Here’s Kenneth Fritch’s projections for Illinois based on the (eventually to be a classic) Annan and Hargreaves model.
Open Thread.
300 thoughts on “Kenneth Cranked out Annan’s model.”
Comments are closed.
Here’s Kenneth Fritch’s projections for Illinois based on the (eventually to be a classic) Annan and Hargreaves model.
Open Thread.
Comments are closed.
I think that this forecast is so wrong. Illinois is still *detecting* over 1000 new cases a day (how many going undetected is anyone’s guess) and this virus is now ripping through nursing homes. Peak can’t happen until mid May at the earliest… and taking Italy as a guide, the graph will need a much fatter tail.
Skeptical,
This has a pretty big tail! We don’t know the death rate. But if it’s 10%, then 100 deaths a day about what you’d expect for 1000 new cases a day.
.
Of course the forecast could be wrong though. I hope it’s not!
Lucia,
Illinois isn’t doing enough testing… death rate should be much lower than 10%, unless you’ve got an exceptionally large number of old people in your state. Your actual death rate is currently about 91 a day, with a 1-2 per day gain, so you’ll be at 100 deaths per day in about a week. Your gain was about 4 per day two weeks ago… so you are reducing the gain but you haven’t hit peak yet.
Want to see a big tail… have a look what’s happening in Italy and you’ll see that the death rate doesn’t fall away very fast, and they have an insanely tight lockdown compared to you.
The problem with trying to find mathematical forecast is that deaths can range from 6 to 41 days after first symptoms appear. People who showed first symptoms yesterday might die in a week or they may not die until the end of May. Very hard to calculate anything when you have a variable like that.
skeptical,
I’m not entirely sure what your point is.
On “the peak”: That is the forecast using Annan’s model based on death data up to a certain date. The pink is a “best fit” to the data. The forecast includes uncertainty intervals. It doesn’t say we’ve definitely hit the peak. We may not have.
Also, it looks like Ken didn’t doesn’t look like it includes yesterday’s data (probably because it wasn’t available yet). If that new data was added, the graph might look different if re-run today. There’s a lot of uncertainty. Notice how flat the upper line is?
(The uncertainty intervals may be a bit low too. I’m not sure how Annan’s code deals with the fact that no shut down is a daily thing, nor how it deals with data possibly being reported in a not-perfectly timely way.)
So if your gripe is merely you think we haven’t hit the peak: We may not have.
On testing:I agree they aren’t doing enough testing. That’s why I think projections based on cases probably won’t work.
On the death rate: Yes, most people (and I ) think the death rate is much lower. You said the graph forecast rate couldn’t be right because new cases are in the thousands. My point is: that’s not inconsistent with a lower and flattening death rate. That death ~ cases*death rate. If the death rates is 10% or less then there being 1000 new cases is consistent with 100 daily deaths or fewer. So there being 1000 new cases does not mean daily deaths must rise higher than 100. Maybe they will… or not. But that’s just not evidence for it.
How do you get a 1-2 per day gain? Are you just fitting a line to some amount of daily data? (Real question.)
Skeptical: BTW, I didn’t say we’ve hit the peak. What’s I’ve said is the quarantine appears to be working. And I’ve said that’s output of a particular model with particular data.
Lucia,
I have been following the deaths per day in Florida (https://fdoh.maps.arcgis.com/apps/opsdashboard/index.html#/8d0de33f260d444c852a615dc7837c86)
There is some ~ 7 day oscillation around the trend (no doubt a weekend reporting issue), but there is now enough data, especially if you look past that 7-day oscillation, to see that deaths per day have dropped by about half in ~12 days. On that trend, daily deaths will reach single digits by early May. Of course, the plan is to start gradually easing restrictions in Florida, but it is unclear to me how much difference that will make. Right now many people are paranoid to the point that they are avoiding virtually all contact with others. Easing official restrictions is not going to make them suddenly less paranoid. People will start going to the beaches, but I suspect they will still keep their distance from others, even on the beach. We have about 8-9 traffic deaths per day in Florida; maybe virus deaths falling below that level will start to calm people down.
I reran the Covid Bayesian Annan code for the US death data that I used previously but this time with different priors. The R Studio report is linked below. I changed the priors means but not the standard deviations. The changes were based on my uninformative view/knowledge of the involved epidemiology. Here are the priors and posteriors (in parenthesis) using the Annan priors followed by my changed priors and posteriors (again in parenthesis):
Latent period in days: Annan 4 (4.1) Fritsch 8 (7.9)
Infectious period in days: Annan 2 (2.0) Fritsch 4 (4.0)
Days to death/recovery: Annan -15 (-14.2) Fritsch -10 (-14.4)
Death Rate Annan 0.0075 (0.0075) Fritsch 0.015 (0.015)
Initial reproductive rate (R0): Annan 3 (3.2) Fritsch 4 (6.3)
Post lockdown reproductive rate (Rt): Annan 1.1 (1.1) Fritsch 2 (1.3)
The fit of the posterior model to the observed data is much the same using the different priors and thus the near same results for the all important Rt posterior can be obtain with largely differing latent and infectious periods, death rates and R0. I was hoping that the Rt posteriors would be nearly the same at the 50% level, but the differences (given the posterior distributions for Rt in both cases) are not significantly different.
https://www.dropbox.com/s/lb2cfgwa4vhr56i/US_Covid_Bayesian_Diff_Priors.docx?dl=0
lucia (Comment #183435)
Yeah, just smoothing daily deaths while semi-preserving total deaths. You can’t use a line fit because the rate is changing.
lucia (Comment #183435): “I agree they aren’t doing enough testing. That’s why I think projections based on cases probably won’t work.”
.
For the U.S. as a whole, daily deaths have been tracking extremely well with daily new positive tests, but with a six day lag. In other words, 6% of the number of new cases today is a very good predictor of the number of deaths six days from now. That backcasts nicely all the way back to March 10 for deaths (March 4 for new cases).
So new cases are a good proxy for deaths. Cases have two big advantages: much better statistics and more up-to-date information. What really matters is the trend in new infections; both cases and deaths are lagging proxies for that, but cases lag by less.
The same seems to work at the state level, although only New York and New Jersey have enough deaths for the noise to not obscure things. Some states might have a smaller lag, but it is hard to be sure.
skeptikal (Comment #183423)
April 22nd, 2020 at 2:46 am
I believe that Annan’s model is accepted here as a first attempt at predicting future deaths given the current data. That was Annan’s purpose in publicizing it; he wanted to motivate the epidemiology people to look at it and improve upon it. His paper linked at this blog showed promise for the predictive power of the model. That is something we can test here as more data becomes available. We certainly need more than anecdotal evidence and people’s views that might be biased with an agenda in determining costs in getting the economy up and running again.
I am primarily interested in the posterior for Rt from the model and its probability ranges and of course the influence that the priors have on the posteriors and especially Rt.
lucia (Comment #183435)
My point is that I think the model is completely wrong. I thought the whole point of you posting the output of this model was to discuss it’s merits.
skeptikal (Comment #183443): “My point is that I think the model is completely wrong. I thought the whole point of you posting the output of this model was to discuss it’s merits.”
.
You have not said anything at all about why you think the model is wrong. You have only complained that you don’t like the results.
skeptikal,
This model is more useful than the IHME Murray model. The projection should improve with another week’s data.
Mike M,
“Cases have two big advantages: much better statistics and more up-to-date information.”
.
Certainly less noisy, although I am not sure about the up-to-date part. But the case rate has another disadvantage: there is a wide range of testing intensity…. meaning that some places are looking a lot harder for cases than others. That could mean that differing numbers of cases with mild symptoms are being reported. Once the number of deaths is high enough to reduce the influence of noise, I suspect deaths are a more accurate measure of total cases in the population, although with a much longer lag, of course.
SteveF
Yes. I’ve seen these oscillations in other data. People are often discussing Sweden… Sweden.. Sweden. It has that same darn oscillation in it. I can’t help wondering if there isn’t some sort of instability that rides on this causing the oscillations? Or maybe it’s a “sub-group” thing with outbreaks bursting out in one nursing home..tailing off. then another later… These oscillations are strong relative to trends in Sweden. You can even see “flat-rise-flat-rise” in the cummulative data Willis post!
Eyeballing Sweden, I honestly can’t guess if things are slowing and will avoid engaging on that until the peaks look flattish for at leasttwo oscillations which hasn’t happened yet.
Kenneth,
Do you specify the lockdown date in the Annan model? Or does it find the optimum one based on the data? I’m asking because it’s hard to really identify a single lockdown date for the US from behavior of various governors and so on.
At least in Illinois it’s pretty clear that “something big” happened between March 18-March 20. A fair number of other states did similar things in that time frame, but some definitely did things earlier and some did not-so-much relative to others.
No matter how the date is determined, I tend to think Illinois (and each individual state) will be a cleaner application of Annan/Hargreaves model just because individual states do have something closer to a “before/after” situation. (Mind you…. none of the states are going to be perfect.)
I should have added: Thanks for running these. Anything you run is interesting. Maybe we can get James to comment at some point (even though I think he doesn’t like me or us. 🙂 )
MIkeM,
You could run Annan/Hargreaves model with cases. 🙂
Although, I guess it would need to be adapted since it really is organized for deaths! It’s got a whole gamma function in there. So, some thought would be required.
Mike M. (Comment #183445)
I have said why I think it’s wrong. I think it peaks too early and I think the death rate falls away too quickly.
I also think that the model relies on a death/recovery time that’s too short… and it relies on setting an arbitrary death rate which most likely won’t resemble reality.
Yeah, there’s a lot of things I don’t like about this model… but I also don’t have anything better to offer, so I might just keep my mouth shut from now on and I’ll just watch how this model compares to the actual numbers as they come in.
Lucia,
I suspect the oscillation with a ~7-day period really is just due to less reporting (cases and deaths) on weekends. The more relaxed Swedes have a more exaggerated oscillation than most other places.
.
I have noted something very strange. The IHME model for Florida (https://covid19.healthdata.org/united-states-of-america/florida) shows the death count history (not projection, history) which appears wildly disconnected from the Florida department of health data (https://experience.arcgis.com/experience/96dd742462124fa0b38ddedb9b25e429). IHME has multiple days with >70 deaths, while the state of Florida show a maximum of 45 deaths on April 6, and all other days lower. So someone has very wrong numbers. The Florida death count is blared each and every day via local news outlets, so if the State was way off, or making giant revisions on death counts long after the fact, then I imagine someone would notice. I can’t find any explanation from IHME where their data comes from, only that the date is “based on the local time zone”.
Kenneth,
I think we are all interested in the posterior Rt (and its reliability). I soooooo want it to be 0.5 or less in Illinois. That would suggest we could relax a little. If it’s 0.9…. not so much!!
Lucia,
“even though I think he doesn’t like me or us.”
.
LOL.
skeptikal
I suspect that either you haven’t read Annan/Hargreaves manuscript or Annan’s blog or don’t understand what their Bayesian analysis does.It doesn’t set a death rate at all. So clearly it doesn’t set an arbitrary one. At most it sets a prior which is updated based on data. It does use values for priors, and these are taken from published papers but given fairly broad uncertainty intervals to allow the the posteriors to take on values from data.
.
Your objections appear to be based on criticizing the model for assumptions it doesn’t make. That makes it hard to see them as important.
.
Yes. The projection may be off for a number of reasons. But the ones you supply don’t happen to be among them!
I think based on recent blog posts Annan has also seen some disagreement among reported values in UK. I wouldn’t be surprised if we couldn’t find some disagreement in Illinois if we poked around. I just don’t know where to poke around!
.
So many of these sites what to provide you nifty graphics of their choice. There’s no easy “find the underlying data” button!!! 🙁
I think one issue with FL death count differences is the health department count if for residents only? There is more going on than that I think though, not sure why the counts are so different. Apparently we don’t care about snowbird deaths, ha ha
.
“Deaths: The total number of Florida residents who have died with positive COVID-19 tests. Includes Florida Residents who died or were tested, infectious and/or exposed while outside of Florida. ”
.
If your region has been in lock down for a month and your death counts aren’t declining yet, then you have a major problem on your hands I think. This basically calls for increased lock downs.
.
Lock down fatigue is going to be a problem. I already feel tired of it and don’t have additional stressors of income and employment. People won’t put up with this stuff for more than another month if they don’t see people getting sick and dying around them.
lucia (Comment #183458)
Okay, I plead guilty.
I’m not overly interested in trying to fix their model so I might bow out of this thread.
Lucia,
The FL department of health has a press release each day reporting total cases and total deaths. They say the dashboard page (with the fancy graphics) is updated each day at 11:30 AM and 6:30 PM. They don’t say anything about data sources.
.
Tom Scharf,
Not counting someone who spends 5 months a year in Florida, and who died in Florida, because they are not officially residents would be beyond bizarre, but I suppose it is possible. The number of part time (winter time) residents is above 1 million, so more than 5% of the state’s population in the winter. The Dept of Health goes out of their way to separately count cases of “non-residents”. Here is the link to the “underlying data explanation….. not very clear. https://fdoh.maps.arcgis.com/home/item.html?id=8d0de33f260d444c852a615dc7837c86
Lucia,
A comment went to moderation… don’t know why.
skeptical,
Obviously, no one suggests you try to fix their model. Thanks for confirming that your objections are based on not knowing what it does– because you haven’t even tried to find out — and making entirely incorrect guesses about what it does.
.
There are assumptions in the model that one could point out are “not true”. That’s the case with any model. Here are things it doesn’t consider:
* seeding rate from infected outsiders traveling into Illinois from elsewhere. That nice Filipino American Lady from Indiana who comes to dance parties in Illinois? Truckers driving across Illinois on I80? The Quad cities spanning Illinois and Iowa? The model doesn’t capture these and other effects.
* people healthy and unhealthy people dieing of natural causes. Babies being born.
* the effect of inhomogeneities in Ro across the many villages, towns and cities in the state. Everything is wrapped up in on R.
It’s a leading order model. The true uncertainty intervals are almost certainly too tight because the assume the under-lying model is “right”, given the appropriate parameters. But of course some effects are missing.
I think it’s still a great contribution compared to lots of other models out there. It’s worth seeing what it shows and I’m very grateful to Kenneth for cranking the graphs out.
SteveF,
I fished the comment out of moderation.
Lucia, I have used an intervention date of 3/20/2020 for the USA, NYC NY state and IL runs. If you think that choice might significantly affect the results I run different lock down dates.
I should note the intervention date in my posts. The plot in the reports I have linked notes the initialized date and number of day’s worth of data on which the model is based and fitted after lockdown . The final 5 days are reserved for testing how well those points fit the model’s predicted values – or at least as far as I can tell at this point in time.
As far as Annan is concerned you would know whether he loved you or not in his first post. As I recall he does not hold back and has a rather pointed approach. Almost all climate scientists from all sides of the aisle that I have dealt have been very polite and especially so with emails.
These Bayesian runs as would be expected take a relatively lengthy time to run and get a R Studio report. I can vacuum most of the house or fix dinner or take a walk while it runs.
Lucia,
Thanks for recovering my comment.
.
My main concern about the Annan/Hargreaves model is that its posterior estimates of key model parameters don’t seem to change much from the corresponding priors…. even when wildly different priors are used. The priors seem “resistant” to changing by the modeling process. This is something of a head scratcher, because it suggests crazy priors are no better or worse than sensible ones in predicting the future…. because they don’t change much in the face of actual data. Maybe the width of the uncertainty in future deaths does change with better selections for priors, but this would have to be shown.
.
BTW, I suspect skeptical will object to any analysis or data which indicates mandated lock downs (business closures, school closures, restrictions on activities, etc) can be eased. Indefinite closure of hundreds of thousands of restaurants and businesses is not an option; skeptical will have to get over it.
Kenneth,
“These Bayesian runs as would be expected take a relatively lengthy time to run and get a R Studio report. I can vacuum most of the house or fix dinner or take a walk while it runs.”
.
Good grief. Interpreters are computer power hogs.
Kenneth,
I have no idea if it affects it a lot! I’d have to be running the model with both to see the sensitivity. Once we have a lot of data, I don’t think it will matter.
I just wondered if that was a hard coded date, of if that was also a little Bayesian. (I thought it was hard coded.) So it sounds like it’s hard coded… right?
I don’t really know if Annan dislikes me or not and it really doesn’t matter. But I think it’s fair to say that very few people would think that support for his model from people at this site is because of some strong historic alliance between us or all our past best-friendness. 🙂
I’m very glad James and Jules came up with this, slogged through the data, tested it, posted forecasts and put their paper out there. It does things models should do:
1) Recognizes that models should hindcast.
2) Is based on some “mechanistic” model that does describe large scale features of contagions. (Here SEIR).
3) Uses a method to tune parameters that makes some sense (not just pickes out whatever the hell he “likes”.)
4) Tries to make it as simple as possible but no simpler. (So, leaves out effects like the seeding from people “outside” the region as not being the leading order thing.)
I think any engineer recognizes these are the right elements in developing a model.
We know Illinois data were not used to develop these thing.
SteveF,
Well… skeptical really jumped out on this model. In the end, his objection does see to be based on the output alone. And even though the traces go through the data, and his objection is based on not liking how it compares to data, he says things like:
Really? Completely?
But then:
He admits he hasn’t read the model and doesn’t have any idea what the recovery time it “relies†on is.
In fact, the model accounts for deaths happening exactly this way!! There’s a whole subroutine computing the death “today†based on the past history of infections! So one would think if skeptical had even the remotest notion of what the model does, he would be applauding it for doing what he thinks it ought to do.
But, of course, the problem is he doesn’t have clue one what it does!
Now to your criticism (which I think is more valid)
I suspect this is because there is not much data. I think (but could be wrong) AH got their priors from some published papers (studies of infectious rates, death rates). They also got some from modeling the Wuhan case. Some numbers should be the same for all cities or countries because they are biology based. (How long you remain infectious should differ much in Wuhan and Chicago.) Others should vary based on behavior.
