Respondent driven sampling: Feasibility, validation and its potential use for estimating health indicators for large general populations

Background. Researchers use many methods to collect population-based health indicator data. Random digit dial (RDD) is one frequently used method however decreasing sample coverage and non-response bias has researchers searching for alternative methods. We hypothesize that respondent driven sampling (RDS) is a cost effective alternative. In accordance with this hypothesis we propose three objectives: (1) assess the mechanics, feasibility, and representativeness of RDS in producing valid estimates of health indicator data (2) evaluate the feasibility in using RDS in comparison to RDD to collect health indicator data from a general population and (3) report on data collected using RDS, specifically the public's perception and acceptance of new HIV testing recommendations.Methods. To address the first objective, a literature review of all published RDS articles was conducted. To fulfill the second objective, we recruited a sample of Albany County, NY adults. Reasons for participating, willingness to refer, and the estimated cost and time needed to collect the sample were recorded. To address objective three, we asked participants about their perceptions of the HIV testing recommendation, whether they would opt-out of testing, and if yes, why.Results. RDS is validated for sampling hard-to-reach populations. However, inconsistent reporting of methodological elements makes determining sample representativeness difficult. These concerns can be alleviated with consistent reporting of methodological issue. When using RDS to recruit from a general population, participants were receptive to RDS. Cross-group recruitment occurred and social network sizes were dense enough to sustain sufficient recruitment. RDS successfully recruited minorities who are often underrepresented in RDD studies. Participants with low HIV transmission risk perceived the HIV testing recommendations to be a good idea. Those who would opt-out were older, had not previously been tested for HIV, and did not perceive themselves to be at risk. More would opt-out in an emergency department compared to a doctor's office.Discussion. Overall, RDS is a relatively new tool that has utility in sampling both hard-to-reach and general populations. We suspect the utility in using RDS to sample general populations, or subsets thereof, will continue to prosper as traditional sampling frames become less inclusive.