The Coronavirus, not the beer

[thethe]

I haven't delved into this. But I have a suspicion that they project hospitalizations and need for ventilators as a multiple of their death projections. Given that their death projections have been much better than their hospitalizations projections, maybe they should have adjusted their multiple as the data came in. If indeed this is the issue.

What could cause the multiple to change? Again I haven't looked into it in any depth so this is just a hunch. It might reflect differences or changes in hospital protocol regarding how sick someone has to be in order to be admitted.

Has the protocol changed materially in the past 3 months that would explain being off by this large of a multiple?

Using the end result as a way to back into the hospitalization is an awful way of doing it especially since hospitalization is the real issue.

I guess we should give them a cookie because through all their various 'What-ifs' they produced an aggregate range that almost certainly contain the actual results?

 

[nsacpi]

Its fun playing in the world of the theoretical and never having to explain why you are right or why you were wrong.

The way I reach that conclusion is by working with the estimated number of infections at various points. Jan 15, Feb 15, Mar 15, etc. Given an R naught of around 2.5 (probably over 3 in NY and closer to 2 in the rest of the country) it has been growing fast and would have been a multiple of what it actually is now absent social distancing and other measures put in place around mid-March.

An R naught of 2.5 implies the number of new cases grows by a multiple of around 15 every 30 days. That's why moving the start of social distancing up or back by just a few days has very big effects.

 

[thethe]

The way I reach that conclusion is by working with the estimated number of infections at various points. Jan 15, Feb 15, Mar 15, etc. Given an R naught of around 2.5 (probably over 3 in NY and closer to 2 in the rest of the country) it has been growing fast and would have been a multiple of what it actually is now absent social distancing and other measures put in place around mid-March.

Those R naughts are not in line with most recent studies done that have it in upwards of 5. Its possible that i missed some more recent studies and if so I'm sorry.

Maybe you are talking about real transmission rates as it rips through a population. If thats the case than that number is not stagnant.

Also - You are estimating the infections on those dates based on a certain multiple of confirmed infections?

 

[nsacpi]

Has the protocol changed materially in the past 3 months that would explain being off by this large of a multiple?

I don't know. As I said that was my hunch.

My guess is the multiple used was based on data from other countries. And maybe our hospitals have adopted somewhat different protocols. I could be completely wrong about this. I'm just offering a guess. It could be I'm completely mistaken about how they project hospitalizations.

 

[nsacpi]

Those R naughts are not in line with most recent studies done that have it in upwards of 5.

The study I saw had to do with Wuhan in its early stages having an R naught of 5. Other work indicates this can't be generalized for all settings. NY city might be pretty high though.

The higher the R naught the more the timing of social distancing will affect outcomes. So an R naught higher than the 2.5 I've been assuming would work to reinforce my point about how catastrophic things would be right now if we had delayed social distancing a few days.

 

[nsacpi]

Also - You are estimating the infections on those dates based on a certain multiple of confirmed infections?

I'm looking at all the data. Cases, tests and deaths. And trying to infer underlying number of infections every 10 days from mid January to mid March. And then trying to infer mid-April infections in a "do nothing" scenario.

 

[thethe]

The study I saw had to do with Wuhan in its early stages having an R naught of 5. Other work indicates this can't be generalized for all settings. NY city might be pretty high though.

The higher the R naught the more the timing of social distancing will affect outcomes. So an R naught higher than the 2.5 I've been assuming would work to reinforce my point about how catastrophic things would be right now if we had delayed social distancing a few days.

Or the high R naught indicates that it already ravaged through a dense urban population much faster than what we now know and therefore your projection for future months is not valid because of the percentage of uninfected population.

This is what I'm saying in the world of the theoretical. I could be 100% wrong and once we open up another 50K will die. Or you could be 100% wrong and if no social distancing measures were applied we would have had a similar outcome to what we are seeing now.

Despite the fact that we have seen studies which indicate larger percentages of the population then we previously believed you are still using the same base R naught which is just plain wrong.There is a big difference between the virus properties and how it behaves in the population.

 

[thethe]

I'm looking at all the data. Cases, tests and deaths. And trying to infer underlying number of infections every 10 days from mid January to mid March. And then trying to infer mid-April infections in a "do nothing" scenario.

What is your multiple on confirmed positive cases to represent the asymptomatic population?

 

[nsacpi]

Or the high R naught indicates that it already ravaged through a dense urban population much faster than what we now know and therefore your projection for future months is not valid because of the percentage of uninfected population.

Possible. But imo unlikely. The testing in NY city has been returning about 50% positives in recent weeks. So if 50% of those being tested are positive then I would guess the overall number for the city is significantly below that. The number we really want is how many are either infected right now AND were infected in the past and have immunity. Your point about herd immunity might be correct by now. But in my judgement not correct as of late March/early April. So we would have had explosive growth in the number of cases in NY city in that time frame without social distancing and other measures. Your argument is not crazy when it comes to herd immunity kicking in. My dispute mainly is about whether it was advanced enough to have kicked in by late March/early April.

 

[nsacpi]

What is your multiple on confirmed positive cases to represent the asymptomatic population?

As I said I'm looking at all the data including cases, deaths and tests. So I'm not going just by a multiple of confirmed cases. Imo that wouldn't make sense given how the testing regimen has changed so much from January to now. When the prevalence of testing changes so much, confirmed cases reflects those changes as much as changes in underlying infections.

I'm working with numbers for underlying cases that looks like this (for the country)

Feb 1 2,000 underlying cases
Feb 15 20,000
Mar 1 200,000
Mar 15 2,000,000

As you can see that is a raging fire.

