Why such a large difference in fatality rates across European nations?

Here is a relevant tweet thread started by Moritz Kuhn, many interesting comments.  For instance Moritz writes: “What is more, it may provide a warning sign for those countries where the elderly and the young live close together, how important it is to contain the virus there early on. These countries are within Europe in particular such as Serbia, Poland Bulgaria, Croatia, or Slovenia.”

Also on Italy, Dan Klein writes to me:

  1. They kiss, hug more, converse longer.
  2. Young people live with their parents, family more.
  3. They smoke somewhat more (packs smoked per capita twice that of Sweden). Smoking weakens the lungs. But also we smokers finger and thumb our cigs and then put them into our mouth. Wash hands first!
  4. For these and whatever adventitious reasons, Italy was early to the problem, and it spread before people learned to adjust behavior.

We will learn more soon.


5. 40-odd prime ministers since 1953, with only Berlusconi and one other lasting more than four years and most lasting less than one year.

6. Differences in testing and reporting. They know fairly precisely how many die. The numbers of infected/no symptoms/no tests are not so precise.

Panic 2020!

Without a shadow of doubt, the main difference is how many people get tested and by what process they are selected. I think that’s a significant omission from the list of possible reasons.

if the u.s. has a higher fatality rate than other countries
then it could be because congress was on the phone with
their stock brokers instead of preparing the rest of the country
for a viral pandemic!

I think the biggest factor is that their healthcare system was overwhelmed by the rapid spread (so mostly #4 with #2 potentially contributing to the rapid spread). I think people are overestimating the age factor. It has certainly not helped that Italy is old, but old people who are not being treated because of limited resources is going to lead to bad outcomes.

But it becomes overwhelmed because of lack of preparedness, not adjusting behavior until too late, and having more people who need to be hospitalized in the first place -> older people.

Agree to a degree. However, a large percentage of the hospitalizations are young people. So, it's not just old people overwhelming the system.

"Large Percentage" .

Bullshit. Numbers, please.


40% according to the NY TImes. Young here is 20-54. The young seem to survive better, but they still seem to be pretty sick. I don't think long term damage is well understood yet.


I think this has to be the biggest factor. They got caught off guard and the hospitals got overwhelmed.

I think this also really hit the elderly pretty hard early on too, when they didn’t know what was going on.

I wonder if this hit the churches hard early on before they even knew what was going on and so spread to the elderly wildly.

I saw this on Twitter:


It says that most the FSSP priests and half the seminarians in France have Corona Virus after a seminarian went to Italy. That’s a very traditionalist group. Trads also tended to hold the viewpoint that this virus was not a big deal.

That Lombardy has twice the hospital beds per capita of the U.S.

Unfortunately (fortunately for Americans) Italy has also less than half ICU beds per capita, needed for the worst cases

But the higher number of ICU beds for gun shot victims drives over use by people jumping in front of bullets or firing guns at themselves.

After all, since Reagan, we know supply increase drives increased demand and consumption.

The actual number of diagnostic tests detecting SARS-Cov2 is only 10% but the demand for COVID-19 medical treatment has been driven up by Trump, economic growth genius at selling everything and creating demand.

Lombardy is a rich area, basically on par with Switzerland or Bavaria.

What is likely to happen to southern Italy will be worse, assuming that the lockdown is insufficient to stop the already existing spread beneath a tolerable limit.

True, but I checked the number of ICU beds given on the newspapers and Lombardia seems pretty much in line with the Italian average.

On the other hand, they can probably stretch a bit more than other regions: hospitals are better staffed, better funded/equipped, etc. They appear to have added more than 50% ICU beds in the past weeks, with significantly more planned in the coming days if no problems are encountered.

Adding ICU beds isn’t that easy-adding hospital beds is. One has to wonder about the quality of staffing and equipment in the erzat ICUs- as an example a recent picture of an Italian nurse in an ICU room w/a CORVID-19 patient shows her wearing her personal protective gear inappropriately

I agree, that's why US will is in a better position if action is timely.
That poor nurse had probably never worked with ICU or infectious disease patients before last week.

