Assessing respondent-driven sampling in the estimation of the prevalence of Sexually Transmitted Infections (STIs) in populations structured in complex networks

Abstract

When a sampling frame for a given population cannot be defined, either because it requires expensive/time-consuming procedures or because it targets a stigmatized or illegal behavior that may compromise the identification of potential interviewees, traditional sampling methods cannot be used. Examples include "hidden populations" of special relevance for public health, such as men who have sex with men (MSM), sex workers and drug users. Since the late 1990s, a network-based method, called Respondent-Driven Sampling (RDS) has been used to assess such "hidden populations".This paper simulates data from hidden populations, in order to assess the performance of prevalence estimators in different scenarios built after different combinations of social network structures and disease spreading patterns. The simulation models were parameterized using empirical data from a previous RDS study conducted on Brazilian MSM. Overall, RDS performed well, showing it is a valid strategy to assess hidden populations. However, the proper analysis of underlying network structures and patterns of disease spread should be emphasized as a source of potential estimate biases. © 2011 Springer-Verlag.