Interesting numbers from a friend.

[Kestryll]

This was posted on FB by a friend of mine:



Several people asked questions about missing items or the criteria used, I've copied the answers since I suspect some here will have the same questions:

When calculating deaths, shouldn't it be a percentage of all resolved cases, not total population or total infected population? We don't know how many will die in the infected but not resolved population. So, wouldn't the most accurate mortality rate be dead/(dead+recovered)?

Why not break it down by age?

Any reason to not include Italy? Or did I miss it?

The Italian numbers are specious. Check the maps. The vast majority of infected and deceased come from areas with large concentrations of immigrants from the Middle East and North Africa. Both infamous for poor nutrition, bad hygiene, crowded living conditions, and eschewing health care facilities. Additionally the burial practices (in the ground by sun down that day) makes it impossible for third party verification of cause of death. Other factors is they have the oldest population in the EU, they have the highest obesity rate in the EU, and they have the highest tobacco use in the EU.

Italy also has a hybrid healthcare system (Medicare for all). With basic healthcare provided by the state. Elective procedures are handled by private insurance. As a result they have an physical infrastructure incapable of dealing with surge events. The lowest per capita bed rate among all industrial nations. An extremely out of balance Doctor patient ratio and last I looked it is impossible to track RN to patient ratio.

By all indications 100% of all non-violent or traffic accident related deaths are now reported as Covid 19.

[Sgt. J Beezy]

[Dirtlaw]

The Janitor comes through. Useful information.

[SAN compnerd]

I don't see it in the data, but does he have any data on what percentage of the population is getting it? The cruise ship in Japan had about 20% infected after a month in close quarters. Does he see that in the broader data set?

[harbormaster]

The US Government began temperature screening at several International Airports on the day China reported 75 cases. You really think there were 75 cases when our government acted? I'm guessing US intel on the ground in China reported the real issue. I don't think we will ever know China's real numbers. I don't think they are as bad as some but I think they are far worse then what was reported.

[SAN compnerd]

The US Government began temperature screening at several International Airports on the day China reported 75 cases. You really think there were 75 cases when our government acted? I'm guessing US intel on the ground in China reported the real issue. I don't think we will ever know China's real numbers. I don't think they are as bad as some but I think they are far worse then what was reported.

China will be doing a victory lap at some point to propagandize the world into the idea that they handled this outbreak better than the western world.

They will certainly be making the point that Communism was again superior to western free markets, it's a huge win for them as they see it.

[theLBC]

the issue is estimating anything based on "confirmed" cases when the actual number of cases could be 10x higher or more.

[maidendeth13]

In regards to all traffic related deaths in Italy now being counted as Covid deaths...

I'd have to think traffic related deaths in Italy are waaaaay down if the majority of the population is on lockdown...

Also, people are sketchy drivers there. Only place I've ever seen a bus overtake a motorcycle around a blind corner

[Sgt. J Beezy]

the issue is estimating anything based on "confirmed" cases when the actual number of cases could be 10x higher or more.

[as_rocketman]

The problem with statistics is... well, they're statistics.

We cannot think of number infected, number killed, and number recovered in the same way. Those metrics are almost totally uncoordinated. When we try to make them into percentages, all meaning is lost.

The number of infected is unknown in every population. What we have are the number with a positive test result, out of a population of those who were administered a valid test. In South Korea, where they've been testing furiously, they have still tested less than 1% of their population.

This is important because who gets tested depends on triage factors -- and that introduces systematic bias. In California, testing so far has been done under a mixture of rules, but the dominant rules are that they only test people who show up at ERs with (often dubious) symptoms. In Italy, the main reason their other statistics look so horrifying is that they only test not only admitted patients, but those already needing advanced life support. As a result their case fatality rate (CFR) looks truly apocalyptic, although it is certainly awful by any measure.

Another problem is that there are many time constants in these numbers. Testing may give a false positive, and it may give a false negative depending on exactly when the test was administered. A subject may be infected but not yet producing enough virus or cofactors to trigger the test. There are also lags in reporting, and there are even examples where the number of cases tracked isn't based on any genomic or antibody test at all, but instead on presumption based on symptoms and community risk.

The fatality rate has some of the same problems. Dead vs. alive is pretty darn definite of course, but in some cases the cause isn't so clear. If you went to the ER with trouble breathing due to something else (e.g., the actual flu, which is bad this year), odds are fair that you might acquire the new germ once you got there -- and no doctor is going to sort that out. On the other hand, fatalities may be linked to the disease without any positive test confirming the presence of the virus (Italy) or, the opposite, only logged as such with firm evidence (Germany).

There is also the lag between these signals. Fatalities are lagging infection by approximately four weeks. Recoveries are lagging infection by six to ten weeks. This is part of the reason this didn't look like a global threat, until suddenly it did -- there was no early understanding of how communicable this bug was, how early, how slow the incubation process was, or how long it takes even recovered patients to be truly rid of it. All those factors are driving this lag, and making it particularly hard to eyeball without experience.

Disclaimer, I have no experience in epidemiology, though I am a practicing data scientist. I am learning this in real time just like so many others.

[theLBC]

You are 100% correct, and this thing is still too early to run probability statistics and we have the information that we have, and that’s what I worked with.

Which is why I capitalized confirmed cases because someone tried to argue with me the probable infected, and, it’s too early to make those arguments.


Sent from my iPhone using Tapatalk Pro

right, not blaming you.
look at the recent test results
i think 10% are coming up positive, with new york up towards 20%?
of the positive tests, 80% have mild or no symptoms?

so out of 330 million, it could be that tens of millions would test positive just in the US.

[colossians323]

I think leaving china out of the equation makes great sense. It will be a far more accurate look at what is going on with real numbers.
Cheena despite all our socialists and msm squealing how great they are, is full of lies and deceit!

[stonefly-2]

the issue is estimating anything based on "confirmed" cases when the actual number of cases could be 10x higher or more.

Crystal ball says "garbage in, garbage out".

https://www.bloomberg.com/news/articles/2020-03-23/china-s-mobile-carriers-lose-15-million-users-as-virus-bites

[LTC-J]

I think leaving china out of the equation makes great sense. It will be a far more accurate look at what is going on with real numbers.
Cheena despite all our socialists and msm squealing how great they are, is full of lies and deceit!

While leaving China may help normalize the data a little bit, there is still too many issues to make comparing the numbers too worthwhile for me:
1) What is being reported a China flu death? I've heard reports of "if a person dies and tested positive for China flu, it is a China flu death". This makes apple/apple comparison difficult at best because I've heard cases people which China flu dying from issues NOT related... such as falling over a balcony... pneumonia complications are getting thrown in a China flu death.
2) China flu cases being reported. Not everyone is diligent about testing. I heard that South Korea was doing extensive testing. So this skews the results downward.
3) Comparing against population is also no good. There is no guarantee that the actual % being infected is even close for different countries. Never mind the urban vs rural.