There isn’t a lot of data. By definition we will never have a lot of data when these things start. In principle, the uncertainty intervals should account for the difficulty in knowing the parameters — and that would include the difficulty in nudging them.
Right now, the Rt Kenneth got isn’t saying we can just open up with impunity. But OTOH, I don’t think anyone was going to do that based on a blog post! There are other factors— including current shortages at hospitals and so on. The model still goes a structure to see how things are going.
lucia (Comment #183472)
Lucia, I am not sure what you mean by hard coded, but it is a date that I put into the code. I have the following choices so far as to what I code into the program:
Start date, intervention date (lockdown), population, data (daily deaths) and priors. And, of course, the label name of the region of interest.
I agree with you about what say about Annan’s modeling plus the fact that I thought he (they) did a great job in explaining it and documenting the code.
SteveF, I think that models with many parameters have similar problems with initial conditions (values/priors) whereby the critical parameter can be reasonably well tracked with variations in other parameters. You really start worrying when one of those fitting parameters comes up with a crazy value.
Kenneth,
I mean… does it have both a prior and a posterior? Or is the number unchanged no matter how much data goes in? (I should look at the code more. I have some tasks though… arghh…)
lucia (Comment #183476
No, it does not have a prior or posterior. The 6 priors were listed in my post and the 6 posteriors are always listed in the R Studio report. They are unnamed in the report but I named them in the first US run report.
SteveF (Comment #183471)
Steve with a compiler my multi-tasking would not allow for vacuuming, walking and preparing meals. I do not know how much faster the analysis would be with a compiler, but I do know that Bayesian analysis was not very feasible without fast computers.
I don’t know about R specifically. Languages that are interpreted in the sense that they execute on a virtual machine are usually pretty close to native code these days.
https://stackoverflow.com/questions/145110/c-performance-vs-java-c
Python is substantially slower;
https://stackoverflow.com/questions/801657/is-python-faster-and-lighter-than-c
but nobody in their right mind would crunch numbers in native Python. There are optimized libraries (basically wrapped up C or C++ compiled code) for that which give much better performance than Python.
Just a meaningless aside. 🙂
mark bofill,
I have had a couple of experiences where I ran the same code interpreted and compiled. Huge differences (3+ times faster compiled). That was years ago; maybe interpreters have gotten better. Of course, hand coded assembly is the fastest, but you need the patience of a saint and have to carefully avoid stack errors, etc.
One issue that jumps out is its reliance on covid deaths data. Experts are arguing over that furiously. I’ve seen arguments that it’s too low by a factor of 2 in Italy and some say it’s 3 times too high because most of the dead had serious health issues that were probably the real cause of death.
With the flu, we just accept a high level of ignorance and up to 60k fatalities per season. The demand for vastly better models and data may be the result of panic and not a huge practical impact.
For healthy people under 65 the chance of dying from covid are infinitesimally small according to Ioannidis and collaborators. That knowledge is quite certain and adequate to design a policy that protects those at risk and could have prevented the disastrous shutdowns. Those who usually die are often seriously ill already and had limited life expectancy. But in future we can focus on keeping them isolated.
And this age differential was clear from the beginning.
Another issue has to do with the lockdown date and it’s use in the model. There are other reasons why the curve might bend down including herd immunity. Nate Silver is postulating that might have happened in nyc. If infection prevalence is 50 times reported cases that is quite likely, There are a few data points that indicate that in early April about 15% had active infections and most were asymptomatic. Mild cases can recover in a couple of days as happened to a colleague of mine’s brother on the east coast.
Another reason is just that most of the truly vulnerable population has either died or become immune.
So this link to lockdown date seems to me unjustified and based on prior conviction that lockdowns ‘work’ whatever that means.
David young,
Do you mean the Ionnidis being cited here:
https://www.statnews.com/2020/03/17/a-fiasco-in-the-making-as-the-coronavirus-pandemic-takes-hold-we-are-making-decisions-without-reliable-data/
That article bases their estimate of mortality on only 7 dying on the Diamond Princess. That was the number back on March 17.
But now, 13 have died
https://www.statista.com/statistics/1099517/japan-coronavirus-patients-diamond-princess/
Doubling Ionnides range, he’s presumably estimate a mortality range of 0.1% to 2% now. Other data may have changed also. (Like perhaps age ranges of people getting sick.) I don’t think I’d call that range “infinitesimally small”. It’s not the plague, but a 1/50 chance of dying isn’t teensie-beensie.
SteveF (Comment #183470)
No, not at all. I’m just looking at the numbers and from what I can see in places where the lock downs happened earlier, the decay in daily deaths is a lot slower than what this model predicts.
When the world first went into lock down, I expected daily deaths to fall dramatically after the lag time… but that hasn’t happened. I still believe that this policy is saving lives, but not as many as I originally thought it would.
You have to remember that I’m living in lock down too… and it’s not a lifestyle choice that I would voluntarily make, but I can see the necessity of it now.
There is growing unrest as the lock down drags on, and the leaders who decide when to ease the restrictions have some really tough decisions to make… balancing economic needs with health needs is not easy.
David Young
Can you point to data that suggest 15% of New Yorkers (or soemone) had active infections in early april?
Skeptical,
Can you show your comparisons of this models predictions to data? (Real question.)
This is the model prediction for Italy when it was first made. (The blue are the forecast) Unlike your gut feeling, it said Italy’s decline would be slow– Rt = 0.9.
skeptical,
“ You have to remember that I’m living in lock down too… and it’s not a lifestyle choice that I would voluntarily make, but I can see the necessity of it now.â€
.
It depends. 8 or 10 states never instituted draconian measures, and still have very low infection rates. Do you suggest they now institute the same policies as Michigan? Real question. Some states have crazy restrictions along with very high infection rates. Which means the policies are nowhere near able to stop the spread, no matter how draconian. The necessity of draconian policies and their economic and social costs are debatable most anywhere, but draconian policies are plainly destructive where infection spread is low.
.
Rational policy should be driven by data and a thoughtful balancing of costs and benefits. Seems to me that in most places, with a few notable exceptions, like Sweden, public policy has been driven by fear and hysteria…. made worse by lefty totalitarian tendencies and infantile demands to never place economic value on human life.
Ioannidis doubled his IFR to account for more people dying. 0.25 was his final number. Also Santa Clara and Los Angeles studies gave similar numbers. In any case, official numbers are totally wrong and are needlessly scaring people.
0.25%vis not negligible. However that’s not what the paper said. For those in good health and under 65 years old the fatality rate is infinitesimally small. Why is this never mentioned by yellow journalists?
One nyc hospital did testing of all expecting mothers who were admitted for delivery. It was in New York post maybe?
David Young,
https://www.nejm.org/doi/full/10.1056/NEJMc2009316
15% positive
I notice that when NBC nightly news reports on lockdown protests in, say, Virginia, they always interview healthcare workers in NYC about why ending the lockdown is a bad idea. It is a bad idea in NYC, but VA is not NY.
Thanks SteveF,
That’s good information. Also: those test are over a span of time. Since we would tend to expect the % infected is increasing, the number positive could be higher. The article mentions possible impact of false negative rage. But there is also false positive rate….so… % infected could be lower. It would be worth knowing the both false pos and false neg rate.
Well… can’t have everything.
BTW: If that’s the infected fraction… and the priors suggested by Annan are true…. it would be good news for NY. It would be very useful to know the false positive rate though. I doubt we do. If it’s 10%, then…. no, not good news. If it’s 1%… good news.
David Young
The impact of age is frequently mentioned.
But I have to say, I don’t see why journalists whether yellow or otherwise should be focusing on the lower bound of an estimate based on lots of data massaging, and then ignoring the upper bound for the population as a while. And even Ioannides estimate for that is not “infinitessimaly small”. All see need to do is look at the numerical range he gave, multiply by two (to account for the fact that he made the estimate when only 7 had died and now it’s 13) and mention the number to see that it’s not necessarily “infinitessimally small”.
It might hypothetically, if you accept a lot of assumptions and data manipulations, be quite small. But it might be pretty large when you consider what you are talking about is “risk of death”.
David Young,
The Santa Clara study was not a random sample. Subjects were recruited on Facebook. That makes the results questionable. It’s also highly unlikely that the asymptomatic infection rate is 50 times the symptomatic rate. Rockland, Westchester and Nassau counties in NY have over 2% of their population as confirmed cases. NYC has 1.7% confirmed. I seriously doubt that 85% of the population of NYC has been infected. For one thing, the new case rate seems a little high for that.
Lucia,
RE false negatives and positives: I believe those those tests were the nasal-swab tests for the presence of the virus itself, not the antibody blood test. It is far less likely the nasal swab would give false positives unless the equipment was contaminated, because the test looks for the virus itself, not an immune response to the virus (which might come from antibodies to other corona virus infections in the past). The nasal swab certainly could give false negatives, especially if the person was very early in the infection and not yet shedding many virus particles. The simplest and most likely conclusion is that there is a very high rate of infection in NYC among asymptomatic people. The authors of the paper suggest false negatives are likely.
.
Two things to keep in mind: 1) pregnant women could be more or less susceptible to catching the virus because they are pregnant (unlikely, but who knows?), and 2) pregnant women are not in the prime age group that tends to get symptomatic illness… that is mostly the over 45 crowd, and there aren’t too many pregnant 45+ year olds. We just don’t know the % of the population that gets the virus and never has symptoms…. but it is likely very significant.
Lucia, the Infinitesimally small statement about “healthy non-elderly people” (<65 years of age) was in another paper Ioannidis wrote with some collaborators analyzing the European data. It had nothing to do with the DP dataset. Ioannidis has been churning out at least 1 paper a week with lots of collaborators. They are all good reading in my opinion.
Ioannidis in his DP analysis took account of the liklihood that there would be further fatalities. His analysis is still valid. If you listen to his youtube videos he is very measured and serious. Vastly better than Annan.
And I'm not focusing on a single number. There are at least 7 or 8 studies around the world using serologic testing by now and all indicate that infections are vastly more widespread than official numbers. There was one in Germany a couple of weeks ago. Iceland has tested quite extensively. There was a town in Italy where everyone was tested. Some of these are a month old. I don't have time to find them right now, but they are easy to find. There are now also the American and French aircraft carriers that I've seen virtually nothing about. All these studies have flaws, but it is virtually impossible that all of them are off by a factor of 10 to 60.
In general the media have been yellow journalists in splashing all over their screens the top line numbers, such as 800K cases in the US and 40K fatalities. Neither number is reliable. The former could be off by an order of magnitude at least. The latter is also an artifact of physicians guesses as to the real cause of death in people with multiple serious illnesses. It satisfies some deep human cognitive flaw to obsess on crises and magnify them. Every hurricane is by some cherry picked measure the "worst ever." Remember Sandy?
Also the whole phoney data driven fears that the hospital system would be overwhelmed in the US was amplified by the media with anecdotal reports of frustrated health care workers. My brother's hospitals are half empty and Oklahoma peaked a week ago. BTW, the UW model was way way off for Oklahoma. Oklahoma does not have a blanket lockdown order. It was comical too. 2 days after the peak they updated their model to say that the peak was 2 days ago. When this is all over and the fatalities are added up, we will be surprised at how many have died because of fear of contacting the health care system and/or prematurely shutting down "elective" procedures and screenings.
DeWitt, See my previous comment.
David Young,
If it bleeds it leads.
.
That said, I believe there is an extraordinary measure of bad journalism here due to the MSM’s animosity toward Trump. What the MSM fails always to explain is that while the overall CFR may be 0.25% (or even 0.5%), that is a number which hides more about the illness than it reveals. First, herd immunity starts to depress deaths long before ever reaching the headline CFR; the USA is not going to have one person in 400 or 1 person in 200 die from the illness; growing herd immunity would likely cut those rates in half. Second, the profile of victims is not adequately explained. The typical victim is a 75 to 85 year old, usually with serious pre-existing conditions, and who’s personal CFR for just being alive, without a coronavirus infection is 3% to 5% per year. Yes, the corona virus adds significantly to the risk of death for the elderly, but other causes of death pose far greater risk. Unfortunately, life is fatal.
DAvid Young
If so, you should provide a link so we can read what they actually wrote.
This strikes me as “baseball commentator” statistics. “His strike out rate on right handed batters during the 3rd inning is….” As such, I can totally understand why journalists (or anyone) would not spend much time reporting this..
.
Right now with no links, I can only go by the article I found. In that article, which merely describes what Ionnides did,
* Ionnides used a death rate from the Disney Princess (or whatever that ships name was) that was off by a factor of 2.
* He did a bunch of fiddle facgtors.
* He came up with a range of possible deathrates. The upper range, when corrected for the factor of 2 is not infinitessimaly small.
.
Given the error of the factor of two and the described massaging of numbers, you will have to link me to an actual paper by Ionnides to read his argument in that paper. Because right now based on the one article I found which read as a very unconvincing tenuous estimate it doesn’t sound like it’s worth paying much attention to what he wrote.
I’m perfectly willing to believe he has something clearer, written in his own word as opposed to garbled by a journalist. But I’m not going to hunt it down under the circumstances.
Dewitt
Heck… 2% * 50 = 100%. That’s either (a) epidimemilogically imposible or (b) getting sick does not result in much immunity.
The latter would be very bad news. It would suggest (1) we will never reach herd immunity and (2) the flattening of cases cannot be the result of a build up of immunity and can only be due to the shut down.
So, I hope 2% known infected with a 50/1 rate of symptomatic is not true because it would mean we are in trouble for a long, long, long, long time with no apparent way out!
Hot off the presses. Preliminary results of testing suggest that 13% of New York state residents have antibodies to covid19 concentrated in NYC and long Island. Apparently Cuomo announced it in his daily press briefing. I’m guessing that means perhaps 20% infection rate in NYC itself. That’s lower than my guess but still much much larger than the official “case” numbers.
I’d be very interested in knowing the numbers by age bracket too since virtually all hospitalizations are among the old and sick.
Another contribution fro Ioannidis.
https://www.medrxiv.org/content/10.1101/2020.04.05.20054361v1.full.pdf
Money quote:
“Overall, this further strengthens the notion that for healthy non-elderly [<65 years of age] people, the risk of dying from COVID-19 this season has been infinitesimally small. This is in stark contrast with many news stories that focus on the demise of young people and the panic and horror that these widely reverberated stories are causing.â€
Here is the link to the article by ioannisis et al: https://www.medrxiv.org/content/10.1101/2020.04.05.20054361v1
.
Seems a reasonable analysis, but his is a tiny voice being drowned out by the MSM megaphone. They are not interested in calm analysis of relative risks.
https://www.youtube.com/watch?v=cwPqmLoZA4s
is recent and a good source of information.
Cuomo’s announced results: 14% of random people in the entire state test positive for coronavirus 19; 3% upstate, and up to 22% in the city region. That puts a reasonably reliable upper bound on the CFR of ~0.5%. Good news indeed. Unlike other studies, this one looks well designed. The big difference between positive rates upstate and near NYC suggests that they are not picking up false positives due to previous exposure to other coronavirus types.
SteveF, Was there any breakdown by age? Probably those who are aged and frail were not able to get to the testing sites so I’m guessing its virtually pretty healthy people.
Wow, 22% already. This is quite contagious, eh? We are going to need to carpet bomb NYC with bleach.
This is positive right now? Virus test? If yes, an other contagion numbers we’ve been flinging round (Ro~3, ratio of time for latency to infectius ~2) … it may well be close to 1/2 over in NYC.
(The dropping side in the natural curse of the whole SEIR thing takes as long and kill as many people as the rising side.)
I’ve been running my SEIR code to have a look.
Oh… it’s antibodies…. I thought viral. That’s different.
DeWitt Payne (Comment #183554): “Rockland, Westchester and Nassau counties in NY have over 2% of their population as confirmed cases. NYC has 1.7% confirmed. I seriously doubt that 85% of the population of NYC has been infected.”
.
lucia (Comment #183561): “Heck… 2% * 50 = 100%. That’s either (a) epidimemilogically imposible or (b) getting sick does not result in much immunity.”
.
There is another possibility. The positive tests tend to be of sicker people, so maybe the ratio of very sick to mildly sick varies from place to place. The severity of an illness often depends on the initial infective dose, which might consist of multiple doses over a short length of time. There are indications of that with the Wuhan virus. It could help explain why there is such an enormous variation between individuals in the severity of infections, why health care workers tend to get especially sick when they catch Wuhan, and why transmission seems to be much more efficient indoors than outdoors.
So if you catch Wuhan from someone who is really sick, you might be more likely to get a severe case than if you catch it from someone with a mild case. And if you have a high exposure, like if your roommate is sick, you are likely to get sicker than if you catch it from someone with whom your contact is briefer. So a place like New York City, especially with people locked indoors, might have a high proportion of severe cases.
The risk for people less than 65 years old is 15x to 70x lower than for the age group greater than 65 according to Ioannidis.
.
Combined with the new denominator NYC CFR data then this suggests that risk to people under 65 is likely less than 0.1%. This is not zero and is serious with the infectiousness of the disease, but perhaps not shutdown the economy serious.
The problem here is that we are already doomed to perhaps 100K deaths this year in my opinion given that a second wave is almost certain this fall. The question is how to minimize overall harm. So far, the discussion and government policy has been mostly focused on how to “save lives” by creating the worst depression in US history. The V shaped recovery can’t happen because people are scared and won’t return to normal behaviour for years especially if government officials and the media continue to scare them.
We should expect a smarter policy aimed at protecting the vulnerable and allowing most people to get back to their normal activities.
MikeM
To be clear: your theory is the ratio of asymptomatic / symptomatic in NYC may be lower, but higher elsewhere?
SteveF
I googled around. The Cuomo numbers appear to be antibody numbers, not viral number. (I had myself wondering given Ro and ratio of time scale for infectious period/ incubation period, how the ‘currently infectious’ could even get to 20%. People go through that pretty fast, so that’s never actually stupendously high– unless viral loads stay high but, for some reason, people are no longer infectious. Dunno….)
