's mostly very young population).
Lack of communication and distrust of local authorities (only praying to G-d will help...)
If the NY subway is anything like London, people will be less than 6 inches apart at rush hour. I imagine that will have played a big part in the spread in both of those cities.
Yes, the subway is the correct answer. That's how this virus spread like wildfire in NYC. Second reason could be the elevators in the tall buildings (there are plenty of those).
A counter argument is that the same didn't happen (or hasn't happened yet) in cities that also have massive public transportation systems like Tokyo or Seoul. There is probably many confounding factors, such as mask use, no talking on the subway, etc, to make pointing to one particular factor very hard.
Those cities see more widespread use of masks generally. And specifically once COVID-19 was common knowledge, so was the use of masks.
Widespread mask usage doesn't happen in NYC, same as the rest of the U.S. It's very recent this is practiced in the U.S., whether voluntarily or by order of a handful of local governments. And in my local area where it's not mandatory I only see about 3 in 4 using masks.
Yep, I agree with you. I was trying to make the point that massive transportation is not the only factor at play. Like you said, there are many others that should be considered.
I don't think that fully explains it. There are plenty of European cities of similar density to NYC that haven't been hit nearly as hard. Naples Italy, for example.
The most likely answer is that NYC was just a good place for it to spread which had plenty of travelers to get things started fast, but that that the death rates aren't particularly high -- everywhere that has fewer deaths has just had less of their population infected ... so far.
Germany has about as many cases in total as NY, yet less than 1/2 of the deaths. The death rate seems high in NY, maybe because hospitals could not cope?
Germany has about as many diagnosed coronavirus cases; the difference could be wholly explained by Germany testing more people. (Germany has administered 2 million tests, New York State [I didn't see a figure for the city], about 700 thousand.)
In my opinion it is more likely that the main cause is that the infection got here very early, and that the mitigation and containment measures started very late in comparison with cities of similar size and density (Seoul for example).
What's interesting is if you look at the rest of NY numbers (-NYC,LI,Westchester). That has infection at 3.6%. Santa Clara was estimated to be 3% and LA county 4%. So it's in strong agreement with those areas.
There also appears to be a difference in the strains. The NYC strain seems to mostly come from the European branch, while the CA infection comes directly from China (and likely much earlier in 2020). There appears to be some difference in virulence between the strains.
NYC and standord are markably different scenarios. But in each case the percentage of baseline infection to deaths is similarish, even though the percent compared to total pop is quite different.
Exaggerated in that the California study predicted a flu like IFR (Ioannidis, one of the authors, has been pushing that as his pet theory for over a month). This shows it is more likely several times that.
it doesn't contradic Ioannidis at all. He might be right in the end, and that would be an incredibly bitter pill to many.
We are getting closer and closer to the flu fatality rate.
from 5% to 3%, then 1% now 0.5% - smart money (and common sense) would bet that the rate will continue to drop,
I would expect that people living in healthier environment than NYC will fare even better. No way NYC IFR is the upper bound for the rest of the country. You could just as well expect it to be 10x higher than other places.
the flu rate is 0.1%, thus we already hit the order of magnitude.
That it's a lower bound for IFR in NYC. With how widely that changes based on age alone, it could also vary based on location. Such that NYC could conceivably have the highest value for that in the US.
A datum that is hard to square when the deaths are still dominated by the older population. If you had convinced everyone over seventy to move out of NYC last year, their CFR would be a fraction of what it is now.
No. I was not claiming it would be lower. Apologies if the framing said it that way. I was just pointing out that we really don't have bounds on this anywhere else.
I'm not sure how that matters. We have a lower bound for an optimistic demographic representation of the country. If NYC is younger and healthier than average and is at 0.2%, why would the lower bound for the country as a whole be lower?
I am actually having a hard time squaring the claim that they are younger than the average. The number of people over 70 that have died in NYC is above the number of people over 40 in many cities across the nation.
That all said, my point is that we don't know the bounds. Period. It could be higher. It could be lower. That is why I said it is conceivable. Probably it is about that value in most places. I am interested in where the data falls.
And the deaths are concentrated on the older population. They do have younger people. They also have more nursing homes and assisted care. They literally have more of everything.
I’m talking about having more number of nursing homes because they have “more of everything”, not making an argument about density and pollution or anything like that.
Ah, yes, I misstook your argument. That said, I do suspect having more nursing homes means they have more people over 60 than most places. Which will skew them to have more deaths, period.
Consider, from all that we have seen, elementary schools could get 100% infection rate on the same population size that NYC nursing homes have and not see the number of deaths they have had. It is not controversial that the IFR is dependent on how many people over 60 you have in your population.
