Identification of Multisystemic Therapy (MST) Subgroups with Distinct Trajectories on Ultimate Outcomes in Norway

The effect of Multisystemic Therapy (MST) treatment for serious behavior problems among adolescents has been established through multiple studies carried out in various countries as well as in meta-analyses (Asscher et al. 2014; Curtis et al. 2004; Henggeler 2011, 2012; Ogden and Halliday‐Boykins 2004; Sundell et al. 2008; Van der Stouwe et al. 2014). MST decreases youth problem behavior including criminal behavior and substance abuse through working directly with families and improving family functioning. However, to our knowledge, except two studies (Halliday-Boykins et al. 2004; Mertens et al. 2017), most of the extant studies evaluated the success of MST in reducing serious problem behavior by examining the average changes in the participants. However, the variations across individuals should also be examined to better understand how MST works or for whom. Halliday-Boykins et al. (Halliday-Boykins et al. 2004), and Mertens et al. (Mertens et al. 2017), on the other hand, examined the trajectories of change in adolescents’ externalizing and/or internalizing behaviors who received MST, and found various subgroups responding differently to treatment. For example, Mertens et al. (Mertens et al. 2017) identified six groups of youth, four of which showed a positive effect of MST with decreases in externalizing behaviors, while two of which displayed poor treatment response by showing either no change or increase in externalizing behaviors. Halliday-Boykins et al. (Halliday-Boykins et al. 2004), on the other hand, identified five trajectories following psychiatric crisis, based on change in externalizing and internalizing symptoms as a result of either MST or psychiatric hospitalization. A substantial proportion of youth representing three trajectories displayed sustained high levels of symptoms, while two trajectories showed a decrease in the symptoms. However, their sample consisted of youth with severe psychiatric symptoms requiring hospitalization, hence not representing the target group of standard MST directed at youth with externalizing problems (Mertens et al. 2017).

In the current study, as a part of continuous quality improvement, we investigated the heterogeneity in how youth with serious problems in Norway did respond to MST in terms of ultimate MST outcomes (i.e., living at home, attending school/work, not in trouble with the law, abstaining from substance abuse, and abstaining from violence) over time. We further explored whether immigrant background, in addition to gender and age of the youth at intake, predicted belonging to the subgroups.

Multisystemic Therapy (MST) is an evidence-based, intensive family- and community-based treatment program targeted at youth aged between 12 and 18 years with serious behavior problems, which may include delinquency and substance abuse (Henggeler et al. 2009; MST Services Inc. 2020). It is based on social-ecological theory (Bronfenbrenner 1979), and addresses known risk and protective factors at the youths’ various social arenas such as home, school, peers, and community. The MST model includes integrated, explicit implementation requirements and strategies, such as directions for organizational support, and routines and systems for evaluation and continuous quality improvement. In Norway, MST has been implemented nationally, guided by the Ministry of Child and Family Welfare, and supported by a national implementation team at The Norwegian Center for Child Behavioral Development (NUBU) tasked with evaluating and maintaining the quality of the implementation.

The MST team and therapist are accessible twenty-four hours a day, seven days a week, and engage the parents and the youth in continuous prioritization of goals, analysis of problem drivers, development and implementation of interventions, and evaluation of outcomes. The treatment goal is to reduce the youth’s problem behavior through empowering parents, increasing family warmth, cohesion, and responsible behavior within and outside the family, hence reducing risk factors for the youth’s further development, averting placement out of home, and setting the youth off on a positive developmental trajectory.

Even though positive outcomes of MST have been found in several studies, the effects might still be moderated by various factors including the characteristics of the youth such as gender, age and immigrant status. While earlier MST studies generally found no moderating effects of gender, age, or ethnicity, researchers are still encouraged to include such variables especially with the move of these programs into communities (Ogden and Hagen 2006).

