In comments, Mike M and I have been discussing simple-ish epidemilogy ideas. I told him:
lucia (Comment #182845): “Anyway,I’ve ginned up a toy model to show a few things. It will be oversimplified, but I’ll show some stuff tomorrow.â€
He then suggested
I suggest you try:
(1) Let it run with R = 3, say.
(2) Reduce R to say, 1.5. That should give you something like the flattened curve in the cartoon with a peak much lower and later than in (1), and much broader.
(3) Let it run a bit with R = 3, then reduce R to 0.5 or so. That should give a peak that is much lower and earlier than (1) but with a width more like (1) than (2). That is somewhat like what has been done *if* what we are doing fits into the claimed framework.
(4) Like (3) but let R go back up to 3 before the cases drop to zero. Uh-oh. That is the issue I am raising.
If we are on the “kill it†curve, then there is a huge problem: IT WON’T STAY DEAD.
Today I’m going to let the model rip and show him the outcome of R0=3.0, R0=1.5 which will address his requests (1) and (2). I think there is sufficient to say about this comparison to stop there and leave the rest for later. (And if Mike M wants this faster, he can do it himself. 🙂 )
The assumptions for this model are:
- The total population to be infected in effectively infinite, and has zero initial resistance to a particular virus
- At time step 0, we seed the population with some initial “sick” concentration, Csick. The number of susceptible people who are sick is Csus=1-Csick
- During each time step, Δt
-
The instantaneous rate at which people who sicken corresponds to
dCsick/dt = R0*Csus*Csick. Where R0 is the reproduction rate in a population with zero immunity.
So during that time step there is an “effective” reproduction rate at any time Reff=R0*Csus - Everyone who was sick at the beginning of the time step either recovers or dies. After they recover or die, they can no longer infect anyone. So dCsus/dt =dCsick/dt. I will arbitrarily say 80% of the sick recover, 20% die during the time step.
- At the end of the time step, all variables are updated by adding the Δ values for that variable.
-
The instantaneous rate at which people who sicken corresponds to
- Note: Time increments Δt will be subdivided to reduce discretization error. (The subdivision of time steps is necessary to avoid having results for larger R0 look worse in the model than they actually are.)
There is nothing fancy in this to deal with amount of time to death and so on. Today’s figures will use R0 = 3.0, R0 =1.5 and throw in 1.1,1.2 and 2.0 because that’s what I’d planned to show before he made his request. If you want fancy, go read James Annan’s stuff.
Results
The following shows a plot of the instantaneous fraction of the original population that is ill during a particular time step:
The main features that are evident by applying the “eyeball” test are
- the fraction of the maximum instantaneous fraction of the population that becomes ill is larger for larger values of R0. In particular, for R0 = 3 more than a quarter population is sick after a few time increments.
- the epidemic is over much more quickly for larger values of R0. For R0 = 3 it is over in roughly 10 time steps.
It is worth nothing that the peak of the “sick” curve occurs when the total number who have already been infected reaches heard immunity. After that, people continue to sicken, just at declining rates. Some of these die.
The following plot shows the cummulative deaths as a fraction of the original population; this number is based on the assumption that 20% of those who fall ill die:
The main features I see using the “eyeball” test is that many more people die in when a disease with R0=3 hits an community with no immunity relative to when the disease with R=1.1 hits the community.
I haven’t had a chance to look at Mike M’s other questions. Heck, I’m not even going to pretend to have proof read this much. I’ll try to look at MikeM’s challenges tomorrow. At which point I’ll put up another barely proofread post.
Time to tutor Physics 2 now. 🙂
lucia,
Those results make sense. The epidemic grows when R > 1, where
R = R0*f_sus
where f_sus is the fraction of the population that is still susceptible. So with R0=1.1, herd immunity is reached with only 10% of the population having acquired immunity by having been infected. But with R0=3.0, herd immunity requires 67% to have acquired immunity.
Lucia,
Very nice. Less virulent diseases cause less death…. who’d a thunk.
.
I this line “ relative to when the disease with R=0.035 hits the community.†I think you mean R=1.1.
So social distancing does save lives…. and lots of them, but only if you never let up. In a sense, it makes the economic argument even more clear…. shutting down the economy does reduce total deaths, at enormous cost…. but only if you never allow it to start back up, at least until a vaccine is available…. but that is impossible. I guess the key question is: What policies hurt the economy the least for the biggest reduction in net transmission rate? Hard question. I just hope those sensible Democrat governors keep blocking the sale of tomato seeds and potted plants…. because that is so critical a public health policy.
The problem, though, is that social distancing and economic shutdown artificially reduces Ro to some Reff < Ro. When that ends Reff increases and we’re back to square one.
Question: Is Ro being underestimated or overestimated if, say, 90% of infected are asymptomatic but still infectious.
That the disease with R=1.1 killed fewer people didn’t surprise me. But oddly, it does matter because when discussing “flatten the curve” in comments below, both Mike M and I were both saying the area under the sickness curves was not changed by changing R. In fact the area is less for lower Ro.
.
Avoiding the argument about whether an particular strategy for reducing Ro is worth it, this does at least establish that a society wide change that makes Ro permanently lower would save lives. It doesn’t merely “flatten the curve”. It is important to establish that at least in some sense, lower Ro does “save lives” before moving on to other strategies.
.
(In some sense we all know this. That’s sort of the point of hand washing, soap, good plumbing and so on and so on. But I think it’s important to make sure we don’t suddenly have people claiming it doesn’t happen if they don’t like a particular strategy for other reasons.
If you don’t like a strategy for other reasons, it’s both perfectly fine and totally necessary to give them. You– SteveF– seem to have been giving the non-epidimiology reasons. Your objects in economic. But we’ve got some others seeming to suggest “stay at home” doesn’t “work” (in some way or another.) I know have a tool to look into that a bit more.
DeWitt,
I think it’s overestimated because a naïve analysis shows an Ro based only on symptomatic cases, where in fact most of the transmission would be by the more numerous asymptomatic carriers. The true Ro would therefore be much lower than the naive analysis indicates. Maybe this is where the U of Washington model goes off the rails.
lucia (Comment #182880): “That the disease with R=1.1 killed fewer people didn’t surprise me. But oddly, it does matter because when discussing “flatten the curve†in comments below, both Mike M and I were both saying the area under the sickness curves was not changed by changing R. In fact the area is less for lower Ro.”
.
Actually, I never said anything about a change in R0 because I never considered the possibility of a permanent change in R0 due to behavior. But I suppose that was implied by the flattened curve, although I never realized that. I agree that a permanent change in R0 would reduce the number of people who eventually get sick and die of the disease. But I see nothing other than acquired immunity that will create a substantial permanent change.
.
I am doubtful as to whether the extreme measures being taken are having a large effect because I suspect that casual contact is not a major vector for transmission. If it is, then yes, what we are doing could matter a lot in the short run, but it is not sustainable. But I suspect it is close contact that is the major transmission route. That is not altered nearly as much; in some cases, what we are doing might even increase that. But I can not prove any of that, it is just guessing gleaned from fragmentary pieces of information.
If an apparent Ro, calculated from only symptomatic carriers is 3, but there are in fact 6 asymptomatic carriers for each symptomatic carrier, then the true Ro is (I think) 1+ (3 – 1)/7 = 1.29. Social distancing then only needs to reduce effective transmission by a factor of 1/1.29 = 0.775 to reach the point where the number of infected hardly changes, until gradually increasing herd immunity very gradually makes the illness die out.
.
Sure looks like the effective R has dropped below 1 in Florida: https://experience.arcgis.com/experience/96dd742462124fa0b38ddedb9b25e429
SteveF (Comment #182884): “If an apparent Ro, calculated from only symptomatic carriers is 3, but there are in fact 6 asymptomatic carriers for each symptomatic carrier”
.
I don’t see that it matters. For simplicity, I’ll make it 2 out of three asymptomatic.
1st generation: 3 total, 1 symptomatic
2nd generation: 9 total, 3 symptomatic
3rd generation: 27 total, 9 symptomatic
The growth of the number with symptoms is 3, 9, 27, 81 … Just the same as if no one was asymptomatic.
SteveF,
It’s going to turn out that temporary social distancing can reduce total deaths somewhat.
.
One reason is that the total deaths when we just have a constant Ro is greater than Deaths=R_death* (1-1/Ro) with R_death the death rate for an infected person. This actually has already been shown if you compare the number of dead in the legend to 0.2*(1-1/Ro) . For Ro=3, social distancing will have a potential of reducing deaths by about 30% over the course of the epidemic.
.
Another is that in the *real world* (but not my current model), the death rate for infected people is affected by whether or not they can get treatment. Also, eventually, we’ll have treatments.
MikeM
What you wrote about the “flatten the curve” graph (or seemed to be saying about the flatten the curve graph in a specific picture.)
.
.
But this isn’t true. All those areas are are similar but curves for different values R0. The areas under them are not the same.
.
I realize you might have been thinking something about lack of permanence. I think one of the purposes of the current post is to get people (including me) to be more specific and clear. Otherwise, a lot is getting jumbled, and people aren’t making the point they think they are making.
.
At least we seem to agree: Lower Ro (i.e. flatter curve) if maintained results in fewer deaths.
Mike M,
That many people are asymptomatic would certainly influence the development of herd immunity. Suppose there were 990 asymptomatics per 10 symptomatics….. the illness begins to diminish when it appears only 10 people in 2000 have become sick, even though half the population has in fact been infected. I was trying to draw a distinction between apparent and actual Ro when apparent Ro is based on only symptomatic people.
SteveF,
Assuming “f” the proportion of symptomatic people is a constant, then both Ro’s are the same. That’s defined based on a population with no pre-existing immunity. What is different is the current fraction of susceptible people over time.
.
I could put a line on my graph to show… I guess I will.
SteveF,
The below is a detail of the fraction of
* the population actually infected (black solid),
* symptomatic
* who die
if “Ro=3.0”. Note: there are no units on time. (The actual “time” corresponding to 1 unit would be equal to the time it takes a person to either recover or die. )
As the all infected can transmit in this model, the “symptomatic” and “die” curves are just the “infected” curve scaled by a constant. (I picked 40% and 20%.)
As the symptomatic and death rate are the only visible manifestation of the epidemic, that’s what we’d see in the data. I don’t think it would make any difference to the value of R0 you estimated based on EARLY data. You can estimate that based on deaths (in fact James Annan is).
.
What happens is that the apparent epidemic would follow the dashed line.
Good work!
.
The real question here is how can we maintain the optimum social distancing outcomes while keeping the economy open, if even in a limited way?
.
We are currently stuck in a “Orange Man Bad, must keep extreme policies” vs “I want to kill old people, full economy” false binary. The first step could be a harsh open economy. You need to make an appointment to shop at Target, and you have to follow certain paths at the store.
.
One can imagine that a proximity boundary count could be made with * mandatory * cell phone location data at businesses monitored by our overlords at Google and Apple. If that count exceeds limits, then the business either rectifies it quickly or is shut down. You could also wear a lanyard that beeps when you have come too close to others. These are quite invasive examples but could be a compromise to get thing rolling again. The smart phone seems to be the one that can be universally implemented in a short period. Privacy advocates will have a bird of course, they can stay home.
Tom Sharf,
Agree. And those aren’t answered directly by an model (not even a better one than this!)
.
I agree discussions are in a false binary. Having said that: We don’t know the full range of what’s possible (legally, socially and so on.)
.
I’m reasonably sure we will not be on lockdown over huge areas for 6 months because people will violate. At that point… what?
