Changes in self-reported and observed parenting following a randomized control trial of Parent-Child Interaction Therapy (PCIT-ED) for the treatment of preschool depression

Abstract

Background:

Parenting in early childhood exerts substantial influence over children’s emotional health and development. Using data from a randomized controlled trial of a novel treatment for early childhood depression, Parent Child Interaction Therapy Emotion Development (PCIT-ED), we explored two broad dimensions of parenting (behavior and affect) to determine whether any changes could be detected following treatment when compared to those in a waitlist control condition.

Method:

229 caregiver-child dyads, 114 randomly assigned to PCIT-ED for preschool-onset depression and 115 assigned to a waitlist completed two structured interaction tasks at baseline and post-treatment. Interactions were later coded by observer’s blind to diagnostic and treatment status.

Results:

Greater reductions were found in self-reported negative parenting behaviors and observed negative affect and greater increases in self-reported positive parenting behaviors and observed positive affect among the caregivers in the treatment group. Increases in the overall positivity of the observed interactional style of caregivers, but no observed parenting behavior change was found following treatment. Discrepancies between self-reported and observed parenting were greater among caregivers on the waitlist.

Conclusion:

Following PCIT-ED treatment, caregivers self-reported improvements in parenting practices and declines in punitive practices along with observed increases in positive affect and decreases in negative affect when interacting with their child. Moreover, coherence between self-reported and observed parenting was higher in the treatment group. These findings highlight the efficacy of PCIT-ED in improving parenting behaviors and the need to use multiple methods to assess parenting in treatment studies.

Introduction

Preschool-onset Major Depressive Disorder (PO-MDD) is increasingly recognized as a serious psychological disorder with long-lasting implications for children’s health and development. PO-MDD leads to chronic functional impairment and is related to poor developmental outcomes across several domains of functioning (Donohue et al., in press; Luby et al., 2014; Whalen et al., 2017). A large community-based study in the United States found the overall prevalence rate of depression in three-year-olds to be 1.7% (Finsaas et al., 2018). Increasing attention has been paid to specific environmental factors such as parenting and the parent-child relationship as key contributors to the onset and maintenance of PO-MDD as well as how changes in these domains may help to ameliorate symptoms (Luby et al., 2019; Pinquart, 2017). In the current study, we investigate caregiver behavior and affect using observational and self-report measures before and after treatment during a randomized control trial of parent-child interaction therapy with an added, novel emotion development module (PCIT-ED).

PCIT is a widely-tested and empirically supported dyadic parent-child treatment for externalizing behaviors in young children (ages 3-7yrs; Brestan & Eyberg, 1998; Thomas, Abell, Webb, Avdagic, & Zimmer-Gembeck, 2017). Standard PCIT consists of two treatment components, the “child-directed interaction (CDI),” which teaches parents to interact positively with their children without criticizing or negatively responding and the “parent-directed interaction (PDI),” which teaches parents how to use nurturing yet firm limit-setting techniques, overlapping greatly with techniques of authoritative parenting. PCIT directly targets parenting in therapy by using a “bug” in the ear technique- where the therapist coaches the parent, who is wearing a small microphone, through a one-way mirror during live parent-child interactions. This allows the therapist to help modify parental behavior as it is occurring. Meta-analyses demonstrate improvements in parenting practices following PCIT, including more praise and reflections, and fewer criticisms and negative talk (Thomas et al., 2017).

Adapting PCIT in the first randomized control trial for treatment of preschool-onset depression, Luby and colleagues (Luby et al., 2018) found promising evidence for the benefits of PCIT-ED in treating PO-MDD. PCIT-ED expands PCIT by adding a novel emotion development (ED) module following completion of time limited versions of the standard modules. The emotion development module teaches caregivers, through in vivo coaching, to enhance their child’s emotional competence and regulation (Lenze et al., 2011), specifically by validating and tolerating the child’s expression of distressing emotions, decreasing reactivity to negative stimuli, and increasing reactivity to positive stimuli. Preschoolers evidenced enormous benefit from the treatment including remission from depression, lower impairment, and enhanced emotional functioning (Luby et al., 2018, 2019), and many of these benefits were sustained at a 3 month follow-up (Luby, Donohue, Gilbert, Tillman & Barch, under review).

In addition to child responses, we also found significant reductions in caregiver depression despite the fact that this was not a direct target of treatment (Luby et al 2018). More expected, and as seen in standard PCIT, was that caregivers who received the PCIT-ED treatment reported marked reductions in parenting stress and a greater sense of positive responsiveness from their child. Relevant to the current questions, following treatment, parents were more likely to self-report more emotionally focused parenting and less minimization techniques at the end of therapy (Luby et al., 2018). Interestingly, although parental depression and stress remained low at a 3 month follow-up, the gains made in adaptive parenting techniques drifted back towards baseline (Luby et al., under review) . For the current study, we utilized unique measures of observed and self-report measures of parenting that were not included in these prior reports.

