>Why would we use the Diamond Princess CFR instead of China's, or Italy's, or South Korea's?
because everyone on the diamond princess was tested. So we know for sure how many cases we are dealing within the sample.
~0.3 and ~0.9% are also not optimistic guesses but the current numbers for Germany and SK. Italy sits at 9%. So the situation is either that Italy is vastly underestimating cases, or Germany and South Korea have lost a magnitude of corpses somewhere. I find the latter less likely than the former.
South Korea's naive case fatality rate (CFR) is already 120 deaths/9037 cases = ~1.3% today, gradually going up from ~0.5-0.6% a few weeks ago. Why? People in a functioning healthcare system take time to die and these people were infected during an expansion phase of spread which rapidly increases #cases (denominator).
SK's cohort CFR is even higher. More properly, we should use the infection number from 3 weeks ago because it takes 3-4+ weeks from exposure to death: 120 deaths/4335 cases = ~2.8%
Germany's current naive CFR at 0.4% will also rise in a similar manner for the same reasons. (You can bookmark this.)
South Korea's and Germany's hospitals were never overwhelmed in the same manner as Italy's. A major reason Italy's fatality number is so high is because doctors there cannot save everyone anymore.
>South Korea's naive case fatality rate (CFR) is already 120 deaths/9037 cases
But the denominator in this formula strongly depends on who and how often you test. In other words: You don't know the number of cases.
Obviously, people who are severely affected are tested more often. People with mild or no symptoms might never be tested, even if they want to (I'm not sure about South Korea but for sure this is happening in Germany).
Based on people in the German parliament and the German soccer league, you can currently guestimate that 1% of the population is already infected (1% of the parliament and 1% of the premier league players are infected. I suspect that they are tested more often and even without symptoms. Maybe they have more contact to other people - maybe not).
Yes, this is a wild guess, but much better than taking the confirmed cases which are heavily biased towards people where the infection causes problems.
Yes, all my figures above are Case Fatality Rate (CFR) and not Infection Fatality Rate (IFR), which includes people with mild or no symptoms. It's much harder to estimate the latter unless one conducts antibody tests on a sufficient sample of a population. South Korea's extensive testing program should bring their two figures closer than those of other countries.
Credible estimates of IFR from noted epidemiologists I've seen are around 1%, assuming that the healthcare system still functions, and much higher otherwise.
COVID-19's CFR & IFR might not even be the biggest problem. High rate of hospitalization and broken healthcare system, with all their ramifications, could be considered even worse.
There is no real consensus regarding IFR yet. I think the best data we have is from Diamond Princess, which is at least 10/712 or ~1.4% and may go up a bit from unresolved 15 serious/critical cases and 100 more active cases. The population there is older, but also have good care.
If the Diamond Princess age group represents just 20% of a population (they are not all elderly), population IFR must be >= 1.4%/5 = 0.28% and likely higher. 0.28% is above the IFR upper range from the paper in your comment.
“Estimated fatality ratio for infections 1%
Estimated CFR for travellers outside mainland China (mix severe & milder cases) 1%-5%
Estimated CFR for detected cases in Hubei (severe cases) 18%”
If you have the raw data you can easily estimate CFR even though the number of infected are increasing by doing a culminate graph over "death share vs time since symptom onset or diagnosis" for the cases.
Up to 8 weeks though I haven't found typical distribution/median. The increase in deaths might exponentially grow for a while after new patient load stabilizes.
John's analysis is cherry picking in many ways. Death rate in Diamond Princess is 1.1% today and 2% is listed as severe. Assuming 50% of severe make it, final fatality rate for the ship may end up closer to 2%. John then adds a 50% discount factor but it is not clear how he picked that number. Also, the 1% of population infected seems to be another number pulled out of a hat. If we are basing our figures based only on the ship with no other assumptions, we have to go with 20% infection rate. Thus, one reasonable estimate of risk from the ship data is 20%x2%x330M = 1.3M deaths if we wait for "evidence" and did nothing. Clearly, this argues for doing something!
