A Systematic Review of Published Respondent-Driven Sampling Surveys Collecting Behavioral and Biologic Data

Abstract

Acknowledgments:We would like to thank Kate Orroth for conducting the literature search for the STROBE-RDS Guidelines and allowing us to use it for this analysis. Abstract word count: 160Word count: 3060 ABSTRACTReporting key details of respondent-driven sampling (RDS) survey implementation and analysis is essential for assessing the quality of RDS surveys. RDS is both a recruitment and analytic method and, as such, it is important to adequately describe both aspects in publications. We extracted data from peer-reviewed literature published through September, 2013 that reported collected biological specimens using RDS. We identified 151 eligible peer-reviewed articles describing 222 surveys conducted in seven regions throughout the world. Most published surveys reported basic implementation information such as survey city, country, year, population sampled, interview method, and final sample size. However, many surveys did not report essential methodological and analytical information for assessing RDS survey quality, including number of recruitment sites, seeds at start and end, maximum number of waves, and whether data were adjusted for network size. Understanding the quality of data collection and analysis in RDS is useful to effectively plan for public health service delivery and funding priorities.

The 12 city HIV Surveillance Survey among MSM in Brazil 2016 using respondent-driven sampling: a description of methods and RDS diagnostics

“…RDS is a chain-link sampling method that permits participants to recruit their acquaintances and uses a mathematical model to adjust for this method of recruitment [16][17][18] . RDS is used in hundreds of surveys [19][20][21] around the world and was recently identified as a major contribution to social, behavioral and economic research by the U.S. National Academies of Sciences 22 . RESUMO: Introdução: Este artigo detalha os métodos utilizados na segunda Pesquisa Nacional de Vigilância Biológica e Comportamental (BBSS) do HIV, sífilis e hepatite B e C entre os homens que fazem sexo com homens no Brasil.…”

Hard-to-Survey Populations and Respondent-Driven Sampling: Expanding the Political Science Toolbox

“…2 A review of peer-reviewed studies collecting biological specimens counted 222 RDS surveys as of 2013 (Johnston et al 2016). A review of RDS for biological and behavioral surveillance of populations at risk for HIV counted 123 surveys as of 2007 (Malekinejad et al 2008).…”

“…In a review of RDS surveys, the WHO (2013) determined the median number of seeds was eight, and that an average of 1.6 were unsuccessful. 27 The average sample size and number of waves in a review of RDS biological surveys were 325 and 9, respectively (Johnston et al 2016). 28 The sample was smaller than the target size of 313, which was estimated as n ¼ pð1ÀpÞ varðyÞ DE where p is the expected proportion, var(y)is the variance, and DE is design effect, calculated with a conservative proportion estimate of 0.50, standard error of 0.04, and a design effect of 2.…”

Reduced Bias for Respondent-Driven Sampling: Accounting for Non-Uniform Edge Sampling Probabilities in People Who Inject Drugs in Mauritius

“…RDS has been widely adopted to estimate the prevalence of disease or risk behaviours within high risk hard-to-reach human populations, including PWID, sex workers and men who have sex with men. It has been used in hundreds of studies around the world (Johnston et al, 2008(Johnston et al, , 2016Montealegre et al, 2013), for surveys of biological behavioural surveillance funded by the Global Fund to monitor HIV prevalence, to assess risk and programme coverage and to measure trends over time (Lansky et al, 2007). Despite its wide use in settings of public health importance, the statistical properties and optimal inferential strategies for data resulting from RDS still require much additional study.…”

“…Several RDS prevalence estimators have been proposed and implemented (Heckathorn, 1997(Heckathorn, , 2002Salganik and Heckathorn, 2004;Volz and Heckathorn, 2008;Gile, 2011;Gile and Handcock, 2015); however, it has been demonstrated that no one estimator is superior in all cases (Tomas and Gile, 2011). We briefly review the three most commonly implemented estimators of μ: the Volz-Heckathorn (VH) estimator, the successive sampling (SS) estimator, and the Salganik-Heckathorn (SH) estimator, and their relationships to the estimator proposed (Johnston et al, 2016). The VH and SS estimators estimate the probability of observing individuals based on their degree and use this probability to perform inverse probability weighting.…”