In reply to jonny taylor:
If you are separating care facilities into 'COVID' and 'non-COVID' sides, then it's probably a good idea to know into which side patients should be sent (false negative rate here is the worry).
In order to identify the asymptomatic, you'd have to keep regularly testing the entire population (less the already positive results). Which we clearly cannot do with the resources we have.
If you assume that prior infection confers immunity (unconfirmed so far, but more normal for such infections), then those found to be previously infected might continue life less restricted, and would need no further testing. Depending, that is, on the false positive rate of the test, and the ability for tests to detect previous, overcome infections (which doesn't seem the case except for more complex antibody tests by the likes of Porton Down), and the social acceptability/public health risk of such a move.
If you want to know the extent of the real problem, and distinguish COVID-19 from other coronavirus or influenza infections, which present similar symptoms, you need a test to distinguish them.
If you're looking to see the extent to which the disease has spread within the population, then carefully-designed randomised testing might be performed (which I think is the case).
Just a few thoughts.
Post edited at 14:13