Although caregivers reported changes following treatment in their behavioral response to their child’s expression of emotions, parental self-report is subject to distortion. Whether these changes can be detected observationally in the caregiver’s actual behavior and affect toward the child is a critical question that has yet to be addressed. That is, it is possible that caregivers parenting beliefs and self-perceptions were altered without true change arising in the way they interacted with their child. Indeed, a recent meta-analysis demonstrates weak associations between parent-reported and observed parenting (Hendriks et al., 2018). Despite self-report being a frequently used method, there are well-documented limitations to parents’ ability to report on their own parenting (Morsbach & Prinz, 2006), including social desirability biases leading parents to under-report negative behaviors such as hostility and physical discipline while overreporting positive behaviors such as praise along with affective dimensions such as warmth (Morsbach & Prinz, 2006; Waylen et al., 2008). On the other hand, observational measures are considered the gold standard in the field and offer superior validity as well as reliability (Gardner, 2000; Repetti et al., 2015). Observational methods offer a more objective approach to studying parental behaviors and affect, and a large body of evidence supports the importance of observing both parenting behaviors and affective dimensions of parenting (positive and negative affect toward the child) as predictors of positive outcomes throughout childhood (Kochanska et al., 2008; Morris et al., 2017).

Given that decades of research have shown that parenting behaviors and parent-child relationships exert substantial influence over young children’s emotional health and development, we sought to explore parenting behavior and affect assessed using self-report and direct observations to determine whether objectively observed change could be detected following PCIT-ED treatment when compared to those on the waitlist. Of particular interest for this study are broad indices of authoritative and authoritarian caregiving, the two most common caregiving styles. An authoritative caregiver balances high levels of demandingness with high levels of warmth and responsiveness, taking a “firm but fair” approach. On the other hand, an authoritarian caregiver expresses high levels of demandingness, but low levels of warmth and responsiveness, taking a discipline-oriented approach.

The aim of the current study is to examine the changes in self-reported and observed parenting following PCIT-ED treatment, while accounting for baseline parenting practices and depressive symptoms. We utilized unique measures of observed and self-reported parenting that have not been previously analyzed and published in this sample. We hypothesize that when compared to the waitlist group, caregivers in the PCIT-ED treatment group will report and display more authoritative parenting behaviors and fewer negative (authoritarian) parenting behaviors at post-treatment. Further, we hypothesize that caregivers in the PCIT-ED treatment group will display more positive affect and less negative affect during interactions at post-treatment when compared to the waitlist group. We also hypothesize increases in the positivity of PCIT-ED caregivers observed interactional style, defined as overall warmth and a sense of connection between parent-child, when compared to caregivers in the waitlist group. As an exploratory secondary aim, given the recent meta-analysis highlighting discrepancies in observed and self-reported parenting practices (Hendriks et al., 2018), we also wanted to determine if discrepancies between observed and reported parenting, a putative measure of accuracy of parenting self-perception, were reduced following treatment.

Method

Overview

Data were drawn from a larger, single-blind randomized control trial comparing PCIT-ED to a waiting list control condition (See Luby et al., 2018 for further details; Figure 1). A modified version of the empirically tested parent-child interaction therapy with a novel “emotion development” module was compared with a waiting list condition in a large sample (n=229) of preschoolers (ages 3-6 years) with depression. Of note, the sample was not required to meet the 127-point ECBI cutoff as in traditional PCIT, even though PDI sessions were conducted. All study materials and procedures were approved in advance by the Washington University School of Medicine (WUSM) institutional review board. Written informed consent was obtained from all caregivers and verbal assent from children. The trial was registered with ClinicalTrials.gov (NCT02076425). Caregiver-child dyads participated in a comprehensive baseline mental health and emotional development assessment at the WUSM Early Emotional Development program. Children who met criteria for early childhood major depression were randomly assigned to either PCIT-ED (n= 114) or the waitlist condition (n= 115), with randomization stratified by gender and comorbid externalizing disorders. Children and their caregivers in the PCIT-ED group completed 6 sessions of CDI and 6 sessions of PDI, for a total of 12 sessions of traditional PCIT as well as an additional 8 sessions of ED (Luby et al., 2019). Children and caregivers in the waitlist group completed 18 weeks of “watchful waiting” before completing the 12 sessions of traditional PCIT and 8 sessions of ED. For both groups, the 20 sessions were conducted across an 18 week period. Of the 114 subjects randomized to PCIT-ED, 6 did not complete any therapy sessions, and 93 completed all 20 therapy sessions. The mean (SD) number of sessions completed was 17.4(6.0).

