Respondent-driven Sampling Sample Research Articles

Mobility and HIV vulnerabilities among female sex workers in Guinea-Bissau: findings from an integrated bio-behavioral survey

BackgroundMobility is an important risk determinant for HIV given the potential for intermittent access to HIV services. Mobility may be particularly relevant among female sex workers, (FSW) who have been shown to be at high risk for HIV in settings around the world. Data regarding the role mobility plays in exacerbating HIV risks among FSW across Sub-Saharan Africa remains limited, and data on FSW in Guinea-Bissau is sparse.MethodsFSW in four regions of Guinea-Bissau were recruited with a respondent-driven sampling (RDS) method and participated in an integrated bio-behavioral survey between September 27, 2017 and January 26, 2018. Associations between reported general mobility, mobility to or residence in Bissau, and social and HIV vulnerabilities among FSW in Guinea-Bissau were assessed using multivariable logistic regression models. Population proportions were weighted for RDS sampling, while logistic regression models were not.ResultsSurvey respondents included 323 individuals in Bissau, 45 in Bissorã, 140 in Bafatá, and 59 in Gabu. Statistical analyses demonstrated that mobility to more than one destination was significantly associated with recent sex without a condom (ie, sex without a condom within the last three sex acts) with both clients (aOR: 2.47 (95% CI: 1.08, 5.64)) and non-paying partners (aOR: 5.39 (95% CI: 2.61, 11.15)) compared to non-mobility. However, mobility to one or more locations was also associated with higher odds of receiving HIV prevention information, and mobility to more than one location was associated with participating in programming with HIV-related organizations.ConclusionsThese results suggest that while some prevention services including HIV prevention information reach mobile FSW in Guinea-Bissau more than their non-mobile counterparts, the higher rates of condomless sex among mobile FSW suggest that HIV prevention needs may remain unmet for mobile FSW in Guinea-Bissau. Additionally, the results suggest a nuanced relationship between mobility, place of residence, and HIV and social vulnerabilities and prevention indicators.

Respondent-driven sampling is more efficient than facility-based strategies at identifying undiagnosed people who inject drugs living with HIV in India

IntroductionInjection drug use drives HIV epidemics in many low-resource settings, yet many people who inject drugs (PWID) living with HIV are not diagnosed. We assessed the ability of respondent-driven sampling (RDS) – which uses peer network connections – to identify undiagnosed PWID living with HIV compared to a facility-based strategy in India. MethodsIn six Indian cities from 2014 to 2017, integrated care centers (ICCs) provided HIV testing. From 2016 to 2017, RDS samples of PWID in these same cities were conducted. Using biometric matching, characteristics associated with identification by RDS only and both RDS and ICC, compared to ICC only were explored. Undiagnosed individuals tested positive and did not report a prior diagnosis. The number needed to recruit (NNR) (average number recruited to find one undiagnosed PWID living with HIV) and the identification rate (average number undiagnosed PWID identified per week) assessed the efficiency of RDS vs. ICCs. ResultsThere were 10,759 ICC clients and 6012 RDS participants; 40% of RDS participants were also ICC clients resulting in 14,397 unduplicated PWID. PWID identified by RDS vs. ICC only were more likely to be male (adjusted odds ratios [aOR] RDS only: 6.8, both: 2.7) and living with HIV but undiagnosed (aOR RDS only: 2.5, both: 1.5). Overall, the RDS NNR was 11 and the ICC NNR was 26. The RDS identification rate (18.6/week) was faster than the ICC identification rate (2.7/week) overall and in all cities. ConclusionsRDS required screening fewer PWID and more rapidly identified undiagnosed PWID living with HIV as compared to ICCs.