.
I’ve read all sorts of things proposed, including registries to authorize “passports” for those who are immune. But that requires a good antibody test with pretty false positives and negatives. (It doesn’t need to be perfect btw. It only needs to be good enough to make the instantaneous collective “Reffective” become less than 1 so that individual infections tend to die out, not grow. )
.
All sorts of things become possible if we manage to have a good antibody test with low false positive and negative rates. And of course, even more things are possible if we have a decent vaccine.
.
I do know that Annan’s modeling suggests the “after shutdown” R for the infection in the UK appears to be below 1. Not a lot below… but below.
Lucia,
I was confusing Ro and Rt.
.
The ratio of asymptomatic to symptomatic cases controls the total number of people who suffer serious illness (or death) by the time herd immunity knocks down the disease (via infection or a vaccine), absent any effective treatment.
.
Speaking of treatments, here is a study from Wuhan of hydroxychloroquine: https://www.medrxiv.org/content/10.1101/2020.03.22.20040758v3.full.pdf
.
Note this is not yet peer reviewed. It was randomized, placebo controlled. All patients in the study (62) had low level pneumonia based on cat scans, those with severe pneumonia were excluded (and presumably all treated with hydroxychloroquine). The study lasted 5 days. The HCQ group showed faster average improvement in fever, cough, and chest cat scan than the control group. Four of the 31 in the control group progressed during the study to severe pneumonia, while none of the HCQ group did.
I guess we are lucky that the lock downs (and potentially seasonality, et. al.) are getting R less than 1.
.
Almost the entire world is in a declining growth stage now (except Russia, India, and a few others). If this thing was so infectious that these policies kept R > 1 then realistically we would be up sh** creek and just have to endure it. That being said, it appears that R is not exactly very much below 1 when examining the daily cases in many places. A policy change that nudged it above 1 would be a big problem. This might be a simple trigger than can be used to re-enter lock down.
.
From examining the trends it would seem that two more weeks of the same outcomes will get things pretty close to under control, especially if you take the reporting lag into account.
.
In the future I think we will go into lock downs much more quickly based on this experience, but based on much better surveillance. Nobody wants to be NYC the next time around.
Tom Scharf,
Yes, in many places (like Florida https://experience.arcgis.com/experience/96dd742462124fa0b38ddedb9b25e429),
the Rt value looks to be well below 1. If that trend continues, then in two weeks the death rate should be down considerably from its peak level. The decisions about what restrictions to ease without driving Rt too high remain to be made, but without a good estimate of the current Rt in different places (and it is going to differ a lot!), those decisions will be arbitrary at best. Will tomato seed sales in Michigan continue to be blocked? Bad data in the hands of nanny-state Democrats makes that kind of lunacy far more likely.
.
James Annan’s Bayesian based model looks a lot more capable of evaluating what the true value of Rt is than the models most politicians are seeing. I would like to see that model applied to data from the individual States, but I don’t have (or know) R!
A bit more on the “death rate†for the virus. Deaths due to the virus are WAY overstated.
.
https://www.foxnews.com/media/physician-blasts-cdc-coronavirus-death-count-guidelines
“Jensen then told Ingraham that under the CDC guidelines, a patient who died after being hit by a bus and tested positive for coronavirus would be listed as having presumed to have died from the virus regardless of whatever damage was caused by the bus.â€
.
and
.
“Right now Medicare has determined that if you have a COVID-19 admission to the hospital you’ll get paid $13,000. If that COVID-19 patient goes on a ventilator, you get $39,000; three times as much. Nobody can tell me, after 35 years in the world of medicine, that sometimes those kinds of things [have] impact on what we do.â€
.
Add in that ventilators are looking like a major CAUSE of death for COVID-19 patients with inflamed lungs.
Ed Forbes,
I doubt it. Excess mortality analysis for New York City between 4 March and 4 April shows three times as many deaths as expected. Unless they had already been tested positive, which currently means they were symptomatic, someone who died after being hit by a bus would not be tested for coronavirus. Sounds like a straw man argument to me.
Lucia, it is always good to see a toy model as you have produced here. I have briefly gone over it and am not sure I understand all the ramifications of it. It certainly makes your point about mitigation not only flattening the curve but also reducing deaths. I have eyeballed your curves and death rates and calculated some results from which I think you can tell me if I am understanding your model. I know you were more or less making a qualitative point, but looking at quantitative results and then applying some different assumptions might give more insights.
First I calculated the expected total deaths in the US with a population of 328 million based on your portion of population deaths: for Ro=1.1, 1.2, 1.5, 2.0, 3.0 would be in millions 11.5, 27.2, 38.4, 52.2, 61.7, respectively, and for those Ro’s the time to peak cases in days assuming a 14 day time for recovery or death would be 70, 140, 420, 1120, 1960, respectively.
I’m trying to understand the mechanism behind the chart in this link:
https://www.nytimes.com/2020/04/14/opinion/covid-social-distancing.html
It claims that intervention two weeks earlier would have reduced deaths to 6,000, and a week earlier would have reduced to 23,000, from 60,000 current projected.
My issue is looking closely, in the 23,000 case with intervention a week earlier, the deaths in March and some way into April are higher than the late intervention case. Why is that?
lucia,
Ro may be the same whether there are asymptomatic infections or not, but the estimate of the CFR is totally dependent. For example, in your toy case of 60% asymptomatic and 20% die (having 20% of the asymptomatic also die seems counterintuitive), the CFR would look like 50%, not 20%.
It’s apparent everyone way under-estimated the initial growth rate, and are now likely way under-estimating the contraction rate. It’s instructive to go back a week or two and view expert predictions. There seems to be two separate calculations needed, before lock down, and after lock down. There shouldn’t be any sharing of this data across the boundary for simple model estimates other than initial conditions. There is some smearing due to reporting lags and such. Estimates are now all over the place but I think they will converge once post lock down data gets to 3-4 weeks.
Tom Scharf,
If you mean that new cases and deaths will drop faster than predicted, then you are wrong. Look at the projections for daily deaths in Italy and Spain at the COVID-19 site. They are wildly optimistic and quite wrong. Everywhere I have ooked where new cases and new deaths have passed the peak, the decay rate is slower than the expansion rate.
The NYT article doesn’t really answer any useful questions other than “Does acting earlier during a certain outbreak where containment is lost reduce the case count?”. Yes, duh. I do find it hilarious that they didn’t bother using NY in their examples, but “randomly” picked three red states. It does make the case that delaying for weeks follows math in an exponential outbreak.
.
Is containment a lost cause?
Is R >> 1?
Is the mortality rate unacceptable?
Will healthcare be overwhelmed without intervention?
.
Then you go to your list of interventions that have more options than (do nothing) and (total lock down of non-essential activity).
DeWitt,
I’m just looking at FL at the moment
https://covid19.healthdata.org/united-states-of-america/florida
.
They have peak daily deaths estimate on May 6th and the peak new reported cases looks to be have been around April 2nd.
https://fdoh.maps.arcgis.com/apps/opsdashboard/index.html#/8d0de33f260d444c852a615dc7837c86
.
This would seem to indicate a delay from case detection to death of over a month. There might be higher math involved here I am not accounting for (total active cases at any one time), but the actual data will clear everything up eventually.
.
Italy looks wrong, but Spain might be right, their reported cases has dropped off sharply in the past week. My eyeball test is assuming the death trend will look like the reported cases trend with a lag. Not perfect of course.
Kenneth Fritsch (Comment #182920)
Lucia, obviously the days to peak in my above post were reversed.
DeWitt,
If you have not already done so, I suggest you read James Annan’s paper on Baysian projections of deaths. https://bskiesresearch.files.wordpress.com/2020/04/operational.pdf
He calculates a current Rt for the UK (based on data through yesterday) of about 0.5, suggesting rapid fall in deaths while Italy has reached an Rt of only 0.9…. which explains the very slow drop in deaths in Italy. France: 0.67, Spain: 0.55.
DeWitt
What does CFR stand for?
Tom Scharf
You need to read Annan’s blog to see how he’s fitting, including a lick down, including some time data and then forecasting.
Case fatality rate or code of federal regulations. I suspect the former. 😉
Tom Scharf,
The time from showing first symptoms to death averages somewhere near two weeks. Hard to see how the Florida death rate peaks much later than a few days from now. The healthdata.org projection just looks wrong.
Steve, thanks for the link to Annan’s analysis. It makes perfect sense to use a Bayesian approach with the Covid-19 data and up dates. He has published his R code but I have not read far enough to determine if he uses a Bayesian program that I have on my computer.
Good that he shows this online for the purpose of motivating those in field making Covid forecasts
MikeN (Comment #182921): “I’m trying to understand the mechanism behind the chart in this link:
https://www.nytimes.com/2020/04/14/opinion/covid-social-distancing.html”
.
What a dishonest article. They write:
“The recent divergence of epidemics in Kentucky and Tennessee shows that even a few days’ difference in action can have a big effect. Kentucky’s social distancing measure was issued March 26; Tennessee waited until the last minute of March 31.”
So let’s look at the numbers:
………….. KY TN
March 26 198 957
March 31 480 2238
April 7 1008 4138
April 14 2048 5823
Cases in Kentucky have been rising *faster* than in Tennessee. So if that proves anything about lockdowns, it is that they make things worse.
Mike M,
“What a dishonest article.”
.
Ummm… it’s what the NY Times does.
Mike M. (Comment #182935)
So let’s look at the numbers:
………….. KY TN
March 26 198 957
March 31 480 2238
April 7 1008 4138
April 14 2048 5823
Cases in Kentucky have been rising *faster* than in Tennessee. So if that proves anything about lockdowns, it is that they make things worse.
__________
True about the faster rise, but the Kentucky case number starts from a smaller base than the Tennessee number. The latter had 4,866 new cases from Mar 24 to Apr 14 while Kentucky had only 1,850. Even adjusting for the population difference (Tenn 47 % greater) would put Kentucky at only 2,720 new cases, compared to the 4,866 in Tennessee. Seems to me that means Kentucky’s lockdown has been more effective that Tennessee’s.
OK_Max (Comment #182940): “Seems to me that means Kentucky’s lockdown has been more effective that Tennessee’s.”
.
Only if you make the inane assumption that the lockdown had an effect weeks before it went into effect.
Mike M. (Comment #182942)
OK_Max (Comment #182940): “Seems to me that means Kentucky’s lockdown has been more effective that Tennessee’s.â€
**Only if you make the inane assumption that the lockdown had an effect weeks before it went into effect.**
______
Never mind, I’ll just try to re-interpret your “”Kentucky have been rising *faster* than in Tennessee. So if that proves anything about lockdowns, it is that they make things worse.”
Here is another way to put what you said:
During Mar 26-Apr 4, Tennessee had 148% more new Covid-19 cases than Kentucky, while having only 47% more population, proving that Kentucky locking down four days earlier than Tennessee made things worse.
You may want to edit that.
Max, suppose hypothetically Tenn’s numbers were starting at 50,000(50x actual) and moved to 55,000(9.5xactual), for a case increase of 5,000, same as actual. Would you still think your critique holds up?
Tom Scharf,
Spain, April 15, 6,599 new cases. That’s the highest number of new cases since April 4. But even before then, the new case bar graph doesn’t look like the precipitous dropoff projected in the IHME page. The daily death bar graph looks even worse. New deaths Spain April 15, IHME projected: 158 (50-411), actual: 557.
OK_Max and others,
What’s important in TN and KY is not what the respective governors ordered, but the behavior of the citizens. The cell phone location data graphed here shows effectively identical behavior in the two states.
