## What is the probability that someone you know will die from COVID-19 this year?

Do you realize that everyone you know someday will die
You realize that life goes fast
It’s hard to make the good things last
You realize the sun doesn’t go down
It’s just an illusion caused by the world spinning round

Chris Sims writes:

I teach a Math Methods class that covers basic probability, mathematical modeling, and data science. For one of my upcoming lectures I plan on applying the very simple tools that we’ve covered so far to COVID-19. As I’m sure you’re aware, there are very dire predictions for the total number of Americans that will die from COVID-19, ranging from 0.5–3M. However, I wanted to connect these impersonal numbers to a very basic question:

* What is the probability that someone you know will die from COVID-19?

You don’t need much to answer this question. You need an estimate of the infection fatality risk (IFR), an estimate of the proportion of the population that will become infected (infection risk, IR), and the number of people that a person knows on average. Coincidentally, in 2013 you published an article in the NYTimes on the latter. Your answer was 600 people.

So, for the remaining quantities. Recent estimates of the IFR range from 0.39 – 1.33%. Recent estimates of the IR range from 40–80%. With this information, the probability of at least one death out of n people, is simply 1 – dbinom(0, n, theta), with theta = (IFR) x (IR). Using “middle of the range” estimates for IFR and IR, we get theta = 0.00516.

This figure shows the probability that at least one person you know (meaning “you” in general) will die, as a function of the number of people you know. If you know 600 people, the probability of at least one death among people you know is greater than 95%:

This is morbid stuff. However, I felt that connecting the overall death toll to the individual level might help the general public understand the importance of taking the threat seriously, and the importance of complying with directives for social distancing.

I haven’t checked the above numbers, but if they’re correct, then I think Sims’s final estimated probability is an overestimate because it ignores correlation in the network. Observed properties of social networks show overdispersion, thus more zeroes and more high values than you’d expect from a purely random selection of people. But it’s a good teaching example in any case, as well as relating to general themes.

Also I think it makes sense to calibrate this by asking, What’s the probability that someone will die from cause X this year?, for different X’s. The survey we analyzed way back when asked respondents a bunch of questions, including how many people they knew who died from homicide, suicide, and auto accidents in the past year. I haven’t accessed this dataset in awhile, but I just pulled it into R and did some quick calculations, and it seems that 11% of respondents said they knew at least one homicide victim in the previous twelve months, 14% said they knew at least one person who committed suicide in the previous twelve months, and 25% said they knew at least one person who’d died in an auto accident in the previous twelve months. That survey was conducted in 1999, a year in which there were approximately 15,000 homicides, 30,000 suicides, and 40,000 people dying in auto accidents in the United States.

1. Dale Lehman says:

I haven’t confirmed the calculations either, but I do have a comment about the intent – to communicate this on a personally meaningful level. I think most people will interpret the result on the basis of the few people among the 600 that they know “well.” When I think of people I know, I generally think of a much smaller group – the 600 would include many people that I forget that I “know.” So, the 95% probability of knowing at least one person that dies will be interpreted as a 95% probability of knowing at least one person I know well.

The other purpose – as a mathematical modeling exercise – is a good one. In fact, my concern can also be addressed using the same example, since the curve does show the probability as a function of the number of people. So, it might be useful to have people estimate the number of people they know as a function of how well they know them.

• zbicyclist says:

A timely point. Someone I worked with some years ago on a local planning project died from COVID-19 earlier this week.

Because he used to head a city department, his death made the news. But my guess will be that even if my death had made the news, he probably wouldn’t remember he’d ever met me — he was running the project, and I was just one of the community members providing input.

2. somebody says:

I’m paywalled so I can’t read the NYT article, but this has made me realize that a pretty solid definition of “knowing” someone is if you notice when they die.

• David J. Littleboy says:

Yes. Really.

A note came through my college reunion mailing list (MIT ’76) that one of us had died from covid-19 (he had just had heart surgery). Not someone I knew, though. I was disconnected from my peers as an undergrad (lived off campus, worked in a lab (the MACSYMA group, for you history buffs)), but have gotten to know quite a few of us through the once-every-five-year reunion bashes that I’ve taken to attending. Now in our 60s, we’re prime fodder for the disease. There’s a good chance I’ll know the next one of us.

Here in Japan, I hang out with folks older than myself (70s through early 80s compared to my 60s), and have lost several friends (fellow musicians/go players) through the usual suspects. I noticed. And I’ll notice if any of them go via covid-19. Music and Go being two of the most efficient ways to spread covid-19, activities in these areas have mostly come to a complete halt.

• Daniel J Reaney Jr says:

Completely off topic, what instrument or instruments do you play?

3. Dave says:

Hmm… When I woke up and someone asked me to rank order the probabilities of where I would read a Flaming Lips’ quote today, I shouldn’t have listed this blog dead last.

4. Chris Sims says:

It’s a good point that the average numbers may not be representative for particular individuals. Right now the embedded assumption is that one’s social network is a representative sample of the age distribution of the population. I imagine that young people probably have social networks that include far fewer older people. One could (with more effort than I’m able to put into this) construct a plot with age on the x-axis, that varies the assumed composition of the social network.

5. Bob says:

The plot thickens, MCMC SIR model from Oxford. I don’t know if it’s better or worse, and I don’t know the impact of a contradictory model at this stage. Commentary below from the FT.

https://www.ft.com/content/5ff6469a-6dd8-11ea-89df-41bea055720b

https://www.dropbox.com/s/oxmu2rwsnhi9j9c/Draft-COVID-19-Model%20%2813%29.pdf?dl=0

• Anoneuoid says:

Our overall approach rests on the assumption that only a very small proportion of the
population is at risk of hospitalisable illness.

These modellers are really annoying. Why don’t they make a prediction we can check now that distinguishes their model from the other one?

• Bob says:

I think epidemic modelling is going to take a hit from this crisis it won’t recover from.

I can think of another domain based on modelling which will suffer a similar credibility problem.

• Anoneuoid says:

In my experience it already lacked credibility years ago in bio to the point where an equation derived from a few assumptions was treated the same as an arbitrary curve fit. People literally couldn’t comprehend the difference.

