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.
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.
This study may well be really good news.