Also, Kentucky has a higher death rate than TN as of yesterday, 26 deaths/million vs 18/million in TN. Also, a larger fraction of confirmed cases have recovered in TN, 36% in TN vs 13% in KY, possibly indicating that TN is at a later stage than KY.
MikeN (Comment #182945)
April 15th, 2020 at 6:15 pm
Max, suppose hypothetically Tenn’s numbers were starting at 50,000(50x actual) and moved to 55,000(9.5xactual), for a case increase of 5,000, same as actual. Would you still think your critique holds up?
___________
MikeN, I presume you mean like the following, with Kentucky’s cases increasing by 1,850 or 934%, and Tennessee’s cases increasing by 5,000 or only 10%, the shutdowns having occurred on Mar 26 in Kentucky and Mar 31 in Tennessee.
March 26 198 50000
April 14 2048 55000
The Question: Given cases in Kentucky have been rising *faster* than in Tennessee, if it proves anything about lockdowns, does it prove they make things worse?
My Answer: No, it strongly suggests Covid-19 infections started much earlier in Tennessee. It also suggests some of the infected Volunteers visited Kentucky.
DeWitt Payne (Comment #182948)
**What’s important in TN and KY is not what the respective governors ordered, but the behavior of the citizens.**
I sometimes have this irrational urge to do the opposite of what I am asked to do. I’m from Eastern Tennessee stock on my mother’s side.
**Also, a larger fraction of confirmed cases have recovered in TN, 36% in TN vs 13% in KY, possibly indicating that TN is at a later stage than KY.**
Makes sense
SteveF,
From Annan:
Yep.
I wish he had integrated the daily curves to get total deaths.
DeWitt,
You could ask him to integrate the daily deaths along the highest probability projection line.
> TN is at a later stage than KY
Despite locking down later…
I decided to dust off my own toy model from a few weeks ago and look at the question of what happens if an intervention changes R0. It is a standard differential equation SEIR model, except that I ignore acquired immunity since that makes no difference until a large fraction of the population has been infected. The model is:
E = number infected but not yet contagious
A = number of actively infected and contagious
R = number removed (recovered, dead, or quarantined)
dE/dt = beta*A – E/L
dA/dt = E/L – A/D
dR/dt = A/D
beta = transmission rate, I used 0.3/day during spin up (t<0)
L = latency period, I used 4 days
D = duration, I used 10 days during spin up
Note that the above choices give R0=beta*D=3.0 and a doubling time of 3.5 days.
At the end of spin up, A = 100 and E = 84. Note that the ratio E/A does not change with time and the absolute values don't actually matter.
Then at t=0, I start an intervention phase in which R0 is reduced to less than unity by some combination of reducing beta (social distancing) and reducing D (quarantine). The slowest response to intervention is obtained by reducing the transmission rate, beta.
Reducing R0 to 0.5 while keeping D at 10 days results in a maximum of 116 active cases on day 4 following intervention. The maximum number of new cases is on day 0.
Reducing R0 to 0.5 while keeping beta constant results in the number of active cases dropping immediately.
Of course, in the real world there is a delay between people getting sick and testing positive. But it still seems that *if* the lockdowns were effective, the result should have been seen within two weeks. That appears to be what the experts expected when the stay at home orders were issued. But it is not what has happened.
As data on asymptomatic cases, symptomatic cases, serious cases, and deaths mount, we can begin to better define what the cost would have been for no public response at all.
.
It is looking like Ro really is somewhere near 3, so assuming nobody has natural immunity, 2/3 would have to be infected to establish herd immunity. Of those who are symptomatic and confirmed positive, about 14% end up in the hospital. A reasonable estimate is that 6 in 7 people are asymptomatic, and probably another fraction have symptoms so mild they are never tested. So perhaps the 14% that end up in the hospital represent about 1.5% of the total of infected people. Among those who are hospitalized, the fraction who die is in the range of 30%, so somewhere near ~0.45% of all infections end in death. The total number of expected deaths if there is zero natural immunity and no effort to control spread is about 2/3 * 320 million * 0.0045 = 960,000 deaths. Those deaths would be dominated by people past 65 and especially those with serious pre-existing conditions.
.
It is very clear that current policies and changes in behavior have reduced the rate of transmission to well below 1 in many places, and the rate of infection is declining in those places.
.
One question I have is if there are reasonable steps to maintain the rate of transmission well below 3.0 while not destroying the economy (and not overloading the health care system). Selective isolation/sheltering of the most vulnerable comes to mind, but perhaps there are other things that could be done. My guess is that widespread awareness of the possibility of infection would alone be enough to reduce the rate of transmission somewhat. If Rt dropped from 3 to 1.5, then herd immunity would be established at 1/3 infection rate, and expected deaths until the disease dies out would drop to ~480,000.
.
The other imponderable is if there exists a significant fraction of the population which is refractory…. they will never become infected. I’m not sure how this could be evaluated.
Mike M,
“But it is not what has happened.”
.
Seems the rate of infection fell in Florida pretty quickly, starting near the beginning of April. The schools were closed on March 17, and the governor’s “stay at home” order (a recommendation really) and social distancing rules were April 3. I noticed a rapid increase in the number of people using masks right around April 3 as well.
Did you read James Annan’s article about estimating transmission rates?
Mike M
Why did you use a latency period of 4 days and Duration of 10 days?
Both seem rather low since estimates based in known exposure and detected symptoms and time to die after exposure are both much higher. If you are going to discuss how long it takes to see anything, you need these.
(My first model just sets both to zero. I want to switch but every time I do, I get interrupted. There is competition for my mac AND I need to try to help Mom get on Zoom… which is horrible.)
Orders by various Governors of the States for blanket house arrest of their citizens, known as “lockdownsâ€, is unconstitutional as the United States are in neither Rebellion or Invasion.
.
US Constitution, Article 1, Section 9, Clause 2.
“The Privilege of the Writ of Habeas Corpus shall not be suspended, unless when in Cases of Rebellion or Invasion the public Safety may require it.â€
.
Michigan Is likely the worst example of executive overreach.
.
Cases are moving into court and preliminary injunctions against such lockdown orders may be forthcoming.
lucia (Comment #182968): “Why did you use a latency period of 4 days and Duration of 10 days?”
.
Because I wanted to use the largest reasonable values so as to slow down the effect of intervention.
The incubation period is definitely not too long. The average time to symptoms is given as 4-5 days and it seems generally accepted that people become infectious before symptoms appear. Note that if the average time to symptoms is 5 days, then in a first order (exponential) model 5% won’t get symptoms until 15 days or longer.
From what I can find, the illness typically lasts a week to 10 days. With an average of 10 days, 5% will be sick for a month or more.
SteveF (Comment #182967): “Seems the rate of infection fell in Florida pretty quickly, starting near the beginning of April. The schools were closed on March 17, and the governor’s “stay at home†order (a recommendation really) and social distancing rules were April 3. I noticed a rapid increase in the number of people using masks right around April 3 as well.”
.
Interesting. How did you determine when the rate of infection started to fall?
———
Addition after looking at the link provided by Tom Scharf (Comment #182926).
https://fdoh.maps.arcgis.com/apps/opsdashboard/index.html#/8d0de33f260d444c852a615dc7837c86
The highest reported number of new cases was on April 2. The were rising rapidly before that, then leveled off or started to decline. That can hardly be due to actions taken on April 3. But it might now be dropping sharply. Is the “new cases by day” plotted by date of the positive test report?
——-
Addition: It seems so, :“Case date†is the date the positive laboratory result was received in the Department of Health’s database system and became a “confirmed case.†This is not the date a person contracted the virus, became symptomatic, or was treated.
The FL Dept of Health Dashboard is what I use to track the cases. Strangely it seems to be both done by the government, and is clear, concise, and useful, ha ha.
.
They finally put a “Deaths by Day” tab in, I think that just showed up. So you can compare the “Recent Cases” and “Deaths By Day”. This is showing only a ~4 day lag between peak case counts and peak death counts.
.
This is very surprising, my guess is that this is due to gravely ill people not getting tested until they are in the hospital and results are not coming back until they are close to death. As of two weeks ago, you could not get tested unless you were having respiration problems.
.
Note that the estimated max deaths per day is 20 days away but the dashboard trend seems to show that it has already passed?! My guess is the model is having big problems with the transition to declining cases. Also the deaths per day counts don’t match up in these two sites for unknown reasons.
https://covid19.healthdata.org/united-states-of-america/florida
.
It says: “The Deaths by Day chart shows the total number of Florida residents with confirmed COVID-19 that died on each calendar day (12:00 AM – 11:59 PM). Caution should be used in interpreting recent trends, as deaths are added as they are reported to the Department.”
.
So I guess the chart isn’t necessarily accurate for recent deaths.
Mike M,
“Interesting. How did you determine when the rate of infection started to fall?”
.
As you correctly note, there is a delay between testing and results being sent to the state health department. I reasonable guess is 2 or 3 days. So the change right around the first of April from rapid increase to flat and then to declining indicates a change in infection rate before early April: somewhere in the range of the latency period plus the period between the start of symptoms and getting a test plus the time for the test results to be reported. I don’t know exactly how long that all is, but I would guess 10 days or more. So to impact the rate of new reported cases, the actual rate of transmission had to have fallen some time around March 20 or a bit before.
.
I note also that people were (on their own) already avoiding personal contact (handshakes, restaurants, etc) well before any state orders. I flew from Sao Paulo to Miami on March 14, and the plane was nearly empty…. that same flight had been 100% full each time I took it, 20+ times over 5 years.
Tom Scharf,
The surprisingly good dashboard is the work of this company: https://www.esri.com/en-us/about/about-esri/overview
with headquarters located in Redlands California. The State of Florida is not likely able to create such a good app.
.
I am glad they put in the daily deaths tab… far more useful (and likely more certain). As to why the healthdata.org is “having big problems with the transition to declining cases”: James Annan said it pretty well:
Ed Forbes,
Yes, the “lockdowns”, especially if any effort is made to actually enforce them, are almost certainly unconstitutional. Just like FDR’s February 1942 executive order that imprisoned US citizens of Japanese descent. The federal courts, having been packed by FDR with his friends and supporters, were unwilling to do anything about it until December 18, 1944. The SC informed FDR that the decision would go against him, so it was announced the day before the Supreme Court ruling that the executive order would be rescinded, effective January 2, 1945. By then the Pacific was was almost over. Like all progressives, FDR didn’t give a sh!t about the constitution.
.
As to the governor of Michigan and a host of others: they also don’t give a sh!t about the constitution. Let’s hope the Federal courts don’t fail us again like they did the Japanese-Americans.
California has been cited as evidence that stay-at-home orders work; IHME says they implemented that on March 19. I think that the Bay area and LA did so even earlier. Florida, on the other hand, did not implement stay-at-home until April 3 and still has not closed non-essential services (according to IHME). Obviously, the epidemics must be following very different paths in the two states.
So I decided to plot daily new cases for Florida and California. Virtually identical.
Addition: Daily deaths are also almost the same for the two states.
Mike M,
The real issue is that Florida has a conservative Republican governor, who can do no right, and California has a socialist governor who can do no wrong.
Many of the hardest hit FL counties had safer at home orders about a week or two prior to the state order.
.
I think these things are working, but am not fond of the increasing zeal the media has for them the more authoritarian they get. Tampa’s mayor put in a 9pm – 5am curfew in recently in response to what appear to be one single large party on Easter. Upon examination the penalties are no different than the existing 24 hr stay at home order, yet she was applauded for the effort.
