Since I am going to spend the day working on something else, let's toss a hand grenade into the Twitter fox hole and then walk away. The number of actual cases of corona virus are likely underestimated by at least a factor 3. This has all kinds of implications but let's 1/n
Keep our focus narrow. First, doctors around the world are discovering large amounts of false negative cases. Studies in Chinese hospitals with multiple marker confirmation report false negative rates from confirmed patients of 30-60%. Put another way, assume we know 2/n
these 10 people have a disease (in this case corona). Our test is going to say that only 5 of them have corona. It is worth noting these studies are being done in hospitals and typically for the purpose of understanding test problems so these are already highly symptomatic 3/n
People that have maybe already tested positive using the standard test with additional testing likely such as CT scans on the lungs or stool tests. In other words, the low hanging fruit is testing negative roughly 50% of the time. Second, this is important mathematically 4/n
Because it skews our understanding of the severity. Assume 100 people have the disease and 10 people die from the disease. Our death rate is 10%. With me so far? However, our testing only finds 50 people have the disease. That gives us a death of 20%. But wait there's more 5/n
Third, these studies were done on cases that were recognized as clinically confirmed cases. One of the primary suspicions about false positive tests is that the viral load (number of virus in the sample) is too low to register positive. Who would that describe? 6/n
Asymptomatic and low severity patients. Put another way, let's assume you just have a runny nose and get tested for corona, doctors think you're more likely to register negative because you do not have a severe case. But wait there is more. Fourth, like any virus there are 7/n
Very likely people who have the virus and registered no reaction to it or it was so mild the developed antibodies to it. These people never even get tested. In summation, this gives us multiple groups of people that have been exposed or should register but do not. 8/n
If we expand the pool even just within the potential range of false positives and add just a little bit for groups 3 and 4 here, this takes out example of a death rate from 20% (using under estimation factor of 3) from 20% to 6.67%. In other words, cute the death rate by 65% 9/n
It bears worth stressing, these numbers come from existing studies of false negative rates about the corona test. The three factor used here I think is pretty conservative in that I could use a much larger number say 4-6 and remain on very solid estimation ground. 10/n
That would reduce the death rate even more. Well, I got other stuff to blow up today. Have fun and try not to resort to cannibalism while I'm gone.
