Longitudinal Outcomes of Child Parent Psychotherapy: Response to Commentaries

Abstract

In response to the commentaries provided by Chu, Gosh Ippen, and Lieberman (2020), Harmon, Hersh, Wei, and Weisz (2020), and McMahon & Maxwell (2020) on our longitudinal follow-up of Child-Parent Psychotherapy (CPP) with mothers with depression and their children, we focus on two domains: accessibility and scalability of CPP and identifying empirically supported mechanisms of change in attachment intervention research. In considering the accessibility and scalability of CPP, we discuss issues related to attrition, length of intervention, and implementation with caregivers with depression. Our discussion of mechanisms of change in attachment interventions explores active comparison conditions, theorized mediators, intervention modalities, assessment methods, and longitudinal research designs. This conversation is intended to highlight important areas for future research in the field of attachment interventions, with the goal of informing clinical and systems-level policies and practices.

In order to understand whether attachment interventions are truly effective in impacting child development over time, the field of attachment research must turn its attention to longitudinal studies. Existing follow-ups to attachment interventions, however, rarely follow children beyond a year or two post-intervention (for a notable exception, see a recent study by Zajac, Raby, & Dozier, 2020). Although our longitudinal examination of Child-Parent Psychotherapy (CPP) begins to shed light on its maternal and child outcomes as far as middle childhood, we know the field has a long way to go in building a strong evidence base for the long-term developmental outcomes of attachment interventions. In light of this, we welcome and appreciate the commentaries by Chu, Gosh Ippen, and Lieberman (2020), Harmon, Hersh, Wei, and Weisz (2020), and McMahon & Maxwell (2020), which can help guide the field as it builds this critical area of research. We focus our response to these commentaries on two main domains we hope future researchers will pursue: 1) the accessibility and scalability of CPP as an attachment intervention, and 2) identifying empirically supported mediators of change in attachment intervention research.

Accessibility and Scalability of CPP

CPP is built on a psychodynamic model that capitalizes on cultivating a strong relationship between the therapist and caregiver to model reflective and sensitive caregiving behavior and provide a corrective emotional experience (Lieberman, Ghosh Ippen, & Van Horn, 2015; National Child Traumatic Stress Network, 2012b). The approach is supportive and non-didactic. As a result, CPP is inherently a time- and resource-intensive intervention. To become a CPP therapist, clinicians are required to attend 42 hours of training and participate in 23 hours of consultation calls. Further, the intervention itself lasts 10–12 months on average. Both Harmon and colleagues (2020) and McMahon and Maxwell (2020) question whether the resource-intensive nature of CPP may create barriers to accessibility and scalability. Further, such a time-intensive intervention incurs ongoing costs while limiting the number of families that clinicians and clinics can serve. In addition, the substantial time investment required of clients may create client-level barriers to services and select for those with greater resources and a stronger intrinsic motivation to change. Even in the middle-class, well-educated, largely self-selected sample in our study, there was a 30% attrition rate in the intervention group. Both Harmon and colleagues (2020) and McMahon and Maxwell (2020) reference alternative interventions that similarly target attachment and parenting, such as Incredible Years (IY; Webster-Stratton, 1981), Parent Child Interaction Therapy (PCIT; Hembree-Kigin & McNeil, 1995), Attachment and Biobehavioral Catch-Up (ABC; Dozier, Higley, Albus, & Nutter, 2002), Circle of Security (COS; Marvin, Cooper, Hoffman, & Powell, 2002), and Video-Feedback Intervention to Promote Positive Parenting (VIPP; Van Zeijl et al., 2006), and wonder what gains CPP offers over these less time- and resource-intensive alternatives. These are excellent points that warrant further discussion to better understand the accessibility and scalability of CPP, particularly with depressed mothers. In response to these points, we will focus on three major issues: attrition, intervention duration, and implementation with depressed mothers.