Progress on the application of respondent-driven sampling in population size estimation

Respondent-driven sampling (RDS) is a method of sampling specifically for hidden populations, often applied to transgender women, female sex workers, men who have sex with men, and other groups that are difficult to identify and contact because of stigma and legal constraints. However, this method is gradually applied to the general population. With the continuous improvement of RDS, studies find that the network size of RDS samples can be weighted to estimate the overall situation and population size. This article summarizes the current application progress of RDS in population size estimation and provides ideas for the development of RDS and the use of RDS to carry out relevant research.

Populations size estimations using SS-PSE among MSM in four European cities: how many MSM are living with HIV?

Although men who have sex with men (MSM) are considered at high risk for transmission of sexually transmitted infections, including HIV, there are few studies estimating the population size of MSM in Europe. We used network data from a survey of MSM in four cities to perform successive sampling-population size estimations (SS-PSE) to estimate MSM population sizes. Data were collected in 2013-14 in Bratislava, Bucharest, Verona and Vilnius using respondent-driven sampling (RDS). SS-PSE uses a Bayesian framework to approximate the RDS sampling structure via a successive sampling model and uses the selection order of the sample to provide information about the distribution of network sizes over the population members of MSM. We estimate roughly 4600 MSM in Bratislava, 25 300 MSM in Bucharest, 7200 in Verona and 2900 in Vilnius. This represents 2.9% of the estimated adult male population in Bratislava, 2.3% in Bucharest, 2.7% in Verona and 1.5% in Vilnius. The number of MSM living with HIV would roughly be 200 in Bratislava, 4554 in Bucharest, 690 in Verona and 100 in Vilnius. Benefits of this method are that no additional information from an RDS survey needs to be collected, that the sizes can be calculated ex post facto a survey and that there is a software programme that can run the SS-PSE models. However, this method relies on having reliable priors. Although many countries are estimating the sizes of their vulnerable populations, European countries have yet to incorporate similar and novel methods.

A Bayesian framework for modelling the preferential selection process in respondent-driven sampling

In sampling designs that utilize peer recruitment, the sampling process is partially unknown and must be modelled to make inference about the population and estimate standard outcomes like prevalence. We develop a Bayesian model for the recruitment process for respondent-driven sampling (RDS), a network sampling methodology used worldwide to sample hidden populations that are not reachable by conventional sampling techniques, including those at high risk for HIV/AIDS. Current models for the RDS sampling process typically assume that recruitment occurs randomly given the population social network, but this is likely untrue in practice. To model preferential selection on covariates, we develop a sequential two-sided rational choice framework, which allows generative probabilistic network models to be created for the RDS sampling process. In the rational choice framework, members of the population make recruitment and participation choices based on observable nodal and dyadic covariates to maximize their utility given constraints. Inference is made about recruitment preferences given the observed recruitment chain in a Bayesian framework by incorporating the latent utilities and sampling from the joint posterior distribution via Markov chain Monte Carlo. We present simulation results and apply the model to an RDS study of Francophone migrants in Rabat, Morocco.

Bias decomposition and estimator performance in respondent-driven sampling

Respondent-Driven Sampling (RDS) is a method of network sampling that is used to sample hard-to-reach populations. The resultant sample is non-random, but different weighting methods can account for the over-sampling of (1) high-degree individuals and (2) homophilous groups that recruit members more effectively. While accounting for degree-bias is almost universally agreed upon, accounting for recruitment-bias has been debated as it can further increase estimate variance without substantially reducing bias. Simulation-based research has examined which weighting procedures perform best given underlying population network structures, group recruitment differences, and sampling processes. Yet, in the field, analysts do not have a priori knowledge of the network they are sampling. We show that the RDS sample data itself can determine whether a degree-based estimator is sufficient. Formulas derived from the decomposition of a ‘dual-component’ estimator can approximate the ‘recruitment component’ (RC) and ‘degree component’ (DC) of a sample’s bias. Simulations show that RC and DC values can predict the performance of different classes of estimators. Samples with extreme ‘RC’ values, a consequence of network homophily and differential recruitment, are better served by a classical estimator. The use of sample data to improve estimator selection is a promising innovation for RDS, as the population network features that should guide estimator selection are typically unknown.

