Background
Since RDS was first developed in the mid-1990’s, this innovative and powerful methodology has been widely applied in more than 120 HIV/AIDS research, surveillance, and prevention efforts in about 30 countries including large-scale National Health Behavior Surveillance by the US Centers for Disease Control and Prevention. The strong demand for RDS is primarily due to its cost-effectiveness as a recruitment tool and the lack of satisfactory alternative sampling design and inference in hidden populations.  However, the initial RDS statistical models were based on strong but unsupported assumptions regarding peer recruitment processes and the structure of underlying social networks. With its increasing applications to a variety of populations in different contexts, serious skepticism has arisen regarding the validity of RDS’s statistical inference models, given the challenges to meet the underlying assumptions during implementation and recent discovery that population estimations derived from the most widely used model are substantially less accurate than generally acknowledged .  Recently, a small group of researchers have been developing new models that are less sensitive to violations of assumptions or are based on more realistic assumptions . These promising procedures, however, are still based on somewhat idealistic recruitment dynamics and require accurate reporting of social network size and composition. Furthermore, the most striking gap in the RDS literature is the failure to address the complexity of the social networks of high-risk populations and factors affecting peer referral behavior and network information reporting. The network members successfully recruited into the study might not actually be representative of their eligible network members reported on surveys, which will undermine the accuracy of estimations derived from current RDS models.