Attrition

Both Harmon and colleagues (2020) and McMahon and Maxwell (2020) cite our study’s 30% attrition rate in the CPP condition as cause for concern regarding CPP’s accessibility, particularly for lower resourced families. Across studies of CPP, rates of attrition have not varied substantially depending on socioeconomic status (SES), suggesting its time commitment does not exclude participation based on income. The lowest rate of attrition reported in a CPP study was 14.3% in a racially, ethnically, and socioeconomically diverse sample of children and mothers exposed to interpersonal violence (Lieberman et al., 2005). This rate was comparable to the 12% attrition observed in the comparison condition, suggesting attrition was not merely a function of participation in the intervention (Lieberman et al., 2005). In a low SES, Spanish-speaking sample of anxiously attached mother-child dyads, attrition was 18% for both the intervention and control groups (Lieberman et al., 1991). Among foster families with children experiencing symptoms of traumatic stress, 28% of families assigned to CPP dropped out of treatment (Weiner et al., 2009). The highest rates of attrition have been observed in the application of CPP as a preventive intervention. In a non-treatment-seeking, low-income sample of maltreating and non-maltreating mother-infant dyads, the attrition rate for CPP was comparable to the rate observed in the current study: 39.6% of mothers randomly assigned to CPP did not participate (Cicchetti, Rogosch, & Toth, 2006). Interestingly, the rate of non-engagement was even higher for mothers assigned to the active psychoeducation comparison condition, with 51% of these mothers declining participation (Cicchetti, Rogosch, & Toth, 2006). Importantly, both Cicchetti and colleagues’ (2006) trial with a maltreated sample and Toth and colleagues’ (2006) trial with depressed mothers are examples of preventive interventions. Therefore, it is important to consider difficulties in engagement in preventive interventions when understanding rates of attrition. Although rates of attrition range from 14% to 40% across studies of CPP, there appears to be little difference in attrition by treatment group, suggesting attrition is not merely a function of CPP itself, but rather of participation in intervention studies overall.

In order to form a clearer understanding of attrition in CPP, however, it is important to examine whether rates are higher for this intervention compared to other parenting and attachment interventions. Unfortunately, in our review of the literature we found that many studies did not provide information on attrition. Further, attrition rates varied widely both across interventions and among studies of a single intervention, suggesting that as observed with CPP, rates may differ based on sample and study characteristics. Comparable or greater rates of attrition have been observed for ABC (22%; Zajac et al., 2020), PCIT (25% to 69%; Lanier et al., 2011; Nixon, Sweeney, Erickson, & Touyz, 2003; Webb et al., 2017), COS (17.5%; Cassidy et al., 2017), and IY (10%; Jones et al., 2007). In order to appropriately interpret these statistics and their implications for the accessibility of different attachment interventions, more research is needed to better understand factors related to attrition and retention both within and across interventions. Further, it is important to consider that high rates of attrition do not mean that an intervention is inaccessible. For example, several families in the Toth et al. (2006) study were lost due to moving out of the area, which does not have the same implications as dropping out of treatment because the treatment is too time-intensive. Therefore, in response to Harmon and colleagues’ (2020) and McMahon and Maxwell’s (2020) important observation, it appears that CPP is not unique in its rates of attrition. However, there may be other factors to consider with respect to its accessibility, which we will consider next.

Duration

We share the question raised in the commentaries about the duration of CPP and its implications for accessibility and scalability. Currently, Toth, Manly, and colleagues are conducting an investigation about the dose requirements of CPP to determine if a 6-month version of the intervention is as effective as the 12-month model (Promotion of Successful Parenting [Project PROMISE]; P50HD096698, sub-project 8262; Cerulli, Cicchetti, Handley, Manly, Rogosch, & Toth, 2021). In this study, CPP is being administered to a sample of low-income mothers pre- and/or postnatally for a period of 6 or 12 months. This service is provided in combination with home visitation by community health workers to address mothers’ concrete needs and reduce poverty-related stressors. The outcomes of this trial will provide important information about whether CPP must be administered for a full year in order to maintain its efficacy, or whether there are certain individuals for whom a shorter dose may be more appropriate.