Edit: Also, Germany does not test dead folks for coronavirus while Italy does. Further, SK death rate has gone up to 1.3% (0.9% is an old number) and many more are in severe category. Thus, the sub 1% numbers seem more like the outliers than the above 1% numbers.
The Dimond Princess was evacuated. It’s passengers where unusually healthy for their age range, and while older than the general population had few people over 85 which is the most at risk population. Further, these people got world class care from experts and whatever minimal care an overworked heath system could provide.
Given all that they still had 9 deaths out of 712 infected with many still in critical condition.
You don't have to be bedridden to have lungs that are one cold away from death.
People with advanced COPD etc are everywhere but can walk short distances etc and prefer cruises to schlepping through airports and whatnot.
Knowing people who go on large long cruises they tell me they've never been on one where they didn't have at least one death. Indeed I know people with serious health issues who go on these knowing there is good on site medical care at hand.
Large ships have a lot of passengers. In the US Men hit a 2% chance of death at 68, which jumps to 3.6% at 75. At 85 that jumps to 9.6%, and by 95 your at 26% and the numbers keep increasing.
This means you can’t simply look at the average age to estimate risk factors. Still a 2% risk of death per year x 3000 people = 1.15 deaths per week ignoring crew. In other words what you’re describing is still a fairly heathy population.
I am pointing out curse ship populations are actually at lower risk than society for this specific disease. The crew is all young and it’s mostly irrelevant if someone is 4 or 40 relative to people being a heathy 80 or sick 90.
That's just not true. It's pretty much impossible to be 80 years old and not be more susceptible to infection generally. Statistically the people dying have an average of 2.7 comorbiditities.
Average is meaningless in this context. US Men hit a 2% chance of death at 68, which jumps to 3.6% at 75. At 85 that jumps to 9.6%, and by 95 your at 26% and the numbers keep increasing. A 50:50 mix of 85 year old men and 38 year old men have vastly higher risk of death than a group of just 62 year olds.
Except those higher odds of death are strongly associated with major heath issues. So, simply excluding the sickest 5% of the population makes a huge difference in survival rates.
That also seems wildly optimistic. 80% seems like a more reasonable assumption than 20%.
Also - 99.9% of those patients (pulled out of a hat) wouldn't have access to health care because the capacity was already overwhelmed, so the death rate will jump markedly.
Because 712 out of the 3711 passengers and crew were infected, and 713÷3713 ≈ 19.2%. So that gives us some sort of vague idea how much of the population from which the ship was drawn will become infected if exposed.
You don't need to be particularly fit or healthy to go on a cruise. Yes, sure, you can't be on life support, but generally 'healthy enough' to travel on a cruise is exactly what I'd expect from any random sample of the overall population.
The problem is the disease is mostly killing off the least heathy. Exclude only 5% of the population and deaths might easily drop by 1+%. Further the Dimond Princess was evacuated specifically because they could not contain the spread. Suggesting their rate of infection is indicative of anything would mean we had somewhere to be evacuated to.
You may not understand what "random sample of the overall population" means if you would expect every single person in the sample to be healthy enough to travel on a cruise. Almost 1% of the population in the US has Alzheimer disease or other dementias, for example.
I noted exactly that in my comment. Yes, not all the population can go on a cruise. No, it's not like it's only the healthiest 20% of the population that can. If I randomly sampled the population I'd expect the majority to be capable of a cruise. What do you think cruises are like?
If you agree that they are not a random sample of the population - they are healthy enough to be fit for travel (like the majority of the population) you are not trying to contradict ant6n's comment as I thought. I misunderstood, my apologies.
and it's not a small travel (e.g. the British tourists that went on board the Diamond Princess).
If you are not in good health at the beginning, you don't adventure yourself 10 hours+ from your home.
So this group is likely in better shape than average population.
Isn't the average age on a cruise ship far older than in the general population? I would've thought the people on board would on average be far more vulnerable.
It’s a narrow band excluding the young and oldest so, the average is older but the maximum is younger. With a very sharp decline in their 80’s, which is when things really get bad.
Considering how quickly the numbers get worse with age and ill heath many countries are at higher risk.
