FIGURES UPDATED 3-15-20
You’ve probably seen various graphs online showing the increase in coronavirus cases in various countries.
I don’t know that I am adding any value here, but I decided to reproduce those graphs using the convenient data from Johns Hopkins Coronavirus Research Center. (I can’t vouch for the reliability of the Johns Hopkins data, but it seems to be what most news organizations are relying on.) Here’s one showing cases for all countries other than China that have reported at least 100 cases. The x axis is days after the 100-case mark was reached.
What we see here is that most countries show a consistent 35-45 percent daily growth in reported cases. Only Japan, at 20 percent, and Singapore and more recently Korea, at around 10 percent, depart significantly from this. It’s also interesting how stable the growth of cases in Japan has been over the past three weeks.
Now here is the same figure, but for US counties that have reported at least 10 cases. The x axis here is days since the first day with a least 10 cases. [UPDATE: I have stopped updating this graph since the Johns Hopkins site now reports cases only for US states.]
What surprises me here is that we basically see the same ~40 percent daily growth rate in cases. I would have thought that given all the insitutional differences and issues around testing, the US picture would have looked somehow different. But it seems like we might reasonably extrapolate from the international experience, that New York or Seattle could reach 10,000 cases in the next two weeks.
I have no expertise whatsoever on infectious diseases, so I am not going to say anything else about this.
Anyway, here is the R code if you want to produce figures like these from the most current data on the Johns Hopkins site. They also give latitude and longitude for every place included, so it wouldn’t be much more work to make an interactive map. Could be an interesting excercise for anyone teaching a statistics or data science course.
UPDATE: I had to change the code because Johns Hopkins is no longer reporting data for US places other than states. Here is the equivalent state-level figure. As you can see, the states with significant numbers of cases all show the same 40 percent daily growth rate.
#install.packages('reshape2')
#install.packages('ggplot2')
library(reshape2)
library(ggplot2)
corona <- read.csv('https://raw.githubusercontent.com/CSSEGISandData/COVID-19/master/csse_covid_19_data/csse_covid_19_time_series/time_series_19-covid-Confirmed.csv',
stringsAsFactors = F)
state.abbrevs <- read.csv('https://raw.githubusercontent.com/jasonong/List-of-US-States/master/states.csv', stringsAsFactors = F)
corona[,-1:-4] <- apply(corona[,-1:-4], 1:2, FUN=as.numeric)
US <- corona[corona$Country.Region=='US', ]
USplaces <- US[grep(',',US[,1]),][,-2:-4]
USstates <- US[grep(',',US[,1], invert=T),][,-2:-4]
USstates <- USstates[grep('Princess',USstates[,1], invert=T),]
states.temp <- USstates
for (i in 1:nrow(USstates)){
abbrev <- state.abbrevs[state.abbrevs[,1]==USstates[i,1],2]
rows <- grep(abbrev, USplaces[,1])
states.temp[i, -1] <- colSums(USplaces[rows, -1])
}
states.temp<- states.temp[!is.na(states.temp[,1]),]
USstates[,-1] <- USstates[,-1] + states.temp[,-1]
corona <- rbind(corona, c('', 'United States', NA, NA, colSums(USplaces[,-1])))
corona[,-1:-4] <- apply(corona[,-1:-4], 1:2, FUN=as.numeric)
china <- corona[corona$Country.Region=='China',-2:-4]
corona <- rbind(corona, c('', 'China', NA, NA, colSums(china[,-1])))
corona[,-1:-4] <- apply(corona[,-1:-4], 1:2, FUN=as.numeric)
countries <- corona[corona$Province.State =='', c(-1, -3:-4)]
makePlot <- function (x, threshold){
cases <- redate(x, threshold)
data <- melt(cases, id.vars=1)
names(data) <- c('place', 'day', 'cases')
p <- ggplot(data, aes(x=day, y=cases, group=place)) +
geom_line(aes(color=place))+
geom_point(aes(color=place, shape=place))
p <- p + scale_shape_manual(values=1:length(levels(as.factor(data$place))))
p <- p + theme(axis.text.x = element_text(angle=45))
p <- p + scale_y_continuous(trans='log10')
p
}
redate <- function (x, threshold) {
out <- data.frame(place='', stringsAsFactors = F)
values <- matrix(nrow=99, ncol=99)
n <- 0
for (i in 1:nrow(x)) {
v <- x[i,-1][x[i,-1] > threshold]
l <- length(v)
if (l > 1) {
n <- n + 1
out[n,1] <- x[i,1]
values[n,1:l] <- v
}
}
values <- values[!is.na(values[,1]),]
cols <- length(colSums(values, na.rm=T)[colSums(values, na.rm=T) > 0])
values <- values[,1:cols]
out <- cbind(out, values)
}
makePlot(USplaces, 9)
makePlot(USstates, 10)
makePlot(countries[countries$Country.Region != 'China',], 99)
Coronavirus isn’t much more dangerous than the flu, and the emergency measures are way out of proportion.
