Assessment of coverage rates and bias using double sampling methodology.
OBJECTIVE: In 1999, a survey of the health status of six rural central New York counties was performed using double sampling. The resulting impact of the methodology on health outcome prevalence estimation was assessed. STUDY DESIGN AND SETTING: Six counties were double sampled using methodology described by Neyman. Prevalence of health outcomes was estimated for the initial survey responders, nonresponders who subsequently responded after being offered a financial incentive, and both groups combined. RESULTS: Although both groups were similar on demographic variables, they differed significantly on several disease and health behavior prevalences. The failure to use double sampling would have resulted in overestimation of the prevalence of chronic disease by as much as 6.2% for females and 2.0% for males. CONCLUSION: Double sampling increased the generalizability of the results from approximately 35% of the population to over 70%, and decreased the bias in the estimation of many of the health end points.