ORIGINAL PAPER Respondent-Driven Sampling and the Recruitment of People with Small Injecting Networks Dana Paquette • Joanne Bryant • John de Wit Published online: 27 August 2011 Ó Springer Science+Business Media, LLC 2011 Abstract Respondent-driven sampling (RDS) is a form of chain-referral sampling, similar to snowball sampling, which was developed to reach hidden populations such as people who inject drugs (PWID). RDS is said to reach members of a hidden population that may not be accessible through other sampling methods. However, less attention has been paid as to whether there are segments of the population that are more likely to be missed by RDS. This study examined the ability of RDS to capture people with small injecting networks. A study of PWID, using RDS, was conducted in 2009 in Sydney, Australia. The size of participants’ injecting networks was examined by recruit- ment chain and wave. Participants’ injecting network characteristics were compared to those of participants from a separate pharmacy-based study. A logistic regression analysis was conducted to examine the characteristics independently associated with having small injecting net- works, using the combined RDS and pharmacy-based samples. In comparison with the pharmacy-recruited par- ticipants, RDS participants were almost 80% less likely to have small injecting networks, after adjusting for other variables. RDS participants were also more likely to have their injecting networks form a larger proportion of those in their social networks, and to have acquaintances as part of their injecting networks. Compared to those with larger injecting networks, individuals with small injecting net- works were equally likely to engage in receptive sharing of injecting equipment, but less likely to have had contact with prevention services. These findings suggest that those with small injecting networks are an important group to recruit, and that RDS is less likely to capture these individuals. Keywords Respondent-driven sampling Á Injecting drug users Á Behavioural surveillance Á Hidden populations Introduction Populations who are at high risk for acquiring HIV, such as people who inject drugs (PWID), are often termed ‘‘hard- to-reach’’ [1]. Their behaviours are illicit or illegal, which makes them less willing to participate in research where they are asked to provide details of their injecting and drug use practices. Furthermore, probability sampling methods are impractical because of the lack of an adequate sampling frame [1]. Respondent-driven sampling (RDS) is a rela- tively new method that was developed for sampling hard- to-reach populations [2] and is increasingly being used as a sampling method for HIV behavioural surveillance [3]. Behavioural surveillance consists of ongoing cross-sec- tional surveys of at-risk populations, and is recognized as a key component in the prevention of HIV and other blood borne viruses (BBV) [4]. RDS is a type of chain-referral sampling, similar to snowball sampling, where participants are asked to recruit their peers (who fulfil the eligibility criteria) into the study. The recruits are then asked to recruit their peers and the process continues until the target sample size is reached [2]. RDS incorporates some design modifications with the goal of minimizing the biases associated with snowball sampling. Specifically, RDS participants are limited in the D. Paquette (&) Á J. Bryant Á J. de Wit National Centre in HIV Social Research, University of New South Wales, Webster Building, Sydney, NSW 2052, Australia e-mail: Dana.Paquette@unsw.edu.au J. de Wit Department of Social and Organizational Psychology, Utrecht University, Utrecht, the Netherlands 123 AIDS Behav (2012) 16:890–899 DOI 10.1007/s10461-011-0032-x