McMahon and Maxell (2020) also reference the meta-analyses by Bakermans-Kranenburg and colleagues (2003; 2005) which found shorter attachment interventions to be more effective. In the 18 years since the first of these papers was published, a number of additional studies have been conducted, including two follow-up meta-analyses. Facompré and colleagues (2018) examined the effect of attachment interventions on disorganized attachment in 16 studies published between 1989 and 2016, including 7 new studies not included in Bakermans-Kranenberg’s (2005) meta-analysis. They found that the number of sessions did not significantly moderate treatment effects (Facompré et al., 2018). A similar study by Mountain and colleagues (2017) examined the effect of only those attachment intervention studies conducted between 2002 and 2015 and identified 4 papers published after Bakermans-Kranenberg and colleagues’ (2003) meta-analysis. They found mixed results with respect to visit dosage and program duration (Mountain et al., 2017). No studies of CPP were included in this more recent meta-analysis, although several were published during this period (Cicchetti, Rogosch, & Toth, 2006; Lieberman, Van Horn, & Ghosh Ippen, 2005; Toth, Rogosch, Manly, & Cicchetti, 2006; Weiner, Schneider, & Lyons, 2009). Together, this more recent research raises the question of whether Bakermans-Kranenberg and colleagues’ (2003) finding that “less is more” still holds true.

Implementation with Depressed Mothers

Harmon and colleagues (2020) ask important questions about how CPP was or was not tailored for implementation with depressed mothers and whether the model can flexibly address the unique characteristics of the participating children, mothers, and dyads. In response to this point, we must first provide a more detailed explanation of the theory behind CPP. CPP is an inherently flexible model that is designed to meet the needs of a variety of clients. It does not attribute challenges in the parent-child relationship solely to a lack of parenting knowledge and skill, but rather to a lack of sensitivity and responsivity stemming from the caregiving environments that mothers experienced in their own childhoods. During the sessions, therapists and mothers engage in joint discussions about perceptions and observations of the participating children. The therapist’s empathic responsiveness with the mother and the child allows the mother to explore her misperceptions of her child. Through respect, empathic concern, and positive regard, the therapeutic relationship provides the mother with a corrective emotional experience. Mothers are consequently able to expand their responsiveness, sensitivity, and attunement to their child, fostering security in the mother-child relationship. Therefore, in response to the question raised by Harmon and colleagues (2020), we argue that no modifications were necessary to implement CPP with depressed mothers and their children. Rather, CPP is designed to meet the needs of diverse parents and children.

Chu and colleagues (2020) highlight that although CPP did not impact maternal depression in our study, it still improved attachment and child and maternal outcomes. Studies of CPP in two separate samples of depressed mothers have consistently found CPP to be effective in fostering secure attachment and other positive child and family outcomes in the absence of significant change in maternal depression (Cicchetti, Rogosch, & Toth, 2000; Cicchetti, Toth, & Rogosch, 1999; Guild, Toth, Handley, Rogosch, & Cicchetti, 2017; Peltz, Rogge, Rogosch, Cicchetti, & Toth, 2015; Toth et al., 2006). Although CPP has proven effective in decreasing symptoms of depression among mothers experiencing interpersonal violence, these mothers differed from the population in our study in that they were not recruited based on depressive status, but rather based on exposure to interpersonal violence (Ghosh Ippen et al., 2011). Further, results from Ghosh Ippen et al.’s (2011) study found differences in CPP’s effect on depressive symptoms based on the number of traumatic experiences to which a mother had been exposed, which suggests depression outcomes may be linked to comorbidity with post-traumatic stress disorder. As Chu and colleagues (2020) remind us, CPP was not designed to alleviate symptoms of depression in parents, but rather to address symptoms of traumatic stress and the impact of these symptoms on caregiving behavior. It is therefore important that the findings of this study are interpreted not as evidence for CPP in treating maternal depression, but rather as evidence for CPP changing the behavior of depressed mothers to improve their relationships with their children and their children’s socioemotional outcomes. However, we share the curiosity of Chu and colleagues (2020) with respect to CPP’s interaction with maternal psychopathology. In light of this, our ongoing Project PROMISE (Cerulli et al., 2021) will examine moderation effects of maternal characteristics at baseline on CPP outcomes, including maternal depression, trauma history, and intimate partner violence. This aim will create an excellent opportunity to answer the questions raised in the commentaries about the fit of CPP for mothers with depression.

