Introduction
Treatments that meet the needs of young people who engage in criminal behavior and reduce reoffending are critically important. This is underscored by the fact that a relatively small group of youth are responsible for a disproportionate number of juvenile offenses, including violent offenses, many of whom recidivate and progress to the adult justice system (Barrett & Katsiyannis,
2016; Garrido & Morales,
2007; Mulder et al.,
2019; Sutherland & Millsteed,
2016). Moreover, alongside their offending-related treatment needs, justice-involved youth often present with other clinical conditions, including mental health and substance use disorders (Beaudry et al.,
2021; Meurk et al.,
2019). Left untreated, these problems can undermine a young person’s responsivity to interventions targeting criminal behavior. Supported by a robust research literature, the Risk-Need-Responsivity (RNR) model identifies the types of treatments that, all else being equal, most likely lead to reduced offending in general, and youth offending specifically (Bonta & Andrews,
2017): namely, treatments that match the young person’s empirically assessed risk level (i.e., more intense and extensive services are afforded to higher risk cases), target dynamic risk factors known to be associated with reoffending, adopt cognitive behavioral principles, and take an individualized, tailored approach that considers relevant responsivity issues (e.g., adapted to learning abilities, motivation level, trauma histories, culture, colonization and social marginalization experiences). Yet, a sizeable portion of justice-involved youth drop out of treatment, and treatment effects for those retained remain modest (Olver et al.,
2011; Pappas & Dent,
2021). These treatment failures underscore the importance of knowing how service providers should deliver evidence-based treatments to maximize benefits to youth, their families, and the community. One therapeutic process variable, the therapeutic alliance, has received little attention in the juvenile justice realm compared to conventional psychotherapy and is the focus of this review.
The therapeutic alliance denotes some of the most important elements of the relationship between a client and therapist. Within the adult psychotherapy literature, the alliance is commonly conceptualized as comprising three interrelated components: (1) mutual agreement and understanding regarding the
goals of therapy; (2) clear definition and negotiation of the
tasks necessary to achieve these goals; and (3) the development of an affective
bond or mutual trust between the parties (Bordin,
1979). Prior reviews consistently link strong therapeutic alliances to positive treatment outcomes across various treatment modalities (e.g., Baier et al.,
2020; Fluckiger et al.,
2018; Horvath & Bedi,
2002).
Alliance research in child and adolescent populations has progressed more slowly than the adult field. To date, two leading youth alliance conceptualizations have guided much of the research and development of measures to assess youth alliance (Karver et al.,
2018). Drawing from psychodynamic perspectives (Freud,
1946; Meeks,
1971), the first conceptualization proffered by Shirk and Saiz (
1992) emphasizes the alliance as a collaborative bond (captured via two dimensions of bond and task collaboration) between the youth and therapist. The second major conceptualization, which derives from Bordin’s (
1979) model, emphasizes the youth alliance as a contractual bond comprising three interrelated dimensions of goals, tasks, and bond (DiGiuseppe et al.,
1996). Although both perspectives propose multidimensional structures, findings have been mixed about whether these dimensions are empirically replicated in youth samples, with some work supporting a unidimensional structure (Faw et al.,
2005; Ormhaug et al.,
2015; Roest et al.,
2016; Shelef & Diamond,
2008). This highlights that the core features of the alliance may be less differentiated for young people than adults (Shirk et al.,
2011). Irrespective of the underlying structure, meta-analyses of child and adolescent clinical populations demonstrate small-to-modest associations between the alliance and treatment outcomes (Karver et al.,
2018; McLeod,
2011; Shirk et al.,
2011; Welmers-van de Poll et al.,
2018).
Several challenges and considerations are relevant to therapeutic work with offending/justice-involved youth that might impact alliance formation and its relation to outcomes. Some are shared with young people in general (e.g., underdeveloped cognitive capacities, involvement of other family members, desire for increased autonomy from adults; DiGiuseppe et al.,
1996; Karver et al.,
2018; Shirk et al.,
2011; Zack et al.,
2007), while others are unique to or magnified by the justice context. For example, offending and justice-involved youth often possess shallow or only external treatment motivation, significant emotion regulation difficulties, high levels of hostility, severe conduct problems, and deficits in social-cognitive skills (Docherty et al.,
2021; Kapoor et al.,
2018; Tarolla et al.,
2002). In addition, rates of child maltreatment, trauma, family disadvantage, disability, and mental illness are all disproportionately high in this population (Baidawi & Piquero,
2021; Beaudry et al.,
2021; Fox et al.,
2015; Papalia et al.,
2022). Such characteristics may represent barriers to forming and maintaining productive therapeutic relationships. Similarly, young people from racially, culturally, and linguistically diverse backgrounds are over-represented in justice systems internationally. The unique contexts and stressors experienced by many of these youth and their families may be consequential for the therapeutic alliance and for clinicians’ efforts to develop trust and (cultural) safety when working cross-culturally.