Research reveals gender related similarities and differences in both manifestations and development of antisocial behavior (see Ogden and Hagen 2009 for detailed discussion). Gender differences in MST outcomes, on the other hand, have not been examined as a central topic in MST efficacy (Ogden and Hagen 2009). Among the limited number of studies, a Norwegian RCT follow-up study (Ogden and Hagen 2006) investigated gender differences in treatment outcomes and showed that generally MST works similarly effective for girls and boys, but is particularly more effective at keeping boys and older ones at home. In another multi-informant study (Ogden and Hagen 2009), gender differences emerged for some treatment outcomes and when reported by some informants. For example, girls showed better results in parent reported externalizing problems and self-reported delinquency, while boys reported higher reduction in internalizing problems at discharge (Ogden and Hagen 2009). On the other hand, teacher ratings revealed equally beneficial outcomes of MST for boys and girls. Hence, gender may emerge as a significant predictor of MST outcomes, for some outcome areas and when reported by some informants.

MST has been argued to be a culturally competent intervention (Brondino et al. 1997). However, cultural differences that families are embedded in, such as culturally appropriate parenting behaviors, youth-parent relations, and acculturation process, may create a variation on how youths and families with immigrant and non-immigrant background engage in and respond to MST (Fox et al. 2017). Van der Stouve and colleagues (Van der Stouwe et al. 2014) conducted a meta-analysis on the effectiveness of MST building on previous meta-analytical reviews (Curtis et al. 2004; Littell 2005), and found a small but significant treatment effect on delinquency, psychopathology, substance use, out-of-home placement and so on. This meta-analysis also examined under which conditions MST was the most effective. While no difference for the effect of MST was found for the proportion of boys across studies, larger effect sizes were demonstrated with an average participants' age of under 15 years and in studies with a larger proportion of Caucasian juveniles (Van der Stouwe et al. 2014). This indicates that MST is more effective with younger and non-ethnic minority youth. With regard to age, possible explanations for effect in favor of younger children was that older children may have a longer history of confrontations with their parents, development of more dysfunctional and coercive interaction patterns in the family, in addition to the increasing academic and behavioral demands in school with older age (Ogden and Hagen 2008). With regard to immigrant background, this meta-analytic effect was inconsistent with that of the primary studies (e.g., Henggeler et al. 2002) testing for moderating effects of ethnicity and reporting no such effects, possibly due to the low power in the individual studies. In another meta-analysis (Jo Wilson et al. 2003), the findings of mainstream programs (i.e., not culturally tailored for minority youth) for juvenile delinquency were synthesized to examine whether these programs had different outcomes on their subsequent antisocial behavior, academic performance, peer relations, and so on. Their results revealed the positive and equal effect of these programs for both minority and majority youth. Hence, more research is needed to examine the differences in treatment outcome between minority and majority youth.

This study is an attempt at further investigating potential predictors of differences in response to MST treatment in a homogenous treatment situation (i.e., the Norwegian Child Welfare System). In sum, we explored subgroups of youth classified by their response to MST for each ultimate MST outcome over time in Norway and whether immigrant background, in addition to gender and age of the youth at intake, predicted belonging to the subgroups.

In this study, extant research on Multisystemic Therapy (MST) was enriched by examining and identifying the subgroups of youth who responded differently to MST in terms of ultimate MST outcomes (i.e., living at home, attending school/work, not in trouble with the law, abstaining from substance abuse, and abstaining from violence). We further investigated whether the immigrant background, age at intake and gender predicted belonging to the subgroups that were identified.

Our results showed that there were two to four subgroups of youth for different MST outcomes. For each outcome, one large subgroup (the sustained improvement group), ranging from 68% to 84% of the overall sample, maintained the positive status or did not deteriorate after discharge as well as over time. Overall, approximately half of the youth (47.8%) belonged to the sustained improvement group across all outcomes, hence showing stable improvement for all MST outcomes. There were also two main groups that emerged for almost of all of the outcomes, a gradual improvement group (except for the ‘living at home’ outcome) and a gradual deterioration group (except for the ‘not in trouble with the law’ outcome) with the youth ranging approximately from 12% to 29%, and from 3% to 10% of the overall sample, respectively. Lastly, there were two small groups with deterioration emerged after discharge for ‘living at home’ outcome (i.e., a group in which more than half of the youth at 6-month follow-up, and all of them at the subsequent measurements, were placed out of home by the Child Welfare Services), and after 6-month for ‘abstaining from substance abuse’ outcome (i.e., a group in which more than half of the youth at 12-month follow-up, and all of them at 18-month follow-up, used substances in a way that impaired daily functioning or led to other serious consequences). These deterioration subgroups consisted of approximately 6% and 2% of the overall sample, respectively. Hence, the overall results indicated a high and sustained degree of improvement across the ultimate outcomes for the vast majority of the youths. However, even though a large group of the youth showed a trajectory with sustained improvement over time, there was still variation in the groups emerged with improvement and deterioration trajectories for various outcomes.