As of now, in Lombardy all ICUs are filled by COVID patients https://twitter.com/emmevilla/status/1241415105100812288

If that were the case, you'd see countries with a high case rate (% of total population) have a high fatality rate per case.

But they don't. Not systematically (judging by the latest statistics). See Switzerland, Austria, Norway, Denmark. All high case counts per head than UK+US (by about an order of magnitude), all much lower death rates (by about an order of magnitude).

More likely its just testing differences and non-tested diagnostic calls. High testers who include and classify lots of asymptomatic individuals have low fatality rates, low testers don't.

People do not die quickly from covid-19 - wait a couple of weeks to see how things are going in those countries.

Probably not the explanation. Deaths relative to recorded cases is completely different in the sets of above countries even at the same point since first case recorded.

I'm not sure you can just look by country. The analysis needs to be more granular. In other words, if Italy experienced a more concentrated outbreak than other countries, it would have a worse experience.

Here's an attempt to test that.

Let's test this by taking Italy by region on deaths/cases (wikipedia data table - https://en.wikipedia.org/wiki/2020_coronavirus_pandemic_in_Italy) and supplement it with latest national deaths/cases from the data links in OurWorldInData.

Plots: https://i.imgur.com/RWZQtRz.png

Small Italian regions with comparable diagnosed case/population to the European countries with relatively low death/case (Germany, Sweden, Belgium, Switzerland) generally have higher death/case.

These plots are also maximally kind to the "concentration" idea, because they account for concentration in region in Italy but not in other countries. Accounting for concentration in Spain, UK, Germany, would likely make Italian regions further outliers in death/case relative to case/population.

So I'm doubtful that "concentration and overwhelmed healthcare" explains the Italian discrepancies. This suggests under-testing in general pop / over-diagnosis of role in fatalities in Italy is likely to work better.

The best site for data I think is Oxford's site which aggregates data from the CDC, WHO and research institutions:


Lots of data and explanations of it.

If you want to understand death rates, case fatality rates, comparative rates of testing, daily counts, etc. it's all there.

Why are people looking at Twitter links for data and discussion?

I can't believe this when there is such extensive and thorough information at https://ourworldindata.org/coronavirus

I understand the value of getting contributions and discussions, don't get me wrong, but I don't know how one could footnote a twitter feed in a research paper. Is there a Chicago Manual of style for this.

Next it will be youtube.

Oh, wait, there is this excellent youtube video on the corona virus that explains it all, and, the people wear white lab coats, which means something.

It was sent to me by a doctor (I'm not saying whether he's a veterinarian or an economist, but the vet probably knows more than the economist)

Here is the viral youtube video: https://youtu.be/Hks6Nq7g6P4

Bill, those are very good sources of data, but your attitude that only doctors (and other similar healthcare health policy professional) should comment on this topic is misplaced. There are other facets to this challenge that do not fall under the domain of doctors.

For one, think logistics.

+1 important emphasis on rational thinking and collaboration.

Then, let them limit themselves to logistics. I am not saying that people should talk about what they know. Far from it.

But, let's not be naive: if you were a foreign actor you might want to have bad health advice disseminated on the internet. I am going to keep pointing to sources that are reliable and relied upon by health professionals or epidemiologist.

Also, I totally support health professionals working with other disciplines BUT you have to know who you are talking to on the other side, not that its just a bunch of economists talking to each other rather than a doctor talking to an economist.

Unfortunately, people do not put their credentials (or reputation) out there when they interact on the internet, and for all you know you could be talking to some guy in St. Petersburg who wants you to get sick or go riot in the streets.


No doubt TC has been fooled by a Russian twitter bot!

