As far as applying the antibody test widely to people who have not had symptoms and no known exposures (and especially who have also practiced social distancing well), the positive predictive value of the test (true positives divided by all positives) will be even worse among than he says because the majority of people who have antibodies will have diagnosed infection or high-risk exposures, thus the prevalence of infection and antibodies among the people with no symptoms and no exposures will be even lower than he assumes in his example). But as he explains a bit later in the thread, the antibody test will be useful in contexts in which the pre-test probability of disease is high enough. Those situations would probably include re-checking people who have symptoms consistent with Covid-19 but a negative antigen test, or people with high-risk exposures but no symptoms such as front-line health care workers (especially with poor PPE), spouses/partners of people with antigen-confirmed infections.
Of course, this Bayesian analysis applies to all medical tests. Patients who push doctors to order various tests when the prior probability is very low, and some doctors who order such tests, may not realize a positive test would probably be a false positive.
Of course, this Bayesian analysis applies to all medical tests. Patients who push doctors to order various tests when the prior probability is very low, and some doctors who order such tests, may not realize a positive test would probably be a false positive.