Figure 1:

Consort Diagram of PCIT-ED Study

Figure reprinted from Luby et al., 2018

During the baseline, pre-treatment assessment, caregivers were interviewed using the Schedule for Affective Disorders and Schizophrenia–Early Childhood (K-SADS-EC) (Gaffrey & Luby, 2012) to assess the child’s psychiatric symptoms and assign DSM-5 diagnoses. Caregivers also completed a battery of psychosocial questionnaires that assessed the child’s emotion regulation and guilt processing, parental psychopathology, parenting practices, and stress. Of greatest relevance to the current study, caregivers and children completed two structured interaction tasks that were video-taped and coded (described in detail below).

During the post-treatment/post-waitlist assessment (e.g., after 18 weeks), the above procedures were repeated. All clinician-administered ratings were completed by independent raters (master’s-level clinicians) who were blind to treatment group and otherwise uninvolved in the study (see Luby et al., 2018 for more details about maintaining the blind). Caregivers and children completed two additional structured interaction tasks that were administered and coded by blinded raters.

Participants

Preschoolers ages (3.0-6.92 years) from the St. Louis metropolitan area were screened and recruited from community sources and mental health facilities using the Preschool Feelings Checklist (Luby et al., 2004), a validated screening measure for early childhood depression. All children who met symptom criteria during this phone screen for preschool-onset depression and who did not have an autism spectrum disorder, a serious neurological or chronic medical disorder, or a significant developmental delay were invited for an in-person assessment during which a full diagnostic interview was conducted (K-SADS-EC). Children and their caregivers were recruited from 2014-2017.

From the original sample of 115 caregiver-child dyads randomized to waitlist, 24 did not complete a post-treatment assessment. From the 114 caregiver-child dyads randomized to PCIT-ED treatment, 14 did not complete a post-treatment assessment. 3 additional dyads randomized to treatment did not complete the structured interaction tasks. Therefore, data from 91 dyads from the waitlist group and 97 dyads from the treatment group had structured interaction data at both assessments and were analyzed in the current report. Analyses comparing dyads who did and did not complete a post-treatment assessment are included in Table S1. Families who did not complete a post-treatment assessment had a lower income-to-needs ratio (t₍₂₂₉₎= −2.26, p= 0.03) and were more likely to have a child diagnosed with conduct disorder (X²= 4.32, p=0.04).

Measures

Preschool depression.

The K-SADS-EC is a semi-structured clinical interview for DSM-5 disorders adapted for use in children ages 3.0–6.9. This measure has test-retest reliability and construct validity, and it generates both categorical and dimensional measures of major DSM-5 disorders (Gaffrey & Luby, 2012). The presence and severity of major depression and comorbid disorders were assessed at baseline and post-treatment. All K-SADS-EC interviews were conducted by master’s-level clinicians and were videotaped, reviewed for reliability, and calibrated for accuracy. Satisfactory interrater reliability was established before the study started, and kappa values during the study were computed on a monthly basis; the overall kappa value during the study period was 0.74 for major depression and 0.88 for all diagnoses. The depression severity score was the number of core depressive symptoms endorsed on the K-SADS-EC.

Caregiver depression.

The Beck Depression Inventory–II (BDI-II), a reliable and valid self-report measure, was used to assess severity of depression in caregivers (Beck et al., 1996). This measure was used as a covariate in our analyses as dyads were not recruited/enrolled based on the presence of caregiver depression.

Self-report of parenting style.

The Parenting Styles Questionnaire (PSQ) is a caregiver-report questionnaire that measures parents beliefs about emotions and their meta-emotion philosophy (Gottman, 2011; Gottman et al., 1996). The questionnaire includes four subscales relating to parental meta-emotion philosophies: dismissing, disapproving, laissez-faire, and emotion coaching. Each of the 23 items is rated on a 5-point scale with (1) indicating strongly disagree and (5) strongly agree. The Parenting Styles and Dimensions Questionnaire (PSDQ) is a 32-item caregiver report measure assessing authoritative, authoritarian, and permissive parenting styles (Robinson et al., 2001). This measure has been widely used in studies assessing children and adolescents across cultures (Olivari et al., 2013). The PSQ and PSDQ were completed during the baseline and post-treatment assessments. See Table S2 for detailed definitions.