For all outcomes, except ‘living at home’, the next largest subgroup emerged was the gradual improvement one. This finding may point to a delayed treatment effect and this is in line with one of the main MST treatment principles. This principle emphasizes generalization of treatment interventions, so the family will be able to maintain and also potentially improve outcomes after treatment has ended. This may also reflect systemic ripple effects of treatment, generating increased contextual reinforcement of positive behaviors. Taken together, the sustained improvement and gradual improvement groups leave only a small proportion of the youths in groups showing deterioration after treatment. For the ‘living at home’ group, on the other hand, there was no gradual improvement group. This is due to the fact that nearly all youths were living at home at the beginning of treatment. The remaining fraction was moving home from a child welfare institution at the beginning of treatment. Nevertheless, a relatively larger proportion of the youths, approximately 16%, displayed some deteriorating pattern for this outcome compared to the other outcomes.

Following the identification of the subgroups, we further explored whether immigrant status, age at intake and gender predicted these various trajectories. Immigrant status did not differentiate the subgroups for any of the outcomes. We expected variation between immigrant and nonimmigrant youth due to a range of factors such as overall socioeconomic level, trust in and skepticism towards child welfare services (e.g., Tembo et al. 2020), language barriers for minority youths, in addition to cultural differences that families are embedded in. The absence of immigrant status predicting membership of any particular class in all of the outcomes, however, may indicate that MST worked just as well for immigrant youth as for non-immigrant youth, as MST is a culturally competent intervention (Brondino et al. 1997). This may possibly reflect the flexibility in tailoring interventions according to the needs of each specific youth and family, which is a fundamental principle of MST. A recent qualitative study revealed that when MST therapists consider cultural differences and acknowledge them sensitively, and act as a cultural broker especially between youth and family, immigrant youth and families engage and change more during MST treatment, hence resulting in successful MST effect (Fox et al. 2017). Our finding is also consistent with the results of a meta-analysis demonstrating the effectiveness of mainstream service programs for various outcomes equally for ethnic minority and majority juvenile delinquents (Jo Wilson et al. 2003), even though the use of different programs and outcomes prevents a direct comparison with our results. It is, however, worth to mention that immigrant youth in our study may be a heterogeneous group, with diverse cultural and socioeconomic backgrounds. However, insufficient power in our study due to the small numbers per immigrant generation status or origin did not allow us to further examine differences among immigrant youth group. So, we should be aware of the fact that comparing a non-immigrant, native group of youth with a mix of various immigrant groups may lead to over-interpretation of the differences, and in addition may conceal large differences amongst immigrants (Anderson and Mayes 2010; Stevens and Vollebergh 2008).

Age at intake, on the other hand, was a significant predictor of belonging to the subgroups for all of the ultimate outcomes except ‘living at home’ outcome. Except for the ‘abstaining from violence’ outcome, older youth were less likely to belong to the sustained improvement trajectory. This is in line with the results of the meta-analysis showing that MST is more effective with younger youth (Van der Stouwe et al. 2014). In our study, we may also infer that while younger youth were more likely to show sustained improvement, older youth displayed a relatively poorer improvement across most of the ultimate outcomes and this may inform the way MST should be delivered. First, it underscores the need for early intervention. A prolonged history of serious behavior problems increases the risk for severe marginalization processes, increased affiliation with negative peers, and loss of positive developmental experiences on the various social arenas – thus also generating a greater risk of entering into a detrimental trajectory during adult life. MST emphasizes the parents as main drivers of change, whereas for older youth, there may be a need for increased work with youth involvement in therapy, and for more individual work with the youth.

