Left untreated, child externalizing behaviors (EXT) (e.g., noncompliance, aggression) and internalizing behaviors (INT) (e.g., anxiety, depression) can lead to significant adverse outcomes across development, including later mental health difficulties, reduced academic achievement, and unemployment (Clark et al.,
2017; Goodman et al.,
2011; Kim-Cohen et al.,
2003). The quality of parenting and parent–child interactions are key risk factors linked to the development and maintenance of these problem behaviors and are modifiable through intervention. For example, reduced parental warmth and sensitivity are associated with both increased EXT (Hoeve et al.,
2009) and INT (McLeod et al.,
2007a,
2007b). In this light, much research has focused on examining the effects of parenting interventions on child and adolescent EXT and INT.
In the literature, parenting interventions are typically categorized as “behavioral” or “nonbehavioral.” Accordingly, we first review evidence on behavioral parent training (BPT), and then review the growing literature on nonbehavioral parenting interventions, with an emphasis on the subcategory of attachment- and emotion-focused parenting interventions (AE). The latter provides the focus of this review, which aims to investigate the effectiveness of individual and group AE on child and adolescent EXT and INT compared to waitlist controls, BPT, and other active comparators.
Behavioral Parenting Interventions
BPT has been the predominant model of parenting interventions since the 1960s. BPT derived from social learning theory, which stipulates that child behaviors are strengthened and weakened through parent reinforcers (Scott & Dadds,
2009). EXT are thought to be maintained through coercive parent–child reinforcement traps, whereby a child and parent engages in aversive behaviors (e.g., yelling) until either person capitulates, ultimately reinforcing the other’s aversive behavior (Patterson,
1982). BPT aims to break coercive cycles by shifting parents’ attention from children’s problematic behaviors to their desirable behaviors through strategies such as praise, clear instructions, logical consequences, planned ignoring, and timeout from positive reinforcement (McMahon & Pasalich,
2020). Examples of well-researched BPT programs include the following: Helping the Noncompliant Child (HNC, Forehand & McMahon,
1981), The Incredible Years (IY, Webster-Stratton,
2005), and Triple P-Positive Parenting Program (Sanders,
1999). BPT is considered the gold-standard for treating EXT, with meta-analytic findings demonstrating its effectiveness across childhood (Kaminski & Claussen,
2017; Mingebach et al.,
2018). While BPT was not originally designed to directly target INT, findings suggest that it may still be effective at reducing INT for children and adolescents (Forehand et al.,
2013; Zarakoviti et al.,
2021).
Although BPT effectively reduces EXT and to a lesser extent, decreases INT, approximately 25–33% of children do not appear to benefit from these programs (Scott & Dadds,
2009), with high dropout and low attendance rates reported. For example, a review of BPT found that 25% of eligible participants did not begin treatment, and 26% dropped out before completing treatment (Chacko et al.,
2016). Research is limited on variables that may influence attrition in BPT programs, or produce barriers to positive treatment outcomes. Notwithstanding this, past findings suggest that baseline variables such as socioeconomic status, parent mental health, family history of trauma or attachment difficulties, and parent attributions may moderate the effect of BPT outcomes for families (Assenany & McIntosh,
2002; Havighurst et al.,
2020; Maliken & Katz,
2013; Scott & Dadds,
2009). Notably, parent mental health problems have been shown to be a moderator of BPT outcomes (Maliken & Katz,
2013). Accordingly, researchers have highlighted the need to focus on parents’ understanding and management of their own emotions in addition to their child’s (Havighurst et al.,
2020), which is not a key focus in some BPT programs.
Another potential barrier to participation in or positive treatment outcomes for BPT includes preference for an alternative theoretical model of intervention. Research in the general field of evidence-based practice with health professionals, including clinical psychologists, demonstrates that practitioners’ preferred theoretical orientation is a significant barrier to implementation of evidence-based treatments (Lilienfeld et al.,
2013; Pagato et al.,
2007). With regard to BPT, the use of timeout—i.e., a core component of BPT, whereby a child is placed in a safe, neutral space for a brief period of time away from parent attention (Kaminski et al.,
2008)—has been the source of ongoing debate among practitioners and researchers, with many parents and professionals re-evaluating its strengths and challenges [see Dadds and Tully (
2019) for a recent review]. Moreover, a recent community-based study exploring why some parents prefer not to use timeout found that they favor alternative parenting strategies believed to value connection, “attachment,” and co-regulation in the parent–child relationship (Canning et al.,
2021). In summary, although BPT has a substantial evidence base and is widely disseminated, there is a timely need to investigate the collective evidence for alternative models of parenting interventions targeting EXT and INT that may be better suited to parents who prefer and value relational and emotion-oriented approaches to parenting.
