“…Screening programmes carried out on a named basis, allowing for self-exclusion, will tend to produce a participation bias that underestimates rates of infection [Hull et al, 1988;Landesman et al, 1987;Carne et al, 1987a,b;Sperling et al, 1988;Connor et al, 1988;Krasinski et al, 1988;Loveday et al, 1989;Centers for Disease Control, 19891. These fallacies of sampling may introduce even more serious error when estimates of HIV seroprevalence are carried out in very low-incidence populations [Bird, 19881. A similar bias may also affect surveys of neonatal blood samples [Hoff et al, (0 1991 …”