I don’t think this virus is it, but watching this has given me zero faith that if a real zombie apocalypse level virus hits they will be able inform and protect us.

Compare to astronomers telling us an asteroid is heading toward earth, those models I would trust.

• Bill says:

Yep, epidemic modelling is having it’s macroeconomics moment.

• Carlos Ungil says:

The model seems somewhat detached from reality. In the three scenarios they provide 60%, 65% or 80% of the Italian population would have been already exposed more than two weeks ago. How come that around half of the confirmed cases, hospitalizations and deaths have happened in Lombardia if it represents only 16% of the Italian population?

• Carlos Ungil says:

The secondary headline in that FT article “New epidemiological model shows vast majority of people suffer little or no illness” is completely unwarranted. As a commenter notes, “It’s important to note that the key conclusion people are drawing from the study is, in fact, an assumption that went into it. On top of page 2 of the study, it states ‘Our overall approach rests on the assumption that only a very small fraction of the population is at risk of hospitable illness.’ The authors are essentially making assumptions about one of the key unknowns about the illness.”

Another interesting comment ends with this “TLDR summary:
Overall, the paper uses an approach to modelling that is fundamentally flawed, rests on assumptions that are shaky and seem to contradict available evidence, and uses far too little data (15 data points!) to draw any meaningful conclusion. I did mathematical modelling for my masters degree, and I reckon I’d have failed if I had submitted this paper. The paper asks good questions about the lack of data but I would never dare to draw any conclusions from this type of reverse engineered model.”

• I am kind of shocked that there isn’t some group of 20 peoplé at the CDC who do epidemic modeling and cost benefit analysis for a living. That they can’t call up Joe and Susan to the podium and have them discuss what the models say is the best approach to balance costs and benefits givent he observed Dynamics of this virus is quite disturbing. WTF ?

• Dzhaughn says:

Well, funding is limited. At \$1M per position, they can only fund 12,000 positions on their current budget. And is controlling disease really the proper focus of the CDC?

• I’ve thought exactly this. It’s really amazing. I’ve often felt that a major part of being an adult is realizing that the people in charge of anything have no grand plan, but even I’m amazed by the lack of planning on the part of those whose job this is supposed to be. How could we *not* be modeling epidemics, health, and economics? Was it really more important to be prattling on about sodas?

• Right? I mean, for \$5M/yr I’ll hire 5 guys off this blog to work for my little LLC and we’ll save the world… Jeezus

• Bob says:

Because sodas and ‘non communicable diseases’ attracts the high tax technocrats, the nanny staters, the ‘nudgers’.

A few weeks ago as this blew up, Public Healh England were tweeting about the ‘hidden salt epidemic’.

That’s what gets the funding and builds the empire.

Even with that though, even if it’s not takem seriously enough, the fact that the field of epidemiology seems to be where everyone have have their own model, with their own parameters, and none of them seem to be applicable, is appalling.

• Thanatos Savehn says:

The CDC is more like an information clearinghouse than a disease battler/quantifier/predictor. One of the things it does really well is to act as a sentinel. The listeria / ice cream incident is a good example. CDC collects info on reportable diseases and looks for patterns. When one is spotted their tracing people get on the ball while their genomics people sequence the bug. They then ship the whole genome sequence and best guesses about sources back to the states. The states then go through their collections of e.g. Listeria monocytogenes (in Texas if a pathologist detects L. monocytogenes some of the bugs are sent off to Austin for safekeeping) and they try to trace the bug (which, like viruses, mutates regularly) back to the rogue Ur-bug. And that’s how they traced it not only back to the ice cream company but to the process line on which it was born and from there to people who had died or been sickened almost a decade before the wider outbreak was detected.

• There’s no doubt that those activities are a good thing for the CDC to be doing. I just don’t see how “plan responses to outbreaks of aggressive diseases” falls outside “Centers For Disease Control and Prevention”.

It my opinion this is one of those things where “hey those guys in the corner seem to be sitting on their hands all day and yakking, let’s fire them” probably happened like 10 years ago, and then … here we are. Of course, to do a good job of planning how to respond to hypothetical diseases, you need a bunch of guys to sit in a corner yakking a lot, as long as what they’re yakking about is ways to model diseases, possible plans for their control, timelines for vaccine production, optimizing time to successful vaccine at acceptable cost and loss of life, ways to fit simulation data to outbreak tracking data, surveillance of diseases in the general population… etc etc etc this is one of those cases where even if they never ever get called on to do anything, you need to have them over there in that corner.

• Thanatos Savehn says:

I fall into cynicism whenever I confront that question head on. Just as the courts hate drawing lines because it makes them look bad when they ignore the line whenever a party they want to win has facts that are on the wrong side of the line, I suspect our politicians aren’t too keen on anticipatory decision analysis because it might conflict with their constituents’ wishes.

I, of course, think war gaming before the war is clearly rational. Why are all the hospital beds already full? According to CDC’s MMWR from last week: “CDC estimates that so far this season there have been at least 38 million flu illnesses, 390,000 hospitalizations and 23,000 deaths from flu.” Also there were five pediatric deaths from the flu just last week and now H1N1 is back in business of all things: “Nationally, influenza A(H1N1)pdm09 viruses are now the most commonly reported influenza viruses this season.” https://www.cdc.gov/flu/weekly/index.htm Oh, and “Laboratory confirmed influenza-associated hospitalization rates … for children 0-4 years and adults 18-49 years are now the highest CDC has on record for these age groups, surpassing rates reported during the 2009 H1N1 pandemic.” So why wasn’t somebody asking the question “Wow, if this flu thingy is heating up how might it affect resources if another thingy comes along?”

• The Wuhan thing happened and was a known commodity in Jan, and clearly a big problem by the time they were in lockdowns around Feb 11…. Some time before that the CDC should have said “Hey, get Joe and Pam revved up to figure out what we should do if this disease starts spreading”… So that by say Mar 1 there should have been some kind of preliminary “here’s 4 or 5 scenarios we think are likely”…

That this wasn’t done is evidence that the federal government is far more sick than I thought it was. And I thought it was pretty sick.