~0.3 and ~0.9% are also not optimistic guesses but the current numbers for Germany and SK.
Fine numbers in the presence of a health system that is not overloaded, or a country that has managed to make effective changes to prevent transmission (how did they do that without science huh?)
By John’s numbers (0.3% die and 1% of population) Italy should get a total of 1800 deaths. Yet Italy is at 6000 and rapidly rising - using real numbers his assumptions are already wrong for a first world country that is a few weeks ahead of the rest of the world.
And why the fuck does he assume 1%? Because some actions have been taken? What actions can be taken since by his own words we lack evidence to make decisions...
Edit: by my calculations the US has 800000 cases at the moment (compared to ~40000 tested positive). 500 deaths with a 0.5% death rate, so three weeks ago there were 100000 cases, but it will have doubled 4 times in 21 days so there is now 800000 cases (already 0.2% of population). Three more doublings (easily realistic) beats John’s 1% within weeks. Ironically, going with his low mortality rate (0.025%) would mean US has 2% infected already...
> By John’s numbers (0.3% die and 1% of population) Italy should get a total of 1800 deaths. Yet Italy is at 6000 and rapidly rising - using real numbers his assumptions are already wrong for a first world country that is a few weeks ahead of the rest of the world.
Just highlighting this bit of the parent’s post. If you want hard evidence that COVID-19 is quantitatively and qualitatively different from other coronaviruses that, as John puts it in his article, “actually infect millions of people every year”… Italy is it.
It's also possible that Italy is vastly unlucky for some generic reason, considering most of the deaths are in a particular region of the country and among an age bracket that is less cosmopolitan and more likely to reflect limited geographic genetic dispersion.
I suppose central China, northern Italy, Iran, Spain, the rest of Italy, and soon enough France, Germany, New York City, Israel, followed by plenty of other places just happen to be vastly unlucky for “some generic reason”.
While age is probably a factor in Italy, it's less of a factor in China, Iran and Spain, which also have high death rates (though the Chinese high death rate was mostly in the early stages). What these cases have in common is a very large surge which overwhelmed local health services.
If it was entirely based on age, you'd expect higher rates of death in Japan and Germany (both very elderly populations), lower in China (less elderly population) and much lower in Iran (young population).
It typically takes 2.5–3 weeks after the start of symptoms for hospitalized patients to die. It only takes a few days to a week for a positive test result to come back.
Up until recently people hospitalized in NYC has had access to doctors and equipment, but NYC hospitals are already on the verge of being overwhelmed, and the crisis is just starting there. 2–3 weeks ago the “CFR” (i.e. deaths to date divided by known positive cases) was also very low in Lombardy.
The mass social distancing interventions they have undertaken in NYC should hopefully start kicking in, and we can all hope that the situation doesn’t get as bad as Lombardy, but in the mean time there are going to be thousands if not tens of thousands of deaths there, and it looks like doctors may soon end up facing choices about who to put on ventilators.
Italian here. The most probable cause for that is that these areas started being affected earlier, so the virus had more time to spread before the lockdown.
That doesn't affect cfr, which was higher even before the hospitals got overwhelmed. I suppose it's possible that Italian doctors are just generally bad, but I doubt that.
CFR is affected by the methodology with which you count the positive cases. Italy is badly underestimating the number of affected people (the head of the Civil Protection service says that we could have 10 times more cases than those accounted for).
I've seen the claim that northern Italy's manufacturing industry is more integrated with China, so there was more cross traffic. I don't know how true this is.
For sure it’s more integrated than southern Italy, but I don’t know and can’t speak about other EU countries. It’s entirely possible the higher traffic brought in the first asymptomatic cases earlier than in other regions.
I think SK's numbers are becoming more reliable by the day though. The new case rate has stabilised to a small number in the range of 50 to 150 per day and the active case count is dropping at a rate of over 200 per day. The daily death rate has been below 10 the whole time I believe. If that trend continues then SK's CFR will be well known soon and I would guess not dissimilar to current estimates. Though it's always possible a false sense of security will set in and people will relax their habits and send it higher again.