Tom Scharf,
The overall CFR for flu is about 0.1 to 0.12, so Coronavirus 19 is 4 or 5 times worse than normal flu. The bigger difference is that many people have partial or complete immunity to the flu, so the number of cases per year is reduced relative to a virus where everyone is (relatively) immunologically naive.
.
Probably 200,000 or more will die in the States before effective treatments and/or a vaccine becomes available. But bringing on an economic depression seems way too high a cost for some marginal change in that number of deaths.
Lucia,
I should have stated that it was an antibody test.
It is unlikely that people stay very infectious for a long time once their immune system develops effective antibodies. If this test was Cuomo’s idea, then he is to be congratulated. When people have better worst case data, then rational analysis becomes more likely. There absolutely are not going to be ‘millions and millions’ of Americans who die from this illness. That is absolutely good news. If you combine the CFR with the now well defined profile of people at real risk, a more sensible approach as we await treatments/vaccines becomes pretty clear: Focus on protecting/isolating those most in danger. The risk to healthy people under 65 is not zero, but it is very low. Fear is a terrible thing, and I do not doubt many people, even younger than 65, will remain terrified. But the more information the public has about the true risks and especially the true risk profile, the sooner ending the current madness becomes politically possible.
SteveF, If you believe that the IFR for people under 40 is 0.02% or less as the Imperial College estimates and we assume there are 150 million of them. Then 30,000 of them could die from covid. That’s not a big excess mortality. Similar calculations for other age groups and with various serious illnesses could lead to smart targeted policies to build herd immunity while minimizing fatalities by selective isolation.
I do believe that this episode conclusively shows that people like Cuomo and Inslee are proud owners of double digit IQ’s. They are also in my estimation political cowards by going for the “no cost is too high if it saves a single life” line of thought. The strong age and pre-existing illness dependence on IFR rates was very well known since January when the Wuhan data was available.
SteveF
Absolutely! They need this data.
It is a bit funny to read reactions that say the high number are horrible news. It’s actually great news. We already knew how many people were in the hospital. The more asymptomatic/recovered etc out there, the better. (At least assuming they can’t get it again.)
I was running scripts trying to figure out how in the heck there could be 20% currently infected given the other things we think we know about Ro and time scales. My curves suggest we never see that. Ever. So other things would need to be wrong! Or there would need to be something I don’t understand about the tests.
The numbers I have suggest 20% “recovered” would mean NYC will still have potential to have case numbers rising without the stay at home, but they’ll have a lot more breathing room than if only 2% of people have been infected and recovered!!
David Young,
Cuomo and a host of others have been horrible at their jobs, and have caused terrible, unnecessary panic. That said, I hope Coumo can wrap his head (regardless of his IQ) around the implications of the NY data. The “each life is of infinite value†crowd are either retarded or politicians trying to get re-elected. Cuomo may be both, but I do think he is a lot smarter than his brother Fredo, who obviously is mentally challenged. I hope the governor sees his way to opening NYC sooner rather than later.
David Young: “They are also in my estimation political cowards by going for the “no cost is too high if it saves a single life†line of thought.”
.
*cough*abortion*cough*
I don’t think NYC’s tourism futures are looking compelling at the moment.
.
I think this is good for the mortality rate, but bad for the contagiousness. It seems too easy to pickup, and thus very difficult to defend against with social distancing. NYC did a lot of testing and has 20x the cases that weren’t tested. >20% of the population in 2 months. Ouch.
.
It’s very contagious and half the transmission occurs before symptoms. It rages through old folks homes. With massive testing seemingly a long way off, it may simply be uncontainable. I just don’t see how this is going to be stopped. Perhaps I am too pessimistic. Let’s hope there is some built in immunity somehow. This is going to be hard.
In Italy, active cases may finally have peaked. The high was 108,257 on April 19. Today it’s 106,848. Daily deaths peaked on 3/27 at 919. on 4/22 there were 437 deaths. The daily deaths are decreasing much more slowly than they increased as are new cases, which peaked on 3/21 at 6,557.
DeWitt,
The whole kissing all the time thing. (even between men) may be part of the problem. When the mafioso kissed you you knew it was fatal…. now any kiss at all.
Relative facts.
.
* 53 percent of all coronavirus patients suffered from hypertension
* Nearly half of adults in the United States (108 million, or 45%) have hypertension
.
* 42 percent of coronavirus patients who had body mass index (BMI) data on file suffered from obesity
* More than 35% of U.S. adults are obese
.
* 32 percent of all patients suffered from diabetes
* 10.5% of the U.S. population, have diabetes
Tom,
Well….eventually contagions stop themselves. That’s the way these things work. If it turns out a lot of people got it but it was so mild they never noticed they had it, that’s actually a good thing.
Tom Scharf,
comment #183596
And some people seem to be surprised that health care costs in the US are higher than in other countries.
In today’s WSJ:
https://www.wsj.com/articles/maybe-the-experts-were-right-about-covid-19-the-first-time-11587659799?mod=opinion_featst_pos1
Max_OK, are you still out there?
DeWitt,
“Max_OK, are you still out there?”
.
‘Out there” is the correct description. But in fairness to Max: he has not a clue about the technical issues. All emotions, all the time.
Tom Scharf,
So if you never get old (or fat, or both), you will be OK. We already knew that.
SteveF,
If you add the collar counties of Rockland, Westchester, Nassau and Suffolk to NYC, and subtract them from the state (using data from here), then 1.9% of the NYC area population have been confirmed infected and 0.3% of the rest of the state. So, at least in New York, the ratio of total infected to confirmed infected is about 3/0.3 = 10 to 22/1.9 = 12.
Using a factor of 11, that puts the lower limit (since some of the confirmed cases still might die) on the CFR or IFR for NY at 0.7%.
Tom, We are an unhealthy country. Usually however, these conditions occur in multiples. Metabolic syndrome usually includes obesity, diabetes, and high blood pressure. If people start paying more attention to their health as a result of this that would be a very good thing.
DeWitt, I hope that Ioannidis does an analysis of this data as lots of adjustments are needed to get a valid output number. It may be true that IFR is higher in New York City than elsewhere. I don’t know the age structure of that population, but I suspect very few young people with children live in the city. My guess is that many more older folks live inside the City.
David Young,
My prediction: this won’t make the fat, hypertensive or diabetic pay anymore attention to their health than previously.
One the thing lockdown will do is create more alcoholics.
DeWitt Payne (Comment #183601)
Max_OK, are you still out there?
_____
Yes, still here. Thank you for your concern.
Yup Lucia, A whole lot of people with little to do but drink, eat, or smoke dope and get into family fights will result in a lot of excess mortality down the line. But when they have no money left for rent or gasoline, the real civil unrest will begin. I guess there are continuing riots in France.
It’s been in my view the worst phony data driven fiasco in US history. If the government continues to print money at the current rate, things could get very nasty by next year. Argentina anyone? It will not be a V shaped recovery and lots of companies will disappear because people have been scared to death.
I just don’t see anyone really thinking about what will happen given current policies.
The media are of course the bottom line culprits here. I can’t stand to watch cable news or read even the New York Times. They really do push false narratives and very misleading data because they want the existing economic and in some cases political order to change and see the crisis as an opportunity. Just as an example, the New York Times had an article about the vaccine expert who was reassigned in which they regurgitated his claims with no fact checking or looking for other information. He is claiming that he was pushed aside because of Hydorxychloroquine. But Politico reports that in reality, he had been arguing with his superiors for over a year and this had been in the works for a long time. Further there is email documentation that this guy supported ordering 20 million doses of the drug. Journalism dies on the pages of the Washington compost and the New Yerk Times.
It was not a random test in New York, but a random test of people who attended big box stores. It is possible that this group is more likely to be infected as they are more likely to be going out.
They need to do a random sample at some point to get a more meaningful number. Ioannidis was saying that in early March and no one paid attention. Government “experts” wanted the tests reserved for sick people and health care workers.
We need to know the demographics of those who were tested. They were probably all healthy and relatively young. It’s also possible that among vulnerable populations such as the elderly, exposure has been a lot higher. With 22% of the younger worker population infected, few nursing homes would avoid an epidemic. Same for hospitals. That would be positive because it would mean that most of those at risk have already been exposed and the 2nd wave would be much less deadly. That would imply a true IFR that is lower too.
Every single study for the background infection rate is coming up with large numbers. None of them are definitive by themselves but the combination of all of them make this 10x to 20x number very plausible. It has to be the assumption at this point.
Tom: Yep. I wish they’d do one out here!!! 🙂
But then, I’ve been saying we need those tests for a while. We need them for many, many reasons.
Not quite right. If 22% workers were currently infectious few nursing homes would avoid an epidemic. If 22% of workers had cleared the virus and were now immune… totally different thing.
Should have phrased it better. If 22% were infectious in the last 2 months and half were asymptotic, it is hard to see how most nursing homes and hospitals could have avoided an epidemic.
“A slide shared by Bryan revealed that the half-life of the virus, in the absence of sunlight, lowers from 18 hours to one hour when the temperature rises from the low 70s to the mid-90s and the humidity rises from 20 to 80 percent.
Adding in sunlight, the virus’s half-life decreases from 18 hours at temperatures of 70 to 75 degrees and humidity of 20 percent to just around 1.5 minutes.”
.
Florida tourism futures are up, ha ha. I hope you don’t mind that all hotels have their air conditioners disabled.
So in a body temperature of 98.6, the half life is even less?
Churchill’s ‘end of the beginning’ might be an appropriate description of todays NY data showing very widespread past infection. The 30% of homeless with asymptomatic infection in Boston was a very real datum, even if it seemed almost impossible just a few days ago. Yes, people will quibble if the NY sample is perfectly representative, and yes, the upper bound for CFR can be argued a bit, But what has become clear is that the virus is not going to end life as we know it; only irrational fear and corresponding foolish and unconstitutional public policies can do that. For the first time in a month, since flying from Sao Paulo to Miami in an almost empty plane, I am cautiously optimistic sanity will return to public policy. It will not be fast, since the many idiotic politicians causing the problems are supported by many millions. of terrified voters. But it is now clear the lunacy will end.
I largely agree SteveF but I am very concerned what the long term consequences might be. I’m guessing a couple years of depression conditions followed by a lot of inflation. I’d put my money in real assets and am going to cash out the pension so as not to be dependent on a single company’s solvency.
The problem of a dishonest and partisan media will not go away either. If it’s possible it will get worse. Civil unrest is a real possibility too. What we have seen over the last 3 years may become the new normal with hyperpartisan and dysfunctional politics and a level of government corruption that’s vastly worse than
Watergate. It’s the biggest challenge since 1860 and who knows if we will find a Churchill or a Lincoln. If Biden is elected, its game over.
lucia (Comment #183520)
I see a whole bunch of blue lines. Isn’t the red line the one we’re supposed to be interested in? (Real question.)
SteveF (Comment #183526)
That’s a fair question. First I would question if they also have low death counts because infection rates are a useless metric in some areas where testing is low. Obviously areas with low population densities are going to have a slower transmission rate than areas with high population density… but even a slow transmission rate can end up with a high death count if left unchecked. Given the fact that every state is doing it’s own thing at the moment and the restrictions are having varying degrees of success, my answer to your question would be no, but every state needs to maintain things like social distancing.
I think it’s too far gone to contain, but it’s also too dangerous to just let it rip through society unchecked… there’s no solution that everyone is going to be happy with.
Skeptical
You’re supposed to pay interested in both the red line (central estimate), the blue bands (uncertainty intervals) and the blue circles which are data that came after the forecast was first made. Ideally, the red line goes through the blue circles. In terms of statistics, the data are supposed be fall around a blue line that stays inside the shaded blue area.
skeptical,
Given the other parameters in an SEIR, a “slow” (by time) transmission rate means a low “Ro” because the the latency period and time for recovery are not affected by human behavior. All other things being equal (especially rate at which infected die) there are fewer total deaths for an epidemic with a lower Ro. If Ro is less than one, the death count as a fraction of total population is trivially small. It’s not zero, but it’s already small.
.
The entire point of the stay at homes is to get the Ro where high down to low levels. If there are geographic locations where it is already low, there is no reason to “check” in those locations.
.
That, of course, leaves questions practical questions which include figuring out what Ro is in different locations and whether you can communicate an maintain a lockdown of “cities only” while letting those in the countryside roam free. We don’t have good data on the geographic distribution of the Ro. And city populations aren’t known for their tendency to obey orders that apply only to them but not their rural neighbors.
But the general principle that you don’t need to “lock down” people who don’t run into many other people anyway is epidimeologically sound.
skeptical,
“ but it’s also too dangerous to just let it rip through society unchecked… there’s no solution that everyone is going to be happy with.â€
.
There is no ‘solution’ at all, only public policies. Places that have not closed down schools, businesses, restaurants, religious services, etc. (say Sweden) may have somewhat higher infection rates, but they are suffering less disruption economically and socially, never mind loss of personal liberty. The best available data now puts an upper bound on case fatality rate of somewhere between 0.5% and 0.75% and replication rate of somewhere near 3 with no behavioral changes (in most places). If nobody changed their behavior at all, herd immunity likely ends the pandemic when about 2/3 of the population has developed antibodies. So the “let it rip†scenario leads to somewhere near 1.2 to 1.5 million deaths, a 40% to 50% increase over the normal annual death rate for the country.
.
But people have changed behavior, even in places like Sweden, and the level of replication is now lower everywhere. Even a modest reduction in rate of replication will cut the total number of deaths by half or more. The problem is that current policies are just stupid. The risk of death is not uniformly distributed, but rather concentrated in a small fraction of the total population (people with serious existing illness, advanced age, or both). I think sensible policies would focus on reducing the risk to that population, not turning the economy and society as a whole inside out, disrupting the lives of the vast majority of people who are themselves at at very low risk. Current policies are a ‘cure’ which I think far more damaging than the disease.
Lucia,
To me, the red line is the prediction. The blue shaded area doesn’t mean very much because of it’s sheer size. The size is masked on the graph by using a logarithmic scale… but it really is massive. I don’t have the model output, but just by eyeballing that graph of Italy it looks like the shaded area for mid April has a range of 50-1000. That’s like shooting at a barn door while standing 6 feet in front of it… pretty easy to claim victory.
If you’re determining the performance of this model by the shaded area, then yeah it’s a great model.
SteveF
If nobody changed their behavior at all, herd immunity likely ends the pandemic when about 2/3 of the population has developed antibodies.
Actually, it doesn’t end there. That’s when we hit the peak infections. Natural epidemics and pandemics end with an excess of exposed relative to the amount required to stop an epidemic. Running the SEIR with Ro=3 and the ratio of latent time to infective time = 2 means about 94% of people will have been infected by the end of the pandemic. That’s a lot more than 67%.
But you can’t end with fewer than 67% infected.
If Ro =3 and you want to end with the number of remaining susceptible at 1/3rd, you need to institute controls before the peak, hold them for a while and lighten the restrictions as the number of exposed grows to the point where heard immunity is just reached. The exact “best” way to do this depends on priorities other than mere exposure (e.g. peak daily death rates etc.) But you’ll have quite a few more infected by never controlling than by controlling.
In the case above about 27% of the population might have escaped infection by controlling at some point. But you can’t do better than that merely by controlling. If you do it perfectly, afterwards, you’ll always have small breakouts though. Population turn over will affect things. Local pockets where the herd immunity is too low will occur, an infected person will arrive.
Bat will continue to do..whatever.
As we are humans, I imagine some degree of ‘control’ always happens provided individuals have some notion what strategies are effective. Everyone recognizes that some people were self isolating, washing hands and so on. Masks by themselve may have helped knock Ro down a little.
skeptikal (Comment #183640): “I would question if they also have low death counts because infection rates are a useless metric in some areas where testing is low.”
.
What do you mean by “where testing is low”? If there are not many cases, there is not much need for testing. So maybe a good measure is the ratio of negative to positive tests. So lets look at the “irresponsible” states. First column is ratio of negative tests to positive tests, second is ratio of positive tests to deaths.
22 103 Utah
23 47 Wyoming
22 47 North Dakota
06 216 South Dakota
08 46 Nebraska
14 17 Oklahoma
06 41 Iowa
12 58 Arkansas
4.5 19 USA
Looks like plenty of testing in those states.
lucia (Comment #183650)
Ro has to be above 1 in all states… if it was below 1 the virus would eventually disappear all by itself. Even an Ro of 1.5 would cause a lot of deaths, but it would just take longer than in an area with a higher Ro.
skeptical
(1) The model line goes smack dab through the early predicted data (which is to some extent all it can really do because behavior in Italy has also been evolving.)
(2) Comparing models along with uncertainty intervals is what you are supposed to do. It’s standard in all fields. That’s what I used to do with climate models. I don’t see any reason to suddenly ignore them for this case.
(3) The model does over predict daily death rates in the later time period. As I noted: People in Italy may be changing behavior somewhat. (For that matter, doctors might be getting better able to save people.)
These are known additional uncertainties for the model. But it still looks amazingly impressive compared to anything else.
I have no idea how you think models are supposed to be tested. I have no idea what forecasting model you think might be better. I get it: you don’t “like” this one… Because … well. most the features of the model you listed as “flaws” were not features of this model. They were just your wild guesses of what a bad model would do. The predictive features you claimed you wanted were actually in this model.
So the model central estimate overpredicts the number of daily deaths in later period. Yeah. But it stated the uncertainty intervals in the first place. Someone using it would know to consider the likelyhood the deaths would follow any trajectory inside the blue bands. That is the standard way to evaluate a model.
skeptical
Huh?
(1) If Ro is above 1 anywhere right now, it would be growing in that place. Being below 1 in North Dakota would not make the virus die out in New York.
(2) In an SEIR model virus will eventually seem to die out everywhere for any Ro. The question is after how many people have gained immunity. If you assume people gain immunity after they are infected, it will die out after the product Ro*Csuscpetible is less than 1. That’s when 1-1/Ro have been infected. (The time to actually die out will result in more infections, but it will be dying out.)
