1%) completed the anonymous questionnaire. Response etc rates were similar across universities (range: 74%�C78%). Our final sample consisted of the 548 students (99.3% of collected questionnaires) for whom we had complete data for our primary outcome, ever smoking of water pipe tobacco. Questionnaire and measures The questionnaire was developed from standard instruments used previously to assess water pipe tobacco smoking (Eissenberg, Ward, Smith-Simone, & Maziak, 2008; Maziak, Ward, Afifi Soweid, & Eissenberg, 2005). The survey was divided into three sections: (a) sociodemographic data, (b) personal history of water pipe and other tobacco use, and (c) attitudes and intentions regarding water pipe tobacco smoking. Sociodemographic variables included age, gender, nationality (Jordanian vs.
other), ancestry, religion, university level, marital status, housing status, and personal income. We divided age into categories based on distribution so that representation within each category was roughly equal. Personal history of tobacco use included current and past patterns of use for both water pipe tobacco and cigarettes. For the outcome of ��ever use,�� we asked an item that translated from Arabic as ��Have you ever smoked tobacco from a water pipe?�� with responses of only ��yes�� or ��no.�� For our outcome of ��use at least monthly,�� we asked only those who endorsed ever use to indicate whether their usual pattern of water pipe was (1) yearly, (2) monthly, (3) weekly, or (4) daily. We assigned anyone who stated a response of 2, 3, or 4 as using water pipe tobacco at least monthly.
Our primary attitudinal measures assessed perception of harm and addictiveness. We framed these items comparing water pipe tobacco smoking harm/addictiveness against that of cigarette smoking. This type of structure is commonly used in the literature in order to obtain data that is not highly skewed (Primack et al., 2008; Smith-Simone, Maziak, Ward, & Eissenberg, 2008; Tamim et al., 2003). A complete questionnaire is available upon request from the authors. Analysis We first computed prevalence for both our primary and secondary outcomes, ever use of water pipe tobacco and use at least monthly. We then computed prevalence of each outcome for each sociodemographic variable. We used chi-square statistics to determine statistical significance of differences in prevalence across sociodemographic variables.
We then used logistic regression to determine the association between independent and dependent variables. For these analyses, we included in our models only independent variables (i.e., sociodemographic variables) with a univariate relationship with outcomes of p < .20. We conducted sensitivity analyses using stepwise backward regression in order to determine the robustness Batimastat of our results. Finally, we used chi-squared analyses in order to assess the association between water pipe tobacco smoking behavior and beliefs regarding harm and addictiveness.