[remus]

This caught my eye in today's CDC update email:

PANAMA IMPLEMENTS GENDER-SPECIFIC SOCIAL DISTANCING In an effort to further enforce nationwide social distancing measures, Panama recently announced that it is implementing gender-specific rules for when people can leave their homes. Women will be allowed to be outside on Mondays, Wednesdays, and Fridays, and men will be allowed on Tuesdays, Thursdays, and Saturdays. On Sundays, everyone must remain indoors. The restrictions will reportedly last for at least 15 days. Officials implemented the additional gender-specific restrictions after observing high rates of noncompliance with the previous policies. These rules will be enforced based on the sex specified on individuals’ national identification cards. Additionally, the announcement stated that Panama has also been invited to participate in the SOLIDARITY Trial, led by the WHO, which aims to assess COVID-19 treatment options.

I know men are at greater risk of severe symptoms than women but how is this measure useful, are they suggesting people are still going out on the pull 

Perhaps it's easier to police than other options? i.e. if the policy was people were only allowed out 3 days a week how would you keep track?

Not suggesting it's a good idea of course!

[mrjonathanr]

I was uneasy with the discussion tone on here at times, and so want to comment - and then I'll sign off for a while.

I  agree that it is important to see as big a picture as possible, and economic impacts are not abstractions (or shouldn't be), but the wreckage of people's livelihoods. Nevertheless the discussion of trade off between health and the economy has seemed a little light at times.

I wonder if there's a perception that the health impacts of CV19 will be largely inflicted on other people; that the climbers here are unlikely to feel their full force. I wouldn't be too sure about that. This virus is novel, and data incomplete.

Only a small number of people die, but that won't console if it happens to both parents in short order.
Few healthy middle aged people die, but a relative or family member might be one of the few.
Or a friend with diabetes, asthma or other condition.
Women seem to be spared more than men, but that's no consolation if the mother of your child, in an overwhelmed hospital, doesn't make it.
Some people won't get their cancer diagnosed or treated till very late. That may be us, our parents, spouses, children.
Some people will get hurt on the road and contract CV19 in hospital in poor condition to fight it off.
Some kids will get ill, and recover, but with permanent lung damage.
As might you, climbing becoming a thing you used to do, before the breathing difficulties.

In short, when pondering the 'tough choices' that come in balancing economic versus personal pain, please consider that the sacrifices being discussed may be made by us.

[Murph]

Didn't mean to offend.

[mrjonathanr]

You haven't offended Murph, my apologies if I came across that way. I am just saying that this is very real and the impacts may be felt by us directly.

The whole thing is so extraordinary I think it is hard to retain a feel for what is really happening.

[Offwidth]

(What Stu said - far too early to make sensible comparisons)

What is a bit infuriating is the lack of normalising deaths by population in the graphs (e.g. we should be plotting deaths per million).

I’ve seen several comparisons of curves after the first ten deaths or first 100 deaths... that’s just bad data manipulation. Comparing Belgium (or our) curve to the USA after 100 deaths is the same as adding or subtracting a week (or more) from the horizontal axis.

Grr.

Worldometer have deaths per million now. As I said before Netherland's rate is currently double ours... Sweden is currently at 3/4 the UK rate despite their first death being almost a week later.

https://www.worldometers.info/coronavirus/

Apologies for the double thread posting but I did think this was a big positive.

[tomtom]

Its been on there for a few days (sneakily added!) - since then we've gone from 8-9 to 42

The site Galpinos linked to is cool - allowing me to plot deaths/mill over time... We'll see but Germany is following the curve - but a couple of weeks behind us....

[Stu Littlefair]

Warning: mathematical and dusty stats discussion will follow this announcement.

Offwidth, I think your preference for deaths per million is not correct.

Consider two countries with exactly the same disease dynamics (Ro, IFR etc), and an outbreak starting on the same day, with the same number of cases. In this case, the deaths per capita, D/N, will follow exactly the same track, but the raw fatality numbers, D, will not.

