What's actual versus what has merely been found? This is why we need to drill down before proclaiming bad headlines, or believing them.
Let's start the analysis with this quote: "The median case fatality rate — the number of deaths divided by the number of cases — at facilities with reliable data is 17 percent, significantly higher than the 5 percent case fatality rate nationwide."
What fatality rate? Well, dividing only the counted deaths by only the counted cases. The actual fatality rate across the board, everyone included, is about 0.85 to 1.01%, not 5%. The actual number of cases in the USA is estimated to be over 12 million (as of early last week's preprint, which used data from before that time), not the 2.5 million cases we see counted as of today. So, we easily explain the discrepancy between the 5% that the NYT uses, and the 1% that epidemiological science is now going with.
The better estimate of actual deaths in nursing homes is the 25% that came out last week, not the 43% that using only counts comes up with. Something we know about nursing homes, is that when they are not creating a scandal by trying to hide deaths in their facilities, they are better at counting the dead by covid-19 than the general population. Any estimate from a given nursing home count will be far better and higher than from the general population.
Granted that we can count the nursing homes that mishandled the spread of the virus, and set them apart from the more responsible facilities, just as we can look at cruise ships and such to point to the situations of superspreading, when dozens of people have congregated indoors. The lesson there is do not congregate like that indoors, like a nursing home or a cruise ship.
To solidy my logic here, the NYT article itself gives the accounting measure. Here's the quote: "In 24 states, the number of residents and workers who have died accounts for either half or more than half of all deaths from the virus.//Infected people linked to nursing homes also die at a higher rate than the general population."
There goes the 43%. People who work at nursing homes are not constitutionally prone to dying from the virus -- unless someone can come up with a really revealing explanation as to why people who work at nursing homes should die at a higher rate. Instead, nursing home associates represent a very general population of workers, young, old, in between, with and without preexisting conditions, etc. If your data shows that people linked to nursing homes die at a higher rate, then both your methods and conclusions are bad, or you just need to look with a cleaner lens for reality.
The first thing that made me relate it to the NYT's article is the use of the phrase "naïve case counts", what the NYT is guilty of -- imagine that, being naïve, versus I suppose scientific.
There is also a very important chart in the PDF, that shows deaths from covid-19 versus excess deaths. This can easily lead one to believe that the death counts are naïve as well. Here is one quote: "Given that deaths caused by COVID-19 may be misattributed to other causes, three different measures of death counts are used: a count of deaths attributed to COVID-19; an estimate of excess deaths attributed to select respiratory illnesses; and an estimate of excess allcause deaths." Then, CTL-F for "Table 1". Table 2 is close below it.
In the "Discussion" header that follows, comes this statement: "While some of these outcome measures may themselves be vulnerable to underreporting, estimates of excess all-cause deaths should not be affected by under-detection of COVID-19 cases, although they may be affected by behavioral changes (such as avoiding emergency room visits) that do not track perfectly with outbreak severity. The fact that this excess death measure tracks well with the prior week’s case counts adjusted for testing coverage in the manner described here supports the case for the validity of this coverage-adjusted case indicator."
In order for Nathan Favero to have a legitimate research paper, he had to get beyond being "naïve" about both death counts and case counts, something the NYT did not do. He had to be scientific. It's a matter for him, of pubic policy and caring about quantitative methodology.
Here in Pennsylvania, 6,603 people have died from Covid-19. About 5,000 of them died in nursing facilities — where the most vulnerable segment of the population would be — and where our brilliant governor ordered them to be taken. Do you see the connection? As far as I am concerned, Gov. Wolf is guilty of at least involuntary manslaughter, and should be prosecuted. Throughout this whole episode the "experts" and their apologists have only confirmed something Richard Feynman — a real scientist — said: "Science is the belief that the experts are wrong.
In that sense, the ignorance of experts is what fires science. It seems a belief of Feynman that he himself must be wrong, that a new Feynman will come along, and then another. The only other reading would be that some who play media politics prop themselves up, as if they are experts to the general public on subjects they are not; that science, in this case epidemiology and statistical analysis, are there to joust them off their high horses. I prefer the first reading, because it relates to the drive of science and all that we will never know. The latter fits too.
What's actual versus what has merely been found? This is why we need to drill down before proclaiming bad headlines, or believing them.