The roughtly 3 percent mortality rate that is being estimated for COVID is too high by an order of magnitude; it’s mainly a statistical artifact of spotty, biased testing and diagnosis of COVID cases. In most countries people aren’t being tested (and diagnosed) for COVID unless they have a serious case or are in high-risk groups, so the sample of cases from which they are guesstimating mortality rates isn’t representative of the typical innocuous COVID infection.
We know this because mortality rates vary wildly in different countries, by well over an order of magnitude. (See Worldometer for up-to-the-hour statistics.) (https://www.worldometers.info/coronavirus/).
In Iran the mortality rate is over 10 percent because they don’t have the resources to do widespread testing for COVID; probably only people on death’s doorstep get a COVID diagnosis. Italy’s testing system has been overwhelmed by the size of the outbreak, so they likewise are only testing and diagnosing the sickest and most vulnerable, and show a mortality rate of 6.7 percent.
South Korea, by contrast, has really cranked up its testing of the general population. It’s therefore diagnosing milder, more representative COVID cases and its mortality rate is just 0.83 percent.
Then there’s northern Europe, where they have relatively small outbreaks paired with massive testing of the general public, including drive-through testing in Germany for people with just a sniffle. There they have extremely low mortality rates, roughly comparable to the flu’s typical 0.1 percent:
Germany: 6 deaths out of 2,745 cases, 0.22 percent mortality
Norway: 1 death out of 797 cases, 0.12 percent mortality
Sweden: 1 death out of 687 cases, 0.14 percent mortality
Denmark, 0 deaths out of 674 cases.
(Stats from late on 3/12/2020.)
Those stats aren’t a perfect measure of mortality, because a few of the current crop of patients will die later; but on the other hand even the Nordic countries are undoubtedly missing a lot of mild cases that don’t bother getting a test. Because there’s always a sampling bias towards testing and diagnosing cases that are more serious than average, we should regard the lowest observed mortality rates as an upper bound on the true mortality rate.
By all accounts COVID mortality is heavily concentrated among elderly patients with chronic illnesses. Very few healthy adults are dying of the disease. And children seem to be pretty invulnerable, few cases and virtually no deaths (unlike with the flu). The closure of schools and universities is crazy, because kids just aren’t at risk from COVID.
So there is no doubt that for the vast majority of the population COVID is no more dangerous than a middling flu season.
It is a serious risk for old and sick people, and we should have a reverse quarantine for them: they should be urged to stay home to avoid catching COVID, and governments should make arrangements to provision them with supplies and financial assistance.
But emergency measures and restrictions aimed at the general population are insane overreactions, and significant human-rights violations. People should organize to oppose them.
Panic is bad!
Coronavirus isn’t much more dangerous than the flu, and the emergency measures are way out of proportion.
The roughtly 3 percent mortality rate that is being estimated for COVID is too high by an order of magnitude; it’s mainly a statistical artifact of spotty, biased testing and diagnosis of COVID cases. In most countries people aren’t being tested (and diagnosed) for COVID unless they have a serious case or are in high-risk groups, so the sample of cases from which we are estimating mortality rates isn’t representative of the typical innocuous COVID infection.
We know this because mortality rates vary wildly in different countries, by well over an order of magnitude. (See Worldometer for up-to-the-hour statistics.)
In Iran the mortality rate is over 10 percent because they don’t have the resources to do widespread testing for COVID; probably only people on death’s doorstep get a COVID diagnosis. Italy’s testing system has been overwhelmed by the size of the outbreak, so they likewise are only testing and diagnosing the sickest and most vulnerable, and show a mortality rate of 6.7 percent.
South Korea, by contrast, has really cranked up its testing of the general population. It’s therefore diagnosing milder, more representative COVID cases and its mortality rate is just 0.83 percent.
Then there’s northern Europe, where they have relatively small outbreaks paired with massive testing of the general public, including drive-through testing in Germany for people with just a sniffle. There they have extremely low mortality rates, roughly comparable to the flu’s typical 0.1 percent:
Germany: 6 deaths out of 2,745 cases, 0.22 percent mortality
Norway: 1 death out of 797 cases, 0.12 percent mortality
Sweden: 1 death out of 687 cases, 0.14 percent mortality
Denmark, 0 deaths out of 674 cases.
(Stats from late on 3/12/2020.)
Those stats aren’t a perfect measure of mortality, because a few of the current crop of patients will die later; but on the other hand even the Nordic countries are undoubtedly missing a lot of mild cases that don’t bother getting a test. Because there’s always a sampling bias towards testing and diagnosing cases that are more serious than average, we should regard the lowest observed mortality rates as an upper bound on the true mortality rate.
By all accounts COVID mortality is heavily concentrated among elderly patients with chronic illnesses. Very few healthy adults are dying of the disease. And children seem to be pretty invulnerable, few cases and virtually no deaths (unlike with the flu). The closure of schools and universities is crazy, because kids just aren’t at risk from COVID.
So there is no doubt that for the vast majority of the population COVID is no more dangerous than a middling flu season.