I think this is why you are arguing for per-capita plots.

However, if infections started at different times in the two countries, plotting versus population will not give comparable curve.  In this case, you need to apply a shift in the x-axis (time), which is what the "days-since-N-deaths" graphs attempt to do. A "days-since-N-deaths" will also correct for outbreaks with different initial numbers of cases: for example the difference between a few infected people arriving on a plane, and a football stadium full of infected people returning from a match in Bergamo.

Now, here's the epidemiology/maths bit. During the exponential growth phase of the disease, the rate of increase of deaths (the slope) does not depend upon the size of the population. It depends upon the number of cases. Later on, the total population size comes in to play, but not during the initial exponential rise.

Therefore, at the start of the infection, all countries with identical transmission dynamics will lie on the same line in a logarithmic graph of "days since X deaths". Here's the proof. In the graph below I use a simple SEIR model to simulate three cases. The "UK" has 60 million people and I run two simulations, starting from 100 cases and 10 cases. "NL" has 6 million people and starts from 10 cases. All other transmission parameters are the same, and the outbreak started on the same date in all sims. Note how the days_since_100 chart lines all the curves up during the exponential growth phase.



Take home message: population only matters in the later stages of an infection, when large numbers are immune. Using a "days_since_X_deaths" plot is the correct thing to do.

Caveat: "days_since_X_deaths" plots are much, much, much better when X is a large number. Even without any randomness, days_since_10_deaths will not line up outbreaks that started with different numbers of cases. If you can find someone plotting raw deaths on a "days_since_100_deaths" plot, that's the one to compare for now.

[tomtom]

Wheres NL on your second chart Stu? Without that I'm not following you...

[Stu Littlefair]

On the second chart, NL is right underneath the UK/N=100 case; they follow exactly the same path on a per capita graph since they are designed to have the same level of infection per person...

If you want to see what a "days_since_100_deaths" graph looks like, I've uploaded one to my work website. Please don't share this widely, it's a mild abuse of my IT privileges.

http://slittlefair.staff.shef.ac.uk/covid/

You can click on the legend to add/remove countries. Shift-clicking allows you to select more than one.

What to look for:

In the exponential phase, all the countries should line up on roughly the same straight line. When you reduce the number of people available for an infectious person to infect, the curves should flatten off. This can occur because the disease has run it's course (infected so many people there are few susceptible people left), or because social distancing means that infected people are successfully isolated.

Countries with larger populations WILL level off at larger numbers of deaths for the same degree of social distancing, which I guess is Offwidth's point?

[Stu Littlefair]

edit: deleted. Somehow managed to quote myself in my entirety. Not even I'm that narcissistic.

[Davo]

I was uneasy with the discussion tone on here at times, and so want to comment - and then I'll sign off for a while.

I  agree that it is important to see as big a picture as possible, and economic impacts are not abstractions (or shouldn't be), but the wreckage of people's livelihoods. Nevertheless the discussion of trade off between health and the economy has seemed a little light at times.

I wonder if there's a perception that the health impacts of CV19 will be largely inflicted on other people; that the climbers here are unlikely to feel their full force. I wouldn't be too sure about that. This virus is novel, and data incomplete.

Only a small number of people die, but that won't console if it happens to both parents in short order.
Few healthy middle aged people die, but a relative or family member might be one of the few.
Or a friend with diabetes, asthma or other condition.
Women seem to be spared more than men, but that's no consolation if the mother of your child, in an overwhelmed hospital, doesn't make it.
Some people won't get their cancer diagnosed or treated till very late. That may be us, our parents, spouses, children.
Some people will get hurt on the road and contract CV19 in hospital in poor condition to fight it off.
Some kids will get ill, and recover, but with permanent lung damage.
As might you, climbing becoming a thing you used to do, before the breathing difficulties.

In short, when pondering the 'tough choices' that come in balancing economic versus personal pain, please consider that the sacrifices being discussed may be made by us.

