Hester van Eeren

| Chapter 5 5 | 110 Discussion Using the PS method to control for the non-random assignment of adolescents to either MST or FFT, this study compared these two interventions on their effectiveness in the Netherlands. In the study sample, target populations were balanced and no differences between the interventions were found regarding externalizing problem behavior. This result proved to be robust in the complete case sample. Some additional results were found, but these were not robust: adolescents assigned to MST were more often engaged in school or work after treatment. This treatment objective likely receives greater emphasis during MST than FFT. Moreover, in the complete case sample, adolescents who had received FFT had less police contact during treatment than adolescents who had received MST. Because the MST sample included a higher percentage of adolescents who had a court order before treatment, they were probably more likely to have additional police contact during treatment. In the present study, the average treatment effect of the treated was estimated and the finding suggests that adolescents who receive MST may display the same treatment effects if they would have received FFT. This treatment effect, however, is only applicable for adolescents and their sample characteristics for whom there were outcome measurements after treatment. Finding only a few robust differences when comparing the effectiveness of MST and FFT in the overall study sample is in accordance with previous findings by Baglivio and colleagues (2014). As the present study demonstrates that adolescents with a court order — interpreted as a possible risk factor following the RNR-model (Andrews et al., 2006; Laan, Slotboom, & Stams, 2010) — were more often assigned to MST (246 MST; 71 FFT), MST could also be expected to be more effective in this subsample. However, due to the incomparability of the FFT and MST subsamples of youth with a court order, the present study cannot confirm this. On the other hand, following the RNR model, FFT could be expected to be sufficiently effective in the subsample of adolescents without a court order, as these adolescents would be expected to have lower risks, and therefore, less intensive treatment would be adequate (Andrews et al., 2006; Laan et al., 2010). However, MST was more effective in reducing externalizing problems in the subsample without a court order. This may be explained by the fact that, although some risk factors were less present in the group without a court order, such as engagement in school or police contact (Table I and IV in Supplemental Material), this group nevertheless reported more problem behavior measured with the CBCL and the YSR (Tables I and IV in Supplemental Material). Another explanation may be that having or not having a court order only provides a rough indication of the risk level of an adolescent, while clinicians assign adolescents to either MST or FFT based on other risk factors as well. The RNR model thus leaves room for interpretation, or a single characteristic cannot fully represent the risk level of an adolescent. Even more, it could be possible that more intensive treatment in a less severe target population is always likely to be more effective, but the question remains as to whether it is appropriate and proportional treatment. For the secondary outcomes, however, no differences were found between the interventions, though these outcomes may be highly relevant to society. This should be taken into account when interpreting the overall effectiveness of the interventions in this subgroup. Furthermore,