A Systematic Review of Therapeutic Process Factors in Mental Health Treatment for Autistic Youth

A systematic review of empirical research examining process factors in mental health treatment for autistic youth was conducted in accordance with the standards described by the Preferred Reporting Items for Systematic Reviews and Meta-Analysis (PRISMA; Moher et al., 2009) guidelines. The review protocol was prospectively registered with the PROSPERO international review database (ID: CRD42021240272; https://​www.​crd.​york.​ac.​uk/​prospero/​). A concurrent search of Ovid MEDLINE, PubMed, and PsycINFO was conducted on June 7, 2021 to identify all articles published to date. The final search strategy is provided in Supplemental Table 1. Based on the previous review of therapeutic process factors completed by Karver et al. (2006), 29 process-related search terms were selected. The search strategy also included autism-related terms (i.e., autis*, Asperger syndrome) and terms describing youth age range (i.e., child*, pediatric, youth, adolescen*, and kid), which were combined with the process-related terms. When possible, terms were modified to align with database-specific index terms (e.g., MeSH terms for PubMed).

Studies identified through the searches were included based on the following criteria: (a) evaluated or described a psychosocial intervention addressing an emotional or psychological mental health-related outcome (as either a primary or secondary outcome; e.g., anxiety, depression); (b) at least a portion of the sample included youth under the age of 18 years, with a previous diagnosis of autism or related disorder (e.g., Asperger syndrome); (c) assessed and/or described at least one process-related factor; and (d) was available in English. Exclusion criteria were: (a) interventions addressing disruptive behaviour only (e.g., opposition, defiance, conduct problems) without consideration of the emotional and psychological aspects of mental health; (b) non-empirical publications (e.g., editorials, commentaries, books, book chapters, theoretical papers); (c) reviews or meta-analyses; and (d) unpublished dissertations or theses. Notably, there is a large body of literature (e.g., studies evaluating applied behavioural interventions) that has focused on reducing “disruptive” behaviours (e.g., self-injurious behaviour; aggression towards others) without evaluating the role of internalized emotional and psychological experiences (e.g., anxiety, depression, emotion regulation) that may explain the observable behaviour. Although it is not uncommon for autistic youth to demonstrate disruptive behaviour that stems from emotional and psychological issues, it is uncertain whether these problems are a targeted treatment outcome of an intervention when there is a lack of direct measurement. For clarity and consistency, the decision was made to focus on studies that explicitly assessed emotional and psychological outcomes related to mental health.

To identify eligible studies, all titles and abstracts of articles retrieved through database searches were independently screened by two authors (CA and NV). For the screening phase, disagreements were resolved through discussion until consensus was reached between the two authors. All articles that underwent full-text review were also independently reviewed by each author. Disagreements for full-text review were resolved through discussion and consultation with the senior author (JW).

The Mixed Methods Analysis Tool (MMAT; Hong et al., 2018) was used to assess the methodological quality of studies that met inclusion criteria, which was developed for empirical studies based on the study design. To determine whether a study is considered empirical, reviewers respond “Yes” or “No” to two screening questions regarding clarity of research questions and adequacy of data collection for addressing research questions. For studies that fail to meet screening criteria, quality appraisal is not conducted. For articles that meet screening criteria, reviewers respond “Yes” or “No” to five methodological-related questions that vary depending on study design (e.g., qualitative vs. randomized controlled trial). If there is insufficient information provided in the article to answer the question, reviewers may respond with “Can’t Tell”. As per the MMAT guidelines (Hong et al., 2018), methodological quality of each study is considered based on responses to individual criterion, versus a global rating or overall score. Included studies were assessed by one of two reviewers (CA or NV), and approximately 30% of studies (n = 8) were coded by the other reviewer as a reliability check. Reviewers were recused from rating articles that they authored. There was substantial agreement across raters (Fleiss κ = 0.77). Disagreements in ratings were resolved through discussion. The MMAT has been used as a quality appraisal tool in other review studies focused on developmental disabilities literature, including autism (Albaum et al., 2021; Doherty et al., 2020).