Estimated Number of People who Inject Drugs in Ho Chi Minh City, Vietnam: Findings from a Two-survey Capture—Recapture Population Size Estimation Exercise

BackgroundHIV/AIDS program managers in Ho Chi Minh City (HCMC), Vietnam have always relied on the police reports and the UNAIDS Estimation and Projection Package for population size estimation of People Who Inject Drugs (PWID).MethodsWe used Respondent-driven Sampling (RDS) to implement a two-source capture—recapture study to estimate the population size of PWID in HCMC in 2017. The study was implemented in seven out of 24 districts and included men and women ages 18 years and older who reported injecting illicit drugs in the last 90 days, and who had lived in the city for the past six months. Estimates of the PWID population size for each of the seven districts were calculated accounting for the RDS sampling design. These were then adjusted to account for the district sampling probabilities to give an estimate for HCMC. Chapman two-source capture—recapture estimates of population size, based on simple random sampling assumptions, were also calculated for comparison.ResultsThe estimates resulted in a population size for HCMC of 19,155 [95% Confidence Interval (CI): 17,006–25,039] using the RDS approach and 17,947 (95% CI: 15,968–19,928), using the Chapman approach.ConclusionThe two-survey capture—recapture exercise provided estimates of PWID in HCMC — based on Chapman estimator and RDS approach — are similar. For planning HIV prevention and care service needs among PWID in HCMC, both estimates may need to be taken into consideration together with size estimates from other sources.

Unweighted regression models perform better than weighted regression techniques for respondent-driven sampling data: results from a simulation study

BackgroundIt is unclear whether weighted or unweighted regression is preferred in the analysis of data derived from respondent driven sampling. Our objective was to evaluate the validity of various regression models, with and without weights and with various controls for clustering in the estimation of the risk of group membership from data collected using respondent-driven sampling (RDS).MethodsTwelve networked populations, with varying levels of homophily and prevalence, based on a known distribution of a continuous predictor were simulated using 1000 RDS samples from each population. Weighted and unweighted binomial and Poisson general linear models, with and without various clustering controls and standard error adjustments were modelled for each sample and evaluated with respect to validity, bias and coverage rate. Population prevalence was also estimated.ResultsIn the regression analysis, the unweighted log-link (Poisson) models maintained the nominal type-I error rate across all populations. Bias was substantial and type-I error rates unacceptably high for weighted binomial regression. Coverage rates for the estimation of prevalence were highest using RDS-weighted logistic regression, except at low prevalence (10%) where unweighted models are recommended.ConclusionsCaution is warranted when undertaking regression analysis of RDS data. Even when reported degree is accurate, low reported degree can unduly influence regression estimates. Unweighted Poisson regression is therefore recommended.

Implementing Respondent-Driven Sampling to Recruit Women Who Exchange Sex in New York City: Factors Associated with Recruitment and Lessons Learned.

Respondent-driven sampling (RDS) relies on productive peer recruitment to capture hidden populations. Domestic studies have identified characteristics of productive recruitment among RDS samples of men who have sex with men and persons who use drugs, but not of women who exchange sex, a group vulnerable to HIV infection. We examined sociodemographic-, behavioral-, exchange-sex-, and protocol-related factors associated with recruitment among seeds (n = 25) and peers (n = 297) in the 2016 New York City National HIV Behavioral Surveillance Study cycle focused on women who exchange sex. Recruiter productivity was significantly associated with not having been recently incarcerated, lower rate of HIV testing, and larger exchange sex networks among seeds, and with HIV-prevention services usage among peers. We describe challenges and lessons learned from implementing RDS in this population. Our study identifies seed characteristics and protocol improvements researchers can utilize when implementing future RDS studies among women who exchange sex.

Population Size Estimations Among Hidden Populations Using Respondent-Driven Sampling Surveys: Case Studies From Armenia.