Compare to Germany: weeks behind SK but already more than three times as many cases, new case rate in the last few days of 2500-4500 (SK's max: 851), daily deaths in the last few days 10-29 and heading north. Actually the death rate must reflect an amazing health care system given 30k cases, but it's early days for Germany. Their pipeline is very full, agreed I wouldn't want to make a prediction there.
OK, SK's new case rate is stabilizing, but they still have 5400 active cases. If even 1% of those active cases die (which is possible, since these longer-lasting cases are likely more severe), that's a total CFR of 2%. And this is for a country that everybody claims has been doing contact-tracing and testing asymptomatic people.
I find it interesting that SK is always brought up in these discussion about CFR and how now action is required. Actually SK is the prime example of a country acting quickly and early (also showing that general lockdown is not necessary in that case). They would have been even better off had it not been for patient 31.
Significant evidence that 50%+ of cases are asymptomatic or very mild and those people are not being tested at all. In the Uk even quite bad cases don't get a test
> Their pipeline is very full, agreed I wouldn't want to make a prediction there.
German here. I assume the recent hard lockdowns will work out pretty much for us... I'm more worried about the US, this is gonna be a mass die-off, and the Trump government's handling of the issue is... let's say abysmal.
German resident here. Why do you assume the recent hard lockdowns will work out? I have discovered in my time here that the German reputation for orderliness and rule following is exaggerated.
South Korea coped with the outbreak by having a test early, test often strategy, but the German strategy seems to be test eventually, test perfectly. That means that there isn't any process to flag essential workers and others as needing a good proper test. Korea's showed it's better to do a test with a high false positive and even a significant false negative many times a day and get the person out of circulation awaiting an accurate test, than to wait for them to find the symptoms concerning and ask for a proper test.
China coped with the outbreak by having actual curfews. Major lockdowns. The sort we couldn't reasonably expect. When I went to do my weekly/fortnightly shopping yesterday, I saw several police officers looking around into restaurants and on the local town square. Not hard to hide from. No-one cared what my business was.
Italy still hasn't really peaked. They did this test-free lock down strategy that Germany is doing. Apparently the amount of intercourse required for viral transmission is ridiculously low.
There's already tens of thousands of sick people here, and the government was very lethargic in their response. The peak will be huge. As I mentioned before, they gave up after Gangelt and seemed to act as if the whole thing would be minor. It took weeks after discovering a major problem existed that needed hard work before German authorities actually agreed to do hard work.
Learning lessons seems to be really hard for authorities at the moment, and I'm genuinely worried. It's like even ideas are subject to the European protectionism - better import a bad idea from Italy than an effective one from South Korea. My goal is to not get ill before there's space in the hospitals again, because any other goal seems unrealistic.
Germany didn't lose control last week. A few days after they realised what happened at the Gangelt carnival, they gave up - that's when they lost control. The spread of cases took a couple weeks to be shockingly high, but this was locked in already in the first week of March.
We must stop confusing the outbreak of diagnoses with the outbreak of cases. That's the mistake every government agency in the West has made, and it's why it spreads faster and better in the west than in South Korea or Taiwan.
Isn't it the circumstances that are being considered, that are the problem? Germany "lost control" of the pandemic. (Germany didn't actually lose control last week, it's just the numbers stopped growing fast in the "ho hum" range and started growing fast in the "I'm scared" range. This is how exponential growth works. Germany lost control when they decided to do nothing for two weeks after noticing community spread.)
It takes ~2 weeks for a person to die after they get infected. So when you are looking at numbers from Germany, you should look at current death count / number of infected somewhere around 10-14 days ago.
While virus is actively spreading, taking "current death count / current total infected", can easily underestimate mortality rate by 5-10x because it takes quite a while from infection -> death.
Go check South Korea's numbers again. As their cases have resolved, death rate has steadily increased, and it's now at 1.33%. It will continue to rise. Same with Germany, it just takes time to die. South Korea had already found majority of their current cases 2-3 weeks ago, but deaths are only now picking up.
Edit: now 1.37%, up again since I checked a couple hours ago.