Although client-therapist collaboration in setting the goals and tasks of treatment is critical to developing a strong alliance, in the case of offending behavior treatment programs, the goals and tasks of therapy are typically pre-determined with community safety considerations at the fore (Kozar & Day,
2012). Relatedly, while some justice-involved youth actively seek help, many do not self-refer to programs targeting offending, substance use, or mental health. Rather, they may be mandated, coerced, or pressured to receive treatment (Hachtel et al.,
2019; Parhar et al.,
2008; Smallbone et al.,
2009). Clinicians working with youth in justice contexts may need, therefore, to balance dual roles: on the one hand being an agent of change, encouraging the young person to develop a trusting and self-disclosing relationship with them, but on the other hand being a figure of control, where they may disclose risk-relevant information to third parties about the young person (Kozar & Day,
2012; Skeem et al.,
2007; Ward et al.,
2015). Even if a therapist does not hold a dual role, the young person may view them as part of a coercive system, undermining the alliance.
Another complicating factor is that treatment approaches for justice-involved youth often include group interventions (or settings/environments) and family treatments (Kozar & Day,
2012; Lipsey,
2006; Pappas & Dent,
2021). While the therapeutic alliance–outcome relationship has been supported in group and family-involved therapies, some evidence suggests that the alliance may interact with other group (e.g., cohesion and climate) and family processes (e.g., parent alliance) in a complex interplay that influences treatment outcomes, both positively and negatively (Alldredge et al.,
2021; van der Helm et al.,
2009; Welmers-van de Poll et al.,
2018). Similarly, it is not unusual for multiple therapeutic (or support) staff to be involved in caring for justice-involved youth, each relationship of varying importance and quality. It is plausible that how the alliance develops and whether it is linked to change is influenced by the complex array of relationships the young person (and their family) forms with their care team.
Recognizing the complexities of therapy in criminal justice contexts, Ross et al. (
2008) revised Bordin’s (
1979) alliance concept to provide a more elaborate theory of how the alliance develops and is maintained in therapy with offending adults. In the revised model, therapist characteristics (e.g., personality, professional/interpersonal skills, biases, expectations of the client), client variables (e.g., irritability, callousness, attachment insecurity, treatment readiness), and their interaction (e.g., (dis)similarity in values, matched cultural backgrounds), are theorized to influence therapist and client cognitive processes and emotional reactions to each other and the therapy process. These processes and responses manifest as in-therapy behaviors that directly feed into the bond, goals, and tasks dimensions of the alliance, and which are also affected by the broader context in which therapy is delivered (Ross et al.,
2008). Contextual factors might include, for example, justice system/organizational policies, availability of pleasant and safe therapy spaces, level of therapist supervision and reflective practice, whether treatment is delivered in a group setting, and so on. Later, Orsi et al. (
2010) extended Ross et al.’s model to elucidate additional factors that might influence the working alliance
1 for adolescents in formal “authoritarian” settings (e.g., child welfare/residential care settings, substance use treatment, probation). Key additions included the potential role of youth developmental stage and their social networks (e.g., families, peers, school) in how the working alliance forms.
Despite the clinical coherence of the abovementioned models, to our knowledge, there are no published systematic reviews of empirical work concerning the determinants of the therapeutic alliance with offending/justice-involved youth. Further, despite the robust association between the alliance and treatment change in conventional psychotherapy, very little is known about the alliance’s role in creating change in justice-involved youth. A review of research on the alliance in adult violent offending behavior programs concluded that while there are clear theoretical and practice grounds for developing a strong alliance, there is insufficient data to determine whether it directly or indirectly impacts treatment outcomes in this population (Kozar & Day,
2012). It is unknown whether conclusions from the broader youth therapy literature or the adult correctional treatment literature about the nature and value of the alliance can be applied to treatments with justice-involved youth.