Search results are illustrated in Fig. 1. An initial search of the databases using the search strategy described above yielded 6,409 articles. After duplicate articles were removed, 4,994 articles remained for review. The two authors screening titles and abstracts agreed on whether studies should be excluded or undergo full-text review for 97% of articles (n = 4849 articles). Title and abstract screening resulted in 4,898 articles being excluded. The full-text for the remaining 96 articles were again reviewed by both authors. Authors agreed regarding eligibility for 82% of articles (n = 79). Full-text review resulted in an additional 74 articles being excluded (see Fig. 1 for exclusion reasons), leaving 22 articles for data extraction and synthesis. Reference lists from included articles were reviewed for potentially relevant studies that were not identified in the original search, which resulted in inclusion of three additional studies (Brown et al., 2015; Jassi et al., 2021; White et al., 2013).

Study characteristics are detailed in Table 1, including study location, study design, sample characteristics, description of intervention, mental health outcomes, and process factors identified. Twenty-five studies derived from 21 unique samples were identified; four of the included articles utilized the same sample (Albaum et al., 2020; Burnham Riosa et al., 2019; Thomson et al., 2015; Weiss et al., 2018). Although there was no lower date bound set for inclusion criteria, all studies identified through the search were published from 2012 onwards, with 68% of studies (n = 17) being published within the past six years (i.e., after 2015). Of the 25 included studies, 64% (n = 16) were conducted in North America and 28% (n = 7) in Europe; one study occurred in Asia, and one study occurred in Australia. In terms of study design, most studies were quantitative-descriptive in nature (36%, n = 9). Five studies (20%) conducted quantitative, non-randomized intervention trials, and five studies were randomized controlled trials. Three studies (12%) employed qualitative designs, and the remaining three studies used a mixed methods approach, involving quantitative and qualitative methods.

Across studies, participants ranged in age from 4 to 29 years. Nine studies (36%) involved adolescents (i.e., 13 years of age or older) and nine included both children and adolescents; five studies (20%) included children only (i.e., 12 years of age or younger), and two did not report on the age range of the sample. The mean age reported across studies was 12.6 years (SD = 3.41; range: 6.2 to 20.6). On average, samples comprised 82.3% males (SD = 13.2%, range: 43% to 100%). Sixteen studies (64%) reported that the majority of participants in the sample identified as White/Caucasian, 11 of which included samples with more than 75% White/Caucasian participants; one study sample comprised 100% Asian participants (i.e., Chinese, Indian, or Malays; Drmic et al., 2017), and one study comprised 100% Latinx participants (Chlebowski et al., 2018). Seven studies (28%) did not report on the ethnicity or racial identity of participants. Most studies (80%, n = 20) reported using autism diagnostic (e.g., ADOS; ADI-R) and/or screening tools (e.g., SCQ; SRS-2), or having participants share a report from a licensed healthcare provider to confirm autism diagnostic criteria were met. The majority of studies excluded participants with a diagnosed intellectual disability or who had limited intellectual abilities (e.g., IQ < 70; 76%, n = 19); five studies (20%) did not specify level of intellectual functioning. One study included a single participant with an intellectual disability, as reported by parents (London et al., 2020).