• Anoneuoid says:

“Hey, get Joe and Pam revved up to figure out what we should do if this disease starts spreading”

This isn’t how “evidence-based” decision making works. You are speculating about if the disease starts spreading.

Once we see evidence for the spreading, then we can act on it:

“There is no reason for measures that unnecessarily interfere with intl. travel & trade.
We call on all countries to implement decisions that are evidence-based & consistent. WHO stands ready to provide advice to any country that is considering which measures to take”-@DrTedros

This has been the ridiculous way the virus has been handled since the beginning and it is 100% relying on “evidence based medicine”. They waited for evidence of human-to-human transmission, then they waited for evidence of nosocomial transmission, then they waited for evidence of international spreading, etc, etc.

It isn’t science. In science you have some model that you can generalize to new situations, that is apparently not allowed in “evidence-based medicine”.

• Carlos Ungil says:

Aren’t you still waiting for evidence that the virus is a nasty one and that people are not just dying at the usual rate?

• Anoneuoid says:

Aren’t you still waiting for evidence that the virus is a nasty one and that people are not just dying at the usual rate?

Why not provide data or interesting ideas instead of snarky comments and media reports of stuff like the incinerator is one town is overrrun, when usually people get buried not incinerated in that area?

Here is something interesting. I heard (but could not verify, maybe due to lack of italian search skills) that there were some TB outbreaks in northern Italy last year:

https://www.medrxiv.org/content/10.1101/2020.03.10.20033795v1?fbclid=IwAR1FcyRZD_jVx8E0MQy9LW9M2mvynKJXOonqCgv05puweFyuyZ8vdemcu4E

• Carlos Ungil says:

I take that to be a “yes”? It’s difficult to say for sure but your reply seems closer to 18th-of-March-Anoneuoid (“I dont think you did understand my point at all. The strongest form of my point is people are dying at the usual rate and also testing positive for this virus.”) than to 23rd-of-March-Anoneuoid (“That isn’t really what I proposed. Its that people with weakened immune systems are dying of respiratory failure instead of from whatever other thing it would have been a few weeks/months down the line.”).

Some more data from the city of Bergamo that I’m sure you will appreciate:

Second week of March: 296 deaths vs. 49 on average over the last ten years.

Third week of March: 313 deaths vs. 45 on average over the last ten years.

That includes all the people dying in the city of Bergamo, notably those from the rest of the province dying in the hospital. Looking at residents in the city of Bergamo only doesn’t change the picture significantly: 154 vs. 26 in the third week of March.

https://www.ecodibergamo.it/stories/bergamo-citta/a-bergamo-decessi-4-volte-oltre-la-medialeco-lancia-unindagine-nei-comuni_1346651_11/

(I tried to send this message before and it never appeared, maybe because there was a second link public.tableau.com/views/LandamentodeidecessinellacittdiBergamo/Dashboard1 which is just one striking chart included in the article above)

• Bob says:

Isn’t that point about the assumption true about all of these models though?

The comments under the FT articles are usually pretty good.

• Carlos Ungil says:

I agree about the quality of the comments, much better than the article in this case. One commenter asks: “In instances such as this, where its readership have comprehensively shredded the logical basis of an article, could not the FT explain, and either defend or repudiate, its reporting?”

Another one makes the point in a more amusing fashion: “Dear FT – Can I send you some of my non-peer-reviewed preprints? I have discovered, assuming only an incredibly low susceptibility to Ebola, that we are all Ebola survivors. Kindly inform your subscribers, the WHO and the world at large. Also, I’m always avaliable for a feature article.”

• Martha (Smith) says:

“hospitable illness” ? — Sound like an oxymoron. (I assume you meant “hospitalizable illness? — i.e., illness requiring hospitalization)

• Carlos Ungil says:

Good catch. I just copy/pasted a comment from the FT and didn’t noticed the error that the author had introduced when transcribing ‘hospitalisable’.

• Carlos Ungil says:

I’ve just noticed that a critical letter from other epidemiologists was published in the FT a few days later ( https://www.ft.com/content/ebab9fcc-6e8d-11ea-9bca-bf503995cd6f ) and the original article was amended “to clarify the fact that the modelling is controversial and its assumptions have been contested by other scientists.”

6. Navigator says:

@Anoneuoid

‘Compare to astronomers telling us an asteroid is heading toward earth, those models I would trust.’

We probably know more about viruses than we do about what’s flying around us, out there. Most info we get is obtained from non-funded amateurs. We can’t see towards the sun either, and they can come from any directions they please (I believe they recently launched a satellite to mitigate this issue though)

Actually, the odds of KNOWN asteroids hitting us are far from low. Don’t even think of the unobserved ones.
Astronomers deal with less noisy data and predicting something that follows the strict laws of Physics, is not a big deal, as long as you have all the data.

It’s much more difficult with viruses and other phenomena in noisier sciences.

• Bob says:

I don’t think anyone is denying viruses are harder than Newtonian mechanics. But my impression of their field is that it’s a shambles.

• Dzhaughn says:

Scientific progress is most easily made on questions thought to be of little importance.

• Anoneuoid says:

We probably know more about viruses than we do about what’s flying around us, out there.

I meant if they predict an object will be somewhere in the future (eg, intersect with the Earth’s orbit) generally it is reliable.

Astronomers deal with less noisy data and predicting something that follows the strict laws of Physics

I don’t think it is that different. You have Newtonian mechanics as a general guideline but you don’t know all the forces, parameters, etc for the object which is why you end up with perturbation theory. Just look how many simplifications and limitations a modern ephemeris has: https://ssd.jpl.nasa.gov/?horizons_doc#limitations

7. Brent Hutto says:

Unfortunately we Americans don’t live in a country where epidemiology is taken seriously as it apparently is in, for instance, Iceland. Without carrying out testing designed to provide data useful for modeling then it’s not about hiring good modelers. The best models imaginable are going to be nearly useless when based on data unsuitable to the task at hand, right?