Attachment- and Emotion-Focused Parenting Interventions
The most researched “nonbehavioral” parenting interventions involve programs based on (1) attachment theory and (2) emotion socialization theory, which are characterized by a focus on affective processes in the parent–child relationship and understanding the meaning of, and/or emotional needs underlying, child behavior. According to attachment theory, a secure attachment develops when a caregiver sensitively responds to their child’s needs and the child can alleviate their distress by seeking proximity to the caregiver (Ainsworth et al.,
1978; Bowlby,
1982). By contrast, when a caregiver inconsistently responds to their child’s needs, the child may develop an ambivalent attachment, which is characterized by an approach and resistance pattern of interaction with the caregiver when distressed. Children whose caregivers are consistently unresponsive to or rejecting of their needs are at risk of developing an avoidant attachment wherein they do not seek proximity to their caregiver when distressed (Ainsworth et al.,
1978). Of most concern for the development of psychopathology, a disorganized attachment is evident in children who do not develop a coherent strategy to assuage their distress and is linked to fearful and frightening caregiving (Lyons-Ruth & Spielman,
2004).
Attachment-based interventions such as Attachment- and Bio-behavioral Catch-up (Dozier et al.,
2011) and Circle of Security (Hoffman et al.,
2006) focus on improving attachment security through targeting caregiving behaviors. These include caregiver sensitivity (i.e., caregivers’ ability to notice and determine their child’s needs) and reflective functioning (i.e., caregivers’ capacity to understand their own and their child’s mental states including feelings and beliefs) (Kobak et al.,
2015). Meta-analytic evidence supports associations between insecure attachment styles and EXT and INT. In particular, disorganized attachment most strongly predicts EXT and avoidant attachment is most strongly associated with INT (Groh et al.,
2017). In addition to attachment styles, caregiving behaviors including caregiver sensitivity and reflective functioning are associated with both EXT and INT (Carlone & Milan,
2020; Dejko-Wanczyk et al.,
2020; Kok et al.,
2013; Wang et al.,
2013). Given the established links between quality of caregiving and attachment, attachment-based interventions have predominantly focused on evaluating changes in caregiver sensitivity and attachment security (e.g., Bakermans-Kranenburg et al.,
2003). However, surprisingly, there has been less research attention on other child outcomes such as EXT and INT, which are linked to attachment and associated caregiving behaviors.
Parenting interventions based on emotion socialization theory—e.g., Tuning in to Kids (Havighurst et al.,
2010)—also focus on the parent–child relationship and aim to support parents in understanding and responding to emotional needs related to child behavior. The rationale for emotion socialization interventions is based in part on associations between greater emotional competence and less EXT and INT (Saarni,
1999). Emotional competence can be strengthened through parents’ “emotion coaching.” Emotion coaching principles include having awareness of children’s emotions and recognizing emotional moments as opportunities for teaching and intimacy, listening empathetically, validating feelings, facilitating children to label their emotions, and helping children problem-solve (Gottman & DeClaire,
1997). Parent emotion socialization behaviors including emotion coaching are significantly associated with both EXT (Johnson et al.,
2017) and INT (Shortt et al.,
2016; Suveg et al.,
2005), and meta-analytic findings show that child emotion regulation is a protective factor for EXT and INT (Daniel et al.,
2020).