BackgroundEstimates of the sizes of hidden populations, including female sex workers (FSW), men who have sex with men (MSM), and people who inject drugs (PWID), are essential for understanding the magnitude of vulnerabilities, health care needs, risk behaviors, and HIV and other infections.ObjectiveThis article advances the successive sampling-population size estimation (SS-PSE) method by examining the performance of a modification allowing visibility to be jointly modeled with population size in the context of 15 datasets. Datasets are from respondent-driven sampling (RDS) surveys of FSW, MSM, and PWID from three cities in Armenia. We compare and evaluate the accuracy of our imputed visibility population size estimates to those found for the same populations through other unpublished methods. We then suggest questions that are useful for eliciting information needed to compute SS-PSE and provide guidelines and caveats to improve the implementation of SS-PSE for real data.MethodsSS-PSE approximates the RDS sampling mechanism via the successive sampling model and uses the order of selection of the sample to provide information on the distribution of network sizes over the population members. We incorporate visibility imputation, a measure of a person’s propensity to participate in the study, given that inclusion probabilities for RDS are unknown and social network sizes, often used as a proxy for inclusion probability, are subject to measurement errors from self-reported study data.ResultsFSW in Yerevan (2012, 2016) and Vanadzor (2016) as well as PWID in Yerevan (2014), Gyumri (2016), and Vanadzor (2016) had great fits with prior estimations. The MSM populations in all three cities had inconsistencies with expert prior values. The maximum low prior value was larger than the minimum high prior value, making a great fit impossible. One possible explanation is the inclusion of transgender individuals in the MSM populations during these studies. There could be differences between what experts perceive as the size of the population, based on who is an eligible member of that population, and what members of the population perceive. There could also be inconsistencies among different study participants, as some may include transgender individuals in their accounting of personal network size, while others may not. Because of these difficulties, the transgender population was split apart from the MSM population for the 2018 study.ConclusionsPrior estimations from expert opinions may not always be accurate. RDS surveys should be assessed to ensure that they have met all of the assumptions, that variables have reached convergence, and that the network structure of the population does not have bottlenecks. We recommend that SS-PSE be used in conjunction with other population size estimations commonly used in RDS, as well as results of other years of SS-PSE, to ensure generation of the most accurate size estimation.

A SIMULATION-BASED FRAMEWORK FOR ASSESSING THE FEASIBILITY OF RESPONDENT-DRIVEN SAMPLING FOR ESTIMATING CHARACTERISTICS IN POPULATIONS OF LESBIAN, GAY AND BISEXUAL OLDER ADULTS.

Respondent-driven sampling (RDS) is a method for sampling from a target population by leveraging social connections. RDS is invaluable to the study of hard-to-reach populations. However, RDS is costly and can be infeasible. RDS is infeasible when RDS point estimators have small effective sample sizes (large design effects) or when RDS interval estimators have large confidence intervals relative to estimates obtained in previous studies or poor coverage. As a result, researchers need tools to assess whether or not estimation of certain characteristics of interest for specific populations is feasible in advance. In this paper, we develop a simulation-based framework for using pilot data-in the form of a convenience sample of aggregated, egocentric data and estimates of subpopulation sizes within the target population-to assess whether or not RDS is feasible for estimating characteristics of a target population. in doing so, we assume that more is known about egos than alters in the pilot data, which is often the case with aggregated, egocentric data in practice. We build on existing methods for estimating the structure of social networks from aggregated, egocentric sample data and estimates of subpopulation sizes within the target population. We apply this framework to assess the feasibility of estimating the proportion male, proportion bisexual, proportion depressed and proportion infected with HIV/AIDS within three spatially distinct target populations of older lesbian, gay and bisexual adults using pilot data from the caring and Aging with Pride Study and the Gallup Daily Tracking Survey. We conclude that using an RDS sample of 300 subjects is infeasible for estimating the proportion male, but feasible for estimating the proportion bisexual, proportion depressed and proportion infected with HIV/AIDS in all three target populations.