(3) We know the SEIR model evetually doesn’t apply for infinite times for a variety of reasons (including that there may be other hosts than humans, that actual population is homomgeneous and so on.) So you need another model once time -> infinity.
Bats, cats and pangonlins at least can be infected by this virus. Like Bubonic plague (carried by marmots and prarie dogs) this virus will likely never truly die out.
skeptical,
“ would cause a lot of deathsâ€
.
A few questions:
What is ‘a lot’.
What economic values to society do you assign to the life of a 20 year old, a 75 year old, an 85 year old, a 95 year old? Real questions, not rhetorical.
A “good” model with a large confidence intervals is telling you something very important, that there is great uncertainty in the outcome and depending on the median prediction is a probably a bad idea.
.
Certainly with a fast exponential progression followed by an unknown and regional change in R that creates a peak and subsequent decline the outcomes are going to be highly variable. Given that we are in an unprecedented time with little prior information to look upon, these models are a total crap shoot. They have been mega-wrong (Imperial) and many have missed their confidence intervals by a mile.
.
They have gotten better with their predictions over time, but so have … errr …. climate models as they get updated and miraculously hindcast with great precision to validate their awesomeness to all.
.
The failure to take seriously the prediction envelope occurs 100% of the time with a media interpretation (hockey stick anyone?). Unless … the media is examining something like sea level rise in which case they will then be kindly directed to the maximum prediction of the envelope and dutifully report that as the most likely outcome.
R greater than 1, daily infection counts increase until the infection burns out
R less than 1, daily infection counts decrease until the infection dies
.
Correct? Obviously R is dynamic and regional and will drive itself to less than 1 as it runs out of hosts. If we knew the real infection count we could gauge R pretty accurately and determine which social distancing rules work. The counts we have now are a proxy for that but are flawed in ways that are hard to understand.
skeptikal (Comment #183657): “Ro has to be above 1 in all states… if it was below 1 the virus would eventually disappear all by itself.”
.
That is only true for isolated populations. I don’t think that any state has sealed off its borders.
The liberal media is going all-in on “Georgia is wrong” to open up. Perhaps they will be right, it’s hard to say, but their certainty is misplaced.
.
Why Georgia Isn’t Ready to Reopen, in Charts
https://www.nytimes.com/interactive/2020/04/24/opinion/coronavirus-covid-19-georgia-reopen.html
.
Somehow the economic charts are missing, probably just an oversight. The NYT has always had an anti-south bias that is barely concealed. I’m not even from the south originally and FL isn’t really the “south” with regards to the typical NE prejudice. It’s still very easy to recognize to me that our coastal cultural betters look at this region with disdain. I mean they find it necessary to warn us hicks not to drink or inject bleach. What would we do without them?
Yea SteveF, 1.2 to 1.5 million deaths sounds like a lot. But a lot of those who die would have died anyway within a year. Ferguson says 2/3 would die within a year anyway. So excess mortality would be 400K – 500K. Given that this is probably inevitable unless a miracle vaccine happens in a couple of months, I think governments really overreacted in a dramatic way.
Imperial College’s report has some interesting numbers. For people over 80, expected annual mortality is 9.5% for females and 12.0% for males in the UK. For those 70-79, its 2.8% for females and 4% for males. 60-69 its 0.95 and 1.45. They go on to calculate a number of months the covid risk for those infected equates to in terms of this expected mortality. It varies a lot but ranges between 2 months for some of the young cohorts to 28 months. Looks to me that the median is about 8 months for men and 13 months for women. Basically, even if everyone got infected, they would on average lose between 8 and 13 months of life expectancy.
This is a grim calculus but it is the only rational basis to judge how “serious” excess mortality is and on which to base policy.
The American media is so superficial and partisan and dishonest, they would never in a million years do thoughtful reporting on this Imperial report. They are enemies of truth in a very real sense. The result is that we are going to take a $10 trillion hit and cause a lot of excess mortality from an emotion based policy and possibly create a scenario where there is widespread civil unrest and violence.
David Young
I’m not convinced that “Ferguson says X” should give us any confidence that X is true.
David Young,
Who is Ferguson, and where did he say that? Can you provide a link to the Imperial college study?
He’s the guy from Imperial College who write papers modeling outbreaks and seems unable to make good forecasts.
Why is it necessary to have an army of healthcare workers to do contact tracing? Why can’t people be told to do this themselves? We don’t need an army of social distancing referees.
.
“Staffers interview people who have tested positive to find out who they were in contact with, then hand that information to tracers who call those contacts, said Joia Mukherjee, chief medical officer of Partners in Health. The contact tracers tell the people who were exposed, “You’ve been identified as a contact, how are you, do you have care, do you have the material resources to quarantine safely?†Dr. Mukherjee said. The goal is “to turn off the tap of community transmission.—
This article extracts some of the relevant data from the Imperial Report. If you believe the numbers Ferguson’s 2/3 statement is correct. I expect their IFR numbers are based on worldwide data but I haven’t delved into it precisely. They look to have been corrected for underreporting of infections. But they look consistent with the probably 10 small studies we now have and the New York data.
Lucia, the 2/3 number is not based on modeling so far as I can see.
https://medium.com/wintoncentre/how-much-normal-risk-does-covid-represent-4539118e1196
BTW, this report also strongly supports Ioannidis’ statement of infinitesimally small risk for healthy people < 65 years old. I think pretty much all the experts agree with this statement because its strongly supported by the data which is a very large data set by now.
I'm constantly surprised that people don't seem to be aware of the very strong age dependent IFR rates. A lot of emotion out there among the "alarmist" sector. It's morphed from "we have 20 years to stop emissions" to "we will all die this year unless we do a permanent lockdown." Emotional tripe. It's very obvious from Annan for example.
David,
Which 2/3rd number?
Obviously, somethings Ferguson says aren’t wrong. Some things lots of people say aren’t wrong.
EVeryone is talking about these. I’m surprised you think people don’t seem aware of something they are all agreeing is true.
David Young,
Thanks for the link to that website. I note that the data being used is probably out of date. The projection of 80% infection rate and 510,000 deaths for the UK indicates a CFR of ~1%, while it is likely to be somewhat lower. Other than that, it is a sensible analysis which provides some perspective about the added risk of death the disease presents.
What I mean is that most non experts seem to be unaware of it. or else their thought process is not logical but dominated by their emotions. Certainly its not gotten big coverage in the media. The reason is obvious. It tends to suggest a more calm and rational approach to a managable problem rather than an emotional response to blame Trump or Repubicans.
The idea that letting the epidemic run its course will result in over a million deaths in the USA is unsupported by the facts.
Yes if you take the 14% infected in New York and assume that means that 5 or 6 times that many will eventually be infected, then multiply the 16K deaths in New York by that and extrapolate to the country as a whole, then you get over a million deaths. But there is a lot wrong with that calculation.
First, many people may be resistant. Only 15-20% of those on the Diamond Princess got infected.
Second, when has an epidemic ever infected 70 or 80% of the population in a single wave lasting a few months? Maybe in some besieged city, but I doubt that happens in a wider population.
Third, that ignores age and vulnerability. It appears that those most vulnerable to getting sick are the one most likely to die; that warps estimates of the fatality rate.
Fourth, New York is atypical. Has anywhere else had the same numbers? Maybe northern Italy.
I have seen 4% as the number with antibodies in California. They have had 1500 deaths, so ignoring my points above and extrapolating like in New York gives 50-60K dead there and maybe 500K for the country. But that is an overestimate, for the reasons I just gave.
So maybe we are “saving” a few hundred thousand lives. In reality, we are taking about saving a few months or maybe a year or so in most of those cases. For that, our rulers are asking 300 million to give up a few months or maybe a year or so of their lives.
It is insane.
David
As far as I can tell lay people are very aware of it. That’s why the term ‘boomer remover’ was circulating earlier on.
You seem to think people not sharing your choice to not value the lives of older people is “emotional” as opposed to “logical”. (Otherwise, I have no idea why you are saying people who disagree with your values are “emotional”.Flinging around accusations of emotionality is not really an argument in favor of your position.
Did anybody else see Dr. Birx’s face while Trump the Chump was prescribing Lysol injections as a miracle cure?
Incompetent. Idiotic. Illiterate. This is what our country has come to?
That’s Fuller dropping by to throw some spears because he wants to understand people.
Mike M,
“So maybe we are “saving†a few hundred thousand lives. In reality, we are taking about saving a few months or maybe a year or so in most of those cases. ”
.
The benefit in lives should be expressed in “quality adjusted life years” or “disability adjusted life years” or similar, not just a toal number of deaths avoided. https://www.eufic.org/en/understanding-science/article/measuring-burden-of-disease-the-concept-of-qalys-and-dalys
If you prevent the death of an 85 year old woman who would remain in otherwise perfect health, then the average benefit would be about 7 QALYs. But of course, is is very unlikely that 85 year old woman would be in perfect health, so the average QALYs would be somewhat lower than 7. If you prevent the death of a 20 year old woman who would remain in perfect health until her death, you save 62 QALYs. Of course the QALYs would be somewhat lower, because she is likely to have less than perfect health until death, especially in old age. QALYs are used in nationalized health care systems to evaluate if a medical treatment is cost effective: a QALY is assigned a value… typically more than US$50,000 and less than US$150,000. This is the “value of a human life-year” those health care systems use.
.
All of which help quantify the utter madness of our national (and most state) reactions to COVID 19….. You have to assign an absurdly high value to a life-year to justify the enormous costs already expended, not to mention the even greater costs of economic damage.
Tom Scharf, are you going to follow Trump’s advice or recommend it to others?
Thomas Fuller,
He didn’t say anything like that.
If everyone would stipulate that Trump is a buffoon, would you stop misrepresenting everything he says?
“THE PRESIDENT: Right. And then I see the disinfectant, where it knocks it out in a minute. One minute. And is there a way we can do something like that, by injection inside or almost a cleaning. Because you see it gets in the lungs and it does a tremendous number on the lungs. So it would be interesting to check that. So, that, you’re going to have to use medical doctors with. But it sounds — it sounds interesting to me.
So we’ll see. But the whole concept of the light, the way it kills it in one minute, that’s — that’s pretty powerful.”
https://www.whitehouse.gov/briefings-statements/remarks-president-trump-vice-president-pence-members-coronavirus-task-force-press-briefing-31/
Fuller,
Are you aware of the actual news that is important here, the research on the virus with high temperature, humidity, and sunlight exposure? I can see how you might have missed that part.
Thomas Fuller,
Where is the Lysol in that very confused statement? I’m not seeing it. I am seeing that Trump, like lots of people, doesn’t know much of anything about health care or treatments. He is a buffoon, after all. You, on the other hand willfully misrepresent what Trump says. It is plain dishonesty.
Bryan, the acting undersecretary of science and technology for the Department of Homeland Security, outlined tests in which he said disinfectants like bleach and isopropyl alcohol quickly killed the coronavirus on surfaces. Bryan also spoke about how the virus was found to be negatively affected by exposure to UV rays and higher temperatures.
SteeF, yeah, I’m a bad guy. I had the nerve to quote the sonovagun.
We are led by a moron.
Here’s part of understanding: Everyone knows Trump is a clown and prone to ridiculous ad-libs. It’s not a revelation and repeating it ad naseum doesn’t change the fact that people still prefer his policies and SC selections to what would have been with HRC. Biden repeatedly drooling during yet another senior moment or allegedly assaulting one of his aides 30 years ago is not likely to swing a MeToo feminist progressive to vote Republican. You can probably understand the latter but cannot seem to understand the former.
.
It’s pointless partisan spear throwing.
So stipulated, for the record.
Cancer diagnoses are down by up to 40%:
https://thefederalist.com/2020/04/24/how-many-americans-will-die-because-cancer-screenings-arent-happening/
I am pretty sure that is not because the Wuhan virus cures cancer.
There are about 35K cancer cases a week in the U.S. of whom about 20K will survive. How many of those 20K/week will now die because of a late diagnosis? And those people would have had a lot more years to live than most of the Wuhan victims.
Trump is not a buffoon. He just plays one on TV. And Twitter.
Well Lucia, In the case of pubic officials its obvious they are being emotional. Cuomo said that if he could save 1 life it would all be worth it. That is not really rational. If you look at the commenters at James’ you will see much worse. There is a lot of anger, name calling, and lashing out. Annan even created a post that was critical by cherry picking out of context and lying about my position. This is everywhere in the media too. You can also see it with regard to Georgia opening as someone else pointed out.
As to older people, its a grim calculus but it is demanded. Every year we condemn 50000 people to die because we refuse to lower the speed limit to 10 MPH. Every year a certain number of older people die because of lifestyle choices and we do little about it. The number is probably huge given the epidemic of obesity and metabolic syndrome. We have done something about smoking but boiling it down to a clash of different values is missing the fact that choices always need to be made. Some people and policy makers are vastly better at sorting through all the long term consequences than others. Most people return immediately to their priors and go into full blown OMBS (orange man bad syndrome) if they are in the media. Just listen to the childish emotionalism of MSNBC talking heads.
I’ve seen virtually nothing on the economic consequences of the lockdown or the excess mortality we are guaranteeing by restricting access to the medical system. No screenings means more death later. No prophalactic masectomies for those with high risk factors means more death. Massive unemployment has never been this high and that usually leads to violence, suicide, and substance abuse. This is sometimes mentioned in the Wall Street Journal or some conservative publications but nowhere else that I’ve seen. I could be just missing it but I’ve not seen it. If you have seen it please point me to it.
Thomas,
Trump was rambling as usual. But he clearly was not discussing lysol. He was discussing the disinfectant discussed just above which was UV light. As usual he’s garbling a lot of things. So: who the hell knows what he meant? I don’t. But one thing is for sure He doesn’t remotely say anything about Lysol!
.
If your going to slam Trump at least do go off into the weeds.
Thomas, Here is the fuller quote including the sentence that followed immediately after what you quoted
Mind you…. I don’t think we are going to kill Corona Virus in lungs by shining UV light on lungs. I think it tends to be bad for living tissue. But it’s clear he wasn’t discussing Lysol.
Tom Fuller, Trump talks way too much. Like anyone who does that you will step in it fairly frequently. He is not a moron. He got elected POTUS. He has unified the Republican party more so than at virtually any time in history.
By trivializing your opponent you are just destroying your ability to effectively oppose him. OMBS explains why Democrats are unifying behind a near octogenarian politician with a double digit IQ who doesn’t know what office he is running for and can’t read off a teleprompter. If you grew up, you would be more effective.
David,
I don’t think his saying that means he’s “emotional”. I disagree that shutting down the economy for 1 life is worth it. But I wouldn’t say it’s “emotional”. It is rhetoric– probably hyperbole even for him…. Politicians do that all the time.
In 2017 Cedars-Sinai Med. Center had a system to use UV internally to treat inflammatory disorders. Evidently something fairly new.
See this from Andrew Revkin.
https://twitter.com/Revkin/status/1253696994343821313
I just was an interesting stat. 1/3 of New York City fatalities were residents of nursing homes. That is not surprising given the high prevalence of infections generally in NYC. I wonder if it has any implications for other states. I also heard that NY health authorities required homes to admit patients who had tested positive. Sounds like a potentially lethal policy.
David Young,
“ I also heard that NY health authorities required homes to admit patients who had tested positive.â€
.
If true, it is effectively mass murder. Do you have any idea where that came from?
.
In Florida, the state has resisted public disclosure of nursing homes and assisted living facilities with coronavirus deaths. (Pressure from the owners of those facilities.). I don’t know if the facilities with coronavirus have finally been disclosed, but I will not be surprised if they have not.
Michael Goodman from the New York post has been talking about it.
In Washington state there was an early cluster in one nursing home. I think 27 died. It’s of interest to me because it’s an obvious place to focus testing of employees that I suspect may have not been acted on. My other interest is how many of the most vulnerable have already been infected. It could be a lot higher than the general population prevalence. If so, it might partially explain why some places have plateaued or reached the peak.
David Young,
Sounds like speculation unless a source is given.
.
Apparently the list of elderly care facilities in Florida with corona virus cases was pulled almost immediately after it was posted. The links to ‘the list’ are all dead, and I can’t find it on the department of health page… a link, but no list. IIRC, the WSJ suggested 1/3 of all deaths nationwide have been in elderly care facilities.
.
As best I can figure from news reports, about 20% of all deaths in Florida have been in elderly care facilities. That may not be an accurate number, because some facilities have apparently not counted people who got the illness and then died in a hospital.
Florida Department of Elderly shows total beds in assisted living and nursing facilities at a bit under 200,000, which I find surprisingly low. Based on news reports, about 200 or a bit more have died at those facilities, so probably a little over 0.1%.
I finally found some better information (Miami Herald). 25% of all Florida deaths have been via elderly care facilities. Those facilities house less that 1% of the population. Elderly (over 75) in Florida total ~8% of the population. An elderly care facility is probably not a good place to live if you are old.
David Young,
There was clustering in some sort of home for elderly in WA and it was early on. I think I read it was one of the residents visitors who brought in the virus. In the early phase, many of the first carriers everywhere were people who traveled for business of pleasure. In a residence, that tended to be relatives of the residents; not the fairly low wage caretakers.
Most homes already don’t permit visitors. The importance of not letting outsiders in became apparent after the first reports of breakouts in WA state. My friends can’t visit their parents in assisted living. In most facilities, the communal dining rooms are closed, and people are more or less stuck in their apartments or rooms. They can take walks like all of us, but still, they don’t have spacious abodes. (My mom is very glad she didn’t move to assisted living in November at our urging. I don’t blame her.)
It’s probably wise going forward to test employees coming and going into nursing homes and assisted living. My guess is almost every one of those places probably already are doing things like taking temperatures. I don’t think they can run virus tests daily– we don’t have enough test kits. That is being remedied. But meanwhile, people can only do what is possible.
It’s all well and good to decree that this sort of advice is sage and unemotional. But it’s only sage and unemotional if you aren’t requesting the policiy be limited to doing the impossible like run tests for which no kits are available instead of doing the things that are actually possible.