I think I would be one of the people whose comments make you uncomfortable. Reading your post I find it hard to disagree with anything you have written to be honest. I think anyone looking at the issues you have raised would agree that all of them are unpleasant and involve suffering and heartbreak. I certainly don’t wish any of those things on anyone and would be deeply upset if they hit me personally. To give you some information my mother is in care home, prone to chest infections and I would guess highly likely to die from any viral infection which I think she will eventually contract despite the lockdown. My own personal feelings on that don’t stop me considering whether overall what we are doing is for the best

I would also say that the choices we are making now are not just saving lives they are costing lives and causing harm. Whether or not we save more and cause less harm from what we are doing seems like a valid discussion topic. I also appreciate that it is emotive which is why I have included a personal detail so that you can see that I am not heartless or lacking empathy but  am just curious as to whether or not what we are doing is the “least worst “ alternative.

Dave

[Offwidth]

Warning: mathematical and dusty stats discussion will follow this announcement.

Offwidth, I think your preference for deaths per million is not correct.

Consider two countries with exactly the same disease dynamics (Ro, IFR etc), and an outbreak starting on the same day, with the same number of cases. In this case, the deaths per capita, D/N, will follow exactly the same track, but the raw fatality numbers, D, will not.

I think this is why you are arguing for per-capita plots.

I'm not actually arguing for per capita plots (see below) but I do want the current overall per capita number to help counter what I see as silly arguments about the NL and Sweden handling things better (we don't know either way yet, but so far from the data we are pretty similar, with if anything in my view the UK looking slightly the best). Even though we don't know yet, most of the epidemiology would favour them having it wrong.

In-country epidemics are actually lots of little exponential outbreaks aggregated. I think Italy and possibly Spain suffered in addition from the exodus of people from the main infected areas avoiding lock-down,  generating lots of new epidemic areas. For initial outbreaks its not unreasonable to assume there are some per capita effects on the overall numbers, assuming reasonably similar geographical spreads of initial infectors (I'm guessing mostly returning skiers from N Italy for us and NL). Yet, in the end, I agree actual deaths will be a better indicator of comparative control measures. Either way, the data will likely show the success of comparative measures from about 2 weeks time, starting most strongly with daily cases and initial indications in daily deaths. I think before then, NL and SW will have changed their plans.

[mrjonathanr]

I think I would be one of the people whose comments make you uncomfortable. Reading your post I find it hard to disagree with anything you have written to be honest. I think anyone looking at the issues you have raised would agree that all of them are unpleasant and involve suffering and heartbreak. I certainly don’t wish any of those things on anyone and would be deeply upset if they hit me personally. To give you some information my mother is in care home, prone to chest infections and I would guess highly likely to die from any viral infection which I think she will eventually contract despite the lockdown. My own personal feelings on that don’t stop me considering whether overall what we are doing is for the best

I would also say that the choices we are making now are not just saving lives they are costing lives and causing harm. Whether or not we save more and cause less harm from what we are doing seems like a valid discussion topic. I also appreciate that it is emotive which is why I have included a personal detail so that you can see that I am not heartless or lacking empathy but  am just curious as to whether or not what we are doing is the “least worst “ alternative.

Dave

I don't disagree with you Dave, the economic cost will hit us hard and it is a valid topic to debate. I am just urging a bit of circumspection- it's easy to drift into the abstract, hard to stay anchored in what the reality of these choices means.

[Stu Littlefair]

I think the epidemiology posts stray into the same territory, but for some of us it's our way of coping, and trying to get our head round things. Faced with the horror, it's comforting to retreat into the numbers...

[Stu Littlefair]

In-country epidemics are actually lots of little exponential outbreaks aggregated ... For initial outbreaks its not unreasonable to assume there are some per capita effects on the overall numbers

This is valid point, but you've drawn the wrong conclusion from it. Initially, each individual outbreak is small compared to it's host population, so the fact that they are fragmented doesn't matter. It's later on that population effects matter.

