Maartje Boer

CHAPTER 2 48 Based on the findings from the LCA (Figure 2.4), we distinguished normative users (endorsement of not more than one criterion), risky users (endorsement of two to five criteria), and problematic users (endorsement of six to nine criteria). Subgroup differences were investigated while controlling for demographic characteristics and with a Bonferroni correction. Results in Table 2.4 show that risky users weremore likely to reportmental health, school, and sleep problems than normative users (OR range = 1.63 (peer problems) to 2.81 (emotional problems), p < 0.001). To an even greater extent, problematic users were more likely to report problems related to mental health, school, and sleep, than normative users (OR range = 2.47 (low sleep duration) to 8.44 (conduct problems), p < 0.001). Furthermore, post-hoc pairwise comparisons showed that problematic users had a higher probability of reporting mental health problems, school problems, and low sleep quality than risky users. Problematic and risky users were equally likely to report low sleep duration. To facilitate interpretability, we transformed odds ratios into marginal effects (M) , which denote effect sizes in terms of probabilities (Williams, 2012). Table 2.4 shows that for each increase in the number of endorsed criteria, the probability of reportingmental, school, and sleep problems increases with 2.10 (peer problems) to 5.45% (perceived school pressure). The subgroups differed most in emotional and conduct problems: Compared with normative users (10.57% and 7.65%, respectively), risky users were more than twice as likely to report emotional problems and conduct problems (24.93% and 18.44%, respectively), and problematic users were four to five times more likely to report emotional and conduct problems (40.23% and 41.13%, respectively). In sum, these findings confirm criterion validity of the test score interpretations, because the higher the level of problematic SMU, the higher the probability of problems related to mental health, school, and sleep. Also, as compared to adolescents in the normative SMU-subgroup, adolescents in the problematic SMU subgroup reported more mental health, school, and sleep problems, followed by adolescents in the risky SMU subgroup. Predictors of Problematic SMU Table 2.5 shows the associations between adolescents’ demographic characteristics and their number of endorsed problematic SMU criteria ( p- values were adjusted with Bonferroni corrections). For girls, the number of