MikeM
Both of your choices are LOWER than those indicated by WHO and lower than I read.
.
.
Who’s estimate for the period is 1-14 days with the “most common” as 5. So your pick of 4 is certainly not “the largest reasonable value”. It’s a lowball. Not high in anyway shape or form.
.
https://www.who.int/news-room/q-a-detail/q-a-coronaviruses
.
And once again, you pick a lowball. Here’s what WHO says.
.
.
Your 10 days is low compared to 2 weeks and it is very low compared to 3-6 weeks.
.
Of course if you stiff short times in your model (as you did), your model will look like response to the lockdown should have been fast. But the numbers those out in the field reports for larger “D” and “L” than in your model. Yours may be low by a factor of 2-3, and are low relative to best estimates.
.
I can show you results for L and D = 0. 🙂
Most people start showing * initial * symptoms in 2-5 days after exposure according to what I read (but the complete range is large), but they also then wait another couple days typically before they are sick enough, or are sure they have a significant illness, before they request a test from a medical provider. Then they have to schedule the test and wait for results which seems to be a highly variable time frame that has changed significantly over the past month. The sickest patients (think pending or possible hospital admissions) are given priority at the labs and jump to the front of the line.
.
My impression is by the time a case is officially counted the person has been sick for at least a week and probably closer to two weeks. It cannot be understated enough that most people can’t get tested with mild symptoms, so who knows how many people had to wait until they had severe symptoms (which takes about 5 days after initial symptoms). There is great uncertainty here.
lucia (Comment #182983),
My use of 4 days to being infectious, and therefore 5 or 6 days to symptoms is in agreement with the WHO statement you cite. I changed by duration to two weeks, to agree with the WHO statement. With R0=3.0, that gives a doubling time of 5 days.
Now with 100 Active and 64 Exposed at t=0, there is a maximum of 115 active cases on day 4.
Tom,
I just going by what WHO reports. 🙂
This paper puts the median at 5 .1 days and mean at 5.5 days:
https://www.jwatch.org/na51083/2020/03/13/covid-19-incubation-period-update
CDC says 2-14 days based on MERS-Cov
https://www.cdc.gov/coronavirus/2019-ncov/symptoms-testing/symptoms.html
At this point, it’s very difficult to estimate the lags “in the wild” because track and trace has become impossible. Also, the US never really did it. But the midpoint estimates appear to be higher than Mike M’s pick of 4 days.
I don’t know the best practice for picking lag in the model based on this. But 4 days is lower than publications are saying.
There are different questions with different answers. I’m probably commenting on the wrong question here.
1. How much lag is there in the reported daily case counts and death counts?
2. What is the replication time of the virus (the delay of virus shedding from contagious person 1 to infected person 2 starts shedding virus)?
MikeM:
On times: there is also evidence the infectious period is LONGER than the recovering time.
.
https://www.sciencedaily.com/releases/2020/03/200327091234.htm
.
.
That would mean that the “recovery time” is a lower bound on the “D” for the model. The lower bound on recovering time after symptoms according to WHO is 2 weeks (which exceeds your 10 days) and if we add the extra time for the formerly ill to actually stop being infective, a more plausible lower bound for D is near 2 1/2 weeks, not the 10 days you picked.
.
CDC says you can leave isolation:
.
If they will not have a test to determine if they are still contagious, they can leave home after these three things have happened:
They have had no fever for at least 72 hours (that is three full days of no fever without the use medicine that reduces fevers)
AND
other symptoms have improved (for example, when their cough or shortness of breath have improved)
AND
at least 7 days have passed since their symptoms first appeared
.
Good luck interpreting “improved”. There is a lot of confusion on the period of virus shedding. You would think that with 2M possible test subjects there might be some more clarity here.
Tom Scharf
Those are good questions.
I think you are trying to get an estimate of “D”. But this time period would underestimate that possibly by a lot.
.
The lag between a person being added to a case count– which happens mostly when they are tested– and when they die isn’t the same as the lag between when they became infective and when they died. The latter is “D” in the model and is likely longer than the former because a positive result to diagnose their infection happens after they are ill. (In our current environment, the test may happen quite a while after they are ill!)
.
This is kinda sorta an attempt at getting to “L”. That is how long it takes form someone who eventually get exposed sicken (and become infectious. In the SEIR model, I is really “is infectious”.) But if we think only those who are sick are infectious, then this should be “L”. Mikes choice for tha tlooks low.
.
Worth noting that the diagnosis of “when” the restrictions could be seen in death statitics would be mostly proportional to (L + D). Based on WHO, CDC published papers I can find, Mikes choice for both L and D are individually low.
lucia (Comment #182993): “But if we think only those who are sick are infectious, then this should be “Lâ€. Mikes choice for tha tlooks low.”
.
But the evidence is that people become infections before symptoms appear. So L should be less than the time to first symptoms.
Increasing L to 5 days would slow things down by maybe a day. No real difference.
The people who are really sick are not the ones doing most of the spreading. So the duration should be lower than mean time to recovery. Also, if people are being isolated (not much point to testing otherwise), that lowers the duration for spreading the virus.
Mike M, Lucia,
Models, with explicit declared constants for relevant parameters, are needed.
MikeM
Sure. But “D” is the time from when they become infectious.
So given the meaning of A and I, this information means you need to add the amount you subtract from “L” to “D”. Because “D” is measured from the time they are infectious to the time they either die or recover. It is not measured from the time they show symptoms. The time from when they are first exposed to death recovery is “L+D”. It is not “L + gap of your choice for reason of your choice + D” with then “oh… let’s pretend the gap is zero for the purpose of modeling.
MikeM
Well…
First: in your model they are. In your model the Exposed don’t infect anyone at all. Then the “A” infect people. That’s in the stage where they are sick.
.
Second: You don’t know whose doing most the spreading. That’s one of the questions no one knows.
.
Third: There is some concern people might be spreading after they seem to have recovered. So no, those people aren’t really sick. But they are spreading long after they were exposed, which your model definitely assumes does not happen.
SteveF
Sure. At least James Annan took a model and found the best fit for the parameters based on actual data from known progress of infection.
.
I intentionally did not put in “time” with units to try to avoid suggesting anyone can predict what is happening. I want to use it as a cartoon to explain qualitative features. The SEIR model can show some features (that’s why it’s used for some things.)
.
My point with respect to MikeM is that even given this choice of model (SEIR) while claiming to be picking number on the upper bound of time delay, he is picking ones that are below the best estimates.
.
So:
* for D, he’s picking 4days , when 5.1 (for a median) and 5.5 for mean are more reasonable. 5.5 is 37% bigger than 4.
* for L he’s picking 10 days when the corresponding values from who are 2 weeks to 6 weeks.
Then after doing this, while also knowing the model is simplified, he is actually comparing to some unspecified data and claiming that using these picks he somehow can claim that this comparison (not shown to us) indicates the stay at homes don’t work!
I am planning to separate my “exposed” from “sick” and rerun (after Jim goes to bed!!) Then I can show some numbers. I wanted to avoid discussing time.. but it’s pretty clear I’m going to have to discuss time.
Lucia,
Thanks for that. I think that the more smart, technically capable people think about this and make their thoughts known, the less likely public policies will go off the rails, and the less likely that know-nothing hacks in government (like Michigan’s governor) will do senseless damage.
.
My hat is off to James Annan: he has made a real contribution to rationally evaluating what seems the ultimate irrational subject: how many will die and when.
.
BTW, James’ little paper probably contributed more to human welfare than anything he has done before. Just sayin’.
SteveF,
Sure. I hope their paper (it’s james and jules’) gets accepted and read. It looks good. I’ve been watching that and (quietly) applauding their work.
.
FWIW: The shape is qualitatively what I can show with a SEIR (because they used a SEIR. Duh. ) — I just have no decent numbers. They thought to do that and did it.
.
They also are releasing the code and other people can apply it to data in other areas. I’m sure people are now doing it. I’m pretty sure at a minimum, it will show that stay at homes are working here just as they seem to be in the UK.
.
Also: the model is out there. So it’s possible to make a testable forcast at this point. It’s great.
SteveF (Comment #182997): “Models, with explicit declared constants for relevant parameters, are needed.”
.
Yes. I did that.
.
SteveF (Comment #182967): “Did you read James Annan’s article about estimating transmission rates?”
.
I can’t seem to find his parameters, except for R values.
——-
OK, from Table 1 it looks like his priors were centered on 4 days for latency and 2 days for duration. But I don’t know what he ended up with.
Lucia,
No slight to Jules intended. It’s just that it has been James out front waving the flag.
Mike M,
As I understand the paper (and I could be mistaken) those parameters are adjusted to get a best fit to the evolving data. Maybe you could ask James to explicitly show how those parameters evolve during a run.
SteveF
I didn’t think you intended a slight. I also tended to say James only. James is definitely the more visible one. Jules blogs less. I just did notice the paper was both of them, so I wanted to be careful I didn’t slight her!
I know from previous posts, he did some development on a number of different data sets. (Italy and so on.) But I do think in the end the UK numbers are from UK historic data up to a point. Then, at a certain point, he made a projection based on data up to a point. Then, a while later, he tested that projection. So he actually made a prediction and then tested before finalizing the current manuscript.
.
Given that these parameters are time constants in a differential equation, exactly what they mean relative to probability distributions is something that needs to be considered. (He has. He discusses using a gamma (?) distribution when estimating the parameters.)
.
Obviously, he can’t wait forever to collect an infinite number of data and also make a forecast, nor can he wait forever to then publish. But I do think his parameters (like D, L ) for the SEIR model are based on estimating the parameters that fit death data (probably for the UK). He then uses that for the UK.
MikeM
As far as I can tell, you picked either the lowest possible values for D and L you could justify– or even picked ones that were lower than could be justified. And, when knocking D even a bit lower, you failed to notice that the reason for knocking it down would mean you need to increase L so that (L+D) is the total time from infection to either recovery or death.
.
So you justify L lower than 5 days by pointing out people are infectious before they have symptoms. But then you don’t make D the time from when they are *infectious* to when the sicken and die, you make it some other time (like perhaps when they are finally diagnosed? Or something? You don’t really say.) Then there is mystery gap of time between when they are *infectious* and *diagnosed*. Given the limitations of an SEIR model, that needs to either be captured in D or in L. It can’t just vanish, poof!
lucia (Comment #183003)
I just finished reading Annan’s paper and obtaining his R code. As a novice at doing Bayesian analysis (I have done a few) I would recommend this paper to anyone who has some basic understanding of Bayesian analysis for its very clear and concise explanation of what went into setting up the program. The code is well documented as far as I can tell after briefly reading through it.
I plan to load the code into R and see how long it takes me to run it correctly. I might ask Nic Lewis if he has read the paper and his opinion of it.
I would hope that this analysis is acknowledged, critiqued and improved upon by those in the field. Annan’s publication, as a layperson in this field, for the world to see in hopes of motivating those in another field would be a great example of the potential usefulness of internet communication in a laissez faire environment.
Kenneth Fritsch,
Let us know how it goes.
.
As to the ‘potential usefulness of internet communication’: sure, communication that decades ago would likely never happen is very useful. I also think it is good that a well known practitioner in a field many people complain of “churning out dross on an industrial scale”, now points to the dross being churned out in a completely different field. Shoe-on-the-other-foot moments are not common, but almost always informative…. to both feet.
Steve, it may take me a while. Even though I am a retired man projects keep popping up at a goodly rate. And then there is my battles with R code whereby I cuss it out until I find I have made a dumb mistake. My efficiency in using R has improved manifold with my first looking for my dumb mistakes.
