The effect of seed origin and interview site location on the geographical dispersion of respondent-driven samples in multiple cities
*Cyprian Wejnert, Centers for Disease Control and Prevention
Keywords: Respondent-driven sampling, HIV, sampling bias
Respondent-driven sampling (RDS) has become increasingly popular for sampling and analyzing hard-to-reach populations. RDS analysis relies on several key assumptions including: 1) recruitment is a random selection of the recruiter’s peers and 2) the sample becomes independent of the initial participants, or seeds, from which it originated. The location of interview field site and geographical dispersion between seeds and recruits may impact these assumptions, but has not been explored. The Centers for Disease Control and Prevention’s National HIV Behavioral Surveillance System (NHBS) employed RDS to conduct interviews with 450 heterosexuals at increased risk for HIV in 21 U.S. cities (NHBS-HET) in 2010. A common protocol is used with flexibility to accommodate local needs, including the selection of seeds and number and location of interview sites. We explored the effect of seed origin, interview site location, and number of interview sites on the geographical dispersion of NHBS-HET RDS samples in two participating cities. Our results underscore the role for formative research in planning studies using RDS. Identifying factors associated with geographic dispersion of respondents may increase the efficiency and efficacy of RDS survey methods.