Income-to-needs ratio.

Income- to-needs ratio was computed as the total family income at baseline divided by the federal poverty level, based on family size at baseline (McLoyd, 1998).

Observational tasks and coding

Baseline (pre-treatment).

Two structured observational tasks designed by Kochanska and Aksan (Kochanska & Aksan, 1995, 2004) were completed by each caregiver-child dyad during the baseline assessment. The Marble Run task asked caregivers and children to build a standing marble run, replicating a picture. The Etch-a-Sketch task asked caregivers and children to work together (each controlling one dial) to make their way through a maze on an etch-a-sketch. Both tasks were designed to induce mild stress, require caregiver assistance for completion, and provoke negative emotions from the child. All tasks were videotaped and parent behavior/affect was coded using procedures described below. Child behavior and affect was also collected, but that is not included in the current report.

Post-treatment (post-treatment/waitlist).

Two different structured observational tasks were completed by each caregiver-child dyad during the post assessment. Two unique tasks were chosen to avoid any practice or carryover effects from the previous completion of the initial tasks, as these were specifically chosen to elicit frustrative, negative emotions in the children. The Magnet Shape task asked caregivers to help their child complete a puzzle with various shaped magnets that was conceptually advanced for the child’s developmental level. During the Drawing task, each caregiver-child dyad was asked to create a picture by taking turns and not lifting their crayon off of a piece of paper. Similar to the tasks administered at the baseline assessment, each was designed to require caregiver assistance for successful completion and elicit negative emotions from the child. All tasks were videotaped and parent behavior/affect was coded using procedures described below. Child behavior and affect was also collected, but that is not included in the current report.

Coding.

All observations were video-taped and coded using the Dyadic Parent-Child Interactions in Early Childhood, PCIT-ED edition manual (Whalen & Gilbert, 2017). This manual was adapted from the Dyadic Interaction Coding Manual (Lunkenheimer et al., 2011) for use in our laboratory assessments. See Table S2 for a detailed description of codes and composites. This system is designed to capture a descriptive landscape of affect displays and functional control/compliance behavior in caregivers and their preschool-age children. A unique feature of this coding scheme is that affect, behavior, and global interactional style are coded independently from one another. For example, a caregiver may verbally discipline the child while maintaining a positive affect display. Alternatively, a caregiver may engage in physical affection with the child while displaying negative affect such as frowning or eye rolling. This feature of the coding scheme allowed for the investigation of affect, behavior, and interactional style separately. The focus of the current report is solely on the caregiver behaviors and affect displays. A team of coders blind to the treatment status of participants, as well as study hypotheses independently watched and scored each video using the Noldus Observer XT software (Zimmerman et al., 2009). Videos were randomly assigned to each coder. All coders were required to achieve greater than 80% reliability with master coders before being able to code independently. Two master coders (DW and KG) coded 20% of random videos to verify inter-observer agreement was maintained over time. Individual codes were then organized into larger categories meant to fit an over-arching theoretical framework: positive behaviors, negative behaviors, positive affect, negative affect, and global interactional style (Table S2). We also computed a positive-to-negative affect ratio by dividing the number of positive affect displays by the number of negative affect displays.

Data analysis plan

Baseline (Pre-treatment) demographic, diagnostic, and severity characteristics were compared in the PCIT-ED and waiting list groups including just those who participated in the structured interaction tasks using t-tests for continuous variables and chi-square tests for dichotomous variables. All observational variables were turned into proportions such that the durations of each code were divided by the length of the interactions (to account for any variation in length of the PCI and therefore opportunity to display parenting behaviors). This was done separately at baseline and post-treatment.

For the primary analyses, caregivers who were randomly assigned to the treatment were compared to those assigned to the waitlist condition using ANCOVAs. The primary outcome measures for this report were self-reported and observed caregiver behaviors and affect at the post-treatment (post-treatment/waitlist) assessment. We calculated variation from pre-to-post using change/difference scores (Cohen et al., 2014; Twisk et al., 2018). All models covaried for the baseline characteristic corresponding to the dependent variable and the stratification variables of child sex and baseline externalizing disorder. All models also covaried for caregiver depression symptoms and child depression severity, both theorized to have potentially significant impacts on observed parenting behaviors/affect (Berg-Nielsen et al., 2002; Lovejoy et al., 2000). Graphical methods (e.g., histograms and q-q- plots) were examined and tests for normality (e.g., Kolmogorov-Smirnov and Shapiro-Wilk) were conducted in the residuals of the dependent variables by group. For variables found to have a non-normal distribution, non-parametric Kruskal-Wallis H tests were also conducted.