We have better data for predicting congressional elections than we do for predicting something like a novel virus pandemic. Which given that political hacks (of all persuasions) are calling the shots and paying the bills, is not surprising.

8. Brent Hutto says:

King County, WA had 13,179 deaths in 2018 (most recent year I can easily find all-cause mortality stats). That works out to 36 people per day dying, averaged over the entire year.

King County, WA has had 100 deaths attributed to COVD-19 up through today. Total.

If you count days since Feb 19 it means over a 37-day period the average number of people who have died in King County was about 1,400 rather than the normally-expected 1,300.

To me that is a very serious occurrence but not a doomsday scenario by any stretch.

• Thanks for this, but I do think most of the deaths are yet to come, expect those totals to double easily in the next 2 weeks.

Still, it’s a lot better than maybe 20k or something. The uncertainties are so large here, that scenarios up to 3 or 4 million Americans dead can not be ruled out, even if the most likely scenarios are more like 10k or 30k or 100k or something.

One thing I can tell you is that the US has WAY more homeless people than Italy does. In absolute numbers it’s about 10x https://en.wikipedia.org/wiki/List_of_countries_by_homeless_population

I expect this to change the infection characteristics a lot in the US. We’ll see over the next few weeks.

• Brent Hutto says:

Once it is fully endemic, I have a hard time imagining COVD-19 killing any fewer people than Influenza. If there’s no vaccine forthcoming (or if the vaccine is of limited effectiveness as the Influenza ones are some years), COVD-19 will probably be even more deadly than Flu.

While an additional ~50,000 deaths per year would be tragic, we as a society certainly seem to that toll from Flu in stride. I know in my state of ~5,000,000 people the 100 or so Flu deaths each year are kind of lost in the noise of heart disease, cancer and accidental deaths. Most years Influenza causes roughly as many deaths as suicide, nationally.

Not to seem cold-hearted, in terms of years of life lost COVD-19 is (so far) much lower in impact than things like cancer and accidents.

• Carlos Ungil says:

> (so far)

That´s the key point. Covid-19 had much lower in impact than things like cancer and accidents elsewhere too, until that wasn’t true anymore. (And King county did take measures weeks ago, the situation would be worse otherwise.)

• Brent Hutto says:

The four leading causes of death in Italy (two heart disease categories, one stroke category, one cancer category) together cause about 700 of the country’s 1,700 deaths per day in a typical year. COVD-19 is killing ~1,000 people per day at the moment so it’s mortality is slightly higher than the next four causes combined.

COVD-19 seems to me likely (but not certain) to be the leading cause of death in Italy for calendar year 2020. But that would require it to end up killing roughly 10x as many people as it has so far, over the next nine months. Which seems entirely possible.

All just back-of-envelope engineer type thinking…

• Anoneuoid says:

COVD-19 is killing ~1,000 people per day at the moment so it’s mortality is slightly higher than the next four causes combined.

Not if most of the people dying are those who were about to die of heart disease in the next couple weeks/months anyway, which is exactly what they have been telling us is happening…

• Brent Hutto says:

William Osler described the reality of pneumonia a century ago. He wrote that it was often considered the natural end of an old person. Ventilators and antibiotics let us for the most part live in a different reality now (at least we lucky few in the developed world) but there is something qualitatively different about a life tragically cut short (by say a 30-year-old woman dying of cancer) versus a person who would likely die from almost any opportunistic infection happening to succumb to COVD-19 today rather than living another month or year.

Especially considering the non-quality of life I’ve observed among relatives and friends who have lingered for a while at that point of no return.

• Carlos Ungil says:

“In #Bergamo, from 1 to 24 March, there were 446 deaths of residents: 348 more than the average in recent years (98).
There were 136 deaths officially due to #Covid19 in the period. There are 212 more.”

• Anoneuoid says:

“In #Bergamo, from 1 to 24 March, there were 446 deaths of residents: 348 more than the average in recent years (98).

Thanks, is this data anywhere to look at besides the average? Also, I looked up who that guy was and this fits with the Italian “hug a tourist” super-spreader explanation:

As the mayor of Bergamo, Gori raised controversy on February 11, 2020 by tweeting that he dined in a Chinese restaurant in solidarity with citizens who were attacked by “alarmists”, and adding that there was “really nothing to fear”

https://en.wikipedia.org/wiki/Giorgio_Gori

NYC had that smiling coronavirus robot spreading the disease too… But across the US, outpatient visits are actually down 30% last week: https://www.cdc.gov/flu/weekly/weeklyarchives2019-2020/data/senAllregt12.html

• Anoneuoid says:

Anyone used to survival via begging are not going to be in a good position right now.

• Brent Hutto says:

“Anyone used to survival via begging are not going to be in a good position right now.”

And once we well and truly crash the world’s economy, the number of people surviving by begging these next couple decades is going through the roof. Potentially including at least a few of we (currently) middle-class first-world lucky ones reading this blog.

9. Martha (Smith) says:

Here is a new study (presented today to the Austin City Council before peer review) of COVID-19 projections for the Austin-Round Rock area: https://utexas.app.box.com/s/7kgikqmev8ex831aq9wcsb3n5kwr7nxh

10. Martha (Smith) says:

Brent Hutto said,
“William Osler described the reality of pneumonia a century ago. He wrote that it was often considered the natural end of an old person. Ventilators and antibiotics let us for the most part live in a different reality now (at least we lucky few in the developed world) but there is something qualitatively different about a life tragically cut short (by say a 30-year-old woman dying of cancer) versus a person who would likely die from almost any opportunistic infection happening to succumb to COVD-19 today rather than living another month or year.

Especially considering the non-quality of life I’ve observed among relatives and friends who have lingered for a while at that point of no return.”

This describes two extremes. There is a lot of territory in between them. What about, say, someone 70 years old who is in good health, and would likely live another 15 or 20 healthy years if COVD-19 had not come along?

• Brent Hutto says:

Martha, you’re describing quite a few of those closest to me. I’m not wishing this virus on anyone, simply pointing out that as a public health problem it is properly assessed in terms of both how many it kills and who it kills.

