Esther Mertens

152 | Chapter 7 and characteristics of the population (Michie et al., 2009). The effectiveness of a component can be affected by interplay among components when combined for intervention. For instance, certain combinations or sequences of components might inflate, or reduce, their individual effectiveness (Collins, Murphy, & Strecher, 2007). In addition, participants’ characteristics might affect a component’s effectiveness. For instance, my meta-analysis identified emotion regulation as a potentially ineffective component, as it was associated with weaker intervention effects. Based on this finding one would suggest to eliminate this component from R&W. However, R&W improved students’ emotional self-regulation, a desired intervention effect, making it counterintuitive to eliminate the teaching of emotion regulation from the program. While teaching students how to regulate their emotions might not be an effective component for the total student population, it could perhaps be effective for prevocational students given that this group of students probably requires a specific intervention approach. Initially, I aimed to take heterogeneity in the student population into account in my meta-analysis by examining whether the association between components and intervention effects was dependent on students’ educational level. Unfortunately, this was not possible as students’ educational level was often not reported in the papers. Thus, also when studying the effectiveness of components, it is crucial to examine under what circumstances and for whom the component is effective (Bonell, Fletcher, Morton, Lorenc, & Moore, 2012). Again, these findings highlight the complexity of generalizing findings from specific circumstances and populations to a larger overall conclusion (Rowe & Trickett, 2018). Strengths and Limitations The findings should be regarded in light of some general strengths and limitations of this dissertation. A strength of this dissertation is the use of two approaches to evaluate universal school-based interventions, that is, an RCT and a meta-analysis. On top of this dual approach, I also identified intervention components related to intervention effects of universal school-based interventions. I did this indirectly in the RCT by examining the involvement of the entire school staff (i.e., Standard condition) and of parents (i.e., Plus condition), representing the structural components of, respectively, a whole school approach and parental involvement. In the meta-analysis, I examined intervention components explicitly by analyzing which components were related to intervention effects. This elaborate approach enabled me to move beyond the evaluation of one specific universal school-based intervention and put the findings in a broader framework of this type of interventions. Another strength is that I assessed intervention fidelity, which is essential information for drawing conclusions about the potential value of an intervention (Durlak & DuPre, 2008). In the RCT all aspects of implementation, as proposed by Durlak and DuPre (2008), were measured: Fidelity, dosage, quality, participants’