The Corona virus experiment enters the next phase: Easing of restrictions. Why? Not sure.
I think we are confused as ever about this disease. Even for a scientist it is difficult to comprehend what’s going on.
Yesterday there were 30,000 new COVID-19 cases in the US and over 1800 humans died from the disease in one day. Why would we ease restrictions? Maybe because we are bored at home, need money, or what?
We have now tested about 1% of the population in the US for the presence of virus. Most of the tests were given to people that were ill or had reasons to believe they were in contact with infected people. So that’s a small fraction of the whole population and a highly biased selection.
No problem. We need to test people at risk.
However, at this point the number of tests is too small to make any claims about the prevalence of the disease.
Why is this important?
Well, some officials think about reopening the country. One of the conditions discussed is a “certificate of immunity”. You get tested for COVID-19 antibodies. Your test comes back positive. You can go to work.
I think there is an issue that, which has to do with specificity, selectivity, and prevalence.
First, none of the antibody tests are perfect, meaning they are not 100% specific nor selective. That’s not unusual.
I thought a test that produces in 95% of truly positive cases a positive result is pretty good. And if the same test has a 5% false positive rate that should make a great test, right? Well, it depends. It depends on the prevalence of the disease. If only a small fraction of the people carries or carried the disease than the predictive value of a positive test is rather low. This is all nicely illustrated here.
Since we don’ know the prevalence of the virus in the general population and the unknown predictive power of a positive antibody, we should not send people back to work, unless we want to tolerate 2000 or more individuals dying every day in this country for the foreseeable future.
Please correct me if I am wrong.
On with the experiment.