But I’m having a hard time finding any aspect of the response to COVD-19 that reflects the stuff I was taught those decades ago in my Intro To Epidemiology classes. Everybody want to wring their hands over the numerators without even a passing thought about denominators.

• A part that people aren’t talking about is also morbidity. It may not kill all the 30 year olds, but if they wind up in the ICU on a ventilator for 25 days and a bill for \$700,000 I think that’s a major problem which is often being glossed over in the discussions. Obviously death is worse, but the morbidity is significant here.

• Curious says:

Simply to provide another aspect not being discussed:

We are currently choosing to save a proportion of the currently ill and probable ill at the expense of the future suicides that will inevitably result from widespread business, home, and career loss.

And before the comments begin — I am as Democratic as they come. This is not a right wing theory as has been described. It is our future reality.

I am not claiming that this decision is wrong, but we should acknowledge the trade-off we are making and choose to do everything we can to prevent that outcome.

• Andrew says:

Curious:

Are you sure about the future suicides? See here and here. The evidence seems mixed.

• Curious says:

Yes. I am sure about the suicides. What creates the reduction of deaths during economic downturns is a reduction in accidents — both automotive and work related.

• Andrew says:

Curious:

Thanks for clarifying. I also found this news article by Linda Qiu with more on the topic.

• Dale Lehman says:

All of a sudden, the numbers don’t matter? I am sure there will be more suicides because of our current actions – but how many? Also, given how much less driving is going on, there will be less highway deaths. But, if people are exercising less, their health will deteriorate and there will be more deaths from other causes. But, wait, their eating habits have probably changed – are they eating more or less healthy now?

It is virtually impossible to get a complete picture. And, our understanding of how many lives will be saved as a result of our severe mitigation measures is similarly highly uncertain. If there was ever a time to have uncertainty and evidence play key roles in our decisions, it is now. And, given the difficulty with the evidence, decision-making should probably look quite different than what we are used to. I’d venture that we need some clear ethical guidelines, consideration of irreversible effects, and as much epidemiological evidence as we can gather in order to make decisions.

• Brent Hutto says:

Dale,

Years from now when the history of this debacle is written the part about “…as much epidemiological evidence as we can gather in order to make decisions.” is going to feature heavily. The very first impulse of public health authorities, when the outbreak was in China well before the first cases showed up here, should have been to provision resources for testing, screening, contact tracing and all the boots on the ground hard word that needs to happen early on.

Now we’re at the end of March and still can’t manage ANY population-based testing. We’re still just doing tests for the sickest people and making assumptions about everything else. The number of “don’t know” aspects of the disease overwhelms any attempt at principled decision making. All we’re left with is political maneuvering to determine which faction’s kneejerk reactions will take precedence.

• Andrew’s second link (an essay in Nature) is a very good one. (I linked it as well in a post I wrote, along with a few other links.) As Andrew and Curious both note, death rates drop during recessions, but this was largely driven by a reduction in driving and workplace accidents. Now, drug use and suicide, both worse under weakening economies, are larger contributors to the overall death rate than in the past. Moreover, quoting from the Nature article, there are “clear negative consequences for individuals facing financial hardships, from stress-induced chronic diseases to mental-health problems.” The whole thing is, of course, extremely complicated. I continue, however, to be struck by (1) how completely we lack either the will or the ability to model any of this, even with large error bars, and (2) how devoid the innumerable discussions of the pandemic are of any sort of broad philosophical or ethical framework. #1 I touch slightly on in my post, though the post in general is not very good; #2 is still a vaguely formed thought.

• Curious says:

Andrew, Dale Lehman, Brent Hutto, Raghuveer Parthasarathy:

There is a fundamental issue at play in our current approach to this crisis. A trade-off between our constitutional rights, which many would argue are fundamental to our economic success, and the rights of the state to subvert those under certain circumstances. The founders intentionally sided with the rights of the individual relative to the rights of the state to intrude, with some obvious exceptions such as war. This crisis is being treated as if it were a war. As such we have decided that we are willing to err on the side of the state relative to the rights of the individual, in order to save lives — certainly thousands and potentially tens or hundreds of thousands of lives.

With this decision we have gone down an economic road at the local and small business level that will likely create a churn that we’ve never seen, certainly not in my lifetime. Landlords owe mortgage, restaurants owe landlords, restaurant owners owe mortgages on their homes and have to feed themselves and their families in the mean time. When we turn the faucet back on — how long will the current landlords continue their generosity relative to what is owed them? Because as soon as that generosity gets shuts down, whether it be a week later or 2 months later, the cascade begins. This is a massive cost to individuals, their families, and their employees that we have decided is a cost they must bear for the benefit of the state and the benefit of those who will not be infected and will not die as a result of these actions.

This may be the right decision for society, but it is important to note that we are taking an approach that is decidedly at odds with our individual approach to governance enshrined in our constitution. If there is a way to mitigate this cascade of business, career, family, and individual losses — we must do it. Will the legislation just passed save these businesses, their owners, and employees from economic ruin caused by our governmental decisions? I don’t know the answer to that. But I sure as hell hope it does.

• Zhou Fang says:

What constitutional rights are violated by effective quarantine?

Broad legal powers with a lot of precedent exist:

• Zhou Fang says:

If western governments do not take effective action to prevent mass deaths using “constitutional rights” as an excuse, this will not defend constitutional rights. In the eyes of those who survive, it would invalidate those rights.

This crisis is a *bigger threat* than war.

• Curious says:

Zhou Fang:

The United States Constitution provides for the right to “Freedom of movement”.

https://en.wikipedia.org/wiki/Freedom_of_movement_under_United_States_law

• Rahul says:

Why skeptical?

Not just suicides, I foresee starvation and riot related casualties too.

I think people are not seeing this with a long enough time horizon.

• Zhou Fang says:

It’s fundamental to *civilisation itself* that people living in it have an expectation of being cared for when they become old and infirm, not be tossed aside once they are no longer productive. Choosing to let a bunch of people die is also not a cost free action.

• Rahul says:

Isn’t that a false dichotomy?

There’s deaths on either side of the descision.

We only need to weigh one vs the other.