Maybe. But at least in Illinois the timing of the peak and the shape of it appears to suggest our peak was mostly due to the “stay at home” order. Same with NYC. (The estimated of numbers already affected only makes the case the current flattening is due to the stay at home stronger.)
It could also now be lower because they don’t get out much.
.
But if it is higher in nursing homes, that means we are not so far along in the pandemic. In which case, the reason the mortality rate for the younger looks low is they haven’t been hit yet. That would be a bad thing.
David young/SteveF,
David, we would get more details if you would do a google search and link. While it does appear homes are force to take patients, it is patients discharged from hospitals. They are presumptively thought to no longer be infectious. The presumption idea might be false, but it’s different from just taking in people who are presumptively infectious.
Details here:
https://www.nbcnews.com/news/us-news/coronavirus-spreads-new-york-nursing-home-forced-take-recovering-patients-n1191811
I certainly don’t blame people from deciding to take their loved ones out of nursing homes and give them care at home. (Presumably with the help of a nurse.) I would worry as Steve suggested that admitting the recovered patients is a death sentence as the recovered might be infective. No one has yet proven they are not– but with most diseases we expect those recovered have cleared virus. So it might not be a death sentence.
Details do matter. So if you (David) think you read a dramatic a story, it might be helpful if you (David) would provide a link to the details as you seem to be tending to leave a lot of context and details during the current situation.
Korean have studied a break out associated with a call center. So it’s not just weddings, funerals and nursing homes. (The do suggest this indicates sustained contact is related to spread. So sharing a big cubical area as opposed to being in the same elevator. With luck, we may see fewer offices thinking cubicals are a great thing. 🙂 )
https://wwwnc.cdc.gov/eid/article/26/8/20-1274_article
I did a standard Bayesian update of the IL Covid daily deaths data for the last 4 days from 04/21 through 4/24. I used the posteriors of the original Bayesian results for start to 4/20 as the priors for this run (standard procedure). The good news is that the model posteriors changed very little except for the all important Rt. The bad news is that 50% Rt level increased from 0.65 to 0.94. Those Rt levels are statistically significantly different for p<0.05. The summary of the original and updated analyses is in the link below.
https://www.dropbox.com/s/nhm8vu3km57cglt/Bayesian%20Update%20of%20IL%20Covid%20Daily%20Deaths.pdf?dl=0
https://www.miamiherald.com/news/coronavirus/article242260406.html
SteveF, you know more about Florida than I. How many fatalities in Miami Dade so far?
Florida Health Dept: 295 deaths. Miami-Dade is the hot spot in FL.
https://experience.arcgis.com/experience/96dd742462124fa0b38ddedb9b25e429
.
Miami-Dade says the under count of infections is 16x according to their antibody tests.
“about half of the people who tested positive for antibodies reported no symptoms in the 14-17 days before being tested”
“UM researchers say their findings are more robust than most because they used Florida Power & Light to generate phone numbers in targeted demographic areas, leading to a more randomized selection of participants.”
.
Another issue I read about is the dose. If you have sustained contact with someone then you get a bigger initial dose (viral load?) and this * may * be more likely to give you a severe case. The level of dose with other viruses is a mixed bag with some showing this behavior and some not.
.
If this were the case then it would probably be good news that many people could be exposed to small doses to limit severe cases, the way some immunizations work.
Tom Scharf,
I can’t help but think dose/response matters with this infection since it generally matters for any infection. I would speculate one of the reasons masks reduce infections is that the drag force on the air exhausted from a breath cough of sneeze makes it travel a shorter distance. The “screening” effect creates smaller eddies that tend to allow greater mixing. So, even if filtering isn’t great, it may reduce the radius in which a near by person can intercept an infective dose. And at any radius, the viral load to a person will be smaller.
Of course, some of the virus could actually be absorbed or blocked by the mask material. But it may not need to be to reduce infectivity from the wearer to a target. (Of course, it can also reduce how much spit you receive from someone not wearing a mask! In that case, it’s still on your mask though. A shield would be better for that!
It appears that 287 have died in Miami Dade county as of April 24. Seriologic testing shows 123K to 221K infections. That’s an IFR of .23% to .13%. Some of those already ill will die over the next couple of weeks. An opposite direction effect is that the testing was performed over a period of weeks in the past. One should also account for the demographics of those tested.
But bottom line is that this result is not inconsistent with what was found in Santa Clara and New York. WHO’s IFR’s are off by an order of magnitude. That’s a sign of incompetance. Fouci’s guess was also too high.
Kenneth….
Thanks for running that.
Yeah, there were a few very low death rates. Now daily deaths are mostly flat. I’m not sure that the uncertainty bands are big enough to deal with the reporting “features” we know exist. (Reported deaths may not be on the day reported. There could be a “weekend/easter” features. This seems to be worse in Easter data, but at home I’ve been highlighting “day of the week” when I run and it does seem to exist in ours too.)
I don’t like 0.94!!!!! Of course, just being knocked down from 3ish would still indicate the stay at home is effective at knocking down R.
One thing that is very clear is that Ioannidis has been consistently right about this matter and the alarmists consistently wrong. It’s a shame that Ioannidis was not head of our CDC or some kind of special advisor to the President. Perhaps we could have avoided this phony data driven panic.
Wall Street Journal has a good article about the Hong Kong flu vs covid. Estimated fatalities for Hong Kong flu epidemic were 100K which I’m sure is a very rough estimate. That is in a population perhaps half what the US has now. No shutdowns and I don’t even remember much press attention to it at the time.
https://www.wsj.com/articles/forgotten-pandemic-offers-contrast-to-todays-coronavirus-lockdowns-11587720625
But if R ~ 0.94 that means an approximately constant rate of fatalities and infections. Curve is flattened but still won’t buy enough time for effective medications or vaccines to be found. Herd immunity will still be ultimate outcome and total excess mortality will be the same. I’m guessing that Illinois hospitals are also half empty with massive unused capacity.
BTW, Authorities in Sweden are saying herd immunity will be reached in Stockholm in a couple of weeks. Their curve looks roughly like that in the UK but probably worse than Norway’s.
David,
It’s not a sign of incompetence. It’s a sign of lack of data when the estimates were made. Assumptions were made based on other respiratory illnesses.
.
Incompetence would be if they don’t update now that they are getting decent repeatable data on actual infection levels and deaths. It does look like mortalities/ infection is low because there do seem to be a lot of assymptomatic.
.
I don’t know how common diseases with such a vast range in outcomes happen. We’ve got people who seem to show no symptoms at all and people who are seriously ill, can’t breath, lungs fill with gunk, really horrible looking scans of lungs and so on. My impression is that huge range of outcomes is rare. Admittedly, I don’t have expertise in this area. But I think no one gets a “light asymptomatic case of measles” while other people have raging cases.
David
Sort of. Flat at a low level of 100 deaths a day is different from flat at 1000 deaths a day.
We have this R with pretty much a full lockdown. I was hoping it would be less and that we could get R~0.94 with a lighter lockdown.
I’d like data from a seriology test in Illinois. It would be very useful for guiding our judgement about the worst case number of daily deaths if we end the lock down.
I claim its incompetence because every epidemiologist knows that IFR’s always come down (in some cases dramatically) as case detection gets better. Even Fauci was smart enough to know that 4% was too high. My memory is getting overloaded but I did read something on this in the last 2 weeks looking at the history of the declines in IFR’s for past epidemics. Sorry I can’t provide a link immediately.
Lucia, Well based on the Miami-Dade results which are probably the best yet because they tried to get a random sample, if everyone got infected in the US, we are talking about 390K to 690K fatalities. These will be heavily concentrated in people with very limited life expectancy. Using Ferguson’s number, that translates into excess mortality over a year of 130K to 230K. A large number, but not a disaster either. All cause mortality in the US is 2.8 to 2.9 million per annum. So a 5-10% increase in mortality. One could also take account that herd immunity is reached when 70% are infected, but the Miami Dade number of deaths might go up significantly too. So perhaps 20% excess mortality at the maximum.
You could do a rough calculation for Illinois too.
Our corrupt media will never do this calculation or write articles on Ioannidis’ research. They want this to look really really bad because they are partisan hacks as even Jonathan Karl for the WH press corps acknowledges.
You are probably right Lucia that this disease was harder to get a handle on because almost unbelievable age dependence.
I will say that Ioannidis had it right in early March. This cements him in my mind as a great scientist. Vastly better than the CDC folks, who I would say did a competent job but were vastly too concerned at minimizing mortality at all costs. I am a little surprised that Trump who usually distrusts experts went along without getting some outside input.
Initial viral dose is likely a big factor in illness severity. But that suggests that our policy may be almost exactly wrong. The sort of casual contact prevented by the lockdowns is unlikely to transmit the disease or, if transmission occurs, is likely to result in a mild case. Essentially natural vaccination. But very close and/or contained contact is likely to lead to severe disease. It is not at all clear that we are reducing that.
.
Nursing homes are a good example. There was an outbreak near where I live that resulted in around sixty cases (that was a couple weeks ago, so it might have gone up). About 40% staff, 60% residents. It was reported to have started with a single asymptomatic staff member who was providing close personal care to residents. It is easy to assume that transmission in nursing homes is from resident to resident, but it might be resident to staff to resident.
.
Think about it. The staff are doing things like bathing and dressing the residents. Lots of very close, sustained, and repeated contact. Very easy for a staff member to catch the virus from one resident and then transmit it to others. And the resulting cases are likely to be severe.
.
Another great place for transmission is public transit. Lots of people are still going to work. They are the people that keep society running. They are essential workers, they can’t work from home, and they tend to be lower income and to use transit to get to work. But cities have cut transit service, so the subways and buses are very crowded. Insane.
From NIH statistics for 2017:
Annual deaths from
Heart Disease: 647,457
Cancer: 599,108
Accidents: 169,936
Chronic lower respiratory diseases: 160,201
Stroke (cerebrovascular diseases): 146,383
Alzheimer’s disease: 121,404
Diabetes: 83,564
Influenza and pneumonia: 55,672
Nephritis, nephrotic syndrome, and nephrosis: 50,633
Intentional self-harm (suicide): 47,173
David Young,
My reaction to the statistics you posed is, “so?”
What we still don’t know yet is if these asymptomatic people actually spread the virus or just safely carry it. I have seen no information on that. Given how infectious this appears I would think they do spread it. There are also differences in how much people shed a virus. People coughing and with other symptoms I assume shed more virus into the local area.
.
If your R=0.94 during a lock down, that is very bad news IMO. Any letup will eventually turn into an outbreak. However it’s important to note that outbreaks look more like little hot spots everywhere than a uniform invasion. Probably more like cancer tumors. Oops, terrible analogy, ha ha.
David Young,
The WSJ has an article about Ioannidis, but it’s an opinion piece by a Allysia Finley who is a frequent participant on the The Journal Editorial Report show on Fox News, not a news story.
Mike M,
Sure. If we want to speculate, what we are doing might be just the right thing, or it might be the entirely wrong thing.
Oh. It’s pretty clear we are reducing it. Lots of people aren’t going to their cubicle farm offices. Lots of people aren’t going to weddings, funerals, big religious services, sporting events.
I don’t know why anyone would assume there is a magic barrier preventing resident-> staff transmission or staff->resident transmission. I certainly always assumed there was no such magic barrier. Above, David Young suggested it first got into WA homes by way of a staff member.
I’d think most people would assume all three happen.
Yes. And we have fewer people going to work and using public transit. That is the bad economic consequence you don’t like. But you can’t vociferously lament the horrors of X when discussing economics and later magically pretend X is not happening when later looking at the effect on the reproduction rate of the virus.
.
I would like to see us come out of quarantine. But it has managed to give us time to create more beds, license more medical staff, build out some capacity for ill. Things like ventilators are still in short supply, but we nave enough because we had time to get more of them.
.
We are also hearing promising treatments daily. ALA might work. Rest..(?) may help. Some slowing of the rate of contagion is saving lives. Perhaps the expense is too high, but it is saving them.
FWIW:
I think the reality is our lockdown is going to fray around the edges all through May. I’m betting clients will cautiously phone their hair dressers and arrange sneak meetings. Kids and their parents will make appointments with their piano, flute, voice teachers. People are going to figure out how to get their dog groomed. Some parents will arrage play dates for their kids…. quietly only doing it with other families who want to have play dates.
.
All the economic and social activity will be very inefficient to remain below the radar. But it’s going to happen. As long as death rates stay flattish, people’s willingness to make exceptions for themselves it will increase of time.
.
Karen’s on Nextdoor will report to the neighbors. Cops are going to become reluctant to do much. Maybe they’ll come give someone “a talk” making sure the arrive to late to witness the actual infraction.
.
DeWitt
At this point, 13 have died. Nine was the correct number in March. Not saying he doesn’t have a point, but those numbers need to be multiplied by 2.
Hey, here’s a media article on anitbody studies, ha ha.
.
Experts demolish studies suggesting COVID-19 is no worse than flu
https://arstechnica.com/science/2020/04/experts-demolish-studies-suggesting-covid-19-is-no-worse-than-flu/
.
Sadly the background infection rate has now become a culture war issue. From a perusal of the comments the Twitterverse is apparently worried a lower mortality rate will be used to justify removing social distancing. Therefore these studies must be “demolished”. Sound familiar climate warriors?
.
Do note that the call is for any of these studies to be dismissed in their entirety, not validated. There is also the framing bias of “no worse than flu”.
With regard to the Diamond Princess analysis, those numbers should NOT be multiplied by a factor of 2. Ioannidis’ already multiplied his numbers by a factor of 2 because partly of the possibility of more deaths. I think even the WSJ quoted a number that in his final analysis was multiplied by 2. Not yellow journalism but they should have been asked Ioannidis to review their story. His final central estimate for IFR was 0.25% in reasonably good agreement with Santa Clara, Los Angeles, and Miami Dade.
I would say Lucia that my statistics show that this is all driven by panic and emotion. Even a BOE calculation shows that number of excess deaths will not surpass the tolls we accept every year from Cancer and Heart Disease and may be comparable to accidental fatalities.
My boss has given me authorization to go into work next week if I follow company guidelines for sanitation. Boeing has reopened virtually all its operations outside South Carolina. They have an excellent protocol for contact tracing and isolation.
I think that many like myself are getting fed up and some Red areas even in Washington will open before Governor double digit IQ says its OK. What’s the doofus going to do, call out the National Guard? What will happen is Red states will open relatively quickly, blue states will lag far behind and may go bankrupt as a results. Hey, stupid policy has real consequences.
Tom, Your linked article shows that yellow journalists are liars and political hacks. Its as bad as the late 19th Century. And these low lives have a messiah complex to boot.
Using my traditional lower math approach, it seems to me that every additional day spent on lockdown has a disproportionately high impact on risk of infection, as the larger number of people infected 5-14 days ago age out of virus shedding.
“So a 5-10% increase in mortalityâ€
.
People will quibble about the correct number, but 5% to 10% increase in overall mortality looks about right. The hard question is what is the value of that increase in mortality? Any rational analysis pretty much demands that the madness end ASAP. Of course, we are not talking about rational analysis.
David Young
Well, you didn’t say what point you thought you were making. That that you have, I think those statistics do NOT show this is all driven by panic and emotion.
“studies suggesting COVID-19 is no worse than fluâ€
.
It is clearly worse than flu. The fatality rate among the most susceptible is at least 4 or five times that of flu. The big difference is that there is apparently no existing population that is resistant to infection, so the infection rages across groups and demographics.
Lucia, The question then becomes why is this relatively small increment to excess mortality treated so vastly out of proportion to everyday policy issues such as what the speed limit should be or what we should do about the epidemics of obesity, metabolic syndrome, heart disease, and opioid addiction, gambling addiction, or cannebis addiction. Most of these things involve elements of choice even though they are in some senses diseases too whereas an epidemic involves perhaps somewhat less choice, even though individual precautions could probably have a big effect. However, if you can answer that question I will consider that there is some rational component to what we are doing.
I think part of it is we have become accustomed to activists and yellow journalists who are in constant hysteria about something that may or may not be serious. It’s become part of the DNA of perhaps 30% of the population skewed toward Bernie Bros and Gals. That contagion can sometimes become a majority with the proper anecdotal distortions of ovcercrowded hospitals in Italy of all places (CNN did this at least once in a story about NYC) or reported CFR’s of 10%. It’s all amplified of course by OMBS (orange man bad syndrome) one of whose symptoms is a desire to make everything as catastrophic as possible and blame the orange man.
In your circle of acquaintances you may not see this often and perhaps you read only sane media (there is still some left) but for people like the MSNBC talking heads, its just completely obvious that they are hysterical. Did you see Chris Cuomo’s recent meltdown in public while he was violating quarantine? And his wife’s insane pseudo science remedies? Morning Joe is like a quivering teenager crouched in a corner.
Yes SteveF its worse than the flu but not hugely worse than the Hong Kong flu in 1969. I’m quite curious about how much background immunity there is to flu in the population. Different strains might still have enough similarity that there is partial immunity.
David Young,
“ Different strains might still have enough similarity that there is partial immunity.â€
.
Almost certainly this is true. Multiple studies show varying individual susceptibility to specific strains based on the first strain encountered in youth.
.
Not clear if exposure to other strains of coronavirus does the same.
David Young,
The answer was until recently we have practically no information on mortality rates, reproduction rate of the virus, number of asymptomatic vs symptomatic and so on. We could see a disease that could be very deadly (symptoms worse than flu), that was spreading rapidly with exponentially rising cases and deaths. Exponential rise is– and should be– a scary thing that needs to be considered.
.
As we get knowledge which indicates mortality rates may be much lower and that asymptomatics are a large fraction of the spread, we can start to take more moderate steps. But that doesn’t mean steps taken when we did not know what we know now were asymptomatic.
.
And it also means your posting mortality rates of known diseases against probably lower estimates for Covid we think we know now amounts to making a silly argument. Because on March 20 when Illinois went into stay at home, plausible mortality rates for Covid were much higher than those.
.
