Where does all the heterogeneity come from?

Here is a Christopher Balding tweet storm, excerpt:

Iceland has done almost 14k tests on an island of 360k so more than 3% of the total population…They have more than 800 confirmed cases, 10k people in quarantine, 800 in isolation, 18 hospitalizations, 6 in ICU, and 2 dead…About how many people SHOULD have corona if the spread etc numbers are accurate. As of March 27, Iceland would be expected to have more than 46k people that have corona. Emphasis this is on an island of 360k and 800 confirmed cases.

What is going on in the Icelandic numbers?  What accounts for this apparent heterogeneity?  Dosage?  Is it that Icelandic clustering is mostly in one easy to control central city and the rest already is “socially distanced,” even in the best of times?

I know there are some MR readers in Iceland, and presumably they read the Icelandic press.  Can anyone shed light on why the death rate is not higher in Iceland?  Is it that the death rate is about to burst a week from now?  Alternatively, you might think the Icelanders have kept their hospitals up and running — important for sure — but that doesn’t explain what seems to be a quite low rate of reported cases.  Or is it that Iceland’s second largest city is so tiny — Akureyri at 18,925 inhabitants — that the virus doesn’t have many easy chances to recirculate once cut off for a while?

Similarly, Sweden hasn’t restricted public life very much and they do not seem to be falling apart?

How much better is Staten Island (less dense) doing than Manhattan (more dense)?

Some reports indicate that in hard-hit Westchester County,. NY, the rate of hospitalization is about one percent (8-10 percent in some other places).  Alternatively, here is serious talk that the death toll in Wuhan is 20x official figures.

How much of the heterogeneity results from the kind of mixing you get?  One account of the low German death rate is the young and the old were never pushed together so much by the policy response.  One account of the high Italian death and hospitalization rate is that the initial quarantine was only regional and thus it spread very dangerous forms of mixing throughout the larger country.

It is possible that Cambodia, Thailand, and Vietnam still will be hit hard, but so far the signs do not indicate as such.  Warm weather may play a positive role, though that remains speculative.  The latest weather paper appears credible and indicates some modestly positive results.  Of course weather won’t explain the relative Icelandic and Swedish success, if indeed those are truly successes.

By the way, on the “everyone already has it” theory, a semi-random sample of 645 from Colorado showed zero positives.

So where is all this heterogeneity coming from?  Is it all just bad data?  That seems hard to believe at this point, and Iceland seems like a plausible source of reasonably good data.

As for concrete conclusions, these heterogeneities should make us more skeptical about any models of the situation.  But it would be wrong to conclude that we should do less, arguably risk-aversion could induce us to wish to do more, including on the lock downs front.

It is also worth pondering which heterogeneities are “baked in,” such as heat and age structure of the population, and which heterogeneities can be altered at the margin, such as forms of social mingling.  It is at least possible that studying these heterogeneities could make policy far more potent.

Overall, I do not see enough people asking these questions.


My guess is we'll find out that there are problems with the swab test. There's a small window where you test positive via the swab test. You have to swab in exactly the right spot and/or at the right time.

Or quarantine/isolation (of an entire household) when symptoms first appear, and contact tracing followed by quarantine/isolation (of an entire household) until one tests negative/following 14 days, is an effective public health strategy that works just as well with coronavirus as all the other infectious diseases handled this way.

Exactly. This strategy constrains the chains of Community transmission. It depends on a large percentage of the population Self-Monitoring for the Onset of symptoms and Self-Isolating. Social Distancing is a safety precaution to protect against the unaware and the non-compliant.

Deaths are often a reflection of the inability to isolate the vulnerable in Long Term Care facilities as reflected in the outbreaks in Washington State and Canada.

why hasn't group testing been tried?

Because we are in a tight OODA Loop. Testing capacity can not outpace the rate at which new cases are imported from hotspots and testing capacity is best used to stop chains of community transmission.

> why hasn't group testing been tried?

It has, but even mature PCR tests have a 5-10% false positive rate. In hotspots like Seattle, the prevalence rate is in 1 in 2500. If you test 2500, the PCR test we have today will likely report 125 to 250 have it.

in vivo early in a viral pandemic
5-10% false positives might actually increase efficacy of public health attempts at isolation/quarantine/handwashing/decreased face rubbing
and reduce spread of disease

The swab based PCR tests work best for symptomatic people. Understanding the percent of Recovered people in a population can wait for the antibody serology tests to reach capacity.

Brave of you to bring up a story which doesn’t promote the concept that COVID-19 is not the end of the world. You hide it well by calling it heterogeneity you may not be punished.

If this hysteria turns out to be ill founded then heads will have to roll. We have spent trillions (maybe 10s of trillions) on this debacle. The math never penciled out for such a response, even under the disaster scenarios.

The media has to be front and center. They started, stoked, and perpetuated the hysteria. We need to fix the glitch.

"The media has to be front and center. They started, stoked, and perpetuated the hysteria"

So you are saying ... it was a hoax to bring down the orange man? Just kidding!

If it bleeds it leads, and hardly anything bleeds like viral pandemic.

Your autocorrect made a mistake - clearly, you meant to write hardly anything breeds like viral pandemic.

EdR is Chinese?

Your view of factors is too narrow.

Calculate long term benefits vs short term costs and you'll find this is actually extremely cheap and beneficial for humanity as a whole. Crash your car at 30kmh and get whiplash and total the car. Crash it at 300kmh and you and your passengers are oil for the next dominant life form on the planet.

The media? The media has just reported the concerns of virologists and epidemiologists. Stop attacking the messenger. People make a choice what media they consume as well.

Actually, the job is to be more than a messenger. We have already seen the problem with the decline of journalism in the politician X said and politician Y said, without doing anything to clarify which, if factual, was correct.

In an ideal situation the journalists would have gone beyond what the epidemiologists were and are saying. In 2009 the European Council investigated the hysteria over the bird flu and concluded that the WHO had depended too much on what big pharma was saying.
Currently we simply have a) case and body counts, b) what the politicians are saying, and c) hoaxes floating around the internet -- with perhaps a decent critique or two thrown in. With good, decently funded (a real problem) journalism we would be hearing reasoned critiques.

Ah, and we have to remember that the press is also reporting on the insane things that Trump says. Even The Guardian has had an article speculating about where his notion of "reopening for business" rapidly came from. Does that matter?

Though it was the Post that referenced the Epstein work (which has, as of today, underestimated the American deaths by 1500 people) as being the source of Trump's thinking about having the churches full for Easter. Epstein's work looks ridiculous as of today, but such is the difference between people making things up and reality.

There are two issues in your comments that need to be identified - 1. Hindsight is 20/20 - once you have the test answer you figure why you were wrong not before. 2. People are much more risk averse than ever before. People are not used to swing deaths around them - as much as they used to when out of every 10 children your mother had she would loose 2-3 at delivery or early on due to contagious diseases.
Great discussion nonetheless

I think those talking about 'hysteria' should look at Singapore. Singapore has probably the most pragmatic and effective government in the world, the last place which would allow policy to be guided by hysteria or twitter mobs or left-right ideological debates. Yet they introduced measures back in mid January and have been gradually tightening week by week and now getting close to a lock down situation as we see in other countries.

Singapore are competent, but so are Sweden and they have almost a 180 set of policies. And there are yet to be any problems there like in Italy. It will all depend if a month's time Sweden is a mess then we acted wisely now. Otherwise we will need to look out for what the worst overreactions were and why they occurred.

Sweden has more death per capita than China. They really should look into that.

Is Tyler using outdated data for Sweden?

As of right now on the Johns Hopkins site, Sweden has 110 deaths out of 10.3 million people. That's more per capita than Germany (482 / 83 million), or Norway (25 / 5.3 million) or Finland (11 / 5.5 million) or the Baltic republics, for instance Estonia (3 / 1.3 million), or Canada (63 / 38 million).

Sweden is better than the UK or France or Spain or Italy but it hardly seems exceptional.

It's pretty clear that there is a public consensus in Sweden not to make societal changes. Just as they declined to make changes in security procedures when Olaf Palme was assassinated, because then it wouldn't be Sweden. In such a small country, the raw number of deaths is not (yet) sufficient to cause alarm, and the public is apparently willing to accept per capita deaths on the same scale as would result from a major flu epidemic rather than take dramatic preventive measures.

Before anyone writes or posts about Sweden, maybe they should look up some information on Sweden


They are worse off than the U.S. at the current time and the situation will probably get worse.

There's just insufficient data to determine an optimal policy right now.

Death toll can reach 100k, or 100M, both scenarios are plausible.

The virus can circulate for several years, or it can be eradicated by late summer.

Maybe we'll get by with cloth masks and disposable gloves; maybe they are placebos.

Perhaps the infection is a non-event for 80% for the people, perhaps the ratio is only 10%.

With so much uncertainty, waiting in a lockdown until we know more is a prudent strategy.

"We have spent trillions (maybe 10s of trillions) on this debacle. The math never penciled out for such a response, even under the disaster scenarios."

At $9 million a life (the standard value used by EPA and DOT for environmental regulation and roadway design cost-benefit analyses), tens of trillions pencil out if you're talking about saving millions of lives. Now maybe you don't believe the claims that millions of lives are at stake based on policy response, but it seems quite plausible that they might be.

So you need a detailed analysis to argue that this "doesn't pencil out".

Perhaps the Icelanders are just healthier. The fatality rate seems closely tied to metabolic dysfunction such as pre-diabetes. 100 million Americans are pre-diabetic; I suspect the comparable number is far, far lower in Iceland. Maybe it's all that fish!

If the Italians are so unhealthy by comparison (orders of magnitude so, apparently!) why do they have the longest life expectancies, particularly so in North Italy.