The general point that the "population" that affects the spread isn't the whole country's population is a good one though. And that you need to be careful about interpreting tail-offs in death rates because of this.

Per-capita plots aren't really useful at all until the very end stages of an outbreak. Look at the bottom plot in my two examples; the two UK models appear in very different positions throughout, despite all the properties of disease transmission being identical. The point at which per-capita comparisons are useful are once all this is over.

[tomtom]

What interests me though Stu - is that looking at the global data - the trends in the per capita death rates are far more similar than any of the other curves... (temporal shifting aside)

[gme]

Warning: mathematical and dusty stats discussion will follow this announcement.

Offwidth, I think your preference for deaths per million is not correct.

Consider two countries with exactly the same disease dynamics (Ro, IFR etc), and an outbreak starting on the same day, with the same number of cases. In this case, the deaths per capita, D/N, will follow exactly the same track, but the raw fatality numbers, D, will not.

I think this is why you are arguing for per-capita plots.

I'm not actually arguing for per capita plots (see below) but I do want the current overall per capita number to help counter what I see as silly arguments about the NL and Sweden handling things better (we don't know either way yet, but so far from the data we are pretty similar, with if anything in my view the UK looking slightly the best). Even though we don't know yet, most of the epidemiology would favour them having it wrong.

In-country epidemics are actually lots of little exponential outbreaks aggregated. I think Italy and possibly Spain suffered in addition from the exodus of people from the main infected areas avoiding lock-down,  generating lots of new epidemic areas. For initial outbreaks its not unreasonable to assume there are some per capita effects on the overall numbers, assuming reasonably similar geographical spreads of initial infectors (I'm guessing mostly returning skiers from N Italy for us and NL). Yet, in the end, I agree actual deaths will be a better indicator of comparative control measures. Either way, the data will likely show the success of comparative measures from about 2 weeks time, starting most strongly with daily cases and initial indications in daily deaths. I think before then, NL and SW will have changed their plans.

"silly arguments" nice way to wind people. I am guessing this is pointed at me as i am the one who keeps mentioning the Netherlands. I only do so as i am talking to Holland every day and on a practical level not just looking at some fucking charts and what they might say in two weeks time.
 
I dont remember saying they are handling the situation better just pointing out they are doing things differently and how that difference is effecting the economy.

[Stu Littlefair]

TomTom - I'm not sure that's true is it?

If you look at the chart I put online and plot all the countries with advanced outbreaks, only Iran (dodgy figure), South Korea (known exception) and Japan (mystery) are outliers. The rest basically follow the same trend.

Is a per-capita graph tighter than this? I've not seen one...

[Offwidth]

In-country epidemics are actually lots of little exponential outbreaks aggregated ... For initial outbreaks its not unreasonable to assume there are some per capita effects on the overall numbers

This is valid point, but you've drawn the wrong conclusion from it. Initially, each individual outbreak is small compared to it's host population, so the fact that they are fragmented doesn't matter. It's later on that population effects matter.

The general point that the "population" that affects the spread isn't the whole country's population is a good one though. And that you need to be careful about interpreting tail-offs in death rates because of this.

Per-capita plots aren't really useful at all until the very end stages of an outbreak. Look at the bottom plot in my two examples; the two UK models appear in very different positions throughout, despite all the properties of disease transmission being identical. The point at which per-capita comparisons are useful are once all this is over.

I disagree. If the UK and NL got roughly the same (geographically spread) initial per capita infections from N Italy skiing (as one might expect) the initial infection numbers will be per capita based but because the distribution will be focussed on the big cities, where infection is easier, growth less constrained and tracing harder, this might move away from per capita as it grows.

[Stu Littlefair]

I don't see what your point is there Offwidth? It seems to be a jumble of mixed points.

Edit: Is your point that *if* the initial infection is the same size per capita, and the outbreaks start at the same time, the disease should follow a similar track? If so, yes, but they'll follow the same track on both graphs at this stage. There's no justification for scaling per capita during an outbreak.
 
