Feasibility of using respondent-driven sampling to recruit participants in superdiverse neighbourhoods for a general health survey.

OBJECTIVES: Respondent-driven sampling (RDS), a modified chain-referral system, has been proposed as a strategy for reaching 'hidden' populations. We applied RDS to assess its feasibility to recruit 'hard-to-reach' populations such as migrants and the unemployed in a general health survey and compared it to register-based sampling (RBS).

METHODS: RDS was applied parallel to standard population RBS in two superdiverse neighbourhoods in Bremen, Germany. Prevalences of sample characteristics of interest were estimated in RDS Analyst using the successive sampling estimator. These were then compared between the samples.

RESULTS: Only 115 persons were recruited via RDS compared to 779 via RBS. The prevalence of (1) migrant background, (2) unemployment and (3) poverty risk was significantly higher in the RDS than in the RBS sample. The respective estimates were (1) 51.6 versus 32.5% (95% CIRDS 40.4-62.7), (2) 18.1 versus 7.5% (95% CIRDS 8.4-27.9) and (3) 55.0 versus 30.4% (95% CIRDS 41.3-68.7).

CONCLUSIONS: Although recruitment was difficult and the number of participants was small, RDS proved to be a feasible method for reaching migrants and other disadvantaged persons in our study.