All the good stuff, caveats, code, and explanations are linked, some in Update III, and the most important in Update II, Update IV, and Bayes Theorem & Coronavirus, so go to them first before asking what-about-this. Skip to the bottom for the latest model.
The thinking seems to be this: It was horrible in Wuhan, it is horrible in Lombardy, and horrible in some other country nobody gives a damn about. Therefore, when it happens to my town—and it’s when, not if—it will be equally horrible everywhere. And worse than horrible, because we are not as prepared or as willing to be as draconian as the Chinese. It isn’t horrible this moment, right now, here, but because it will be horrible means it is already horrible.
That seem a fair summary? One survey found “1 In 5 Americans Expect They’ll Be Diagnosed With Coronavirus“.
None of it is right.
In Wuhan itself, the City of Doom, some 2,446 souls departed their fleshly existence earlier than expected. Google tells us the city has between 11 and 19 million, depending on whether you count the entire metro area as “the city”.
The city had 49,995 cases. The case rate was 0.26% to 0.45%, depending on what China called “the city”. The total dead rate was 0.01% to 0.02%. The case dead rate was 4.9%.
People fixate on that last number, forgetting you first have to get the bug before you can die from it. But everybody now seems to believe they’ll get it with certainty. Review Bayes Theorem & Coronavirus!
If everywhere will eventually be as bad as Wuhan, then, given 7.7 billion of us now speak with authority on “social distancing” and “flattening the curve”, as if we’ve been using these neologisms from birth, from 20 million to 35 million the world over will get the bug, and from 1 million to 1.7 million will croak from it. (Fifty million died from 1918’s Spanish flu.)
That will beat the yearly regular flu total, which, on an average year, only performs half that good. Poor flu! It tries hard, but not every disease can be above average.
Did somebody say flu? CDC on week ending 7 March (again, they’re always a week behind on these):
CDC estimates that so far this season there have been at least 36 million flu illnesses, 370,000 hospitalizations and 22,000 deaths from flu.
That’s in the United States alone, friends. Mostly pneumonia and old folks (God bless them). Worldwide at least ten times that number.
If the Wuhan numbers apply globally, losing a million or two of us worldwide is not welcome news, but it’s not panic-level, end-of-the-world, buy-every-roll-of-toilet-paper-in-sight news. It’s wash-your-damned-hands, stay-at-home-if-you’re-sick news.
Right across the Yangtze from Wuhan is a city named Huanggang, still in Hubei province. They had 2907 cases and 125 dead. Google says 7.5 million population. That’s a 0.039% case rate, and 0.0017% dead rate. And again a high 4.3% case dead rate.
Applying Huanggang to the world gives about 3 million cases and maybe 130,000 dead bodies.
Shanghai, you ask? The city about 500 miles east—and downwind—of the City of Doom? Shanghai, clustered and claustrophobic, had 352 cases and 3 dead. Shanghai has 24.2 million souls. That’s a 0.0015% case rate, 0.000012% dead rate, and 0.85% case dead rate.
Applying those numbers to the world gives 112,000 cases and 955 dead bodies. We’re at more for both, so obviously applying Shanghai to the world doesn’t work.
Neither does Wuhan!
There is no evidence to suggest that everywhere will be as horrible as Wuhan or Lombardy. There is lots of evidence it will not be as horrible. There is instead evidence that many places will be like Huanggang and Shanghai, and somewhere in between, and that there will be only a few Wuhans and Lombardys. Look all around Wuhan and the rest of China and East Asia for confirmation.
Not only that, but look at China as a whole. Their case number has barely budged since last week, and the number of dead has fallen to about 0 (some days none, some days a few). And it wasn’t like every single city in China welded people into the apartments, or whatever.
China has 1.386 billion (Goog), 80,860 cases and 3,213 dead (these are mid-morning Monday numbers). That’s a 0.0058% case rate and 0.00023% dead rate, and a 4% dead case rate. Applying to whole world (China included): 450,000 cases and 18,000 dead bodies. Applying to just the USA! USA! USA! gives 19,100 cases and 750 dead bodies.
China is not homogeneous, neither is the USA, and neither is the world. So none of these extrapolations taken without considering uncertainty should be trusted.
“Briggs, you fool. I tried to tell you last week that China is lying about their numbers.”
No kidding? How do you know?
