SERVING THE QUANTITATIVE FINANCE COMMUNITY

 
User avatar
Alan
Posts: 10167
Joined: December 19th, 2001, 4:01 am
Location: California
Contact:

Re: Models for Covid-19

April 2nd, 2020, 9:07 pm

We can sort of see the lag between cases and deaths in the worldometers data. South Korea is therefore interesting because cases have been level for a few weeks, as have deaths. From this it looks like the death rate is about 6%. Of course it could be less if not all cases are reported or if death is attributed to the virus in "error" ("with" rather than "from"). But equally it could be higher (again distinction between "with" and "from").

Maybe I am reading it wrong but the graph of  Outcome of Cases says to me the death rate is trending down from 2.88% (Apr 1 value), so likely below that level. If people take at least a month to die, then the current deaths from the chart you reference are more likely coming from the surge period when daily new cases were around 500. That would put the death rate around 1.5%, a more plausible ultimate asymptote to the Outcome of Cases chart.  
 
User avatar
Paul
Posts: 10775
Joined: July 20th, 2001, 3:28 pm

Re: Models for Covid-19

April 3rd, 2020, 12:45 am

Could be. I didn't know it took that long.
 
User avatar
Alan
Posts: 10167
Joined: December 19th, 2001, 4:01 am
Location: California
Contact:

Re: Models for Covid-19

April 3rd, 2020, 3:34 am

Here's a report. It says
Clinical Progression

Among patients who developed severe disease, the medium time to dyspnea ranged from 5 to 8 days, the median time to acute respiratory distress syndrome (ARDS) ranged from 8 to 12 days, and the median time to ICU admission ranged from 10 to 12 days.5,6,10,11 Clinicians should be aware of the potential for some patients to rapidly deteriorate one week after illness onset. Among all hospitalized patients, a range of 26% to 32% of patients were admitted to the ICU.6,8,11 Among all patients, a range of 3% to 17% developed ARDS compared to a range of 20% to 42% for hospitalized patients and 67% to 85% for patients admitted to the ICU.1,4-6,8,11 Mortality among patients admitted to the ICU ranges from 39% to 72% depending on the study.5,8,10,11 The median length of hospitalization among survivors was 10 to 13 days.1,6,8


I had to look up 'dyspnea': labored breathing. It's kind of ambiguous. If the times are for events that occur one after the other, then median time to ICU admission is 23 to 32 days after developing a severe case. Then, I saw somewhere else that the median time in the ICU for non-survivors, was 4 days. Under that reading, the median time to dying after severe symptoms, for those who do die, is about a month. On the other hand, if all the times given are from 'developing a severe case', then its only the reported 10 to 12 days to ending up in the ICU. Adding another 4 for non-survivors yields only half a month of time for those who die.     
Last edited by Alan on April 3rd, 2020, 3:48 am, edited 1 time in total.
 
User avatar
Paul
Posts: 10775
Joined: July 20th, 2001, 3:28 pm

Re: Models for Covid-19

April 3rd, 2020, 3:45 am

So that’s, er, good news!
 
User avatar
Alan
Posts: 10167
Joined: December 19th, 2001, 4:01 am
Location: California
Contact:

Re: Models for Covid-19

April 3rd, 2020, 3:48 am

See my edit.
 
User avatar
Paul
Posts: 10775
Joined: July 20th, 2001, 3:28 pm

Re: Models for Covid-19

April 3rd, 2020, 3:53 am

It's all meaninglesss anyway:

NOTE ON FRANCE DATA:

On April 2, France reported 884 additional deaths that have occurred in nursing homes over the past days and weeks [source]. The French Government did not include these deaths in their official count, as their count only takes into consideration deaths of hospitalized patients. Following international standards of correct inclusion, our statistics will include these deaths, and will add them to the April 2, 2020 count following the attribution criteria of date of report.
If and when the French government determines and communicates the correct distribution of these additional deaths over time, we will adjust the historical data accordingly. A similar issue took place on February 12, when China reported an additional 13,332 new cases in a single day due to a change in how cases were diagnosed and reported in Hubei.

The French only count you if you die in the right place. Can't say I'm surprised. 

