Estimates of population smoking prevalence: self-vs proxy reports of smoking status.

Abstract

In the face of rising costs of surveillance systems, it is time to reexamine the feasibility of including proxy respondents in surveys designed to provide population estimates of smoking prevalence. Data are from the California. Tobacco Surveys, which are random-digit dialed telephone surveys. One adult provided demographic information and smoking status for all household residents. Additionally, some adults were selected for in-depth interviews that also included smoking status questions. We matched information from proxy respondents and self-respondents and evaluated smoking status discrepancies between them relative to demographic and other factors (n = 2930 matched pairs) in 1992. We address the potential bias these discrepancies might introduce into the population estimate of smoking prevalence. Overall, the discrepancy between proxy report and self-report was 4.3%, and it increased particularly when the self-respondent reported nondaily smoking or recent quitting. Discrepancies acted in both directions, and the net bias was that the screener survey overestimated smoking prevalence by 0.1% in 1992 (0.3% in 1990). Smoking status questions can be added to ongoing surveys such as the census or labor force surveys; one adult could provide smoking status for all household members.