Researchers have recently considered the implications of attachment theory for emotion socialization theory and vice versa, as in part, the attachment relationship develops through caregivers’ responses to child emotional cues. For instance, many of the skills within attachment-based interventions (e.g., caregiver sensitivity and reflective functioning) require parents to have capacity for emotion regulation (Hajal & Paley,
2020). Given similarities of theoretical origin and intervention targets between interventions guided by emotion socialization and attachment theory, for the purposes of this review, we combine these two interventions as AE. We define AE as those that go beneath behavior and aim to strengthen the emotional quality of the parent–child relationship by helping caregivers understand children’s attachment and emotional needs expressed through their behavior.
Emerging evidence of randomized controlled trials (RCTs) investigating the effects of individual and group AE on EXT and INT has revealed significant improvements for children and adolescents (e.g.., Baker et al.,
2015; Havighurst et al.,
2019; Moretti et al.,
2018) and potentially comparable effects to BPT (Duncombe et al.,
2016; Högström et al.,
2017). Furthermore, AE may produce sleeper effects, whereby positive intervention effects may be enlarged at follow-up compared to post-treatment, hence requiring longer follow-up assessments to identify changes (Bakermans-Kranenburg et al.,
2003). However, other studies have failed to find statistically significant changes in EXT or INT for AE (e.g., Adkins et al.,
2021, Rolock et al.,
2021; van Doesum et al.,
2008). Thus, without a systematic review and meta-analysis of pooled RCTs, we cannot be certain of the effectiveness of AE on EXT and INT for children and adolescents.
Previous Reviews of Parenting Intervention Effects on Externalizing and Internalizing Outcomes
Several prior systematic reviews and meta-analyses have investigated the effectiveness of parenting interventions broadly on both EXT and INT; however, AE appear to be frequently missing. For example, a meta-analysis of past meta-analyses examining the effectiveness of any parenting intervention for the treatment of EXT in clinic-referred children aged under 13 (Mingebach et al.,
2018) found a significant moderate effect for reduced EXT (effect size = 0.46, 95% CI 0.35–0.55); though, only 2 of 26 meta-analyses included nonbehavioral interventions (Leijten et al.,
2018; Lundahl et al.,
2006). The first meta-analysis compared behavioral to nonbehavioral interventions on EXT but the nonbehavioral studies did not reference attachment or emotion socialization theory (Lundahl et al.,
2006). The second meta-analysis examined the integration of “relationship enhancement” with BPT in a single intervention for reducing EXT; however, it did not consider the isolated effects of AE without behavioral components (Leijten et al.,
2018). Although these past studies increase our understanding of BPT, the specific effects of AE on child EXT and INT remain unclear.
In addition to the predominant focus on BPT in past research, it is possible that AE were often excluded from prior reviews of parenting interventions due to participant eligibility criteria. BPT programs were specifically developed to treat EXT and secondary problem behaviors, whereas many AE were originally developed to primarily strengthen caregiving and the parent–child relationship in targeted populations, such as children exposed to maltreatment and/or in out-of-home care (Dozier et al.,
2011; Slade et al.,
2018). To illustrate, a systematic review of 64 RCTs of psychosocial treatments in children 12 and younger concluded that group and individual BPT demonstrated the most support for reducing EXT; yet only one emotion-focused intervention and no attachment-based interventions were included (Kaminski & Claussen,
2017). This was most likely due to the study’s inclusion criteria requiring participants to exhibit clinically elevated EXT behavior at baseline, which is not typical for AE research. Furthermore, children in out-of-home care—who are commonly targeted in AE—were also excluded from this review.
Finally, although a recent narrative review of AE research suggests promising intervention effects for child and adolescent EXT and INT (Havighurst et al.,
2020), this study did not involve a meta-analysis and is limited to studies published January 2019 through June 2020. In summary, previous research examining the effects of parenting interventions on EXT and INT have not comprehensively and systematically examined the effectiveness of AE on child and adolescent EXT and INT. This is a significant gap in the literature with considerable importance for research and practice concerning parenting interventions and child wellbeing.
