Use of voter registration records to recruit a representative sample.

Research samples are not often compared to broader community samples to evaluate their representativeness, a critical factor in determining the generalizability of study findings. This study evaluated the use of voter-registration records for recruiting a representative sample of community-dwelling, older, and overweight participants for research on improving measures of diet and physical activity. County voter-registration records were used to identify individuals between 45 and 75 years of age and living in the two cities closest to the research lab. The data were collected from July, 2007 through November, 2008. Prospective participants were mailed an introductory letter and opt-out postcard, and received a follow-up recruitment phone call in which they underwent further screening if interested in participating. The representativeness of the final voter-recruited sample (N = 191) was evaluated by comparisons of demographic variables with Behavioral Risk Factor Surveillance System (BRFSS) data at the county and state levels. The voter-recruited sample was only partially comparable to that of the BRFSS sample, with expected differences in variables related to race/ethnicity, the proportion of women, employment status, and educational attainment. Voter-registration records are a relatively low-cost ($75 per participant) method of recruiting a community sample that avoids some biases of other recruitment methods, but may not achieve a fully representative sample.