Steve, I have found I can learn from the work of those with whose conclusions I do not necessarily agree and especially when they are transparent in presenting their analysis and willing to discuss it with laypeople. I think that is one of the great advantages of the internet.
China “revises” its Wuhan death toll by almost +50%. For anyone who was giving China the benefit of the doubt on its numbers, I think that ship just sailed. NYC did something similar with a deaths at home increase but I think China is guilty, guilty, guilty at this point of controlling all its data releases through an internal propaganda viewpoint.
.
For anyone who hasn’t seen this WHO/Taiwan news clip, it is absolutely hilarious.
https://www.youtube.com/watch?v=UlCYFh8U2xM
.
Any international organization has to dance around the China/Taiwan madness, but this was a bit crazy.
Trump: I have total authority!
Media: Aaaaaggghhhh!!!! Dictator! Ignore Trump everyone!
Trump: Have it your way, the governors decide.
Media: Ummm, what?
.
This little dance is happening over and over. Trump baits the media by inferring the opposite or extreme version of his plan. The media reflexively and emotionally enters Orange Man Always Wrong mode. Trump announces actual different plan, and has now defused all the predictable knee jerk criticism with a simple bait and switch tweet. In this case the screaming would have been “no leadership from the Feds, Trump avoids responsibility”. Maybe this is all just coincidence or Trump operating from gut instinct without any plan.
Tom, I see three possibilities: Trump lied , changed his mind, or just has poor short-term memory. I suppose all three are possible.
In response to T Scharf post #183044
Here is a link to a blogger (laowhy86) who discovered a web posting that Chinese govt has dictated that “Any paper that traces the origin of the virus should be strictly managed.” See 3:35 of this video See 3:35 of this video. https://www.youtube.com/watch?v=a-GVcfP1zrg
Laowhy86 has a Chinese wife who is a doctor and although he has turned vehemently anti-Chinese Govt in last 9 months, he has been a reliable, truthful source along with his fellow Chinese blogger Serpentza (who taught Chinese doctors English) About 2.5 weeks ago Laowhy scooped the mainstream media with a post about how the virus may have originated from an accidental transmission from a lab worker at the virology lab to the nearby wet market. https://www.youtube.com/watch?v=bpQFCcSI0pU&t=2s
More news on Lag time
[…]
They think this could be the infection just lingering, which
If this is true, the “L” in any SEIR model could be quite large. Even yesterday, there was plenty of reason to expect it was bigger than MikeM’s choice of 10 days. But now, even more reason to think even bigger.
https://www.wsj.com/articles/south-koreas-new-coronavirus-twist-recovered-patients-test-positive-again-11587145248?mod=e2tw
Lucia,
“More than 160 South Koreans have tested positive a second time for the coronavirus, a development that suggests the disease may have a longer shelf life than expected.”
.
You just gotta love the MSM. Is that 160 out of 180, 160 out of 1,600, 160 out of 16,000? It matters. They also say these 160 have not passed the infection to anyone else. Is the virus load so low that they are not infective at all? They never even ask the right questions, never mind answer them.
“South Korean health officials and advisers, based on their initial review of the results, don’t suspect inaccurate testing to be a culprit. Test-kit makers and laboratories say South Korea’s kits have a 95% sensitivity to the coronavirus.”
.
Based on this statement, I would conclude the opposite here if there was a significant number of people tested who did not retest positive. From what I read the tests are likely to miss up to 30% of active infections if you have a negative outcome, but if it comes back positive the chances are very high that you have it. Just not doing good swabs can miss an infection.
OK_Max,
I don’t have the energy to track this particular media hyperventilation down, but there is a 4th option that Trump was simply pointing out that Fed law overrides state law in many instances (as state law overrides city law, etc.) and the media interpreted this in the most malevolent way possible, as is common.
SteveF,
They don’t ask questions is certainly true! 🙂
I assume “More than 160” means between 161 and 169.
.
Tom,
Yep. I’d be tempted to think the test-kit makers estimate of the false negative rate is wrong. 🙂
This is surprising: very high incidence of asymptomatic cases among Boston homeless.
https://www.wbur.org/commonhealth/2020/04/15/boston-homeless-population-coronavirus-asymptomatic-universal-testing
A new model update from a few days ago is in. The updated projections for FL changed … ummmm … very significantly.
https://covid19.healthdata.org/united-states-of-america/florida
.
It went from:
4700 projected deaths to 1363 deaths.
Peak death in 17 days from now, to peaks deaths 15 days * ago * (really it is saying peak deaths are now and there was a spike a couple weeks ago).
.
Some of this is the model getting a grip on when the peak really is and is forgivable. The future is hard to predict I hear. It obviously did a poor job of predicting the peak so isn’t taking into account social distance timing and history of the effectiveness of social distancing.
HaroldW,
Yikes!
“Of 397 people tested, 146 — or 36% — came up positive, BHCHP President Dr. Jim O’Connell says. But another finding jumped out at clinicians.
Every one of these folks were asymptomatic. None of them had a fever, and none of them reported symptoms,”
.
36% of the population positive but asymptomatic! Were there a bunch of symptomatic who had already been removed due to symptoms? If not, then the obvious way to avoid complications of severe corona virus disease is to live on the street.
.
Poor countries have consistently low case rates. A Brazilian colleague tells me that the disease in Rio appears to be mainly an illness of the rich and middle class…. the poor people in a shantytown a mile or two away from a wealthy neighborhood are not getting it very much….. and it’s not for lack of possible exposure. Plenty of poor people work directly for rich people as gardeners, maids, car washers, etc. It is to me a mystery.
Tom Scharf,
The models would be funny if they did not have such serious implications: the modelers absolutely do not know what they are doing, but policy is being based on their projections. Every projection they have made and will make is without value. It seems to me unlikely the rate of deaths in Florida will drop off as fast as they now project, but at least they are not now hyping ‘impending doom’ like they have been for the last 6 weeks.
.
The projected death total in Sweden (those bad, bad Swedes who want nothing more than to kill sick old folk!) has suddenly dropped almost in half…. even though the Swedes didn’t change their policies at all. They are sitting at sidewalk cafes, drinking beer and conversing. Does the current projection of deaths for Sweden mean anything? Hell no!
It may be that people who have been exposed to a lot of previous illnesses have some kind of immunity to the new disease that they don’t understand yet. These types of people don’t run to the doctor to get antibiotics every time they get sick, and they are around a lot of sick people. They may have more muscular immune systems or have a very specific type of immunity that maps directly to clamping down hard and early to the coronavirus.
.
Poor communities in India and Brazil should catch on fire, and so far they haven’t.
Yes, the new projections now fall into “too fast decline” that DeWitt has been talking about. It shows the deaths declining at about the same rate as they increased, which doesn’t seem accurate.
Tom Scharf,
“It may be that people who have been exposed to a lot of previous illnesses have some kind of immunity to the new disease that they don’t understand yet.”
.
That would be my guess. There are four relatively common strains of human corona viruses which are known to cause common colds (and in some cases lead to pneumonia). If you had an antibody screening test for these other four corona viruses, then it would be possible to see if people with the antibodies for the other corona viruses are less susceptible to the Covid 19 virus.
Tom Scharf (#183065)
Yes, just took a look at the Florida projections from the IHME model. From claiming that the curve was still rising — and clearly overestimating case rate/death rate/resources — to claiming that the peak is past. From watching the new case data, it seems to me that the new case rate is past the peak, but it doesn’t seem to be coming down very fast. I wouldn’t care to guess whether their current death estimate is high, low, or pretty close.
Tom Scharf (#183069)
Yes, I was also thinking that the homeless may have more robust immune systems than the average American. It’s hard to believe that there are ~40% asymptomatic infected across the board.
Covid-19 deaths per million people in Nordic countries as of 04/17/2020
Sweden 136
Denmark 58
Norway 30
Finland 15
If pub crawling and shagging strangers is more important than saving poor old granny from Covid-19, then the Swedish way of dealing with this virus is the way to go. Sweden embraces the disease rather than implementing ways to limit it’s spread.
Well, to be truthful, Sweden’s high schools and colleges were closed down because of Covid-19, and the government said social distancing might be a good idea. But the Swedish government does not seem aware of our need for intimacy. If you ever try chatting up a babe at a distance of 6 ft in a pub, you likely won’t get even to first base.
Please spare me the lecture on the destruction of the economy. The economy has never been destroyed (obviously) and never will be destroyed unless mankind is destroyed.
Tom Scharf (Comment #183050)
April 17th, 2020 at 10:19 am
Trump: I have total authority!
Media: Aaaaaggghhhh!!!! Dictator! Ignore Trump everyone!
Trump: Have it your way, the governors decide.
Media: Ummm, what?
________
Trump: No, don’t have it your way, if your way isn’t my way.
Tom. he’s doing if again, he continues to revolve.
Trump Foments Anti-Restriction Protests, Alarming Governors
Says today’s NYTimes
Comparing TN and KY again, IHME puts projected total deaths in TN at 231 and KY at 407 and TN has a higher population. So much for how much better the social distancing orders in KY are better than TN. Of course I don’t have very much faith in IHME projections.
Speaking of IHME, they no longer have deaths falling off a cliff for Italy and Spain and have increased the projected total deaths significantly, but I think projected deaths are still declining faster than what is likely to happen. For one thing, new cases aren’t declining very fast and recoveries are increasing very slowly.
In the WSJ:
Also in today’s WSJ another hypothesis on why the racial disproportion in COVID-19 deaths:
On Sweden in Vanity Fair:
Granny can save herself, provided she doesn’t live in a nursing home, by not pub-crawling and shagging strangers.
DeWitt,
I am not sure what to make of the claim of 50 to 85 times as many infected as confirmed (symptomatic) positives. If that were true, then the true fatality rate is very low, and total expected deaths until herd immunity somewhere near 100,000 to 200,000. Not sure that is credible. Perhaps their tests were not selective enough and they are picking up people with antibodies to other common corona viruses.
Max,
Typical.
The economy will go down the tubes temporarily and a lot of people will suffer, and some people will suffer extravagantly, and some increased percentage will commit suicide.
When people say ‘the economy will be destroyed’, it is perverse and stupid to pretend not to understand what they mean. Obviously they don’t mean the economy will be destroyed forever; Christ, has a bigger strawman ever been constructed? NOBODY thinks or means that when they talk about the economy being destroyed.
I get it that there is no persuading you, because I believe you do this deliberately. Tonight I didn’t feel like ignoring you and allowing you to pretend here that the reason nobody was answering was that you were speaking sense.
mark bofill,
Not worth it.
shrug.
Every once in a while, I’ll remark to register a protest on the record, so to speak. For my own peace of mind. I’ll let it go now.
Re DeWitt Payne (Comment #183086)
DeWitt, driving can be dangerous, but driving is not a contagious disease. A better analogy would be STD’s. Promiscuous people believe hopping in and out of bed with lots of different partners is worth the risk of catching a social disease. But even that’s not a very good analogy because promiscuity is voluntary behavior and there’s reward for the risk. People don’t voluntarily get Covid-19 and there’s no reward involved.
Sweden’s three Nordic neighbors, Denmark, Norway and Finland, have a combined rate of 35 Covid-19 deaths per million population compared to Sweden’s rate of 136 per million. So Sweden’s half-assed anti-coronavirus policy is killing grannies at almost four times the rate of it’s neighbors.