Even now, if we don’t do “stay at home”, plausible mortality rates remain high and in the range of many of those diseases. Also, obviously we don’t do “stay at home” for heart disease,Alzheimers or most of those. They are contagious. That’s one of the reasons they don’t suddenly rise exponentially the way a contagion can.
.
So, my polite reaction when I saw you post those was “So”. My less polite reaction was: He thinks he’s making a point. It’s a silly unfounded one. The latter appears to have been corect.
SteveF
There aren’t as many out there. I think we know of 4. MERS ad SARS seem bad enough that I suspect the background immunity from those is not large. But who knows. Maybe it will turn out the be similar to Covid. But I think we might have had a clue about that if it were true.
.
The other one is mild, so it’s possible having gotten that gave some people partial immunity. Or at least it gave their systems a head start on reacting to Covid. Maybe someday we’ll know.
.
While we can speculate that it might have, obviously, it is not something we could count on in February or March. We can’t even go much beyond speculating now. We do know all sorts of things seem to matter. Being vitamin D deficient seems to correlate with getting a bad case.
Jim and I started taking a multivitamin in February specifically to avoid vitamin D deficiency. I read enough to think that was a good precaution with low downside risk. I’m going to need to refill my daily vitamins soon.
There is now a study that found ALA — a common supplement– seems to help with critical cases. I suspect that supplement will now be flying off the shelves. Did I buy a bottle? Yes. (I will not be taking the full dose they give the critical ill. Just the amount people who are into supplements took probably for stupid reasons.) The bottle should arrive from amazon next week. (You can probably find bottles on the shelf at Walmart. Or could two days ago. 😉 )
lucia,
According to the NIAID, there are seven coronaviruses known to infect humans. Four of them cause only mild to moderate disease. There are hundreds of coronaviruses in the field in animals like pigs, camels, bats and cats.
Ahh! Ok. I thought 4 four humans.
I knew there were some that infect animals. I read vitamin D has been found to be useful when treating coronavirus in pigs. 🙂
Well Lucia, I think what you say about how little we knew early on is patently untrue. Ioannidis did his DP analysis in early March. There were other datasets too such as returning expats. In every previous epidemic IFRs declined over time, in some cases precipitously so. Imperial college’s early analysis assumed a IFR of 0.1% but Oxford was saying lower.
We knew from the very beginning that healthy people under 65 were at very minimal risk. Even who data showed this very early. There was absolutely no rational justification for locking down everyone.
What did happen was an over reaction and the best scientists were saying so at the time. The more mediocre ones differed but a lot of this was driven by ignorant non scientists who stupidly thought the 10% top line fatality rates were real. The cdc is absolutely culpable that we didn’t do random testing in February.
The biggest phony data driven panic in American history.
You didn’t contradict my observations about the media and the Democrats (but I repeat myself).
David Young
Ionnides does not multiply the 7 deaths by anything in his article here:
https://www.statnews.com/2020/03/17/a-fiasco-in-the-making-as-the-coronavirus-pandemic-takes-hold-we-are-making-decisions-without-reliable-data/
He wrote that himself. And he quote clearly says he used the case fatality rate of 1%=7/700 and then projected.
.
There have now been 13 deaths. His estimate used 7. The person who says Ionnides used 7 is Ionnides.
.
He then does some projecting onto the age range of the US. But that process has some of the “strike out when pitching to left handed batters in the 7th inning” flavor to it. It might be appropriate or not.
.
But like it or not: HE says he used 7 deaths and he does NOT say he considerd the possiblity that more Diamond princess passengers would die when projecting onto the age range.
David Young (Comment #183805): “What did happen was an over reaction and the best scientists were saying so at the time. The more mediocre ones differed but a lot of this was driven by ignorant non scientists who stupidly thought the 10% top line fatality rates were real.”
.
Indeed. We have burned trillions accomplishing nothing. A far smaller amount put into contact tracing and protecting the most vulnerable would have meant fewer deaths and far less collateral damage.
.
David Young: “The cdc is absolutely culpable that we didn’t do random testing in February.”
.
They were certainly culpable for the testing fiasco, but random testing would have accomplished absolutely nothing. What was needed was targeted testing in support of contact tracing and protecting the most vulnerable.
Even with the greatly expanded testing we have now, I don’t see what random testing would accomplish other than providing data that might inform better decisions.
lucia (Comment #183806),
You are simply wrong.
Ioannidis: “But since this estimate is based on extremely thin data — there were just seven deaths among the 700 infected passengers and crew — the real death rate could stretch from five times lower (0.025%) to five times higher (0.625%). It is also possible that some of the passengers who were infected might die later, and that tourists may have different frequencies of chronic diseases — a risk factor for worse outcomes with SARS-CoV-2 infection — than the general population. Adding these extra sources of uncertainty, reasonable estimates for the case fatality ratio in the general U.S. population vary from 0.05% to 1%.”
Lucia, Please look near the end where he lists a number of reasons why his analysis might underestimate mortality. He then states he is multiplying by a factor of 2. Final estimate is 0.25%.
This is the what Ionnides wrote here
https://www.statnews.com/2020/03/17/a-fiasco-in-the-making-as-the-coronavirus-pandemic-takes-hold-we-are-making-decisions-without-reliable-data/
No. I am not wrong. First: 13/7* 0.0625 = 0.1160714. So multiplying by actual deaths already puts us over his upper bound on uncertainty range. And his claimed expansion also supposedly included other factors. Among these is the fact that people who take cruises are generally not on deaths door and are likely healthier than their cohort of elderly who live in assisted living facilities, are too unwell to travel and so on. So, you need to account for the fact that, being healthy, they were less likely to die.
We don’t know how he partitioned these, but we now know the 13/7 is a certainty not a mere possiblity. And the reasons for expanding beyond that value still remains. So you still need to multiply by 13/7. He did not multiply by 2 to account for deaths. He may have had some small (unknown and unknowable factor to us.) But it was clearly less than 2 because it would have been only some portion of what he did multiply by and he expanded is intervals by an amount less than two.
Claiming he multiplied by two to account for future deaths is simply wrong.
David,
I did. And I quoted it.
Having read it, I can see he does not say 2. If you do the math, he used a value he multiplied by is less than 2. AND, that value he came up with supposedly included not only more deaths, but other factors. So no, he did not account for the actual doubling of deaths after he did that calculation. Not even if you wish he had.
I get you want to believe he did, but to do that, you have to think the additional uncertainties reduced the uncertainty due to the possibility more people would die. I doubt Ionnides thought that.
For crying out loud, lucia.
Ioannidis says: “the real death rate could stretch from five times lower (0.025%) to five times higher (0.625%)”.
Then he says: “it is also possible that some of the passengers who were infected might die later”.
Then he says: “reasonable estimates for the case fatality ratio in the general U.S. population vary from 0.05% to 1%”
Twice as high. Well 1 is not exactly twice 0.625, but to one significant figure, 1 is twice 0.625. And those final numbers are no better than one significant figure. That he gives more than one significant figure for the earlier results is probably just that he is doing what I always taught first year chemistry students to do: Don’t round until the end of the calculation.
What we know now, based on more than the cruise ship data, is the CFR is almost certainly more than 0.2% and almost certainly less than 1%. It is much, much lower for young people, and significantly higher for the very old and the very unwell. It doesn’t really matter what Ioannidis said.
Lucia,
“There is now a study that found ALA — a common supplement– seems to help with critical cases.â€
.
Link or reference?
In the mean time, I am going to inject myself with Lysol, or something stronger if I can find it, since Thomas explained to us that is exactly what Trump wants us to do.
Lucia, I think you are nit picking here. It’s been a month since I read the article. I do recall his central estimate doubled from 0.125% to 0.25% (just from memory). As you point out its not quite a factor of 2 for the upper limit. Multiplying by 13/7 has no justification whatsoever given that he already almost doubled his IFR’s. He’s not specific enough about which factors contributed how much to the estimate perhaps because its an expert judgment.
In any case, this is irrelevant now that we have lots of other results from serologic testing. His final central estimate is within the ranges from those other datasets.
New York seems to be an outlier. My theory is that New York had a much more than 25% of the vulnerable exposed via nursing homes and hospitals and other care facilities which will inflate the IFR. Also New York testing was of a fairly select group of young otherwise healthy people. More careful analysis will be needed which I hope someone in New York is doing. It does appear that fully 1/3 of the fatalities were among nursing home residents.
MikeM
The multiplication is for all sources, not just possible increases in the death. You keep leaving that out. Also 1%/(0.625%)< 2. So
(a) He did not multiply by 2. He multiplied by less than 2.
(b) If you don’t know multiply by at least 13/6, then you are acting as if the other sources of uncertainty were negative.
(c) He doesn’t tell us the portion of extra uncertainty he must have attributed to future deaths.
So: he did not multiply by 2 to account for extra deaths. In fact, whatever he thought that extra uncertainty was at the time, it was a very small fraction of 1.6.
If you want to know use the numbers, you either have to ask him how much the other factors accounted for the spread in uncertainty and how much the extra deaths in, then back out the difference to multiply the value by some number slightly less than 2, or just mulitiply by 2.
Mike–
That you have not read the article in a long time doesn’t make correcting what you said a “nit pick”.
SteveF
Last sentence of article.
https://www.medrxiv.org/content/10.1101/2020.04.15.20066266v1
“Eligible patients were randomly assigned in a 1:1 ratio to receive either ALA (1200 mg/d, intravenous infusion) once daily plus standard care or standard care plus equal volume saline infusion (placebo) for 7 days. ”
Obviously, I’m not going with the intravenous transfusion. But ALA is an common supplement taken by people who like to waste their money on supplements. Ordinarily I don’t take supplements… but I did buy some. 🙂
https://www.amazon.com/NOW-Alpha-Lipoic-Acid-Capsules/dp/B000I1YJQC
SteveF
Of course, if you want to get your ALA from food, you can eat a lot of liver or broccoli etc.
SteveF (Comment #183815): “What we know now, based on more than the cruise ship data, is the CFR is almost certainly more than 0.2% and almost certainly less than 1%.”
.
That would be the CFR so far. It should continue to go down. The more vulnerable have been over represented. In NYC, 23% of cases have been of people over 65, compared to 12% of the population. Given what has happened in nursing homes there, the more vulnerable elderly are probably even more strongly over represented. 18 and under are only 2% of cases but 20% of the population. So *if* a large fraction of the population ends up getting infected, those still to be infected will have a lower CFR than those already infected.
.
My numbers are from:
https://www1.nyc.gov/site/doh/covid/covid-19-data.page#download
and
//www.baruch.cuny.edu/nycdata/population-geography/pop-demography.htm
My percentages may be off a bit due to being too lazy to go get my calculator.
The comments at Medscape on what appears to be a poorly done VA retrospective study shed a good bit of light on Hydroxychloroquine with supplements. Virtually all of the health care providers thought that the study (showing no utility for Hydroxychloroquine was deeply flawed) and in so doing they provide valuable insight into the treatment. Doctors and health providers left their names when they commented on Medscape, which gives their comments some additional credibility in my mind.
Here are some of the comments.
”
“Dr. albert tartini| Nephrology
Absolutely useless study early intervention has been the key to successfully treating Covid 19 with plaquenil. The patients I have used it on early all improved by the 3rd dose. The authors, the NIH guidelines for Covid need themselves to be reviewed.”
************
Dr. David Rogers| Internal Medicine
Doctors all over the world are using it and have deemed it effective when used properly with azithromycin and zinc at the appropriate timing. The problem is you can’t use real time patient treatment results as a research study without IRB approval, patient consent, etc. Sermo did a survey and almost 3000 doctors worldwide deemed this medication regimen to be the most effective regimen we have based on their own experience.
*****************
Prof. Robert Stevenson, FHRS, Senior Scientist, Greatbatch Medical.
The following summarizes what I see emerging as a consensus from the scientist and MD members of the UCLA Gerontology Research Group (UCLA GRG) of which I am a long-term member…..
c.HCQ only works well early on in Covid-19.Once a cytokine storm develops, along with bilateral lung ground glass opacities on CT, HCQ administration can’t be nearly as effective
d.HCQ reduces viral load even more dramatically with Azithromycin on board (Z-Tab).The science as to why this is true still evolving.A few weeks ago, French physicians published graphs on this reduced viral load, this but missed the Zinc.
e.With HCQ on board, serum Zinc levels drop rapidly as Zinc is being transported into cell nuclei.So, with HCQ administration, it is vitally important to monitor Zinc levels and supplement Zinc as needed (caution, way too much Zinc can be dangerous).
f.The best approach is early administration of HCQ plus Azithromycin plus Zinc (HAZ) and also make sure Vitamin D and C levels are in range.Also watch electrolytes….
h.HCQ dose very important.One failed HCQ trial did cause cardiac arrythmias, but the dose was ridiculous; 600 mg twice a day for over one week.As stated above, get HAZ on board quickly as soon as any Covid patient presents with dry cough, temperature above 100, and particularly if short of breath. Add a Z-Tab packet and some Zinc. Now you have HAZ.
i.Physicians I know personally (and who are keeping all their Covid-19 patients out of hospital) are using an initial first day loading dose of HCQ (Plaquenil) 400 mg 2X followed by 200 mg 2X daily for four days.” https://www.medscape.com/viewarticle/929253 [To access must give email, but there is no charge]
Mike M,
“So *if* a large fraction of the population ends up getting infected, those still to be infected will have a lower CFR than those already infected.â€
.
Big if. Yes, the numbers of confirmed cases are skewed toward older people. The question is if all cases are skewed the same way, and if so, why. Seropositive incidence is what you have to look at across a representative cross section of the entire population. Even if it were to turn out a lower rate of people in the youngest cohorts are seropositive, that could just mean they are so resistant they don’t often get enough of an infection to become seropositive. I don’t think you can yet say the CFR will go down from the range it is most likely in based on available data.
Lucia,
I wasn’t impressed by the article. Eight patients? They also failed to describe what normal treatment was (HCQ? Azithromycin? O2?)
.
WRT food sources: Wikipedia says the pills give you thousands of times higher quantities than foods; it is a coenzyme that you synthesize yourself. If you want higher blood levels, the supplement is the only option. BTW, the active form is the (+) enantiomer, the other can’t participate in the enzyme reaction. Since it is very likely only half the synthetic stuff is the active form, you are only getting half as much as you might think.
lucia (Comment #183473)
Yes, exactly… and have a look at how much the model output has changed with Kenneth adding 4 additional days of data. Rt went from 0.65 to 0.94 just by adding 4 extra data points. That kind of change can only occur if the model is extremely sensitive to individual data points. It’s crunching a lot of numbers to produce an output which I would describe as something not much different to a basic poly fit… and the uncertainties are HUGE. In Kenneth’s latest run, eyeballing the shaded area seems to give a range of 1 to 900 by the end of June. That’s so wide you could drive a truck through it.
For some reason you seem to love this model, but I really can’t figure out why. If you add some of those blue circles to the original graph you posted, you’ll find that half of the circles won’t even fall in the shaded area… the model output which produced that original graph was seriously wrong.
SteveF (Comment #183835): “I don’t think you can yet say the CFR will go down from the range it is most likely in based on available data.”
.
I agree *if* the number of cases stays a relatively small portion of the population. But if 70% of the population ends up seropositive, then the distribution will have to be more uniformly spread across people of different ages and health status. And that will lower the CFR.
Oh… I know it’s a small sample. Honestly, to some extent, I think the reason people take supplements is to feel in control of something. It’s as much psychological as anything. I’m pretty sure that’s going on with me buying a bottle. I figure if one pill a day only gets me half what it claims is in there, that’s fine. If it doesn’t do anything… that’s ok too.
I started Vitamin D way back in Feb. There were nohad low vitamin D. His prescription ran out, but his on low edge of normal. So he’s taking it.) I’m sure he won’t be taking the ALA. 🙂
Today I’m going to be making more masks. We’ve been wearing them, but next week it’s required. I need enough to not have a “waiting around” period while they are drying after washing. To be safe, that’s actually 4 masks a person.
Most days, we at most go to the grocery store. But I like to wash after every trip, and every now and then something happens and we need to go somewhere else. We might need them for walks too. I’m not sure.
MikeM,
With the already highish seropositive rates in NYC, I don’t think the idea that older people are exposed at a greater rate is reasonable. I think they just get sick more easily.
I suspect what we might find in nursing homes is close to all/nothing. If one person in the nursing home gets sick, then it spreads easily. Many residents sick; quite a few die. If no one has been sick, then probably no one in that home has been exposed.
I bet there are lots of nursing homes and assisted living facilities with zero deaths. They don’t make the news, but there’s a “dog that did not bark” issue in lack of news. If most or even a large fraction of nursing homes had been hit, we’d be hearing calls to do “something”. What exactly, I don’t know. But we aren’t hearing people talking about everyone taking their old loved one out and moving them into their own living room and so on. Some individuals are doing it– but it’s not widespread talk.
Here is a curious datum relevant to the question of deaths vs, cases for tracking the epidemic.
The New Mexico Department of Health reports outbreaks at 12 “congregate living and acute care facilities”. Here are the number of those facilities in each county and the number of deaths reported in those counties (I don’t have the breakdown by facility):
5 in San Juan County, 33 deaths
4 in Bernalillo County, 33 deaths
1 in McKinley County, 9 deaths
1 in Sandoval County, 11 deaths
1 in Santa Fe County, 0 deaths
Those counties represent just over half the population of the state and about 85 percent of the cases. There have been 7 just deaths in the rest of the state, out of 93 total. The most cases, 708, have been in McKinley County, nearly 1% of the population, but they have a CFR of just 1.3%. Bernalillo is next with 670 cases, followed by San Juan (435), Sandoval (365), Dona Ana (110, 1 death), and Santa Fe (95).
The correlation of deaths with nursing home outbreaks seems better than the correlation of deaths with cases.
SteveF (Comment #183660)
A lot is a lot… it’s a numberless value which indicates a significant quantity, used when you can’t put an actual number on something.