It is a serious risk for old and sick people, and we should have a reverse quarantine for them: they should be urged to stay home to avoid catching COVID, and governments should make arrangements to provision them with supplies and financial assistance.
But emergency measures and restrictions aimed at the general population are insane overreactions, and significant human-rights violations. People should organize to oppose them.
Panic is bad!
I admit this was my initial response too. But you have to admit, most medical and public-health professionals are saying this is much worse than a seasonal flu. Is there a reason not to believe the experts here?
By all accounts COVID mortality is heavily concentrated among elderly patients with chronic illnesses. Very few healthy adults are dying of the disease. And children seem to be pretty invulnerable, few cases and virtually no deaths (unlike with the flu).
Yes, this is consistent with everything I’ve read too. But a disease that mainly kills old people, is still a deadly disease
Another thing about this is, the virus itself doesn’t have to cause a lot of fatalities to be a big problem. If a large number of people have to be hospitalized, that can overwhlm the capacity of the health care system and leave other people with dangerous illnesses unable to get care in a timely way. That seems to be what’s been happening in Italy.
Will Boisvert- if in 3 weeks it turns out that deaths have not doubled or tripled or much worse in Germany, Sweden, Norway, or the US for that matter- then I will admit to panicking. It seems like it really sucks in Italy right now though. The ‘remain calm- it only mostly kills old people’ really isn’t comforting if you care about some of those old people and don’t manage to forget that it also gets some much younger people.
I think reaction to the possibility of it being as bad as quite a few doctors have said it could be is the smarter thing to do. If you are right, and they are wrong, well how many died because of that mistake.
It would be better to panic, honestly.
I live in Lombardy and I’m in lock down since a pair of weeks. The problem is that politicians react to cases of infection but the virus takes two weeks to have effect, so that politicians react to the situation as it was two weeks earlier.
But the lock down is very expensive economically, and many people will lose 100% of their income from it, and really it is a pain when you can’t go to your girlfriend house or similar, so in practice it means politicians will act only when it is evident that the situation is dire.
But this means that they will act two weeks too late, and with cases doubling every 4 days this is a very big problem.
If you accept that you’ll have to lock down eventually it is obvious that it is better (and cheaper) to lock down earlier than later. Once you are already locked down this becomes obvious.
@ JW:
“The virus itself doesn’t have to cause a lot of fatalities to be a big problem. If a large number of people have to be hospitalized, that can overwhlm the capacity of the health care system and leave other people with dangerous illnesses unable to get care in a timely way. That seems to be what’s been happening in Italy.”
No, the alarmism about hospital systems getting overwhelmed is even more overblown than the exaggeration of mortality rates. US hospitals will easily handle any plausible COVID outbreak.
Lately there have been apocalyptic claims that COVID will overwhelm US hospitals. The New York Times and others have cited a CDC estimate that there could be up to 21 million hospitalizations, swamping the nation’s million or so hospital beds.
But numbers like that are obviously wrong, and demented. Let’s look at the stats. (From Worldometer, late 3/13/20.)
1. Chinese data show that the total number of COVID diagnoses, and therefore hospitalizations, will be pretty small.
Alarmists aren’t talking much about this, but in China the COVID epidemic is over. New cases and deaths have dwindled to a handful each day. So we can get a good sense of the ultimate scale of the epidemic from the Chinese example—and it’s much smaller than the alarmist forecasts.
Total cumulative COVID diagnoses stand at about 80,000 in China, and will likely never make it over 100 thousand. Deaths are about 3100, and will likely never exceed 4,000. This in a country with four times the population of the US. Most of those cases were never hospitalized. So the notion that the US might have to *hospitalize* 21 million COVID patients–250 times more COVID patients than the total number of COVID cases in all of China, hospitalized and unhospitalized combined–is sheer lunacy.
2. South Korean data show the same thing, specifically the numbers for “serious / critical” cases, a proxy for patients that need hospital care.
South Korea’s COVID outbreak is also waning, with daily new cases down 87 percent from their peak on March 3; soon the epidemic will gutter out there as well. Currently there are 7300 active cases, out of a total cumulative caseload of 7869. I can’t find exact figures on hospitalizations, but we do have numbers for “serious / critical” cases, a good proxy for patients that need hospital care: there are only 93 of them, about 1 percent of total cases and an order of magnitude below the 10 percent or higher hospitalization rates that alarmists are positing. If South Korean COVID diagnosis rates applied to the US proportionately, and if we hospitalized every single diagnosis, we would still have only about 51,000 hospitalizations at the peak of an outbreak, easily absorbed by the reserve of about 300,000 vacant hospital beds the US usually has on hand.
3. Italian data show higher serious/critical case rates, but still easily manageable.
Currently Italy has about 1,328 serious/critical COVID cases out of an active caseload of 14,955. Scaling up proportionately to the US population, that would be 7,304 serious / critical patients. Even at ten times that number the US hospital system would easily handle them. (I have no idea why Italy would be having trouble; it may just be overtreating a lot of mild COVID cases in the hospital that don’t need to be there.)