Here’s a thought. Those American’s protesting the lock-downs should volunteer to be infected with Covid-19. Apparently, they aren’t very worried about the virus anyway. By acquiring it, staying home a few weeks, and recovering, they would be helping develop herd immunity. Think of this like volunteering for military service in time of war. Patriotic !
mark bofill (Comment #183089)
**When people say ‘the economy will be destroyed’, it is perverse and stupid to pretend not to understand what they mean. Obviously they don’t mean the economy will be destroyed forever; Christ, has a bigger strawman ever been constructed? NOBODY thinks or means that when they talk about the economy being destroyed.**
______
Oh, I know the “destroy the economy” is hyperbole, an exaggeration for effect.
Mark, I believe what got you going was my counter hyperbole as quoted below:
**Please spare me the lecture on the destruction of the economy. The economy has never been destroyed (obviously) and never will be destroyed unless mankind is destroyed.**
Obviously, you find my hyperbole more objectionable than the hyperbole I find objectionable. That’s OK. We can have the same feelings, but for different reasons.
Perhaps I’m being pedantic, but I believe “destroy the economy” is not only an exaggeration, but goes beyond what the word “destroy” actually means, suggesting a misunderstanding of definition.
Webster’s definition of destroy: to cause (something) to end or no longer exist : to cause the destruction of (something) : to damage (something) so badly that it cannot be repaired.
Now for a cheap shot. Your “they don’t mean the economy will be destroyed forever,” suggests you lack understanding of the word “destroyed.” Perhaps you had the resurrection in mind.
It’s not the hyperbole that bothers me Max. It’s that it’s dishonest to pretend misunderstanding of something obvious in order to make a misleading statement. It’s essentially a lie, it’s bad faith, and it’s counterproductive except to mislead people.
To be fair, I have done exactly what you are doing now; I remember arguing that climate models are primarily propaganda based on the literal definition of propaganda. It was an asinine position to take, just as the one you’re taking now is.
I don’t think this is hyperbole:
https://www.forbes.com/sites/sarahhansen/2020/03/24/the-great-depression-vs-coronavirus-recession-3-metrics-that-will-determine-how-much-worse-it-can-get/#6916fe1b15bd
I see little evidence that Sweden’s strategy has been less effective than Denmark’s. Sweden has 4 times as many deaths as Denmark, with about twice the population. But it seems that has been roughly the case since late March, which was before the restrictions in Denmark could have mattered. Denmark’s rate of growth of deaths dropped below 10% on April 6, Sweden on April 9. It looks like cases per capita in the two countries have been quite similar.
I don’t know where I can download data to do a more careful comparison.
———-
I ran across the following in the Wikipedia article on Denmark:
“In late March, authorities acknowledged that the strategy of mitigation had partially worked, but had been less successful than the mass testing in China and South Korea.”
https://en.wikipedia.org/wiki/2020_coronavirus_pandemic_in_Denmark#Lockdown
——–
Addition: It looks like new cases in Norway have been flat since a little before mid-March: https://en.wikipedia.org/wiki/File:Rplot19-1.png
Obviously not due to the lockdown that started on March 12.
Edit to Comment #183115: The log scale on the Norway graph is misleading. This one is more consistent with the lockdown having done something: https://en.wikipedia.org/api/rest_v1/page/graph/png/2020_coronavirus_pandemic_in_Norway/0/30ee11bb5ef054c6fe6e10db97e0a046ba9a69f9.png
But the noise make it hard to tell.
It’s OK to argue and disagree what level of risk is worth it to open the economy, it’s not OK to pretend a position of zero risk is viable and then get on your high horse and look down on others (including 20M unemployed people) who are arguing in good faith what the appropriate risk might be.
.
It is unsurprising and not a revelation that increased risk leads to increased downsides, it also leads to increased upsides (hopefully). Public health officials are incentivized to only argue one side of the ledger, there is another side. As with the immigration debate the media has failed to balance this coverage with anecdotal human interest stories from both sides. It’s a false presentation of greedy corporations versus healthcare heroes.
.
The unemployed might be endangering live-in grandma by going back to work, but they also face financial ruin sitting at home without health insurance and a public health risk that still exists anyway. If the economy was open they can still choose to not work, now they have no choice. The difference is liberty. This shutdown was necessary, but so is opening things back up with an honest debate about trade-offs.
.
I would say that once deaths per day have reached half their previous peak the economy can start being opened again. This looks like it’s 2-4 weeks away in most places.
The Santa Clara study is only 1 of a handful of such studies. There is one from Germany showing similar results. Analysis of the Diamond princess data gives an almost identical result. That’s why I find this result credible.
OK_Max,
Trump is deflecting blame from the fed to the states by giving them the authority. The question is why the states were screaming for that responsibility to begin with, ha ha.
.
Different regions have different problems, and the city/states are in the best position to make those assessments. It’s the right policy, even if Trump thinks so.
OK_Max,
I strongly disagree. Car accidents are not always single car into fixed object crashes. Other cars, drivers, passengers and pedestrians can be involved, so yes, it is somewhat like a communicable disease. And ridiculously low speed limits might temporarily minimize deaths and injuries, but, like the 55mph speed limit, they will be ignored more and more frequently. The analogy to lockdown orders is apt.
Coronavirus Testing Needs to Triple Before the U.S. Can Reopen, Experts Say
https://www.nytimes.com/interactive/2020/04/17/us/coronavirus-testing-states.html
.
The very last people on earth I want placing hard restrictions on the economy reopening are Harvard “experts”, well maybe I’ll trust the Asian ones ha ha. It’s OK to want more testing, but the entire world wants more testing simultaneously with a lot of pending outbreaks in the southern hemisphere. I want $1 test-at-home kits that take 2 minutes.
.
The question is what happens when this desire doesn’t materialize? You can’t wish a testing infrastructure that large into existence in two weeks. The biggest lesson learned here is having a flexible testing infrastructure in place for the next pandemic.
In looking at Annan’s code I see that he used RStudio with rmarkdown to produce an interactive program where it will be a simple task to enter different data, change parameter values, etc. and immediately see the results in graphs and tables. I attempted to run his program without RStudio but that quickly became complicated and thus I downloaded RStudio. I believe the interactive program that Annan intended is what would be ideal for the discussions here. I will attempt to run as time permits his code in RStudio even though I am not familiar with that application. I believe that from his published code could be extracted code to run a Bayesian and even a frequentist analysis but it would not be as pretty as what Annan presents. I am open the suggestions on this matter.
Density is destiny… Policies, not so much.
To continue flogging the deceased equine wrt KY and TN, based on yesterday’s data on worldometers.info, the active cases per million for TN and KY were 497 and 462. TN is going down and KY is going up. Anyone want to bet whether there is a followup article in the NYT?
Speaking of New York, they say intubations are down in NYC, but based on the new case data not decreasing much, I’m wondering whether that is due to a change in medical practice rather than an actual decrease in acute cases.
Absolutely clueless.
.
Without More Tests, America Can’t Reopen
https://www.theatlantic.com/ideas/archive/2020/04/were-testing-the-wrong-people/610234/
.
“How can we close this gap between our needs and current capacity? We need a national strategy over the next 10 weeks, one that draws on the many strengths of our research system.”
.
“Researchers” are going to find it quite distressing when they can’t order 1M swabs from Amazon delivered in 2 days, adhere to HIPAA requirements, get FDA approval for their lab, setup a massive collection and reporting infrastructure, and a thousand other details that researchers have never even thought about.
.
I have dealt with a lot of people from industry and the academic R&D world with my job and they are on different planets, no, they are in different universes.
.
Academic R&D couldn’t put in place an infrastructure to staple 2 pieces of paper together a million times in a 6 months timeframe.
.
I laugh hard when I see articles about how 3D printers are being used! Do people not understand what mass production really is? These people need to watch a few episodes of “How It’s Made”.
https://www.sciencechannel.com/tv-shows/how-its-made/
Kenneth,
I wish I could help, but I am R-mute. I think Willis Eschenbach runs Rstudio.
SteveF,
What we really need is a web based model like David Archer’s MODTRAN page.
https://coronavirus.jhu.edu/us-map > U.S. Map > Confirmed By Population
.
This gives a useful map of the rate of infection normalized to population. That strange hotspot in rural GA is home to one of the largest toilet paper plants in the USA, ha ha.
https://www.wsj.com/podcasts/the-journal/why-you-still-cant-find-toilet-paper/ac752fad-b207-46f2-b3ef-251d578f9cb5
Tom Scharf,
“Do people not understand what mass production really is?”
.
If someone hasn’t ever worked in an industrial environment, they probably do not really understand what is involved. Getting lots of pieces (and people!) working as needed to produce in volume is a non-trivial activity.
DeWitt Payne (Comment #183130): “To continue flogging the deceased equine wrt KY and TN, based on yesterday’s data on worldometers.info, the active cases per million for TN and KY were 497 and 462. TN is going down and KY is going up. Anyone want to bet whether there is a followup article in the NYT?”
.
Indeed. Remember how much Louisiana was in the news a few weeks ago? When was the last story you heard about what is happening there? I’d say, a couple of weeks ago. The Wikipedia article about the epidemic there has maybe two sentences under April.
So here is what has happened: In the last two weeks, reported new cases have dropped by a factor of FOUR. I have not been able to find anything as to why. That ought to be front page headline news.
New cases have also dropped quite a bit in Washington state, but not as dramatically.
Tom Scharf (Comment #183122): “As with the immigration debate the media has failed to balance this coverage with anecdotal human interest stories from both sides. It’s a false presentation of greedy corporations versus healthcare heroes.”
.
In both cases the clerisy is promoting their interests while throwing working people under the bus.
Toma Scharf,
I looked at both maps (death rate and cases confirmed, by population); the cases by population has higher resolution because there are so many more confirmed cases than deaths in most places.
I just figured the death rate is more reliable. But either map tells the same story: nearly all the differences in infection across the USA are driven by population density, not policies, and not by the politics of local elected officials.
Mike M,
“In the last two weeks, reported new cases have dropped by a factor of FOUR.”
.
I have read that Louisianans have the greatest average number of sex partners of any state in the country…. maybe that has changed a bit. Just sayin’. 😉
Mike M.,
I noticed when I updated my spreadsheet that Massachusetts had passed Louisiana in total cases, total deaths and total cases/million. LA is still ahead on deaths/million, but even that may not last. It’s annoying that LA doesn’t report recoveries so the active case number is less meaningful. My guess would be that sorting on true current active cases/million, LA would move even further down the list.
I think the rate of infection is a bit like an SN2 chemical reaction: the rate depends on the product of the concentrations of the two reacting species: infected and not-infected individuals. Once the pandemic is under way, the rate of growth should be proportional to roughly the square of the population density, all else equal.
Edit has an annoying feature: It reverts to the original post, not the edited version, if you need to edit twice.
Tom Scharf (Comment #183131)
Tom, in my former life I worked and managed in many areas of R&D, engineering, marketing, production support, QA and with CEOs and presidents of companies. Being clueless of other operations and fields is an individual trait but also an outcome of specialization. Specialization is a positive feature of a free market economy and not a detriment if utilized properly. It does require communication across fields and a personal understanding of the importance of all these fields and how the fields interact. Clueless for me would be a personal failure to communicate and understand.
“How can we close this gap between our needs and current capacity? We need a national strategy over the next 10 weeks, one that draws on the many strengths of our research system.â€
I believe that governors and mayors realize that, after correctly claiming in our federal system it is they who have control over restarting the economy, there is great political risk in their starting up the economy on their own. Regardless of the trade-offs between getting the economy going and a potential second surge in Covid cases and deaths, the media will emphasizing the negative and will be mostly “clueless” about any positive aspects. The governors and mayors therefore have to get a connection back to the federal government in order to have finger pointing target. Having the bad orange man as a proxy for the feds is an added advantage and especially with the MSM.
