N_, August 15, 11:29, #57654:
– “Only a small number of tests (say a few thousand) are needed to estimate incidence (“total infected so far” per population), assuming the test works reasonably well.”
(a) I would expect this to be true for an infection that is near equilibrium in a population, ie. it has been around long enough to be distributed throughout the population – a proportion having sufficient immunity, a proportion having insufficient immunity but not a recent chance to be infected, and a proportion with the infection; all three groups distributed among each other.
(b) I would expect it to be untrue for a novel infection (1) to which the population begins with no immunity and (2) which spreads very fast.
A “few thousand tests” is a sample of the population; hopefully a random sample. In case (a), a random sample should yield a result representative of the randomly distributed infection, but in case (b) infection will be very clustered around the initial seeding infections, and by definition, the locations of these are unknown.
The fast spread is also significant, as the timescale for performing the tests and collating the results is similar to or exceeds the doubling time of the infection; the results will be out of date by the time they are ready for publication.
– “With the new SARS strain this happened through the biowar defence network and was done covertly. All sorts of tests are done covertly all the time.”
N_, do you have insider knowledge of this or are you guessing? To me this seems an unlikely approach, since the type of pathogen to test for would be unknown. As a first guess, if I were designing such a biowar detection/monitoring system, I would suggest monitoring of health databases to detect any sudden surges in a broad variety of symptoms, and to produce distribution maps of them to indicate clusters.