Thus far, we have primarily highlighted the significant gap in the literature on AE effects on child EXT and INT, though there is also limited research around its effects on parent mental health and wellbeing outcomes. Previous reviews of AE have investigated intervention-induced improvements in parent skills (e.g., caregiver sensitivity; Bakermans-Kranenburg et al.,
2003) but not parent mental health, which may also be impacted by some AE (e.g., Moretti et al.,
2018; Weihrauch et al.,
2014), particularly given that AE predominately target parents’ behavior, thoughts, and feelings (Havighurst et al.,
2020; Kobak et al.,
2015; Maliken & Katz,
2013). A meta-analysis on one type of AE—Circle of Security—found a medium effect on parent depressive symptoms following intervention; however, only three studies were included (Yaholkoski et al.,
2016). In light of these limitations, further research on parent mental health and wellbeing outcomes following AE is needed.
Discussion
To date, BPT has been by far the most established model of parenting intervention for reducing child mental health problems, particularly EXT, and the predominant focus of treatment outcome research in the area. Notwithstanding this, the popularity of AE appears to have grown among clinicians, researchers, and parents. This systematic review and meta-analysis aimed to conduct the first examination of the effectiveness of AE on EXT and INT compared to (1) waitlist comparators, (2) BPT, and (3) any active comparators. We also conducted additional planned analyses on risk of bias and length of follow-up, as well as analyzed parent mental health as a secondary outcome.
Meta-analytic findings supported our first aim and suggest that AE are more effective at reducing EXT and INT compared to waitlist conditions. Although we were unable to test our second aim using meta-analysis, findings from our narrative review of two RCTs suggest that AE may produce comparable effects to BPT for decreasing EXT. No retrieved studies directly compared AE to BPT on INT; thus, it is unknown how the two interventions compare in this regard. AE did not show to be more effective than active comparators for EXT or INT, not supporting our third aim. However, additional analyses found that at 6-month follow-up or greater, effects for INT were sustained and non-significantly increased for EXT, such that at 6-month follow-up, AE showed a significant small effect relative to active comparators for EXT. Sensitivity analyses also found that when high risk of bias studies were removed, effect sizes for EXT increased relative to waitlist comparators and remained consistent for INT across comparators. Interestingly, no statistically significant effect was observed between AE and any comparator for parent mental health outcomes.
The current findings both concur with and diverge from results from previous meta-analyses investigating intervention effects on EXT and INT. A meta-meta-analysis of various parenting interventions for clinic-referred children 13 and under found an effect size of 0.46 for EXT (Mingebach et al.,
2018), which is higher than the effect size observed in our study. This difference in effect size could be a true estimate, though it could also relate to several other factors. First, our review paper spans across birth to 18, whereas this prior meta-analysis only looked at preadolescence, missing an additional 6 years where problem behaviors can increase (Moretti et al.,
2018). Second, considerable heterogeneity (
I2 = 84.56%) was observed in the Mingebach et al. (
2018) study, whereas we did not conduct a meta-analysis if heterogeneity was considered high (
I2 > 59%) (Higgens et al.,
2021). Third, we included both community and clinical samples, rather than only clinical samples, and past research demonstrates larger effect sizes in clinical samples as higher baseline levels are subject to greater room for change post-treatment (Leijten et al.,
2013; McMahon et al.,
2021). Fourth, although AE derive from similar theoretical orientation, they are more heterogeneous than models of BPT, which may have led to more varied effects on EXT and INT, reducing the overall effect size. As BPT research on mental health outcomes is far more extensive than AE research, we did not have the same degree of flexibility to consider various subgroup and sensitivity analyses to tease apart differences. Therefore, caution needs to be applied when comparing effect sizes to previous review papers; the only way to truly compare these interventions is for participants to be randomly allocated to BPT or AE within the same study.
Conversely, previous meta-analytic research on parenting interventions for INT, which analyzed studies of children from birth to 18 and excluded effect sizes with high levels of heterogeneity, found an overall effect size of 0.12 for INT (Yap et al.,
2016). This effect size is smaller than the one observed in our study and appears comparable in terms of population and statistical method. Hence, this suggests that AE could be a more promising choice of parenting intervention for reducing INT in children and adolescents. Nevertheless, without directly comparing AE to BPT—or other parenting interventions—through a meta-analysis of high-quality RCTs, we cannot assume AE is less effective for EXT or more effective for INT.