4. How about US data? Very low serious/critical case rates, therefore low hospitalization rates.
In the US itself, out of 2,190 active cases, a grand total of 10 are currently listed as serious/critical. That number would have to rise by a factor of 30,000 to outstrip America’s reserve of vacant hospital beds. But then we’d have about 60 million active cases, giving a total case rate of 181,000 per million. That’s 3,200 times higher than China’s peak case rate of 56 per million, 1,617 times higher than South Korea’s peak rate of 155 per million. That’s just not going to happen.
So there’s no support for forecasts of a scary hospital burden when you look at the evidence. Realistically, a plausible high-end estimate for the hospital roll from an extreme COVID outbreak would be maybe 20,000 hospitalizations at any one time, not a serious problem.
“A single death is a tragedy; a million deaths is a statistic. When one dies, it is a tragedy. When a million die, it is a statistic” (c) Stalin
It is so weird to read your comments…
Your cynicism, which disguise under pragmatism, leaves the most vulnerable, i.e. poor and elderly, on their own. Maybe they are lucky, maybe they are not, but surely statistically speaking we should and will be fine.
And as a bonus thing, we will also solve our retirement funding problem. I mean for those who are lucky not to be in those two categories today.
Alarmists aren’t talking much about this, but in China the COVID epidemic is over. New cases and deaths have dwindled to a handful each day. So we can get a good sense of the ultimate scale of the epidemic from the Chinese example—and it’s much smaller than the alarmist forecasts.
I don’t understand. I thought you were arguing against a draconian response to the corona virus — closing schools, prohibiting large gatherings, discouraging travel, etc. But these are exactly the things that China did!
If I understand your argument it goes like this:
1. Some people think the coronoa virus is a public health crisis.
2. China treated it as a public helath crisis, and it is now under control.
3. Therefore, we should not treat it as a crisis.
Is that really your argument?
(Or is your argument that the outcome in China has nothing to do with the policy response there?)
You are of course right that, so far, the US has a modest number of cases and only a handful of deaths. But are you sure it makes sense to wait until large numbers of people are already dying, before taking steps to limit the spread of the disease?
@ JW:
“I don’t understand. I thought you were arguing against a draconian response to the corona virus — closing schools, prohibiting large gatherings, discouraging travel, etc. But these are exactly the things that China did!….
2. China treated it as a public helath crisis, and it is now under control.
3. Therefore, we should not treat it as a crisis.
Is that really your argument?
(Or is your argument that the outcome in China has nothing to do with the policy response there?)”
1. Good points. I am skeptical that the emergency measures were very effective. China didn’t really “shut down.” Yes, schools and concerts and (not all) travel may have been banned, but people still have go to shops and grocery stores and work and ride subways and buses. It’s not possible to live in China and not mingle in crowds where viruses spread.
So the small Chinese case count has to call into question the assumption that COVID is as infectious as the flu. To put that assumption in perspective, CDC estimates that in the US this year there will be 36 million to 51 million flu cases, with 370,000 to 670,000 hospitalizations and 22,000 to 55,000 deaths—for a disease against which over half the population was vaccinated. So if COVID is as infectious as flu, the Chinese emergency measures must have been astonishingly effective to hold it to just 80,000 cases and 3-4,000 deaths in a country with four times the population of the US, and with no vaccine. I don’t think that’s plausible. COVID is likely less infectious than the flu.
But it’s hard to tell because now every country is treating COVID as a public health crisis. There’s no business-as-usual control outbreak that we can use as a comparison to assess the effectiveness of the containment regimes.
There may be another way to estimate that. The general anti-COVID measures—school closures, bans on travel and public gatherings, self-isolation for symptoms without testing—will presumably also have the effect of suppressing the spread of the flu virus at the same time and to a similar extent. Flu seasons are pretty well characterized, so epidemiologists may be able to estimate what the flu toll would have been in the absence of anti-COVID measures, and compare that to the observed flu toll during the COVID emergency. That should give us a measure of how effective the emergency measures were against COVID as well.
2. I think there’s a strong case for treating COVID about as seriously as the flu.
–The fact that the Chinese and SK epidemics subsided with such a low impact suggests that COVID is less infectious than the flu.
–The data on serious-critical cases suggest that COVID is not more virulent than the flu. CDC put the flu hospitalization rate at about 1.3 percent. For Covid my proxy hospitalization ratio of (deaths + serious-critical cases)/(total cases) is 1.6 percent in the US, In Germany it’s 0.46 percent, in Sweden 0.36 percent.
–Consistent with the previous point, the mortality rates for COVID in Northern Europe, likely an upper bound for the true mortality rate, are in the 0.1 to 0.2 percent range, comparable to the flu.
So COVID is a virus that’s probably less infectious than the flu, about as virulent as the flu, and about as lethal as the flu. So why treat it differently from the flu?