 

[thethe]

Possible. But imo unlikely. The testing in NY city has been returning about 50% positives in recent weeks. So if 50% of those being tested are positive then I would guess the overall number for the city is significantly below that. The number we really want is how many are either infected right now AND were infected in the past and have immunity. Your point about herd immunity might be correct by now. But in my judgement not correct as of late March/early April. So we would have had explosive growth in the number of cases in NY city in that time frame without social distancing and other measures. Your argument is not crazy when it comes to herd immunity kicking in. My dispute mainly is about whether it was advanced enough to have kicked in by late March/early April.

I think the timing of when (I guess if applies as well) we hit herd immunity is valid. But using confirmed positives as your proxy is dangerous and contains a lot of noise. We can see the rate of growth in confirmed cases (again terrible proxy based on what we are learning about asymptomatic population) spiked in mid to late march. That kind of tells us that the immunity wall began to form 2-3 weeks prior to that and the rate of growth plummeted. I understand this isn't a case closed piece of evidence but its pretty ****ing damning.

 

[thethe]

As I said I'm looking at all the data including cases, deaths and tests. So I'm not going just by a multiple of confirmed cases. Imo that wouldn't make sense given how the testing regimen has changed so much from January to now. When the prevalence of testing changes so much, confirmed cases reflects those changes as much as changes in underlying infections.

Nobody is trying to use aggregate amounts though. Of course the aggregates were increasing. I definitely presented what my intention was incorrectly. There has to be some sort of approach to take the confirmed case percentage and apply that to the complete active population. Otherwise, you are missing a huge swatch of the data.

 

[thethe]

I'm working with numbers for underlying cases that looks like this (for the country)

Feb 1 2,000 underlying cases
Feb 15 20,000
Mar 1 200,000
Mar 15 2,000,000

As you can see that is a raging fire.

But you see this is where I think you are wrong when you project outwards. You are reaching a critical mass point where the true transmission rate falls off a cliff.

 

[nsacpi]

I think the timing of when (I guess if applies as well) we hit herd immunity is valid. But using confirmed positives as your proxy is dangerous and contains a lot of noise. We can see the rate of growth in confirmed cases (again terrible proxy based on what we are learning about asymptomatic population) spiked in mid to late march. That kind of tells us that the immunity wall began to form 2-3 weeks prior to that and the rate of growth plummeted. I understand this isn't a case closed piece of evidence but its pretty ****ing damning.

I think the main criticism I have of your position (if I understand it correctly) is you want to use a high R naught and low fatality rate in NY city to transition from very few cases to herd immunity without any carnage in between. I don't think that's possible even with an optimistic case fatality rate. To get to herd immunity you have to go through a massive number of deaths (far higher than what we've seen). Which unfortunately (if I'm right) means there is a risk of some large number of deaths later this year if we get a secondary wave in NY city.

 

[thethe]

I think the main criticism I have of your position (if I understand it correctly) is you want to use a high R naught and low fatality rate in NY city to transition from very few cases to herd immunity without any carnage in between. I don't think that's possible. To get to herd immunity you have to go through a massive number of deaths (far higher than what we've seen). Which unfortunately (if I'm right) means there is a risk of some large number of deaths later this year if we get a secondary wave in NY city.

I think what we saw in NYC was carnage. Especially when you compare it to everywhere else in the country.

The point that you make regarding we would have seen a massive number of deaths to acheive herd immunity is again your belief as is mine that we have already gotten there. Death is a by product of who is contracting the disease and not a general point.

 

[nsacpi]

But you see this is where I think you are wrong when you project outwards. You are reaching a critical mass point where the true transmission rate falls off a cliff.

I understand your argument but we differ in timing. Imo it is pretty unlikely that there was any significant herd immunity in late March/early April, either nationally or in NY city.

 

[nsacpi]

I think what we saw in NYC was carnage. Especially when you compare it to everywhere else in the country.

The point that you make regarding we would have seen a massive number of deaths to acheive herd immunity is again your belief as is mine that we have already gotten there. Death is a by product of who is contracting the disease and not a general point.

So as I understand it there are 3 components of your argument:

1) very high R naught for NY city

2) low case fatality rate generally

3) a disproportionate number of infections in NY city among those parts of the population that are healthy and young

 

[thethe]

I understand your argument but we differ in timing. Imo it is pretty unlikely that there was any significant herd immunity in late March/early April, either nationally or in NY city.

4.3M people a day take the subway.

Lets ignore that I believe the virus has been here since Q4 last year since that is a completely unproven hypothesis.

You still have 2.5 months (March 15th date) of a virus active in a population of 9M people where 4.3M per day take the subway. When the virus was fresh in the population the R naught could have been 10 at that point. Maybe higher.

 

[thethe]

So as I understand it there are 3 components of your argument:

1) very high R naught for NY city

2) low case fatality rate generally

3) a disproportionate number of infections in NY city among those parts of the population that are healthy and young

1) Absolutely
2) Been my position for almost a month
3) Yes but the question is how disproportionate. The majority of the commuting population is younger but within that there is a subset of at risk pre-existing conditions. My stance has been the subway is the main conductor on the virus and my experience taking the subway system for 3 years when I was working with PwC is driving my analysis on who is there to be infected.

 

[The Tariff King]

1) Absolutely
2) Been my position for almost a month
3) Yes but the question is how disproportionate. The majority of the commuting population is younger but within that there is a subset of at risk pre-existing conditions. My stance has been the subway is the main conductor on the virus and my experience taking the subway system for 3 years when I was working with PwC is driving my analysis on who is there to be infected.

Just curious what your route was.