Regarding the interventions, the majority of studies (64%, n = 16) involved treatment programs that were based on CBT. Other forms of psychosocial treatment included psychoeducation (Backman et al., 2018; Gordon et al., 2015), social skills interventions (Kang et al., 2021; Lordo et al., 2017; White et al., 2013), mindfulness-based interventions (Brewe et al., 2021; Pahnke et al., 2014), counselling or psychotherapy (Brown et al., 2015; Chlebowski et al., 2018), music programs (Hillier et al., 2012), and animal-assisted occupational therapy (London et al., 2020). Fifteen studies (60%) involved individualized treatment and 36% (n = 9) used group formats; one study involved a combination of individual and group sessions (White et al., 2013). Parents were fully involved in treatment (i.e., were present for the entire duration of all sessions) in 40% of studies (n = 10), and partially involved for another 40% of studies (e.g., present for only a portion of each session; attended some, but not all sessions; participated in parallel parent-only sessions). For 20% of studies (n = 5), the extent of parent involvement was not reported. Mental health outcomes that were commonly targeted included anxiety (60% of studies, n = 15), emotion regulation (28%, n = 7), internalizing and externalizing symptoms broadly (16%, n = 4), depression (8%, n = 2), and obsessive–compulsive symptoms (8%, n = 2). Studies also examined other mental health-related outcomes such as general psychopathology and psychiatric comorbidity, psychological distress and stress-related behaviour, emotional and behaviour problems, positive and negative affect, and self-esteem.

Study quality was appraised using the MMAT (Hong et al., 2018). All studies were deemed to have clear research questions and collected appropriate data to address the research questions, and thus met MMAT screening criteria to proceed for further appraisal. Eight studies (32%) met 100% of the MMAT criteria for the respective study design; 10 studies (40%) met 80% of criteria; four studies (16%) met 60% of criteria; one study (4%) met 40% of criteria; and two studies (8%) met 20% of criteria. Summary of responses to questions regarding methodological quality are provided in Table 2.

Among studies that employed quantitative methods (n = 22; including randomized and non-randomized, descriptive, and mixed methods designs), the following process factors were identified: family accommodation, parent-therapist alliance, treatment adherence, treatment expectations, treatment readiness, treatment satisfaction, youth involvement, and youth-therapist alliance. Details regarding the measures used to assess each process factor are provided in Table 3. As shown in Table 1, approximately 62% of studies described process factors as indicators of treatment feasibility, without examining process-outcome associations. For example, several studies described homework completion as an indication of treatment adherence (e.g., McNally Keehn et al., 2013; Thomson et al., 2015), but did not report correlations between homework completion and treatment outcome. Eight studies explicitly evaluated the relation between process factors and treatment outcome, and six studies assessed relations among process factors. An overview of study results for each process factor follows below.

Four of the included studies conducted qualitative analyses to evaluate participants’ experiences with mental health interventions (Chlebowski et al., 2018; Drmic et al., 2017; Edgington et al., 2016; London et al., 2020), which resulted in several emerging themes of process-related factors. Three studies identified themes related to parent involvement in mental health treatment for autistic youth. For example, strong, consistent support from stakeholders (including parents), such as parent enthusiasm for their child’s treatment, was described as an important factor for facilitating and implementing CBT for anxiety in autistic adolescents (Drmic et al., 2020). Themes related to challenges with parent involvement also emerged, such as difficulties with obtaining information from adolescents when parents are not directly involved in treatment (Edgington et al., 2016), and different perceptions between therapists and parents in terms of the expected role parents play in mental health treatment for children (e.g., expecting child to take part in sessions independently, without parent involvement; Chlebowski et al., 2018). Youth motivation was a relevant theme in considering the feasibility of implementing an intervention program, and participants’ acceptability and skill application. Identifying motivated adolescents to participate in treatment was important for establishing positive rapport and facilitating enjoyable sessions (Drmic et al., 2020), and self-motivation was relevant for unprompted use of coping strategies (Edgington et al., 2016). Relatedly, a lack of readiness from adolescents to participate in groups was described as a barrier for coaches to facilitate sessions smoothly (Drmic et al., 2020). Other process factors that were mentioned in qualitative themes included group cohesion (e.g., importance of group dynamics; Edgington et al., 2016), therapist direct influence skills (e.g., ability to carry out program delivery; Drmic et al., 2020), the relationship between parents and therapists (Chlebowski et al., 2018), and youth engagement in relation to treatment progress (London et al., 2020).