Mechanisms of Change in Attachment Interventions

As Harmon and colleagues (2020) aptly highlight, the future of intervention research is directed toward identifying core elements of treatments that are directly linked to their effects. In other words, how did CPP change attachment? Identifying mechanisms of action would not only advance understanding of developmental processes, but would also create the opportunity to distill the intervention to its active ingredients. As a result, interventions can become more efficient, which can address concerns related to time and resource investment, as well as accessibility and scalability. We agree that this is a critical direction for future research that can be best addressed by focusing on the following domains: comparison to active comparison conditions, empirically testing theorized mechanisms of change, identifying effective intervention modalities, examining methods of assessment, and conducting longitudinal studies to understand developmental trajectories.

Active Comparison Conditions

Harmon and colleagues (2020) explain that a key step toward understanding mechanisms of change is to compare the treatment in question (i.e., CPP) to an active control. The gold standard approach to determining CPP’s unique effects would be comparing CPP to not just any active control, but rather to another attachment intervention, such as one of the interventions mentioned above. Unfortunately, no such study exists. The lack of a direct comparison between attachment interventions is not unique to CPP; to our knowledge, the first study of this kind, which is comparing PCIT to COS, is currently underway (Kohlhoff et al., 2020).

In the absence of a study comparing CPP to another attachment intervention, we must instead examine existing studies comparing CPP to an active control. In a sample of maltreating families, CPP was compared to a psychoeducational parenting intervention and both treatments were found to increase rates of secure attachment (Cicchetti, Rogosch, & Toth, 2006). However, CPP emerged as more effective than the psychoeducation condition in decreasing maladaptive maternal representations and negative child self-representations, and in increasing positive mother-child relationship expectations (Toth, Maughan, Manly, Spagnola, & Cicchetti, 2002). Although these studies begin to provide important information about the effects of CPP in comparison to an active control, we agree that a direct comparison of CPP to another attachment intervention would be more illuminating regarding its unique effects. We would like to highlight this suggestion as a critical future direction for research.

Theorized Mechanisms of Change

There is a lack of empirical evidence on mechanisms of change in attachment interventions broadly, not just within studies of CPP. Attachment interventions are designed to target parents’ theorized behavioral precursors to secure attachment, including sensitivity, responsiveness, appropriate stimulation, interactional synchrony, warmth, involvement, and mutually responsive interactions (see Fearon & Belsky, 2016). Equally important is preventing or changing parents’ behavioral precursors to disorganized attachment, which include insensitivity, frightened, frightening, and atypical parental behavior (see Lyons-Ruth & Jacobvitz, 2016). Although a recent study has identified decreased levels of parental withdrawal as a mechanism of change for ABC’s effect on attachment security (Yarger, Bronfman, Carlson, & Dozier, 2020), there is little additional empirical support for these proposed mechanisms in the context of any attachment intervention, including CPP (Toth, Alto, & Warmingham, in press). As a result, we enthusiastically agree that this is a critical avenue for future research.

McMahon and Maxwell (2020) highlight working models of attachment as one potential mechanism of change warranting further attention. We agree that representational models are important to consider in the context of attachment interventions and have found that CPP is effective in improving them in both children and mothers from maltreating families (Toth et al., 2002). However, in Toth and colleagues’ (2002) study, change in representations was not examined in association with change in attachment security, which would have provided important information about mechanisms of action in CPP. As a result, representational models continue to be a focus for Toth and colleagues in their examination of CPP. Another aim of Project PROMISE (Cerulli et al., 2021) is to examine internal representations (as well as parenting cognitions and stress responsivity) as a mechanism of change in CPP.