“Look at everybody running in circles! This proves it’s worse than they’re saying.”
That so? How does prove the numbers are wrong?
“Because they’re lying about their numbers!”
Maybe, but the numbers are all we’ve got.
I do find it touching that so many people have such a simple, child-like faith in medical statistics. Many seem to think they are as accurate as your counting the number of cars in your driveway. It’s much more like trying to guess the number of pennies in a jar that’s been covered by a sock, and which you can only see from the couch on a high bookshelf.
To be fair, this simple faith is shared by a large number of researchers. Which accounts for all the breathless headlines we constantly see. And the studies we regularly take apart.
Here’s some sobriety from reminiscing about the 1957 pandemic. (Believe it or not, there were pandemics before this one.)
Flu pandemics recur reliably but unpredictably every decade or so, and their extent and intensity varies. With COVID19, we may be in midst of a once-every-50-years event, perhaps similar to 1957 pandemic, but not as bad as the 1918 pandemic. Let’s talk about the 1957 pandemic. 1/
— Nicholas A. Christakis (@NAChristakis) March 14, 2020
And how about that 2009 pandemic? Forgotten already? Why didn’t we freak out then and do now? Twitter et al. wasn’t what is was then, maybe.
The H1N1 flu killed about 12,500 Americans in 2009 (there’s some +/- here, considering we also have the regular flu); so far, coronavirus has slaughtered fewer than 100 (79, Monday evening). Curiously, Washington was also disproportionately represented in the H1N1 breakout, as it is in this one.
The ordinary, and so far much deadlier than the coronavirus, flu follows a seasonal pattern for the reasons you already know: when it’s warm and sunny outside, people stop clustering together inside and get out in the healthy air, a natural germicide.
No, it is not guaranteed sunlit warm air and outdoor living will kill or slow this virus, but it’s where to put your money.
Here’s the thing: I’ve seen dozens of explanations for the origin of this new bug, more dozens of explanations of how it spread, theorizing from everybody on how to stop it. Nobody is blaming—or crediting—nature. Everything bad is some person’s fault, and every success is because of some person or persons. Many are sure their personal level of concern is inversely proportional to the rate of spread. There’s too much credit going around on either side. If you think the politics now is bad, wait until this is over. The blamestorming and trophy-giving will be truly nauseating.
We don’t get this with the flu, or the common cold (sometimes also a coronavirus). These outbreaks always stop, even without active intervention, except for the initial vaccinations (for flu) and admonitions to wash hands. Natural causes account for most effects (good and bad), and we’re happy to say so. Because flu and colds are mundane, and coronavirus is new and panic-worthy.
“Briggs, you’re an uncaring ass. Your advice could make people not care, and not caring kills. You are personally putting old people and children at risk!”
Well, no kids anywhere have died from coronavirus. How about you? You drive, right?
Some 37,000 people a year die horribly and painfully in motor vehicle accidents. The only way to prevent this is a complete shutdown of the nation’s roads, highways, and byways. Every time you get into a car you are selfishly putting children and old people at risk.
That’s just the USA. WHO estimates “1.25 million people die in road crashes each year, on average 3,287 deaths a day.” On top of the deaths, in the USA “An additional 2.35 million are injured or disabled”. Driving is deadly.
About 22,000 people this year died wheezing and strapped to machines of the flu. The only way we could have prevented this was a complete shutdown of all social gatherings, all restaurants, all schools, everything. Every time you go outside you are selfishly exposing your neighbors.
“Those are different.”
“You don’t understand. This is new.”
It seems no one is willing to admit to being responsible for having somebody even get sick in their jurisdiction—the threat of lawsuits in American is always huge—whether this is a political or commercial jurisdiction. Shutting it all down removes blame. Forget who it ruins or what it costs.
Also, part of the terror comes from the clustered nature of the deaths, as well as the newness. Flu is spread out. This virus, being new, hits and runs.
On that newness, this wonderful graph of media-induced panics, contributed by reader Harry Goff. Remember how we were all going to die from ebola? Zika? Bird flu? SARS? MERS?
One last word about good numbers, or all numbers. All numbers are conditional. The Wuhan numbers were condition on Chinese behavior, biology, geography, economics, food, everything that it causally involved. Extrapolating Wuhan numbers to everywhere says everywhere will be like Wuhan in these causal aspects.