Relational factors comprise constructs that describe different aspects of the relationships between therapists, clients, and caregivers, both within and outside therapy sessions. Consistent with the non-autism literature (Fjermestad et al., 2009; Karver et al., 2006), relational factors appear to be the most well-understood aspect of the therapeutic process in mental health treatment for autistic youth. Several studies examined therapeutic alliance between therapists and autistic youth. Therapeutic alliance refers to the working relationship between therapist and client, based on bond, collaboration on therapeutic tasks, and agreed upon treatment goals (Bordin, 1979). Preliminary psychometric evidence suggests that youth-therapist alliance can be validly assessed through different informants (e.g., youth, therapist, independent-observer) at multiple points during treatment, making it possible to conduct longitudinal analyses using multiple perspectives. Within the general child literature, variation in alliance-outcome associations has been attributed in part to reporting source (e.g., youth vs. therapist; Karver et al., 2018). Researchers should, therefore, consider incorporating measures that rely on multiple perspectives when evaluating alliance with autistic clients to help determine the predictive validity of ratings provided by different informants. Findings across studies indicate that it is not only possible for clinicians to form a strong working relationship, but that therapeutic alliance may be an important contributor to improvements in anxiety (Kerns et al., 2018; Klebanoff et al., 2019) and emotion regulation (Albaum et al., 2020; Brewe et al., 2021) for autistic children and adolescents. Therapeutic alliance with autistic youth was also found to be positively associated with youth involvement during therapy sessions, and treatment adherence outside of sessions (Burnham Riosa et al., 2019), though further research is needed to understand the transactional pattern that occurs between these factors over the course of therapy. Given the inherent social-communication challenges associated with autism that can make it difficult for youth to form meaningful relationships, it is critical to consider therapist factors that may contribute to establishing and maintaining alliance with these clients. For example, clinicians’ lack of knowledge and experience, poor competence, and low confidence about working with autistic adults have been identified as barriers to treatment for these clients (Maddox et al., 2020). A recent study also found that therapists are less likely to treat autistic youth compared to youth with ADHD, which was partly explained by differences in therapist attitudes about working with each population, their knowledge about mental health for autistic clients, and the normative pressures they felt about treating autistic youth (Roudbarani et al., in press). In light of the association between therapist attitudes and intention to treat, and the proposed theoretical connection between therapist attitudes and process factors (Karver et al., 2005), it is important to evaluate how these attitudes may influence relational aspects of the therapeutic process that are relevant for treatment outcome. For instance, researchers may aim to answer questions such as: Do more favourable therapist attitudes predict stronger therapeutic alliance? Does the formation of therapeutic alliance strengthen therapist attitudes about working with autistic clients? Are more favourable therapist attitudes related to greater improvement in treatment outcome, and is this association mediated by the quality of therapeutic alliance? Addressing these research questions may involve assessing therapist attitudes prior to beginning therapy, and then repeatedly measuring attitudes and therapeutic alliance over the course of treatment to determine whether there is a shift in therapist attitudes, and if there is bidirectional link between attitudes and alliance. In addition, therapists may find it beneficial to have youth and parents complete brief measures of therapeutic alliance (e.g., TASC-R) to monitor their perspectives on the relationship, and then actively work to strengthen the relationship with autistic clients by addressing areas of concern that have been indicated.