To address these knowledge gaps, we conducted a systematic review of studies investigating the therapeutic alliance in treating offending/justice-involved youth. Our primary objectives were to synthesize quantitative research on (1) the determinants of a positive therapeutic alliance; and (2) the relationship between alliance quality and treatment outcome. In addition, we sought to synthesize qualitative research on the perspectives of young people, their caregivers, and treatment providers about the nature and role of the alliance in this context. This qualitative strand was more exploratory and aimed to include, for example, views about the features of a positive alliance, factors that facilitate or hinder alliance formation, and the relevance and value of the alliance in generating therapeutic change.
Methods
This review was prepared using the Preferred Reporting Items for Systematic reviews and Meta-Analyses guidelines (PRISMA; Page et al.,
2021) and the Enhancing Transparency in Reporting the Synthesis of Qualitative Research statement (ENTREQ; Tong et al.,
2012). Although the review was not registered, review methods were determined a priori and written into a protocol, available from the authors.
Eligibility Criteria
Population
The target population comprised adolescents (i.e., aged 10–17 years, or present in a juvenile facility or service) with justice involvement and/or a history of criminal behavior (i.e., behavior that would be grounds for arrest, irrespective of whether criminal sanctions were imposed). In most instances, justice involvement and offending were derived from official sources (e.g., incarceration, convictions), with a minority of studies using other sources (e.g., self- or informant-reported delinquency and violence). Where studies used mixed samples (e.g., justice-involved and child welfare-involved youth), we required that ≥ 75% comprised justice-involved/offending youth or that analyses were reported separately for this subgroup. The following studies were excluded: samples with a mean age below 10 years; samples defined by general or composite constructs like externalizing/disruptive behavior, aggression, and substance misuse, that did not clearly meet population criteria; and samples with primarily developmental status offenses or “soft delinquency”, that is, non-criminal offenses that are law violations due to minor status (e.g., truancy, running away from home, violating curfew).
Study Designs/Settings and Evidence Sources
All records were required to describe original empirical research. Studies that used quantitative or qualitative methods were eligible for inclusion, but we excluded case studies. All types of study/treatment settings were eligible. Both peer-reviewed research studies and dissertations were eligible for inclusion, providing they were in full-text form and written in English. No limits were placed on publication year.
Therapeutic Alliance
Quantitative studies must have included a measure of the therapeutic alliance (or treatment alliance, helping alliance, working alliance) or one or more of its key components (e.g., mutual affective bond, collaborative goals, and/or task orientation). Furthermore, the associations relevant to the review objectives had to be tested statistically. For qualitative studies, eligible studies were required to either directly inquire about the alliance or identify the alliance (or one or more of its components) as a resultant theme from participant responses. We did not place restrictions on the method or timing of alliance assessment.
Only studies where the target therapeutic alliance occurred in the context of therapeutic/clinical treatment were eligible for inclusion, i.e., we excluded non-therapeutic contexts (e.g., mentoring programs, employment/educational programs). Although we did not place limits on target treatment problems (e.g., offending behavior, mental health, substance use), there must have been a broadly stated aim to address problematic/antisocial behaviors, reduce psychological distress, or increase prosocial and adaptive functioning. Eligible target alliances were between the practitioner(s) providing the treatment and the young person or parent/caregiver receiving the treatment (i.e., youth–therapist alliance or parent–therapist alliance). Correctional relationships (e.g., with correctional officers, youth justice case managers) that served a surveillance function rather than, or alongside, a rehabilitative role were not eligible for inclusion.
Alliance Determinants and Outcomes
For studies with analyses relevant to our first review objective (factors associated with alliance quality), we took an inclusive approach and considered all potential factors reported in studies. Only factors measured prior to or concurrent with the alliance were eligible. Studies relevant to our second objective (alliance–outcome relationship) were required to measure and report at least one outcome variable. Primarily, we were interested in ‘treatment’ outcomes; that is, those factors directly targeted by treatment or might otherwise be reasonably expected to change because of treatment. However, given the anticipated limited literature concerning the alliance in justice-involved youth, along with the challenges associated with effectively engaging this group in treatment, we also included studies examining ‘process’ outcomes (e.g., dosage, treatment completion, therapist adherence). We did not require temporal separation between alliance and outcome measures; both prospective and concurrent associations were eligible.