The same rise in death rate was observed way back when SARS happened. At first people estimated death rate at 2-3%, and it was continually revised upwards as cases resolved.
The opposite happened with H1N1 (swine flu) in 2009, though. The CFR was estimated to be significantly higher during the pandemic but afterwards was estimated at 0.02% with 60M American cases. [0]
Policymakers should use conservative estimates to be careful, but we still just don't know how deadly COVID-19 is.
The unanswered question of the GP is why you'd expect only 1% of the US population to be infect?
I can't see the slightest basis for such an assumption. This an extremely infectious, quickly spreading disease. 30% of the US population seem like a more likely estimate.
SK has a 1.3% fatality rate at the end but that was with a functioning health case system. If even 1% of the US population get infect, the health system won't be function and you'll have a higher fatality rate.
Once the hospitals run out of ventilators and supplies and once the doctors are all sick themselves. Then sick-with-serious-viral-pneumonia = dead, for a large portion of the sick. Infected != sick, no doubt but we know the ratio of infected to sick over time, despite Ioannidis' disgraceful efforts to cast doubt here.
>The unanswered question of the GP is why you'd expect only 1% of the US population to be infect?
Ioannidis addresses this in the article. Extensive community spread is actually unlikely to be the case for this virus, epidemic development is hard to discern from simply increasing rates in testing and sensitive populations seeking testing, and maybe most importantly there is little evidence that lockdowns and other extreme measures have significant impact on reducing this sort of respiratory infection, he cites this paper. [1]
" The highest quality cluster-RCTs suggest respiratory virus spread can be prevented by hygienic measures, such as handwashing, especially around younger children.[...]Global measures, such as screening at entry ports, led to a non-significant marginal delay in spread. There was limited evidence that social distancing was effective, especially if related to the risk of exposure."
But talk about lack of data. This is looking at the spread of a different virus - a virus that certainly does not have the infectiousness of the Covid virus. So the this pure speculation.
It's a meta-study looking at over 67 papers that address the question of the spread of respiratory viruses. Obviously, every virus is a new virus, that is always a problem.
The question remains however why we ought to treat concerns about data about the virus different than concerns about data about the response to the virus. Why do we treat the virus like a black swan event, but not the unprecedented response of shutting economic and civil liberty down to a degree maybe not seen in 100 years?
It seems ironic that people critize Ioannidis for a sort of first-order error in thinking by not considering uncertainty. Yet causing damage and applying first-order thinking to disruption of global supply chains that likely will drive entire nations into deep recession and instability is apparently adequate.
It's a meta-study looking at over 67 papers that address the question of the spread of respiratory viruses. Obviously, every virus is a new virus, that is always a problem.
Yeah, and if any of those other viruses caused a world wide pandemic, I missed it. The article is a specious disaster.
Why do we treat the virus like a black swan event, but not the unprecedented response of shutting economic and civil liberty down to a degree maybe not seen in 100 years?
An extreme provokes an extreme response? Of course?
Human lives are more important than economies. And economies can't function with massive loss of life anyway. Even more, this is a massive exogenous shock. Once it's done, the various players can pick up the pieces. Until then, it should be treated like a war. Society trumps economics (hopefully, otherwise both are headed for disaster, 1 million deaths+ was the Imperial College Report estimate for what happens if the US does nothing).
>Yeah, and if any of those other viruses caused a world wide pandemic, I missed it.
They do actually. Several of the outbreaks studied among the papers were influenza pandemics, coronavirus pandemics, and SARS. (page ~110-120)
This rhetoric you're starting here about bringing out the war drums to fight invisible enemies is exactly what Ioannidis is afraid of. It is not scientific, it is not based on evidence, and it does not, weigh the tail risk of a global economic breakdown. Which may, in fact, be literal war in some places.
There is a trade-off between the economic effects and response to the virus. It is not a binary question.
This rhetoric you're starting here about bringing out the war drums to fight invisible enemies is exactly what Ioannidis is afraid of. It is not scientific, it is not based on evidence, and it does not, weigh the tail risk of a global economic breakdown.