Review findings suggest that family accommodation may also be a relevant relational factor to consider when addressing mental health concerns with this population. Family accommodation refers to parent and other family members’ behaviour that abets the child in avoiding anxiety-inducing experiences (Lebowitz et al., 2012, 2014). Previous reviews of the general child literature (e.g., Karver et al., 2006) have not considered family accommodation as part of the therapy process, though it may be a pertinent factor to consider when parents and family members are involved in a child’s treatment. For youth without autism, family accommodation is positively related to child anxiety levels, and reductions in family accommodation have been linked to greater symptom improvement (Lebowitz et al., 2014; Merlo et al., 2009). Three studies identified through the review indicated progressive reduction in parent-reported family accommodation over the course of therapy (Jassi et al., 2021; Jones & Jassi, 2020; Storch et al., 2015), with one study finding a negative association between family accommodation and treatment response (i.e., treatment responders reported lower levels of accommodation compared to non-responders; Storch et al., 2015). Shifts in family accommodation may be particularly relevant to the therapeutic process when parents are involved in their child’s treatment, as is often the case for autistic youth (Reaven, 2011). However, based on literature currently available, it is still unclear whether changes in family accommodation are related to changes in mental health outcomes, and if family accommodation is associated with other aspects of the therapeutic process, such as parent-therapist alliance or parent beliefs about treatment.

Client expectations regarding the efficacy, relevance, and importance of treatment, as well as their readiness or motivation to participate, are thought to be key factors related to treatment attendance and adherence (Karver et al., 2005; King et al., 2014). Although treatment expectations have been considered pre-treatment characteristics (Karver et al., 2005), evidence from the current review suggests that beliefs about treatment may not necessarily be fixed, as Backman et al. (2018) found that participants’ perceptions of treatment credibility shifted over the course of therapy. Within the context of interventions for children and adolescents without autism, both youth and parent expectations and readiness may have a dynamic association with treatment participation, in turn influencing therapeutic outcome (Karver et al., 2006). Minimal research has explored these processes for autistic youth, either in terms of construct validity or treatment outcome. None of the studies identified in the review assessed parent expectations and willingness to participate in their child’s treatment. Qualitative analyses indicate that youth motivation may be relevant to therapist capacity to establish rapport and facilitate sessions, and treatment adherence on the part of the youth outside of therapy sessions. Parent expectations regarding their role in their child’s treatment was also described as pertinent to the provision of mental health treatment for autistic adolescents (Chlebowski et al., 2018). As noted above, Backman et al. (2018) quantitatively assessed youth perceptions of treatment credibility during the receipt of psychoeducation about autism, and found that perceived credibility improved over the course of the intervention. However, the authors did not examine whether credibility was associated with changes in youth anxiety or depression. Similarly, a second study assessed children’s readiness and willingness to take part in treatment prior to participating in CBT and found a non-significant relation between treatment readiness and child-therapist alliance (Albaum et al., 2020). The authors reported a considerable range in the degree of child readiness across participants, suggesting that autistic youth who take part in mental health intervention likely vary in their willingness and commitment to participate; however, they did not assess the association between treatment readiness and changes in mental health outcomes following treatment completion. Thus, there is a lack of empirical evidence available on the expectations autistic youth and their parents have regarding therapy, and how motivation or readiness for treatment is related to the therapy process and positive outcomes. It may be worthwhile for researchers to evaluate client beliefs about therapy prior to starting treatment, and track expectations and motivation during treatment. Findings from research of this nature could help to determine whether there are common patterns in the ways autistic clients’ beliefs about therapy change over the course of treatment, and if certain trajectories are associated with better outcomes. Establishing an evidence base on the connection between clients’ beliefs and treatment outcome could help inform adaptations to therapeutic techniques, such as motivational interviewing (Feinberg et al., 2021; Rogers et al., 2019), which can be used by clinicians to promote expectations, motivation, and treatment readiness for autistic youth and their parents.