On the other hand, older youth had a greater probability of belonging to the sustained improvement trajectory for the ‘abstaining from violence’ outcome. This may be a reflection of increased neurological maturation, associated with improved executive control of attention and self-regulation capacity – which may play a greater role regarding violence than for the other outcomes (Gardner et al. 2008; Spear 2000). However, we do not have any data to examine this, hence future studies may explore this. Also, we cannot rule out the fact that some of the youths in the dataset belong to the ‘late onset’ developmental trajectory (Moffitt and Caspi 2001), where problem behavior spontaneously decreases as the youth gets older. Moreover, our data is limited, as this variable covers a range of phenomena, such as not distinguishing between physical violence, and threats of physical violence. This may confound potential systematic changes with age. So, future studies are welcomed to investigate this in detail.

While there was no difference for boys and girls in predicting group membership for ‘living at home’ and ‘attending school/work’ outcomes, girls, on the other hand, were more likely to belong to the sustained improvement group for the other three outcomes (i.e., not in trouble with the law, abstaining from substance abuse, and abstaining from violence). It is interesting to note that these are the three outcomes which are more likely to be related to externalizing behaviours, and in the literature several findings suggest that girls tend to display lower levels of behavioral problems than boys (e.g., Bongers et al. 2004; Moffitt and Caspi 2001; Ogden and Hagen 2009). Since our study is the first exploring various subgroups of ultimate outcomes and the role of gender in predicting these trajectories over time, we do not have any literature to compare our findings. Moreover, as gender differences in MST outcomes have not been examined as a central topic in MST efficacy (Ogden and Hagen 2009), our finding is in line with the Norwegian RCT follow-up study (Ogden and Hagen 2006) which demonstrated gender as a significant predictor for some MST outcome areas and when reported by some informants.

Some limitations to this current study should be mentioned. In this explorative study, we examined a limited number of demographic predictors; and we were not able to investigate various individual (e.g., social competence, aggression), contextual (e.g., parental factors, peer characteristics, family characteristics) or implementation-related (e.g., the site and/or the therapist characteristics, alliance with therapist) factors subject to change in predicting various trajectories. Examining the role of these changeable factors may provide valuable information on how to improve treatment response by addressing and manipulating these factors early in treatment (Mertens et al. 2017). Moreover, we did not examine the potential interaction among our predictors such as the interaction of age and gender. Even though our sample size was relatively large, due to the LCGA analyses and emergence of small groups, there was still a lack of power in our dataset to detect interaction effects for any of our small groups. The ultimate goals were evaluated as dichotomous variables, with an either-or response; hence, variation and range in these outcomes (e.g., a youth who breaks the law many times versus another one who does once) cannot be captured. With regard to the data in this study, it should also be mentioned once again that the data for this study was not originally collected as research data, but as the program data collection registered by different MST teams, as a part of the continuous quality improvement system for MST in Norway. The program data were also collected from different data sources for the different time points. That is, while data at intake and discharge involved MST team assessment, follow-up data involved parents’ assessments. Moreover, the ultimate outcomes were measured at all-time points only for MST completers. Hence, it is important to emphasize that the potential generalizability of the study pertains to completers (i.e., 88.4% of the cases ending treatment in MST Norway during the period in question), and not to all youth and families going through MST treatment. Non-completers were not included as they were not asked to submit follow-up information due to ethical considerations. However, additional attrition analyses between completers and non-completers revealed that even though there was not a significant gender difference between these two groups, there was a tendency for a reduced completion rate for older youths, and immigrant families tended to complete treatment at a lower rate than non-immigrant families. Hence, our results should be interpreted with caution in the light of these limitations of our sample’s representativeness. Last but not least, the generalizability of our findings is limited by a lower prevalence of antisocial disorders in the Nordic countries compared to other countries worldwide (Skogen and Torvik 2013).

Despite these limitations, this study still adds valuable information to our knowledge about how youth respond to MST. Our results suggest that not every youth responds similarly to MST, and there is heterogeneity in their response over time as well as the predictors of the subgroups of youth across different outcomes. More studies examining heterogeneity will possibly help us to identify factors that may be targeted in order to better tailor the MST interventions for youth with serious problems.