All analyses were stratified by sex, as previous studies have shown clear differences in adolescent smoking between boys and girls [7, 8]. Initially, the prevalence and 95% confidence interval (CI) of smoking was estimated by each SES indicator and each survey year. Cochran-Armitage test for trend was used to assess time trends in smoking prevalence over time by each SES category. Chi-square test was performed to evaluate differences in smoking prevalence among different categories for each SES indicator by each survey year. When the sample size was small, Fisher’s exact test was conducted. Next, socioeconomic inequalities in smoking between low and high SES groups were assessed by absolute and relative measures. Both absolute and relative measures were estimated with 95% CIs for each SES indicator and each survey year. For the absolute measures, prevalence differences (PDs) [10, 12, 15, 27] and the slope index of inequality (SII) [2830] were calculated using generalized linear models with binomial distribution and identify link function. The coefficient yields an estimate of the absolute inequality. When this binomial model failed to converge, a generalized linear model with normal distribution and identify link function was used [31]. For the relative measures, prevalence ratios (PRs) [15] and the relative index of inequality (RII) [2830] were calculated using generalized linear models with binomial distribution and log link function. The exponentiated coefficient yields an estimate of the relative inequality. The SII and RII were estimated using ridit score for each SES indicator as an independent variable. The PDs and PRs are simple measures of inequality, which are pairwise comparisons of smoking prevalence between low and high SES groups [27]. The SII and RII are summary measures of inequality as the changes in smoking between the bottom and top points in the SES hierarchy while accounting for the cumulative distribution in each SES [27]. Time trends of absolute and relative measures were assessed by the inclusion of the interaction term between each SES indicator (ridit score for the SII and RII) and survey year [27, 28]. In the model, survey year was treated as a continuous variable coded as 1 for 2008, 2 for 2012, and 3 for 2016 [28] and Wald test was used to test if the interaction term was statistically significant. Finally, these models were adjusted for sociodemographic factors, such as grade and region, which were considered as potential confounders.

Note: The content above has been extracted from a research article, so it may not display correctly.



Q&A
Please log in to submit your questions online.
Your question will be posted on the Bio-101 website. We will send your questions to the authors of this protocol and Bio-protocol community members who are experienced with this method. you will be informed using the email address associated with your Bio-protocol account.



We use cookies on this site to enhance your user experience. By using our website, you are agreeing to allow the storage of cookies on your computer.