No, "Hearty Nordic people can survive that which is the slayer of Italians" is not the answer.

(Not is "Italians touch each other to much, so the spread would be much higher", or other such dorkery.)

22% of Icelanders are obese, whereas 21% of Italians are, but it looks like in Italy obesity is skewed toward younger adults as they have moved away from the Mediterranean diet.

follow the tweet storm/model carefully
the fello/a made at least 5 big bold assumptions
and then wonders why they don't add up to 46,000 cases!
and this is even before considering
the sensitivity and specificity of the test
we say ask the Melania model or mebbe the South Korean
Dr. Woo Joo

The young South Korean journalist interviewing Dr. Woo Joo calmly
asks many pertinent rational questions eliciting beaucoup bucketfuls of good information and doesn't mischaracterize
his answers. So unlike the American "media"!

We have reached a point where we either have to accuse many countries of egregious lying, or assume the models should be "heterogeneous" as you have politely described it.

My personal bet is that the disease transmits mostly through spit droplets and R0 is dramatically affected by personal-space and mask-wearing conventions.

Let's not forget that our susceptibility/resistance is genetic, and Iceland does not have the same genetics as Vietnam. Let's also not forget that both culture and (spoken) language have effects. Is Icelandic a "spit it out" language? I wonder if anyone has measured the difference in aerosol production between different language (and culture) groups while speaking? Of course, randomness occurring in clumps isn't nor ever was a 'thing', right?

Do you think it’s easier to model the climate or a new virus?

A model is as good as the data you use to validate it

In terms of hot countries, the Philippines went from 251 confirmed cases to 848 over the past week. With the number of deaths standing at 68, we can imagine this 848 confirmed cases number is an underestimate.

The bottom line is that the Philippines now looks like where the U.S. did just three weeks ago, although its stricter containment measures may yet bend the curve. Thailand looks to be in a similar position but since it didn't react as quickly as the Philippines did, it might have higher growth rates and worse outcomes.

The evidence that a hot, humid climate plays a significant role in slowing the spread is looking less and less likely by the day. The simple, Occam's Razor model to explain these differences is that Southeast Asian countries simply got lucky and I believe took early measures to stop people from Hubei province from visiting. But once the virus spread to Australia, Korea, Taiwan, Singapore, Japan and Italy, it was just a matter of time before a critical mass of people from those countries visited and infected enough people to start a conflagration.

Also the median age in the Philippines is 24... Wikipedia happens to have a long list of patients from this country.

They didn't stop people from China from visiting. The first country to stop Chinese flights was Taiwan which started to look at inbound Chinese travelers on Dec 31 and I believe banned all Wuhan flights by Jan 23.
The first case in Wuhan was on November 17. The virus had free rein more or less for 66 days. The lockdown started on Jan 23.
Thailand has a huge number of Chinese tourists. Its first case should have been a long time ago.
There's a probably a temperature sensitivity, all coronaviruses have it including SARS-Cov1 but we will have to see.

Again, we are talking about probabilities. A single infected person is not necessarily going to start an epidemic in a country just as a single lit match or unattended campfire does not start a forest fire 100% of the time. You need the combination of recklessness and bad luck played out over time.

The Philippines banned people with a history of visiting China from entering the Philippines on Feb 2, shortly after it registered its first case. The overwhelming majority of the Chinese population was not infected and of those who were infected and chose to travel, it is quite possible a number of them did not infect any locals while on overseas trips.

Perhaps it's not warmth per se that is bad for the virus, so much as your nose runs more in the cold: e.g., see the big role that ski vacationers have played, whereas golf vacationers have played less of a role (so far).

@SS - but you have it backwards? I read somewhere that a runny news helps keep away the cold, though I may have it backwards. Timeout for quick Google search...no hits on page 1. The thesis then is mucus from a running nose is a large viral contagious load? Speculative but plausible.

If you have a runny nose you might brush it with the back of your hand, or you might blow your nose into a tissue and accidentally brush it with your fingers. So then having a runny nose is primarily a danger to yourself.

The droplets remain airborne longer in dry air. That's why humidity matters.

The droplets are less airborne, they also dry faster on surfaces as surface tension decreases with temperature, so only smaller droplets can stick on surfaces ( as droplets).
Mucus is also thicker when temperature rises and that's the first defense against entry into epithelial cells of the respiratory tract

*Mucus is also thicker when temperature and humidity rise

For the Wuhan figures, as per the Twitter thread the large number of urns ordered by mortuaries could just be for the regular non-Covid deaths if mortuaries were closed during the shutdown.

For the Wuhan figures, a conspiracy but not completely loony theory is that there's 21M fewer cell phone accounts in China, never happened before and cell phone accounts are both required and hardly ever canceled in China, says the Epoch Times and Bretb* site. 21M inactive (dead) accounts, by all accounts? CHN CCP spokesman Han Xia (what a name) on March 25 dismissed this by saying these are quarantined businesses cancelling their cell phones; unlikely, for various reasons not least of which like in Greece it's hard to cancel a phone (you don't just do it over the phone like in the USA, but have to show up and prove that you are the legal owner of the phone, I'm sure it's the same in CHN where the govt tracks citizens using their phone).

More info here: https://www.theepochtimes.com/the-closing-of-21-million-cell-phone-accounts-in-china-may-suggest-a-high-ccp-virus-death-toll_3281291.html

21 million deaths? You’d think someone would notice. My Chinese suppliers reopened after quarantine and got back to work, they didn’t send me a message asking me to wait while they recruit new workers to replace all the dead ones.

I am somewhat suspicious of The Epoch Times. Are there other sources or other reporting on the cellphone issue in China?

I read somewhere that the normal death rate would have been 22k considering the population size and the time frame of the crematorium closure. They ordered 45k urns. The official stats are 2800 covid death in Wuhan. Surely they want to keep an inventory of at least a few days on hand. Another factor is that the health system was overwhelmed. Many people who would have died later this year died during the crisis. This would indicate a multiple of 2 to 5x real death count vs official.

Or as TG Reaper might say, no reason to be skeptical of any models of the situation if they involve handling the dead.

'and which heterogeneities can be altered at the margin'

An effective public health system that invested in having the resources to handle a pandemic? Maybe there is an advantage to having a credible threat in the short term of being exposed to a crazed regime's weapons to being able to handle a much rarer, though also more certain, event.

Ceará state, in Northeast of Brazil, is a warm state. Temperatures rarely go lower than 23 degrees celcius. Yet is the third state with more cases in Brazil. 322 cases so far and four deaths. It is growing fast there. We had hope that high temperature in Northeast would save us, but it seems will not.

What about indoors? Yes, a place can be hot, but if people are indoors in air conditioning, doesn’t that negate all the positives? Air conditioning:

Humid? No
Hot? No
Negatively charged air? No

The prevalence of time indoors in groups with air conditioning is a factor I’d loom to first.

I don't know the specifics of how the disease is spreading. However, note that it's a poor state in Brazil. This means that most (like 95%) doesn't spend time in facilities with air conditioning. And it's quite humid.

The '10k quarantined' is probably doing a lot of work here. If you test the 3% of the population with symptoms and quarantine the close contacts of those who test positive, you've done a lot of containment.

I'm warming more and more to the idea that cultural differences are making a lot of difference to the r0, by the way -- both Italy and Iran have the kiss-on-the-cheek greeting. Japan is perhaps closer to the opposite end of the spectrum.

By the way, on the “everyone already has it” theory, a semi-random sample of 645 from Colorado showed zero positives.

I am going to have to call bullshit on this since the link is to tweet that doesn't appear to contain an actual source for the claim. Additionally, when I tried various Google searches to capture an actual source, none showed up in the search results. You got anything better than the tweet to link to?

There is a source, which seems serious, given in the second tweet:

Unfortunately the only thing the source says about how the sample was selected is that it was a group of "first responders and their family" of the San Miguel County.

This is not really a serious evidence against the "everyone has it" theory, because "everyone has it" is a strawman. I have suggested in this comment section, for instance, that as much as 100 million persons could have it -- not that I believe in this number (I really don't know), but that I didn't know what evidence I had against that. Of course, since it is a contagious illness, the 100 millions would be very heterogeneously placed geographically, with paces full of infected and other places (rural ones mostly) almost void. So this poll provides little evidence about the number of infected in the US. A serious poll, as I suggested in my email to Tyler last week, would be a real randomized sample chosen in a large but relatively homogeneous locality, such as New York City. This would give us information about the number of infected people in NYC, not in the US in general, but that would be enough to get, finally, a decent estimate of the mortality rate and hospitalization rate (since we know the number of deaths and hospitalized in NYC).

In San Miguel County, according to the same source, the number of deaths due to coronavirus is 0. So if we believe this poll, the mortality rate there is 0/0. We had an uncertainty of a factor 1000, now it is a factor infinity. Great progress !

For what is worth, and I am not going to give a source for this other than my own authority, I have a good friend and colleague who is a deputy (representative) in France and a good friend of Macron. I asked him about those polls, and he told me some have been done by the government (but not published) and showed very high number of infected people in France.
He didn't give me a figure though.

And that is meant on several levels.

However, as pointed out by a couple of Israeli tourists in NYC on 9/11, America does everything in a spectacular way. So in a couple of weeks, you might just understand why the French government has been acting as it has, because what is going on the Alsace is a catastrophe - and there, the numbers are not only not good, they are very, very bad. And they are growing worse, day by day. And if one is an American, the fact is that the Alsace is much closer to American style density than Paris or the NYC region should be deeply concerning.

One of the advantages of a truly free press is that as this pandemic spreads through NY, for example, people will be able to be informed of what happens as it happens. Though without the luxury of saying that Italy, Spain, France, the UK and so on are different and what happens there will not also happen here. And by then, many will be asking for more time to get prepared, but time cannot be bought for any amount of money.