Literature Search Strategy
To locate eligible studies, we searched several databases from inception to March 3rd, 2021: PsycINFO; MEDLINE; Criminal Justice Abstracts with Full Text (EBSCOhost); ProQuest (Social Science Premium Collection); and CINCH Australian Criminology Database. For each database, we used the following search terms: “offen*” OR “violen*” OR “aggress*” OR “delinquen*” OR “criminal behavio?r*” OR “justice-involved” AND “therapeutic alliance” OR “working alliance” OR “helping alliance” OR “therapeutic relationship” AND “youth*” OR “juvenile*” OR “adolescen*” OR “young people*” OR “young offender*” OR “teenager*”. After the initial search, the references of previous systematic reviews on the youth alliance–outcome relationship were examined as well as references from potentially suitable papers. A forward citation search (in Scopus) was conducted of three existing reviews and two oft-cited studies relevant to the alliance in forensic and justice settings (Florsheim et al.,
2000; Holmqvist et al.,
2007; Kozar & Day,
2012; Orsi et al.,
2010; Ross et al.,
2008). The search strategy was updated on February 4th, 2022, to include any eligible studies published since the original search.
Study Selection
Study selection was undertaken by two authors (N.P. and A.D.) using Covidence systematic review software. After removing duplicate records, both reviewers independently screened 93% of titles and abstracts to remove irrelevant documents, with 90% agreement (the remaining 7% of records were identified via the updated search and screened by N.P. only). Disagreements were resolved through discussion, with decisions favoring an inclusive approach at this review stage. Next, full-text articles of screened-in records were obtained and assessed against inclusion criteria. A sample of 24 full-text records (21%) was independently reviewed by both authors, with 92% agreement. Decisions on the remaining full-text records were made by one author (N.P.). Where there was insufficient information to determine study eligibility, attempts were made to contact authors; where unsuccessful, ambiguous records were excluded.
Two data extraction forms were developed (one for quantitative and qualitative studies), including all variables for which data were sought from primary studies. To establish the forms, content areas of interest were identified and items were developed to assess these areas. Next, pairs of reviewers (N.P., A.D., N.M.) independently piloted the coding forms using a small subsample of included records (n = 3), with 88% agreement. All items were reviewed, discrepancies were resolved, and poor coding items were revised to improve accuracy and agreement. The final coding forms examined several content areas: author and study descriptors (e.g., year, location, design); sample descriptors (e.g., youth age, gender, race/ethnicity); treatment descriptors (e.g., target problem, treatment model, setting, duration); alliance descriptors (e.g., alliance dimension(s), target relationship, rating source, measure(s), timing); alliance determinants and outcomes (e.g., variables/outcomes assessed, rating source, measure(s), timing); and a summary of key analyses/findings and authors’ conclusions. Finally, authors N.P. and N.M. extracted data for all remaining quantitative and qualitative studies, respectively.
Quality Assessment
The Mixed Methods Appraisal Tool (MMAT) Version 18 (Hong et al.,
2018) was used to appraise included studies critically. The tool consists of separate sections for quantitative and qualitative methodologies. Quantitative analyses were assessed across five methodological criteria: sample representativeness; measurement of the independent and dependent variables; completeness of outcome data; confounding influences; and, where applicable (longitudinal studies only), changes in exposure (independent variable) status over the study period and potential co-exposures. Each criterion was rated as ‘Yes’ (indicating sufficient methodological quality such that plausible issues would be unlikely to alter results seriously), ‘No’ (indicating insufficient methodological quality such that plausible problems seriously weaken confidence in the results), or ‘Cannot Tell’ (the study did not report appropriate information to answer ‘Yes’ or ‘No’ or reports unclear information). The appraisal of qualitative studies involved assessing the congruence of the aims of each study to five domains, namely, qualitative methodology, methods, analysis, findings, and reporting.
In accordance with MMAT recommendations, we developed a list of specific indicators for each domain and applied these uniformly across all quantitative and qualitative studies, respectively. Quality criteria were rated at the review objective level. Two authors independently rated quality for four quantitative studies (N.P. and A.D., with 80% agreement) and two qualitative studies (N.M. and N.P., with 100% agreement). Disagreements were resolved through discussion, and rating indicators were clarified and expanded where necessary to improve consistency. The remaining studies were appraised by one author (N.P. or N.M.).
