We developed an algorithm to identify COVID-19 onset using data collected by a commercially available wearable device. The resultant algorithm had high sensitivity (82%), with moderate specificity (63%). In developing this algorithm, we placed greater emphasis on sensitivity than on specificity, as our goal was to develop an algorithm that could effectively identify individuals who should obtain laboratory-based diagnostic testing. In this context, lower sensitivity would result in fewer people with potential COVID-19 receiving diagnostic testing, which poses a more serious problem in most screening settings than lower specificity, which would result in more people without COVID-19 receiving diagnostic testing.
You need to be very interested in the subject to read all of this, but it is fascinating how various metrics can be used in place of a focussed test to ascertain likely infection.