I presume that this is result of too late reaction to the outbreak but perhaps somebody knows better.
I presume that this is result of too late reaction to the outbreak but perhaps somebody knows better.
Look at: recorded deaths, excess deaths, hospital admissions, ICU admissions, serological studies.
Together they can give a good picture of what's happening.
Please don't look at serological studies. Many of the serological studies have frighteningly high false-postive rates that can lead to faulty assumptions that 20% or 50% of people have been infected already. This pretty much isn't true anywhere.
That's not helpful.
Once you have at least 10 or 20% immunity sero studies can be used with some confidence, and there are now a couple of tests that claim 98% or better sens/spec, so using the earlier poor ones should be avoided now.
Even New York State as a whole is surely over 5% now.
Obviously if a poor test with 85% or 90% specificity is used and you have low prevalence then you get a poor result. But if you can peg the floor of the prevalence at several percent and use an accurate test then what’s making the results useless?
But even disregarding sero studies, you can draw some conclusions of immunity from death numbers. E.g if assuming a 0.5% IFR with 1600 dead in 2100000 you would have 15% immunity (assuming everyone infected is immune of course). IFRs so high that they yield under 5% immune (basically IFR over 1.5%) aren’t consistent with observation.
But they are. The IFR in places like China, South Korea, and NZ that have things mostly under control at this point, the CFR is consistently above 2% (and as high as 5%), and those places are and were doing significant viral testing, so wouldn't have missed 50% or 150% of the infected. So a CFR of 2% or 1.5% isn't unreasonable (unless you're relying on sero studies as "observation").
Unless you're assuming that there's a huge swath of people (like the significant majority) of Covid-19 cases that are entirely asymptomatic (and we don't have any reason to believe this, we have upper bounds on pre-symptomatic cases at like 50% from randomized viral presence studies, and the asymptomatic numbers should be notably lower), the IFR and CFR should be reasonably close, in places that have things under control and that makes 1.5% IFR more likely than .5%.
> In the case of Stockholm there was 2.5% active infection over a month before the sero study that showed 10%.
Source for this? Sweden has 26,322 confirmed cases today. You're saying that 24000 of those were confirmed in Stockholm well over a month ago?
The same high specificity sero test that estimated 10% in the Stockholm pop., found 20% with antibodies among hospital staff in one hospital (https://lakartidningen.se/aktuellt/nyheter/2020/04/en-av-fem...)
> Unless you're assuming that there's a huge swath of people (like the significant majority) of Covid-19 cases that are entirely asymptomatic
I'm not. 50% sounds reasonable. But in Sweden, only people who are ill enough to be admitted to hospital have been tested so far (especially around the time of the earlier studies - now it's expanding to more general testing) and other than those it has been mostly hospital staff. So the missing number is all asymptomatic AND all the cases that aren't severe enough to require emergency care. And that's a significant number.
I thought there was a broad consensus now that is centering in on IFR of 0.3% in many places while others seem to have higher numbers like 0.5% or 0.8%. I haven’t seen many reports that suggest over 1%. CFR doesn’t really make any sense if you only ever confirm a fraction of the cases.
> But in Sweden, only people who are ill enough to be admitted to hospital have been tested so far (especially around the time of the earlier studies - now it's expanding to more general testing) and other than those it has been mostly hospital staff. So the missing number is all asymptomatic AND all the cases that aren't severe enough to require emergency care. And that's a significant number.
Right, but I'm not talking about sweden, I'm talking about places that were doing more significant testing.
> The same high specificity zero test found 20% with antibodies among hospital staff in one hospital
This would be unsurprising. There have been studies showing that certain at risk groups have very high prevalence (meat packers, homeless, etc.). These can't really be generalized across the broader population.
We can infer that, for example, the hospitalization rate in NYC and San Francisco and Stockholm will be similar per-case, and per fatality, and draw conclusions from there. Those conclusions don't lead to 10 or 20% of the population infected anywhere. They might lead to 5% infected in the worst hit areas (NYC, Lombardy, etc.).
Edit: I take this back partially: Stockholm apparently could reasonably be hit as hard as NYC at the moment, so it could also be at a 5% infection rate. Sweden as a whole almost certainly isn't though.
> I thought there was a broad consensus now that is centering in on IFR of 0.3% in many places while others seem to have higher numbers like 0.5% or 0.8%. I haven’t seen many reports that suggest over 1%. CFR doesn’t really make any sense if you only ever confirm a fraction of the cases.
I haven't seen this. It's certainly possible, but low estimates I'd seen were .5%, centered closer to .8 or 1%.
https://www.nejm.org/doi/full/10.1056/NEJMc2009316
Verity et al. estimated IFR around 0.6% back in March, based on stuff like the Diamond Princess, again no antibody tests:
https://www.thelancet.com/journals/laninf/article/PIIS1473-3...
This is broadly consistent with the latest serology. NYC implies a bit worse if we divide excess deaths by antibody positives (though they haven't published their methodology as far as I know). Preprints from many European countries imply a bit better.
