Re: COVID-19 – Sensitivity, specificity, and prevalence in testing is key information.
This commentator has yet to acquire Italian as another language, so the insights that follow might have already been considered in the original piece, if so, apologies in advance. Risk of repetition aside:
The problem with blanket testing, especially of an isolated village of 3,000 people, is that even with very good tests the possibility of false positives remain high. Assuming for the moment that nobody has the virus at issue in the village (0 in 3,000 prevalence) and the test is 90% sensitive and 90% specific. Then of those three thousand, 2,700 will come back accurately as negative, but 300 would come back as positive despite the patient not having the virus. If we alter the sensitivity and specificity to 95%, 2,850 will come back accurately as negative, but 150 will come back as falsely positive. If we alter this yet again, making 98% the sensitivity and specificity, 2,940 will be accurately identified, but 60 will be false positives. This is simply because the test is imperfect. It is unclear to this commentator the actual sensitivity and specificity of the tests involved, nor is the prevalence in the village known.
Of course, if one were to serially repeat these tests on this population, they would not come out with mathematical precision each time. It would vary and could have more false positives on one measurement, but less on the next. This problem affects all low-prevalence disease testing. One must be mindful of this when it comes to blanket testing in isolated areas, or anywhere for that matter.
So the possibility exists that those tested who were asymptomatic, might not have had the virus at all. And that the difference on the repeat measurement was just what happens in the World of Small Numbers.
Irrespective of this point, the bottom-line recommendations of isolation and the full suite of measures remain sound.
Competing interests: No competing interests