You don't have to dedicate a huge number of tests to this approach. Pull a random sample of 10-20K individuals where you test for current infections (identifying symptomatic and asymptomatic persons) as well as anti-bodies if that test is available yet (identifying persons who had it already but recovered) and have each complete a survey of their daily routines over the preceeding week and plans for the upcoming week. Re-test and survey everyone in the sample a week later. That testing would drastically increase our knowledge about the total infection rates as well as hospitalizations, mortality, etc., compared to using the "confirmed cases" numbers that are riddled with selection bias. Right now, our sole focus in testing is to react to the virus that's already infected someone for treatment purposes. That's not a bad thing, but it means we will continue shooting in the dark in terms of our policy response on this virus for quite a while.