I do wish all countries would get their acts together in just something simple as agreeing what constitutes dying from and dying with.
 
User avatar
Alan
Posts: 10167
Joined: December 19th, 2001, 4:01 am
Location: California
Contact:

Re: Models for Covid-19

April 3rd, 2020, 4:22 am

Looked up a couple of the references in my post. If Fig. 2 here is representative, it looks like the more pessimistic reading is correct: median time from onset to ICU admission was less than two weeks (Wuhan). Then, this reference says "the survival time of the non-survivors is likely to be within 1–2 weeks after ICU admission". So, in Wuhan anyway, looks like it killed in less than a month from onset.
 
User avatar
Paul
Posts: 10775
Joined: July 20th, 2001, 3:28 pm

Re: Models for Covid-19

April 3rd, 2020, 7:05 am

If you half close your eyes while looking at the data from Italy you could say that there's one week between the peak of cases and peak of deaths.

But for each country we don't know how many people are lying at home ill, not bothering to tell anyone. Or if they do, whether they are included in cases. Perhaps they don't count until they get admitted to a hospital. Or maybe it has to be the intensive care unit of a hospital before they are recorded. I mean, if France can ignore nearly 900 deaths then I suspect most data from most countries is garbage.
 
User avatar
Cuchulainn
Topic Author
Posts: 62126
Joined: July 16th, 2004, 7:38 am
Location: Amsterdam
Contact:

Re: Models for Covid-19

April 4th, 2020, 10:56 am

Here is a paper on a conceptual SEIRNDC model that seems to subsume the protean and emerging models on this thread.
https://www.researchgate.net/publicatio ... tal_action
 
User avatar
Paul
Posts: 10775
Joined: July 20th, 2001, 3:28 pm

Re: Models for Covid-19

April 4th, 2020, 5:40 pm

A problem is that every reasonable model will fit the data. That's the nature of these models. The vast majority of research is to get promotion for the academics. (This is from China so their motivation is not getting shot in the back of the neck or disappearing.)
 
User avatar
Cuchulainn
Topic Author
Posts: 62126
Joined: July 16th, 2004, 7:38 am
Location: Amsterdam
Contact:

Re: Models for Covid-19

April 4th, 2020, 7:25 pm

Speaking of censorship, I was listening this morning on the radio sender BNR to Prof. Ron Fouche of Erasmus University, Rottterdam on corona etc. He was the person who first analysed the Mexican virus a few years back and came up with a vaccine or something. Anyways, he claimed that the Americans wouldn't let him publish his work. ... biological warfare.

At least the Chinese have published something that you are free to criticise from a scientific viewpoint It is difficult with unreleased, undocumented C code from 2003.

It will never be published, I suspect.
 
User avatar
Cuchulainn
Topic Author
Posts: 62126
Joined: July 16th, 2004, 7:38 am
Location: Amsterdam
Contact:

Re: Models for Covid-19

April 6th, 2020, 11:06 am

Image
 
User avatar
Cuchulainn
Topic Author
Posts: 62126
Joined: July 16th, 2004, 7:38 am
Location: Amsterdam
Contact:

Re: Models for Covid-19

April 6th, 2020, 1:38 pm

 
User avatar
Cuchulainn
Topic Author
Posts: 62126
Joined: July 16th, 2004, 7:38 am
Location: Amsterdam
Contact:

Re: Models for Covid-19

April 7th, 2020, 7:26 pm

'The models show hundreds of thousands of people are going to die and you know what I want to do? I want to come way under the model,' he said. 'The professionals did the models and I was never involved in a model. At least this kind of a model.' 

Guess who said that?
 
User avatar
Cuchulainn
Topic Author
Posts: 62126
Joined: July 16th, 2004, 7:38 am
Location: Amsterdam
Contact:

Re: Models for Covid-19

April 8th, 2020, 11:17 am

ABOUT WILMOTT

PW by JB

Wilmott.com has been "Serving the Quantitative Finance Community" since 2001. Continued...


Twitter LinkedIn Instagram

JOBS BOARD

JOBS BOARD

Looking for a quant job, risk, algo trading,...? Browse jobs here...


GZIP: On