Economic values to society… okay. Let’s look at 20 year olds. Some people leave school and go straight into the workforce while others go on to higher education before entering the workforce…. so let’s say that the ‘average’ age someone enters the workforce is 20. That means that a 20 year old has never produced anything of economic value to society and has never paid any taxes… therefore the economic value to society of a 20 year old is zero. 75 year old, 85 year old, 95 year old… they’ve all contributed economic value to society their entire working lives, both in work output and in taxes paid. I don’t think you can assign a dollar value to that but I would assign a ‘full’ value to them.
When you go to bid granny goodbye, don’t forget to thank her for her generation’s hard work which has allowed you to enjoy the affluent society you live in. Also, be sure to let her know that she’s being sacrificed because a bunch of 20 year olds want to go night-clubbing.
skeptical,
“ I don’t think you can assign a dollar value to that but I would assign a ‘full’ value to them.â€
.
And that is the crux of the disagreement. You refuse to assign an economic value (to society) for a human life, under any circumstances. It is exactly why certain governors refuse even to consider easing restrictions.
.
I note that concepts like quality adjusted life years are routinely used to judge the societal cost/benefit ratio for “life saving†medical procedures. I expect you would object to that calculus as well. I have absolutely not doubt that as a 30 year old I was much more valuable to society than I am today. When I am a minute from death I will be of zero value to society, as is everyone. That does not means friends and family do not feel loss, they do, of course.
.
Thanks for finally clarifying why you do not want the economy opened. I believed you would refuse to assign a societal value to a life, but I am happy you confirmed that. I doubt any further exchanges with you will be constructive.
lucia (Comment #183842),
From what I can find, it looks like about 0.5% of the population is in nursing homes and another 0.3% in assisted living:
https://www.statista.com/statistics/323196/number-of-licensed-nursing-home-beds-in-the-us/
https://www.asccare.com/assisted-living-statistics-a-deeper-dive-into-the-demographics/
If we apply that to NYC, we get a total of a little under 70K in such facilities. Given the reported ~3500 deaths in such places, it seems that 5% of the residents have died (but I might be mixing New York state stats with NYC stats). I have not heard of any nursing homes where all residents got sick, let alone all died. If the CFR among such people is 30%, then perhaps 20% of all residents in NYC senior living facilities have been reported cases.
steveF,
You will be interested in this:
https://www.medrxiv.org/content/10.1101/2020.04.17.20061440v1
It’s about the possibility that common cold viruses result in some immunity for Covid-19.
You’re right in a way, Skeptikal, but that’s not quite the way society is structured. It is the younger generation who pay for the benefits of the older generation (unless they are completely self sufficient, of course). They contributed to their own well being, and that of those who came before, and generally they put future generations into increasing debt.
.
Of course, that’s as simplistic an outlook as your own, but I thought I’d just throw it out there to balance.
.
The real point is that this is not a battle between letting the old die or throwing the young into destitution. A compromise can be found where measures are taken to protect the vulnerable while the young continue to keep the gears of life grinding.
.
Having those least at risk exposed to the virus to develop immunity is probably the fastest, and most effective, way to protect Granny long term, because while transmission vectors exist, so will the threat. This was always supposed to be the plan. Flattening the curve was the name of the game to prevent medical facilities from becoming overwhelmed, not locking people away until no trace of the virus can be found, because obviously, this is unworkable for everyone. My parents are pulling their hair out being stuck at home with nothing to do.
Lucia,
Very interesting, thanks. So 38% of non-covid19 people carry T-cells reactive to one of the two covid19 spike proteins, almost certainly from the common cold covid virus strains.
.
The authors note a prospective study is needed to see if those T-cells are protective against covid19, in part or entirely. Looks like a very important result, since any existing resistant population reduces the size of the susceptible population and the potential total number of deaths. Differences in the frequency of reactive T-cells in different populations might explain some of the surprising differences in rate of spread.
Lucia,
One other interesting point: If a prospective study showed the presence of reactive T-cells is clearly protective, then serotesting of people for the presence of those cells could be used as a screen for identifying people at lower (zero?) or higher risk. That would be hugely helpful.
DaveJR,
I agree, and I have said multiple times on this thread and elsewhere that sensible policy is to protect those most at risk and allow those not at risk to carry on, until either herd immunity is established (perhaps via vaccines), effective treatments are developed, or both. People like skeptical are part of the problem, not part of the solution.
SteveF
Yes.
Also: if we knew for sure it gave immunity or even just thought it was pretty likely, and we could be sure someone hasn’t recently been infected with the covid causing corona, we could expose them to the other virus. Sort of like cowpox vacinnes.
I’ve been staying home. I’ve been wearing masks. I’d risk it and get a cold to avoid Covid 19 or at least get a weak case. I wouldn’t want to have both at once though. 🙁
That said: I’ve already showed myself willing to try things that might not work. I’m taking D and ordered ALA. (Not going to bleach my lungs though. 🙂 )
“.. As mass coronavirus testing expands in prisons, large numbers of inmates are showing no symptoms. In four state prison systems — Arkansas, North Carolina, Ohio and Virginia — 96% of 3,277 inmates who tested positive for the coronavirus were asymptomatic, according to interviews with officials and records reviewed by Reuters. That’s out of 4,693 tests that included results on symptoms…â€
.
https://mobile.reuters.com/article/amp/idUSKCN2270RX?
.
Again showing that we are destroying the US economy without a good reason. It is a total panic of mobs.
Lucia,
“ Not going to bleach my lungs though. 🙂â€
.
You know, they sell those mist generators…. Just load one up with household bleach solution. And while you’re at it swallow some fish tank chemicals and burn yourself with UV light. 😉
DaveJR (Comment #183848)
Dave, the problem is that nobody has found that compromise yet. Even with the current lock down, the vulnerable are still not adequately protected. Have a look at how many nursing homes are taking a direct hit from this virus. Law makers are counting on a vaccine or a cure being found sometime soon… I’m not very optimistic of that happening. I’m also not very confident that herd immunity will ever be reached… nobody knows what percentage of infected people will develop an immunity or how long that immunity will last. This could well end up becoming a constantly circulating virus until a cure is found. Not a very good prospect.
I don’t want to be in lock down any longer than I have to be, but I also understand the risk of opening up society too soon and ending up with the mass deaths that we’ve been trying so hard to avoid.
Ed Forbes,
Prison populations tend to be younger than the whole of the country.
Ed,
On the immunity…. I tutor kids. I was exposed to sniffing teenagers last fall. I got a cold. then Jim got a cold. With luck… they protected me.
.
Honestly, because of some in person tutoring, I do get exposed to colds a lot. I probably wouldn’t have if I didn’t tutor. Maybe this will turn out to be lucky. 🙂
Skeptical (and Steve),
The economic value we assign to a life (as in, how much economic damage are we willing to incur to save a life) is not necessarily tied to the economic contribution of that life to society. It seemed to me Skeptikal’s argument was based on this premise, for example:
The economic value to society does not necessarily determine the economic value of an individual life. Still, this does not mean we should value an individual life infinitely more than the economy.
Basically, I’m objecting to smuggling in the concept that economic contribution to society is the standard by which we should judge how much we are willing to spend protecting individual lives. This isn’t necessarily so (and I don’t think it’s so at all, actually).
It is sort of interesting that most reports will highlight how this means we have a severe under count of cases but not also note that if this is true, (a) the case fatality rate is proportionately lower and (b) most people who get it will ever notice.
.
Serology tests are looking more and more important at this point. If it’s 20% affected in NYC, we aren’t done with the contageon by any means, and that’s not the reason for the flattening in cases. BUT we are a hell of a lot of closer to the end than if only 2% have been affected. That also means the peak in symptomatic cases won’t be as high as predicted by an SEIR model that assumes the “infectious” is more or less comparable in size to the “visible cases”.
.
Knowing how many have been affected is a BIG chunk of utility in estimating the progress of a disease within a simple SEIR!
SteveF,
6,200 new cases in Brazil yesterday. Only the US had more. It looks like it may be cranking up in Russia too.
Steve: “ Prison populations tend to be younger than the whole of the country.â€
.
Not all systems are younger
.
“They started with the Marion Correctional Institution, which houses 2,500 prisoners in north central Ohio, many of them older with pre-existing health conditions. After testing 2,300 inmates for the coronavirus, they were shocked. Of the 2,028 who tested positive, close to 95% had no symptoms.“
SteveF, although prisoners on average used to be younger than the gen pop, mandatory sentencing and three strikes provisions mean that there are enough older prisoners to drag up the mean and median to something roughly equal to the rest of the country.
Given some of the comments here, I am surprised we are not advancing the solution portrayed cinematically by Charlton Heston and Edward G. Robinson. Love me some Beethoven’s Sixth…
Lucia,
If you had, say, 38% of people already resistant due to a previous infection with a common corona virus infection, and 22% who already had the covid19 infection (mostly asymptomatic), then there are 60% already resistant. The really weird thing is the very odd profile of severity. I mean, the vast majority of people who get covid 19 either asymptomatic, or with symptoms so mild they hardly notice, yet among many of the very old (and especially very old and very unwell) the virus has a high case fatality rate compared to the common flu.
.
I wonder too if those with existing reactive T-cell populations are in part responsible for the asymtomatic cases…. especially asympotmatic cases among older people… and there are plenty of these. Much to learn.
Thomas,
Well… that’s cryptic.
Notwithstanding your apparent “surprise”, evidently no one has advocated that horrible thing which it, evidently, surprises you to observe they are not advocating. So the fact that you are “surprise” merely suggests that your interpretation of what people are saying is likely out of whack. That doesn’t reflect badly on others. You can decide if, perhaps, it reflects badly on your ability to interpret or understand what others are saying.
.
If there are specific statements you object to, you might consider actually stating them so we can discuss the merits of your view. That you love Beethoven’s Sixth is, perhaps, interesting, but I don’t plan on discussing the relative merits of Beethoven’s various pieces, nor Beethoven vs. other people.
Lucia, The factor of 2 horse is dead and we should stop beating it. The bottom line is that Ioannidis’ study was well done and was prescient in that later larger data sets are giving very similar IFR’s. My bottom line is still that the best scientists knew in early March that corona virus was more deadly than flu but much less deadly than the top line statistics were showing (by a factor of 10-50).
We should also bear in mind that most experts said that flattening the curve was the goal of mitigation. It was never claimed that total fatalities would be reduced (so long as the health care system was not overwhelmed). Perhaps 200K excess deaths were inevitable at the time China allowed international flights out of Wuhan while lying about human to human transmission. There is no point in letting our empathy with victims blind us to this reality. The only adult question is what is the best path to minimize total harm, both human and economic harm. It’s pretty hard to argue that our policy choices are anywhere near to that minimum. My observation is that the chorus of “we are all going to die” is vastly loader than the adult debate.
SteveF
Yes. That strikes me as weird. I admit lack of expertise on immunology. But it seems like the serology tests that show high numbers of affected would suggest that people with nearly no symptoms were “infected” in the sense their immune system’s responded. It seems they also likely became infectious. This would be in contrast to being exposed and the body reacted as if nothing ever happened.
.
I have no idea what is normal. So, for example: If someone had cowpox, we know they didn’t get symptoms of smallpox. That’s good enough for Jenner’s time. But I can’t help but wonder if those who had cowpox also developed antibodies specific to smallpox without getting symptoms? And perhaps, without ever being infectious (perhaps because while their body made antibodies, their viral load of small pox did stay really small?)
.
I can’t help but think some virologists and epidemiologists know the answer to about some other diseases, but I don’t! And maybe they don’t because antibody tests and virus tests based on DNA are newish. That means some of these studies couldn’t be done back when smallpox was still a scourge.
Aw, Lucia, it isn’t that cryptic. In Soylent Green the antidone to aging was, well, dying in a booth with an appropriate music video playing as you were euthanized. Edward G. Robinson died in such a booth with Beethoven’s Sixth playing as background music to bucolic images flashing before him.
After which he was mulched into Soylent Green… sigh… how quickly we forget.
Thomas Fuller,
” although prisoners on average used to be younger than the gen pop, mandatory sentencing and three strikes provisions mean that there are enough older prisoners to drag up the mean and median to something roughly equal to the rest of the country.”
.
Always grinding the political spear points I guess.
.
That suggestion is misleading. There are no young kids in prison, so to compare the age means and medians for the whole non-incarcerated population with the incarcerated population is misleading. The comparison that is reasonable is adult populations only. Here are the current Federal prison numbers:
Ages 18-21 1,866 1.1%
Ages 22-25 8,366 4.9%
Ages 26-30 21,931 12.8%
Ages 31-35 28,182 16.4%
Ages 36-40 31,206 18.2%
Ages 41-45 26,545 15.5%
Ages 46-50 19,979 11.6%
Ages 51-55 13,749 8.0%
Ages 56-60 9,475 5.5%
Ages 61-65 5,467 3.2%
Over 65 4,746 2.8%
Both mean and median look like ~38-39. State percentages may be slightly different, but not a lot. The total US population, excluding those younger than 20, has a median somewhere near 50 and a mean a bit higher. The numbers for greater than 60: prison 6%, general population ~22%. The people truly at high risk for covid19 (>65 years old) represent 2.8% of the prison population but 16% of the general population.
.
Whether mandatory sentences and three-strike laws make sense or not has nothing to do with covid19.
Thomas, it’s great that you read the first sentence of Lucia’s response. The rest of her response is even better. Read it again!
I’m not saying anything about whether those policies are appropriate. I’m saying the mean and median ages of the prison population have moved steadily upwards to being closer to that of the general population.
David
You could do that by just stopping. 🙂
First: It was always claimed the goal of flattening the curve was to save lives. That was the point of putting horizontal lines indicating hospital capacity on the curve. Those were always on the graphse with people discussing that the cases above “capacity” curve were much more likely to die from lack of treatment.
.
Saving lives of the infected means they were absolutely positively claiming the goal was to save total lives. If you didn’t understand this at the time and still don’t, well… that’s on you. It was clear then and still is now. That this was communicated and done so clearly doesn’t change even if, perhaps, it was not the absolute optimum path, or you prefer some other path (which may also not be optimum.)
.
Second: That flattening the curve also reduces the total number infected under a simple SEIR model. So it does reduce total death. This is a simple truth. Even if someone, somewhere didn’t make this more subtle point when communicating one advantage of flattening the curve, that doesn’t mean they aren’t allowed to point it out now. It also doesn’t make pointing out this simple truth “emotional”.
Grousing about people observing this suggests you are the emotional one who doesn’t want people to point out true facts if they happen to not support your preferred policy.
Honestly, as far as you accusing others of emotionality: Pot. Kettle. Black.
I see. You now want to change the subject to criticizing China. I know you might be frustrated and feeling emotional about their behavior. So you want to go off on a bender blaming them. And indeed, there is some blame to be apportioned there.
But what China did way back in January is irrelevant to the issue of whether our flattening our curves was
(a) advocated to save lives in addition to not overburdening hospitals. (It was as I noted above.) and
(b) whether it does save lives. It would under a simple SEIR model. You can run one and see.
.
.
Sure. And yet you are going off on emotional benders to slam the Chinese rather than sticking to discussing minimizes harm.
You are also denying reality about what people advocated about flattening the curve.
So once again: given your tendency to go off on irrelevant rants again china… with regard to your accusation that those who disagree with you are “emotional”: Pot. Kettle. Black.
JD Ohio,
I think you should have posted the complete comment from Professor Stevenson. I’m glad I went to Medscape to see it for myself. Here are the missing sections:
So make sure you’re getting enough zinc in your diet. For us old folks, getting serum zinc level checked would probably be a good idea.
DeWitt
My multivitamine gets me 53% of the RDA of zinc. Perhaps not enough… but I honestly think it’s a good time to take a multivitamin. Ordinarily, the main objection to doing so is wasting money and creating expensive urine. But there are enough people finding some deficiencies that may make a difference that it seems like good insurance now.
.
It also avoids the danger of self-prescribing megadoses and/or missing something.
.
Oddly, i suspect many peoples diets have improved during the stay at home. Ther’s more home cooking. There’s less temptation to buy Reeses peanut butter cups at the train station on the way to work.
Sure, some of the home cooking might not be all the low fat or low carb or whatever some advocate. But generally speaking, people eat less cheese cake, tiramisu, chocolate brownie sundaes and so on when they cook themselves. OTOH: I am making cinnamon rolls ever evening.
Lucia,
“But I can’t help but wonder if those who had cowpox also developed antibodies specific to smallpox without getting symptoms? And perhaps, without ever being infectious (perhaps because while their body made antibodies, their viral load of small pox did stay really small?)”
.
Good questions. I do not know the answers, and maybe nobody knows. Modern serology testing will no doubt answer WRT to corona viruses.
.
Smallpox and cowpox are closely related viruses in the same virus family. (Like the corona viruses.) They are the same size and shape (I have characterized the size distribution for vaccinia particles many times…. they are quite big for virus particles), and almost identical in structure. So it is not very surprising that there is cross immunity. Interesting also the dose-severity angle: before vaccinia became widely used, people dosed themselves (and their kids!) with relatively small numbers of smallpox particles to protect against severe illness. There was a fatality rate of a few % even with low doses of the smallpox virus, but that was deemed at the time a prudent choice, since fatality was over 30% with a full blown case…. plus permanent disfigurement and/or blindness for the survivors.
SteveF,
People used to have measles parties. I don’t know what the protocol was. I went to one. . .
As I mentioned, people are going to be twarting social distancing. This is evidently a party in Chicago recently
https://www.tmz.com/2020/04/26/1-000-people-attend-crowded-house-party-in-chicago/?fbclid=IwAR1H9A8vYGjcCO5TltNMXzzYaE_T7HVc_rriHS_limEg_hdbghtLuqJV-jQ
lucia,
I remember chickenpox ‘parties’. Not so sure about measles. They used to quarantine you for measles.
Thomas Fuller,
“Given some of the comments here, I am surprised we are not advancing the solution portrayed cinematically by Charlton Heston and Edward G. Robinson.”
.
Wow. Maybe you could point to those comments and explain how those comments lead you to expect advocacy of euthanasia, and subsequent cannibalism. I always can count on progressives to accuse their political opponents of horrible things, based on zero evidence; I guess the ends always justify the means.