Related to client expectations is the extent to which these expectations are met, and whether they are satisfied in their therapeutic experience, based on perceived helpfulness, relevance, and enjoyment. It has been hypothesized that treatment satisfaction may contribute to active involvement and better compliance, though the mechanistic pathway between these process-related variables is poorly understood (Karver et al., 2005). Within the autism literature, it appears that treatment satisfaction is often reported as an indicator of overall feasibility or acceptability of the intervention. Satisfaction was often analyzed descriptively, with various studies reporting high ratings of satisfaction from both parents and youth. Only two studies reported on relations between treatment satisfaction and symptom improvement (Swain et al., 2019; Walsh et al., 2018), with mixed results. Researchers who are evaluating interventions and choose to include measures of treatment satisfaction should consider exploring shifts in satisfaction that may occur throughout treatment, and possible interactions with other key therapeutic processes (e.g., compliance, adherence) that may contribute to variation in treatment effectiveness.

Engagement in treatment is considered critical for successful outcomes (Becker et al., 2018). Within the general youth literature, the term “engagement” tends to be used interchangeably with “involvement” or “participation”, and refers to the active, effortful, and collaborative role that youth and parents have during and between therapy sessions (Karver et al., 2005). Research has yet to explore treatment engagement in relation to outcome within the context of mental health intervention for autistic youth. Several studies identified through the review provided quantitative descriptions of treatment adherence, generally operationalized as homework completion outside of sessions (e.g., Gordon et al., 2015; Pahnke et al., 2014); however, homework completion or between-session practice was typically dichotomized as “complete” or “incomplete”, failing to consider other factors, such as ease or difficulty of assigned practice, that may be relevant to the overarching construct of engagement. Youth involvement within therapy sessions was also described in several studies (e.g., Weiss et al., 2018; White et al., 2013), but relied solely on global therapist ratings on a single-item. Developing valid and reliable measures that consider nuanced behavioural indications of in-session participation can lead to greater understanding of how engagement contributes to treatment success, and can equip clinicians with the knowledge and skills needed to promote participation during sessions with autistic clients.

Parent involvement was an overarching theme in multiple studies that employed qualitative methods (Chlebowski et al., 2018; Drmic et al., 2017; Edgington et al., 2016). Direct parent involvement may be particularly important for parents to learn how to best assist their child in practicing the skills being taught (Drmic et al., 2017; Edgington et al., 2016). For younger children, parents may be expected to take on various roles in their children’s therapy, such as co-therapists. Incongruence between therapist and parent expectations about the extent of parent involvement can potentially hinder the quality of parent participation both within and outside of sessions (Chlebowski et al., 2018). Prior to beginning therapy, establishing clear, agreed upon expectations between therapists and parents about participation may augment engagement in their child’s treatment. Some studies have empirically examined therapeutic alliance between therapists and caregivers of autistic youth who participate in their children’s therapy. There is emerging support for the valid measurement of parent-therapist alliance; however, findings are inconsistent in terms of process-outcome associations. Klebanoff et al. (2019) found parent-therapist alliance was associated with greater reduction in anxiety post-treatment, whereas Kerns et al. (2018) did not find a significant process-outcome association. Given that parent involvement is a recommended modification to psychosocial intervention for autistic youth (Reaven, 2011; Walters et al., 2016), it is important to establish stronger empirical knowledge about how parents contribute to their child’s treatment success.

Review findings should be interpreted within the context of the study’s limitations. The current review only offers a narrative synthesis of process factors; meta-analysis of effect sizes was not calculated, thus preventing comparison of potential process-outcome associations. This review only included studies published in English, and additional empirical evidence regarding process factors may be available in other languages. In regard to client characteristics, several studies included samples for which a portion of participants were over the age of 18 years (Backman et al., 2018; Brewe et al., 2021; Hillier et al., 2012; London et al., 2020; Pahnke et al., 2014), but did not conduct analyses to evaluate potential effects of age (e.g., separating < 18 vs. 18 +; including age as covariate). Samples were also limited in terms of ethnic diversity, with most studies reporting that the majority of the sample comprised participants who identified as White/Caucasian. Finally, results are based on samples that almost exclusively involve autistic youth without co-occurring intellectual disabilities and may not be equally relevant for mental health treatment for youth with limited cognitive abilities and adaptive skills.