Learning Objectives: Explain how declining response rates may affect population estimates of smoking prevalence.
Abstract: Objective: Random digit dial telephone surveys have been experiencing declining response rates throughout the US. This study assesses bias in two statewide tobacco surveys, one conducted in 1993 with a 78.4% response rate, and one conducted in 2000 with a 64.3% response rate.
Methods: Data are from the household screening interviews of the 1993 Massachusetts Tobacco Survey and the 2000 UMass Tobacco Study conducted by the Center for Survey Research, UMass Boston. The interview is conducted with one adult informant who provides demographic and smoking status information for each adult in the household. These data are used to estimate statewide smoking prevalence. Since smoking prevalence tends to differ according to age, sex and education, if the lower response rate of the 2000 survey is associated with increased discrepancies from the demographics of the adult population of Massachusetts, we can expect that survey estimates of smoking prevalence would be less accurate. The demographic characteristics of the 1993 and 2000 survey samples were compared to the US Census data for Massachusetts in 1990 and 2000 respectively.
Results: Preliminary analyses indicate that the discrepancies in gender and education distributions were almost identical for the two surveys; the discrepancy for age distributions was somewhat lower for the 2000 survey than the 1993 survey. One group, those with less than a high school education, are slightly better represented in the later study with a lower response rate.
Discussion: Response rate alone is not a sufficient indicator of the representativeness of a population-based survey sample.
Back to Evaluation and Surveillance Posters
Back to Evaluation and Surveillance
Back to The 2002 National Conference on Tobacco or Health