Preliminary Evaluation of a Multi-Modal Early Intervention Program for Behaviorally Inhibited Preschoolers

Abstract

Objective

Approximately 15–20 percent of young children can be classified as having a behaviorally inhibited (BI) temperament. Stable BI predicts the development of later anxiety disorders (particularly social anxiety), but not all inhibited children develop anxiety. Parenting characterized by inappropriate warmth/sensitivity and/or intrusive control predicts the stability of BI and moderates risk for anxiety among high-BI children. For these reasons, we developed and examined the preliminary efficacy of the Turtle Program: a multi-modal early intervention for inhibited preschool-aged children.

Method

Forty inhibited children between the ages of 42–60 months and their parent(s) were randomized to either the Turtle Program (n = 18) or a waitlist control condition (WLC; n = 22). Participants randomized to the Turtle Program condition received 8 weeks of concurrent parent and child group treatment. Participants were assessed at baseline and post-treatment with multi-source assessments, including parent and teacher report measures of child anxiety, diagnostic interviews, and observations of parenting behavior.

Results

The Turtle Program resulted in significant beneficial effects relative to the WLC condition on maternal-reported anxiety symptoms of medium to large magnitude; large effects on parent-reported BI; medium to large effects on teacher-rated school anxiety symptoms; and medium effects on observed maternal positive affect/sensitivity.

Conclusions

This study provides encouraging preliminary support for the Turtle Program for young behaviorally inhibited children. Importantly, the effects of the Turtle Program generalized to the school setting. Future studies should examine whether this early intervention program improves long-term developmental outcomes for these at-risk children.

Fifteen to twenty percent of young children can be classified as behaviorally inhibited (BI) during infancy, and roughly half continue to display socially reticent behaviors throughout childhood (Degnan & Fox, 2007). Prospective studies demonstrate that stable BI across infancy and early childhood is associated with the development of later anxiety, particularly social anxiety disorder (SAD; Chronis-Tuscano et al., 2009) – suggesting a need for early intervention.

Parenting characterized by inappropriate warmth, a lack of responsive supportiveness and high levels of control, intrusiveness, and overprotection predicts both the stability of BI over time (e.g., Rubin, Burgess, & Hastings, 2002) and the development of later anxiety (McLeod, Wood, & Weisz, 2007). Moreover, maternal overcontrol moderates risk for anxiety, such that children with stable BI who also experience maternal intrusive control and a lack of responsiveness are at greatest risk for adolescent social anxiety (Lewis-Morrarty et al., 2012).

Within our theoretical model (Rubin et al., 2009), parents of children high in BI perceive them as vulnerable and thus respond to them in an unresponsive, unsupportive, and intrusive manner. Over time, these children become increasingly dependent on their parents and come to believe they are unequipped to deal with anxiety-provoking situations on their own. When parents respond to inhibited preschoolers with appropriate warmth and sensitivity, their children follow a healthier developmental trajectory (e.g., Hane, Cheah, Rubin, & Fox, 2008). Following from this transactional model, we developed an early intervention program for preschoolers displaying high BI and their parents. Unlike other programs for inhibited children (Rapee, Kennedy, Ingram, Edwards, & Sweeney, 2010), the “Turtle Program” targeted parent and child behavior.

The parent component was derived from adaptations of Parent-Child Interaction Therapy (PCIT), an evidence-based treatment for externalizing disorders in young children that utilizes in-vivo coaching to teach parents skills to enhance the parent-child relationship and implement effective discipline (Eyberg, Nelson, & Boggs, 2008). Adapted PCIT may also be an effective treatment for anxiety. Pincus and colleagues (2005) adapted PCIT for separation anxiety disorder (SAD) by including a “Bravery Directed Interaction” (BDI) module. A more recent adaptation, Coaching Approach Behavior & Leading by Modeling (CALM), was developed to treat 3–8-year-olds with anxiety disorders (Comer et al., 2012). These PCIT adaptations appear promising, based upon a preliminary RCT (Puliafico, Comer & Pincus, 2012) and open trial (PCIT-CALM; Comer et al., 2012), but have not been tried specifically with inhibited children.