If you don't want to take my word and evidence for it, here's some support from the stats team at the FT:

https://twitter.com/jburnmurdoch/status/1244380095164420101?s=21

Quote John burn-Murdoch:

"Plenty of things indirectly related to population: can make a difference. Population density can increase the rate of spread. Countries with major travel hubs may end up with multiple outbreaks. But population alone is a bad thing to adjust for if wanting to assess how bad an outbreak is"

[Offwidth]

"silly arguments" nice way to wind people. I am guessing this is pointed at me as i am the one who keeps mentioning the Netherlands. I only do so as i am talking to Holland every day and on a practical level not just looking at some fucking charts and what they might say in two weeks time.
 
I dont remember saying they are handling the situation better just pointing out they are doing things differently and how that difference is effecting the economy.

To be 100% clear I did not aim this at you (nor anyone else here). I think you have been pretty balanced in the discussion. I meant it when I said this is the most sane bubble I know on the internet... people can nearly always disagree here without it becoming a highly aggressive flame war.

I was referring to the fact it seems to be all over social media and suits the libertarian viewpoint favoured by people who care more about the economy now than the potential outcomes later (much worse in terms of deaths and likely so even in in terms of the ecomony if we dont lockdown) and the denialism of many others, who are less guilty. Sadly charts in the coming months will be real, not best estimates and the epidemiology is hard to challenge. The 0.2% point is similar.....the very best data we have from extensive tests is 0.3% where the lockdown was early and hard and way off health response capacity. UKB also discussed this sensibly.

[Offwidth]

I don't see what your point is there Offwidth? It seems to be a jumble of mixed points.

Edit: Is your point that *if* the initial infection is the same size per capita, and the outbreaks start at the same time, the disease should follow a similar track? If so, yes, but they'll follow the same track on both graphs at this stage. There's no justification for scaling per capita during an outbreak.
 
If you don't want to take my word and evidence for it, here's some support from the stats team at the FT:

https://twitter.com/jburnmurdoch/status/1244380095164420101?s=21

Quote John burn-Murdoch:

"Plenty of things indirectly related to population: can make a difference. Population density can increase the rate of spread. Countries with major travel hubs may end up with multiple outbreaks. But population alone is a bad thing to adjust for if wanting to assess how bad an outbreak is"

As I said, I don't want per capita death graphs. I do want the per capita data as a balance to total mortality  graphs. I'm saying if  the main initial infections were geographically distributed, and approximately per capita, the initial growth will have had some strongly per capita factors.

[Oldmanmatt]

To me, andI know it’s probably naive, but I’m interested in the per capita numbers, because in nations of varying sizes, the individual has a varying value.
Given the degree of economic parity amongst western nations, relative to the global spectrum; a Dutch citizen is worth “more” than a Brit. That swings both ways, individuals within the “productive” bracket contribute “more” individually and those “no longer productive” cost more, individually.
I know those terms sound provocative, that’s not the point, it’s just the only/best way this Engineer could think of delineating as clearly as possible. It’s not representative of “deserving”.

It the same for any resource isn’t it? The value of a unit increases with it’s ubiquity?

So, the impact of a larger relative loss, on a smaller nation, is relatively greater? Economically at least. So, the Dutch ought to have taken greater care of that resource?

Unless the “resource” sits on the other side of the seesaw...

That’s just cynical musing, not a serious condemnation of the Dutch position.

[Offwidth]

A report of secondary deaths due to a stressed  health system not being able to cope:

https://www.bbc.co.uk/news/world-us-canada-52137160

[Stu Littlefair]

As I said, I don't want per capita death graphs. I do want the per capita data as a balance to total mortality  graphs.

You need to define what you mean by "data" then. Are you talking about cases?

I'm saying if  the main initial infections were geographically distributed, and approximately per capita, the initial growth will have had some strongly per capita factors.

No. It really won't. The initial stages are completely and utterly blind to the size of the population. They do depend on things like concentration of the initial cases, and population density in outbreak locations. Total population is a very poor proxy for any of those things.