what we are using for policy decisions.

Number of hospitalizations would be a meaningful metric. But when the only people being tested are those who are sick or at high risk (and many people who are sick can't get tested unless they tick enough of the symptom boxes), test results tell you almost nothing about whether the spread is getting better or worse. We would either need extremely widespread testing (like 70%+ of the population), or true random sample testing for test results to tell us anything meaninful.