“…For other studies, healthy volunteer bias 19 may underestimate seroprevalence, attracting people with symptoms 26 may overestimate seroprevalence, and studies of employees, 14 , 21 , 25 , 32 , 66 grocery store clients 23 or patient cohorts 11 , 14 , 27 – 30 , 36 , 38 , 40 , 50 , 51 , 56 , 59 , 62 , 64 , 67 risk sampling bias in an unpredictable direction.…”