Coronavirus

[jhm]

US nCFR is ~3%, not 6%. And Iran just makes numbers up.

Yesterday there were 33,331 new cases and 1970 new deaths. That's a rate of 5.9%. When the rate for new cases and deaths is higher than nCFR = total deaths / total cases, it mean that nCFR is climbing.

 

[LN1_Casey]

It's just not inevitable that we have to shut down the economy again. There will likely be an ongoing low-level transmission sparks, and that is unavoidable, most likely. But we need those sparks to decay, not grow. To keep them low/decreasing, we'll need constant testing - for everyone. In addition to contact tracing, phone tracking of all people who choose to leave their homes, etc.

I am concerned that the gov't right now is not focusing enough on this - they are of course focusing on therapeutics and such to help people who are ill NOW. But in order to reopen the economy successfully, there will need to be a very well defined plan that we have a high confidence will work. That's going to require doing something that seems like overkill, and well in advance of actually releasing the lockdown.

Otherwise we'll be stuck in the second wave - and we'll just have to shut everything down again. As I've said before - if I am not sufficiently assured of the success of the screening & testing methods, I will not be returning to work - it's just too risky with this virus, even though I am a fit & healthy 42 year old with no co-morbidities. I will not be the only one making that type of difficult decision.

Yeah, but what they really need to do is start doing antibody tests and not the current active tests. This will show the true status of people and the disease; people who caught it unknowingly, and finished unknowingly can be identified, not just people who don't currently have the disease.

The economy will be in dire straights if we have to go on lockdown again after releasing the masses, but I don't think that will happen. There will be outbreaks, there will be responses, but the populous would be less reactive to the disease as we become more familiar with it, how it spreads, how to mitigate spreading, and our abilities to respond to it. As Overlord Musk has said, panic is dumb. Unknown causes a lot more panic than the familiar, no matter how dangerous.

 

[AlanSubie4Life]

They've had almost exactly 3500 cases over the last month. That means that cases are taking an average of a month to resolve. That seems a lot longer than I thought.
It still doesn't explain why there was no peak in deaths from the peak in cases. My only explanation is that the peak of cases was almost all low risk groups (younger and more female).

Seems like it could be as long as 4 weeks from positive test result to death. Might be longer in Korea due to pre-emptive testing. There's a lot of uncertainty about this, surprisingly. But it also depends on how long it takes a death to be recorded, whether the reporting reports the onset of symptoms or the positive test date, etc.

To me it looks like there is a small peak in late March in the Korea data, but as you say it's probably lower due to the young demographic primarily involved.

The constant low ongoing death rate seems reasonable for a 2% CFR with an average of 75 cases a day, plus a little adder from the people from the initial surge who are occasionally dropping off. Looks like it should settle to a steady state of a couple deaths a day in the next couple weeks.

 

[deonb]

3M N95 costs about $1 to manufacture (i.e. labor, raw materials, machines). Then you factor in shipping and transport costs from overseas. I would say $5 sale price makes sense...

I bought 10 of them (3M N95) last year including shipping for $15.

$5 each doesn't make sense at quantity.

 

[SageBrush]

NY successfully lobbied to allow COVID patients.

I don't know about that, but there is Covid-19 on the ship

 

[AlanSubie4Life]

Yeah, but what they really need to do is start doing antibody tests and not the current active tests. This will show the true status of people and the disease; people who caught it unknowingly, and finished unknowingly can be identified, not just people who don't currently have the disease.

Sure, it's important to do that too - but at the current time it's just a couple % of the population so won't be much help (though definitely it is something to be done). We'll of course need to know that people who are immune cannot be carriers, how long the immunity lasts...etc.

The economy will be in dire straights if we have to go on lockdown again after releasing the masses, but I don't think that will happen.

If we don't use a data-driven approach to properly conduct the reopening, and put the infrastructure in place to deal with the problem, we won't have much choice but to lock down again. There's no way to scale the hospital capacity enough to deal with the flood of cases, and in any case that would not be humane, and would have enormous impact.

The populous would react very poorly to an unmitigated spread. It's not so much panic, as it would be fear. The familiarity with the disease I do not believe would help - there's very little evidence right now that we know how to stem the spread short of complete lockdown. Masks and stuff will help, but there's no evidence it does anything more than reduce the R0 value a little - which only slows the exponential growth.

 

[omgwtfbyobbq]

The deaths per day is a function of the true number of cases. Their model says that there will be 67 deaths on April 15th with a 95% confidence interval of 7-372. That means they have huge uncertainty about the true number of cases right now. If you don't know the number of cases now how can you possibly know the number of cases in the future? Unless you're assuming that the R0 drops to zero.

I'm not sure if R0 will drop to 0, but I think it's safe to say it'll converge to a smaller range of values than in the distant future than it's range of values in the near future. The same likely applies to demographics because the CFR depends on the interaction between them and the local R0.

Because we don't have a lot of certainty about R0 in the near future across the country, the estimate for the CFR in a given region is going to be a larger range of values than if we had more certainty about it. As the virus moves through the whole population, the range of values for the CFR will be constrained because R0 will become more constrained and we already know about the demographics of the country as a whole. In addition, our understanding of how demographics affect the CFR will likely improve with time/more info.