We report results of an RCT examining preliminary effects of the “Turtle Program” compared to a waitlist control (WLC) condition. Although other groups have developed early intervention programs for young children with inhibition and/or anxiety disorders (Hirshfeld-Becker et al., 2010; Kennedy et al., 2009; Rapee et al., 2010), our study builds on previous studies by also incorporating in-vivo parent coaching within the peer context and reporting observational and teacher-report outcomes. Also, unlike other PCIT adaptations for anxious children, the Turtle Program allows for in-vivo parent coaching within the peer group context.¹

Method

Participants were recruited from local preschools, daycares, pediatricians, and media advertisements. Inclusion criteria included child age of 42–60 months and a Behavioral Inhibition Questionnaire (BIQ) score ≥ 132 (Bishop, Spence, & McDonald, 2003). Exclusion criteria included a Social Communication Questionnaire (SCQ; Berument, Rutter, Lord, Pickles, & Bailey, 1999) score > 15 to rule out autism². Prospective participants who met basic entry criteria on a telephone screen attended a clinic visit at which informed consent was obtained.

Interviewers uninformed of group membership administered the Preschool Age Psychiatric Assessment (PAPA; Egger, Ascher, & Angold, 1999) to parents³. Mothers completed the BIQ (α = .89 in this sample); Child Behavior Checklist (CBCL; Achenbach & Rescorla, 2000); and Preschool Anxiety Scale (PAS; Spence, Rapee, McDonald, & Ingram, 2001) total (α = .72) and social (α = .69) anxiety subscales. Teachers completed the School Anxiety Scale (SAS; Lyneham, Street, Abbott, & Rapee, 2008) total (α = .92) and social anxiety (α = .93) subscales.

An observational taxonomy was used to assess Positive Affect/Sensitivity (i.e., the parent’s ability to respond to the child’s verbal and nonverbal requests for attention or assistance with warmth and positive enjoyment of the child) and Negative Control (i.e., parental behavior that is ill-timed, excessive, or inappropriately controlling relative to what the child is doing) during free play and Lego model building (Rubin, Cheah, & Fox, 2001)⁴. Cohen’s kappas ranged from .80 – .83 for Positive Affect/Sensitivity and Negative Control.

Three cohorts, each consisting of 5–6 families, were randomly assigned to the Turtle Program (n = 18) or WLC (n = 22); see Table 2 for participant characteristics. The Turtle Program included 8 weekly, 90-minute concurrent parent and child group sessions, each led by two therapists⁵. Post-treatment assessments included the PAPA anxiety module, parenting observation, and parent and teacher questionnaires. WLC families received a 6-week parent psychoeducation program after post-treatment assessments were completed. There were no significant differences between families who did and did not complete post-treatment assessments (Figure 1).

Table 2.

Individual and Family Characteristics at Baseline Assessment

	Treatment (n = 18)	Waitlist (n = 22)
Child’s Age in Months M (SD)	50.81 (9.37)	54.27 (10.19)
Child’s Gender n (% Male)	9 (50)	8 (36)
Maternal Age M (SD)	35.56 (4.72)	38.47 (4.12)
Race n (%)
Caucasian	10 (55.6)	11 (50.0)
African-American	2 (11.1)	5 (22.7)
Other	5 (17.8)	2 (9.0)
Education n (%)
High School or Less	1 (5.6)	--
Some College	1 (5.6)	1 (4.5)
Bachelor’s Degree	3 (16.7)	7 (31.8)
Master’s Degree	9 (50.0)	5 (22.7)
Doctorate	4 (22.2)	5 (22.7)
Marital Status n (%)
Never Married	--	2 (9.1)
Married	17 (94.4)	16 (72.7)
Separated/Divorced/Widowed	1 (5.6)	--
Child Baseline Diagnoses n (%) a
Social Phobia	13 (72.2)	10 (45.5)
Any Anxiety Disorder	14 (77.8)	10 (45.5)
Selective Mutism	2 (11.1)	-
Specific Phobia	1 (5.5)	1 (4.5)
Separation Anxiety	3 (16.7)	1 (4.5)
Major Depressive Disorder	2 (11.1)	1 (4.5)
ADHD	1 (5.5)	-
Oppositional Defiant Disorder	1 (5.5)	-
Annual Family Income M (SD)	115,118 (56,227)	137,733 (67,832)

Note. M = Mean; SD = Standard Deviation; ADHD = Attention-Deficit/Hyperactivity Disorder.

^aCurrent = symptoms present in last 3 months.

Figure 1.

CONSORT Diagram.

The Turtle Program: Helping Shy Preschoolers Come Out of their Shells

Parent Component

The parent group manual was modeled after PCIT for SAD with two primary exceptions: (1) exposures occurred within the clinic so that parents could be coached in the moment; and (2) group (rather than individual) format. Children were pulled from the child group for dyadic parent-child coaching, while other parents observed for the purpose of vicarious learning. Session content is presented in Table 1.

Table 1.