“ One of the advantages of a truly free press ... people will be able to be informed”

Allow me to retort: fuck you.

If there is one hospital overwhelmed due to some random concentration of cases the media will descend upon the hospital like locusts exaggerating as much as possible while making maters worse through their presence. They will add insult to injury by then constantly implying, if not out right saying, that the entire system is overwhelmed.

They are a plague. We must fix the glitch.

@Reason - troll alert... raise your game; at least I am fairly factual and informative. BTW, I destroyed your last response. Go read my rebuttal, buttal... Bye, I won't be replying to you anymore.

In light of what is happening in Italy, Spain, and the Alsace, this is really, really amusing, in a black humor sort of way - "If there is one hospital overwhelmed due to some random concentration of cases the media will descend upon the hospital like locusts exaggerating as much as possible while making maters worse through their presence."

No journalist in those regions have any desire to visit any hospital at any time for the next several weeks, minimum. And the hospitals are all overwhelmed anyways.

Pay attention, and you too will notice just how reduced the reporting from hospitals become. If you have the chance, look at how Italian churches handle those who have left a local hospital in the last few days. The caption is "Army medical staff clears the coffins in the church of San Giuseppe in Seriate, Italy. Photograph: Carlo Cozzoli/REX/Shutterstock " and that should be good enough to see how well the Catholic Church and Italy handle something they do have some historical experience of.

Only partly true as a) journalists are collecting information by telephone and b) governments are releasing information which can be culled through for the most spectacular headlines. Here in Madrid, finding two dead people in their beds in a nuring home became headline news and, in the last English-language article I read those two had become "some".

The real point is nobody is bothering to report that the hospitals are completely overflowing, as that is inescapable public knowledge. At least for people living in Italy, Spain, and the Alsace. Even the reporting about the morgues being completely full has recede too, that also being public knowledge.

The journalists are manufacturing precisely zero of the current death tolls.

Nice retort for someone calling themself “Reason”. [says gesturing scare quotes in the air]

San Miguel County, Colorado is also one of the most physically isolated places in the United States. If I had to pick one county that I thought would be untouched by a pandemic, this would be in the top 50 (of 3000 counties). It only has one real town, Telluride.

I didn't connect Telluride with the original report of serological testing. Makes a lot of sense then as a ski resort they were particularly susceptible - remember the Vail export to Mexico. Looking up San Miguel county data, they only have one rRT-PCR-confirmed positive, although an Atlantic article I just read about this testing indicates their testing has been limited due to lack of swabs (sterile collection devices).

Looking into UBI further the company has had experience with SARS antibody testing, and has been exporting their SARS-CoV-2 ELISA test to Taiwan and Hubei. Glad to see that they submitted EUA (along with data they generated from samples from China).

Am awaiting the FDA to approve their first serological test, hopefully sooner than later, and then additional EUAs after that. Need to get a handle on the scope before policy gets ahead of facts on spread.

Crazy to think how different the rate of infection is within Iceland, given its genetic homogeneity. For sure DECODE (now owned by Wuxi in China) is matching their positive data with what they already know about the country's genetic structure.

Several people caught the virus at a ski resort in Switzerland.

The hotspot in Idaho is Sun Valley, not the Boise metro area.

Looking at the data on reported cases, it looks like there wasn't really an upward trend until March 9th. That could have just been lack of access to tests, but it's also possible they just got lucky and the infected they did get for the first week after the first confirmed case didn't spread it to many people.

The alternative is that it's much less deadly and much more contagious than they expected, and it made it over really quick and spread fast. But the problem with that theory has always been that you should have seen a massive spike in reported health cases with flu-like symptoms but a negative flu diagnosis.

There's nothing unusual about the Iceland data; many other countries show a similar pattern. If anything, the one surprise is that Iceland's caseload per capita is unusually high, second only to Luxembourg among countries with over 100,000 people. So the claim that Iceland would be expected to have 46,000 cases seems bizarre.

The Wuhan rumor should not be taken seriously. It is likely higher than official estimates, but not 20X.

Sweden's data is interesting. It might reflect the fact that the amount of social distancing that people do is relatively uncorrelated with government "restrictions". And that's important if true.

Vietnam is doing better than Thailand, perhaps because Vietnam is (culturally) East Asian, whereas Thailand is Southeast Asian. Warm weather may help both.

Iceland and Luxembourg should be compared to cities, not countries, if you are going to be comparing caseload per capita. In any case, Iceland's sub-1% fatality rate is worth noting compared to other affected regions where CFR often exceeds 4%. Obviously more testing of less severe cases is a factor, but that obvious factor may be obscuring other factors at play.

Why is 20x out of ballpark for Chinese lying about fatalities? China Daily says for the month of January, "The total number of flu deaths last year [2018] was 144, compared with 56 in 2016 and 41 in 2017." (https://www.chinadaily.com.cn/a/201902/22/WS5c6f11daa3106c65c34eaaf7.html) That bald faced lie is at least 20x off, why not for COVID19?

Vietnam's response has been excellent and largely overlooked. I was skeptical at first that they could handle it but the Government has both introduced clear measures early, been willing to take drastic action without Chinese style brutality, and been able to enforce. There has been a lot of community cooperation also.

Concerning the death toll being 20x higher: a response to that tweet points out that in a city of 11M a 2 month backlog of regular deaths is already ~20k, so 40k urns does not seem an usual amount to order.

That cuts two ways. Covid19 can also increase the death total from heart attacks, strokes, etc because the health system is overwhelmed. The death rate from Cov19 might be lower but it increases the overall death rate by demanding so many resources. China's numbers on direct Cov19 deaths could be accurate yet largely ignore secondary deaths.

Plus a lot of the spread in the disease seems to be in health care workers. A country like North Korea just leaves the sick to die without help and the spread slows. Perhaps Iceland and Sweden don't have health care workers rushing into burning buildings without proper safety measures.

In Ohio, I think 16% of cases are in health care workers. If that is mostly mild cases in 2-4 weeks they can return to fight. I pray. Taking measures to protect health care workers could bend the curve the most in some communities.

Plus what kind of nursing homes do Iceland and Sweden have? Old people in clusters do very bad and might inflate the early numbers.

Keep in mind that per capita numbers are not the most informative guide. If the epidemic starts the same day in two places, and spreads the same way, but one location is 100 times the size of the other. The per captia rates will 100X different.

Heterogeneity looks a lot bigger when you're in the middle of it.

If you had told people on February 27th "Your hospitals will be overflowing with coronavirus patients 4 weeks from now" or "Your hospitals will be overflowing with coronavirus patients 6 weeks from now", those would've sounded pretty similar to them.

Right now, they seem pretty different.

If Iceland's daily growth rate has been 20% rather than 30%, that would reduce the estimates by an order of magnitude: 4,600 cases rather than 46,000. If it's had a daily growth rate of 13%, that would fit an increase in cases from 23 to 800 over the course of a month (Feb 27 - Mar 27).

A 13% daily growth rate means that the number of cases doubles every 5.7 days. That's a pretty fast-moving crisis, which needs to dealt with quickly before it gets out of hand. It just doesn't look so fast when the next country over has a 20% or 30% daily growth rate.

At least according to this WaPo article.

"On Feb. 5, with fewer than a dozen confirmed novel coronavirus cases in the United States but tens of thousands around the globe, a shouting match broke out in the White House Situation Room between Health and Human Services Secretary Alex Azar and an Office of Management and Budget official, according to three people aware of the outburst.

Azar had asked OMB that morning for $2 billion to buy respirator masks and other supplies for a depleted federal stockpile of emergency medical equipment, according to individuals familiar with the request, who spoke on the condition of anonymity about internal discussions.

The previously unreported argument turned on the request and on the budget official’s accusation that Azar had improperly lobbied Capitol Hill for money for the repository, which Azar denied, the individuals said."

But remember loyal MR readers and commentors all the blame goes to the CDC and FDA, not the people actually running things at the top of the Trump Administration.

+1 . (Sarcasm on): But the Deep State (FDA, CDC) is at fault?! Trump is fighting for us! (Sarcasm off)

Of interest NY governor Cuomo, who fought the NYC mayor about quarantine and no doubt helped contribute to the spread of Covid-19, is praised as a libertarian hero, when just this morning he is reported to have threatened Trump against a quarantine of NY state (which is badly needed IMO). Some kind of hero...

Prior_approval, autistic slayer of strawmen.

I have read that this former GMU employee was fired 30 years ago for either sexual harassment or sexual assault.

In the spirit of the metoo movement, I think that should be noted every single time you identify this person. Personal courage is important - people who have personally experienced sexual harassment or sexual assault, possibly including yourself in such a traumatic situation, need to be supported unconditionally.

Iceland and Luxembourg should be compared to cities, not countries, if you are going to be comparing caseload per capita. In any case, Iceland's sub-1% fatality rate is worth noting compared to other affected regions where CFR often exceeds 4%. Obviously more testing of less severe cases is a factor, but that obvious factor may be obscuring other factors at play.

Why is 20x out of ballpark for Chinese lying about fatalities? China Daily says for the month of January, "The total number of flu deaths last year [2018] was 144, compared with 56 in 2016 and 41 in 2017." (https://www.chinadaily.com.cn/a/201902/22/WS5c6f11daa3106c65c34eaaf7.html) That bald faced lie is at least 20x off, why not for COVID19?

Extremely low population density cities. We kind of defeats the purpose of comparing to cities. If you want to compare cities, you need to compare Reykjavik or the city of Luxembourg.

The best samples we have all imply the IFR is probably well under 1%.