Intervention Modalities

In addition to identifying behavioral mechanisms of change, it is important for future research to identify intervention modalities associated with change. For example, CPP’s intervention modalities include 1) promoting developmental progress through play, physical contact, and language; 2) offering unstructured reflective developmental guidance; 3) modeling appropriate protective behavior; 4) interpreting feelings and actions; 5) providing emotional support and empathetic communication; and 6) providing crisis intervention, case management, and concrete assistance with problems of living (Lieberman et al., 2015). In contrast, video feedback is an important intervention modality for ABC, VIPP, and COS. PCIT is characterized by live bug-in-the-ear coaching, and IY is designed around didactic lessons, including video vignettes, worksheets, and group discussion. Each of these interventions takes a different approach to targeting similar underlying mechanisms of change. Future attention to differences in effectiveness among intervention modalities will help answer the question of how to target mechanisms of change in the most efficient and effective way.

Assessment Methods

Another important part of understanding mechanisms of change is examining the methods used to assess those mechanisms. Harmon and colleagues (2020) wonder whether there was overlap in our study between what mothers learned in CPP and the behavior assessed in the Strange Situation (Ainsworth, Blehar, Waters, & Wall, 1978). Classification of secure, insecure-ambivalent, and insecure-avoidant attachment in the Strange Situation is based on infants’ scores in four major domains, each characterized by specific behaviors: proximity-seeking (e.g., active approach, reaching, etc.), contact maintenance (e.g., clinging to the parent, remaining in physical contact for an extended period of time, etc.), resistance (e.g., hitting, screaming, pushing, throwing toys, etc.), and avoidance (e.g., failure to acknowledge the parent upon re-entry, actively ignoring a parent who is trying to get the child’s attention, failure to interact with the parent throughout the course of the paradigm, etc.; Ainsworth et al., 1978). In CPP, mothers are not taught specific behaviors (i.e., responding to their baby’s outstretched hands), but are rather supported in developing a general reflective capacity to understand what her child needs when he or she is stressed. Although mothers are not taught specific skills, their increased awareness of the impact of their own history on their parenting and perceptions of their children is theorized to impact parenting behavior. Therefore, in response to Harmon and colleagues’ (2020) comment, there was no explicit overlap in what parents and children learned and practiced in CPP and the behavior assessed in the Strange Situation that might complicate interpretation of study findings. We agree, however, that this is an important question to raise in the context of intervention research and encourage future studies to thoughtful choose methods of assessing mechanisms of change.

Longitudinal Outcomes

Even outside the context of interventions, the field of attachment research lacks empirical evidence supporting longitudinal mechanisms of change across development. As McMahon and Maxwell (2020) point out, Thompson (2016) argues that evidence is still needed to support the core argument of attachment theory, which posits that changes in caregiver behavior associated with improvements in attachment security are sustained, and this stability over time explains improvements in children’s socioemotional development. Although one interpretation of our study’s findings is that CPP fostered positive maternal behavior that both improved attachment security and was stable over time, the lack of data in the intervening period between the end of the intervention and the follow-up 6 years later means this interpretation remains a conjecture without solid empirical evidence. We acknowledge that our study does not contribute to this gap in the field and agree that a rigorous empirical investigation of these developmental processes over time is sorely needed.

Summary and Conclusions

Chu and colleagues (2020), Harmon and colleagues (2020), and McMahon and Maxwell (2020) all raise excellent questions about CPP and the field of attachment intervention research in general. In order to advance the field and create a broader impact on children and families through effective dissemination and implementation of these interventions, it is important for future research to focus on implementation factors and mechanisms of change. Special attention should be paid to implementation factors related to treatment dropout and study attrition, intervention dosing, and modifications for special populations. In addition, research should focus on building empirical support for theorized mechanisms of change by conducting trials that include active comparison conditions, examining theorized mediation effects, identifying effective intervention modalities, thoughtfully choosing assessment methods, and conducting long-term follow-ups with multiple intervening assessment points to determine developmental trajectories.

As Chu and colleagues (2020) emphasize, continuing to build the evidence base around the importance of the attachment relationship in child development is critical for informing clinical and systems-level policies that cater to the parent-child relationship rather than the child in isolation. CPP is just one of many empirically supported attachment interventions that have the potential to support children and families. By continuing to focus research not only on how attachment interventions work and for whom, but also on factors that can make treatments scalable and accessible, we can start to shift systems of care toward more targeted and effective services for children and their families.