Coronavirus

[SageBrush]

The IHME projections of Covid mortality for NYS are stable for the past 48 hours. If the national projection is increasing it is from elsewhere.
People who take the estimate changes as some sort of indication that the stat analysis is wrong should review the modeling essay over at 538.com.

As one example, IHME bakes into their analysis that every state in the Union will invoke SAH edicts this week. To the extent that is not true the estimates shift upwards.

 

[SageBrush]

The mean onset-to-death

When does counting start ? I will not be surprised to hear from hospital admission, but it could be from symptom onset.

 

[Daniel in SD]

When does counting start ? I will not be surprised to hear from hospital admission, but it could be from symptom onset.

Onset of symptoms. Infection to symptoms averages an additional 5.1 days in their model.

 

[AlanSubie4Life]

Sorry, you lost me there.

Just saying there is massive undercounting of cases, resulting in a smaller denominator. The numerator is also probably smaller than it should be, but on the other hand the numerator is exploded by a factor of nearly 4x, or more, due to the overloading of hospitals. (At least 4% of patients will die without medical care, as opposed to the normal 1-1.5%.)

So this apparent IFR should not be extended to other countries.

 

[AlanSubie4Life]

Oh yes … wrong by a mere factor of 10.

So, really nothing much to go by … even the hospitalization data is spotty at best (31k cumulative).

That said, our 14 day lagged CFR right now is not 2%, but 55% (5099/9197)....

14 days doesn’t work nor does 18 days (onset to death mean). This is because of 1) Delay on Presenting 2) Delay on diagnosis, 3) Delay of reporting of diagnosis. There is likely only a very short delay (1 day at most) on reporting of death which would go the other way.

According to the doctor typically patients don’t get sick enough to require admission until about 6-7 days after sx, so entirely possible the delay on presenting is often at least 6 days. So a total of 10 days subtracting from that 18 seems reasonable.

I think 6-7 days works best, but it is really just a wild shot in the dark (totally empirical, and looking for a specific result, based on my priors - the apparent near-steady-state Korean result).

 

[wdolson]

No. And thank God No!

This blogger quote a couple of small Chinese studies showing very low survivor ratios for COVID-19 patients on ventilators (14% and 3%). He also points to a larger UK study that showed 33%. A small Seattle study included 18 patients on ventilators:
9 died
6 were extubated
3 were still on the ventilator when the study ended.

I'll go with 33% survival. Maybe it improves a little with experience, but still much lower than I expected.

The 1.1 million includes both public and private labs.

200k * 0.02 = 4k

That said, our 14 day lagged CFR right now is not 2%, but 55% (5099/9197)....

The results may be different because the Chinese fudged the numbers, or the criteria about when to put someone on a ventilator is different in China. They may only have put people on ventilators who they thought had a good chance of survival or may have put people on ventilators sooner. We can't be sure all the variables were the same.

On NPR today I heard an interview with an ICU doctor who said with normal pneumonia, they put someone on a ventilator for 3 days, give them antibiotics, and they often recover. But with COVID-19 there is nothing they can do to make people better except wait and hope. She's seen a 50% mortality rate on ventilators and she said another 25% or so survive but may never be able to breathe on their own again.

I think the more accurate data is what we're gathering here in the US based on US diagnostic criteria. If possible we should be consulting with Chinese doctors who treated this to see if they found anything that helps improve survival rates.

 

[SageBrush]

Onset of symptoms. Infection to symptoms averages an additional 5.1 days in their model.

Yes, although they mention that in ~ 70% of their data set the onset of symptoms was not actually known so they 'imputed' it.

 

[Daniel in SD]

Yes, although they mention that in ~ 70% of their data set the onset of symptoms was not actually known so they 'imputed' it.

Yeah, it seems like there should be much better data now. When they wrote the paper in early March the only data was from China.

 

[Cosmacelf]

Wow, this is a great interview. The researcher is from California, and the interviewer is a New Zealand reporter. The reporter asks intelligent questions and the researcher provides rapid fire informative answers. Oh, and this is a promising antibody therapy to cure COVID-19 patients.

 

[Daniel in SD]

At least 4% of patients will die without medical care, as opposed to the normal 1-1.5%.

I'm very skeptical of your 4% number but as you know I'm also skeptical of medical care in general. It presumes that everyone who requires hospitalization would die without it. I presume the treatment for most patients is oxygen and fluids? Gov. Cuomo said today that only 20% of people put on ventilators survive. Anyway, the answer is not something we want to find out.

 

[SageBrush]

I was mulling over the possible CFR of the 40 - 60 year old demographic without major risk factors (obesity, diabetes, CV disease, tobacco.)
One approach could be to use the biologic age difference that a risky lifestyle conveys since it is known that about 10-15 years of life are lost to risk factors. That would suggest that a low risk 65 y/o may have the save morbidy/mortality risk from Covid as a 50 - 55 y/o average American. By the same reasoning a 65 y/o with multiple risk factors would be biologically equivalent to an 80 y/o American without risk factors.

This implies that the 40 - 60 age group without risk factors will have a low CFR. My WAG is under 1/1000 instead of the 1/100 reported overall for this age cohort.

 

[SageBrush]

I'm very skeptical of your 4% number but as you know I'm also skeptical of medical care in general. It presumes that everyone who requires hospitalization would die without it. I presume the treatment for most patients is oxygen and fluids? Gov. Cuomo said today that only 20% of people put on ventilators survive. Anyway, the answer is not something we want to find out.