Diamond Princess data imply an age adjusted IFR under 0.2% given age distribution of U.S. Ioannidis has mentioned this.

Iceland, around same.

Vo, Italy tested all 3000 citizens the third week of February and found that 3% had C19. Repatriated Norwegians, same %. If you apply that level of prevalence to the entire region of Lombardy, say, and assume the established growth rate until mid-March, you'd need to adjust total cases by perhaps a factor of 100x, which would push the IFR under 1%. Still quite bad, but their own health minister has said they arent bothering to distinguish between "died with" and "died of" C19. If they have had breakouts in hospitals - an ER doctor in WA was fired by his hospital for saying they had poor C19 controls, so we know it probably happens "even here" - that will have inflated death counts, perhaps by a lot.

And further to that point, and importantly, excess mortality data to date from Europe do not show higher than avg deaths. Even Italy isnt showing exceptionally high excess mortality up to the end of last week. See here: https://www.euromomo.eu/outputs/zscore_country_total.html

There are unusual details, like the 40% mortality rate in the Kirkland nursing home, and the reports of 4x mortality in Bergamo, Lombardy. I do suspect heterogeneity in death rates that requre a special causal answer, like viral dose. But the best data we have suggest C19 isnt that deadly in most cases.

That doesnt mean it wont overrun some hospitals, or that it doesnt matter, or that we shouldnt be preparing for the worst. I have to add this. You know why.

“ excess mortality data to date from Europe do not show higher than avg deaths” hilarious if true. I looked at the graphs, they correspond with the basic intuition: about 10k people die per week in Italy. So far c19 has killed about that many in Italy over a few weeks. It’s a bad virus, but hardly the end of the world, or the death of “all old people” as the media has actually been claiming the last few days.

The shutdown of the economy OTOH is quite serious.

South Korea has been testing its citizens at a rate of 10,000 per day. There is still no evidence that they have undercounted cases by an order of magnitude. Their current case fatality rate is 1.59%. It has actually been steadily increasing which is exactly what you would expect if the disease is relatively deadly and the infection is past its peak. The results are emphatically not consistent with the theory that we have massively overestimated mortality or underestimated infections.

The glimmer of hope from South Korea, at least for people under 50, is that they have only two fatalities out of more than 2,000 cases in the 30-49 age bracket and continue to have 0 fatalities among under-30s.

South Korea has contained it well

But always adjust CFR for the age profile of the population. The age distribution of those testing positive skews older than the South Korean population. At a glance, over 11% of positive cases are in those over 70, almost twice the baseline percentage of S.Korean pop. This is a major issue since mortality increases by orders of magnitude in groups over 70..

If C19 were to become more prevalent - and it certainly is because % of positives among those under 20 is implausibly low - the IFR would fall by a lot, definitely well under 0.5%.

Fortunately, the Korean authorities are publishing data on a daily basis and already provide CFR by age bracket. The numbers are not consistent with IFR "well under 0.5%" unless you are talking about a very young population.

Again, they are getting 10,000 negative test back per day -- where are all of these hidden carriers of the disease and why is there no evidence of their existence to date?

80+ 17.51%
70-79 6.77%
60-69 1.74%
50-59 0.56%
40-49 0.08%
30-39 0.10%
0-29 0%

Wouldn't the continuous steam of ~100 daily new cases imply the persistence of hidden carriers in the population, given the tracking work done for each positive? By the time a small cluster is discovered, there could already have been viral clearing from some of the asymptomatic carriers in the cluster.

Sigh. If you test 10,000 people a day a get 100 positive new case, this suggest that the infection rate may be as high as 1% of the population, or may be jus 0.1%, who knows. But the official rate of infection *which is the one used to compute the death rates you gave us" is around 10,000 or 0.02% of the population. We do not know, but we may have to divide the death rates you gave us by 5, or maybe by 50, to get the true rates. In any case, these rates are just **an upper bound** of the real dest rates, in no case an (imprecise but unbiased) estimate.

we have heard that in South Korea they don't have a lot of nursing homes
and aged care facilities?

If anything, South Korea data supports the undercount argument, assuming their containment and mitigation work is as advertised. US and SK had first reported case at same time. Assume we had similar trajectory at beginning. On Feb 29, the peak of daily cases in SK, they reported 813 new cases, and 3000 total. US shows 8 new cases, and 68 total. If US also had 3000 cases on feb 29, we could have 4,000,000 today with 30% daily growth.

Whilst the Colorado data is bad for a narrowly defined “everyone has it” theory, it is consistent with a rapid endemic transmission theory where n% infection scales rapidly from 0% to say 20% once introduced. It is a bad outcome however for conventional wisdom, which accepts that transmission rates throughout the world all sit at 0%<n<0.5%. For this to have been supported, the infections needed to be single digits given the population size. It is an important but little discussed requirement for the true CFR to be 1%+. Jason y above’s views are entirely consistent with the data as we know it.

Jason y makes good points. I don't see what is contentious about them. The CFR data, such as it is, paints a very clear picture about what is happening: the risk to those under 60 is minimal (not zero, but minimal <1%) and it increases with age such that those over 80 have a more serious risk of death (15% to 20%), and that the risk is increased by certain pre-existing conditions. Furthermore, we can reasonably expect that the IFR will be lower than the CFR but by how much we don't know.

Is there an expectation that these CFR numbers will change as more cases develop?

...infectious disease spread and epidemics/pandemics ain't something new in human experience -- we have lots of information on this already.

No need to re-invent the wheel and ignorantly wonder about the patterns of disease spread among world populations.
Epidemiology is a well established science.

At least by economists, who seemingly feel that only they can determine what is a well established science.

And like any science it has all manner of limits.

In the present circumstance, it is reasonable to ask, if the rather blunt instruments of lock-down are necessary, being correctly applied, or even really effectively addressing the pandemic. (Depending on which numbers you see and where, things are looking quite grim for NYC and CA....)

It is much more reasonable to ask why back in January, if the rather blunt and expensive instruments of testing, quarantining/isolation and contact tracing would have allowed the U.S. to look more like South Korea or Germany today, as compared to what is actually happening.

To be honest, just forbidding spring break activities in Florida, based on Italian and Spanish experience, would have been of immense public health benefit, as will likely be demonstrated in the next four weeks.

I seriously doubt the US has the required state and civil society capacity that a real quarantine would require.

We’re about to run an experiment with the tri-state area and my guess is that we fail miserably.

is that the number of dead in the U.S: will be at least two orders of magnitude greater than the number of 500 dead oriignally predicted by Epstein hoover.org/research/coronavirus-isnt-pandemic

This was the sort of leading edge thinking that was circulating among a certain echelon of the Trump Administration who were apparently arguing for letting up on the then existing shut down measures - a week ago.

Yes, why didn't the Administration take action in January. What was going on in January? Oh, that's right: the impeachment trial. Gosh, maybe the endless Democrat tantrum has negative effects!

For a long time, I think we all assumed that anybody exposed to enough HIV would get it.

And then it was found that some fraction of the population of Europe, maybe 10%, have resistence to it. Various questions arise about where this comes from, but "accident of history" is a fine catchall.

Or consider
- suggesting that vulnerability to leprosy maybe genetically mediated.

Regarding covid-19 - the widely quoted statement "we have no immunity to it" is of course the correct place to start.

But maybe that's not true.

Maybe lots of people have partial or complete immunity (due to some historical accident) and in fact are immune (totally asymptomatic cases) or nearly immune (very mild disease.)

So antibody and perhaps even some kind of serum challenge test may be required.

(None of that makes any of the victims less dead, and a whole lot of work is required to explain the mid aged Seattle transit driver, or apparently, an infant in Illinois. There is nothing like 100% immunity among people under 70.)

There is clear evidence that in fact there is no "natural immunity" for SARS-CoV-2. This is not 100% established yet, because the research hasn't been published yet, but it growing increasingly clear. Several preprints are circulating showing that researchers have looked, repeatedly, and it's not there.

Yes, some people have few symptoms, but those people are infected, and the virus is replicating in their body, and they are contagious, and they don't have antibodies against SARS-CoV-2 until they are infected. Being resistant to the symptoms is very different from being immune. They people (who end up showing minimal symptoms) do not add to the potential herd immunity, for example, at least until they are actually infected, and recover.

One hospital in Brussels tests all incoming patients (broken leg, car accident, finger cut,....)
And they currently report some 8-10% infection rate. They do this by CT scan.

Because using a ratio of 1 confirmed case to 100 unknown cases, Italy currently has an infected rate of 15% today.

The 1 to 100 ratio is a total fantasy of course, but it is difficult to believe that Belgium currently has an infection rate as high as Italy's a week ago.

Yes, a source would be very useful.

Vo, Italy had a 3% prevalence rate a month ago. Norwegians brought home from Lombardy, same %. Italy did lock the country down 3 weeks ago, but 15% is absolutely within reach given best estimates of doubling time.

Vo Italy was pretty much the epicenter of the entire pandemic, and the area with the earliest and strictest lock down. Which gives something of a best case upper limit when looking at the effects of testing everyone while making sure basically everyone remained quarantined.

Do you honestly think that this will happen in the NY region? Vo was a test case, one that has fairly well proven that everyone has taken far too long to go to a total shutdown to avoid the (inevitable?) collapse of the health care system.

"There is a place in Italy where the coronavirus has already been defeated. Its name is Vo', a small settlement of 3,000 inhabitants in the countryside of the Veneto region, about 70km from Venice.

It suddenly became famous around the world on 21 February, after the news of Italy's first virus-related fatality: one of its residents, pensioner Adriano Trevisan.

After the ministry of health defined it as "a cluster of infection", Vo' was placed under quarantine on February 23.

