Maartje Boer

VALIDATION OF THE SMD-SCALE 49 2 endorsed criteria was 1.28 times higher than for boys. For lower and medium educated adolescents, the number of endorsed criteria was 1.42 and 1.27 times higher than for higher educated adolescents. Post-hoc pairwise comparisons showed that lower educated adolescents also endorsed more criteria than medium educated adolescents. Compared to 12-year-olds, the number of endorsed criteria was 1.16 times higher for 15-year-olds. Post-hoc pairwise comparisons showed that 12- and 15-year-olds were the only age groups that differed significantly in the number of present criteria. For adolescents with a non-Western immigrant background, the number of endorsed criteria was 1.20 higher than for native adolescents. Post-hoc pairwise comparisons showed no other differences by ethnic background. Inaddition,werepeatedpreviousanalyses, butusedriskyandproblematic SMU as outcome conducting multinomial regression (using normative SMU as the reference category). Table 2.5 shows that girls and adolescents who attended low or medium education were more likely to report risky SMU and problematic SMU than boys and adolescent who attended high education, respectively. For example, 4.06% of all girls were likely to report problematic SMU, compared to 2.89% of all boys. Compared to 12-year-olds, 13- and 15-year- olds had a higher probability of reporting risky SMU (30.81% vs. 36.98% and 36.67%, respectively). Problematic SMU did not vary significantly by age. Risky SMU did not vary across ethnic background, but non-Western adolescents had a higher probability of reporting problematic SMU compared to native adolescents (5.05% vs. 3.10%). Discussion Using a large-scale, nationally representative sample of Dutch adolescents, the present studydemonstratedgoodpsychometricproperties for theSocialMedia Disorder (SMD)-scale (Van den Eijnden et al., 2016), that measures problematic SMU. Multiple assessments of structural validity showed a solid unidimensional factor structure, whereby all nine items substantially contributed to the factor. The test scores showedgood internal consistency, but theyweremost reliable at higher levels of the scale’s continuum. The factor structure was measurement invariant across gender, educational level, age, and ethnic backgrounds. The data yielded three subgroups of users that were distinguished by low, medium, and high proportions of positive scores on all criteria rather than on particular