Table 2 presents the results of the multivariate model for ever initiating smoking and for smoking continuation.
Multivariate models always contained age and ever/never smoking status.
In our multivariate models with ever having suicidal ideation or suicidal thought as response variables, psychiatry was a significant predictor, while in the model with ever having serious suicidal thoughts, there was no difference between medical specialties, suggesting that seriously thinking on suicide is less dependent on specialty.
However, middle-born sibling position was not in the final multivariate model because it was associated with ever having had commercial sex.
Since they are correlated, only one of these variables was ever included in the multivariate model.
Several of the indicators were eliminated in the final multivariate model including household income (p = 0.533), whether ever homeless (p = 0.749), family composition (p = 0.433), main source of income (p = 0.638) and labour force status (p = 0.929).
History of STI diagnosis ever was not included in the multivariate model due to collinearity with the outcome of interest.
Although skin color, skin reaction to first summer sun, and sunbathing between ages 16 to 25 were univariately associated with ever indoor tanning, these did not remain in the multivariate model after adjustment for other factors.
Because of potential synergistic effects between arsenic and smoking on lung cancer [ 6, 11, 18, 32], we then tested effect modification by smoking status (never and ever) through the addition of cross-product terms into the multivariate model.
Univariate models were constructed initially, then age and sex adjusted; finally a multivariate model was developed adjusting for age, gender, baseline smoking status (categorised as current, ever or never smokers), inflammatory marker (ESR in EAC or CRP in NOAR) level, and year of enrolment to the cohort as a proxy for changing treatment strategies over time.