It is a good site but it has some of the common problems other sites have. Comparing raw number of cases between countries is very deceiving. We should use per-capita numbers to determine how good/bad each situation is (Italy is a lot worse in that view, China and US not so much). Also, my biggest pet peeve is the "flatten the curve" graph. You should never, ever, imply that the flattened curve will end up accounting for less infections. The flat curve needs to be thick and long, while the spike curve cannot be that tall and sharp.

While I mostly agree with this comment in that graphics shouldn't overpromise, strictly speaking, flattening the curve likely will reduce the number of infections.

Plagues percolate, they don't wash over the landscape like a flood. There are always people and areas left behind untouched, and if you slow the spread, there will be more people and areas left behind.

Right now the virus is spreading in a completely vulnerable population. Absent countermeasures, we wouldn't have meaningful immunity until it's too late. Slow things down and areas where outbreaks were bad will serve as firewalls for the areas where things weren't bad. I'd be curious to know the magnitude of the effect, but my understanding is the best modeling efforts show this. I think the Imperial College model everyone's scared about shows this.

There looks to be a geographical severity component radiating out from northern Italy, although Spain and the Netherlands may be other points of radiation.

Also, size matters. Small countries generally have much higher "case loads" (i.e. testing), but modest deaths to date. May be early days, but Switzerland's timeline is similar to Spain's, except for deaths.

Tiny Slovenia looks like it is doing very well, given its proximity to northern Italy.

And of course Germany defies the "big country dysfunctional" rule.

"And of course Germany defies the "big country dysfunctional" rule."

No, that's not really true. Germany has a high case count. They just are showing a low Death rate. At this point, I'd say it's more likely a case of elderly deaths not be classified correctly.

Switzerland isn't notably more dysfunctional than Germany and yet its Death rate is around 1% vs Germany's 0.25% rate.

And the graph above shows Sweden equal to Germany, but that's not true (at least with current data). Sweden's rate is 0.77%.

The tweet thread indicates that Germany's cases have been slanted towards the young (<60). That would certainly explain part of the differential.

The answer to that pretty much answers the question in terms of restricting the disease as much as possible. This will likely not last, but Germany has done the necessary work of at least getting systems in place for tracing and quarantining, and in attempting to keep the disease distant from vulnerable groups. This will undoubtedly break down at some point.

The denominator ("cases") is unreliable.

In terms of per capita deaths, Sweden is 1.6 per million, Germany is 0.6.

Does anyone believe 3.4% of Japanese that contracted the virus died? In terms of death per million, Japan is among the lowest in the world at 0.3.

Maybe Sweden is "riper" than Germany in terms of deaths, but the timing of deaths does not suggest so. And Japan is ripe indeed.

The only other explanation I can think of is large differences in the classification of deaths.

Germany may have done little post-mortem testing for Covid19, the assumption being that they've tested the sick so much that they catch nearly all the cases before they die.


I'm going off of someone else's translation.

If you're doing post-mortem only testing, how are you supposed to identify if a person actually died *of* cov19, rather than died *with* cov19?

The only count that actually makes sense for computing a death rate for cov19 is if you count cases that follow a clearly identified progression of symptoms related to the disease.

Otherwise you are just identifying people who have the virus and who are dead, but for whom it may not have mattered at all, may have been the straw that broke the camel's back but not the sole cause, and which in some cases it may actually have been the undisputed heavyweight champion cause.

Slovenia only borders Friuli which hasn't been affected anywhere near as badly as Lombardia or Emilia-Romagna. By the time the Italian-Slovene border was closed to non-essential traffic Friuli had only a handful of cases.

great point and understanding of political geography

Why are we spending so much time trying to come up with explanations for phony-baloney data? The denominators being used to calculate these are worse than useless.

I agree completely. The denominators could easily be false by a factor 100. And this has been discussed many times here.

Moreover, the chart is false. It says "CFR in %" but then the numbers on the vertical axis should be 2, 4, 6 and not 0.02, 0.04, etc. But who cares of another factor 100 error?

