Statistical analysis

Analysis was performed in SAS, Version 9.3.⁵⁴ To answer research question 1, descriptive statistics were calculated for each abuse exposure variable.

To answer research question 2, participants were categorized by smoking status (i.e., never, former, and current) and descriptive statistics were calculated for each demographic variable. Chi-square tests were performed for categorical variables and Kruskal–Wallis tests were performed for ordinal variables to determine if the distribution of each demographic and abuse exposure variable differed across smoking status.

To answer research question 3, and mirroring the modeling for smoking status on the published risk factor for smoking paper among this population,¹⁶ a multivariable logistic regression model was fit with the binary outcome of smoking status (1 = current smoker, 0 = former, and never smoker). Univariate analysis with binary smoking status was performed for all known risk factors for smoking among this population¹⁶ contained in the dataset and all abuse exposure variables, using LOGISTIC procedure in SAS. All univariate terms found to have significant associations (p = .05) were then added to a multivariable logistic regression model to allow for the independent effect of each variable, controlling for other known risk factors. The model fit was examined using the Hosmer–Lemeshow test and area under the receiver operating characteristic curve.⁵⁵