An eye opener for me - test for antibodies. I don’t have COVID-19 antibodies in my blood. I was very sick starting March 10th, but testing was not available yet in my area. Recovery was very long and symptoms suggested COVID-19. I guess it was something else what gave me 105.5 F temp followed by terrible dry cough for weeks.
There is a very common statistical misconception that is made in circumstances such as this. I did it in a post a couple of months ago after I got tested for antibody. The point is that with a single test, it is more likely than not that the test is wrong if a positive result is obtained. Conversely, the odds that a false negative is incorrect are quite small.
Bear with me through some arithmetic. Let's assume testing of 100,000 people in a population where 1% of the people have been exposed and have the antibody. The test I had done (at the University of Washington virology lab) is quite sensitive and skews towards false positives rather than false negatives. So for my case assume 5% false positives and 1% false negatives. (Those are pretty good numbers, by the way.)
So 1% infected out of 100,000 people total means 1000 people have the antibody and should test positive. With a 1% false negative rate, those 1000 people generate 10 false negatives and 990 true positives.
Then there are the 99,000 people who don't have the antibody and should test negative. With 5% false positives, those people generate 4950 false positives and 94,050 true negatives. Below is table summarizing the outcomes. As indicated, the odds lie greatly against an inaccurate false negative. Whereas with a positive result, the odds are high that it is a false positive. The underlying conclusion is that one test really doesn't mean that much, particularly if the test is positive. You really need two tests that agree to be able to conclude with reasonable certainty.
The table above simply draws from an entire pool of people, whereas testing more likely to skew towards people who have had symptoms (and thus are more likely to have actually been infected). So, let's assume that limiting the population to only those who have exhibited one or more symptoms reduces the pool to 1,000 people (1% of the pool), and of those 1,000 people 25% were were infected and should test positive. Continuing to use 5% false positive and 1% false negative error rates, the chart above changes as shown below. But note that even under these conditions, the odds of a false negative are still less than 1%.