Observed and self-reported parenting behaviors and affect at each assessment were then classified and dichotomized into theoretically based groups: positive self-reported, positive observed, negative self-reported, negative observed (Table S2). 2(baseline/post-treatment) X 2(positive/negative) cross-tabulation tables were used to compare the proportion of parents in each group who either self-reported or were observed to display each parenting behavior/affect. For each parental behavior and affect (e.g., positive self-reported, positive observed, negative self-reported, negative observed), non-parametric tests for repeated samples (McNemar’s test) were used to: (1) examine the changes from baseline to post in behaviors/affect and (2) examine the discrepancies between behaviors/affect that caregivers reported and behaviors/affect that were observed during caregiver-child interactions. The McNemar test assesses for consistency in responses across two variables while accounting for random change. For example, the test compares whether a caregiver reports negative parenting behaviors before and after PCIT-ED treatment (or waitlist) while accounting for the fact that some participants will randomly switch between authoritative/positive and authoritarian/negative groups. This tests whether the PCIT-ED treatment worked to change caregivers from “high” to “low” levels of authoritarian/negative parenting behaviors, for instance. If the PCIT-ED treatment had an effect on negative parenting behaviors, the number of caregivers who move from high to low will be higher than the number that move from low to high.

Power analysis

Sample size for an ANCOVA of two levels, with five covariates was determined using a power analysis (Faul et al., 2007). The power analysis was conducted using an alpha of 0.05, a power of 0.80, and a medium (f= 0.25) effect size. The sample sizes needed was 100, thus we are well-powered to detect medium effects.

Results

Baseline demographic, diagnostic, and severity characteristics among the PCIT-ED (n=97) and wait list (n=91) groups used in the current study are found in Table S3. The groups did not differ in terms of demographic characteristics, child diagnoses, child depression severity, or caregiver depression.

Baseline (pre-treatment) Parenting Behavior and Affect Across Waitlist and Treatment Groups

As shown in Table 1, the groups did not differ in terms of self-reported parenting on the PSDQ. The emotion coaching subscale of the PSQ was significantly different between the two groups at baseline (t₍₁₈₈₎= −2.27, p= 0.02) with the waitlist group (M=17.59; SD= 2.260) reporting fewer emotion coaching parenting practices than the treatment group (M=18.41; SD= 2.35). The groups did not differ in observed parenting behaviors or affect or task length during the baseline assessment.

Table 1.

Baseline (Pre-treatment) parenting characteristics

Measure	Wait List (N= 91)		PCIT-ED (N= 97)		Comparison
	Mean	SD	Mean	SD	t	p
Self-Report
PSQ: Dismissinga	9.98	3.80	10.12	3.31	−0.28	0.78
PSQ: Disapproving	6.87	3.84	6.71	4.18	0.27	0.79
PSQ: Laissez-Faire	6.34	1.20	6.37	1.27	−0.17	0.87
PSQ: Emotion Coaching	17.59	2.60	18.41	2.35	−2.27	0.02
PSDQ: Authoritative	105.97	11.54	106.21	9.48	−0.16	0.88
PSDQ: Authoritarian	39.63	6.90	39.02	5.97	0.64	0.52
PSDQ: Permissive	33.23	6.39	32.72	6.11	0.56	0.58
Observational
Negative behaviorsb	0.03	0.04	0.03	0.03	0.41	0.68
Positive behaviors	0.47	0.04	0.47	0.04	−0.26	0.80
Negative affect	0.04	0.65	0.05	0.07	−0.23	0.82
Positive affect	0.21	0.17	0.21	0.15	0.04	0.97
Positive to Negative affect ratio	1.16	0.19	1.16	0.17	0.12	0.91
Global Positive interactional stylec	2.85	0.95	2.73	0.92	0.91	0.36
Global Negative interactional style	1.68	0.77	1.70	0.85	−0.15	0.88
Task Length	12 minutes, 47 seconds	3 minutes, 11 seconds	12 minutes, 56 seconds	3 minutes, 6 second	−0.31	0.76

^aPSQ- Parenting Styles questionnaire; PSDQ- Parenting Styles and Dimensions questionnaire

^bObservational parenting behaviors represents proportions taking the frequency of each code divided by the total duration of the interaction. Observational parenting affect represents proportions taking the duration of each code divided by the total duration of the interaction

^cGlobal ratings are coded on a Likert scale with 1-Very Low; 3-Moderate; 5-Very High

Changes in Parenting Behavior and Affect Following Treatment/Waitlist

Changes in self-reported and observed parenting behaviors and affect following treatment were calculated via ANCOVA using simple difference scores (post minus baseline). All models covaried for the baseline (pre-treatment) characteristic corresponding to the dependent variable and the stratification variables from the larger RCT of sex and baseline externalizing disorder. In addition, caregiver depression symptoms and child depression severity were covariates theorized to have potentially significant impacts on observed parenting behavior and affect¹.

