Maartje Boer

VALIDATION OF THE SMD-SCALE 31 2 these validation steps, we examined the association between adolescents’ demographic characteristics and problematic SMU. Methods Sample Analyses were carried out using cross-sectional data from the HBSC-study, conducted in the Netherlands. The study is part of a WHO-collaborative cross-national study carried out every four year since 1983 and investigates adolescents’ well-being and health behaviors in their social context. We used the Dutch HBSC-sample collected in 2017 among secondary school students (Stevens et al., 2018). The sample consisted of 6,718 adolescents (51.16% boys) aged 12-16 years ( M age = 13.94, SD age = 1.37). The sample comprised different educational levels (46.32% pre-vocational, 25.34% general higher, and 28.34% pre-university) and ethnic backgrounds (78.27% native, 16.59% had at least one parent born in a non-Western country, and 5.15% had at least one parent born in a non-DutchWestern country). Although the sample closely resembled the adolescent population in the Netherlands, the data included sample weights toadjust for sampledistributiondifferenceswiththepopulation. Theseweights included gender, educational level, school year, and urbanization degree of participants. The HBSC-sample was therefore nationally representative for the Dutch adolescent population in secondary schools (Van Dorsselaer, 2018). For analytic purposes, the sample was randomly split into two subsamples, which we labelled as ‘calibration sample’ ( n = 3,359) and ‘validation sample’ ( n = 3,359). Respondents who did not respond to any of the items on the SMD- scale were excluded from these samples ( n = 92), which yielded a final sample of n = 6,626 ( n calibration = 3,310, n validation = 3,316). The HBSC-data had a hierarchical structure, where adolescents were nested in school classes ( n = 328) and schools ( n = 85). Schools were randomly selected from a list of schools provided by the Dutch Ministry of Education, after which three to five classes per school (depending on the number of students per school) were randomly selected. The response rate on school- level was 37%. The main reason for not participating was that schools were already approached for other research. School non-response was somewhat higher among schools in urban than in rural areas ( χ ²(5) = 15.6, p < 0.01). Participating and non-participating schools did not differ regarding their