It's kind of like those big funnels you put coins/etc into at malls/etc. Even if there's variation in the object placed into the funnel, and variation of it's position at times Ta through Tn, we know that at some sufficiently large time Ts, the vast majority of objects will have fallen into the bottom of the funnel and we can accurately predict their position. This isn't quite that neat, but the range of likely values for R0 and the CFR as a function of demographics will almost certainly contract over time, which leads to greater confidence in predictions at some point in the future compared to predictions between now and that point.

 

[SageBrush]

I am not surprised that they had another wave. I will not be surprised when the various "recovering" states have another wave later.

Until enough people have been infected and survived, or given a vaccine, and herd immunity can take place--this virus will continue to roll in and out like the tides. The whole "flattening the curve" isn't to not get it the disease, but to get it later, after all.

That is not a fair analysis. Consider e.g. SK with a population of ~ 60M and ongoing cases of ~ 100 per day. That works out to 2.7M cases in the country over a lifetime.

 

[AlanSubie4Life]

I'm not sure if R0 will drop to 0, but I think it's safe to say it'll converge to a smaller range of values than in the distant future than it's range of values in the near future. The same likely applies to demographics and other risk factors. Because we don't have a lot of certainty about R0 in the near future, the estimate for the CFR in a given region is going to be larger than if we had more certainty about it. As the virus moves through the whole population, the range of values for the CFR will be constrained because R0 is constrained and we already know about the demographics of the country as a whole.

It's kind of like those big funnels you put coins/etc into at malls/etc. Even if there's variation in the object placed into the funnel, and variation of it's position at times Ta through Tn, we know that at some sufficiently large time Ts, the vast majority of objects will have fallen into the bottom of the funnel and we can accurately predict their position. This isn't quite that neat, but the range of likely values will almost certainly contract over time, which leads to greater confidence in predictions at some point in the future compared to predictions between now and that point.

Well, the IHME model changed its estimates for deaths overnight by 26% (a reduction). So we can clearly tell it is a quality model. We're down to an estimated 60k deaths now, so I guess this CV is just like the flu. All a bunch of panic over nothing.

Currently it estimates the peak in daily deaths will be in 4 days! I find that to be...dubious. This is based on the fantastic rate we've been adding cases over the last week or so, and the known delay from identification to death. Looks to me like the peak death rate (assuming we actually have reached a national peak in daily cases, which is questionable) is more likely to occur in about 8-10 days from now.

I think their model is being screwed up by the incomplete data from Italy. Who knows how much weight they gave that vs. something from a better tested location, like Germany. Germany very likely has not quite reached their peak death rate yet, though they may be close.

Something is wrong with that model. Not sure what, but it seems detached from reality.

 

[SMAlset]

Newsom was asked about whether Calif's contract for securing 200M masks was commandeering other states/hospitals/etc supply and getting in the way of the Federal government's efforts (saw today they..FEMA included..were commandeering orders paid for by State and local healthcare facilities--article mentioned reports from 7 different state facilities who were bewildered after being told fend for yourself and pay the going rate*). Newsom said quite to the contrary; California has been working closely with federal agencies and these PPE supplies were not doing that. He went on at some length to explain how there were joint efforts.

*LATimes article: Hospitals say feds are seizing masks and other coronavirus supplies without a word

 

[SageBrush]

Yeah, but what they really need to do is start doing antibody tests and not the current active tests.

I'm willing to wager that the national exposure has only reached 1 -2 percent.

 

[jhm]

The graph I posted is daily deaths. This is Switzerland from the link you posted

View attachment 530264

You can see that the daily reporting is choppy, but if you smooth it out it appears to be that they are at the top of a predictable 'bell' curve. The Worldometer is also under-reporting compared to ECDC

The daily death peak is important as it should be about the time half of ultimate deaths are observe. Conservatively, I'd use April 4 as the peak. At that point total deaths were 666. So perhaps Switzerland will see only about 1332 ultimate deaths.

It is tempting to think that maybe the death peak was on March 31. But here there are 433 death. Twice that is 866, but total death are already at 895 and going strong. So clearly ultimate deaths is substantially more than twice 433.

 

[LN1_Casey]

Covid-19: death toll by age groups in Brazil, Italy, Spain and the US

Here is an interesting website that compares Italy, Spain, USA, and Brazil in reference to the age group in relation to COVID-19 infections and deaths. It was published on the 3rd, and USA numbers are from the 23rd--but it shows a better depiction of correlation between age and the virus than the Wordometers did with their flat percentages.

Italy, for example, 95% of the deaths were ages 60 and older (2 April). The jump from 50-59 to 60-69 is three times as many deaths (479 to 1448), and 70-79 triples again to 4196, and 80-89 doesn't grow exponentially again (only to 5029) because by then there just isn't as many in that age range to begin with. 60-69 range is a bit over 7 million, 70-79 is just under 6 million, and 80-89 drops noticeably to 3.5 million.