Sign at an anti-lockdown protest in California: “We Deem Our Governor Non Essentialâ€.
SteveF
It is in an SEIR model (which Annan uses.)
The lack of domain knowledge is bidirectional. I was shocked how hard production was when I first had to deal with it.
.
3D Printer:
Bzzzzzzzzzzzzffffffffffffffffffffffffffzzzzzzzft ……….
Bzzzzzzzzzzzzffffffffffffffffffffffffffzzzzzzzft ……….
Bzzzzzzzzzzzzffffffffffffffffffffffffffzzzzzzzft ……….
…. repeat 1000 times ….
Bzzzzzzzzzzzz. Clunk. One unit of fragile plastic complete.
.
Production Line:
chunk, chunk, chunk, chunk, chunk, chunk, chunk, chunk, chunk, chunk, chunk, chunk, chunk, chunk, chunk, chunk, chunk, chunk, chunk, chunk, chunk, chunk, chunk, chunk, chunk, chunk, chunk, chunk, chunk, chunk, chunk, chunk, chunk, chunk, chunk, chunk, chunk, chunk, chunk, chunk, chunk, chunk, chunk, chunk, chunk.
1,000 robust units completed.
.
Getting to the first “chunk” is a difficult, long, and expensive process.
Luica,
“It is in an SEIR model (which Annan uses.)”
.
I don’t doubt that. But with very broad-brush treatments (like the whole of the UK or even the whole of a State), the value of Ro varies in a very non-linear fashion with the (very) local population density. The value of Ro likely scales roughly as the square of the local density. So in a country where most of the population is located in a few very densely populated cities, yet on average has relatively low density, the value for Ro will be surprisingly high based on the average population density for the country. Places like Manhattan are basically screwed.
.
What I am trying to say is that it should be possible to account for how variation in population density impacts Ro.
SteveF,
Yes. The SEIR model treat the population a some sort of single well mixed vessel with properties homogeneous through out the vessel. As far as the model is concerned, it’s one big single “thing”. That’s useful for predicting somethings; not others.
.
He does have different “R” for different counties. He does have it change based on strategies like “lock down”.
.
You could compute and compare R before lockdowns in countries and see how that varies with population density. Of course, other cultural factors are going to matter too.
Here is a very sensible criticism of the extreme lockdown strategies:
https://www.spiked-online.com/2020/04/17/theres-no-direct-evidence-that-the-lockdowns-are-working/
mark bofill (Comment #183113).
**To be fair, I have done exactly what you are doing now; I remember arguing that climate models are primarily propaganda based on the literal definition of propaganda. It was an asinine position to take, just as the one you’re taking now is**
_______
And to be fair, it isn’t hyperbole to say “destroy the economy” if you don’t understand exactly what “destroy” means. It’s just ignorance, and I don’t mean that in a bad way. I’m ignorant of the exact meaning of many words, “empathy” being a recent example.
Still, I find it hard to believe most people here do not know “destroyed ” means no longer exists. Those who know the meaning and say the economy will be destroyed are being dishonest because they know damn well our economy can’t be destroyed.
Hyperbole causes counter hyperbole, and discussions suffer.
mark bofill (Comment #183114)
April 18th, 2020 at 6:36 am
I don’t think this is hyperbole:
“This is an economic tsunami,†Mark Zandi, chief economist at Moody’s Analytics, told Vox’s Ezra Klein. “We’re about to see dizzying decline in economic activity,†he said. “There’s no analogue to it in the modern era.â€
_______
No, not hyperbole. As for analogue, WWII in a few ways, unless you consider it not modern era.
Number in WWII armed forces comparable to number unemployed because of Covid-19 and greater in duration? We will see.
Total dollar cost, WWII compared to Covid-19? We will see.
GDP going to military compared to loss of GDP from Covid-19? Again, we will see.
BTW, when referring to the impact of Covid-19 above I mean both the virus and the measures to control it.
Lucia,
I should have said the value of Ro in excess of 1 should vary roughly in proportion to the population density squared.
DeWitt Payne (Comment #183125)
April 18th, 2020 at 9:16 am
OK_Max,
…driving can be dangerous, but driving is not a contagious disease.
I strongly disagree. Car accidents are not always single car into fixed object crashes. Other cars, drivers, passengers and pedestrians can be involved, so yes, it is somewhat like a communicable disease.
________
True, you can be the victim of another driver or a Covid-19 carrier. But if you see the other driver coming, you might avoid him. You never know who is carrying the virus.
Another difference is Covid-19 infections have increased extremely fast and potentially the entire population could be infected, while motor vehicle injuries and fatalities have pretty much leveled off.
I drive having some idea of the risks. Comparatively, I feel kinda in the dark about the risk of catching this virus when I go to public places.
Automobile deaths are quite random if you are not at fault. You can offset the risk with safe driving practices. You can virtually eliminate the risk by staying at home. This is not a perfect analogy, nor is it meant to be. The point is that the public accepts a certain amount of risk for the convenience of transportation and its economic benefits. It understands that they could pay twice as much for highly regulated cars and road infrastructure and be even safer, but chooses a trade off between economics and public safety that evolves with current conditions. People have a certain amount of agency in the cars they buy, where they choose to live, and how much non-essential driving they do.
.
The “killing grandma” crowd refuses to acknowledge this reality and almost universally refuses to set any threshold for this trade off, instead using their self righteous moral superiority as a bludgeon on their opponents. It’s very transparent and tiring.
Further info on the CDC initial test failure, due to lab contamination:
https://www.washingtonpost.com/investigations/contamination-at-cdc-lab-delayed-rollout-of-coronavirus-tests/2020/04/18/fd7d3824-7139-11ea-aa80-c2470c6b2034_story.html
.
“The Washington Post separately confirmed that Food and Drug Administration officials concluded that the CDC violated its own laboratory standards in making the kits. The substandard practices exposed the kits to contamination.”
“The troubled segment of the test was not critical to detecting the novel coronavirus, experts said. But after the difficulty emerged, CDC officials took more than a month to remove the unnecessary step from the kits”
“A spokesman, Benjamin N. Haynes, provided a statement Friday that acknowledged substandard “quality control†in its manufacturing of the test kits.”
The CDC decided to manufacture the test kits themselves. The FDA eventually instructed them to stop doing so because the CDC couldn’t figure out what was wrong with it. The CDC sent out faulty kits that their own quality control guidelines and procedures should have prevented from ever being shipped. The state labs had to tell them it wasn’t working, as water samples were testing positive. A f***up of epic proportions here, heads are going to need to roll.
.
“Jeffrey Shuren, the FDA’s director for devices and radiological health, tells the CDC that if it were subjected to the same scrutiny as a privately run lab, “I would shut you down.—
Heads should roll at the CDC, but that does not mean they will roll. The heads in question sit on the shoulders of our Lords and Masters, a.k.a., civil servants.
OK_Max
Uhmm…. sometimes you know.
Mike M,
Well, civil servants really do have the kind of job security that Trump bragged about (shooting people in the streets without consequence). It is essentially impossible to fire a civil servant. Congress could easily change that by lowering the GS level at which someone can be fired immediately. They won’t. People will get the kind of government administration they deserve if they elect representatives who refuse to change the GS system
200 people on the streets of Chelsea, Massachusetts (a Wuhan virus hot spot) gave researchers a drop of blood for a Wuhan virus antibody test. None thought they had had the disease, although about half reported having at least one symptom (only half?). 32% tested positive for the antibodies.
https://www.dailywire.com/news/massachusetts-researchers-tested-people-on-the-street-for-coronavirus-antibodies-one-third-had-them
MikeM–
Yep. That’s similar to the article I found about homeless people in…?(Boston)?
The tenor of the article was we had to do more for the homeless people. But in fact, none had symptoms.
The possibilities I can think of:
1) This tests has lots of false positives that trigger on another covid-virus.
2) Lots of people have gotten this. It will be over soon, and yes.
lucia (Comment #183173): “The possibilities I can think of:
1) This tests has lots of false positives that trigger on another covid-virus.
2) Lots of people have gotten this. It will be over soon, and yes.”
.
Those would seem to be the options. But the antibodies are in the blood of the test subject and the virus proteins are provided by the test. So I think that (1) means that people have had a different coronavirus infection that has given them antibodies that react to the Wuhan virus. I *think* that means they should already be at least somewhat immune to Wuhan.
.
I am no virologist, but I think that antibodies always react with the structural proteins in the virus shell. Coronavirus has three of those (so says Wikipedia). I would think that two strains might have essentially the same structural proteins but differ as to the non-structural proteins produced when it invades in a cell. Those control viral replication, so they might very well determine how virulent the virus is.
Tom Scharf (Comment #183165)
This is not a perfect analogy, nor is it meant to be. The point is that the public accepts a certain amount of risk for the convenience of transportation and its economic benefits. It understands that they could pay twice as much for highly regulated cars and road infrastructure and be even safer, but chooses a trade off between economics and public safety that evolves with current conditions.
_______
Yes, I accept the risk of the road when I drive and I accept the risk of infection from Covid-19 when I go out in public. As you said “This is not a perfect analogy, but there is a similarity.
There’s also an important difference. Being on the road will always put us at risk for having an accident. Being in public will not always put us at risk for Covid-19, the population eventually becoming immune through infection and/or vaccination when one is developed. I want to be one of the latter, so naturally I’m for buying time
OK_Max,
You have, as usual, completely missed the point and wasted a lot of bandwidth nitpicking irrelevancies. The point was not whether driving a car was like a catching a communicable disease, but how the government might reduce a known risk to a very low value by changing the rules, specifically for auto accidents by lowering the speed limit to a ridiculously low value. The analogy was comparing reducing the speed limit to stay at home orders, not driving a car to catching a communicable disease.
We already know how reducing the speed limit worked out. We are already seeing large public protests about draconian lockdown orders.
DeWitt,
Most of the governors will only act to remove restrictions when they believe it is politically expedient to do so. Vocal protests are but one means to tilt the balance. I suspect letters and emails from voters will further tilt the balance. But the biggest factor will be polling which shows their policies are unpopular with voters. Of course, there are some states where there is no real political opposition (oh say, California, New York, etc). Count on the economic damage to be maximized in those states. Then there are a few governors who are just nuts. I will bet that Michigan’s numbskull governor is not going to be re-elected.
OK_Max,
“You have, as usual, completely missed the point and wasted a lot of bandwidth nitpicking irrelevancies. The point was not whether driving a car was like a catching a communicable disease, but how the government might reduce a known risk to a very low value by changing the rules, specifically for auto accidents by lowering the speed limit to a ridiculously low value.”
__________
Oh, I understand the point of the analogy is the risk tradeoff. But I don’t think it’s a good analogy. As I said the risk of one ( Covid-19) can be eliminated, the risk of the other( road accidents) can be lessened but not eliminated. The former is achieved through buying time at the expense of the economy, the latter is achieved through time loss (lower speed limits) and also has a lower fuel consumption trade off.
Again, I understand the point of your analogy. What I don’t understand is why you think it’s pointless and a waste of time to discuss why it’s a flawed analogy.
Mike M
Maybe. I’m optimistic. Cautiously optimistic… but optimistic!
DeWitt Payne (Comment #183145)
April 18th, 2020 at 10:57 am
Edit has an annoying feature: It reverts to the original post, not the edited version, if you need to edit twice.