Results Synthesis
Regarding quantitative research, we anticipated heterogeneity in the types of treatment programs under consideration and alliance determinants and outcomes. As such, we planned to narratively synthesize findings rather than conduct meta-analyses. Results were synthesized first according to review objective and then by key variable domains. Qualitative research was synthesized using thematic analysis. This was initiated with line-by-line coding of results/findings text of primary studies. Free codes were then organized into descriptive themes. Interpretation resulted in the development of overarching analytical themes across studies that went beyond the interpretation in the primary analyses. Further organization resulted in a synthesis of findings represented by themes and sub-themes, illustrated in tabular form, and elaborated on in a descriptive narrative.
Discussion
This review provides the first systematic synthesis of research concerning the therapeutic alliance in treating justice-involved and offending youth. We identified 23 independent studies meeting inclusion criteria, most of which were quantitative studies (61%) conducted in the US (74%). Males and racial/ethnic minority groups were over-represented, consistent with their disproportionate contact with juvenile justice systems. The synthesis of nine qualitative studies generated meaningful themes related to the development of a constructive alliance and the potential role of the alliance in initiating treatment change. The perspectives of young people were included in two-thirds of qualitative studies, with a small number drawing on therapist and parent views. The synthesis of 14 quantitative studies of determinants of alliance quality and its association with outcomes yielded mixed findings overall. Nevertheless, drawing from both syntheses, the review highlights several pertinent findings.
Factors Linked to Alliance Formation
Contrasting with prior work that notes externalizing problems may hinder the alliance (Shirk & Karver,
2003), we found limited evidence to suggest antisocial problem severity predicted alliance difficulties in multivariate analyses. This discrepancy likely reflects differences in the reference group. In prior work, externalizing children are typically compared to internalizing children (e.g., Ayotte et al.,
2015) whereas in this review, we focus only on offending/justice-involved youth. Interestingly, there was some evidence for a positive link between antisocial behaviors and youth/parent alliance quality, including among youth higher in CU traits. While some suggest this could indicate superficial alliances among high CU-youth (Simpson et al.,
2013), others suggest it may reflect their increased interpersonal proficiency, verbal abilities, and emotion regulation skills (i.e., fewer barriers to alliance formation; Mattos et al.,
2017). The review also draws attention to potential differences in the impact of problem severity on family-therapist emotional bonding across race and ethnicity (Ryan et al.,
2013).
Consistent with theory on alliance formation (Orsi et al.,
2010), the review highlights the potential relevance of the young person’s relational frameworks and social systems to therapeutic relationships. Positive peer and parent attachments were linked to positive youth perceptions of alliance quality (Bovard-Johns et al.,
2015), and deviant peer relationships were associated with negative perceptions (Florsheim et al.,
2000). Further, parents’ involvement in the young person’s life (e.g., in school, monitoring, limit setting) and therapy seemed to relate to better early alliances (Hogue et al.,
2006; The MVPP,
2014), as did including a treatment focus on peers (Dauber,
2004). Thus, justice-involved youth with poor parental attachments, who are deeply embedded in deviant peer groups, and with family-related criminogenic needs, may be among the more resistant to treatment and the development of an alliance. These findings broadly align with the evidence on the effectiveness of multi-systems approaches for antisocial youth (Pappas & Dent,
2021) and extend this to suggest potential positive flow-on effects of these approaches for alliance formation. No studies examined the impact of youth attachment/interpersonal styles on the therapeutic bond, despite the disproportionately high rates of attachment-related abuse/neglect and interpersonal difficulties in this population (Modrowski et al.,
2021). One study found that higher sexual abuse trauma symptoms predicted poorer alliances (Bovard-Johns et al.,
2015), potentially via mechanisms of diminished trust and safety in relationships. The synthesis of qualitative studies highlighted the importance of attending to trust, safety, and security to facilitate an affective bond.
