Predict the number dead from COVID-19 in America by end of 2020

The Quarantine Zone
121

I’ve been hesitant to make a prediction but I’ve been very confused by the predictions I’ve been seeing. If we take 200K deaths as the number and a 0.5% death rate the lowest I’ve seen any where that would imply that 40mm people got the disease. What I don’t understand is why the other 290mm people don’t catch it. The way I see it the only way that happens is we social isolate to the point that everyone who has the disease can’t pass it on to someone else and the infection stops or we figure our a vaccine and are able to get the rest of the country inoculated before it can get to them.

All of that assumes that our health care infrastructure stays in top condition and they are predicting Colorado will be out of ventilators in the next 7 days and out of hospital beds in the next 14 so even 0.5% seeps optimistic.

My understanding is flattening the curve just drags out the deaths over a longer period of time and the hope is to keep it low enough that we don’t add extra deaths. So even predicting 40% of the population catches it before the vaccine is developed and we maintain a 0.5% death rate due not over loading hospitals it seems we’re looking at 660K deaths between now and next June. If we assume that the death rate stays flat over the next 8 months I don’t see how we have less than 350K deaths between now and the end of 2020 but that has the stay at home orders in place until a vaccine is developed which doesn’t seem likely.

My honest guess is 60% infection rate and 0.5% death rate or right at 1mm deaths and I think the stay and home orders will end this summer and not be able to be reinstated this fall so I’d guess that occurs this year.

122

No reason to worry. It’s only a couple more weeks until Easter. That’s when Jesus will return to save the world. Err, provided one is Christian of course.

123

DSeid, you keep referencing the shape of curves and so on, but do you acknowledge that your prediction requires 96% of Americans to never contract even an asymptomatic case of the virus?

124

How many Easter bunnies will be sacrificed on the altar of… oh, forget it.

I’ve been in massive Holy Week celebrations in Amalfi Italy and Antigua Guatemala. The latter, with up to a half-million visitors jamming that shakened town’s narrow cobbled streets, hosts the fattest Semana Santa fest outside Santiago Spain. I’ve not found church notices banning gatherings generally and I’m not sure the faithful and curious would all stay away anyhow. Specific dioceses have halted parades and public mass. But I fear poor communities will seek the usual comfort of closeness. The Vatican “suggests” postponing processions until September. Will the faithful wait?

125

No I do not.

In fact using the sets of assumptions that I have felt are most likely true I come with not dissimilar results.

Diamond Princess Cruise was not average adult demographics but way overweighted to a senior demographic. People want to argue that maybe a higher SES and healthier senior demographic, well maybe, but age is an independent risk factor for mortality and severity of infection. The fact that in a cohort overweighted to a higher risk elderly demographic the infection mortality rate was, per your previous post, 0.66%, tells us that that is the number for a high risk demographic which is likely to have many fewer asymptomatic and mild infections than younger lower risk groups are.

I also do not believe that the rate of infections in Hubei with cases believed to have begun in late November, with no major public health intervention for two months from its likely onset, resulted in a grand total of only about 0.1% of the province becoming infected (67,802/59,000,000). Maybe two to three orders of magnitude off seemed more realistic at the time and now.

How can we get a sense of what the number might be?

Well you look at a flight of 126 German nationals who returned from Wuhan on 2/1/20. Of those 126 there were two that tested positive as acutely infected (1.6%), none who had symptoms. No way to know how many had had mild to asymptomatic infections already resolved. So minimally acute asymptomatic infections are not so rare that going two for two is unlikely.

Using an SIR model right now requires picking assumptions mostly pulled out of the air because actual data that can be believed is lacking.

IF I was to attempt a model I would at least model it under a range of possible assumptions. I would ignore “confirmed infections” having any model meaning at all.

One set would include that the death rate of those with infections across all with identifiable disease but untested and not labelled is 0.1%, and that for each of them there are maybe 9 asymptomatic to unsuspected mildly symptomatic cases.

I would in this set also posit that the 24% of Americans who are children function as if they are Resolveds in the SIR model by virtue of low contagiousness. That herd immunity comes into play for this disease at 40% of the population functioning as Resolved. And that under social distancing guidelines younger lower risk individuals are more likely to among those less compliant and among those in the “essential” workforce, experiencing the larger share of infections first, with higher risk individual less likely to be noncompliant with social distancing rules and less frequently part of the “essential” workforce.

We need to add 16% more of the total population to get the total in the Resolved bucket to 40%, which means 21% of all adults. As noted it seems probable to me that young adults be in groups to get infected more, and to have higher asymptomatic rates, but lets just go with the overall numbers above. Adults are 76% of the U.S. 330 million population. 0.760.21330*0.01=53,000 (rounding up). Seems low.