Authorship Correction: Sampling Key Populations for HIV Surveillance: Results From Eight Cross-Sectional Studies Using Respondent-Driven Sampling and Venue-Based Snowball Sampling.

[This corrects the article DOI: 10.2196/publichealth.8116.].

Sampling Key Populations for HIV Surveillance: Results From Eight Cross-Sectional Studies Using Respondent-Driven Sampling and Venue-Based Snowball Sampling.

BackgroundIn using regularly collected or existing surveillance data to characterize engagement in human immunodeficiency virus (HIV) services among marginalized populations, differences in sampling methods may produce different pictures of the target population and may therefore result in different priorities for response.ObjectiveThe objective of this study was to use existing data to evaluate the sample distribution of eight studies of female sex workers (FSW) and men who have sex with men (MSM), who were recruited using different sampling approaches in two locations within Sub-Saharan Africa: Manzini, Swaziland and Yaoundé, Cameroon.MethodsMSM and FSW participants were recruited using either respondent-driven sampling (RDS) or venue-based snowball sampling. Recruitment took place between 2011 and 2016. Participants at each study site were administered a face-to-face survey to assess sociodemographics, along with the prevalence of self-reported HIV status, frequency of HIV testing, stigma, and other HIV-related characteristics. Crude and RDS-adjusted prevalence estimates were calculated. Crude prevalence estimates from the venue-based snowball samples were compared with the overlap of the RDS-adjusted prevalence estimates, between both FSW and MSM in Cameroon and Swaziland.ResultsRDS samples tended to be younger (MSM aged 18-21 years in Swaziland: 47.6% [139/310] in RDS vs 24.3% [42/173] in Snowball, in Cameroon: 47.9% [99/306] in RDS vs 20.1% [52/259] in Snowball; FSW aged 18-21 years in Swaziland 42.5% [82/325] in RDS vs 8.0% [20/249] in Snowball; in Cameroon 15.6% [75/576] in RDS vs 8.1% [25/306] in Snowball). They were less educated (MSM: primary school completed or less in Swaziland 42.6% [109/310] in RDS vs 4.0% [7/173] in Snowball, in Cameroon 46.2% [138/306] in RDS vs 14.3% [37/259] in Snowball; FSW: primary school completed or less in Swaziland 86.6% [281/325] in RDS vs 23.9% [59/247] in Snowball, in Cameroon 87.4% [520/576] in RDS vs 77.5% [238/307] in Snowball) than the snowball samples. In addition, RDS samples indicated lower exposure to HIV prevention information, less knowledge about HIV prevention, limited access to HIV prevention tools such as condoms, and less-reported frequency of sexually transmitted infections (STI) and HIV testing as compared with the venue-based samples. Findings pertaining to the level of disclosure of sexual practices and sexual practice–related stigma were mixed.ConclusionsSamples generated by RDS and venue-based snowball sampling produced significantly different prevalence estimates of several important characteristics. These findings are tempered by limitations to the application of both approaches in practice. Ultimately, these findings provide further context for understanding existing surveillance data and how differences in methods of sampling can influence both the type of individuals captured and whether or not these individuals are representative of the larger target population. These data highlight the need to consider how program coverage estimates of marginalized populations are determined when characterizing the level of unmet need.

NEW SURVEY QUESTIONS AND ESTIMATORS FOR NETWORK CLUSTERING WITH RESPONDENT-DRIVEN SAMPLING DATA.