The day after, one of the most iron-clad sanitary cordons in Italy's history was built around it: no one could enter or leave the town, and goods (only medicines and food) could only reach Vo' if authorised by the Prefetto (central government representative in Padua).

Alessio Turetta, 30, owner of a brewery in Vo' and town councillor said to EUobserver: "The pillars of our strategy to manage the health emergency were quarantine and testing".

The town implemented a similar model as South Korea, though with a less hi-tech approach, and on February 29, 97 percent of its population had been tested (Vo' residents living elsewhere were of course excluded)." euobserver.com/coronavirus/147848

My point was only that if you apply the prevalence in Vo in late Feb, roughly corroborated by the prevalence of C19 in Norwegians returning from Lombardy, and you use the doubling rates from estimated r0, even when account for the lockdown, a 15% attack rate to date in the region is not implausible. That would substantially lower estimates of IFR.

Vo shows what happens with a region experiencing community spread in mid-February going on total lock down by February 23. In that case, after having stopped the virus's further spread by that date, you end up with a total of 3% infected. Vo stopped all new infections basically by being quarantined from Feb. 23, something that has only been truly attempted in China, and nowhere else apart from Vo.

Tyler, it comes from the fact that there are no two humans alike. What's your next question?

I hope you are not upset about our diversity. Try again to cope with it.

The Iceland 30% continuing daily growth is just an assumption based on early numbers. All we can say is :
They administered 12615 tests, found 802 positives or 6.3%
if the test is truly random we can assume then that 22.9K icelanders are infected.
1/2 showed no symptoms. The data does not show if these were mostly the youngest.
they have 2 deaths , we can estimate the adjusted deaths using a doubling time of 6 days and assume it takes 3 weeks from infection to fatality, so perhaps a lag factor of 2^3.5 or 11.3.
This gives 2*11.3/12615 = 0.18% fatality rate in the general population.
The number of deaths is small today ( 2), so there are significant error bars in this simple estimate.
Their median age is 36.5 years, significantly younger than italy (47). They're generally healthy. This might explain the low death rate.
We know the death rate varies with
1- availability of competent health care in time ( Italy and Spain failed spectacularly)
2- age structure of the population
--We need more time to evaluate Sweden
--The urns in Wuhan don't mean much. People die every day regardless of coronaviruses. Perhaps they're just refilling inventory as supplies were not coming in. these kind of news are best ignored unless there's independent confirmation that something really is going on.
--In Westchester county the rate of hospitalization is 1% of the positives, not 1% of the population. 1 % of the positives is not out of line, It's actually higher than 10% for confirmed cases. Hospitalization is not ICU, see CDC report here on 2449 cases:
frequent random testing and serological tests will clarify things. And yes, we might see that the death rate is not homogeneous and has variations country to country.

I would add that yes, we should always be skeptical of models. They can be useful , they can be misleading. They appear more authoritative than they are especially to the media.
We've had 40+ years of climate change modeling and the models all run hotter than the actual data, some of them spectacularly so. They depend sensitively on feedbacks such as clouds and these are not well understood.
On the positive side it's a relief that with Covid-19 saturating the news , we're not daily bombarded anymore with apocalyptic climate change predictions

Hard to do this correctly on a cell phone in the middle of the night, the denominator above, should be 6.3% of 360k = 22,68K =true number infected in Iceland
This gives the fatality rate at ~ 0.1% of the population , incidentally the common flu rate. Thought provoking. Needs verification

I’m currently living in Vietnam (Ho Chi Minh city) and have some additional observations as to why the current case load is very low (less than 200 confirmed cases, zero fatalities to date) despite a population of over 90 million.
Firstly, the government took drastic measures early on, in January, to stop travel to China and Korea. It has also embarked on a widespread information campaign with at one point daily updates via SMS.
I don’t subscribe that much to the weather effect because in north (Hanoi, cold winters) the case load is only marginally higher than in the south (year round warm weather) and mostly due to foreign visitors.
What is of note though that Vietnam has an exceptionally young demographic with (according to 2017 data) less than 7% of the population over 65 years of age.
Life expectancy is only around 73 years which is 5 years lower than the reported age of the Italian fatalities.
Perhaps it’s the demographic profile that makes the country relatively resilient.

I live in HCMC too. I think it's the masks. As in Japan, everyone where's a mask when they get the sniffles. Some say this specific virus seems to be transmitted most efficiently by spit droplets. I got a lot of push back when I was skeptical of the utility of masks. People here swear by them, if only to reduce inhaling the exhaust from the motorbike in front of them. Only infected I know of are a couple of English teachers who caught it from a pilot at the Buddha Bar.

I am the CEO of the Healthcare Institute of the Westfjords of Iceland. We run all primary care in the region, two small hospitals and three nursing homes, albeit in a sparsely populated area. My PhD is in health economics.

I will not try to resolve the supposed conundrum, but let me just point a few things out that might help:
1. First, I‘d like to point out two resources, on which I base my later points:
a. https://www.covid.is/data has data updated daily, categorized by age, region and many other variables.
b. https://covid.hi.is/english/ has a model updated twice a week, with the projected caseload and impact on healthcare.
2. The testing has been done both asymptomatically and symptomatically. The former was done by deCode genetics (now a subsidiary of Amgen). The results from that exercise stunned everyone claiming that the virus had already spread; only 1% of tests came positive. Since then, restrictions have been in place on much of social life, although by no means a lockdown. Thus, the hypothesis that a big chunk of Icelanders are walking around with asymptomatic infection is highly unlikely.
3. The age distribution of cases has been most prominent in the ages between 40 and 49, partly because the initial infections came from people on skiing holidays in the Alps. This group has been quite resistant to the virus. Spread among older people has been extremely low, in part because nursing homes around the country have been in lockdown.
4. As you point out, Iceland is almost a city state with 2/3 of the population living in the capital area. In fact, the region where I lead the charge, no case has been confirmed (the two cases you can see in the data got infected abroad and have received care in Reykjavik). It can therefore be said that the spread has been halted significantly. Also, Iceland is taken together a homogenous well-off, well-educated, monolinguistic nation with a strong single-payer healthcare system and as of late robust state finances. Also, the populace trusts the elite, in this case the directorate of health and in general the people in charge.

Appreciate the local knowledge. However, at root we're talking about a disease with approx 850 confirmed cases, probably within a shadow infected population at least 3x larger. And substantial local transmission.

It is simply not very plausible for example that few elderly folk or even >60s have not been infected, and that "age luck" explains very low fatality and hospitalization rates.

(It doesn't seem plausible either to me that's it's lag... But to understand that will take real understand of introduction rates and time to death from infection from strain analysis, not the messy proxy of using cases that are identified and confirmed).

Actually, Germany has a similarly low proportion of known cases among the old, and also a low CFR: https://experience.arcgis.com/experience/478220a4c454480e823b17327b2bf1d4

Median age of German cases matches that of the country (as logically would be expected), median age of Italy's is about 20 years older than the median age of the country!

The question is not "Is Iceland / Germany's data younger than Italy / Spain?" , of course it is, but "Compared to Italy and Spain, is Iceland / Germany's data more representative of reality and the real IFR?" and matching the age distribution of the country seems to be a good indicator of a qualified yes to that latter question.

But why just talk about Iceland's age distribution of cases when we can *look* at it.

Population histogram of cases by age - https://twitter.com/alexandreafonso/status/1243557013759700997 (courtesy of Michele Tizzoni's twitter again).

The Icelandic histogram (reasonably broad testing) mirrors the population age distribution (https://www.populationpyramid.net/iceland/2017/), albeit with kids underrepresented...

... while the Dutch histogram overlain by the side, where only symptomatic suspected cases have been tested, does not.

Again, the issue here is not "Are Iceland's cases younger?", it is more like "Are Iceland's cases representative?" (E.g. are they consistent with expected community spread in a situation where the virus is randomly distributed among age classes? They are.).

(One other thing, it is funny to see people go "Oh, Iceland's successfully sheltered the elderly, and that has led to a low death rate", among those who rejected sheltering of the elderly as infeasible and impossible a few days earlier.)

Thank you for taking the time. Interesting.

Thank you Dr. Ólafsson, that was very helpful.

Plenty of people are asking exactly these questions.

There's heterogeneity in how many infected people from China went to different parts of the world.

Has that been studied? Or at least some proxy for it (e.g. People entering a country/state/area from China since December 2019?)

The virus may not like low temperature:

Stockholm: today's minimum -2 deg C
Reykjavik: 3 deg C, far colder on recent days
New York: minimum 8 deg C

Indeed. There are preprints of data analyzed regarding temp, humidity, and wind, and it indeed appears that a medium level of humidity, a fair amount of wind, and cold-but-bitter-cold, are best for spread.

Florida by County

Many counties have -0- deaths
Miami is outlier
Collier County population 380,000, heavy tourism and non-resident seasonal travel, RSW Airport nearby, median age 51, 11 nursing homes, 10 retirement communities, 83 cases, many travel related, -0- deaths, daily count decreased post March 25.

The Icelandic data is here: https://www.covid.is/data

There are two test centers. One (NUHI) tests people who have plausible symptoms and/or a known connection to an infected person. The other, deCODE genetics, tests anyone who shows up. The positive rate at deCODE is only around 1%. And this is still not a random sample since people are more likely to show up and ask for a test if they have some symptoms or some reasonable suspicion that they might be infected.

The fatality rate in Iceland is low because the age distribution has been extremely fortunate so far. And because most of the infections are relatively new and it takes weeks to die of this thing. There is nothing in the Icelandic data that is particularly inconsistent with data from elsewhere.

How dare you come here and bring actual data! Can't you just let people with no expertise speculate wildly and spread nonsense? Have you no manners?!