• Carlos Ungil says:

How many people do you expect to die of starvation and commit suicide (a voluntary act, by the way)? Experts forecast covid-19 deaths to be in the hundredths of thousands.

• Zhou Fang says:

The “do-nothing” covid19 deaths would be in the millions.

• Curious says:

Carlos:

Suicide is “voluntary” only in a very narrow sense. Certainly not the sense in which someone steeped in causal reasoning and modeling would consider “voluntary”.

• Zhou: the COVID deaths worldwide from what we’ve done already could be in the millions. In 1918 about 50-100M died worldwide. This would almost certainly be worse than that in the do-nothing case. I would argue a true “do nothing” like if we were all robot drones going about our daily lives unconcerned… would be in the several hundred million to a billion.

• Curious says:

Daniel Lakeland:

I am convinced you are among the smartest people who comment on this blog and so if you have to come to that estimate you are more than likely correct. I am trying to understand this and have a couple of questions if you have the time to answer them:

1. How are you getting to that 3 mil number?
2. What is the number given current restrictions?
3. Are there softer restrictions that result in something close to all out quarantine?

• Curious, thanks for your kind words. I will freely admit to having large logarithmic error bars on my estimates…

My estimate of ~ 3M cases today comes from combining several lines of evidence. The first thing we can say is that as of today it’s basically 100k confirmed cases in the US according to Johns Hopkins, so this is the *absolute minimum* number of infections.

Next we know that across most of the countries similar to the US it’s been deaths growing at rate doubling every 2 to 4 days:

deaths don’t have the testing problem as much, so I think it’s a reasonable estimate that the virus spreads in general population of cities etc at about doubling every 3 days say… which means 10x every 10 days.

Next, we have the fact that it seems 5 days is median time to onset of symptoms, and symptoms don’t get bad enough to go seek medical treatment until 7 or 10 days for many. So whatever is confirmed, it’s both an underestimate because they’ve been withholding tests, and an underestimate because it reflects new infections from say 7 to 10 days ago…

So, round numbers we expect the confirmed case number to be lagging the actual infections by ~ 10 days worth of growth… Which takes our 100k today and makes it 1M or so right away.

There are additional issues regarding that some people will have mild symptoms and not seek treatment, etc. Put together it seems 3M is a decent number at the moment but obviously if it’s still growing exponentially even after the isolation measures, it’s hard to know because what rate should we use?

Now, CA implemented safer at home 8 days ago… So some time between 8 days ago and today it should have slowed quite a bit, but it’s not like it will just turn off…

It’s very hard for me to figure out how well the isolation stuff is working, because it hasn’t been in place long enough. So I really don’t know but if you look at that site above it seems places getting a handle on things are seeing something like doubling every 10 days. Which is good because probably the duration of the severe illness is around 10 days, so they can kind of reach steady state in hospital utilization.

If ~ 10% of people get a critical care case, we’d see 300k critical care patients, and that’s well above what we can care for, so we might see 150 or 200k deaths just built-in to our current state.

I hope I’m wrong about the deaths and the critical care numbers. It’s hard to figure out what fraction of infected need critical care. It could be only 2% for example. No one is doing sufficient widespread surveillance to figure out the *real* prevalence in the population in US, Italy, Spain, UK etc.

11. Brent Hutto says:

Carlos,

I can “forecast” any number I care to make about the economic fallout. Just like the “expert” forecasts of deaths, there is so little data underlying those predictions that the error bars are a couple orders of magnitude wide.

If we stay in the current state of partial shutdown for a month or so, then gradually ease it off, we’re probably looking at a few million unemployed and something like a short-term 10-20% contraction in the GDP.

If we stay in the current state until May then crank it down indefinitely to a more draconian lockdown when (go figure) the virus has not magically disappeared then we’re looking at something way worse than the Great Depression.

Where it falls in within that range of scenarios depends on a bunch of things for which we have zero information. Pure guesswork.

The range of coronavirus death forecasts are similarly dependent on what guesses you make about several parameters. What did the U-of-Washington epi people just publish? Anywhere from 30,000 to 160,000 this spring and summer? Yeah, that about covers the waterfront.

The difference in speculations about economic fallout and speculations about COVD-19 deaths is that we acting as though the disease fallout worst case scenarios are true and totally ignoring the economic fallout. Because 10,000 people dying of pneumonia seems far, far, far worse than 1,000,000 people unemployed.

• Carlos Ungil says:

Brent, my question was about to Rahul regarding his remarks about ”starvation and riot related casualties” and weighing the “deaths on any side of the decision”. Do you expect those deaths to be also in the hundredths of thousands? Comparing deaths to unemployment is not comparing deaths to deaths, is it?

• Andrew says:

Brent:

I had the idea that we would have regular universal testing, and then only the people who test positive, or who live in a household with someone who tests positive, would be required to stay at home. This is still lots of disruption but nothing compared to locking down the whole country.

• Carlos Ungil says:

Andrew, even if test capacity in the US was 1mn/week (it’s still well below that) it would take years to test the whole population. It’s too late for tracking and tracing and focused testing as they do in South Korea, at least in some regions.

• Brent Hutto says:

But the same people who failed to make provision for testing and tracking are the “experts” telling us to believe their models based on supposition and conjecture. They could not have more comprehensively dropped the ball on this.

To get back to the stuff this blog is about, it’s an interesting question (to me at least) which is more useful a modeling approach in a fluid situation like this…

On one extreme, you do the kind of mechanistic Bayesian models that mean making (informed) suppositions about many physical parameters. Even though in this case prior information is terribly lacking. I think those techniques are pretty powerful and through many iterations of modeling, checking, revising and adding new data they can be insightful even if in the begining they start from mostly guesswork.

Or you do the kind of “field level” modeling that is basically curve fitting. You look at the early shape of the developing trend, maybe match your parameters to some external data from areas where it’s been building longer but basically just keep refining a trend extrapolation as each day’s new counts arrive.

The latter approach isn’t sophisticated but it might be the case that the former approach takes a long time and many iterations to yield better predictions than the basic curve fits. Or maybe not, I am not able to judge that sort of thing myself.