One-way ANCOVAs to test for treatment group differences in self-reported and observed parenting behavior and affect while controlling for child sex, child externalizing disorder, caregiver depression symptoms, child depression severity, and the baseline characteristic corresponding to the dependent variable were conducted using difference scores as the dependent variables. There was a significant effect of treatment group on all self-reported parenting variables as well as observed negative affect, positive affect, the PA/NA affect ratio, and global interactional styles (Table 2). Specifically, reductions from baseline to post-treatment were found for the PCIT-ED group in dismissive, disapproving, laissez-faire, permissive, and authoritarian parenting styles when compared to the waitlist group. Increases from baseline to post-treatment were reported for the PCIT-ED group in emotion coaching, and authoritative parenting styles.

Table 2.

ANCOVA tests using difference scores comparing self-reported and observed parenting behavior and affect between the waitlist and treatment groups

	Wait List (N= 91)		PCIT-ED (N= 97)
	Mean	SD	Mean	SD	F(df)	p	Partialη2a
Self-Report
PSQ: Dismissingb	−1.20	0.29	−3.15	0.28	24.20 (188)	.000	0.12
PSQ: Disapproving	−1.43	0.31	−3.27	0.30	16.61 (188)	.000	0.01
PSQ: Laissez-Faire	−0.38	0.13	−0.78	0.12	5.41 (188)	0.02	0.03
PSQ: Emotion Coaching	0.62	0.22	1.24	0.22	4.13 (188)	0.04	0.02
PSDQ: Authoritative	1.33	0.78	4.11	0.75	6.90 (188)	0.01	0.04
PSDQ: Authoritarian	−1.50	0.52	−6.23	0.50	45.77 (188)	.000	0.20
PSDQ: Permissive	−1.41	0.47	−4.45	0.45	23.23 (188)	.000	0.12
Observational
Negative behaviorsc	0.02	0.01	0.01	0.01	0.63 (184)	0.43	.00
Positive behaviors	0.60	0.05	0.69	0.05	1.58 (184)	0.21	0.01
Negative affect	0.07	0.02	0.03	0.02	4.84 (184)	0.03	0.03
Positive affect	0.26	0.05	0.47	0.04	11.35 (184)	0.001	0.06
Positive to Negative affect ratio	0.18	0.04	0.41	0.04	15.75 (184)	.000	0.08
Global Positive interactional styled	−0.08	0.10	0.36	0.10	10.88 (184)	0.001	0.06
Global Negative interactional style	0.25	0.08	−0.09	0.08	8.73 (184)	0.004	0.05

^aPartial η²- measure of effect size reflecting the percentage of the variance in each dependent variable explained by treatment group

^bPSQ- Parenting Styles questionnaire; PSDQ- Parenting Styles and Dimensions questionnaire

^cObservational parenting behaviors represents proportions taking the frequency of each code divided by the total duration of the interaction. Observational parenting affect represents proportions taking the duration of each code divided by the total duration of the interaction

^dGlobal ratings are coded on a Likert scale with 1-Very Low; 3-Moderate; 5-Very High

During the caregiver-child interactions, the duration of negative affect displayed by those in the PCIT-ED group significantly decreased from baseline to post-treatment when compared to the waitlist group. The duration of positive affect that was displayed significantly increased from baseline to post-treatment when compared to the waitlist group. The PA/NA ratio significantly increased for the PCIT-ED group from baseline to post-treatment. There was a significant increase in observed positive interactional styles (e.g., overall warmth and a sense of connection between parent-child) and a significant decrease in observed negative interactional styles (e.g., overall negativity and a sense of low mutuality/connection between parent-child) from baseline to post-treatment in the PCIT-ED group only. There were no significant increases or decreases in observed positive behaviors or negative behaviors from baseline to post-treatment in either group, however it should be noted that the rates of these behaviors were low.