Jackpot. I think the usually excellent marginal revolution has fallen short in terms of basic math on the Coronavirus coverage. Almost all obvious cross-sectional patterns in raw data are explained by the cross sectional variation in testing patterns.

Agreed. According to "cases", Luxembourg leads the world with 773 cases per million inhabitants, Norway is 4th with 427.

Meanwhile, the UK has only 48 cases per million and Japan 8. Guh.

Sure. Everyone should just wait until all the data comes in then we can solve it. Just hope that the disease is still around.

The question isn't the numbers. They indicate something, anything from differences in testing regimes, reporting regimes, to a dramatically different strain of the virus. Or none of the above, simply an indication of the training of the health professionals. The people trying to manage this thing are looking at all this data and trying to make sense of it.

You are missing the point. The numbers we have now are mostly useless. We are putting the cart way before the horse here and trying to figure out nonsense from garbage data. We should be spending essentially all of our time now figuring out how to get the data right so we can get some true sense of what is going on and what the right strategy is. But, of course we are months behind on this and so we are screwed either way.

The Case Fatality Ratio is such a terrible measure, confounded by dozens of factors, omitting much else (like years of life), and easily inflated artificially. See here: https://drive.google.com/file/d/1v6by5x1owlLExFrmIOXjzl03cBmlF-WH/view?fbclid=IwAR33OPRhyJbWWKX_rsJEj-mqXRmYJCXixY_cl11sxSypv0Y5TxL58a9Z5Mo

Yes. One should try to make guesses about infection fatality rates (IFRs), and those guesses need to account for testing procedures.

Isn't there an inflection point when areas run out of ventilators? The elderly also live with the young in Spain, but in most of the country, today they can still get ICU treatment, so many still live. Yet in Madrid, those in senior living facilities are already being left to die, and won't even be taken to the hospital. Needless to say, that gets us to facilities where one worker is infected, they infect a dozen elderly, and they get a dozen deaths.

Italy also has a noted history of excessive deaths from influenza, so it shouldn't be a surprise that they're having excessive deaths from a flu-like virus.

Whether it's from Italy having an older population, excessive smoking or a health care system not up to snuff for dealing with even yearly influenza occurrences, the county is a definite outlier so excessive deaths form this outbreak shouldn't be a surprise..

You're right, good point. They have both higher by far the number of flu cases compared to US and higher mortality because of older age.

I think the author is looking at the data wrongly.

1) To analyze Italy vs. other European countries, you have to understand that Italy has given up widespread testing (they simply cannot keep up). The tested are now mostly the hospitalized, which are obviously the most severe cases (typically old or with pre-existing conditions, but not only). The actual number of cases in Italy is often suggested to be between 5 and 10 times larger, which would mean a fatality rate 5-10 times lower. Germany, in contrast, seems to have a better estimate of the number of true cases, not to speak of South Korea.

2)The deaths lag the infections of about two weeks. Therefore, the fatality rate appears to *increase* during the epidemic until the peak is reached. In this phase, it would be better to compare the dead of today with the infected of two weeks ago. Italy is ~10 days in front of France, one week in front of Spain.

3) On Italy vs. China: Italian share of population aged 65+ is double. Aged 80+ is four times. Note that most other western countries are younger than Italy but not as much as China.

Addition "They kiss, hug more, converse longer" could explain higher rate of contagion (which there is little evidence of), but certainly not higher mortality.

I'll bet 100 clams with this guy. Especially the time lag.

Only a week ago we were hearing that Hubei's 4% mortality rate (vs. 1% overall in China) was eviidence of people dying without proper treatment. But, this week, it turns out their mortality rate was always 1% anyway.

The graph does spawn some questions, but that Twitter thread... Yikes.