About 1/4 US admissions for Covid require ICU care,
80% of ICU care go on to a vent,
and 20% survive the vent

You are quite right though: those hospital systems that will be overwhelmed would be much better off figuring out how to meet the supply requirement of O2 to outpatients rather than screw around with ICUs and vents.

 

[SMAlset]

Any statistics of whether people who get CV19 fair better and don’t progress to pneumonia if they’ve had an pneumonia shot? I know most assisted care and nursing homes require patients to be up to date on them but also know the pharmacies won’t give you one if you aren’t 65 or older. Helpful or not with this virus?

 

[Norbert]

14 days doesn’t work nor does 18 days (onset to death mean). This is because of 1) Delay on Presenting 2) Delay on diagnosis, 3) Delay of reporting of diagnosis. There is likely only a very short delay (1 day at most) on reporting of death which would go the other way.

According to the doctor typically patients don’t get sick enough to require admission until about 6-7 days after sx, so entirely possible the delay on presenting is often at least 6 days. So a total of 10 days subtracting from that 18 seems reasonable.

I think 6-7 days works best, but it is really just a wild shot in the dark (totally empirical, and looking for a specific result, based on my priors - the apparent near-steady-state Korean result).

For the US, nothing works because of the extreme lack of testing especially 2 weeks ago (probably less than 10%), and also we are still early in the curve. For China, 2 weeks works very well. As it does for Iceland, Germany and SK if applied with some discretion for early times.

In the early time of an curve, nothing works well because you either can have lots of positive tests but no deaths yet, or lots of deaths but no tests yet.

I got better results in predicting trends by using 14 days, and being careful about whether that makes sense in a specific situation. In the case of Germany, the result was high, 9.6%, but I could already tell from the context. I knew that the CFR would go not just to 1%, but above. In the case of SK I said recently it would go towards 1.9% or 2%, and that is still what I am saying.

 

[EVNow]

14 days doesn’t work nor does 18 days (onset to death mean). This is because of 1) Delay on Presenting 2) Delay on diagnosis, 3) Delay of reporting of diagnosis. There is likely only a very short delay (1 day at most) on reporting of death which would go the other way.

According to the doctor typically patients don’t get sick enough to require admission until about 6-7 days after sx, so entirely possible the delay on presenting is often at least 6 days. So a total of 10 days subtracting from that 18 seems reasonable.

I think 6-7 days works best, but it is really just a wild shot in the dark (totally empirical, and looking for a specific result, based on my priors - the apparent near-steady-state Korean result).

Yes - 8 to 10 days from case confirmed to death among 70+, longer for <70 yrs ?

But we have 200k cases and 31k hospitalizations. So, possibly a lot of people are being sent back home after testing - and they would come back to hospital if they get serious.

BTW, from this - we can see length of stay in hospital, ICU and Vent in NJ.

 

[SageBrush]

Any statistics of whether people who get CV19 fair better and don’t progress to pneumonia if they’ve had an pneumonia shot? I know most assisted care and nursing homes require patients to be up to date on them but also know the pharmacies won’t give you one if you aren’t 65 or older. Helpful or not with this virus?

The Pneumococcus serotypes in the vaccine you are thinking about are rarely the causes of ventilator associated pneumonia

 

[dhanson865]

Yes - 8 to 10 days from case confirmed to death among 70+, longer for <70 yrs ?

But we have 200k cases and 31k hospitalizations. So, possibly a lot of people are being sent back home after testing - and they would come back to hospital if they get serious.

BTW, from this - we can see length of stay in hospital, ICU and Vent in NJ.

am I reading that right, they've already run out of ICU beds (reached capacity in ICU) and are about a week away from running out of any kind of beds in the hospital for the general population (non ICU patients)?

 

[Norbert]

The IHME projections of Covid mortality for NYS are stable for the past 48 hours. If the national projection is increasing it is from elsewhere.
People who take the estimate changes as some sort of indication that the stat analysis is wrong should review the modeling essay over at 538.com.

As one example, IHME bakes into their analysis that every state in the Union will invoke SAH edicts this week. To the extent that is not true the estimates shift upwards.

What did change, for example, is their predicted peak death per day (nationwide). By 18% in a single day, upwards.

Of course many deaths are from NY, but I wouldn't know what exactly changed their prediction. I think one of their notes says it is data from elsewhere. In any case, it seems their predictions are very sensitive to incoming data of a single day, and perhaps made too early, in a sense. But of course we need something.

See my previous post:

The IHME forecast for the peak deaths per day number already seems to have gone up, since the White House task force presentation, 1 day later:

In the WH presentation, the peak deaths per day were expected to be:
2,214 on April 15.

The current number I find on the website is:
2,607 on April 16.

I'll non-scientifically predict that it will be pushed further back and further up.

 

[SageBrush]

am I reading that right, they've already run out of ICU beds (reached capacity in ICU) and are about a week away from running out of any kind of beds in the hospital for the general population (non ICU patients)?

If you are talking about the IHME projections, then probably not because the ICU availability is based on pre-Covid data. Actual availability today e.g is improved by cancelling elective surgeries.

That said, NYC is on the ropes and they have a steep climb ahead.

 

[Norbert]

Onset of symptoms. Infection to symptoms averages an additional 5.1 days in their model.

I think it is going to take you a while to figure out what exactly different reports are referring to, if they they have clarity about that themselves.

Just for clarity about my posts, I am simply talking about the lag between reported "total cases" and "total deaths", while this is only one of several strong factors affecting the larger CFR picture.

It seems to me that these models spend too much talking about arcane details and not enough time talking about basics like the testing level and other limitations to the data that they work with.