3. “So far, the US has a modest number of cases and only a handful of deaths. But are you sure it makes sense to wait until large numbers of people are already dying, before taking steps to limit the spread of the disease?”
Well, apply that argument to the flu. You seem to be fine with waiting around while 50,000 Americans die of the flu without calling for any steps at all to limit its spread; yet at 50 deaths from COVID you’re demanding emergency mobilizations. Does that make sense?
It’s not about taking no steps to limit the disease, it’s about making sure the steps are rational and proportionate.
That’s not where we’re headed. The scientific consensus is that some of the most onerous emergency measures, like school closures and Trump’s travel bans—he’s now talking about domestic travel bans!—are useless. Why are we doing them? Because of panic.
A good rule of thumb to counteract that is to ask, would I impose this emergency measure to combat the flu?
–A COVID vaccine? Yes! I get flu shots.
–Banning public gatherings? Nope. If we don’t do it for the flu, why do it for COVID?
–I’ve heard some people say we should let prisoners out of jail and stop arresting criminals so they don’t catch COVID in their cells. Should we do that during flu season as well? I say no.
–I saw on Youtube that the Italian police are fining anyone who goes out for a walk. Should we do that for the flu?
–Social distancing? Sure. I can see the government doing a campaign to persuade people to social-distance during a bad flu season. Not too disruptive, makes sense. Urging people to stay home if they have symptoms, or are old and chronically ill? Yes, that will help vulnerable people without hurting anyone else much.
The rule works well. If you wouldn’t do it for the flu, don’t do it for COVID.
Suffice to say that the various Italian doctors that I hear on TV do not agree with WB. In practice we (in the main areas of infection, not in all Italy) are already in a triage situation where almost every doctor who works in an hospital is redirected to patients with covid, preferential treatment is given to those who have most chances of surviving.
Maybe you’re right, MR2. Italy is something of an outlier.
Could the travel restrictions be causing bottlenecks, i. e. patients can’t travel to hospitals where there is room for them?
Have you seen any up-to-date statistics on the number of COVID patients that have been hospitalized, the number placed in intensive care or on ventilators, compared to the normal number of hospital beds?
I would like to get a quantitative look at the medical situation in Italy.
I looked up some stats on the COVID hospital situation in Italy.
The problem is that 1) Italian hospitals are severely undersupplied with “critical care” beds and staff, the kind patients with severe respiratory illness need; and 2) in Italy COVID is more virulent than in other places, with much higher mortality rates and, presumably, higher rates of severe cases that need critical care.
In Lombardy, epicenter of the Italian outbreak, the provisioning of critical care beds is about 7.2 per 100,000 population. In the US it’s 4 times higher, about 30 per 100,000. If Lombardy were provisioned at US levels, it would have an extra 2,300 critical care beds, likely enough to accommodate all the severe COVID cases.
And, for unknown reasons, in Italy COVID is more virulent and lethal than in the rest of the world, so there are more patients who need critical care. The mortality rate is 6.8 percent in Italy, the highest of any major outbreak, higher than Iran (4.8 percent), China (3.9), Spain (3.1) the US (2) Germany (0.2 percent) etc. COVID critical care cases presumably mirror those differences. The ratio of patients requiring critical care in Italy is probably triple what it is in the US or more.
So there is a medical crisis in Italy, but Italy is a special case. US hospitals will have a much easier time coping with COVID patients than Italy has had.
You know Will, unless you have some knowledge that no one else seems to have, arguing that Covid19 in Italy is some especially dangerous variation of the virus that we should not consider is just plain irrational and irresponsible.
As far as my personal experience- I went to a scheduled appointment with the Ear Nose and Throat doctor that monitors my recovery from cancer surgery. I called ahead and told them I had had a fever, a sore throat, and a bad cough just in case they preferred me not to come in or wanted to take precautions. They told me to come in anyways. No test for coronavirus but did have the pleasure of some instrument stuck through my nose to look at my sore throat. I don’t think they had the test for it available.
Talked to my primary care doctor a couple of hours ago, he happens to be my brother. Asked about what was going on about the corona virus and he told me there has been NOTHING from the CDC or the Feds in general about it. He runs several urgent care clinics around here by the way. This is ridiculous even if everything Will B. has said turns out to be true. Major failure from Federal government.
@WB
I also heard that there is a study that say that there is a Covid-A (milder) and a Covid-B (deadly), and that the high death rate in Italy is due to a prevalence of Covid-B.
The study is based on some cases where a patient got Covid, apparently got healthy again and was released, but then got a worse Covid infection later. So if the study is correct, Covid-A doesn’t build immunity to Covid-B, which sucks, and unless you are sure that Covid-B will stay only in Italy then there is few room for complacency.
Also, the idea of the study (if I remember correctly, I read a summary some days ago so I’m not sure) is that the Wuhan numbers are due to an average of A and B deaths, so are an overestimate for A but an underestimate for B.
As far as I know it is true that Covid patients in infected areas are not shuttled around to less stressed parts of Italy, however I guess that the reason is to not break the quarantine. Also note that Lombardy (where most cases are) is probably the region in Italy with the best healthcare (basically because it is the richest and most urbanized region).