The graph of the regression results where Singapore, South Korea, and Japan seem to not be included among the industrialized economies was a wreck. The comments on that were interesting, and very harsh.

current CFRs are largely an artifact of testing protocols. Take Italy for instance. The region of Veneto, which has tested more people per capita than South Korea, has a CFR below 3%. Lombardy, which is the epicenter of the pandemic in Italy (at least for now) and with more conservative testing protocols, it's nearly 11%. It's not like people in Veneto don't kiss or hug less than Lombards. For more, see: https://twitter.com/MicheleZanini/status/1240716828436869120

sorry, typo... should have concluded with "Don't think people in Veneto kiss or hug any any less than Lombards." And BTW other ratios like age structure are different because of diagnostic approach. For more, see: https://twitter.com/MicheleZanini/status/1240716828436869120

It strikes me that CFR is really a misleading metric for COVID-19 (in addition to being hard to estimate.) COVID-19's important method of action would be better captured by CIR (case ICU ratio) as COVID-19 strikes by overwhelming the medical delivery system.

For instance, if COVID-19 progressed by either (a) being completely asymptomatic or (b) killing you in 60s, no one in Italy would be too worried now. Conversely, consider the 2009 H1N1 epidemic in the US. There were 60 million infections and over 12000 deaths and no one cared. However, if, in addition to the 12000 deaths there were 6 million ICU admissions, the epidemic would have been catastrophic.

> It strikes me that CFR is really a misleading metric for COVID-19

Yes, this is correct. Washington state has roughly 1376 sick, and that was determined by 19.3K tests for a postive rate of 6.6%. State wide, that is one in 5000.

These tests (PCR tests) generically have a false positive rate of 5-10% for mundane things like meningitis. It's really hard to know at this stage what it is for this disease.

So, if you broadly test 10,000 people that aren't sick, you end up with 500 people that supposedly have it it even though they aren't sick. And so you test again to winnow that down.

But there is zero value in comparing deaths to cases at this stage when everyone is at a different place on the testing curve with tests of dubious quality.

Deaths, on the other hand, are usually confirmed at least in the early stages of a pandemic.

Cross country comparisons may also encapsulate their culture's attitude towards the value of life.

So, a country that values life less may act differently than a country that values life more.

Similarly, a country that has more resources may act differently than a country that has limited resources.

Look at the country on the very left. Kudos to the Austrian Chancellor (= Prime Minister). He is 33 years old and has handled the situation with exceptional maturity and calmness.

After reading about that Italian-American family in New Jersey who lost 4 adult members in the last week, all the result of a March 3rd family dinner, I think there may be something unique about Italy's culture. This Freehold family is first and second-generation Italian-American, probably with Old Country customs. I can picture a lot of cheek kissing at this get-together, a lot of talking, wine and laughing at the dinner table. Plenty of droplets flying around the room. I wish a good reporter should sit down with the survivors and ask (respectfully) what exactly happened that night, where were people sitting etc. We could learn some lessons that apply to all of us.

Countries with very high rates of testing might as a result have very high rates of false positives, thereby depressing the apparent death rate. In fact, depending upon the accuracy of the test, the false positives might eventually swamp the number of actually sick people...

Add to that, different countries use different tests and test procedures, and no government run testing authority has given any estimates about the selectivity of their testing process.

I was amazed by the data of how long the virus seems active on surfaces. It was (I go by memory) 6 hours on copper, more than 24 on cardboard, and 2 days or more in aluminum or plastic.

Maybe one difference is that there are very few nursing homes in Italy, older people live at home, even when alone. Usually, they go out every day, often more than once, to buy food, to have coffee, to play cards, to read the paper in the piazza. Even when they are with people of the same age, they share facilities frequented also by the rest of the population. Unless surfaces are sanitized effectively every 5 minutes, which are not, they might get the bug from those.

I think France and Spain might be similar, while it might not be the case in Northern Europe.

With effective medical care the marginal mortality rate is near 0%. When the hospital system can no longer provide critical care to the newly infected, the marginal mortality rate goes to 6% or so, and over 10% for the elderly, average mortality rates then start rising quickly.