Considering our findings within the context of past reviews on BPT, both AE and BPT may effectively reduce EXT and INT; however, there may be unique mechanisms underpinning these different intervention approaches. It is suggested that proximal targets of AE, such as attachment security, reflective functioning, caregiver sensitivity, and parent emotional awareness, may be mechanisms accounting for improvements in EXT and INT (Carlone & Milan,
2020; Havighurst et al.,
2020; Kobak et al.,
2015; Kok et al.,
2013; Wang et al.,
2013). For example, individual studies included in the review found that parent sensitivity post-intervention mediated the effect of AE on child EXT (Lind et al.,
2020); and improvements in parent emotion socialization mediated youth INT following an AE (Kehoe et al.,
2020). Further research should shed light on these and other mechanisms accounting for the effects of AE on EXT and INT, as has been done for BPT with regard to behavioral parenting practices (Forehand et al.,
2014). If AE and BPT continue to produce comparable results in future RCTs akin to the two observed in this paper, researchers and clinicians may also consider the extent to which AE and BPT are compatible and complementary. For example, it is theorized that individual differences in attachment patterns may be in part learnt through behavioral principles of classical and operant conditioning (Bosmans et al.,
2020), and BPT has been shown to improve attachment-based parenting domains such as caregiver sensitivity (O’Connor et al.,
2013). Although there have been recent attempts by researchers to integrate these theoretical models in parenting interventions, they have not always found superior effects on EXT (Leijten et al.,
2018; O’Hara et al.,
2019) and are yet to be examined for INT. Thus, the current review highlights several research directions to examine both possible underlying mechanisms of AE as well as their compatibility with other parenting interventions.
Our review explored several possible factors that may influence or moderate the effectiveness of AE on EXT and INT. Firstly, follow-up analyses demonstrated a non-statistically significant increase in EXT from post-treatment to 6-month follow-up or greater, and effects at post-treatment held at follow-up for INT. This suggests that at minimum, meta-analytic evidence demonstrates that the effects of AE remain beyond immediate post-intervention for both EXT and INT. Previous research suggests that AE may produce sleeper effects (Bakermans-Kranenburg et al.,
2003). For example, a study in our review that compared an AE (Connect; Moretti et al.,
2018) to several BPT interventions found that while BPT showed the greatest effect post-treatment, the AE continued to show improvements at 2-year follow-up such that there were no longer differences in effect sizes between BPT and AE (Högström et al.,
2017). If future research confirms sleeper effects, this could be unique to AE relative to the broader parenting intervention literature. For example, a review of 40 RCTs found that BPT had sustained effects—effects sizes remained the same—for up to 3 years without further improvements (van Aar et al.,
2017). Regardless, at present, our meta-analysis of 38 studies supports comparable sustained effects at follow-up for AE.
Secondly, study risk of bias appeared to impact AE outcomes, such that low-risk studies produced results that are more favorable for AE compared to waitlist comparators on EXT and INT. Only 13 studies within this review were considered low risk of bias indicating an important limitation in the existing literature, which may also understate the overall effects of AE on EXT and INT.
The current review also attempted planned subgroup analyses including sample type, age, caregiver type, delivery method, and baseline levels of EXT and INT; however, heterogeneity and/or limited studies prevented consistently quantitatively synthesizing results. Individual study findings suggested trends toward clinical samples and higher baseline levels of EXT/INT showing greater change following AE relative to comparators. These trends are comparative to previous research that has found that parenting interventions are more effective in decreasing problem severity when initial problem severity is higher (Leijten et al.,
2013; McMahon et al.,
2021). A greater proportion of studies with non-birth parents relative to birth parents showed more favorable results, though there were very limited studies to determine a pattern. Previous research has been inconclusive as to whether socioeconomic characteristics are potential moderators of intervention effectiveness (McMahon et al.,
2021). Regarding our qualitative results, it appeared that AE may have showed more favorable treatment outcomes for preschool and school-aged children relative to infants and adolescents. However, we are cautious in interpreting this potential pattern of results as there were limited studies with children aged under 3 or over 12 years. Thus, this is an important area for further research. It is also important to note that previous research has been mixed on whether age is a moderator on psychosocial interventions for children and adolescents (e.g., McMahon et al.,
2021). Finally, no clear patterns emerged for the impact of delivery method on treatment outcomes.