Turtle Program Session Content

Session	Parent Group	Child Group	Coaching7
1	Psychoeducation	Learning to introduce yourself	Coaching during separation and pick-up
2	CDI Teach	Making eye contact Relaxation (Balloon Breathing)	Coaching during separation and pick-up
3	CDI Coach	Communicating to Keep Friends	Individual CDI coaching
4	BDI Teach	Facing your Fears (Lizzy the Lamb book)a	Coaching during separation and pick-up
5	BDI Coach 1	Expressing Emotions Group activity: Sharing about oneself game during snacktime	Individual BDI coaching: Bravery challenge
6	BDI Coach 2	Dealing with disappointment Group activity: Show & Tell	Individual BDI coaching: Show and tell
7	PDI Teach	Working together Group activity: Scavenger Hunt	Coaching during separation and pick-up
8	PDI Check-in and Wrap-up	Group activity: Graduation Party	Coaching during graduation party

Note. CDI = Child Directed Interaction; BDI = Bravery Directed Interaction; PDI = Parent Directed Interaction.

^aScheffler, A. (2011). Lizzy the lamb. London: Pan Macmillian.

Session 1 (psychoeducation) was modeled after Being Brave: A Program for Coping with Anxiety for Young Children and Their Parents (Hirshfeld-Becker et al., 2010) and described our theoretical model. During Child Directed Interaction (CDI), parents were coached to stay “a step behind” (as opposed to adopting a controlling or intrusive style) and provide labeled praise for child behaviors such as independence, sharing his/her own ideas, and appropriate social behaviors. During BDI, parents learned to apply CDI skills in the context of anxiety-provoking situations (e.g., asking another child to play, participating in “Show and Tell”) and were coached to remove attention for avoidant or clingy behaviors while praising social approach behaviors. Parents were also instructed in planning and implementing out-of-session exposures. During Parent Directed Interaction, parents learned to distinguish anxious and oppositional behaviors, and to implement discipline strategies for the latter. In the final session, a “graduation party” was held which also served as an exposure task in which parents were coached to use their acquired skills.

Child Component

The child group was adapted from SSFP (Coplan, Schneider, Matheson, & Graham, 2010), which employs social skills training in a manner that is developmentally appropriate for preschool children. The didactic portion was brief, with sensitivity to the attention span of young children, and incorporated puppets and games as age-appropriate ways of conveying content. After the didactic portion, children engaged in free play and group activities, during which SSFP leaders utilized systematic modeling, guided participation, and reinforcement of relevant social skills, as well as facilitation of social interaction and social problem solving. Activities were incorporated to allow for exposure to feared social situations (Table 1). Children were praised for approach behaviors and discretely encouraged to use SSFP skills as appropriate.

Analytic Plan

Generalized estimating equations (GEE; Hardin & Hilbe, 2003) was used to examine treatment effects on outcomes. GEE is an extension of the generalized linear model that allows for the analysis of correlated observations in repeated measures designs. Users can specify model distributions, the structure of correlated data over time using working correlation matrices, and conduct full factorial models using categorical and continuous predictors. GEE is robust to small sample sizes and missing data. In this study, unstructured correlation matrices were selected for each parameter based on the lowest Quasi Likelihood Under Independence Model Criterion (QIC) value and a-priori hypotheses. To examine change in anxiety diagnoses, binomial distributions and independent or exchangeable correlation matrices were specified. Gamma distributions with logit link functions were specified for observed parenting and other non-normally distributed outcome variables, and identity link functions were used with normally distributed outcome variables. For each outcome, main effects of time, treatment group, and the interaction effect of time x treatment group were estimated. To control family-wise error, estimated marginal means were calculated using simple and pairwise comparisons for main and interaction effects. Hedge’s g (1981) was calculated as an indicator of the magnitude of time x treatment group effects⁶.

Results

Baseline group differences were examined in preliminary analyses. Seventy-seven percent of children in the treatment group met DSM-IV criteria for an anxiety disorder relative to 47.6% of children in the WLC (Table 2); however, this difference was not significant, χ² (1)= 3.73, p = .054. Similarly, the two groups did not differ on baseline BIQ scores (t(35) = −1.012, p = 0.319) or PAPA total anxiety symptoms (t(37) = .0682, p =0.135). No other significant baseline differences between groups were observed (Table 2). Descriptive statistics for all outcome measures are presented in Table 3.

Table 3.

