Weighted percentages and survey-adjusted standard errors measured the distribution of the covariates (demographic characteristics and lifetime psychiatric variables) reported at Wave 1 for the sample that completed the Wave 2 interview. Univariate and multivariate logistic regressions with survey-design corrections and sampling weights were used to estimate the likelihood of suicide attempts occurring between Wave 1 and Wave 2. The resulting estimates are presented as unadjusted (OR) and adjusted odds ratios (AOR), respectively, with 95% confidence intervals (CI). All multivariate analyses included all demographic and lifetime psychiatric diagnostic variables (control variables) and a previous suicide attempt (before Wave 1), as shown in Table 1. The comparative strength of DSM-IV nicotine dependence versus tobacco use status (current, past, or never) and change in smoking status was tested by the Bayesian information criterion (BIC), the covariances between estimated parameters, and likelihood ratio (LR) tests on the improvements in the -2 Log-Likelihood among two nested models [28]. Data were analyzed with SAS 9.2 (Cary, NC, USA) using the procedures included to correct for complex survey design.