I am going out on a limb by extending a theory that I have heard about prevalence of auto-immune conditions in populations. Some have speculated that communities that have moved into cities hundreds of years ago tended to select for more active immune systems - living in a crowded medieval city was likely a filthier environment and a small town/countryside. Thus, a place like N Italy might have many people with these strongly active immune systems with respect to a place like Iceland (my wife is from Italy, and has an auto-immune condition. Thus, my interest). As Covid mortality is related to an over-zealous immune system, maybe there are real differences between ethnic groups based on their settlement histories. Again, very speculative.

10 deaths on the Diamond Princess now
CFR is now at 1.4

One of the things that is puzzling here in Spain is the heterogeneity in mortality rates. The rate has not stopped going up in all the regions (CCAA), but the spread currently is from 1.9% to 13.6% here in Madrid. The first rate I noted in Madrid, which I expected would fall, was 7%.

I know no obvious explanation for the spread. At first I thought this was a testing issue as only the symptomatic are tested, but I would have expected the rate to have declined rather than risen over the last week. The closest guess I can imagine at the moment is that the higher number of cases is overwhelming hospitals to the point that care is more seriously compromised. However, as hard pressed as healthcare workers are here, I have a hard time believing that we have been at that point for more than a week.

Almost forgot the link to cases by region: https://www.eldiario.es/sociedad/mapa-evolucion-coronavirus-expansion-Espana-28-marzo_0_1005099739.html#mapaccaa

Two things to consider - data is very noisy, esp. with smallish numbers early on, and so it's best to look at log graphs of data over time. Second, death rates trail infection rates, by a week or two, so you expect to see the new case rate peak well before the death rate peaks. A good example of this is the Diamond Princess - when the world stopped paying attention, 7 had died of about 700 infections. Over the subsequent weeks, 3 more died of C-19, bringing the total to 10.

It's interesting to see the international media now taking an interest in the Icelandic response. What the foreign media doesn't mention is that our authorities underestimated the virus every bit as much as other European nations. Even well into March our chief epidemiologist was saying that even with no state response at all the death toll in Iceland would only be 15 people or so (while an estimate based on the Imperial College reports would yield over 2000 deaths in Iceland with no action). The authorities here were saying that this was an annoying but reasonably harmless epidemic, on par with the swine flu back in 2009. The explicit policy was to slow the epidemic a little bit to make it easier to cope with but otherwise let it spread through the population and create herd immunity. The authorities felt there was no need to close the borders - indeed, you are to this very day welcome to come to Iceland as a tourist from any country in the world and you can enter with no quarantine or even a health check at the airport. That is, if you can find a plane to carry you, which you mostly can't anymore.

Social distancing measures were planned and then put on hold for days explicitly because implementing them "too early" might risk stopping the infection altogether and then we wouldn't get herd immunity, which would be a bummer. When measures were finally taken they were relatively mild compared to the European-style lockdowns. Indeed, Icelandic schools and kindergartens still haven't been shut.

And despite all this, Iceland is looking like it's had a more effective response than most European countries. Since Iceland is a bit of a biotech hub, the authorities found themselves sitting on a giant pile of testing capacity and figured they might as well use it. Those that test positive are put in isolation and their recent contacts are put in quarantine. This has been going on since the first detected cases in late February.

It's still not clear if Iceland has the epidemic under control but it's possible. If so, it's a testament to the importance of testing and contact-tracing. A strong response on that front can make up for a lot of slack everywhere else.

I need to correct myself a little bit here. The Schengen shutdown of external borders does affect Iceland, even though the Icelandic authorities fought this tooth and nail.

There have been 39 hospitalizations so far, including 11 ICU cases. Icelandic hospital data can be seen here: https://www.landspitali.is/default.aspx?pageid=b629a8e0-b262-49e0-b842-0f776cb4241e

Your last two paragraphs are very important. With lots of testing, great contact tracing, isolation of contacts, and quarantine of positives, (and you're a small island nation now virtually cut off from the world) you can ignore a lot of the measures being done elsewhere and have a handle on things.

- genetic - both in virus and humans
- diet
- initial load related to social customs

We need to stop treating data other than daily deaths like it means anything outside of a few on-the-ball countries.

A few Western countries, Germany and Iceland at least, seem to have stayed ahead of it with mass testing and contact tracing. Just like Korea. They'll get the lowest death rates because they won't overrun medical care.

For the rest it's just a matter of how early it started to spread and how fast. The speed likely has something to do with density and social customs. Google "Swedish bus stop meme". In places like Spain and Italy and the US (due to testing fiasco) the true number of cases dwarves the detected cases because it got out of hand.

swampr is right. Deaths are counted correctly in most countries. Cases depend on how many tests are done, who is tested, and the accuracy of the test.

Deaths lag infections by perhaps 4 weeks. Deaths also depend on whether ICU patients can get on a ventilator, with death rates higher if they cannot.

healthweather.us shows fevers in the U.S. nationwide and by county. We see that fevers fell by a factor of 3 after social distancing became widespread around March 15. Fevers are counted accurately enough and they will lead deaths. Of course fevers include flu, but I think some of the anomalous fevers after March 1 were Covid-19.

We need more random tests to get a current count of the disease. In Colorado, most tests are being conserved for hospital patients according to my doctor’s office. So most tests are done on people with terrible symptoms. There are no treatments so there is no action to take after a positive test. We need to use more tests to get real-time data not lagging data.

Do those questions matter? Healthcare workers probably like to believe that the main determinant of mortality is their own indidividual skill and grit. Do healthcare workers make a difference?

On topic of "heterogenity", a tweet highlighted by Lourenco, linking an arxiv preprint on the Lombardy Regional Task Force's time distribution of identified cases -

Seems pretty important because national figures which aggregate data don't show this local picture. Regional "heterogenity" within Italy and time heterogenity within region.

2x Notes:

- First symptomatic case in Lombardy (shows symptoms, tested positive) on 01/01/2020.

This of course implies that if this not the first true infection in the country, and more that the actual infection must have occured at least 2-14 days earlier (December 2019).

Assume infections grow from this introduction date on the German exponential curve, and on minimum infection lag, 10 millions cases would be expected today, while on the maximum, 80,000,000 (of course impossible)

This is a noisy approximation, as it is not certain at all that this will be the rate; and low level initial infections may die out by chance, etc.

But would say that even an earlier start date does not change too much if you look at the broader picture with infections in mid-January, if we assumed the first infection recorded was misleading. Still a range of 1.5 to 12 million cases would be expected from observed growth rate in confirmed cases and implied start date of identified confirmed cases with symptoms (and our estimated incubation time frame of 2-14 days, median 6).

I would add that people who want to claim there was is both a >1 month effect in death lag after cases are identified and a heightened level of deaths in early March, will then need an early-mid Jan date for first infections, or their argument may become internally inconsistent.

- The growth rate of curve of confirmed cases within Lombardy flattens at 26/02/2020, then seems to diminish thereon.

This of course coincides with a switch to testing only symptomatic subjects, but it is hard to see why that itself would lead to a plateau and decline of rate of increase in cases.

Sitting in London and comparing how things are playing out in the UK and in Iceland I can see clear differences in how people respond, how government responds, how companies respond, and in environmental factors.
1. Iceland has practically no public transport, which reduces risk of exposure.
2. Retirement homes and nursery homes in Iceland started excluding visitors many weeks ago. Many government offices and corporations closed their receptions pretty early, to limit traffic in and out.
3. Icelandic politicians have stayed out of the way of the Department of Civil Protection and Emergency Management, so the public message has been very clear, with nationally broadcast daily briefings. This department is trusted by over 90% of the population.
4. The government immediately passed a number of bills to expand benefits and make it financially possible for people to stay home and still pay the bills.
5. Offices and buildings tend to be better ventilated (and more spacious) than workplaces in e.g. the UK because heating costs a lot less. It might be a small factor, but being from Iceland and working in London I often feel like fresh air is a real premium product in a UK office.
6. The police have a unit that is dedicated to tracking your contacts if you have tested positive. They try to contact the individuals you might have passed the virus to, and those people are asked to self-isolate even if they have no symptoms.
7. As for the low number of deaths, my compatriots like to complain about the Icelandic health services but I can tell them that an elderly person getting a respiratory illness now in Iceland is going to get tested and treated a hell of a lot faster than someone in the UK. The NHS is impressive, but it often takes a lot of effort, patience and skill at dealing with bureaucracy to get treatment for things. This "threshold" could be causing late treatment of vulnerable people - I know for sure that I think twice before going to my GP with anything that doesn't look deadly.
8. Iceland is an "early adopter" when it comes to any tech, so solutions for working from home are more widespread than in many other European countries. Almost everyone under 70 is IT literate/capable.

That doesn't mean that the pandemic is "under control" (however you define that) or that the response has been flawless, but it looks like the curve is being flattened enough to let the hospitals cope with the serious cases.

I mean, the thing is that the growth rate and the growth in cases and the estimates of cases per capita, all do not seem low. So "They are flattening the curve of infection and that is why the death rate seems low", does not really work.

The number of new confirmed infections per day peaked on the 25th of March and has been going down for 5 days.
The rate of infections in the over 60's cohort looks low, and that might be part of the answer.
The official Icelandic dashboard is here https://www.covid.is/data

I don't think that's too strong to draw a trend over. The whole trend is more interesting (as much of a noisy way to look at reality as is). Infections over 60s looks (see tweet above), proportionate to their representation in populations as a whole.

What about the possibility that a low-probability 0.1%(?) super-spreader event causes an explosion in a given region, and the seemingly unaffected areas have just been lucky enough not to have had such an event?

This could be an asymptomatic, highly contagious public service worker, or someone with a massive social network, with lots of hugs, handshakes and kisses (obviously more likely in Italy than Iceland).