On the idea that italian NHS has few intensive care beds relative to other health systems I don’t know, generally italian NHS is considered good: in the last WHO ratings of worldwide health systems it came second just after France, but it was many years ago (the WHO stopped publishing these votes after the matter became politically contentious in the USA because the USA ranked below Cuba). It is possible that on the specific question of intensive care bed the italian NHS is underserved, but what I hear most often is a lack of ventilators to help Covid patients who have problems breathing.
Another possible explanation is that Italy has an aging population (both because of high life expectancy and low birth rates), so as Covid kills mostly old people this could cause an higer rate of deaths. But this could justify a death rate that goes from 4% to 6%, not from 0.2% to 6%, so I would assume a normal death rate at 4% from this.
[note: I have problems posting my comments because the blog’s systems believes I’m a bot, apologies if sometimes I don’t answer]
I don’t think there’s any reason to believe that it’s more lethal in Italy than elsewhere. Seems more parsimonious to assume that what we’re seeing is some combination of 1) the disease is much more dangerous for older people, and Italy has one of the oldest populations in the world; 2) the outbreak in Italy started earlier, so more cases have had time to progress to the lethal stage; and 3) more aggressive screening may be picking up more cases in Germany and elsewhere.
More broadly:
Evenif this turns out to be only as bad as the flu, which seems very optimistic, the flu is actually a serious problem. 50,000 deaths is a lot – it’s about 2 percent of all deaths in the country. And the big difference is, we’re already stuck with the flu. The example of China suggests that with aggressive enough containment measures, it might be possible to prevent this one from getting established. If we have the chance to prevent a second seasonal flu from being established, I think that’s worth some pretty major measures. Don’t you?
So even if this were equivalent to the flu, it’s not clear to me that current reactions are excessive. And some of the things that are being proposed now, actually should be done in a bad flu season, like releasing low-level offenders and people being held pre-trial from jail, and halting evictions and heat shutoffs.
But I don’t see how you can be confident at this point that it’s just going to be equivalent to adding a second seasonal flu. The reported mortality rates may be distorted in various ways (not all of them upward) but they certainly suggest a much higher number. And there’s also the question of how widely it spreads. If everyone is infected with something with even an 0.2 percent mortality rate, that’s a catastrophe. The all-causes mortality rate in the US is only 0.7 percent.
Obviously we don’t know the relevant parameters at this point, so some of this is just about how one evaluates highly uncertain dangers. But the thing I can’t get away fromis that steady 30-40 percent daily increase in cases, almost everywhere there’s a significant number of infections. That means that, in a sense, it doesn’t matter how accurately we are counting the cases — whatever the situation is today, it’s going to be 10 times worse in a week, 100 times worse in two weeks, 1,000 times worse in three weeks.
In most countries, from what I can tell, deaths seem to be following that same curve, just 10-15 days behind. If we have five deaths so far in New York State, I expect we’ll have 50 in a week, 500 in two weeks, 5,000 in three weeks, 50,000 in four weeks. Given how it’s progressed elsewhere, I don’t see any reason to think the curve will flatten out in the absence of drastic measures to limit its spread. To think otherwise just seems like wishful thinking.
Given that we can observe a death rate rising at the same exponetial rate as reported cases, some of these questions of infection and mortality rates — while we would obviously like to have a clearer sense of them — seem beside the point. Whether we are measuring it correctly, or whtehr we’re overestimating the lethality but underestimating the number of cases, the bottom line is the same. 50,000 deaths is equally bad whether it’s the result of 200,000 cases or 2 million.
@ JW:
“I don’t see how you can be confident at this point that it’s just going to be equivalent to adding a second seasonal flu. The reported mortality rates may be distorted in various ways (not all of them upward) but they certainly suggest a much higher number. And there’s also the question of how widely it spreads….If everyone is infected with something with even an 0.2 percent mortality rate, that’s a catastrophe. The all-causes mortality rate in the US is only 0.7 percent…..But the thing I can’t get away fromis that steady 30-40 percent daily increase in cases, almost everywhere there’s a significant number of infections….If we have five deaths so far in New York State, I expect we’ll have 50 in a week, 500 in two weeks, 5,000 in three weeks, 50,000 in four weeks.”
No, nothing like that will happen. The mortality rate is about 0.1 percent among people who get infected, and most people—80 percent or more–will never get infected anyway.
There is crucial evidence in this regard in the data from the Diamond Princess cruise liner, which had a severe outbreak and was quarantined in early February. It’s the best evidence we have of infection and fatality rates in a sizeable population who were all tested, thus eliminating the sampling bias towards serious cases that’s distorting national case-fatality rates.
So of the 4,061 people aboard the Diamond Princess, about 700 tested positive for coronavirus. Of those, 7 died, for a case fatality rate of 1 percent.