The numbers should either be fractions (".02") or percents (2%) but not both, as they are now (".02%").

"We will learn more soon."

627 Italian deaths today, that's bad news. Unless the March 10th lockdown starts to bear fruit in terms of lower death counts soon, Italy is on track for 20,000 deaths.

A similar trajectory in the USA would produce 100,000 deaths. That's equivalent to a really, really bad flu season. Given the disruptive measures taken to contain the virus on top of even such a death toll, it's a pretty big deal.

It's still possible that the Italian death numbers top out around here and decline, which would be welcome news that would likely play out in other countries' trajectories.

For a variety of reasons, other countries will also likely suffer less than Italy, but I'd still guess America will be lucky to get away with under 10,000 deaths and 25,000 is more likely.

20,000 deaths would be a pretty average flu season for Italy.


10,000 deaths from flu in the US is well below an average year. 25,000 is about average.

20,000 deaths plus lock the country down for a month or more is kind of a big deal though. Not the end of the world, for sure. Not the Visigoths cresting yonder ridge.

You suggested that Italy was on track for 20,000 deaths if the lock down DIDN'T bear fruit. I pointed out that that was just an average flu season for Italy.

I am genuinely interested to know what the projections are lock down vs no lock down, relative to normal flu death numbers in Italy. That seems an extremely important question for which we need an accurate answer, given the severity of the consequences of the lock down.

No lockdown is of course conjecture, but we are dealing with exponents here. Daily deaths continue to increase almost 20%, where they have plateaued over the past week. At this pace, the number of deaths doubles every four days, which would put Italy at 100,000 dead in about three weeks.

If the seasonal flu proceeded in a purely geometric progression, there'd be no one alive on earth to have this debate. It doesn't, of course, even in the absence of special measures to prevent spread. There's a good Marginal Revolution video posted just a few days ago illustrating this.

There are a lot of unanswered questions here. Perhaps the most important of which is "is this really worse than the seasonal flu, and, if so, but what factor?"

This virus is something like 10x-20x as lethal as a typical flu, so another estimate is 200,000-400,000 deaths.

Assuming no knock-on effects from overburdened health care facilities. Let her rip would have been very ugly.

We actually don't know that at all. It might be that lethal, but we don't have the data to say at the moment, because we don't know whether "confirmed cases" is under-representing actual cases by a factor of 10 or a factor of 100?

I'm not saying that you're wrong, but I'm certain that you don't have a good basis for concluding that actual lethality is between 10x and 20x seasonal flu. The current state of the data isn't good enough to make that conclusion with any degree of confidence.

That's fine. Do you have a best guess as to lethality? Flu is 0.1%.

As far as the whole planet being wiped out, in the case of Italy, total deaths would top out around 1-2 million (depending on lethality) if the whole country got infected, because the virus doesn't kill most people.

Maybe you think this is no deadlier than the regular flu. If I thought that, I'd agree with you.

This thing has killed 3,800 Italians in the past 13 days, and the daily death toll keep ticking up. This is not the flu.

When you started, you guessed Italy was "on track" for 20,000 deaths. You're not even very good at guessing.

You are guessing. I am refusing to guess. I don't think we should be guessing when the current best policy response is certain to cause a severe global economic depression. I think we should do our best to get it right.

Mr. President, I'm not saying we wouldn't get our hair mussed. But I do say no more than ten to twenty million killed, tops. Uh, depending on the breaks.

Dead bodies
Don't lie.

Coronavirus deaths are on top of the usual flu numbers, not a substitute for them.

If each virus competes to kill the same subset of people, they kind of are.

Hardly anyone is talking about strains. There are over 900 strains:


The US has been seeded from multiple sources. Individual states have been seeded from multiple sources:


These strains will differ in infectiousness, severity, mortality.