And ~0.2% of NYC is dead from coronavirus by now. So if less than 5% of the city has had it, you think IFR is >4% there? And you think NYC's announced serology data is ~15% false positives? Even though the total positive rate--which would include both true and false positives--in serology studies in most other cities (which haven't been hit nearly as hard as NYC) is far less than 15%? This makes no sense at all.
The simple explanation for Korea's high CFR is that for a disease where most countries have failed to trace ~100% of the cases, they're failing to trace about half. I think I've seen their public health authorities speculate exactly that, though I can't find the link now.
Inciting public panic is a different kind of harm from inciting public complacency, but it's harm nonetheless. I urge you to correct the misinformation you've posted above.
ETA: And really any single IFR is inherently misleading for this disease--the IFR climbs so steeply with age that it will vary greatly with population age structure, and with whether the disease disproportionately hits the most vulnerable (like when it spreads in nursing homes) or least (which should be our goal, since if we're unable to avoid herd immunity from recovered cases before a vaccine comes then that will cost the fewest deaths). But that's what was discussed above, so I stuck with it.
You're confusing NYC and NY State numbers. .1% of NYC is dead from Covid at the moment. Which leads to an IFR of 2%, an IFR which falls in line, if a bit above those expected in South Korea etc.
> Inciting public panic is a different kind of harm from inciting public complacency, but it's harm nonetheless. I urge you to correct the misinformation you've posted above.
I agree. Can you explain what about what I've said might incite a public panic? I believe all I've said is that everything I've said can be traced back to two basic assumptions:
- I think it's reasonable to believe that the IFR is among the high end of expert estimates
- Naive reading of many serology studies leads to the conclusion that something like 80% of Covid cases are completely asymptomatic. This is likely untrue and believing it leads to dangerous conclusions.
> Even though the total positive rate--which would include both true and false positives--in serology studies in most other cities (which haven't been hit nearly as hard as NYC) is far less than 15%?
Indeed, in most it is. And then there's the study that showed 60% positive serology in a city in France, and 30% positive serology in Massachusetts. So yeah a 10% or 12% false positive rate in a serology study is completely feasible, given that the French study probably had a 40% false positive rate or something equally ridiculous.
I don't believe a binary symptomatic/asymptomatic distinction is terribly useful. I agree that most cases that are tracked eventually report symptoms, just symptoms that are mild enough they wouldn't normally have sought medical care (in normal times, because they wouldn't bother; or now, because they're afraid to go to hospital). That seems entirely consistent with the serology studies.
You originally said:
> There are basically no (larger than a single building) places that can legitimately claim 5% or higher infection rates
I believe that's very, very false--NYC's official numbers are around 4x that already, and those imply an IFR consistent with (and even higher than) estimates like Verity's. To claim even 2% IFR seems borderline-irresponsibly high to me, let alone the >4% that implied.
If you post links to the studies you mentioned, I can take a look. I believe the Massachusetts one you mention is Chelsea, which indeed found a very high rate; but subjects were approached in a public place (thus oversampling people with higher-risk behavior), in a particularly hard-hit neighborhood. So I'd guess its results are correct for the people sampled, but not representative of the overall city.
Have you studied how the specificity of antibody tests is validated, like on blood banked before the emergence of the virus? Why don't you think that validation is adequate? I agree that the correctly-calculated confidence interval for something like the Santa Clara study is near-uselessly wide, but I don't see any reason to doubt serology from harder-hit areas.
1. https://www1.nyc.gov/site/planning/planning-level/nyc-popula...
Yes, my understanding is two-fold. One, that independent verification (like https://covidtestingproject.org/) found less specificity than the manufacturer reported results. And it found that some tests had false positive rates of well over 10%, which is just completely useless. The Santa Clara study is useless with pretty much any test, but even some of the NYC numbers have error bars down to zero, depending on the serology test used.
> in a particularly hard-hit neighborhood. So I'd guess its results are correct for the people sampled, but not representative of the overall city.
I can and do also totally believe conclusions like this are possible: I'm sure that there are population subsets that have achieved near total immunity. I don't particularly doubt the conclusions of the Boston homeless study, for example, although I expect if we check back in on them, many will have reported symptoms and some will be hospitalized. But again, drawing conclusions about "we should reopen" from that kind of thing is irresponsible.
> I believe that's very, very false--NYC's official numbers are around 4x that already, and those imply an IFR consistent with (and even higher than) estimates like Verity's. To claim even 2% IFR seems borderline-irresponsibly high to me, let alone the >4% that implied.
It looks like you're correct, the data I was looking at was wrong/outdated/I can't read or something. Given that, I would agree that NYC basically has to have 5% infected, and 10% is reasonable, although 20 and 25% I'd still have doubts about (and even more doubts about them given that the data is from ~3 weeks ago). So yes, I'd agree that an IFR of 4% plus is irresponsible to suggest, and I didn't intend to do so.