Do I have any solid basis for those assumptions? No of course not. But neither are they unbelievable ones out of keeping with what we do actually know. Until seroprevalence studies are done we have no real idea about how many cases of SARS-CoV-2 infection are asymptomatic/very mildly symptomatic to diagnosably having COVID-19 and what the true infection mortality rate is. We do not know for sure that kids are not contagious (but we have some evidence to suggest that they are not very contagious at least). We don’t really know at what point for this disease herd immunity would play a role and to what degree under social distancing being in effect the bulk of those infected would be in the lower risk groups and to what degree the higher risk groups can be protected.

I cannot assume my chosen assumptions are true or false, maybe herd immunity for this disease needs 45%, no one really knows, and we cannot assume the assumptions chosen by the modelers are true or false either. We are simply missing the key critical inputs. I’d bet mine are not exactly right. I’d also bet against the other sets of assumptions as well. But I do believe the growth curves similarities because they are data: day 12 after 1 death/million Italy was increasing at 26% a day; day 12 for New York it’s 27%.

Just like kids don’t always follow the exact percentile growth curves I don’t expect these to stay lock step from here. But given that all the Western countries before us have followed the general shape I would be very surprised if we did not

126

Speaking of Italy, New York just passed Italy in deaths per million.

The US death toll is still doubling faster than every 3 days.

127

8,
possibly 9 on that list

128

?

No it didn’t.

In NYC the current death number is 1374 with a population of 18.8 million, 73 deaths per million. New York State is running 79.7. Italy is 206. It’s ahead of where Italy was for that many days into being past 1 death per million though, if maybe that’s what you meant. But it’s been there literally from day one.

129

The total for the state is 1941 dead. But you’re right. I had the city population plugged in. So they’re at 99.8 deaths per million vs Italy’s 217.8. And there is some question as to Italy’s accuracy.

130

Okay, DSeid: so you think the true death rate is far, far below 0.66%. Interesting! Okay, that’s clarifying.

I do wish, even if they don’t have the capacity to just go test everyone everywhere, they would try to do some kind of experiment somewhere like on one of the Hawaiian islands or somewhere people are relatively “captive”. Maybe even among prison populations? Or even in a relatively isolated but not “captive” community. Just go around and try to test everyone and then get a sense for what the true death rate is.

Because if your hunch is right, it will be very good for the people who don’t die from this, obviously. But it will actually be troubling in terms of the impact it has on our political and public health systems. The people crying “hoax” or “overreaction” will have some justification for saying so (it’s highly questionable as to whether we should be on this level of lockdown if you’re right, TBH); and probably worst of all, we won’t get really prepared like we need to be for the potential of a really nasty future virus, one that has a SARS-like death rate but is far more contagious. And experts say that is likely on the horizon, and that we are not ready for it. I have been feeling that a silver lining of Covid-19 is that it will be just bad enough to serve as a wakeup call, without being so bad that the wakeup call has come after we’re already really screwed. So if this ends up somewhat fizzling out, we could learn all the wrong lessons and millions could die next year, or next decade, as a result.

131

AFAIK, the historical evidence is that viruses do mutate and a subsequent strain can be milder or more dangerous. Pace the Spanish 'flu, where the first wave was bad but tolerable, the second wave was the killer, and it tailed off in the third wave. This may happen again, but I’ll leave the details to the virologists.

This pandemic would slow down if a less infectious strain emerges, but I am not betting on that, or that it will become milder.

On looking again at the graphs yesterday I see the main European countries now on a straight line but still climbing, and the USA appears to still be an exponential curve, i.e. growing at an increasing rate instead of a straight line. South Korea is still increasing slowly, but on a straight line. What are they doing right in a very crowded country?

The graph indicates to me that the numbers will climb rapidly in the USA and the question is whether the medical facilities of the big cities will be overwhelmed; from what I have read, the main area affected in NY. The country boys can ride out the epidemic in splendid isolation, until they have to go shopping.

One thing about estimating the death toll is that the estimates for the number of who died of the Spanish 'flu vary hugely. What is 25 million here or there? Why the gross lack of precision?

132

No it would not, unless this new strain is more infectious but less lethal than the current one(so that everyone gets IT before they get the current virus), AND it still provides immunity to the current one after recovery.

Both of these requirements are highly unlikely.

Much more likely, it would just give you the opportunity to catch both covid-19 and covid-19b at the same time.

133

Predictions are pretty much guesswork since the administration has been slow walking testing from day one. They don’t want the numbers to be high, that would make Donald look bad, so their solution is to put as many roadblocks in the way of testing as possible. That lack of data hampers the fight against the virus, but it is politically expedient and that’s all that matters. The tests are simply never going to be as available as they should be.

134

Scientific fact is that mutations happen all the time, but the vast majority of them are neutral or deleterious. In other words, the mutant either doesn’t survive, or doesn’t change in any real way. In fact what we mostly see is that every year, the flu mutates just enough to dodge the vaccine we made last year.

It is possible that we could see a virus with different characteristic emerge, but it is not guaranteed. And it’s a roll of the dice to even guess what those different traits might be.

This is false. If a less infectious strain emerges, it would be crowded out by the more infectious strain, unless there was cross-immunity with the other strain, which also is not a guarantee.