Balding seems to be assuming that the large number of tests were done randomly, while Eiríksson's comment describes very effective contact tracing, testing, and isolation. So maybe the main difference is Iceland has a competent government which reacted to the disease early with an effective implementation of traditional public health responses. Not coincidently, that seems to be what Singapore, Taiwan and the other "outlier" countries have done.


The Fresno County Department of Public Health announced 12 new cases of the coronavirus on Saturday, bringing the county total to 43 confirmed positive tests.

Health officials projected the number of COVID-19 cases will continue to rise as more lab work is processed and confirmed.

Cases up by a third in one day, definitely entering phase two.
LA is reporting 360 new cases and 6 deaths.


LOS ANGELES - Due to the limited availability of coronavirus testing kits, the city of Los Angeles is only offering testing to residents who are most at risk of contracting COVID-19, it was announced Monday.

At this time, the only individuals in LA that are eligible for testing include those with symptoms who are 65 and older; those with symptoms who have underlying chronic health conditions; and those who are subject to a mandatory 14 day quarantine period due to confirmed exposure to COVID-19.
The death rate in LA thus represents the death rate for symptomatic patients, and is holding at about 1.5%.

Calling a tweet that misinterprets an extremely speculative Chinese article about funeral urns as "serious talk" has really lowered my opinion of Tyler's willingness to actually vet his sources. Go to the tweet, then to the linked article. Their "investigation" revealed that a mortuary in Wuhan, upon reopening after being closed for some time, received shipments of 5000 urns over the first few days they were open. The writers assume that ALL of these urns are meant for COVID-19 deaths (What is the baseline number of bodies handled by the mortuary for this time period in previous years? The article doesn't ask this question.). They then write this: "Wuhan has seven other mortuaries. If they are all sticking to the same schedule, this adds up to more than 40,000 urns being distributed in the city over the next 10 days."

Talk about a slight of hand! One mortuary orders a bunch of urns after being closed for a while. Therefore all other mortuaries in the area have done the exact same thing, and all of these urns are for coronavirus victims. And this merits being passed along as "serious" talk that should make us rethink the actual fatality numbers?

As a general note, in a world where the virus has hit every country in a different way and with different timing, "heterogeneities" should be easy to find. What might be more instructive is to look middle of the bell curve - at the utterly predictable patterns we are seeing in the rate of spread and rate of hospitalizations across numerous countries. And in the way the rate of spread changes with interventions. Looking at western Europe, the biggest distinction you can make is that Italy got hit hard first, and that Germany is doing a stellar job of testing and so is finding many times the numbers of minimally symptomatic cases as elsewhere in Europe. Otherwise it's all been unfolding with amazing consistency.

Meanwhile, the shit has hit the fan in the US, in an utterly predictable way. Naval gazing on the early and imperfect data from a few outlier situations isn't going to make the reality on the ground in New York and New Jersey and Washington State and Louisiana go away.

I am wondering if Tyler is deliberately posting some questionable sources to see how his own readers react. A sort of secondary filter to adjust his own weighting of his commenters' opinions.

Seems an odd strategy to me. Some of these links and sources are incredibly flimsy. It's not like deep prior knowledge and lots of reading and research are required to rip them to shreds.

One explanation for heterogeneity in fatality rates:

Each year in the US alone, 100,000 patients die from infections caught in the hospital. The rate of hospital-acquired infections is very heterogenous across countries and regions. Italy, for example, has a much higher rate than Germany because of poorer standards in hospitals.

With the systems currently stretched, different rates in hospital-acquired infections matter even more. I am convinced that many patients that died with the Coronavirus may actually have died from other infections, but they are all classified as Coronavirus victims.

You make a good point, but you are wrong in saying that "they are all classified as Coronavirus victims." Different countries attribute cause of death differently. My understanding is that in both China and Germany deaths were not counted if there were major contributing factors apart from the viral infection. I know that was a big issue early on in China - death certificates only listed a single cause of death, so a patient with chronic congestive heart failure, who got the virus and crumped because of the pneumonia and died of heart failure would be listed as a death due to CHF, not C-19. I have read that Germany has a similar approach. And in many countries, there were clearly deaths happening to patients who hadn't been tested, and the only country I have seen reports of doing posthumous testing is Italy, which would inflate their numbers.

Again, this is not something I am able to find straight answers on. But there are lots of anecdotal reports coming from doctors and nurses that fatalities are not always being counted even in patients known to be infected.

Thank you for your response.

I believe your observations illustrate further why there is heterogeneity. Political reasons influence the reporting of causes of death. You make the point about China. On the other hand, I would suspect that a hospital in Italy would rather report a death as caused by C-19 than by a hospital-acquired infection.

Sometimes it's both. Realize that a massive number of the cases are hospital-acquired infections, with doctors/nurses/respiratory techs being the victims. In the Netherlands, before their testing capabilities were maxed out, they were randomly testing health care workers who were treating C-19 cases, and found that a significant percentage were infected. In Italy, dozens of doctors have died from the virus, having acquired it from contact with their patients.

When all his written about this, the failure to protect health care workers seems to be a source of the worst outcomes. What is the connection, if any, between protecting health care workers and death rates?

What measures would work best to protect them? Antiviral, blood transfusions, etc. Are we not aggressive enough with alternative treatments. Success in the Spanish flu was often a function of quality nursing, but they had the advantage that some age groups that could fill that role were to a significant degree immune.

Plus getting a high viral load might affect your outcome, given that the demographics for health care workers seem to indicate worse than expected outcomes based on demographics, is viral load exposure a key indicator of the outcome?

So much of the data is cloudy.

One obvious connection to death rates is that every health care professional who tests positive or develops symptoms needs to isolate themselves, degrading the care further. Suddenly there are 2-4x the usual number of critical care patients, and 5-10% of the workforce has to go isolate themselves, with some percentage of those adding to the critical care patient numbers.

Add in the incredible amount of stress that comes from working with inadequate equipment and supplies while facing the risk that you are making yourself vulnerable to a potentially fatal illness. That stress is on top of the usual stress of sleeplessness, overwork, and exhaustion. All of that degrades decision-making and leads to compromised care.

What measures would work best? That's obvious: PPE. Lots of it, of high quality. After that, lower patient loads, a chance to rest, to keep up with the accumulating literature. When you're a doctor or nurse and you see patients in distress sitting in chairs in the hallways and waiting rooms, you can't just walk away from it. And so you make compromises, to take chances, and you ultimately are likely to get sick.

Treatments? That's going along as fast as possible, and hopefully will continue to be refined and more readily available. Alternative treatments? No. Blood transfusions? If you mean convalescent plasma treatment, there's not much of that around, and it's going to the sickest patients in the few places it's available. Antivirals? I haven't seen any reports that the few antivirals that are generally used in an ambulatory care situation are of any utility. Vaccines? Yes, of course, and sooner the better.

No one has any idea if higher viral exposures make people sicker. So far it doesn't appear that health care professionals who get sick are having worse outcomes, so I think this is either unlikely or a minor factor.

Just read the tweet storm from Balding. In one of the first, he writes this: "I have not been able to find an Iceland specific R0 so let's use the daily 30% growth rate."

Is he serious? He's doing the math as if there is an inevitable 30%/day increase in cases, which has not been the case in any country that has initiated ramped up testing, isolation, quarantines, and public health measures. His assumptions about how many people "should" be infected at idiotic! By his logic, China should have 80% of the country infected by now, because in his world the virus spreads at 30%/day regardless of any intervention. Tyler, why are you continuing to give credence to such weak sources? Why do you have such high standards for evidence for things you understand, and such low standards for things you don't?

Better data, better models. Better models, a better sense of how and where to look for the best data. When models are applied to new and developing data then noise often mistakenly gets amplified.
Covid epidemiological models, in an abstract and general way, make salient a persistent difficulty in macroeconomics and in forecasting generally. Models are only as good as the data, and the data that is most interesting and consequential is the data that is new and noisy.

Most of the heterogeneity comes from different testing regimes. This is not hard to understand. Second order might be social structure, density, weather, etc.

It’s surprising that few people have talked about this from an applied math background and parameter instability. It’s an exponential growth function with a time constant in days. Estimating the true death rate is unstable: You just move the start of the curve backward or forward a few days or weeks and the results change considerably. Further just changing the duration of illness by a day drastically changes the case fatality rate — you are correcting for the number of fatalities in ratio to the number infected on the right date right?

So yes, of course the data looks inconsistent.

It's early days in Chile yet, but they lead South America in cases (probably due to testing) yet, with only 6 deaths so far, they are holding up among the best in the world. Chile I think also leads the world in type-O blood type.

Law of small numbers. Iceland is an outlier due to it being a small sample relative to the population of the world. Also it is an island.

Is Iceland necessarily really different from Italy? Already a week ago, Decode Genetics estimated that 1% of Icelanders had the virus. Now there are probably many more.

In Italy, the epidemic is at least 2 months old. If just 5% of Italians have been infected, this means 3 million infected. And 10,000 deaths for 3 million is the infection fatality rate of 0.3. Add to this that Italians probably exaggerate the death count because they categorize everyone who dies with Covid19 as dying from it, and Italy ceases to look all that different from Iceland.