• Carlos Ungil says:

Do you think that the epidemiology and public health professors and researchers behind these models are the ones who dropped the ball? They have been calling for testing and preparation for what was coming for quite some time!

To give just a couple of examples, this was published seven weeks ago: science.sciencemag.org/content/367/6478/610.full

Will novel virus go pandemic or be contained?

[…] Based on what they have seen so far, many researchers think it’s probably too late to contain the virus. “As the virus continues to spread in China, the risk of exportation to other countries grows and sooner or later we will see it spread in another country,” Aavitsland says. So far there has been no sustained transmission outside of China, but Lipsitch expects that to change: “I would be really shocked if in 2 or 3 weeks there wasn’t ongoing transmission with hundreds of cases in several countries on several continents.” […]

and this was published four weeks ago: http://www.propublica.org/article/cdc-coronavirus-covid-19-test

Key Missteps at the CDC Have Set Back Its Ability to Detect the Potential Spread of Coronavirus

The CDC designed a flawed test for COVID-19, then took weeks to figure out a fix so state and local labs could use it. New York still doesn’t trust the test’s accuracy.

[…] So far, the United States has had only 15 confirmed cases, a dozen of them travel-related, according to the CDC. An additional 45 confirmed cases involve people returning to the U.S. having gotten sick abroad. But many public health experts and officials believe that without wider testing the true number of infected Americans remains hidden.

“The basic tenet of public health is to know the situation so you can deal with it appropriately,” said Marc Lipsitch, professor of epidemiology at the Harvard T. H. Chan School of Public Health. He noted that Guangdong, a province in China, conducted surveillance testing of 300,000 people in fever clinics to find about 420 positive cases. Overall, Guangdong has more than 1,000 confirmed cases. “If you don’t look, you won’t find cases,” he said. […]

• Brent Hutto says:

Andrew,

Honestly, as recently as a couple weeks ago when I first heard that this was coming I thought exactly what you’re describing. My training is in public health but I’ve never worked in the infectious disease side so maybe I’m just naive. But I’m truly shocked that nobody is even TALKING about that kind of classic public health approach.

It’s like no preparations whatsoever were made during the months when we should have known it was coming, then we skipped over the rational measures and straight to the panic. We still have no real information on which to assess how many people have been exposed, the relationship between exposure and infection, the temporal relationship between infection and shedding virus, all the basic parameters that determine the course of the disease through the population.

• Carlos Ungil says:

> What did the U-of-Washington epi people just publish? Anywhere from 30,000 to 160,000 this spring and summer? Yeah, that about covers the waterfront.

By the way, note that that’s assuming the social distancing measures are maintained over the next four months (i.e. late July) and “this number could be substantially higher if excess demand for health system resources is not addressed and if social distancing policies are not vigorously implemented and enforced across all states.”

• Brent Hutto says:

Exactly.

Whatever measures are taken, it isn’t the “two week” at a time B.S. that local and state and national governments keep repeating. There won’t be any less COVD-19 in circulation at any foreseeable future time than there was two weeks ago. It’s working its way toward endemic steady state. So if schools and universities are 100% online today, then they aren’t going to be back in classrooms any time soon. If 20% of the economic activity in the country is shut down by fiat today, that activity isn’t being resumed when the weather warmed up.

Which is why some of us are so hung up on the economic consequences. If every retailer, restaurant, hair salon, health club closes for a month some of them will never re-open. If they all close for six months (or a year or until a vaccine arrives or any other vague date in the far future) then none of them are coming back.

I don’t see any of lasting just until May or June or July. And if I’m guessing correctly about it being far longer than that, the economic dislocation is going to be horrific.

• Anoneuoid says:

A cross-sectional study with short4term follow-up was conducted in the 700-bed Amphia Hospital
(Breda, the Netherlands) and the 800-bed Elisabeth4TweeSteden Hospital (Tilburg, the Netherlands),
employing 9,705 HCWs, 17.7% of whom are male. Between March 7, 2020 and March 12, 2020,
HCWs in both teaching hospitals who suffered from fever or mild respiratory symptoms in the last 10
days were voluntary tested for SARS-CoV42 infection, in accordance with the local infection control
policy during outbreaks.

[…]

A total of 1,353 HCWs were screened, 86 (6.4%) of whom were infected with SARS4CoV42.

[…]

Two weeks after the first Dutch patient with COVID419 was reported, the prevalence of COVID419 in
HCWs with fever or respiratory symptoms in two Dutch hospitals in the southern part of the
Netherlands was 6.4%. This unexpected high prevalence supported the hypothesis of hidden
community spread of SARS-CoV42 and is considered a minimal estimate of the prevalence in all
HCWs at the time of screening. Only HCWs with (recent) symptoms were screened, and
oropharyngeal swabs were used for testing, which may have a slightly lower sensitivity than a
nasopharyngeal swab (Wang). Another possible explanation for the unexpectedly high prevalence
would be hospital4acquisition. However, all patients with fever or respiratory symptoms in both
hospitals were routinely tested for SARS-CoV42. At that time, a limited number of infected patients
was nursed under strict isolation precautions, and only three SARS-CoV-2-infected HCWs mentioned
exposure to an inpatient known with COVID19. There was no clustering of infected HCWs in specific
departments.

https://www.medrxiv.org/content/10.1101/2020.03.23.20041913v1

So two weeks after the first reported case 1353/9705 (14%) of healthcare workers in the population had symptoms and 86/9705 (0.9%) tested positive.

Thats about 10% of the population qualifies as a suspected case and 10% of them (1% of the population as a whole) testing positive in a given week. This is pretty similar to what was seen elsewhere.

In Iceland they saw about 1% of the general population test positive: https://nordiclifescience.org/covid-19-first-results-of-the-voluntary-screening-in-iceland/,

In the US 10-20% of the people tested are positive: https://covidtracking.com/us-daily/

• Carlos Ungil says:

What does “widespread” mean, 1% prevalence? That leaves substantial room on the upside.