There is no single unique case dead rate! Or there is, but it’s 0% or 1% for each person who has the disease. Please be careful in extrapolating.
The faith folks have in numbers is nothing next to the absolute conviction that medical tests are perfect.
One of Trump’s advisers said they were developing point-of-care test kits, which might be made widely available. In the hands of a real doctor these can be fine. But they are more inaccurate than lab tests. Trump’s advisers rightly said that not everybody should rush to be tested. The number of false positives and boost in panic that these would cause can only be imagined.
Some guy on Twitter points us to this Italian study of swab testing. They had (Table 3) 32 people tested twice. Results: 18 tested negative both times, 6 tested positive both times, and 4 each inverted the results, a 25% mismatch rate; 12.5% switching either way.
I reiterate: mass testing asymptomatic people is nuts because of false positives, and would lead to even more panic.
False negatives are also a problem. But you have to account for disease severity: ~ 80% experience cold-like symptoms, requiring no treatment. ~15% need treatment, and ~5% die even with it. Tests probably correlate with this, meaning least sick are most likely to test false negative. So the false negatives only directly affect 15%; however, people said to be clear could also pass on the disease. It gets complicated, but the effect is small than false positives, especially in the face of over-reaction.
There is also the distinct possibility of the tests being confused with flu, in the sense that some people dying from the flu will also have coronavirus in them. The deaths can then sometimes wrongly attributed, making coronavirus seem worse. One German doctor is convinced this is the case.
The tests are so new — and have so little history behind them — that Maron said he’s been told they are 90% accurate at best. And that’s for patients presenting with the known symptoms: fever, respiratory issues, gastrointestinal distress. For those who aren’t, Maron said, the accuracy is much lower…
So, if you’re symptomatic, meaning you have a fever and cough, and I swab you and send that sample in, the accuracy of that is 90%. There’s a 10% chance it’s wrong, which is a pretty big window. This is opposed to the flu test, which is 99.999% positive. We’ve had enough testing on those and they have refined that the processes that they know for sure.
“This is complicated. It’s complex science. And, the best we can tell, it’s only 90% accurate. And there is no evidence whatsoever to indicate it’s otherwise.”
None of this is new: every person working regularly with medical data knows these things. How long before this is assimilated by the media, the Lord only knows.
Many headlines are like this next one, designed for clicks and panic inducement: The Worst-Case Estimate for U.S. Coronavirus Deaths
Between 160 million and 214 million people in the U.S. could be infected over the course of the epidemic, according to one projection. That could last months or even over a year, with infections concentrated in shorter periods, staggered across time in different communities, experts said. As many as 200,000 to 1.7 million people could die.
This is wrong, not even close to being right. You want the worst case scenario? This: everybody is infected and reinfected with the coronavirus until they die from it. Almost eight billion dead!
Now, with that idiocy out of the way, can we talk about the more likely scenarios?
Incidentally, anybody remember how news media make their money and politicians win their support? Clicks, eyeballs, followers. The more freakish the headline or pronouncement, the bigger the profit the larger the following. Now, honestly, do you think any editor or politician has succumbed to the temptation to exaggerate the coronavirus threat?
Another: “Half Of America Will Get Sick”: Here Is What Goldman Told 1,500 Clients In Its Emergency Sunday Conference Call“. Of all the BS out there, this is the BSiest. Are they short selling?
Here’s a guy (thanks to C-Marie for the tip) who had millions of hits on his ACT NOW OR WE’LL ALL DIE article. A series of impressive looking plots, soberly put advice, and the call for immediate China-style crackdown. Here’s the chart he uses to prove this is necessary:
He shows the “Lockdown” on 23 January, and the day after he adds the label “True new cases immediately plummet”, by which he could have only meant “True new cases continued to increase and even to spike, eventually falling well after the lockdown.”
Here’s an expert: “Ohio health official estimates 100,000 people in state have coronavirus”. At the time this official made this estimate, there were 5 confirmed cases. You’d think an expert mind addled by panic would still know the difference between 5 and 100,000. But, no. They have since walked this number back, but not before increasing the frenzy and closing all restaurants and bars. In Cincinnati, the mayor has threatened to use the police to break up gangs of more than two people on the sidewalk. Mom has to walk the kid alone, or in shifts, if it’s kids.