Descriptive Statistics

Measure	Statistic	Treatment (n = 18)			Waitlist (n = 22)
		Baseline	Post-treatment	Baseline	Post-treatment
Diagnostic Interview		(n = 18)	(n = 17)	(n = 22)	(n = 17)
PAPA
SAD Symptoms	M (SD)	2.61 (0.78)	1.0 (1.37)	2.1 (1.1)	1.64 (1.45)
SAD Diagnosis	n (%)	13 (72.2)	5 (27.8)	10 (45.5)	7 (31.8)
Total Anxiety Symptoms	M (SD)	32.28	14.76 (10.12)	26.48 (12.81)	22.29 (14.07)
Any Anxiety Diagnosis	n (%)	14 (77.8)	7 (38.9)	10 (45.5)	7 (31.8)
Parent-Report		(n = 18)	(n = 17)	(n = 22)	(n = 17)
BIQ	M (SD)	169.50(3.71)	144.79 (4.14)	163.46 (3.81)	159.76 (6.30)
CBCL Internalizing	M (SD)	60.83 (1.19)	51.02 (2.19)	58.71 (2.10)	58.05 (2.25)
Preschool Anxiety Scale	M (SD)	63.94 (2.31)	45.00 (4.04)	60.64 (3.91)	52.85 (4.49)
Social Anxiety	M (SD)	19.84 (1.07)	14.30 (1.33)	17.52 (1.16)	17.20 (1.26)
Teacher-Report		(n =15)	(n =16)	(n = 21)	(n = 18)
School Anxiety Scale	M (SD)	12.91 (1.95)	10.97 (1.88)	14.09 (2.11)	17.67 (2.80)
Social Anxiety Subscale	M (SD)	8.57 (1.50)	8.16 (1.45)	8.80 (1.21)	9.59 (1.43)
Generalized Anxiety		4.34 (0.59)	2.81 (0.55)	5.06 (1.17)	7.90 (1.69)
Observed Parenting		(n = 18)	(n = 17)	(n = 22)	(n = 17)
Positive Affect/Sensitivity-Free play	M (SD)	1.45 (0.18)	1.49 (0.22)	1.45 (0.16)	1.35 (0.17)
Positive Affect/Sensitivity – Lego	M (SD)	1.46 (0.29)	1.53 (0.3)	1.51 (0.24)	1.43 (0.29)
Negative Control- Free play	M (SD)	1.02 (0.03)	1.01 (0.01)	1.03 (0.05)	1.02 (0.04)
Negative Control- Lego	M (SD)	1.00 (0.13)	0.96 (0.18)	1.03 (0.13)	1.01 (0.14)

Note. M = Mean; SD = Standard Deviation; PAPA = Preschool Age Psychiatric Assessment; SAD= Social Anxiety Disorder; BIQ = Behavioral Inhibition Questionnaire; CBCL = Child Behavior Checklist.

On the PAPA, significant Time x Group interactions of large magnitude were found favoring the Turtle Program on total anxiety symptoms (Table 4). Treatment effects on social anxiety symptoms and diagnoses were marginally significant and of medium magnitude (.55 and .62, respectively). Significant Time x Group interactions also were found on the BIQ, CBCL Internalizing Problems, and PAS social anxiety subscale, such that children in the treatment group demonstrated greater improvements from pre-treatment to post-treatment in parent-rated BI, internalizing problems, and social anxiety symptoms relative to WLC children; these effects were large in magnitude, ranging from .84–1.06 (Table 4). Teachers similarly reported significant pre-post treatment reductions in total and generalized anxiety of medium to large magnitude for Turtle Program, relative to WLC, participants. A significant Time x Group interaction on maternal positive affect/sensitivity during free play of medium magnitude was found, also favoring the treatment group. No treatment effects on maternal negative control were observed.

Table 4.

Results of GEE Analyses

	B	SE	OR	95% CI	95% CI	P	Ga
Diagnostic Interview
SAD Symptoms	−0.71	0.41		1.52	0.09	0.081	0.55
SAD Diagnosis		0.97	1.91	0.003	3.81	0.05	0.62
Any Anxiety Symptoms	−0.59	0.21		0.99	0.18	0.005	0.88
Any Anxiety Diagnosis		0.96	1.68	0.21	3.57	0.081	0.55
Parent-Report
BIQ	−21.019	7.14		−35.01	−7.03	0.003	0.93
CBCL Internalizing	−9.15	2.72		−14.71	−3.82	0.001	1.06
Preschool Anxiety Scale	−11.15	6.21		−23.32	1.01	0.072	0.57
Social Anxiety Subscale	−5.22	1.97		−9.09	−1.35	0.008	0.84
Teacher-Report
School Anxiety Scale	−5.52	2.76		−10.92	−0.11	0.045	0.63
Social Subscale	−1.204	1.59		−4.33	1.92	0.45	0.24
Generalized Subscale	−4.37	1.52		−7.35	−1.38	0.004	0.91
Observed Parenting
Positive Affect/Sensitivity
Free play	0.14	0.06		0.018	0.25	0.024	0.73
Lego	0.09	0.07		0.047	0.23	0.196	0.40
Negative Control
Free play	−0.01	0.016		0.04	0.02	0.447	0.23
Lego	−0.03	0.07		0.16	0.1	0.66	0.13

Note. SAD = Social Anxiety Disorder; BIQ = Behavioral Inhibition Questionnaire; CBCL = Child Behavior Checklist; B = Beta; SE = Standard Error; OR = Odds Ratio; CI = Confidence Interval; G = Hedges’ g.

