Most people show symptoms of COVID-19 at around 5 days and 97% show before the end of 12. Consequently, you would expect the number of new cases showing symptoms to continue to spike after a lock down for 5 days and then to drop off for the next seven. The numbers wouldn't drop to 0 because there would be new cases that didn't pre-exist the lock down which would start to show up two days after the first day and since we are in the upward phase of even the flattened curve would rise even as the lock down proceeds. As of the second day onward there would be more cases than those merely including the previously infected and the difference between the two numbers would grow, but after the 5th day both the number of previously infected newly exhibiting symptoms and the entire number of those newly exhibiting symptoms should drop.
Let's see how that played out in Virginia. Remember, that the 17th of March was when the governor closed down restaurants, theaters, and gyms. More draconian and questionable actions would come later, but the 18th was the date when things really began because the closure of restaurants was the biggest chunk of the social distancing requirements. Let's see how things played out:
Orange is the line which should have been the number of people infected prior to the lock down on the increase for 5 days. As is obvious from above, the number of newly symptomatic infected was on an upward trajectory prior to the governor's lock down. This should have continued through 5 days as more people were revealed to be infected. Then, the numbers of newly symptomatic that had it prior to the lock down drops as is represented by the yellow line until the end of day 12 when almost all of them had developed symptoms.
The green line starting two days in represents those who were not infected when the lock down started, but became so thereafter, and the red is everyone who is displaying new symptoms whether they became infected before the lock down or after.
It's a crude diagram, and I don't claim it to be 100% accurate (I have extrapolation, not actual numbers). Still, it gives you some food for thought because nothing in there is going the way it's supposed to. Almost immediately upon the lock down there was a significant drop in numbers followed by a spike when the time came for a drop. Huh?
I suspect that this may have more to do with testing availability than it does with the actual spread of the disease. Perhaps some places had a lack of tests for a period of time and then when tests became available there was a surge of people taking the test? It's the best explanation I can come up with. Yes, I understand that statistical dispersion or variability (or whatever the correct term is) is out there. I'm not now nor never will be Bill James, but I get that numbers become more accurate with a larger sample size to find medians and averages and a particular day is generally meaningless. However, that's the sample size we have and it does seem to show general trends and they don't follow the path they should.
On the other hand, by eyeball the average from 17 March thru 25 March appears to be about 100. Could the level of spread have plateaued? It's still not what should have happened, but it's an interesting possibility.
In reading various articles over at SSRN, I've run across the assertion that there is no way to measure COVID-19 except by the number of deaths. The rationale goes something along the lines of you can't trust the number testing positive because that is reliant upon the number of tests available and how they are being utilized - whether to chase down infected or placate some guy who shows up at the hospital with the sniffles. You can't rely on the number of hospitalized because different places will have different standards for hospitalization. If you are in a major metropolitan area with 1,000 people hospitalized with the disease the standard for admittance is going to be higher than it is if you are the only person with the disease in a three-hundred mile radius of your town. This leaves us with only one dependable metric: how many have died?
I think there are probably some flaws in that. On the other hand, as distasteful as it is, when I look at the chart above I think they might have a strong argument.
Let's see how that played out in Virginia. Remember, that the 17th of March was when the governor closed down restaurants, theaters, and gyms. More draconian and questionable actions would come later, but the 18th was the date when things really began because the closure of restaurants was the biggest chunk of the social distancing requirements. Let's see how things played out:
Orange is the line which should have been the number of people infected prior to the lock down on the increase for 5 days. As is obvious from above, the number of newly symptomatic infected was on an upward trajectory prior to the governor's lock down. This should have continued through 5 days as more people were revealed to be infected. Then, the numbers of newly symptomatic that had it prior to the lock down drops as is represented by the yellow line until the end of day 12 when almost all of them had developed symptoms.
The green line starting two days in represents those who were not infected when the lock down started, but became so thereafter, and the red is everyone who is displaying new symptoms whether they became infected before the lock down or after.
It's a crude diagram, and I don't claim it to be 100% accurate (I have extrapolation, not actual numbers). Still, it gives you some food for thought because nothing in there is going the way it's supposed to. Almost immediately upon the lock down there was a significant drop in numbers followed by a spike when the time came for a drop. Huh?
I suspect that this may have more to do with testing availability than it does with the actual spread of the disease. Perhaps some places had a lack of tests for a period of time and then when tests became available there was a surge of people taking the test? It's the best explanation I can come up with. Yes, I understand that statistical dispersion or variability (or whatever the correct term is) is out there. I'm not now nor never will be Bill James, but I get that numbers become more accurate with a larger sample size to find medians and averages and a particular day is generally meaningless. However, that's the sample size we have and it does seem to show general trends and they don't follow the path they should.
On the other hand, by eyeball the average from 17 March thru 25 March appears to be about 100. Could the level of spread have plateaued? It's still not what should have happened, but it's an interesting possibility.
In reading various articles over at SSRN, I've run across the assertion that there is no way to measure COVID-19 except by the number of deaths. The rationale goes something along the lines of you can't trust the number testing positive because that is reliant upon the number of tests available and how they are being utilized - whether to chase down infected or placate some guy who shows up at the hospital with the sniffles. You can't rely on the number of hospitalized because different places will have different standards for hospitalization. If you are in a major metropolitan area with 1,000 people hospitalized with the disease the standard for admittance is going to be higher than it is if you are the only person with the disease in a three-hundred mile radius of your town. This leaves us with only one dependable metric: how many have died?
I think there are probably some flaws in that. On the other hand, as distasteful as it is, when I look at the chart above I think they might have a strong argument.
No comments:
Post a Comment