“…For these reasons, the study was not based on a random sample, but rather on a sample of women from each class who were willing to be interviewed. The sampling methods used are compatible with the two-stage cluster (Tran, Detels, Long, Phung, & Lan, 2005b), modified snowball (Todd et al, 2007), refined chain referral (Mills, Saidel, Magnani, & Brown, 2004), and respondent-driven (Johnston et al, 2006) strategies currently recommended for studying hard-toreach populations. Nonetheless, the characteristics of study participants may have differed from those of women who were not invited to complete the interview and those who declined the invitation.…”