Hester van Eeren

| Chapter 5 5 | 92 Because referral practices and treatment populations differ between countries (Asscher et al., 2013; Sundell et al., 2008), the relative effectiveness of MST and FFT is unknown in other countries. In the Netherlands, youth are referred to MST or FFT by various referral agencies, including the Child Protection Council, juvenile judges, local referral institutions, and primary health care providers. To allocate adolescents and their families to either one of the treatments, a well-known model, the Risk-Need- Responsivity (RNR) model, is often used. Following this model, the intensity of the treatment is matched to risks and characteristics of the adolescent. The higher the risk of delinquent behavior, the more intensive treatment should be (Andrews & Bonta, 2010; Andrews, Bonta, & Wormith, 2006). The model implies that adolescents should be assigned to FFT unless there are indications that MST would be more suitable, such as serious delinquent behavior, a high risk that the family cannot provide a safe environment, and an increased risk of recidivism (Oudhof, Ten Berge, & Berger, 2009). In practice, this assignment procedure is followed by clinicians, assigning youth to either FFT or MST. A previous Dutch study comparing both treatment populations found that more youth receiving MST had a court order than youth receiving FFT, and that youth receiving MST had more risk factors than those receiving FFT (Hendriks, Lange, Boonstoppel-Boender, & van der Rijken, 2014). This finding is in accordance with the results of a Swedish study which demonstrated that youth receiving MST had more behavioral problems than youth receiving FFT (Gustle, Hansson, Sundell, Lundh, & Lofhölm, 2006). However, although both European studies showed that the most at risk youth received the most intensive treatment (i.e., MST), the model leaves room for interpretation and may be subject to chance. In fact, the target populations of MST and FFT show substantial overlap (Hendriks et al., 2014). Therefore, it appears that criteria used to allocate adolescents and their families to either one of the treatments are not fully mutually exclusive. Because these studies only looked into treatment populations and did not consider treatment effects, it remains unknown which intervention is the most effective for these overlapping target populations. Therefore, the current study aimed to investigate the relative effectiveness of MST and FFT in the Netherlands. Because interventions are compared in their everyday practice settings, a quasi-experimental design was used, meaning that youth were not randomly allocated to one of the interventions. Without controlling for pre-treatment differences, a difference in outcomes is either caused by the intervention itself, or by pre-treatment characteristics of adolescents and their families. Therefore, a propensity score (PS) was estimated and used to control for this ‘allocation bias’. Using the PS, the treatment arms can be balanced from a large set of observed, pre-treatment characteristics (Austin, 2011; Rosenbaum& Rubin, 1983; Rubin, 2001). When all important covariates are measured, applying the PS to achieve the balance of the treatment arms enables controlling for allocation bias and may even yield results equivalent to randomized studies (Austin, 2011; Shadish, 2013; West et al., 2014). If randomization is not feasible in clinical practice, the use of a PS is a valid solution (Shadish, 2013; West et al., 2014). It should, however, be noted that, in contrast to randomized studies, a PS can only control for overt bias (i.e., bias due to observed pre-treatment differences) and not for hidden bias (i.e., bias due to unmeasured or unobserved differences) (Rosenbaum,