The objectives of this thesis were to explore how effectiveness of school-based anti-bullying interventions (SBABI) evolves over time and to assess the possibility to predict the medium-term or long-term effectiveness of SBABIs on the basis of their short-term effectiveness. The first step included a literature review in order to understand the study designs and evaluation techniques that researches used to assess the effectiveness. This literature review described the methodologies based on which researchers collected evidence and concluded on the effectiveness of their SBABIs. In order to address the thesis objectives, a collaborative project was established, named SET-Bullying (“Statistical modelling of the Effectiveness of school based anti-bullying interventions and Time”). The above-mentioned literature review was used to identify potentially eligible studies. After addressing a call for collaboration to the corresponding authors of these studies, this project included data from two of them, the DFE-SHEFFIELD study from United Kingdom and the RESPEKT study from Norway. Both of these studies have used pupil self-reported frequencies on being bullied and bullying others as an effectiveness measure, but using different instruments to elicit this information. Thus, the subsequent step of this thesis was to harmonize the data from these studies using polychoric principal components analysis, in order to be able to perform the same analysis with the data from both studies. The data from both studies were analysed using mixed effect models in order to take into account the hierarchical (i.e. the responses of pupils from the same school may be more correlated with each other as opposed to the responses of pupils from different schools) and the longitudinal structure (i.e. same pupils are more likely to respond in a similar way in the repeated measurements of each studies) of the data. With regard to the primary objective of the thesis, it was observed that effectiveness (where it is observed) may evolve either in a linear fashion or a “delayed effect” may be observed. This refers to a minimal evolution of effectiveness over the first study measurements and a sharper evolution at the later study measurements. This finding is only hypothesis generating at this point. Would this be confirmed in future studies, it will have important implication of the design, implementation and evaluations of SBABIs. About the secondary objective of this thesis, there were some preliminary findings of the possibility to predict the medium-term or long-term effectiveness based on the short-term effectiveness. However, these predictions in some cases seemed to be very variable. Future research should focus on how to make these predictions more accurate in order that this allows for dynamic and adaptable delivery of SBABIs. / Doctorat en Santé Publique / info:eu-repo/semantics/nonPublished