Using respondent driven sampling to assess tobacco use in an urban American Indian population: Survey implementation and sampling results

Background: While large numbers of American Indians reside in urban areas in the U.S., often no sampling frame exists to identify American Indians in urban areas for data collection. Respondent driven sampling (RDS), when implemented properly, can reduce the bias associated with other chain-referral methods. Purpose: To describe the implementation of RDS to sample an urban American Indian community for in-person interviews on tobacco use. Methods: An academic-community partnership developed and implemented the project. American Indians age ≥ 18 living in Hennepin and Ramsey Counties, Minnesota were eligible for in-person interviews, and were identified via referral by previous participants. Software designed for RDS analysis was used to draw recruitment networks. Using data from eligibility screening, respondent characteristics were determined. Community and academic partners reflected on the experience to identify lessons learned. Results: Over nine weeks, 965 eligible respondents were interviewed, starting with five initial respondents (seeds), and extending through 12 waves of referral. Recruitment networks illustrate the variation in referral patterns. Analysis of screening data reveals slightly more younger than older individuals, and more women (64%) than men (34%) among respondents. Most (85%) were from Hennepin County, reflecting greater population size and network density. Respondent distributions will be compared to 2010 Census data. Lessons learned and applicability to other cultural communities will be discussed. Conclusions: RDS is a sampling method that can be used to recruit survey participants from communities for which lists are not available if culturally appropriate staff and methods are used, and the networks are sufficiently dense.