In addition to planned analyses, exploratory subgroup analyses found that AE specifically developed to target child and adolescent mental health outcomes showed more favorable outcomes for EXT and INT. These AE tended to include components or modules that focused on parent–child conflict and/or parenting skills to manage difficult child behavior. Future research should investigate which specific strategies included in AE may fuel greater reductions in EXT or INT, in addition to understanding what mechanisms may explain these effects. Microtrials of common components in AE to isolate mechanisms of effects could be utilized to this end (e.g., see Leijten et al.,
2015).
In addition to child EXT and INT, our review investigated secondary parent mental health and wellbeing outcomes, which included depression, anxiety, and stress. Parent mental health outcomes—most typically, depression—were quantitatively synthesized, though surprisingly, no statistically significant effects were observed for AE against any comparators. This is inconsistent with a previous meta-analysis of an AE that found reductions in parent depressive symptoms (Yaholkoski et al.,
2016); however, other AE studies and parenting program research have also found null findings (Baradon et al.,
2018; Jeong et al.,
2021). Previous research suggests that parent training programs may significantly improve parent variables that are most proximal to the intervention such as parenting stress and perceived parenting competence rather than distal outcomes such as depression (Colalillo & Johnston,
2016). Hence, it could be hypothesized that if interventions more closely targeted parent mental health and wellbeing, greater changes may be observed. Connect is one example of an AE that considered mental health outcomes in its development and has shown positive improvements in this domain (Osman et al.,
2017).
Although there were insufficient studies to pool parent stress quantitatively, five of seven studies were in favor of AE reducing parental stress relative to comparators, suggesting potential benefits for improving parental stress. This also highlights an important gap in AE research, in that similar to child outcomes, parent measures are usually focused on attachment-related constructs such as caregiver sensitivity (Steele & Steele,
2018), and effects on parental mental health are scarcer.
Limitations
To the best of our knowledge, this is the first quantitative synthesis of the effectiveness of AE for child and adolescent EXT and INT. Our review overcomes shortcomings of previous studies by taking a comprehensive and inclusive approach to ensure all controlled trials of published AE that assessed EXT and INT post-treatment were included. We were intentionally broad in our included presentations, caregivers, age range, and other demographic variables.
Notwithstanding these strengths, this study has several important limitations. First, many studies we included did not require participants to have baseline clinical levels of EXT or INT. Previous research has demonstrated that baseline EXT moderates treatment effects, such that greater reductions in EXT are linked to higher baseline EXT (Leijten et al.,
2013; McMahon et al.,
2021). Due to insufficient studies, missing baseline data, and heterogeneity, our review was unable to quantitatively investigate whether baseline clinical level was a moderator across comparators. Second, five studies meeting inclusion criteria for this review were unable to be used in quantitative analysis due to missing data that were not provided by authors when requested. Third, 12 included studies were considered to have a high risk of bias, which is a limitation of intervention research. In our study, removing studies with high risk of bias, including non-randomized studies, did not appear to reduce overall effect sizes; however, this is still a potential issue that should be considered in future research. In particular, intervention fidelity was rated the highest risk of bias meaning that we were unsure of the extent to which some studies may have delivered the intervention as program developers intended it. Fourth, the current review chose a common recommended statistical approach for meta-analysis that relies on the assumption that the control and intervention groups have similar baseline characteristics (Cuijpers et al.,
2017). Given the risk of bias found, this may not have consistently been the case and could have affected the pooled effect sizes. Since lower risk of bias studies showed more favorable outcomes, it is possible that treatment effects in this review are underestimated. Fifth, while this meta-analysis did not report results if statistical heterogeneity was high, there was clinical heterogeneity among types of active comparators. We followed planned protocol by pooling all active comparators, though there were insufficient studies to comprehensively investigate subgroups by type of active comparator. Further research needs to consider relative efficacy of AE to other treatments, including BPT as more RCTs are published. Finally, this meta-analysis primarily relied on parent-report outcomes (75% of included studies), though child-report, teacher-report, or observational data may have shown different results (e.g., see Fearon et al.,
2010).
Publisher's Note
Springer Nature remains neutral with regard to jurisdictional claims in published maps and institutional affiliations.