Examination of Q-Q plots and results from the Kolmogorov-Smirnov tests indicated that the residuals of several dependent variables within each group were not normally distributed (negative behaviors, positive behaviors, and negative affect). Therefore, we conducted non-parametric Kruskal-Wallis H tests to ensure the validity of our results, as these tests are distribution free and robust to nonnormality. Replicating the results of the ANCOVA, a Kruskal-Wallis H test showed that there was not a statistically significant difference in negative behaviors between the groups, χ²₍₁₎ = 0.91, p = 0.34, with the mean rank negative behaviors of 89.24 for the waitlist group and of 97.06 for the PCIT-ED group. A similar result was found for positive behaviors with the mean rank of 93.47 for the waitlist group and of 92.56 for the PCIT-ED group (χ²₍₁₎ = 0.05, p = 0.83). A final Kruskal-Wallis H test showed that there was a statistically significant difference in negative affect between the groups, χ²₍₁₎ = 3.94, p = 0.04, with the mean rank negative affect of 101.24 for the waitlist group and of 85.36 for the PCIT-ED group.

Discrepancies between reported and observed parenting behavior and affect

Self-reported and observed parenting variables at baseline and post-treatment were theoretically grouped: self-reported positive, self-reported negative, observed positive, and observed negative (Table S2). McNemar tests were then conducted on the dichotomously classified parenting variables. The number of caregivers who reported negative and positive parenting at baseline and those who reported them at post-treatment were identified, with tests conducted separately for the treatment and waitlist groups. McNemar’s test showed significant differences between self-reported parenting at baseline and post-treatment (Figure 2). Caregivers in the PCIT-ED group were significantly more likely to be classified as reporting more positive aspects of parenting after treatment when compared to baseline (OR= 7.5; p<.001). Caregivers in the waitlist group were significantly more likely to be classified as reporting fewer positive aspects of parenting at post-treatment when compared to baseline (OR= 3.84 p=.007).

Figure 2.

McNemar’s tests: Changes in Self-Reported Parenting from Baseline to Post-Treatment [N(%)]

The results also identified the number of caregivers who were observed to exhibit negative and positive parenting at baseline and those who were observed to exhibit the same parenting at post-treatment. McNemar’s test showed significant differences between observed parenting at baseline and post-treatment but only for the treatment group (Figure 2). Caregivers in the PCIT-ED group exhibited significantly more positive parenting at post-treatment when compared to baseline (OR= 2.67; p=0.02). Caregivers in the waitlist group did not exhibit any differences in positive parenting at post-treatment when compared to baseline. 24/36 (67%) caregivers in the PCIT-ED group exhibiting negative parenting at baseline exhibited positive parenting at post-treatment.

An additional McNemar’s test was conducted to determine if discrepancies between observed and self-reported parenting differed from baseline to post-treatment. A discrepancy would be indicated if caregivers self-reported overall positive parenting but were observed to exhibit overall negative parenting or vice versa. Significant discrepancies emerged for the waitlist group, such that less consistency between self-reported and observed parenting were evident (X²= 22.69; OR= 5.85; p<.001). The PCIT-ED group evidenced greater consistency between self-reported and observed parenting (X²= 1.09; OR= 1.5; p=0.29). 7/27 (26%) caregivers on the waitlist exhibited discrepancies between their observed negative and reported positive parenting at the post-treatment, whereas 13/67 (19%) caregivers in the PCIT-ED group exhibited discrepancies between their observed negative and reported positive parenting at the post-treatment.

Discussion

The purpose of this study was to examine the changes in self-reported and observed caregiver behaviors and affect following PCIT-ED treatment, while accounting for baseline caregiving practices and depressive symptoms among both children and caregivers. It was hypothesized that when compared to the waitlist group, caregivers in the PCIT-ED treatment group would report and display more positive parenting and less negative parenting at post-treatment. We also hypothesized increases in the positivity of PCIT-ED caregivers observed interactional style (e.g., overall warmth and a sense of connection between parent-child) when compared to caregivers in the waitlist group. These hypotheses were mostly supported with greater reductions seen in self-reported negative parenting and observed negative affect and increases in self-reported positive parenting and positive affect. Increases in the positivity of the observed interactional style of caregivers in the treatment group was also found. Of note, directly observed positive and negative parenting behaviors were not found to be significantly different between the treatment and waitlist groups at post-treatment, however the rates of these behaviors were low.

Consistent with previous research on PCIT (Thomas et al., 2017), self-reports of parenting behaviors significantly changed across treatment with caregivers reporting less dismissive, disapproving, and laissez-faire styles as well as less authoritarian and permissive parenting. In addition, caregivers in the treatment group reported increases in emotion-coaching and authoritative parenting at post-treatment. This finding is in line with other studies examining the impact of PCIT as this therapy directly targets parenting techniques, and meta-analyses demonstrate that parenting practices improve (Thomas et al., 2017). The novel emotion development module added in this study specifically focuses on teaching caregivers to validate and tolerate their child’s distressing emotions (Lenze et al., 2011), and our findings indicate that caregivers do, in fact, report significant changes in these dimensions of their behavior (Luby et al., 2019) as well as decreases in authoritarian and increases in authoritative strategies following treatment.