These differences are unlikely to explain all of the differences we see in CFR between countries and regions, and age groups. But they are a factor, at this point not quantifiable. They could also help to explain some of the inconsistencies in the overall numbers many others have noted.

If we are lucky, covid-19 is far more widespread than we suspect due to the spread of relatively benign strains. And herd immunity is being established, unbeknownst to us. The only way to know is to greatly expand testing, both PCR and serological, as AT advocated in another post, and to perform random testing, as advocated by Ioannidis. Without knowing we will continue down this economically disastrous path, thinking, after a peak, that we are "mitigating" and "suppressing" when in fact many may already be immune.

Speculative, yes.

"covid-19 is far more widespread than we suspect due to the spread of relatively benign strains"

Think about what you are saying. If there were strains that were relatively benign and those strains gave protection against the more virulent strains then we would simply isolate one of those benign strains and infect a large portion of the population. This would end the panic today.

There are smart people attempting to deal with this issue. If such a simple and low cost solution were available it would be done.

Note the "if we are lucky" qualifier.

> If there were strains that were relatively benign and those strains gave protection against the more virulent strains then we would simply isolate one of those benign strains and infect a large portion of the population.

How would you know? Serological tests have only now been developed. Beyond the very limited existing PCR test results for recovered patients, we don't know who has been infected and has recovered. And the existing PCR tests do not identify the strain of each infection. You have to work up a full genome sequence to know the strain at this point.

There is nothing simple or low cost about isolating a strain and then infecting a large portion of the population. You would need to do at least a perfunctory clinical trial to establish safety.

Wuhan Still Finding Symptom-Free Virus Cases, Caixin Reports


Scaling up whole genome sequencing of COVID-19

"Deep and broad sampling would provide data needed to determine incubation times, spread mechanisms, and prevalence, to gauge the extent of herd immunity, and to create predictive models that may guide containment efforts in this and future pandemics. As therapeutic options come available, we want to spot any emerging drug resistance or vaccine escape mutations."


Italy has annual mortality from seasonal flu as high as 25,000, which is comparable to the US, although it has more than 5 times Italy's population.

This translates to, on average, 68 deaths each and every day from seasonal flu in Italy. Obviously, daily deaths are much higher in peak flu season than at other times of the year. How do those numbers compare with the present situation (after correcting from deaths attributable to lack of available hospital care, which I presume is not the case during a regular flu season)?

Are there parallels/correlations in flu versus COVID-19 mortality rates when taking into consideration age, gender, risk factor (e.g. autoimmune disease, smoker, already comprised by cold/flu/recent infection, other underlying conditions)? For example, if flu mortality rate in Italy is 0.5% (notional number as an example) and COVID-19 is 5% (again, notional but would/should exclude excess deaths as a result of lack of ICU beds). Can we look at data to see a correlation that a 0.1% flu mortality in another country/area could be 1% (i.e. 10x0.1%) COVID-19 before excess deaths?

It's probably not very different, the denominator is just very wrong. Someone should send 5 000 random test kits to people in every country to know the real spread

Yes, it's likely that we aren't getting good estimates on the spread.

I think one reason for the high death rate in Italy is their good healthcare system.
In the past it has kept many old people alive way past their "sell date". They should have died earlier from [insert chronic condition here], but did not.
Covid-19 was a tipping point for the individual.

Death rates correlate strongly with ICU beds. Everyone that dies is first in the ICU.

Below are deaths per million population and in parenthesis are ICU beds per 100K. You'll often see people like Rachel Maddow ding the US on hospital beds per 1000. The US performs poorly there. Why? Because our medicine has advanced to the point that things like knee surgery is done on an outpatient basis. In other words, as you get really advanced at medicine, you need fewer hospital beds.

ICU beds, on the other hand, are for the sickest of the sick. The richer the nation is, the more of those you have. Because rich people don't want to die. The US and Germany both have a lot of ICU beds.