If young people (and their families) do not believe the treatment will benefit them, they are unlikely to enter a treatment relationship with enthusiasm (Kozar,
2010). It is unsurprising then that the review found youth with low readiness to change and negative perceptions of the utility of treatment and its “fit” with their ideas about change generally reported poorer alliances (Holmqvist et al.,
2007; Savicki,
2008; The MVPP,
2014). This is consistent with research with justice-involved adults, where readiness is shown to be one of the strongest predictors of alliance (Taft et al.,
2004). One interpretation is that alliance difficulties may partly mediate the link between lower treatment readiness/motivation and poorer treatment outcomes (Higley et al.,
2019). Another perspective is that stronger alliances may improve young people’s perceptions of the value of treatment, their capacity to benefit, and their intrinsic motivation. A related theme arising from the syntheses was the importance of young people perceiving treatment goals and therapeutic tasks as helpful. Central to this was the collaborative and practically oriented aspects of the therapeutic relationship, where adolescents (and families) held more positive views of therapists who included their ideas about what they felt needed to change in goal setting, who helped them practically, and who maintained a focus on the ‘work’ of treatment (Holmqvist et al.,
2007; see also Theme ii of qualitative results synthesis).
Despite the lack of attention paid to the role of therapist characteristics in quantitative studies, several qualitative studies identified the importance of therapists’ traits and skills in promoting positive alliances (see Theme ii). For example, qualities like empathy, genuineness, honesty, humor, flexibility/responsiveness, being ‘straight up’ (direct) with them, and a non-blaming, strengths-reinforcing approach were generally valued by adolescents.
Alliance–Outcome Relationship
There was considerable variation in the impact of alliance on treatment outcomes for justice-involved youth. Both substantive (e.g., treatment modality and dose, types of outcomes, alliance type) and methodological characteristics (e.g., source of alliance and outcome assessment, timing of measures) appeared relevant to such variation. Most studies focused on antisocial outcomes (e.g., externalizing behaviors, substance use, institutional misconduct, recidivism), where most alliance effects emerged. More rigorous research is needed to understand the relations between the alliance and other clinical outcomes (e.g., mental health symptoms, parent functioning) and treatment processes (e.g., attendance, drop out).
A common finding was that a stronger youth-therapist alliance early in treatment rarely predicted post-treatment declines in externalizing behavior and offending and at times predicted worse outcomes. Evidence suggested this may be explained by how the alliance evolves: youth who began with poorer alliances that improved during treatment (or had high stable alliances) showed reduced antisocial outcomes, whereas deteriorating (or low stable) alliances were linked to problem escalation (Florsheim et al.,
2000; Hogue et al.,
2006). Thus, youth who enter treatment with low readiness/motivation (and hence poorer alliances) who are supported to become more collaborative and work-ready may achieve comparable outcomes to youth who maintain consistently strong alliances (Hogue et al.,
2006). Conversely, alliance decliners, whether youth who experience an alliance rupture that is unrepaired or those who impression-manage early but become less inclined or able to sustain positive alliances as treatment demands increase, may be at risk of treatment failure (Florsheim et al.,
2000). Although the potential impact of alliance change is clinically intuitive and broadly consistent with prior work (Polaschek & Ross,
2010; Shirk & Karver,
2003; Welmers-van de Poll et al.,
2018), alliance shifts may be consequences rather than causes of therapeutic progress.
Only three studies examined specific features of the alliance in relation to outcome, with mixed findings. Simpson et al. (
2013) found no alliance effect (overall or for tasks, goals, and bond components) on violent institutional incidents. Holmqvist et al. (
2007) found staff ‘close feelings’ toward youth (i.e., affective relationship elements) were linked to worse offending outcomes. In contrast, youth perceptions of the treatment as useful (i.e., collaborative relationship elements) were associated with reduced offending. Another study found that the strength of the youth-therapist bond was not significantly related to any measured treatment outcomes (Glebova et al.,
2018). Although these findings might suggest differences in the relative importance of the collaborative and affective aspects of the youth-therapist alliance in generating change for justice-involved youth, the evidence base is too small and limited to draw firm conclusions. In the broader youth therapy literature, meta-analyses find that the alliance-outcome effect is not significantly moderated by alliance dimension (McLeod,
2011; Welmers-van de Poll et al.,
2018).
