Comparison of methods for identifying a population-based sample of Filipina women for a health survey

This paper describes and compares 2 random-digit dialing (RDD) methods that have been used to select minority subjects for population-based research. These methods encompass the census-based method, which draws its primary sampling units from census tracts with a high proportion of minority persons, and the registry-based method, which derives its primary sampling units from a population-based cancer registry. Our study targeted Filipinos living in 10 Northern California counties, where they constitute 4% of the total population. Eligible participants (Filipina women, at least aged 20, who spoke 1 of 4 interview languages) were asked to complete a short telephone interview. Both the census and registry methods located Filipino households with comparable efficiency and with a higher yield than would be expected in a non-targeted population survey, such as the Mitofsky-Waksberg RDD method. No systematic pattern of responses was evident that would indicate that either method sampled women who were systematically less acculturated or less likely to use cancer screening tests. Although both methods offer substantial gains in efficiency, their utility is limited by generating samples that tend to over-represent high-density areas. The degree to which these methods are considered viable depends on further refinement to limit, or eliminate, their inherent selection biases without sacrificing their increased efficiency to locate minority populations.