Science is a means of discovering the most likely state of things and an always uncertain one. Other human institutions have to come into play when it is necessary to act. Those institutions make the trade-off rather simply calculating them. In the current context, the institution of a war is appropriate (more appropriate than all the semi-wars we've had over the last 50 years in fact, better than "war on drugs" or "war on terror"). We confronted by tiny semi-living creature that happens to be very good at killing us. We should band together and engage in unified, determined action to protect ourselves.
As far the economy goes - the economy is a phenomena of society. The productive machinery should kept going as much as practical and the entire process managed by the government, essentially a machinery akin to WWII needs to be in place for the duration of this. Such war measures kept things running at that time and there's little reason to think they wouldn't work today. Now, as far as lots of people losing their investment. Well, sorry, investments aren't life. This, in a sense, very quick trip from 1929 to 1948 for y'all.
But the government isn't aiming to keep the productive machinery going as much as possible. They're doing precisely the opposite, shutting down productive machinery as much as possible.
Bogus wars against viruses would've been totally unnecessary if leaders had prepared for this eventuality and not shutting down long-term preparedness. Look at this as payback with interest to that debt. Waiting for evidence would lead to worse handling, and shut everything down during panic time instead. Ensuring even worse outcomes.
> Italy sits at 9%. So the situation is either that Italy is vastly underestimating cases, or Germany and South Korea have lost a magnitude of corpses somewhere. I find the latter less likely than the former.
Both could be true. For example, Germany doesn't systematically test dead old and/or hospitalized people for Coronavirus infections while Italy apparently does. Germany has more resources to test potentially infected people than Italy, where all resources are needed for treatment of the hospitalized cases.
In reality, both Italy and Germany will have infections in the 100.000s, most of them with mild or no symptoms.
That's a false dichotomy. The situation could also be that access to treatment is a huge factor in mortality, and that Italy and Wuhan faced a large enough caseload that people were unable to get healthcare, which drove up mortality significantly.
This is a much more likely factor than the idea that Germany and South Korea are somehow 10x-30x more effective in testing their population than Italy.
I'm not sure about Germany, but South Korea was and is extremely effective at testing their population. They also did and are doing a fantastic job tracing infections (although part of this is due to being able to mostly focus on a single super-spreader rather than many simultaneous outbreaks. I think that South Korea probably confirmed 70-80% of their actual number of cases, while Italy probably has at least 2-3 actual cases for each confirmed case. Testing and tracing en masse is really, really hard, and is basically impossible if you don't catch it soon enough. I think that hospital overloading is definitely able to double or triple the death rate, though.
Deaths are not instantaneous, add exponential growth and it taking up to 9 weeks from infection to death and it’s really deceptive. All deaths tell you is how many people where infected weeks ago.
Just look at the lag in China’s rate of infections vs deaths. They had 22 deaths on March 8th and 40 new infections. Further, new cases drops off vastly faster than infections with the sickest talking longer to get better and staying at risk of death for weeks.
Italy is a very elderly population. If, as seems to be the case, death rate ramps up dramatically over 70 and especially over 80, then this may explain part of it (though clearly testing in Germany and SK has been more effective than in Italy, too).
Prof. Drosten who developed the test says that cannot happen in practice. One would get a false positive result with SARS-1 (de facto extinct) and related coronavirus strains in Asian bats (that have not crossed into humans).
How many HIV deaths can be directly attributed to the virus? Italy’s way of counting may be preferable to the method in other countries (which I assume will be revised at some point when the full history of covid-19 is studied). And even Italy is undercounting!
I know this may sound ridiculous, but Kimchi and Sauerkraut? Are any other countries with a high consumption of low pH fermented foods also showing a lower CFR?
because everyone on the diamond princess was tested. So we know for sure how many cases we are dealing within the sample.
~0.3 and ~0.9% are also not optimistic guesses but the current numbers for Germany and SK. Italy sits at 9%. So the situation is either that Italy is vastly underestimating cases, or Germany and South Korea have lost a magnitude of corpses somewhere. I find the latter less likely than the former.