Are we flattening the curve?
I read this article. And it makes some good points. But, also misses the facts.
It claims that the restrictions to getting a test could be hampering the rates and providing a false sense of security.
And while some things claimed are absolutely true. Like the true rate of infection being higher than reported and that restrictions are limiting who gets tested. That doesn't mean that testing is driving the number of detections down.
In fact, I would argue that the reverse is actually happening. The way testing is being done is actually inflating the reported rate if anything, not the other way around.
While testing rates may not be ideal, across the board they are increasing. The number of tests taken and the numbers processed per day are both on the rise. This alone combats the notion that testing would be getting worse. Simply having a higher output should drive up number of detections.
On top of that, the restrictions in place, are designed to limit the testing to those more likely to come into contact with it (essential service workers) or those already presenting symptoms. This, will also drive up the number of detections.
To understand this, let's say that around 1% of the population is infected and we can test 1000 people per day. If the sampling is random, we'll expect around 10 people to test positive from that day. Which... hey would be an accurate 1% detection rate.
But, since our population is greater than 1000, if we restrict it to groups that are more likely to test positive. Let's say 200% more likely. Then we'll STILL test only 1000 people a day. But we would expect to see 20 positive tests instead of 10. Guess what? That is 2%. It is HIGHER than 1%. Restrictions drove UP the number.
Furthermore, while the data is admittedly imperfect, it is biased in a pessimistic way. This makes it more likely that optimistic trends observed (over time) reflect true trends at large, while making pessimistic trends less likely to be true.
And that is actually a VERY good bias to have for what is going on.
If you had randomized testing which is incomplete, then it would be much harder to spot the biases (because, whether we know it or not, just about any incomplete data set contains biases). In turn this makes it harder to interpret trends.
When you KNOW you can't get complete enough data, what you can do is try to control your biases to emphasize certain kinds of trends.
And, whether intentional or not. That is what is going on with COVID-19 testing. We are biasing the data towards finding positive test results. This is being done (perhaps unintentionally) by increasing testing capacity. And as well by controlling who gets tested to higher risk groups.
This is a good thing, because it makes it MORE likely that we trend toward more positive daily cases. Which will drive home the message that we need to practice social distancing and self-isolation. The lag time on this virus really needs people to be scared enough to comply with the emergency procedures being put in place.
It also means that by the time results start looking better with in any consistency that we're more than likely to be well past the point when this truly started happening, and also less likely that it is a statistical anomaly. Which, again, is VERY important to reduce the odds of a relapse or second wave.
We want this as much under control as possible before restrictions are lifted.
I absolutely agree that more testing would be better. But, until capacity is greater than demand, restricting testing is actually better.
It claims that the restrictions to getting a test could be hampering the rates and providing a false sense of security.
And while some things claimed are absolutely true. Like the true rate of infection being higher than reported and that restrictions are limiting who gets tested. That doesn't mean that testing is driving the number of detections down.
In fact, I would argue that the reverse is actually happening. The way testing is being done is actually inflating the reported rate if anything, not the other way around.
While testing rates may not be ideal, across the board they are increasing. The number of tests taken and the numbers processed per day are both on the rise. This alone combats the notion that testing would be getting worse. Simply having a higher output should drive up number of detections.
On top of that, the restrictions in place, are designed to limit the testing to those more likely to come into contact with it (essential service workers) or those already presenting symptoms. This, will also drive up the number of detections.
To understand this, let's say that around 1% of the population is infected and we can test 1000 people per day. If the sampling is random, we'll expect around 10 people to test positive from that day. Which... hey would be an accurate 1% detection rate.
But, since our population is greater than 1000, if we restrict it to groups that are more likely to test positive. Let's say 200% more likely. Then we'll STILL test only 1000 people a day. But we would expect to see 20 positive tests instead of 10. Guess what? That is 2%. It is HIGHER than 1%. Restrictions drove UP the number.
Furthermore, while the data is admittedly imperfect, it is biased in a pessimistic way. This makes it more likely that optimistic trends observed (over time) reflect true trends at large, while making pessimistic trends less likely to be true.
And that is actually a VERY good bias to have for what is going on.
If you had randomized testing which is incomplete, then it would be much harder to spot the biases (because, whether we know it or not, just about any incomplete data set contains biases). In turn this makes it harder to interpret trends.
When you KNOW you can't get complete enough data, what you can do is try to control your biases to emphasize certain kinds of trends.
And, whether intentional or not. That is what is going on with COVID-19 testing. We are biasing the data towards finding positive test results. This is being done (perhaps unintentionally) by increasing testing capacity. And as well by controlling who gets tested to higher risk groups.
This is a good thing, because it makes it MORE likely that we trend toward more positive daily cases. Which will drive home the message that we need to practice social distancing and self-isolation. The lag time on this virus really needs people to be scared enough to comply with the emergency procedures being put in place.
It also means that by the time results start looking better with in any consistency that we're more than likely to be well past the point when this truly started happening, and also less likely that it is a statistical anomaly. Which, again, is VERY important to reduce the odds of a relapse or second wave.
We want this as much under control as possible before restrictions are lifted.
I absolutely agree that more testing would be better. But, until capacity is greater than demand, restricting testing is actually better.
Comments
Post a Comment