Evaluating Respondent-Driven Sampling in a Major Metropolitan Area: Comparing Injection Drug Users in the 2005 Seattle Area National HIV Behavioral Surveillance System Survey with Participants in the RAVEN and Kiwi Studies

Purpose

To empirically evaluate respondent-driven sampling (RDS) recruitment methods, which have been proposed as an advantageous means of surveying hidden populations.

Methods

The National HIV Behavioral Surveillance system used RDS to recruit 370 injection drug users (IDU) in the Seattle area in 2005 (NHBS-IDU1). We compared the NHBS-IDU1 estimates of participants' area of residence, age, race, sex, and drug most frequently injected to corresponding data from two previous surveys, the RAVEN and Kiwi Studies, and to persons newly diagnosed with HIV/AIDS and reported from 2001 through 2005.

Results

The NHBS-IDU1 population was estimated to be more likely to reside in downtown Seattle (52%) than participants in the other data sources (22%–25%), be older than 50 years of age (29% vs. 5%–10%), and report multiple races (12% vs. 3%–5%). The NHBS-IDU1 population resembled persons using the downtown needle exchange in age and race distribution. An examination of cross-group recruitment frequencies in NHBS-IDU1 suggested barriers to recruitment across different areas of residence, races, and drugs most frequently injected.

Conclusions

The substantial differences in age and area of residence between NHBS-IDU1 and the other data sources suggest that RDS may not have accessed the full universe of Seattle area injection networks. Further empirical data are needed to guide the evaluation of RDS-generated samples.

Introduction

Injection drug users (IDU) are a population at elevated risk for several infections of public health importance, including HIV, hepatitis B, and hepatitis C 1, 2. Epidemiologic surveys of drug-injecting populations are important in measuring the prevalence of these diseases, evaluating public health measures to control them, identifying unmet needs and noting opportunities for prevention efforts. However, obtaining an unbiased sample of IDU has proved problematic because of the illegal nature of drug injection and the social marginalization of many IDU. Most methods of IDU recruitment contain well-recognized sources of bias, the effects of which are difficult or impossible to quantify 3, 4.

Respondent-driven sampling (RDS) has been proposed as an advantageous means of accessing hidden populations 5, 6, 7. In RDS, participants are given coupons with which to recruit their peers and offered payments when new recruits bring in the coupons. Theoretical reasoning and mathematical modeling propose that RDS methods can produce a study population from which unbiased estimates of the properties of a target population can be calculated. RDS offers further advantages in the ease of its implementation and the standardization of its methods, advantages which have made RDS attractive in international studies with limited resources 8, 9. While RDS is being used in a growing number of settings 10, 11, 12, 13, 14, 15, 16, 17, 18, 19, 20, 21, 22, 23, 24, there remains a need for empirical data evaluating how well RDS fulfills its promise in practice.

One means of evaluating RDS is to compare RDS results for a population with data on the same population obtained through other means (25). In 2005, IDU were surveyed for the National HIV Behavioral Surveillance system using RDS in 23 U.S. cities (the NHBS-IDU1 survey), including Seattle (25). Two previous studies, the Risk Activity Variables, Epidemiology and Network Study (RAVEN) and the Kiwi study, recruited Seattle area IDU using differing sampling strategies. Data on IDU diagnosed with HIV/AIDS during the period 2001–2005 were available through the HIV/AIDS Reporting System (HARS). We compared these different populations in terms of area of residence, age, race, gender, and drug most frequently injected, variables commonly used to characterize IDU populations 15, 18, 20, 26.