Dewitt…
I was in El Salvador when I had measles. My first communion was post-poned. I don’t know if we were quarantined, but perhaps half the point was to stock the larder?
.
My mom injected us with gamma gobulin in preparation. It was supposed to boos immune response.
https://en.wikipedia.org/wiki/Gamma_globulin
Thomas,
It is cryptic. Because there is no way for the reader to guess whose comments or which make you think someone would suggest that. In anycase no one has.
Your “surprise” doesn’t mean anything about what your intended targets might think, do or advocate. It only says something about you.
Thomas Fuller,
Based on the extreme comments made by many liberal lock down advocates, I am surprised there are not calls for those in violation of the orders to be simply executed by beheading on the spot for the attempted murder of elderly people. After all, anyone who thinks the rules are plainly unconstitutional needs to be made an example of.
I think in another context the phrase ‘The constitution is not a suicide pact’ was used frequently. The U.S. has had experience with quarantines, martial law and other arbitrary measures infringing on freedom of movement, expression and interaction in the past without permanent damage. Some of those measures were grossly inappropriate, e.g. internment of Americans with Japanese ancestry during WWII. But we learned from that, made amends as best (?) we could and moved on.
If lockdowns prove to have been an overabundance of caution I am sure we will learn from it, make amends as best we can and move on. Unless you have been eavesdropping on our evile Democratic bat channel and heard about our secret plan (Thank you Bill Gates and George Soros) to make the lockdowns permanent.
x I think in another context the phrase ‘The constitution is not a suicide pact’ was used frequently. The U.S. has had experience with quarantines, martial law and other arbitrary measures infringing on freedom of movement, expression and interaction in the past without permanent damage. Some of those measures were grossly inappropriate, e.g. internment of Americans with Japanese ancestry during WWII. But we learned from that, made amends as best (?) we could and moved on.
If lockdowns prove to have been an overabundance of caution I am sure we will learn from it, make amends as best we can and move on. Unless you have been eavesdropping on our evile Democratic bat channel and heard about our secret plan (Thank you Bill Gates and George Soros) to make the lockdowns permanent.
Lucia, You are descending into writing long comments that take things I’ve said (omitting critical qualifiers) and then saying something different that may also be true as if it contradicts what I am saying. You are quoting out of context which strikes me as meaning you are losing the argument.
My statement about flattening the curve is true. Yours is also true. Is this worth more than those 12 words? No its not.
I mentioned China very tangentially and what I said was true. You are reading my mind when you say I was changing the subject. Are you feeling well today?
I’m not sure the age profile is so “odd”. I saw something in the last month comparing to the flu and it didn’t look that different from the flu. I can try to find it again.
Old folks who are dying slowly have always been quite likely to get an infection that finally kills them or contributes to their death. That’s why pneumonia is so common as a cause of death in the elderly. My father in law got a mersa infection that lasted a year and a half. Ultimately he died from liver cancer that may have been caused by the strong antibiotics he had been taking. But no one will ever know what factors contributed how much to his death.
Tom Fuller,
I forgot to add: their heads should be placed on pikes outside the Governor’s mansion as well.
I don’t know how, but my comments are now in lockdown. I’ll put masks on the next ones. Maybe spray them with disinfectant. Maybe shine a light on them…
From Hullabaloo: “Dr. David Katz, a public health expert at Yale, tells Friedman health professionals he knows want this balanced approach to addressing the COVID-19 pandemic. Economic costs have human impacts too. But we must be smart about it. “The moment you stop respecting this virus, it will kill someone you love,†Katz says.
That recalls what characters say in movies about space. Coincidentally, I rewatched Ridley Scott’s The Martian last night. Safely back on Earth, Mark Whatney (Matt Damon) tells a class of astronaut candidates:
It’s space. It’s filled with chance, circumstance, and bad luck. It doesn’t cooperate. At some point, I promise, at some point every single thing is gonna go south on you, and you’ll think: this is it. This is how I end. (then) And you can either accept that… or you
can get to work.
You begin. Solve one problem. Then the next. And the next. Solve enough and you survive.
Yeah. That’s what this time feels like. Except we are not solving enough problems and more keep cropping up. The biggest impediment is the problem in the Oval Office whom cowards in his party refuse to solve. They are desperate to retain power, save the economy … oh, and save lives. Their boss cannot focus on one problem long enough to solve it and move on to the next. Neither can they. Not to mention their priorities are screwed up because their ethics are.
In a deleted scene, astronaut Watney ponders the hundreds of millions of dollars it cost to save his “sorry assâ€:
They saved me because that’s what we do. Every human being has a basic instinct to help each other out. If a hiker gets lost in the mountains, people coordinate a search. If an earthquake levels the city, people all over the world send emergency supplies. This instinct is found in every culture without exception. And because of it, I had an entire planet on my side.”
I think in another context the phrase ‘The constitution is not a suicide pact’ was used frequently. The U.S. has had experience with quarantines, martial law and other arbitrary measures infringing on freedom of movement, expression and interaction in the past without permanent damage. Some of those measures were grossly inappropriate, e.g. internment of Americans with Japanese ancestry during WWII. But we learned from that, made amends as best (?) we could and moved on.
If lockdowns prove to have been an overabundance of caution I am sure we will learn from it, make amends as best we can and move on. Unless you have been eavesdropping on our evile Democratic bat channel and heard about our secret plan (Thank you Bill Gates and George Soros) to make the lockdowns permanent.
yyy
Thomas Fuller,
“Dr. David Katz, a public health expert at Yale, tells Friedman health professionals he knows want this balanced approach to addressing the COVID-19 pandemic. Economic costs have human impacts too. But we must be smart about it. “The moment you stop respecting this virus, it will kill someone you love,â€
.
Yale? Please.
.
A public health expert from Yale is no more capable of balancing societal costs and benefits than anyone else is. The moment you stop recognizing that the cushy lifestyle of the folks at Yale require a functioning economy, that cushy lifestyle is put in danger.
I have linked the 4/25/2020 New York state Annan Bayesian analysis results. I did not update the original one using the posteriors for that run because the daily death data for New York state used in the original run had changed. Although the total deaths from New York have been considerably larger than Illinois, New York is in a better place for Rt with a value of 0.48 versus 0.94 for Illinois. Any conjectures on why that is the case. Deaths per million population for New York currently is 851 and for Illinois is 149.
https://www.dropbox.com/s/bwzcttcga8deuzq/NYState_Bayesian_Covid_2_19_2020_4_25_2020.pdf?dl=0
David,
Sure. And you probably sprinkle in things like in “yellow journalists” and “bottom line culprits here” tangentially, or failing that perhaps you use that sort of invective in cool dispassionate unemotional ways.
Thomas. Sorry, I’ve released the comments.
David Young (Comment #183889): “I’m not sure the age profile is so “oddâ€. I saw something in the last month comparing to the flu and it didn’t look that different from the flu.”
.
Compared to the flu, Wuhan is much less bad for the very young and quite a bit worse for the very old.
I suspect that influenza would be much worse than Wuhan in a population that had never experienced it. That might be why the flu can be deadly in the very young: their immune systems are encountering it for the first time. As you age, your immune system gets better at fighting off the flu due to repeat exposure, so that most cases are asymptomatic or minimally symptomatic. Then even an elderly person with a declining immune system can usually fight it off.
.
David Young: “Old folks who are dying slowly have always been quite likely to get an infection that finally kills them or contributes to their death.”
.
I wonder how many deaths every year would be due to “whatever is going around” if the same criteria were used as with Wuhan. Not as many as Wuhan is causing, but probably not a negligible number either.
Yellow journalism is a well used phrase usually used to describe Gilded Age “journalism.” I think I first saw it in Samuel Eliot Morrison great history but I’m not sure. I’m using it as a descriptive phrase. If you like the Washington Post or the New York Times that’s fine with me, but even prominent people within the profession (Jonathan Carl) are saying that they are behaving like an opposition party. In the 19th century malefactors of great wealth owned most newspapers and they were frankly partisan, often nastily so. Everyone knew it and it was at least honest. The difference today and why the media is corrupt in my view is that they still maintain the fiction that they are “objective” and “non-partisan.” The Kavanaugh witch hunt (an accurate description) proved they are either disconnected from reality or lying, or maybe a little bit of both.
Using strong language does not imply lack of truthfulness or accuracy.
BTW, I perhaps didn’t make clear earlier that I was accusing the press and politicians of reacting emotionally. That was not about you.
Thanks Mike M. for the interpretations. If you have a reference where I could read up that would be great.
It took Italy about a month to half their deaths per day. They turned the corner but it’s a pretty slow draw down, slower than people want to see no doubt. The UK has been almost flat for 3 weeks. The virus is “finding a way”, even during lock down.
.
This is increasingly looking like a choice between * many * more months of lock down or letting the virus burn through the population. It’s unclear whether we can find a way to run the economy and keep contact rates to being manageable.
.
Over the next month we will see how infection rates change with different re-open strategies. Although we will be treated to “new spike!, new spike!!, new spike!!!” on a daily basis we won’t be able to gauge actual rates for weeks.
.
The virus has unfortunately left us with no easy way to isolate it, too many asymptomatic carriers, and a lot of virus shedders prior to symptoms. It almost seems bio-engineered in this sense (I’m not spreading conspiracy theories, just commenting on a very effective design even if by happenstance).
.
Very tough decisions ahead I suspect. Let’s hope there is some seasonality.
Thomas Fuller,
” Unless you have been eavesdropping on our evile Democratic bat channel and heard about our secret plan”.
.
Nah. The foolish plans are all quite out in the open…. you know, with comments like a Governor refusing to put an economic value on a human life, and another Governor who claims protecting the Constitution he is sworn to uphold is “above my pay grade”. Talk about economic suicide pacts.
.
Yes, the draconian measures will be seen as a needed starting point, which Democrats then refused to end, even when their immense cost relative to their modest benefits became obvious.
Kenneth,
I hope the model is right, but today’s deaths, 367, are already over the projection and the deaths for 4/25 are close to the upper boundary of the error range.
Question: The cumulative deaths through 4/30 are the same as through 5/31. Do the deaths through 5/31 include deaths through 4/30 or are they additional?
Ken, Interesting plot. I assume that a lot more of the vulnerable have been infected in New York than Illinois just because there are much higher case numbers in New York. I’d still like to see a really good statistician do an analysis of infection rates by age group and fatality rates by age. That would give us a much better idea of what is going on. Just confirms for me that this kind of model is a very blunt instrument to use to predict the course of any epidemic. Particularly for this virus, taking account of the age structure of the population would seem very important to me. Does NYC have more elderly than Chicago? Don’t know.
Kenneth,
“Any conjectures on why that is the case.”
.
Maybe the combination of people who may have some T-cell based immunity from earlier infections with normal corona viruses (34% in Lucia’s link up thread), combined with people who have developed antibodies do to covid19 via asymptomatic infection (~14% statewide, ~22% NYC area), plus those who survived symptomatic infections (~1.5% of total population), is enough to drive down Rt in NY. I don’t know what the consequence of having a potential pool of only 43% of the population would do to your model, but I have to believe it would be significant.
DeWitt, I notice the same numbers from the plot as you did. I suppose the area under the curve would tell us. I was eyeballing it and then had to mow my grass before it rains tomorrow. I’ll have look after my martini and dinner.
Looks like the model estimate of 60,000 deaths in US will have to be revised upward.
MikeN,
Sure. Looking at the numbers, it is hard for me to see total US Covid19 deaths by year end being a much less than 150,000. That could change if a very effective treatment is found, of course.
The Democrats will end lock downs because it is not possible to continue with them indefinitely. If both sides operated in alternate universes the release from lock downs would probably occur within a couple weeks of each other. The bigger forces at play make it inevitable.
.
The very tiring palace intrigue aspect is that the democrats are hyper focused about appearing humane and displaying their virtue while performing the same behavior. The right has little patience for this show. Given these priorities the left hopes they can be “forced” by the right to perform the same behavior they would have done on their own anyway, thus deflecting blame for killing grandma.
.
Don’t believe it? I put up as evidence 2009 in which the left had full control of the government and a progressive paradise failed to materialize because the structural constraints of the economy and other factors are dominate over alleged good intentions.
.
The left will open the economy in their states, because they have to. They will do so with a great and wondrous empathy on full display. The end effects will be the same. A state’s treasury cannot cash empathy checks. The militant do-gooders will find a way to blame Orange Man for decisions made entirely by their blue governments anyway. Yawn.
SteveF,
Social distancing and flattening the curve does not create herd immunity. We’ll see another peak in the fall, if not earlier, maybe not in NYC, though. It would be miraculous if there were an effective vaccine before the end of the year. I have serious doubts that there will ever be an effective vaccine. Treatments, OTOH, are a distinct possibility. It would be amusing to see the reaction if a treatment using HCQ was found to be effective. If that happened, Trump would probably be re-elected by a landslide. I suspect that factors into the resistance to using it.
David Young, the media treatment of Kavanaugh can be usefully compared to that of Biden regarding Tara Reade. In the Reade case, however, there seems to be actual evidence an incident took place ie her mother called in to Larry King shortly afterwards after Reade told her about it.
Tom Scharf (Comment #183901): “This is increasingly looking like a choice between * many * more months of lock down or letting the virus burn through the population. It’s unclear whether we can find a way to run the economy and keep contact rates to being manageable.”
.
There is a middle ground: Gradually backing off.
.
Two reasons were given for the lock downs: Crushing the virus and keeping the health care system from overloading. The first has long since ceased to be viable and overloading the system has not even come close to happening, except maybe in New York. And even there they never needed the hospital ship or the emergency hospitals in Central Park and the Javits Center.
.
So let people get their cancer screenings and hip replacements and dental exams. Let people decide if they want to reopen their businesses, or go to work in those businesses, or patronize those businesses. Make it so that people have the option of staying out of work for a while and make very sure that people don’t go to work sick but still get paid. Encourage people to keep working from home, if possible. Keep emphasizing the importance of hygiene.
.
Make sure that testing is readily available and that positive tests are followed up with contact tracing and quarantine. Work hard to protect the most vulnerable in nursing homes.
.
People can be allowed to act responsibly while still keeping this under control, at least outside of the deepest blue states. That is demonstrated by Sweden, the states that never imposed lock downs, and the fact that transmission rates started dropping well before the lock downs.
.
So let’s bet on that. If new cases start to threaten the capacity of the health care system, it might be necessary to tighten things up. But I doubt that will be a problem, at least not in most of the country.
MikeN (Comment #183907): “Looks like the model estimate of 60,000 deaths in US will have to be revised upward.”
.
Why?
DeWitt
Theoretically it could. You’d just be slowly reduce the contraints of “stay at home”. It would take a long, long time though.
.
We’ve now built up hospital capacity in Illinois. I won’t like seeing an increase in daily cases … but we actually have more beds, ventillators etc than we had a month ago. So we could deal with it.
.
Hospitals probably can’t take a “let her rip” rate of cases yet though.
.
There does seem to be progress in treatments… all tentative. Early treatments given to people who are diagnosed early would be best!
DeWitt,
“I have serious doubts that there will ever be an effective vaccine.”
.
Maybe, but if the presence of reactive T-cells from other corona virus infections turns out to be protective for covid19, that will be a very good sign an effective vaccine can be developed.
.
Mike M.
“Why?”
Are you joking?
SteveF (Comment #183916): “Are you joking?”
.
No.
But it appears that my mind reading skills are severely deficient.
On the WSJ web site here:
There’s a lot to say in the article about Sweden, but if I copy that, I will have copied almost the entire article. So I’ll paraphrase. Sweden only asked people over 65 to shelter in place. They didn’t close elementary and middle schools. Even the Volvo plant is back in business. While the death rate is higher than in the other Scandinavian countries, it’s lower than much of Europe. Since Sweden’s containment measures are less onerous than most, they can be maintained longer if they work. We should hope they do.
I saw where Tom Scharf noted that Italy’s daily deaths were declining at a slow rate and I thus ran the current Italy daily deaths with Annan’s Bayesian model and went back to Annan’s original runs to compare Ro and Rt values:
3/29/2020 Ro= 2.81+/-0.43; Rt=0.81+/-0.35
3/12/2020 Ro=2.93+/-0.28; Rt=0.90+/-0.08
4/25/2020 (my run) R0=2.72+/-0.50; 1.14+/-0.08
The results of my run are in the dropbox link below and the link to Annan’s early runs is below that link. Obviously it is of interest to see how new data changes the all important model projection for the Rt value. Tomorrow I will look at the UK where Tom noted a leveling of the daily deaths. Annan’s early runs for UK showed a low Rt.
https://www.dropbox.com/s/lul0ika1xpt8ug2/Italy_Daily_Deaths_2_15_to_4_25_Bayesian.pdf?dl=0
https://bskiesresearch.files.wordpress.com/2020/04/operational.pdf
Kenneth,
How can the deaths be going down if the Rt value is greater than 1?
SteveF,
I think it’s supposed the be the Ro “after” the intervention. So Reffective is Rt*Csusceptible.
James talked about trying to identify when the contageon first started.
I have linked the latest US daily death Annan Bayesian model results below. The projected number of deaths by May 31 is 88K to 150 K with the current Rt at 0.88± 0.1. I went through the Annan code and only was able to clean up the readability of the plot. The code for the plot function is over 50 lines long and refers to several previously coded functions.
https://www.dropbox.com/s/12n41z07fbvgqjh/US_Daily_Deaths_Bayesian_02_19_2020_to_04_26_2020.pdf?dl=0
“How can the deaths be going down if the Rt value is greater than 1?”
Is that just a temporary artefact of data winding down through the infection rates? But how good are the data? Do we really know the levels of infection. An app in the UK suggests that the country has less than 1 %of symptomatic carriers. And it it has decreased 25 % in a week.
Kenneth Fritsch (Comment #183988)
While I’m not overly fond of the model you’re using, I do appreciate all the effort you’re putting into providing us with these updates.