To drive that home some, here in WA, fully 92% of the deaths have been in people over the age of 60. It is quite ridiculous how deadly this is if you are older.
All I am saying people there breath the NYC air every single day. Can't possibly be good for them, especially in the light of some chronic pulmonary inflammation induced deaths.
The point on the lower bound of 0.2% is informative. I did not know that.
There are huge error bars on that ratio, because "the population of NYC" is not something easily defined, and the death count (at this time) includes a lot of "excess deaths" that almost certainly have nothing to do with the virus (e.g. untreated cardiac arrest).
That's some major revisionism. No credible source was ever suggesting a 5% IFR. For example the Imperial model was using 0.9% given UK's age distribution. That looks likely to be spot on.
And Ioannidis? He was quite certain that the CFR was going to be a little higher than 0.1%. Yes. CFR, not IFR. So he's off by two orders of magnitude.
At various points in interviews and articles he used Diamond Princess, South Korea, Germany, Iceland as strong evidence of miniscule fatality rates, in every single case selectively ignoring that deaths lag symptoms which was already well known at the time. All of them had their death rates double or more after he used them, and it was easily predictable based on recent exponential growth and death lag.
In his stat article he was saying it is conceivable if we didn't know about it we wouldn't have been able to even detect it in the death numbers after it ran its course (he has since walked that back).
Even the other day after his serology preprint he was saying it doesn't seem to have a higher chance of killing you than seasonal flu for each person infected: https://www.youtube.com/watch?v=cwPqmLoZA4s&t=1h9m50s
And he claims the WHO said 3.4% of people who get infected would die:
His data from the Diamond Princess is completely outdated. He cited 7 deaths. We are now up to 13 with 7 more on critical condition. He has been completely wrong in each of his predictions.
I would respect him more if he just argued from am economic perspective that the economic damage is greater but his wild hypotheses about Covid being comparable to the common cold or flu have been completely refuted by all data.
0.1% is the flu cfr. 0.6ish% seems to be the corona ifr. these are comparing chalk and cheese. how many people who get the flu never rock up to a doctor? the ifr for flu is closer to 0.01%.
the worst case scenarios are disproven it is true - but so is the idea that it's just a flu.
let us be grateful this trial run of a deadly global pandemic was only moderately bad.
.1% is flu IFR but if I remember right the number doesn't include true asymptomatics which are estimated at up to 75% (which could bring it down to 0.025%). I'm not sure on this, that was on a CDC page I saw.
first you say 0.1% is the CFR for flu then, in the same sentence you claim that it is probably closer to 0.01% because people don't go to doctors with the flu.
Are you arguing that after all this time we still don't know what the actual CFR for flu is? And that the reported CFR is a gross overestimation? - I find that hard to believe.
To me, this feels that once this disease hits the reported flu numbers people start arguing that oh wait, the flu is actually even less dangerous ...
Oh, wait. You don't understand that the CFR and IFR are not the same thing? That explains a lot. I thought you were just being disingenous when comparing the early CFR statistics to the current IFR estimates.
The CFR is, by definition, computed from known cases. It's thus trivial to determine exactly: just divide the confirmed deaths by confirmed cases. So we definitely know the CFR of flu. The problem is, of coures, that it's highly likely to be an over-estimate.
On the other hand, the IFR is hard to determine, since we don't know which cases we missed, nor whether the unresolved cases will end up living or dying. Which is why all we have is estimates.
I'm sorry but 0.1%-0.2% directly contradicts 0.6%-0.9% (or higher because that doesn't account for the lag between infection and death). That's a 3 to 9 times higher death rate.
The major methodology problem with the Stanford study was not that they recruited participants through Facebook.
The major problem in the Stanford study was that they ran a test that has a 3% false positive rate, and found that 3% of the test-takers tested positive. (And that apparently the asymptomatic COVID-19 rate is 90+% - which does not square with the Diamond princess data).
They could have ran that same study back in 2018, and would have gotten the exact same garbage results.
This study is good news... I'd be interested to project the rate of infections WITHOUT shutting down the economy for what's now going on 6 weeks. Seems like it was all in vain.
This still is not a random sampling. It only samples from shoppers at grocery stores and big-box retailers. Imagine doing the same study, but of people who ordered groceries online only. Would you expect to see big differences in exposure rate? I think so.
The study itself isn't linked anywhere, nor have I seen it elsewhere. Science is all about the details. It's not hard to imagine half a dozen ways that the bottom line result of this study could have been skewed by decisions made by the study authors and ground team.