But that 1 percent rate is undoubtedly much higher than the rate in the general population. The population aboard the vessel was heavily skewed towards old and sick people. Most of the passengers were elderly, and many had chronic illnesses; of the dead all but one were in their 70s and 80s. In a younger, healthier general population the true case-fatality rate would be dramatically lower than the Diamond Princess’s 1 percent. So Germany’s observed fatality rate of 0.26 percent is quite plausible as the true fatality rate.
But another striking finding from the Diamond Princess means that true fatality rates are much lower still. Of the 700 people who tested positive for the virus, about half of them never developed any symptoms over an observation period of a month or longer. So the rate of asymptomatic infections in the general population is undoubtedly 50 percent or higher (again, we should expect a higher asymptomatic rate in a younger, healthier general population).
That means that, in national testing numbers, for every reported positive test there is at least one case of infection that isn’t reported because it’s asymptomatic. (Unlike on the Princess Diamond, asymptomatic people aren’t given tests in national screening programs, except for contact tracing, and that goes by the wayside in a mass outbreak.)
So to account for unreported asymptomatic cases we should cut the observed German fatality rate at least in half, for a true fatality rate of 0.13 percent. It’s quite likely even lower. The literature indicates that the asymptomatic rate for the flu is up to 75 percent, and that’s a plausible estimate for the coronavirus asymptomatic rate in a population that’s younger and healthier than the Diamond Princess oldsters. If so, the true fatality rate would be about 0.07 percent, very comparable to the flu’s 0.1 percent. And that observed German fatality rate of 0.26 percent is itself overstated because there is still some sampling bias in the German testing regime, with serious cases and vulnerable elderly and sick demographics being over-represented in those who get tested and diagnosed.
Then there’s the question of infection rates. In the Diamond Princess cohort the infection rate was 18 percent, which agrees well with seasonal flu infection rates of about 20 percent for unvaccinated populations. But the coronavirus infection rate would likely be lower in the general population, because the Princess Diamond was an unusually infectious environment, with 4,000 people aboard a cruise ship for over a month, a sky-high population density in very close quarters. Even after the quarantine was imposed, the crew, 14 percent of whom were infected—naturally, crew members are younger so they had a lower infection rate—continued to work throughout the ship, preparing meals and serving them to passengers in their cabins, thus functioning as ubiquitous vectors for the virus.
The Princess Diamond outbreak confirms what we see in the German data: COVID is no more lethal or infectious than the flu. The doomsday scenarios of millions of deaths are flat-out deranged, yet they are being used to justify the imposition of a police state and a severe recession that will throw millions out of work, with many deaths of despair. Sheer madness.
@ JW:
“Even if this turns out to be only as bad as the flu, which seems very optimistic, the flu is actually a serious problem. 50,000 deaths is a lot – it’s about 2 percent of all deaths in the country. And the big difference is, we’re already stuck with the flu. The example of China suggests that with aggressive enough containment measures, it might be possible to prevent this one from getting established. If we have the chance to prevent a second seasonal flu from being established, I think that’s worth some pretty major measures. Don’t you?”
–Are you saying that we can prevent COVID from becoming a recurring seasonal illness if we suppress it now?
No, it doesn’t work that way. The flu is seasonal because it retreats to natural reservoirs, mutates and returns the next year in a variant that people are not immune to. We don’t know whether SARS-CoV-2 will do that, but if it does this year’s containment won’t prevent next year’s pandemic.
In fact the opposite may be true. By preventing innocuous infections (the vast majority), the emergency measures may hinder the development of herd immunity that could snuff out later outbreaks.
–And no, we’re not “stuck” with the flu. If we impose the “lockdown” during next year’s 4-month flu season we will likely have just as much success in suppressing it as we do with COVID. We could do it every year and maybe save thousands of lives every year. Are you proposing that? If not, why not?
Or maybe we could just have forcible vaccination of the entire population every year. That would prevent many flu deaths (not all because the vaccines aren’t completely effective). Are you proposing that?
People fear the huge flu death toll so much that half of them don’t even bother getting vaccinated. We don’t impose a flu lockdown because it’s expensive and inconvenient, and nobody cares how many people die as a result. Why is COVID different? It isn’t. There’s just no coherent scientific or moral justification for the panic.
Yeah Will and you know all this exactly how? How have you managed to come up with a mortality rate of .1% when no one else seems to know it exactly and many estimates are at least ten times higher?
I have been following stats for your preferred countries you posted deaths for March 12. Germany-6 now 44, Norway-1 now 7, Sweden-1 now 11, Denmark-0 now 6. 1 week in time interval.
Sure that’s only 68 people out of about 8 billion in the world. But come on you really think we are overreacting about it? What if your estimates are wrong?
Yeah, I could be wrong.
The number to watch right now is total deaths in Italy. With a flu-like 20 percent infection rate and 0.1 percent case fatality rate, Italy’s total COVID deaths would come in at 12,000. But Italy is already a third of the way there at 4,032 total deaths, including 627 new deaths today. Unless the daily new death tally declines soon, and sharply, Italy will blow past 12,000 total deaths in a week or two. That would definitively prove me wrong and show that COVID is a different kettle of fish from the flu.
