Overlooked Threats to Respondent Driven Sampling Estimators: Peer Recruitment Reality, Degree Measures, and Random Selection Assumption

Abstract

Intensive sociometric network data were collected from a typical respondent driven sample (RDS) of 528 people who inject drugs residing in Hartford, Connecticut in 2012-2013. This rich dataset enabled us to analyze a large number of unobserved network nodes and ties for the purpose of assessing common assumptions underlying RDS estimators. Results show that several assumptions central to RDS estimators, such as random selection, enrollment probability proportional to degree, and recruitment occurring over recruiter's network ties, were violated. These problems stem from an overly simplistic conceptualization of peer recruitment processes and dynamics. We found nearly half of participants were recruited via coupon redistribution on the street. Non-uniform patterns occurred in multiple recruitment stages related to both recruiter behavior (choosing and reaching alters, passing coupons, etc.) and recruit behavior (accepting/rejecting coupons, failing to enter study, passing coupons to others). Some factors associated with these patterns were also associated with HIV risk.

Multiple strategies to identify HIV‐positive black men who have sex with men and transgender women in New York City: a cross‐sectional analysis of recruitment results

“…incentivized referrals from purposively selected seeds), a rate similar to a Chicago based study of young MSM aged 16 to 20, which achieved a rate of 11.5 enrolments per month only by adding several cohorts of additional seeds . Researchers have pointed to multiple factors not measured in established RDS analytic approaches that can impact on RDS recruitment, including how participants interpret eligibility criteria and apply them to peer groups and the frequency with which peers receiving referral coupons redistribute them within their own networks rather than enrolling in the research . Detailed investigation of how RDS participants understand and operationalize incentivized peer referrals is needed to optimize RDS and will also contribute to our understanding of dynamics that shape network‐driven participation in HIV prevention research .…”

“…The RDS approach also includes analytical methods to produce statistically valid estimations of the targeted population by accounting for non‐random influences on the likelihood of an individual being recruited, primarily the size of the peer network from which participants draw additional recruits and the prevalence within networks of given characteristics . Studies have identified additional sources of non‐random bias that may limit the ability of RDS to produce statistically valid characterizations of the population engaged, including non‐random bias deriving from individual‐level characteristics associated with successfully recruiting members of one's social network ; cultural factors influencing how network peers are selected for referral ; and different ways in which recruits engage in the recruitment process . Several authors have concluded that RDS inference models currently available do not adequately account for these types of bias, calling into question the utility of RDS to construct valid population estimates of hard to reach subpopulations .…”

“…Studies have identified additional sources of non‐random bias that may limit the ability of RDS to produce statistically valid characterizations of the population engaged, including non‐random bias deriving from individual‐level characteristics associated with successfully recruiting members of one's social network ; cultural factors influencing how network peers are selected for referral ; and different ways in which recruits engage in the recruitment process . Several authors have concluded that RDS inference models currently available do not adequately account for these types of bias, calling into question the utility of RDS to construct valid population estimates of hard to reach subpopulations . Nonetheless, as a recruitment method, RDS has been widely used to engage US racial and ethnic minority MSM and TGW in HIV research and testing .…”

Estimation and correction of bias in network simulations based on respondent-driven sampling data

“…After a simulated person was recruited into the sample, (s)he was provided the number of coupons determined by the coupon parameter, but (s)he could choose not to give out all the coupons 36 . The number of coupons a person intended to give out was determined by a discrete uniform distribution between 0 and the number of coupons provided.…”

Risky injection practices and HCV awareness in Chiang Mai Province, Thailand: a respondent-driven sampling study of people who inject drugs

“…While RDS has been widely used in the last two decades and adopted by leading public health organizations [52,53], the quality of estimates derived from these data has been challenged [54,55]. To mitigate such concerns, we monitored potential sources of participation bias by asking participants about their motivation to participate in the study and if they had difficulties giving out coupons when they came to collect their compensation.…”