Respondent-driven sampling (RDS) is a popular method for sampling hard-to-survey populations that leverages social network connections through peer recruitment. While RDS is most frequently applied to estimate the prevalence of infections and risk behaviors of interest to public health, such as HIV/AIDS or condom use, it is rarely used to draw inferences about the structural properties of social networks among such populations because it does not typically collect the necessary data. Drawing on recent advances in computer science, we introduce a set of data collection instruments and RDS estimators for network clustering, an important topological property that has been linked to a network's potential for diffusion of information, disease, and health behaviors. We use simulations to explore how these estimators, originally developed for random walk samples of computer networks, perform when applied to RDS samples with characteristics encountered in realistic field settings that depart from random walks. In particular, we explore the effects of multiple seeds, without replacement versus with replacement, branching chains, imperfect response rates, preferential recruitment, and misreporting of ties. We find that clustering coefficient estimators retain desirable properties in RDS samples. This paper takes an important step toward calculating network characteristics using nontraditional sampling methods, and it expands the potential of RDS to tell researchers more about hidden populations and the social factors driving disease prevalence.

An Immunization Strategy for Hidden Populations

Hidden populations, such as injecting drug users (IDUs), sex workers (SWs) and men who have sex with men (MSM), are considered at high risk of contracting and transmitting infectious diseases such as AIDS, gonorrhea, syphilis etc. However, public health interventions to such groups are prohibited due to strong privacy concerns and lack of global information, which is a necessity for traditional strategies such as targeted immunization and acquaintance immunization. In this study, we introduce an innovative intervention strategy to be used in combination with a sampling approach that is widely used for hidden populations, Respondent-driven Sampling (RDS). The RDS strategy is implemented in two steps: First, RDS is used to estimate the average degree (personal network size) and degree distribution of the target population with sample data. Second, a cut-off threshold is calculated and used to screen the respondents to be immunized. Simulations on model networks and real-world networks reveal that the efficiency of the RDS strategy is close to that of the targeted strategy. As the new strategy can be implemented with the RDS sampling process, it provides a cost-efficient and feasible approach for disease intervention and control for hidden populations.

Modeling and Analyzing Respondent-Driven Sampling as a Counting Process

Respondent-driven sampling (RDS) is an approach to sampling design and analysis which utilizes the networks of social relationships that connect members of the target population, using chain-referral. RDS sampling will typically oversample participants with many acquaintances. Naïve estimators, such as the sample average, will thus be biased towards the state of the most highly connected individuals. Current methodology cannot estimate population size from RDS, and promotes inverse probability weighted estimators for population parameters such as HIV prevalence. We propose to use the timing of recruitment, typically collected and discarded, in order to estimate the population size via a counting process model. Once population size and degree frequencies are made available, prevalence can be debiased in a post-stratified framework. We adapt methods developed for inference in epidemiology and software reliability to estimate the population size, degree counts and frequencies. A fundamental advantage of our approach is that it makes the assumptions of the sampling design explicit. This enables verification of the assumptions, maximum likelihood estimation, extension with covariates, and model selection. We develop large-sample theory, proving consistency and asymptotic normality. We further compare our estimators to other estimators in the RDS literature, through simulation and real-world data. In both cases, we find our estimators to outperform current methods. The likelihood problem in the model we present is separable, and thus efficiently solvable. We implement these estimators in an accompanying R package, chords, available on CRAN.

Social networks of men who have sex with men and their implications for HIV/STI interventions: results from a cross-sectional study using respondent-driven sampling in a large and a small city in Tanzania

ObjectiveMen who have sex with men (MSM) in sub-Saharan Africa remain hidden and hard to reach for involvement in HIV and sexually transmitted infection (STI) services. The aim of the...

Potential and Challenges in Collecting Social and Behavioral Data on Adolescent Alcohol Norms: Comparing Respondent-Driven Sampling and Web-Based Respondent-Driven Sampling