By the way, (under-)reported covid-19 deaths in Lombardy are on track to end the month well above 7000 (5400 as of today, up from 3800 four days ago). Compared to all-cause monthly deaths in the region being in the 7400-8800 range in 2003/2017 that seems quite remarkable.

• Brent Hutto says:

The situation in Lombardy is a tragedy, no doubt. I can’t imagine what the public in general but especially their health care workers much be going through day after day.

To put the number in perspective, the deaths in that state attributed to COVD-19 will soon reach 0.1% of the population with about 10x that number diagnosed as cases. During the peak of the outbreak it far outstrips any other cause of death but still not quite as much of a doomsday scenario as it may seem from the media coverage.

• Carlos Ungil says:

> but still not quite as much of a doomsday scenario as it may seem from the media coverage

Where do you put the line between a doomsday scenario and an acceptable outcome?

I don’t know what media are you talking about but I doubt they forecast say fifty million deaths in the US. But single-digit millions would probably die and many more would be severely ill if nothing is done. Not that much in a country with 330mn people, quite a lot compared to less than three million deaths per year.

On the other hand, this simple intervention (not doing anything) would improve noticeably the (non-age-adjusted) mortality rates in the coming years!

• Anoneuoid says:

What does “widespread” mean, 1% prevalence? That leaves substantial room on the upside.

Within two weeks of the first reported patient 1% of healthcare workers were already infected (if you trust the tests)…

Keep in mind these are not antibody tests, they should only reflect active infection. So, eg if on average people are positive for 2 weeks and it has been around for 2 months (8 weeks) then (assuming a constant 1% positive) 4% of the pop would already have been infected.

• healthcare workers are of course at much higher risk, and are often used as sentinels in surveillance because they get things early.

• Anoneuoid says:

Yea, but it matched with the results from general population in Iceland.

Also, 14% had fever/cough symptoms at the time of the study and it said most were still going to work. Is that normal?

• Healthcare workers are receiving a LOT of pressure to be available, in an all-hands-on-deck way at the moment. Lots of them who aren’t normally involved in critical/ER care are thinking twice and/or quitting. If you’re a nurse who does allergy tests all day or something, and they suddenly want you to risk your life to get out of the allergy clinic and come help with rolling beds around the halls with hacking coughing patients who have a deadly disease, using no personal protection because they ran out… you’re probably thinking twice about that.

• Anoneuoid says:

SUre, but 14% of the employees (not even the patients) have a cough and fever at the hospital? That does not sound like somewhere you want to go.

12. Brent Hutto says:

Carlos,

Is anyone seriously proposing that “nothing be done”? Even our dysfunctional public health system is capable of the basics of isolating people known to be contagious, treating the most seriously ill and so forth. Of course until a couple weeks ago I’d have thought them capable of screening for dangerous viruses known to be spreading rapidly worldwide and they blew that one.

As I keep pointing out, like the Flu has done for a century this coronavirus and its descendants is going to kill a *lot* of people in absolute terms over the next years and (as far as we know) decades. There’s nothing we can do to stop that.

The key is to not crash the entire world’s economy for a long time in a futile effort to keep something inevitable from happening. Slowing it down is great. But the measures we take to slow it down have a cost and I think we have already surpassed any rational tradeoff between our ability to slow down the inevitable and the damage being done by those measures.

• Carlos Ungil says:

> Even our dysfunctional public health system is capable of the basics of isolating people known to be contagious

Apparently not. What makes you think you won’t see Lombardy’s tragedy replaying elsewhere?

They stopped commercial activity on March 11, when there were around 500 death, and all non-essential activity on March 21 (schools and universities had been already been closed for weeks). Despite that the number of deaths has increased by more than an order of magnitude and we don’t know yet how far will it go.

The population of NYC is similar (15% less) and have reported 366 deaths by now. Should they follow Lombardy on their lockdown policy for maybe a similar level of tragedy or should they keep business activity as usual even though the outcome is likely to be much more tragic? Both are valid policy choices, mind you. But I’m not sure what are you proposing precisely.

• Brent, what exactly are you proposing?

As I said above, in a true “do nothing” scenario, you’d see 500M or 1B people die by christmas. Cost of that happening is easily 70 Trillion.

What is needed now is what South Korea is doing, which is to keep the growth rate down so that d/t ~ 1 where d is the duration of severe illness (time from onset of medical care to either death or recovery sufficient to exit hospital) and t is the doubling time of the infection.

If d/t < ~ 1 then hospitals can stay at a moderate capacity in more or less steady state while we buy time to figure out how to reduce severity with drugs, and/or find a vaccine.

To get there, we need to first remain closed until the current pulse of infections works through the medical system, so at a minimum probably 4 more weeks, and during that time frame dramatically more logical and effective surveillance and testing needs to come online.

That is LITERALLY our BEST option. What will actually happen, who knows, but I have *zero* confidence in federal capability to do literally *anything* good at the moment.

all the good stuff will happen at the state levels, and will vary widely from state to state.

• Anoneuoid says:

So far (up to March 21) the excess mortality is undetectable at a national level in Europe:

Over the past few days, the EuroMOMO hub has received many questions about the weekly all-cause mortality data and the possible contribution of any COVID-19 related mortality. Some wonder why no increased mortality is observed in the reported mortality figures for the COVID-19 affected countries.

The answer is that increased mortality that may occur primarily at subnational level or within smaller focal areas, and/or concentrated within smaller age groups, may not be detectable at the national level, even more so not in the pooled analysis at European level, given the large total population denominator. Furthermore, there is always a few weeks of delay in death registration and reporting. Hence, the EuroMOMO mortality figures for the most recent weeks must be interpreted with some caution.

Therefore, although increased mortality may not be immediately observable in the EuroMOMO figures, this does not mean that increased mortality does not occur in some areas or in some age groups, including mortality related to COVID-19.

https://www.euromomo.eu/index.html

13. Anonymous says:

I will no get COVID-19 AT ALL

14. theresa chiles says:

so what is the math here. If I don’t know one person that has died in my circle of 30 (friends and family) what would be the Math if I can find 800 people that are saying the same ie: they don’t know one person in their circle of 30 ????