There was this Washington state doc on Twitter who said his colleagues were reaching the end, exhausted from all the death surrounding them. He made it sound like they were in an active war zone. I called him out on it, saying, “So then Seattle is lying to the rest of the country? Because only 65 people in the whole USA have died from coronavirus (so far). Or maybe they’re dying from fright? And thus not being reported.” He didn’t answer and deleted his tweet.
Everybody wants to be part of the story.
Here’s a random collection of headlines, a small sample taken on just one day; there are hundreds of examples since our last update (I won’t do links): “We’re not going back to the way life was before”, “China bought the west time. The west squandered it”, “The Internet can’t save us from loneliness in pandemic”, “China bought the west time. The west squandered it”, “You will adjust to the new normal”, “Complete shutdown on the table.” On and on.
Our problem is that we, our culture, lives only in the moment. The past is long forgotten, the future an impossibility. This now, is it. And it must be perfect, for there is nothing else. Safety above all is the favored slogan, but this is effeminate.
Just like SARS (remember?), a few areas outside the hot zone had higher case death rates, like in Lombardy with coronavirus.
You have to hand it to Italians, though. They maintain their soul while imprisoned.
Monday, the White House mentioned they were doing “modeling”, inputing various parameters such as R0 and “social distancing”. These models said “Panic!” Now I’ve told you many times that what comes out of these models is exactly what you put into them. They are built to say what they say. This does not mean they are wrong, because they might make good predictions. But they’re not used for predictions primarily. They’re used for guidance, to decide what to do before the predictions come in to be verified. Meaning what to do was, in a sense, preordained.
Wasn’t it always obvious that if there is no contact between people that any virus would have a difficult time propagating?
Forget about all that R0 mumbo jumbo. That kind of number is for modelers. Just as there is no single unique case death rate, there is no single unique R0 either. Concentrating on it is like concentrating on the case death rate as if it applies to you as your own personal chance of dying. We’ll just look at actual numbers, and nothing else.
The numbers above suggest a better model than the naive one we have been using is one which models outbreaks in population centers, where the population size varies by how it does in real life. Somebody want to pay me a lot of money to do that? No?
Then we’re stuck with the naive model, which only assumes this infection will spread like most infections do. Small initial increases, then GOING EXPONENTIAL!, then tailing off and puffing out, all this in waves or peaks, and finally the news cycle moving on to new hobgoblins.
Given all that, it’s clear the naive model has been under-predicting the second peak. This is due in part because of the delays and choppiness in reporting, especially from Italy. They’re not exactly running on a 24-hour precise reporting schedule.
The naive model is also biased to the initial slope, and is thus over-certain. We have to keep that in mind. Meaning, since it’s under-predicted before, it’s a good bet it will continue to do so, until the second peak becomes obvious, because the model did well in China (initial peak) after that point, too. Also, the model cannot see any future peaks, which of course might happen, maybe even as late as the fall.
All right, enough of that, here we go, the latest projections, which numbers current as of Monday 8 PM EST.
This says by about the 15th of April, there will be 825,000 total cases, and 14,000 total deaths. That’s conditional on the model; adding in under-prediction says these totals will be higher. How much higher? Depends on how far from the second peak we are.
The daily new cases:
Probably not there yet, though the model has it coming in about two weeks. The choppiness in reporting, however, is now readily apparent.
Incidentally, this choppiness, even averaged over the whole world, shows why models on a country level would be even more variable. The worldwide totals averages out some of the variability.
Here are the daily new deaths:
If we’re not at the case peak, and you can’t die before having the bug, the closer predicted death peak can’t be right. Or if it is right, then that means the new cases likely have a healthy dose of false positives. Or that many have a less deadly strain of the bug.
This is doubtful because of the simple case death rate, pictured here:
Recall that this is just the simple dividing of total deaths by total cases. That inflection point a little more than a week back speaks Italian. The individual city case dead rate in the most troubled areas of China were about 4%, and this total is pushing that direction. Just like SARS (as I have been repeating endlessly), there were areas outside the Chinese hot zone that had higher case dead rates in the secondary peak.
Recall too the initial dip in this is caused after the first peak forms, which is after the initial sickest people meet their maker. Same thing for the second smaller dip.
Do focus on the total numbers of deaths to date: 7,113 (as of Monday evening). That seem lower than you might have guessed given the media and political frenzy?
To support this site and its wholly independent host using credit card or PayPal (in any amount) click here