NYC appears to have used an in-house IgG assay for which unfortunately no paper exists; but the Wadsworth Center seems reputable, with no obvious incentive to overstate the prevalence. To any extent they're under political pressure, that would probably go the other way (to show the success of the lockdown, and justify its continued effect). So perhaps they made a simple mistake, but their IFR ends up consistent with serology performed by other means in Europe, and with the few populations universally tested by PCR--actually on the high end, especially if we include probables or use full excess death.
The harm of the coronavirus certainly shouldn't be minimized. The harm of the continuing lockdown shouldn't be either--a generation of schoolchildren may lose months of public education, one of very few social equalizers in American society. Given the harm on both sides, I believe we should be making policy based on our best estimates of the IFR, and not either extreme of a confidence interval. That seems unquestionably <2% to me, and probably <1% based on European serology (though the NYC may be slightly higher).
https://www.google.com/amp/s/www.wsj.com/amp/articles/corona...
pay attention to hospitalizations. that metric is disconnected from the new cases and expanded testing relationship.
As soon as hospitals start allowing elective surgeries again hospitalizations become less of an indicator and you have to go by "suspected" or "confirmed tested" covid hospitalizations.
It'd be disconnected if you had samples that are comparable (ideally random, representative of the population), but that's not the case in most countries. When you start with very limited testing capacity, you prioritize "important" groups of people that are often at higher risk of infection (healthcare workers, older people, ...) so you can expect higher percentage of people to test positive. But as you increase the capacity / number of tests, the sample structure will change - people with lower risk of infection will get tested, so it's natural to expect the percentage positive to decrease.
If percent positive of all tests is the same as before then the infection rate is not decreasing, or increasing - its constant. That is the case in many areas although some are seeing decreases or increases. E.g. some are increasing (e.g. Maine, South Dakota), some are decreasing (e.g. NYC, NJ), some are constant (e.g. DC/MD/VA have been constantly at 20% positive for the last couple of weeks now).
Most states are looking for a decreasing rate of positive tests for two weeks especially of its the percent positive case is under 10%. Generally under 10% is considered under control or manageable by experts.
So at first tests were basically only given to people who had multiple symptoms, to be sure. Now you can get a test in my area if you're asymptomatic but working at a job that requires you to come into contact with other people.
If the percent positive of all tests is constant even as you test more people, that means that even as you test less obviously symptomatic people, the number of infected is constant, which could mean the infection rate is going up. Or not, but it's not certain that it's complex.
> Generally under 10% is considered under control or manageable by experts.
Can you cite this?
https://www.npr.org/sections/health-shots/2020/04/22/8405263...
There are several other sources out there that state that for example the WHO also recommended it as a guideline. Yes not everyone is being tested but the 10% does account to some degree of the "severity" of those being tested. If you're only able to test the most likely cases you'll get a much higher % positive.
Asking interesting questions is good. Asking commonly answered questions is...
Hospital beds, like percent positive tests, are only one metric for determining health care capacity and how close we are to having an outbreak that overwhelms the all resources needed by the health care system to properly function. Globally there is a shortage of the medication used for sedating which is used when doing intubation or another medical procedure that requires sedation. Globally there is a shortage of PPE. PPE is at an extreme shortage globally still, and without PPE you have a shortage of medical workers (who then get sick or potentially spread it to other patients who aren't covid patients) - without health care workers those hospital beds become pretty useless. A lot of the initial cases in Wuhan came from cross exposure of non covid patients to covid patients at the hospital so canceling elective surgeries And mandating PPE made sense to prevent hospitals from becoming infection hotspots.
Why is everyone so hyper focused on hospital beds? They also literally canceled all elective surgeries as part of these stay at home orders which freed up something like 70% of beds that were being used. Additionally there aren’t as many trauma victims because car accidents, tourism, etc is down.
They know severity of cases and poor outcome when contracting the disease comes from initial viral load - being in close, confined places with someone who has it for a substantial period of time increases the severity of the disease. This is true for this virus and has mountains of evidence behind it in the study of virology to be true in general with viruses. It’s increasingly important here because the virus has no treatments in terms of medications, severe cases seem to be complex and the progression of the disease from this virus is not well understood. These guidelines and requirements for wearing a mask in public and need for health care worker PPE is a large part of the effort to help people have a less severe case of the disease if and when they do get exposed.
How myopic and superficial to not think about something more than hospital beds as a metric when evaluating these shutdown policies.
From tomorrow the devolved parliaments (certainly the Scottish government) will be diverging from tonight's confused "Stay Alert" messaging as announced by Boris. For at least the next three weeks the message in Scotland is still "Stay at Home".
I'm guessing it's because mitigations together with an increasing immunity is lowering the reproduction under 1.
In many large cities as many as 10 or 20% are now immune and that's a huge number if the theoretical herd immunity is somewhere around 60, and the practical one is somewhere even lower.