BackgroundRespondent-driven sampling (RDS) is a method successfully used to research hard-to-access populations. Few studies have explored the use of the Internet and social media with RDS, known as Web-based RDS (WebRDS). This study explored the use of combining both “traditional” RDS and WebRDS to examine the influences on adolescent alcohol use.ObjectiveThis paper reports on the recruitment processes and the challenges and enablers of both RDS and WebRDS. It details comparative recruitment data and provides a summary of the utility of both methods for recruiting adolescents to participate in an online survey investigating youth alcohol norms.MethodsProcess evaluation data collected from research staff throughout the study were used to assess the challenges and solutions of RDS and WebRDS. Pearson chi-square test (Fisher’s exact test if applicable) was used to compare the differences in sociodemographics and drinking behavior between data collected by RDS and WebRDS.ResultsOf the total sample (N=1012), 232 adolescents were recruited by RDS and 780 by WebRDS. A significantly larger proportion of Aboriginal or Torres Strait Islander (P<.001) participants who spoke English as their main language at home (P=.03), and of middle and lower socioeconomic status (P<.001) was found in the RDS sample. The RDS sample was also found to have a higher occurrence of past 7-day drinking (P<.001) and past 7-day risky drinking (P=.004). No significant differences in gender, age, past month alcohol use, and lifetime alcohol use were observed between the RDS and WebRDS samples. This study revealed RDS and WebRDS used similar lengths of chains for recruiting participants; however, WebRDS conducted a faster rate of recruitment at a lower average cost per participant compared to RDS.ConclusionsUsing WebRDS resulted in significant improvements in the recruitment rate and was a more effective and efficient use of resources than the traditional RDS method. However, WebRDS resulted in partially different sample characteristics to traditional RDS. This potential effect should be considered when selecting the most appropriate data collection method.

Network Structure and Biased Variance Estimation in Respondent Driven Sampling.

This paper explores bias in the estimation of sampling variance in Respondent Driven Sampling (RDS). Prior methodological work on RDS has focused on its problematic assumptions and the biases and inefficiencies of its estimators of the population mean. Nonetheless, researchers have given only slight attention to the topic of estimating sampling variance in RDS, despite the importance of variance estimation for the construction of confidence intervals and hypothesis tests. In this paper, we show that the estimators of RDS sampling variance rely on a critical assumption that the network is First Order Markov (FOM) with respect to the dependent variable of interest. We demonstrate, through intuitive examples, mathematical generalizations, and computational experiments that current RDS variance estimators will always underestimate the population sampling variance of RDS in empirical networks that do not conform to the FOM assumption. Analysis of 215 observed university and school networks from Facebook and Add Health indicates that the FOM assumption is violated in every empirical network we analyze, and that these violations lead to substantially biased RDS estimators of sampling variance. We propose and test two alternative variance estimators that show some promise for reducing biases, but which also illustrate the limits of estimating sampling variance with only partial information on the underlying population social network.

Assessing the geographic coverage and spatial clustering of illicit drug users recruited through respondent-driven sampling in New York City.

We assess the geographic coverage and spatial clustering of drug users recruited through respondent-driven sampling (RDS) and discuss the potential for biased RDS prevalence estimates. Illicit drug users aged 18-40 were recruited through RDS (N = 401) and targeted street outreach (TSO) (N = 210) in New York City. Using the Google Maps API™, we calculated travel distances and times using public transportation between each participant's recruitment location and the study office and between RDS recruiter-recruit pairs. We used K function analysis to evaluate and compare spatial clustering of (1) RDS vs. TSO respondents and (2) RDS seeds vs. RDS peer recruits. All participant recruitment locations clustered around the study office; however, RDS participants were significantly more likely to be recruited within walking distance of the study office than TSO participants. The TSO sample was also less spatially clustered than the RDS sample, which likely reflects (1) the van's ability to increase the sample's geographic heterogeneity and (2) that more TSO than RDS participants were enrolled on the van. Among RDS participants, individuals recruited spatially proximal peers, geographic coverage did not increase as recruitment waves progressed, and peer recruits were not less spatially clustered than seeds. Using a mobile van to recruit participants had a greater impact on the geographic coverage and spatial dependence of the TSO than the RDS sample. Future studies should consider and evaluate the impact of the recruitment approach on the geographic/spatial representativeness of the sample and how spatial biases, including the preferential recruitment of proximal peers, could impact the precision and accuracy of estimates.