ReplyDeleteLet's start the analysis with this quote: "The median case fatality rate — the number of deaths divided by the number of cases — at facilities with reliable data is 17 percent, significantly higher than the 5 percent case fatality rate nationwide."
What fatality rate? Well, dividing only the counted deaths by only the counted cases. The actual fatality rate across the board, everyone included, is about 0.85 to 1.01%, not 5%. The actual number of cases in the USA is estimated to be over 12 million (as of early last week's preprint, which used data from before that time), not the 2.5 million cases we see counted as of today. So, we easily explain the discrepancy between the 5% that the NYT uses, and the 1% that epidemiological science is now going with.
The better estimate of actual deaths in nursing homes is the 25% that came out last week, not the 43% that using only counts comes up with. Something we know about nursing homes, is that when they are not creating a scandal by trying to hide deaths in their facilities, they are better at counting the dead by covid-19 than the general population. Any estimate from a given nursing home count will be far better and higher than from the general population.
Granted that we can count the nursing homes that mishandled the spread of the virus, and set them apart from the more responsible facilities, just as we can look at cruise ships and such to point to the situations of superspreading, when dozens of people have congregated indoors. The lesson there is do not congregate like that indoors, like a nursing home or a cruise ship.
To solidy my logic here, the NYT article itself gives the accounting measure. Here's the quote: "In 24 states, the number of residents and workers who have died accounts for either half or more than half of all deaths from the virus.//Infected people linked to nursing homes also die at a higher rate than the general population."
There goes the 43%. People who work at nursing homes are not constitutionally prone to dying from the virus -- unless someone can come up with a really revealing explanation as to why people who work at nursing homes should die at a higher rate. Instead, nursing home associates represent a very general population of workers, young, old, in between, with and without preexisting conditions, etc. If your data shows that people linked to nursing homes die at a higher rate, then both your methods and conclusions are bad, or you just need to look with a cleaner lens for reality.
As I do at least daily, I just went through the pre-prints for today, and found this abstract: Adjusting confirmed COVID-19 case counts for testing volume. The entire PDF is accessible for download near the upper right of the abstract.
ReplyDeleteThe first thing that made me relate it to the NYT's article is the use of the phrase "naïve case counts", what the NYT is guilty of -- imagine that, being naïve, versus I suppose scientific.
There is also a very important chart in the PDF, that shows deaths from covid-19 versus excess deaths. This can easily lead one to believe that the death counts are naïve as well. Here is one quote: "Given that deaths caused by COVID-19 may be misattributed to other causes, three different measures of death counts are used: a count of deaths attributed to COVID-19; an estimate of excess deaths attributed to select respiratory illnesses; and an estimate of excess allcause deaths." Then, CTL-F for "Table 1". Table 2 is close below it.
In the "Discussion" header that follows, comes this statement: "While some of these outcome measures may themselves be vulnerable to underreporting, estimates of excess all-cause deaths should not be affected by under-detection of COVID-19 cases, although they may be affected by behavioral changes (such as avoiding emergency room visits) that do not track perfectly with outbreak severity. The fact that this excess death measure tracks well with the prior week’s case counts adjusted for testing coverage in the manner described here supports the case for the validity of this coverage-adjusted case indicator."
In order for Nathan Favero to have a legitimate research paper, he had to get beyond being "naïve" about both death counts and case counts, something the NYT did not do. He had to be scientific. It's a matter for him, of pubic policy and caring about quantitative methodology.
Here in Pennsylvania, 6,603 people have died from Covid-19. About 5,000 of them died in nursing facilities — where the most vulnerable segment of the population would be — and where our brilliant governor ordered them to be taken. Do you see the connection? As far as I am concerned, Gov. Wolf is guilty of at least involuntary manslaughter, and should be prosecuted.
ReplyDeleteThroughout this whole episode the "experts" and their apologists have only confirmed something Richard Feynman — a real scientist — said: "Science is the belief that the experts are wrong.
In that sense, the ignorance of experts is what fires science. It seems a belief of Feynman that he himself must be wrong, that a new Feynman will come along, and then another. The only other reading would be that some who play media politics prop themselves up, as if they are experts to the general public on subjects they are not; that science, in this case epidemiology and statistical analysis, are there to joust them off their high horses. I prefer the first reading, because it relates to the drive of science and all that we will never know. The latter fits too.
ReplyDelete