Note in the data below Italy is running about 20X China for deaths normalized for population. Spain is 10X. France is 2X

Deaths from COVID Per Million Population (ICU Beds per 100K):
Italy: 56 (12.5)
Spain: 22 (9.7)
France: 5.6 (11.6)
China: 2.4 (3.6)
UK: 2.2 (6.6)
USA 0.66 (34.7)
Germany: 0.58 (29.2)


Mani Puliti?
It destroyed Italy.

I wouldn’t be surprised if inhaling diesel passenger car fumes for years make things worse in certain European cities.

Depends a lot on who is keeping and how accurate the numbers are. GIGO.

The graph says Y axis is in % but then has nominal decimals. This is incorrect, no? Should multiply values by 100 if calling it a percent, otherwise can mislead people to think this is 100X less deadly.

The obvious key is how widely testing being done. I've seen reports (without specific numbers) that Germany is testing many more people than other European countries. In some countries only clearly ill patients with serious symptoms are being tested, even if they've had contact with known positive cases. Only a few countries are actively testing people who are not symptomatic (e.g., they had close contact with an known COVID-19 patient). Those countries are picking up many more asymptomatic patients. Many of those symptomatic patients never get ill enough to have been tested otherwise, but they add to the denominator of known cases. Hence, the CFR is far lower. This seems to be what's happening in Germany - a few days ago, they had a similar number of known cases to France, but many fewer deaths. In the last few days, Germany has detected far far more new cases than France has, but very few of those cases are listed as severe cases, and their CFR continues to fall. They are clearly testing far more extensively than other European countries.

Several countries have so overwhelmed their testing capacity that they are have raised their criteria for testing several times, and are now testing only the sickest patients. They are simply not picking up those who will go through an infection with mild or no symptoms.

Another confound is that some countries were reportedly not testing the deceased of the virus if they weren't known to have been infected. I've seen reports that Italy was testing deceased patients, so that may have inflated their numbers somewhat. And of course now their system has been overwhelmed for a couple of weeks, and people who are dying who would likely have survived if they'd have been treated a few weeks ago.

The confounding factors are so severe that these comparisons are pointless. Virtually every country has changed their testing criteria at least once or twice, sometimes in dramatic ways. The best we can do to understand the serious illness and fatality rates for this virus is to look at countries that have been dealing with COVID-19 for the longest time, have had the infection rate peak and settle down, and have done extensive testing. That excludes all of Europe and the US.

Forgot to add - it's pointless to look at CFR in countries where the known cases are doubling every 2-3 days. By the time the sick patients in the first detected cohort die, the known case rate as doubled 3-5 times. Remember that from onset of the illness to death is usually 2-3 weeks. Until the new case rate stops growing exponentially, the numerator and the denominator are out of sync. In a country doing a lot of testing, this will make the CFR look really low (I think Germany is a case of this, and expect their CFR to "catch up" to some extent, while the CFR in, say, France, will fall as they test more widely).

Thank you. You might also look at the Johns Hopkins site for additional info: https://coronavirus.jhu.edu

Might matter if countries caught their first cohorts (and subsequent cohorts) near onset of disease, then you'd get a few weeks lag.

But of course they don't. Not particularly as a rule.

In truth, this data is simply misleading.

The CFR in the chart for “Australia” shows about 0.02% as of March 12. What it meant was, in fact, 2% not .02%. There were 5 fatalities out of 157 cases. A VERY small number and, as Ioannidis suggests, almost certainly nonsense.

A of 21 March it’s the CFR in Australia is 0.7% (7 deaths out of 876 cases). And that’s likely indicative of nothing, too.

I would expect the most important explanatory variable by far is the number of people entering the country with COVID19.

Does that data exist?

What about air pollution. Northern Italy and especially Lombardy is the most polluted area in the western world in terms of pm2.5 and pm10 and this is obviously as bad or worse for lungs than smoking.

Comments for this post are closed