______
It’s a tradeoff between looking like a dummy or wasting time.
I tried editing one post four times before finding that out. From now on, I will opt for looking like a dummy.
Chelsea is where Logan Airport is.
Mike M, Lucia,
I asked my wife (who used work in clinical chemistry and developed ELISA test kits many years ago) the chance of the test being misled by the presence of an antibody to a related antigen (eg another coronavirus). She said: “It depends on how well the ELISA was validated, but assuming a reasonably specific test, if a cross-reactivity is detected, then there is a good chance there is some existing immunity.â€
.
I linked to a paper in an earlier thread that reported on a SARS scare at a nursing home years ago when an ELISA test for SARS picked up many positives for SARS, when in fact one of the common corona viruses had spread among the residents and staff…. killing several residents. So it seems to me reasonably likely we are seeing the same thing now with the very high positive test rates. Whether that means there is a substantial population already resistant to the COVID 19 virus, only time will tell, but if I were a betting man, I would bet there is.
MikeN,
Chelsea is near the airport, but separated from the airport by a tongue of Boston Harbor. I think the official address for the airport is East Boston. https://www.google.com/maps/place/Chelsea,+MA+02150/@42.387624,-71.0316494,12z/data=!4m2!3m1!1s0x89e371b266e21107:0x76239b15fba2b915?hl=en-us
Max,
It’s not the common usage. Google ‘cities that were destroyed and rebuilt’ and you’ll find plenty of articles. I find clips on destroyed cars that were rebuilt. I also find articles both on the right and left sides of the spectrum, from Ben Shapiro to the New York Times, talking about the economy being destroyed.
.
Now I *do* actually believe words are important and that it’s a good idea to adhere to the established meanings. For example, I object strongly to playing word games with the meaning of ‘socialism’, even though in common usage ‘socialism’ doesn’t appear to mean what it historically has meant. So, I don’t know how to sort that out exactly.
shrug.
OK_Max (Comment #183158): “Still, I find it hard to believe most people here do not know “destroyed †means no longer exists. Those who know the meaning and say the economy will be destroyed are being dishonest because they know damn well our economy can’t be destroyed.”
.
I find it interesting that Max thinks it an exaggeration to say that Hiroshima and Nagasaki were destroyed by atomic bombs.
Technically, “destroyed” might overstate what is happening to the economy as a whole. But large swaths of the economy are being destroyed along with the livelihoods and life savings of many people.
SteveF
I’m hopeful. But I have to admit that part of my optimism is just wanting to be optimistic.
.
Having said that: I do think that the Illinois quarantine is working. I put up a post with an image of death rates.
.
I think it is very important to separate the two question of
* is the quarantine working and
* is the cost of quarantine worth it?
.
But it does appear to be working.
After updating my R program and all the libraries and running R studio for the first time, I was able a few minutes ago to run the rmarkdown program in R studio which produces those pretty graphs and more details than Annan published. My next step is to attempt to run some Worldometer data that was not included in Annan’s analysis. I won’t be putting any pressure on myself to get it done, but if anyone here who has some familiarity with R and wants results quicker or to just play around with the data, I could give suggestions on how to setup R to run Annan’s program.
Mike M. (Comment #183193)
**I find it interesting that Max thinks it an exaggeration to say that Hiroshima and Nagasaki were destroyed by atomic bombs.**
_________
Saying Hiroshima and Nagasaki were destroyed by Atomic bombs is not as much of an exaggeration as saying Covid-19 measures are destroying our economy. Parts of these cities and parts of their populations were destroyed, but recovery was remarkably fast. Hiroshima, for example, had its streetcars running again three days after the bombing.
**Technically, “destroyed†might overstate what is happening to the economy as a whole. But large swaths of the economy are being destroyed along with the livelihoods and life savings of many people.**
Technically, it would be correct to say your savings have been destroyed ( no money left in your account). It’s happened to me a number of times. Your business could also be destroyed, no longer exist. Livelihood destroyed? That would depend on what you mean by livelihood.
OK_Max,
All analogies are flawed, but some are useful, to paraphrase Box. I don’t think your nitpicks are particularly relevant. Obviously, I think the analogy of draconian and less draconian speed limits is a useful analogy to strict and less strict lockdowns. Btw, a really draconian speed limit decreases fuel efficiency. Fuel efficiency approaches zero as mph approaches zero. Note I used the limit approach to avoid the nitpick that with the engine off at zero mph, the fuel efficiency is undefined. Note that this also applies to electric vehicles.
Tesla efficiency vs speed for different models:
https://qph.fs.quoracdn.net/main-qimg-fdbc7de54fdab7f44de4b90891750920
Can we stop nitpicking the use of ‘destroyed’? It’s really boring.
mark bofill (Comment #183192)
**Now I *do* actually believe words are important and that it’s a good idea to adhere to the established meanings.**
_______
YES, and I could do a better job of that myself.
DeWitt Payne (Comment #183226)
OK_Max,
**All analogies are flawed, but some are useful, to paraphrase Box. I don’t think your nitpicks are particularly relevant. Obviously, I think the analogy of draconian and less draconian speed limits is a useful analogy to strict and less strict lockdowns.**
_____
Nitpicks are cumulative. While I get your point, another thing about the speed limit analogy that I don’t like and even you may not like is this: No Covid-restrictions at all would be the equivalent of no speed limits at all.
A no Covid-restriction policy is where my STD analogy applies. You are welcome to nitpick it.
____________________
DeWitt Payne (Comment #183228)
**Can we stop nitpicking the use of ‘destroyed’? It’s really boring.**
DeWitt, I’ll agree to that if you will agree to stop talking about the speed limit analogy.
Lucia,
Sure, wanting to be optimistic is quite normal (except for the profoundly evil, of course … 😉 )
.
“is the cost of quarantine worth it?”
Almost certainly, IMO, not worth it. The idiots of the world (AKA OK_Max et al) will object, of course, but any rational analysis will show the shutting of the economy is NOT worth it. The issue, at bottom, is if someone is pre-predisposed to having the world revolve around government edict. If so, this is close to nirvana. If not, this is very close to a nightmare.
.
SteveF, I always thought that was just a postal issue. I would go on 16 east all the way to Chelsea, and then take 1A south to the airport. I never realized it because I never went over the water from Chelsea to the airport. Before they put it underground, I think the exit would say airport and East Boston, but it might have been Chelsea. If you weren’t careful, you could miss the airport exit pretty easily and be lost.
MikeN,
It is a complicated area. Certainly the airport is not really in Chelsea… although in truth a good driver of a golf ball could move a golf ball from Chelsea to East Boston or vise versa. Not me, of course, but someone a bit younger and more flexible.
Luica,
“But it does appear to be working. I guess the rational question is “Compared to what?” Not a trick question.
.
Yes, it is clear that social distancing can reduce the rate of transmission dramatically. But the bigger question is if the social distancing rules 1) are worth the cost, and 2) if it is possible to continue without destroying (there is that word again!) the economy.
STeveF
“But it does appear to be working. I guess the rational question is “Compared to what?†Not a trick question.
Compared to having instituted no change at the time of the quarantine and just waiting and seeing.
.
This working doesn’t imply something else might not have worked.
.
But, yes, I mean “worked” relative to just letting it rip. And when they said that, I pointed out the obvious– that it has to “work”. And, I’m no pointing out that the data are confirming that it is working.
.
You may think it is unimportant to point out that it is working relative to doing nothing. I have read people literally saying it “won’t” work, and who say “isn’t working”. Well… it is working.
.
Whether it is working better than some alternative we might have done? Dunno.
Lucia,
Versus just letting it rip.
.
Complicated. And that is not an attempt to avoid the issue. There are lots of places where the ‘cure’ is clearly worse than the illness. What may make sense for Chicago and environs will absolutely NOT be suitable elsewhere. I note the city where my two youngest live (Maceio, Brazil) has a handful of cases (I think 135 as of yesterday) in a city of a million, but poor people in that city and the surrounding region are actually starting to starve for lack of work and income. Really, all of the efforts to control the illness are having terrible impacts on poor people who are not themselves suffering from the disease, only from the only from the government response to it.
.
In case I have not said it before: The entire government response has gotten out of hand almost everywhere, and is us hurting more than helping in many places.
Ohio after two weeks of 300-400 new cases daily has had 900,1100, 1300 new cases this weekend. Just a bit of an Easter bump hopefully.
As a pre-emption to a$$holes like OK_Max, who think everyone is as shallow, venal and selfish as they are: my kids in Maceio are perfectly secure. My concern is for the many very poor kids, in Maceio and elsewhere, who are not secure in the face of idiotic government policies.
SteveF,
I agree that it opens up a lot of questions related to inhomogeneity. Given those, it opens a lot of questions about whether we ever needed to do something state wide. But I do think it’s important to discuss that from the baseline of admitting that at least relative to do nothing across the state, this has reduced the deaths in the aggregate over the state.
.
I think the other questions are important and need to be discussed.
.
But it is also important to recognize that deaths are down, they are down in the time frame related to the lock down and so on. Not.. oh.. it might have been the “concern” that someone felt on March 9!
.
I am also concerned about the economic impacts. But I think we will get nowhere in conversations about ending this if we don’t admit various facts on the ground. Social distancing, enforced by the government, has knocked deaths down. (It’s also knocked the wind out of the economy. I’m also not going to let people say it hasn’t. And I’m also going to point out that poverty kills. See Venezuala.)
New York City is putting a floor on the Case Fatality Rate for SARS-CoV-2 infections. According to the latest count (not including the extra ~4,000 deaths reported recently), 0.12% of the population of the city has died. If we add in those extra deaths, then the CFR is at least 0.17%. Also, it’s putting an upper limit on the ratio of asymptomatic infections. Since 1.64% of the population of the city are confirmed as infected, then the upper limit of the ratio would be 60:1, assuming that everyone in NYC has now been infected, which seems unlikely. Of course if we look at Rockland County, where 3.0% of the population is confirmed infected, that puts the upper limit at 32:1.
DeWitt,
Good to know bounds. But… 60:1 might as well be ∞!
Lucia,
For sure efforts to inhibit spread have reduced deaths, at least in the short term. The question is: at what cost? Places like Chicago, London, Paris, etc may be able to tolerate the economic downside, but many places, like Chelsea, Mumbai, Maceio, and a thousand others may not so easily absorb those costs. So long as even discussing the cost side of the cost/benefit equation remains verboten (and yes, in the MSM it is basically verboten) the policy choices will remain tilted against and directly injurious to those who need to work to live.
DeWitt Payne (Comment #183247): “According to the latest count (not including the extra ~4,000 deaths reported recently), 0.12% of the population of the city has died.”
.
Are those deaths of New York City residents or deaths in New York City? Hospitals, especially critical care hospitals, are often concentrated in the urban core. So the numbers might be inflated by suburbanites being transferred into hospitals in the city for care.
Still, those numbers do seem to support the idea that there was a potential catastrophe.
SteveF,
Discussing the cost side is not verbotten here. But pointing out lockdowns do reduce deaths is also not verbotten.
.
I think we need to discuss both.
I look forward to discussions of both.
I’m trying to figure out what kinds of things trigger moderation.
My post of the following comment went to moderation minutes ago:
“I look forward to discussions of both.”
I suspect it went to moderation because it’s too short. I will now see what happens with more words
Is 10,000 deaths for NYC or for the metro area?
MikeN,
It’s for the five boroughs, not the surrounding counties in New Jersey, Long Island and north of the city.