^aValues of 0.2 correspond to small effects, values of 0.5 correspond to medium effects, and values of 0.8 correspond to large effects.

Discussion

This small RCT was, to our knowledge, the first study of an early intervention program targeting children on the basis of early BI which included concurrent parent and child groups and which also measured treatment effects across parent ratings, teacher ratings, and observed parenting. Our novel, theoretically- and developmentally-grounded treatment approach, adapted from PCIT and SSFP, involved in-vivo coaching of parents in the use of behavioral strategies as their inhibited children participated in a peer context. Results suggest that the Turtle Program holds great potential to improve child anxiety symptoms across home and school settings, and perhaps more importantly, to increase observed maternal positive affect and sensitivity.

The treatment group demonstrated significant pre-post treatment improvement relative to the WLC on maternal-reported anxiety symptoms and diagnoses, of medium to large magnitude. Consistent with other studies finding effects on parent-reported (Kennedy et al., 2009) and laboratory-observed BI symptoms (Kennedy et al., 2009; Hirshfeld-Becker et al., 2010), we found large effects of treatment on parent-reported BI. This is particularly promising given that baseline BI negatively predicted treatment response to Hirshfeld-Becker’s (2010) Being Brave intervention for 4–7 year olds with anxiety disorders. Thus, we selected a very challenging population and yet demonstrated change on both inhibition and anxiety.

As an index of generalization, teachers reported improvements in school anxiety symptoms for children in the treatment group relative to the WLC. This was encouraging given that no intervention took place in the school and teachers were uninvolved in treatment. Given that researchers who have completed intervention studies of young children with BI or anxiety disorders have rarely, if ever, collected teacher questionnaires, it has been unclear whether these earlier interventions decreased anxious behavior in the school setting. Given the social nature of BI and demonstrated associations with later social anxiety, it is critical that effects of early interventions designed for inhibited young children are demonstrated at school.

We were also encouraged by the significant treatment effects on observed maternal positive affect/sensitivity. Given that maternal warmth and sensitivity both predict the discontinuity of BI over time and protect against future maladjustment in at-risk children (Coplan et al., 2008; Degnan & Fox, 2007; Hane et al., 2008), this finding was particularly noteworthy. Observational measures are often considered the gold-standard in treatment research, given the potential for parent reports to be biased and influenced by expectations that treatment will work. This may be particularly the case when parents are involved in treatment and the comparison condition does not receive active treatment, as was the case in our study and other studies of young children with high BI and/or anxiety (Hirshfeld-Becker et al., 2010; Rapee et al., 2010). We were surprised by the lack of treatment effects on negative/intrusive control. In contrast with studies showing that parents of inhibited and/or anxious children tend to engage in negative control (Hudson & Rapee, 2000), few mothers in this study demonstrated negative control at baseline. Characteristics of our sample or observational context may have contributed to this finding.

This study was limited by a small sample size and WLC (rather than active comparison) condition. We also did not observe child social behavior in the laboratory for both conditions at pre- and post-treatment. Finally, the BIQ was used both as a selection and outcome variable. Despite these limitations, the findings were encouraging in that the Turtle Program demonstrated effects on parent and teacher ratings of anxiety as well as observed maternal positive affect/sensitivity, supporting the need for further evaluation. Future studies should include larger samples that are block randomized based on the presence of baseline anxiety disorders. Future studies should also include observations of child behavior and follow-up assessments to better characterize the effects of early intervention on the trajectory of anxiety in this at-risk group. A larger, socioeconomically diverse sample and multiple assessment points during the course of treatment will allow for an examination of mediators (e.g., parenting, social skills) and moderators (e.g., baseline anxiety severity, culture, physiological reactivity, socioeconomic status) of treatment effects. Future studies may also examine single and additive effects of various treatment components (e.g., in-vivo coaching, parent- or child-only group), given that studies have reported effects of a parent-only intervention on diagnostic outcomes (Rapee et al., 2010) and no differences between parent-child and parent-only interventions (Waters et al., 2009).