Latrinsorm
10-03-2020, 11:02 AM
I've talked before about our discovery of more (http://forum.gsplayers.com/showthread.php?126122-On-the-Cure) and less (http://forum.gsplayers.com/showthread.php?125754-On-Time&p=2167441#post2167441) successful treatments for COVID-19, and how seroprevalence (http://forum.gsplayers.com/showthread.php?126253-On-Being-Halfway&p=2172016#post2172016) studies allow us to control for the fact that our testing has accelerated dramatically over time. So I thought I'd put that information together to ask the question: is there any evidence that we are getting better at keeping people with COVID-19 alive?
No.
.
The CDC has continued to perform seroprevalence studies (https://covid.cdc.gov/covid-data-tracker/?CDC_AA_refVal=https%3A%2F%2Fwww.cdc.gov%2Fcoronav irus%2F2019-ncov%2Fcases-updates%2Fcommercial-labs-interactive-serology-dashboard.html#serology-surveillance) through mid-July for a few areas. Here is the mortality rate (deaths per total cases, not deaths per confirmed cases) that that data implies for three localities listed in alphabetical order:
0.7% - Minnesota
1.3% - Missouri
1.0% - New York City metro area
While NYC had a much earlier outbreak (87% of its deaths on the CDC measurement in July by the CDC measurement in May) than Missouri (73%) and Minnesota (67%), none of them saw anything other than 1% to a statistically significant degree. It's also important to note that these numbers aren't impacted by the inarguable fact that deaths lag behind cases: at these points in their respective outbreaks the changes in the numerator just aren't big enough compared to the denominator. With that said, it's useful to get a bead on how the numbers change if we were looking during a big increase, so if we look at the April 1st numbers for the NYC metro we get a reading of 0.3% for simultaneous deaths, 0.9% for deaths a week later, and 2.2% for deaths two weeks later. Since we know the NYC metro area was really 1.0%, if we're looking at another locality during a big increase we should compare deaths about 9 days later ("about" since some places don't release data on weekends).
.
Okay, but obviously we'd really like to compare to a Florida or Texas, a really big state with a much later peak. Unfortunately the previous study was only on very specific areas that didn't include them. (There is one on South Florida but it was discontinued in May, so it's not helpful for this.) The CDC does have a broader set of data (https://covid.cdc.gov/covid-data-tracker/?CDC_AA_refVal=https%3A%2F%2Fwww.cdc.gov%2Fcoronav irus%2F2019-ncov%2Fcases-updates%2Fcommercial-labs-interactive-serology-dashboard.html#national-lab) which necessarily means shallower as well; for example, we only have one measurement for Texas on August 15th that gives a mortality rate of 0.5% to 1.0%. Similarly, our only measurement for Florida works out to 0.8% to 1.8%. Again I want to stress that this data is a lot less reliable than the above so the error bars are even bigger than those listed here, but when even the Gaussian error bars are sufficient for the purposes of making it impossible to prove a sub 1% mortality rate it doesn't matter.
.
One of humanity's greatest strengths is the belief that we can control our environments.
Sometimes, that belief is wrong.
1% of New Yorkers who contracted coronavirus, tested or otherwise, died.
1% of Texans who contracted coronavirus, tested or otherwise, died.
1% of Floridians who contracted coronavirus, tested or otherwise, died.
When their peaks happened didn't change this.
The demographics getting infected didn't change this.
Their nursing home policies didn't change this.
New treatments (real and imagined) didn't change this.
The only thing that has ever changed the number of people dying is the number of people getting infected - all people.
When we bring infections down, deaths go down too.
When we don't, they don't.
It is the only way.
No.
.
The CDC has continued to perform seroprevalence studies (https://covid.cdc.gov/covid-data-tracker/?CDC_AA_refVal=https%3A%2F%2Fwww.cdc.gov%2Fcoronav irus%2F2019-ncov%2Fcases-updates%2Fcommercial-labs-interactive-serology-dashboard.html#serology-surveillance) through mid-July for a few areas. Here is the mortality rate (deaths per total cases, not deaths per confirmed cases) that that data implies for three localities listed in alphabetical order:
0.7% - Minnesota
1.3% - Missouri
1.0% - New York City metro area
While NYC had a much earlier outbreak (87% of its deaths on the CDC measurement in July by the CDC measurement in May) than Missouri (73%) and Minnesota (67%), none of them saw anything other than 1% to a statistically significant degree. It's also important to note that these numbers aren't impacted by the inarguable fact that deaths lag behind cases: at these points in their respective outbreaks the changes in the numerator just aren't big enough compared to the denominator. With that said, it's useful to get a bead on how the numbers change if we were looking during a big increase, so if we look at the April 1st numbers for the NYC metro we get a reading of 0.3% for simultaneous deaths, 0.9% for deaths a week later, and 2.2% for deaths two weeks later. Since we know the NYC metro area was really 1.0%, if we're looking at another locality during a big increase we should compare deaths about 9 days later ("about" since some places don't release data on weekends).
.
Okay, but obviously we'd really like to compare to a Florida or Texas, a really big state with a much later peak. Unfortunately the previous study was only on very specific areas that didn't include them. (There is one on South Florida but it was discontinued in May, so it's not helpful for this.) The CDC does have a broader set of data (https://covid.cdc.gov/covid-data-tracker/?CDC_AA_refVal=https%3A%2F%2Fwww.cdc.gov%2Fcoronav irus%2F2019-ncov%2Fcases-updates%2Fcommercial-labs-interactive-serology-dashboard.html#national-lab) which necessarily means shallower as well; for example, we only have one measurement for Texas on August 15th that gives a mortality rate of 0.5% to 1.0%. Similarly, our only measurement for Florida works out to 0.8% to 1.8%. Again I want to stress that this data is a lot less reliable than the above so the error bars are even bigger than those listed here, but when even the Gaussian error bars are sufficient for the purposes of making it impossible to prove a sub 1% mortality rate it doesn't matter.
.
One of humanity's greatest strengths is the belief that we can control our environments.
Sometimes, that belief is wrong.
1% of New Yorkers who contracted coronavirus, tested or otherwise, died.
1% of Texans who contracted coronavirus, tested or otherwise, died.
1% of Floridians who contracted coronavirus, tested or otherwise, died.
When their peaks happened didn't change this.
The demographics getting infected didn't change this.
Their nursing home policies didn't change this.
New treatments (real and imagined) didn't change this.
The only thing that has ever changed the number of people dying is the number of people getting infected - all people.
When we bring infections down, deaths go down too.
When we don't, they don't.
It is the only way.