That won't work. There is a high false negative rate with those tests, and it's likely that false negatives are more common in those who have only mild infections. A study of patients with only positive tests would be a skewed sample.