Abstract

۱٫ Introduction

۲٫ Data and methods

۳٫ Results

۴٫ Discussion

۵٫ Conclusion

Funding

Declaration of competing interest

APPENDIX

References

Abstract

     This study examined associations between characteristics of the residential neighbourhood and the school and adolescent mental health, including the moderating role of family socioeconomic status (SES) and family support. Nationally representative Dutch data from adolescents aged 12–۱۶ (N = 6422) were analysed through cross-classified multilevel models. Findings showed that school characteristics are more strongly linked to adolescent mental health than residential neighbourhood characteristics. More specifically, higher levels of school SES were associated with more hyperactivity-inattention problems, while higher levels of school social disorder were related to more conduct problems and more peer relationship problems. Further, higher levels of school SES were associated with more emotional symptoms only for adolescents with a relatively low family SES. Higher levels of neighbourhood SES were associated with fewer peer relationship problems. Overall, there was little evidence for the moderating role of family SES or family support.

Introduction

     With a global prevalence of about 13% of adolescents suffering from mental health problems (Polanczyk et al., 2015), it is of vital importance to identify both risk and protective factors for these mental health problems. In addition to person-level characteristics such as gender and socioeconomic status (Afifi, 2007; Rivenbark et al., 2019), there is increasing evidence that contextual factors matter. Ecological theory posits that environmental contexts influence mental health problems among adolescents (Bronfenbrenner, 1977; Cohen et al., 2009; Minh et al., 2017). The residential neighbourhood and school context represent the most important out-of-home environmental contexts, within which adolescents spend a substantial amount of their daily life given compulsory education and mobility restrictions (Allison et al., 1999).

     Two often studied contextual characteristics within these contexts include physical factors such as the amount of greenspace and air pollution (Jennings and Bamkole, 2019; Mueller et al., 2019; Vanaken and Danckaerts, 2018; Zhang et al., 2020) and social factors, particularly socioeconomic status (SES), social environment and social disorder (Aldridge and McChesney, 2018; Visser et al., 2020). It is important to consider both physical and social factors, not only because these are both related to adolescent mental health, but also because they are interrelated (Sugiyama et al., 2008). For instance, nearby greenspace can stimulate social activities within neighbourhoods (Ruijsbroek et al., 2017).

Results

     Descriptive statistics for the variables are displayed in Table 1. Correlations between the variables are provided in the Appendix (Table A1). With a largest variance inflation factor (VIF) of 8.04, all VIF values were below the critical value of 10 (Hair et al., 1995), indicating no multicollinearity (Table A2).

۳٫۱٫ Associations on the neighbourhood and/or school level

     Table 2 showed the results of the school-only and neighbourhood-only MLMs and the CCMM per mental health indicator. The school-only and neighbourhood-only MLM showed comparable and significant ICC values at the neighbourhood and school level (e.g., for hyperactivity-inattention problems ICCneighbourhood = 2.7% and ICCschool = 3.1%), indicating that the proportion of variance in adolescent mental health was rather similar at both levels. However, when assessing both contexts simultaneously using CCMMs, the neighbourhood-level ICCs for all mental health indicators were considerably smaller than the school-level ICCs, implying that the between-level variance in adolescent mental health was largely driven by schools rather than neighbourhoods. Furthermore, for all mental health outcomes, the ICCneighbourhood values decreased when including the school level, indicating an omitted context bias. Additionally, the CCMM for emotional symptoms and hyperactivity-inattention problems revealed that the neighbourhood variances were nonsignificant. Although these results suggested that we could drop the neighbourhood level from the models and use MLMs with only the school level context for both emotional symptoms and hyperactivity-inattention problems, we performed CCMM for all four mental health outcomes for reasons of consistency.