Is there a simple correction factor for comparing adolescent tobacco-use estimates from school- and home-based surveys?

Abstract

School surveys conducted to monitor the prevalence of adolescent smoking generally yield higher estimates than home-based surveys. Our goal was to examine the consistency of this mode effect over age for different measures of adolescent smoking behavior. If consistent, it would be feasible to define a simple adjustment factor for comparing overall results across survey modes. We identified three pairs of school- and home-based surveys (two national pairs and one California pair), each pair conducted in the same year, that allowed us to compare three key measures of adolescent smoking across age: (a) Ever tried, (b) smoking in past 30 days, and (c) regular smoking. We also examined male experimentation with smokeless tobacco, for which a mode effect has not been investigated. School surveys yielded higher estimates than in-home surveys for all the tobacco-use measures examined. In general, the mode effect was greater in younger age groups. Estimates for the two modes appeared closer for current and regular smoking measures compared with the ever-tried measure. Considering report of smoking in the past 30 days, the mode effect was much greater for infrequent smoking (1-9 days) than for more frequent (10+ days) smoking. Finally, school surveys yielded much higher estimates for male smokeless tobacco use than did home surveys. Because mode effects were related to age and level of reported use, no simple correction factor for comparing across modes is feasible. Social desirability bias of self-reported behavior in the different settings may account in part for the observed mode effects.