Adjusting for Changes in Mode of Administration in Repeated Cross-Sectional Surveys

Abstract

Objective: The present study compares changes in mean consumption and hazardous drinking between 1992/3 and 1997 in Switzerland, first without adjusting for changes in mode of questionnaire administration, and then by adjusting for mode of administration by use of a methodological split-sample study. Changes were compared for the total population and for subgroups of sex and linguistic regions. Method: The samples of the first and second Swiss Health Surveys (SHS), conducted in 1992/3 and 1997, respectively, were analyzed. Both surveys had used a combined telephone/self-administered questionnaire technique. Individuals had first been interviewed by telephone and were subsequently sent by mail a questionnaire containing additional questions. Whereas in the first SHS the questions on alcohol consumption had been included in the mailed questionnaire, in the second they were administered by telephone. In addition, to evaluate possible effects of different questionnaire versions, a split-sample study conducted in 1996 administered the version of the first SHS to one split of the sample and that of the second to the other. For the present study, only current drinkers were analyzed, resulting in sample sizes of N=9,464 (first SHS), N=9,774 (second SHS), and N=658 (split-sample study). Results: When mode ofadministration was not taken into account, mean consumption and percentages of hazardous drinkers decreased overall and in all but one ofthe eight subgroups examined. Controlling for mode of administration, however, resulted in (a) an overall increase in mean consumption and in the percentages of heavy drinking and (b) both increases and decreases for separate subgroups. Conclusions: Methods of data collection may influence within-country comparisons of repeated cross-sectional surveys of alcohol consumption.