Family involvement and caregivers’ perceptions of alliance quality appeared relevant to the alliance–outcome relationship for justice-involved youth. Some work found that early/mid-treatment parent-therapist alliance better predicted improved antisocial outcomes than youth-therapist alliance (Glebova et al.,
2018; Hogue et al.,
2006). Other findings suggested the effect of youth alliance on antisocial outcomes may be enhanced in family-based treatment and when a strong parent alliance is developed (Hogue et al.,
2006; Shelef et al.,
2005). The risk of premature dropout may also be reduced where youth and parent alliances become increasingly balanced over time (Cosgrove,
2020). The parent alliance–outcome relationship has been demonstrated in the wider youth therapy literature (Karver et al.,
2018; McLeod,
2011). However, it may have particular importance in youth justice settings, where a parents' lack of involvement and negative beliefs about treatment are among the most common barriers to service provision for justice-involved youth (Kapoor et al.,
2018). Although there have been increased efforts to involve parents/caregivers, many youth justice and forensic mental health services lack resources or frameworks to actively engage families in treatment (Robertson et al.,
2019). Therefore, while the review findings require replication, they suggest increased research and practice focused on the role of parent alliance in treating justice-involved youth may be worthwhile, including how therapists and organizations can best facilitate these alliances.
Few quantitative studies compared the relative contribution of treatments’ technical and relationship aspects to outcomes with justice-involved youth. One exception showed that both were important in predicting outcomes following FFT, with alliance being more salient for clinical externalizing symptoms and fidelity for program retention and adjudicated arrests (Cosgrove,
2020). Although these findings reflect a single study, the synthesis of qualitative studies also supported the complementary role of treatment techniques and alliance in initiating and sustaining change (i.e., Theme iii). The bidirectional link between alliance and therapist fidelity found in this review further underscores the need to include technical and relational process measures to understand better how they interrelate and their unique and interactive effects on outcomes in this population.
Limitations and Future Directions
There are several limitations associated with our review. First, the number of included studies was small, and the evidence that emerged was often inconsistent. A greater volume of research is needed, which would be assisted by therapists and program evaluators including measures of the alliance in treatments with justice-involved youth. Second, the quality appraisal of studies identified several issues, particularly problems with sample representativeness (e.g., exclusion of treatment dropouts), confounding influences, limited consideration of whether a change in alliance reliably relates to a change in outcome, and a lack of detail about treatment fidelity. These issues could in part be addressed by researchers striving for random sampling, considering key confounds of their hypothesized relationships and controlling these through design/analysis, employing longitudinal designs with repeated measurements of both alliance and outcomes, and sufficiently detailing methods and treatment protocols. Third, alliance measures used were diverse (nine different measures across 14 quantitative studies) and not always widely validated. This diversity mirrors the variation in how researchers define youth alliance (Karver et al.,
2018), which likely contributed to the inconsistency of findings. More empirical and conceptual work is required to examine which measures and what features optimally capture the therapeutic alliance in youth justice settings. While this work progresses, researchers’ selection of alliance measures should ideally be guided by theory, how reliable and valid the tools are with youth (offending) populations, and prior research demonstrating their sensitivity to change. Assessment from multiple perspectives is also recommended. Overall, these methodological issues mean the review findings should be viewed as important preliminary data about the alliance in justice-involved youth that must be re-evaluated as more high-quality evidence emerges.
The review also highlighted some important knowledge gaps and avenues for future research, several of which have already been discussed. Regarding determinants of alliance quality, more research is needed examining the role of therapist characteristics and in-treatment behaviors, parent characteristics, and other treatment setting/contextual factors (e.g., group vs. individual therapy, telehealth vs. in-person interventions, level of actual or perceived coercion, institutional climate and safety, quality of relationships with other involved therapeutic or support staff). Justice-involved youth are likely to come from diverse cultural backgrounds, and so greater attention should be given to understanding whether and how culture, language, and perceptions of cultural safety affect the therapeutic alliance. Also lacking from studies was in-depth information from therapists' perspectives about how they form productive alliances with diverse clients and their families. Increased research focus in these areas would advance theory and knowledge and provide practical information for clinicians about strategies to build the alliance. Finally, more research is needed examining the mechanisms through which the alliance might work to influence outcomes, the features (e.g., affective versus collaborative) that may be most relevant to treatment change, and the conditions impacting whether this occurs.
Publisher's Note
Springer Nature remains neutral with regard to jurisdictional claims in published maps and institutional affiliations.