Hi, Bill,
But same issue as earlier arises. If Italy’s ‘flu-like’ Covid rates do result in only 12000 deaths (pretty likely to occur, at a minimum), that’s in a situation with a huge lockdown, which we’ve never had for flu. So that means Covid is a lot worse than the flu, unless you tame it with a massive lockdown.
” How have you managed to come up with a mortality rate of .1% when no one else seems to know it exactly and many estimates are at least ten times higher?”
See my comment above, March 19 at 7:11 PM. (Sorry it’s a bit involved.)
I didn’t just pull that out of a hat. It’s based on German case fatality rates of 0.26 percent, then adjusting for asymptomatic COVID cases that aren’t being tested for or diagnosed, which are at least 50 percent of all COVID cases per the Diamond Princess data, and probably more for a general population. So divide the German CFR by 2 or 3 to account for undiagnosed asymptomatic COVID cases and you get a CFR of roughly 0.1 percent.
But the proof’s in the pudding, so we’ll see if the Italian death count stays under 12,000. If it goes far above that, then I’ve gotten things wrong.
And finally, if I did get it wrong, one way may be by not factoring the lag time between the emergence of COVID infections and death, so that the current total deaths should be compared not with current total cases but with total cases a week or so previously.
I’m not sure that will make a huge difference ultimately, because the Diamond Princess data seem to put a firm cap on elderly case fatality rates at about 1 percent, so all-age case fatality rates must be much lower. But the data will tell in the end. And even a 0.2 case fatality rate is twice as bad an epidemic as 0.1 percent, so I guess small differences matter.
So, there is this site that has a lot of interesting stats per country for the coronavirus:
https://www.worldometers.info/coronavirus/country/uk/
If you look at the death/recovery stat for closed cases, you’ll see that:
UK and USA are really bad, although there are still few cases;
Italy is also quite bad;
France and Spain are better than Italy but not by much;
South Korea, China and most interestingly Iran are WAY better than all the above.
So my question is: why exactly?
This is a link to the research that speaks of two branches of coronavirus, since we spoke about it earlier in the comments:
https://academic.oup.com/nsr/advance-article/doi/10.1093/nsr/nwaa036/5775463?searchresult=1
The Italian daily death toll may be receding at last, let’s hope. But it’s clear that the COVID epidemic is much worse than I thought, and I apologize for underestimating it above and for disparaging the containment measures.
Hi, Bill,
I read Italian avg. flu deaths have recently often averaged as high as 18K per year, and Italian Covid deaths will likely come in under that. So total deaths may be no worse than avg. to bad flu year, and the same is fairly likely to be true, perhaps within a factor of 2-ish, both in the US and globally. But again that’s with now-massive suppression, and had that started earlier many lives would have been saved.
One good thing to come out of the massive economic wreckage is that we’ll all be MMTesians now. Effectively printing cash (or equivalently fungible free credit) lost a lot of stigma in 2008, and it’s basically the instant go-to solution now, even for Republicans. Making it statutory rather than de facto will probably take some time, but the zeitgeist has changed.
How well does MMT work in the context of a drop of 50 percent in GDP because of the mandatory idling of the workforce?
Will all the extra deficit and stiumulus spending by the government, accompanied by the collapse of production and imports, cause inflation to soar along with the 30 percent unemployment the Fed is predicting? Does printing cash not cause any inflationary pressure even when output and supply cannot increase because of the lockdown?
Wondering if the lockdown is economically and politically sustainable for very long.
Will, the question should be how well does anything work in that context. I would argue MMT will work better than many other paradigms. The whole thing still sucks of course and MMT doesn’t provide any medical cures in and of itself.
But actually what we are likely to see is that the situation shows the fundamentals of MMT are true as far as how the economy actually works. Even if half of us can’t actually work right now.
Which still doesn’t do anything about the virus. But it does explain how the Feds are going to ‘afford’ to spend 2 trillion dollars so far.
Agree with Jerry re MMT. The “descriptive” part of MMT, which is UNDENIABLY true, will now be much harder for the many people who, um, deny it to continue, um, denying it… 🙂
Re inflation during the covid crisis, MMT totally acknowledges that inflation is a (actually, the only) problem, and democracies need to decide how much they want to tolerate in return for keeping demand at a given level. MMT just says whatever that decision is, you can just print as much cash as you need to get to it. (You can also issue debt if you wish, largely to provide safe returns on people’s voluntary savings, but you don’t have to.)
But for people who lost all their income when fired due to the virus, even 30% (very unlikely, we will still be in a major recession) inflation on the few items people can spend on would be a lot better than having no money at all without the cash drop (and augmented unemployment benefits). For the rest of us who are better off, what modest inflation we might actually get might be a cut in our real living standards, but that is the beauty of any inflation resulting from equal (or progressive) cash drops–it automatically redistributes real levels of living standards from those who are better off to those who are worse off. What’s not to like?