The website shows they're 56.1% of the infected, but maybe that's just because they went in to get tested because they felt the symptoms/needed to be incubated. Is Italy doing the antibody test, or the active test? Maybe there are far more asymptomatic people than otherwise assumed.

 

[jhm]

I bought 10 of them (3M N95) last year including shipping for $15.

$5 each doesn't make sense at quantity.

Would you be willing to sell your masks for $5?

 

[omgwtfbyobbq]

Well, the IHME model changed its estimates for deaths overnight by 26% (a reduction). So we can clearly tell it is a quality model. We're down to an estimated 60k deaths now, so I guess this CV is just like the flu. All a bunch of panic over nothing.

Currently it estimates the peak in daily deaths will be in 4 days! I find that to be...dubious. This is based on the fantastic rate we've been adding cases over the last week or so, and the known delay from identification to death.

Something is wrong with that model. Not sure what, but it seems detached from reality.

All models are detached from reality.

In general, I think most epidemiological models will have more variance in predictions earlier in an event and less later, which may seem counter-intuitive, but should be normal. I would be more concerned if the change in predicted deaths increased over time because that indicates a problem with the model assumptions, assuming of course everyone continues to act rationally. If everyone decides to F the shelter in place idea and instead have huge foam parties, all bets are off.

The CV is just like the flu in terms of deaths if large parts of the country shelter in place for months with the CV and we don't do anything for the flu. The problem is, because CV is so contagious and appears to have much higher CFRs/IFRs than the flu, if regions stop sheltering in place too soon it can and will blow up in those regions and lead to more deaths than the flu would have with no sheltering in place.

We can try to eyeball where that dividing line is, but if we get it wrong we can easily see situations where we have many more deaths from COV than from the flu, which appears to be what's happening in Italy, Spain, the UK, New York, and so on. They didn't shelter in place soon enough and now they're getting hammered.

Someone can try to come out of shelter in place sooner to lessen economic damage, but if they misjudge that and come out too soon they'll get hammered by COV sickness/deaths and still get hammered economically anyway compared to if they had kept up the shelter in place for another couple weeks or months and had more near term economic harm, but next to no long term sickness/death. If they come out too soon they could get more sickness/death and more long term economic harm.

 

[AlanSubie4Life]

The website shows they're 56.1% of the infected,

but maybe that's just because they went in to get tested because they felt the symptoms/needed to be intubated.

Yes, that's likely why. Italy is very probably extremely undertested. There's no reason to think that the true IFR for this disease is any higher than 2-3% (I would guess it is closer to 1%, but it depends on the demographics of course - Italy has an old population). So undertesting is a huge problem.

All models are detached from reality.

Some more than others. Would be good to have a mechanistic model in this case.

I would be more concerned if the change in predicted deaths increased over time

That seems to be what is happening. We should start tracking the daily death prediction deltas in absolute & % terms over time, I guess.

 

[LN1_Casey]

That is not a fair analysis. Consider e.g. SK with a population of ~ 60M and ongoing cases of ~ 100 per day. That works out to 2.7M cases in the country over a lifetime.

SK also had a competent government response very early on, and a populous not as headstrong/diverse as America. Further, they are basically an Island (NK being an non-traveling landmass) and thus have more control over the diseased travelers that can enter the country by either airlines or ships--much more easily controlled than by checking every passenger and every car to drive over like it can in the EU or US state to US state/Canada/mexico.

Still, I think once other countries let up on their controls, SK may have more incidents popping up as travelers think "they're fine" but are really asymptomatic, or their own residents travel overseas for work and return as carriers.

 

[N5329K]

More than one way to test for community infection rates:
New research looks at wastewater to detect community spread of Covid-19
Robin

 

[sparcs]

N.J. supermarkets, stores must limit customers to 50% of capacity, all inside must wear masks under new coronavirus restriction

Gov. Phil Murphy announced Wednesday he is requiring all employees and shoppers at the businesses still open in New Jersey to wear face masks or coverings until further notice to help fight the coronavirus outbreak and has limited the number of customers allowed inside stores to a max of 50% of their capacity.

Workers will also be required to wear gloves if they’re interacting with customers, Murphy said. Businesses are required to give workers masks, coverings and gloves at the businesses’ expense.

 

[Daniel in SD]

In general, I think most epidemiological models will have more variance in predictions earlier in an event and less later,

This may be true but for a novel virus where the ultimate number of deaths is a function of our mitigation efforts it makes zero sense.
The deaths two weeks from now is function of how many cases there are today. If there is uncertainty about how many deaths there are two weeks from now that mean that there is uncertainty about how many cases there are today. A change in the number of cases there are today will change how long it takes for the number of cases to drop to zero.
And that's neglecting the fact that if the number of cases today turns out to be on the high end of their range it probably means that their estimate of the basic reproduction rate is wrong which would further extend the curve.
Any model that says there will be ZERO deaths from COVID-19 in CA on May 20th 2020 is garbage. I will give anyone even odds on the other side of that bet (which is a great deal since they say that it has a 95% chance of happening!).