Don’t bet on a mutation being an ace in the hole. Most mutations won’t have any impact to humans. The few that do, will make things worse.

135

Let’s continue that table of yours.(including correcting the last line)
3/29/2020— 363
3/30/2020— 558
3/31/2020— 912
4/01/2020— 1049

You were definitely seeing an anomaly, largely by using “today” figures before the day is done.

The curve has slowed down a bit, but this is so small it is likely swamped by the noise and granularity of the data.

Over your period the growth rate was +27.7% per day
Over the same period plus including the following 3 days, this ‘dropped’ to +27.3% per day.

Look at the graph on United States COVID - Coronavirus Statistics - Worldometer
Specifically, look at the graph labelled “Total Coronavirus Deaths in the United States”, with the display set to “logarithmic”

As long as that line is at a constant angle, the death rate is growing at a constant factor. if the curve flattens out, death rate growth is slowing down. If it curves up, the growth is accelerating even more.
Right now, that is a very, very straight line. I.E. no change in the death rate growth over the past month.
(actually micro worsening around middle of month, micro improvement last 2 weeks… but very very slight changes only)

136

That type of data is what informs my prediction.

And scares the daylights out of me.

137

In reverse order - I again feel the need to emphasize that 85Kish, mostly all occurring in one month, would be a VERY BIG DEAL that is far above healthcare system capacity. It is a nationwide Cat 5 hurricane.

I do think that very substantial action is justifiable if that results in flattening that surge to over 4 months instead of all in one. Knowing for sure which precise action is actually required to do that requires accurate information not in anyone’s hands, and we have seen its initial rate of rise is overwhelming without early action, so responding fairly all out while awaiting and actively collecting that information is completely the right thing to do. We have the information we have at the time we have it and need to act (and inaction is still an act) in the face of uncertainty and doubt. We can pull back from a stronger response as (and if) evidence comes in that less restrictive measures will slow it down enough from the point we are at; we can’t put them in early when they would have mattered most after the fact.

And I realize that selling the somewhat nuanced position above is hard to do, especially in the current political clime.

So publicizing models run with only assumptions that make the case most starkly may be the only way to get the required action. But yes there is risk in overstating certainty in those assumptions and the model results based on them.

Testing everyone with the swab still wouldn’t give you the information you want. That tells you how many are actively infected at that precise time, but not how many have BEEN infected or how long it took them to clear it. If the asymptomatic and mildly symptomatic stop shedding sooner than the symptomatic (likely), then it would also still underestimate the fraction of all infections that are in the former groups. Again a random sampling of even just several hundred in different age cohorts, best done in a place that is not in the rapid increase phase, correlated with histories of what, if any, symptoms they recalled having, would be enough to give the critical information we need within some acceptable confidence intervals.

Again though the prediction method I am proposing does not need that absent critical data. It doesn’t care if the germ infects fewer and spreads less broadly but is more deadly, or infects many more but is is much less deadly. It is based exclusively on seeing what has come before us in countries similar to us that have responded with similar sorts of actions in similar sorts of timings.

138

Yes and at the same point, 10 days after hitting 1 death/million, every Western country has been on the near straight line portion of the curve. At that mark the exact rate has varied but it was mostly straight for each.

And every one has started to have that curve begin to flatten by 5 days later gradually flattening more from there.

You really cannot look at a baby’s growth from birth to 3 months and just draw the line straight from there to predict its future height and weight. You do better if you look at the growth pattern of other kids before them to inform your expectation.

139

Well, the US hit 1 death/million around March 22 - so about 11 days ago. So the curve should be flattening in the next 4 days or so if you’re correct. We’ll see. I think the highly politicized response and geographic realities of the US are going to make its curve a little different.

140

Exactly. The prediction is falsifiable in the near term.

Meanwhile New York is ahead of the rest of the U.S. in having hit that mark by 3 days and has been showing some flattening between days 10 and 13. Not sure it will drop to the 20% mark by day 15 that I had predicted or not but it is flattening, right on schedule with the other Western outbreaks.

Please note - this does NOT state whether this happens due to natural history of the disease or because Western nations all seem to follow the same time course of social distancing responses both by individual choices and by edicts. Either could be true, or some combination. In that regard the difference in curves between Sweden’s contrarian approach

and the rest the Western nations, and very similar neighbor Denmark which intervened hard and early, will be informative, and over the next several days. Sweden is so far, at day 13 past that mark, NOT slowing down, still at 33% growth. Maybe they will in the next several days but if they get much beyond day 15 without slowing down then it is strong evidence that the fuller social distancing measures are very much necessary. If they DO slow down it gives us some reason to consider relaxing measures as we get under more control. And some sense of what the payoff is on the full out social distancing side of the calculations. Denmark for their part may be showing how being ahead of the curve is valuable. It looks like they began to slow down on day 9 after the 1 death/million mark and now at day 13 they are at 16% growth. A few more days will demonstrate if that is real or just noise.