Second, our results may have been affected by sampling bias resulting from nonrandom selection, artificially strengthening the associations found in this study, particularly for women using only telephone referral or pimp/brothel venues as they would be theoretically less accessible to outreach workers.