Of greater interest, our findings suggest that significant changes occurred in observed affect displays and interactional styles for parents receiving the PCIT-ED treatment versus those on the waitlist. Specifically, reductions in negative affect and increases in positive affect as well as increases in the positivity of the caregiver’s interactional style were observed by coders blind to treatment status. These increases in positive affect displays and positivity of the interactional style and decreases in negative affect displays held even after accounting for caregiver and child depression severity, child sex, and externalizing co-morbidity. The quality of the affective relationship between caregiver and child is arguably one of the most important across development. The emotional bond formed between the parent and child, the sense of connectedness, closeness, and mutuality sets the stage for many other relationships in the child’s life (Lamb & Lewis, 2005). These findings point toward one domain of improvement in this relationship following treatment that may have significant implications for the maintenance of treatment gains.

Although there were no significant findings regarding changes in observed caregiver behavior among those caregivers in the treatment group, a finding inconsistent with prior research, it likely that these aspects of parenting are still important within the parent-child relationship. This lack of findings may be a result of the specific coding scheme that was used, the nature of the structured interaction tasks and/or the infrequency of some of the caregiving behaviors that were coded and therefore, the lack of opportunity to observe such behaviors during the limited observational task. Interestingly, while caregivers reported changes in authoritative and authoritarian caregiving behaviors, none were directly observed. Therefore, as a secondary aim, we also investigated whether discrepancies between observed and reported parenting practices were reduced following treatment. Caregivers on the waitlist were more likely to exhibit discrepancies between their observed and reported parenting than caregivers in the treatment group. This increased coherence between self and observed parenting is intriguing and may suggest that following treatment, parents are more aware of their own behaviors and affect displays. This may also indicate that the treatment also brought about more accurate self-perceptions of parenting in addition to tangible observed changes in parenting. This is important as parents’ increased accuracy of their own parenting styles may be associated with a greater ability to modify their parenting as their child’s needs change with development. This may also lead them to be more sensitive and responsive to their individual child’s needs. Future work understanding how these different methods of assessing parenting cohere together is important, as relatively few studies report associations between methods.

The results of the present study should be considered with the following limitations in mind. First, this RCT used a waitlist as opposed to an active control treatment, and therefore, the results may be different if parents on the waitlist group were also receiving some sort of control intervention focused on parenting practices. Findings also need to be interpreted in light of the larger treatment study inclusion/exclusion criteria. Although sampling ensured equal distribution of specific characteristics across the two-groups, other criteria may have selectively excluded certain groups, as the final sample was relatively high in socio-economic status and primarily white. The use of global, as opposed to micro-codes may have influenced results since observers may have been swayed by overarching positive or negative impressions of the dyad or a few prominent events that occurred during the interaction (Gardner, 2000). Finally, we chose to focus on the behavior and affect of the parents only. Both child and dyadic effects may be equally important in predicting treatment outcomes and should be the focus of additional research. Despite these limitations, this study also makes a significant contribution to the literature on parenting changes following treatment for children’s depression. The observational assessment conducted in this study is likely to be more robust and possess superior validity when compared to self-reports of parenting, which are commonly used. The results of this study highlight the efficacy of PCIT-ED at improving parenting behaviors based on both parent’s self-report and their observed interactions with their child. Further, those who engaged in PCIT-ED showed improvements in the accuracy of their self-perception of parenting, a key skill for flexibly parenting to meet the changing needs of the child. PCIT-ED treatment improved aspects of parenting that were directly observed and crucial for a healthy, supportive parent-child relationship. It is hoped that future research can replicate and extend this work with additional focus on the aspects of parenting that may be the most relevant for improvement in children’s depressive symptoms.

Supplementary Material

Supp TableS1-3

Table S1. Baseline (pre-treatment) demographic and diagnostic characteristics of participants who completed a post-treatment assessment vs. participants who did not.

Table S2. Brief definitions of self-reported and observed parenting.

Table S3. Baseline (pre-treatment) demographic and diagnostic characteristics of participants.

Key points.

• Parenting behaviors and parent-child relationships exert substantial influence over young children’s emotional health and development

• After treatment, reductions in observed negative affect and increases in observed positive affect were found among caregivers. Higher coherence between self-reported and observed parenting behaviors and affect were evident among caregivers in the treatment group.

• Findings highlight the importance of targeting parenting behaviors in the treatment of child depression and using multiple methods to assess parenting styles in the context of treatment studies