Regarding San Miguel County and Telluride. Compare to Vo Italy. Vo Italy had the first recorded death in Italy from Covid on Feb 21 (I think) and found ~3% of the population infected around that time. They were a relative hotspot. San Miguel County reported their first *case* on March 20. One week ago.
Regarding ELISA test and Colorado –since the county just reported their first *case* a week ago, it may not be surprising that the first batch of Antibody tests found nothing. There may not be that much to find at the moment. Also, I would also have less confidence in the very first field deployment of new test for a new disease in a small sample size.
Regarding Iceland – the current death rate is not inconsistent with IFRs in the 0.1% - 0.5% range. Although they have done a lot of testing, Iceland is relatively small sample compared to the whole world, and I would argue that they are not representative of the dynamics in play in hotspots around the world. Density seems to play a role in spread severity.
Sweden numbers are going up. Deaths are starting to look parabolic. They may just be behind the curve. Not sure anything really *different* is happening there.
Perhaps a lot of this heterogeneity happens in any pandemic? And we are just watching it real time, instead of seeing it in large post hoc investigations? Also, maybe details on past outbreaks are going to be less robust? Even from 10 years ago? We just have so much more data now?

Finding the cause to 'heterogeneity', is like finding the root cause to poverty, you'll never find out the root cause to poverty.

Heterogeneity means that no one knows what's going on, and since this issue is important, what we often do when confronted with heterogeneity, just making something up and faking that we know, won't do.

Not sure what to do with this, but one of the ordinary boring pieces of hedge wisdom with breaking in a college kid who is doing their first real work at a real job is to get them to say 'I do not know' when 'I do not know' is the right answer. That's sometimes difficult since 'I do not know' is never the right answer in school.

Regarding Christopher Balding. There is *a lot* being assumed there.
Also, Iceland's own website says they had one confirmed case on Feb 27. He says 23. His opening premise is flawed -- by a factor of 23. starting with one death you get 2000 cases instead of 47k.

thinking a little bit more about heterogeneity. Considering the speed at which this apparently spreads and grows, many seemingly minor interventions could have drastic effects further down the curve timeline. So yes, let us discover and exploit all the heterogeneity we can! E.g., assuming an adequate mask supply, (if and when that ever happens), lets get everyone wearing them!

Anyone have a link to the Iceland data?


Glad I saved that link. Google surprisingly does not find it easily.

I will eat my hat with zero seasoning if this doesn't have anything to do with it:


Consider that in the age bracket of 25 to 29, which is when young people are very active in traveling and have money to do it, in the Nordic countries basically NOBODY lives with their parents. In the worst hit Spain and Italy, MOST people in that age bracket live with their parents. Germany falls somewhere in between.

I bet you could find a decent correlation between mortality as percent of cases diagnosed (older people disproportionately die and young people disproportionately remain undiagnosed) and millenials living with their parents. In fact, the 25-29 age bracket probably understates the difference, since it is far from abnormal for Italians and Spanish living with their parents 30-39, and their parents would be in their late 60s and early 70s, right at the steeply increasing part of the hockey stick mortality rate.

Anything else is probably explained by dumb chance (wrongly timed soccer match), policy delay and errors, some cultural differences (Swedes don't spend much time with their parents. Finns don't spend much time with anybody. The Frech kiss and Italians have family dinners).

If you want to have comparison between POLICY DIFFERENCES, look at Sweden and Finland. Basically open-border, liberal social democracies nearly identical in health care capacity, income and culture. Totally different approaches: Sweden has flock immunity strategy all but explicitly with minimal intervention, run by PM Stefan Löfven (62) and chief epidemiologist Anders Tegnell (63). They are talking about maybe closing skiing resorts in the future. Finland has a green-left government of mostly millenial women, who are getting criticism from the constitutional experts for cutting off a third of the population amongst other things. Nobody was surprised that they also mobilized the army, after closing the borders, schools and universities 2 weeks ago, banned gatherings over 10 people, and forced those over 70 stay inside. As of now, Sweden has 110 deaths. Finland 11. Finland has 1240 confirmed cases, Sweden 3700. Finland is rapidly increasing test capacity, currently per capita at over South Korea's and 3 to 4 times that of USA's. Sweden reportedly can't test health care staff in Stockholm anymore (severe lag in diagnosed cases).

Within the following weeks the results of these differences should be very clear and apparent. Finland made a major policy mistake of not informing returning travelers of 2 week quarantine in early March, which shows in the rising case and fatality numbers. This muddles the picture currently. We get the antibody tests within 2-3 weeks, so we see prevalence then. Until that just look at the fatalities and realize that Finland got its first confirmed case before Sweden. It is mitigation vs suppression.

On Sweden:

While it is too early to evaluate outcomes, you should keep in mind that Sweden is in practice doing what others are doing, its mostly a rhetorical difference and a few symbolic issues.

Travel has largely stopped, most people are following social distancing, many shops and establishments are closed and even many schools.

In Stockholm, movement is down 72% based on mobile data, less than some other european cities, but not far from the middle of the pack.


The foreign news stories exaggerate how different Sweden is behaving, they are mostly doing things a few days later than other countries and relying on social consensus rather than laws, informal rather than formal institutions.

There are important symbolic exceptions, including ski-resorts, but overall it’s a question of degree.

Quantitively, Sweden is closing down a bit less than other Northern European countries, so if Sweden end up doing does a bit worse in terms of deaths, there is no anomaly.

Those "rhetorical differences" and "symbolic issues" in Finland, Norway and Denmark are serious laws and policies which were implemented, because while most people follow recommendations, some do not out of confusion and yet there is a small group of people who simply don't take public security into any consideration. It's not the most who follow recommendations that laws are written for. It's the minority who don't or can't understand. Even a minority is well enough to keep the disease smoldering, ready to burst in flames after they try to lift restrictions. Small differences in policies, small delays etc, can mean a difference an order of magnitude in the final count.

Currently they are going to close the western border in Finland to even work related traffic because the prevalence in Sweden's side is so high that it's now spilling to Finland's side. This is nearly unheard of.

Also, any referring to Sweden's cultural special status vs other Nordics is a bit silly, because they are nearly identical. That's why I say that this is a very good comparison. I point out Finland in particular, because traditionally Finland is following Sweden in many issues. But now they've taken completely different approach.

The tweet storm asks about the heterogeneity between Vo and Iceland, vs. Wuhan and Italy. But the obvious inference to the best explanation is right in his own tweets: testing at large scale relative to pop allows islation and slows growth. This is zero support for the author's desire to shut fewer places down: the places needing shutting down are those incapable for now of testing at scale. The author derides "fear mongering" about mass deaths but the best explanation in answer to the question supports the conclusion that, unless we shutdown until more testing, there would be mass deaths. No one is saying that there will be. There will be if not serious, or if we give in to wishful thinking not supported by the evidence. Further, the author seems concerned about economic tradeoffs, but Tyler was right to say that not shutting down (in places with inadequate testing) will be more economically costly, not less.

Don't you "smart" people ever stop and think? You failed to predict two of the greatest economic calamities of the last 100 years. Yet you twitter on flailing in the dark. A little humility required. Truly a "dismal science".

Which best describes the relationship between COVID-19 and heterogeneity...
A. The infective organism is heterogeneous
B. The infected organisms are heterogeneous
C. The environmental context is heterogeneous
D. B and C only
E. A, B, and C

Surely momentum leans toward D, but has A been ruled out? Given the evidence, Occam's Razor would likely prefer a premise that Wuhan/Lombardy are circulating a more toxic variant of the virus.

The comparison between Finland and the brewing disaster in Sweden is mirrored with the comparison with my own province of Québec in Canada. We live in similar climate and despite speaking French we are descendants of the Viking invaders of Normandy and of similar genetic makeup so that leaves some factors constant. We are in total lockdown. Sweden has five times our casualties and a much steeper slope.

One other possible factor is blood type. Several researchers have reported that blood type of O is less likely to contract the disease and experience complications from it. Iceland has a higher percentage of O blood type than Italy - 48% to 40%. Other nations with elevated levels of infection include the US, Spain, and France.... all with lower levels of their populations with O blood type - 36%. Countries with relatively low levels of infection per capita include Peru, Bahrain, Egypt, Chile, Ireland.

This is some of the best discussion of the Iceland numbers and heterogeneity I have seen. TobiD, extremely good comment.

Here is some important data, I believe: https://healthweather.us/ . It tracks all "Influenza-Like-Illnesses" (ILI) against historical data, in the US. This now includes covid19.

You can see the ILI counts for each county as a percentage of total population. For virtually all counties, the incidence plummets when the lockdown goes into effect. I live in Santa Clara County, CA which is a great example. We are now far below (~20%) of average. I think our lockdown is working.

In the worst hit areas around NYC, the rate is still above the long term average, but even here the rates are now dropping.

Thanks Dennis. That link is very good.

My hope is that the places in US which implement strict measures early and which manage their infection load the best, will actually get through this with LOWER overall mortality, because less traffic, less accidents, and other contagious diseases also spread less due to social distancing, hand washing. USA especially has high traffic, accident and violence mortality. I may be a bit optimistic here. Need more data.

Seems to me that Christopher Balding's tweet is misleading.
He states they registered 23 positive tests on February 27 when gov data says they only got there by March 4th. If he'd projected 30% growth using this date,it would only go to 10,000 cases and not 46,000 by March 27th.

He also conflates results from private company deCODE, which targets mostly asymptomatic people + those w unclear symptoms while gov testing targets people based on contact tracing + clear symptoms. Both have conducted similar number of tests, but have very different results.
For deCODE: 0.8% are positive, whereas gov has 13% positive. One can be seen as somewhat closer to random, other half is very well targetted.

But for me the most interesting insight from Icelandic data comes from the day when the more random testing started on 13th Jan. By that day gov testing had uncovered 143 cases, while if we assumed deCode was nationally representative it would suggest there were 3,000 cases out there on same day. That is clearly exaggerated, but still the 20 fold difference is huge, goes way beyond missing the 50% asymptomatic. To me, the conclusion from this that identifying cases through track and trace only goes so far...
Data is on: https://www.covid.is/data

Comments for this post are closed