Trajectories of Dysregulation in Children with Autism Spectrum Disorder

Abstract

Objective:

This study determined whether child and family environment factors are associated with differences in developmental trajectories of emotional and behavioral dysregulation in children with autism spectrum disorder (ASD).

Method:

Participants included 186 families of a child with ASD (5-12 years old at baseline; 86% male; 83% non-Hispanic Caucasian; 35% comorbid intellectual disability). At each of the four time points (each spaced 12 months apart), mothers and fathers within each family completed well-validated measures on their own mental health, their child’s dysregulation, their parent-child relationship, and their parent couple relationship. Longitudinal multi-level modeling was used to describe trajectories of dysregulation across three years and test whether parent depression, closeness in the parent-child relationship, and positive parent dyadic coping were associated with differences in child trajectories.

Results:

On average, child dysregulation decreased across time. Closer mother-child and father-child relationship quality was associated with lower baseline dysregulation. More severe child restricted and repetitive behaviors, fewer maternal depression symptoms, and more positive parent dyadic coping were associated with declines in child dysregulation over time.

Conclusions:

On average, children with ASD become less dysregulated across time. However, there is important variability in dysregulation trajectories of children with ASD. Children with ASD who have a high (versus low) severity of restricted and repetitive behaviors appear to be at risk for greater dysregulation. The family environment, and specifically a closer parent-child relationship, better maternal mental health, and more positive couple coping, may contribute to a pattern of improved child regulation across time in ASD.

Dysregulation in childhood, broadly defined as patterns of emotional, behavioral, and cognitive responses that interfere with appropriate goal directed behavior, is considered to be a transdiagnostic risk factor for mental health problems (Beauchaine, 2015; Beauchaine & Cicchetti, 2019; Mick et al., 2011), and is associated with increased risk for engagement in risky behaviors, aggression, non-suicidal self-injury, substance use, depression, and anxiety into adolescence and adulthood (Robson, Allen, & Howard, 2020). Although not a core diagnostic feature of autism spectrum disorder (ASD), dysregulation is common in children with ASD and is a primary reason parents seek treatment (Mazefsky et al., 2013; Samson, Wells, et al., 2015). Understanding how dysregulation changes across childhood and factors associated with differences in level and trajectory of dysregulation among children with ASD is critical for prevention efforts. The current study examined initial level and change in parent-reported emotional and behavioral dysregulation in a sample of 186 children with ASD across three years and tested for individual- and family-level factors that altered the course of these trajectories.

ASD is a heterogeneous neurodevelopmental disorder characterized by impairments in social-communication and restricted and repetitive behaviors that emerges early in life and persists across the lifespan, affecting 1 in 54 children in the U.S. (American Psychiatric Association, 2013; Maenner et al., 2020). Dysregulation, including challenges in the modulation of affect, behavior, and cognition (Mick et al., 2011), is a common experience for children with ASD (Conner, Golt, et al., 2021; Patel, Day, Jones, & Mazefsky, 2017; Samson et al., 2014; Uljarević et al., 2018). As early as infancy, parents of children diagnosed with ASD report that their children exhibit more negative and less positive affect, have more difficulties soothing, and retrospectively report more self-regulation difficulties at one year of age compared to children without ASD (Gomez & Baird, 2005; Zwaigenbaum et al., 2015). Impairments in self-regulation have been linked to increases in emotional and behavioral challenges, including anxiety and mood problems, and declines in social skills in children with ASD (Aldao et al., 2016; Berkovits et al., 2017; Conner, White, et al., 2020; Li et al. 2020). While it remains uncertain whether dysregulation is an inherit component of ASD itself (Gomez & Baird, 2005) or is better conceptualized as an indicator of mental health comorbidities, (Mazefsky & White, 2014; White et al., 2014), it is clear that dysregulation is a prevalent and clinically significant problem for many children with ASD and thus an important treatment target.

A substantial body of work has focused on examining the developmental course of self-regulation in neurotypical children (e.g., Robson, Allen, & Howard, 2020). On average, self-regulation improves as children age; however, research suggests that there is also important variability in the trajectories of (dys)regulation across childhood (McQuillan et al., 2018). This research has informed prevention and intervention efforts aimed at promoting optimal development and reducing risk for later mental health challenges (Smith et al., 2019; Stormshak et al., 2018). To-date, research on dysregulation in children with ASD has largely been cross-sectional and focused on group-level (i.e., mean) differences in ASD relative to neurotypical comparison groups (e.g., Conner, Golt, et al., 2021; Samson et al., 2014). Virtually nothing is known about the developmental course of dysregulation across childhood or factors associated with within-group variability in level and/or trajectory of dysregulation across time in ASD. Yet, this information has important implications for understanding when and for whom prevention and intervention efforts are needed.

Evidence from cross-sectional studies suggests that the heterogeneity in ASD itself may be associated with differences in dysregulation in children with ASD. For example, dysregulation has been associated with more severe ASD symptoms including level of social communication difficulties, restricted behaviors/repetitive interests (RRBs), and sensory abnormalities (Samson et al., 2014), as well as broader markers of social functioning (Jahromi et al., 2013; Nehaus, Webb, & Bernier, 2019). However, research to-date has yet to examine whether ASD symptom severity is associated with varying trajectories of dysregulation across middle childhood and into early adolescence in ASD.

The family environment may also be a critical predictor of dysregulation in children with ASD. Indeed, self-regulation has been posited to be an experience-based, dynamic process that emerges within and in response to a specific environment (Thompson, 2019). Family is arguably the most central and enduring environmental context in childhood, and thus, is thought to strongly shape emotional development (Morris et al., 2007). Characterized by a variety of intrafamilial processes including the parent-child relationship, the parent marital or couple relationship, and parent mental health, the emotional climate of the family offers children a context to observe and understand others’ emotional responses in the face of challenging or stressful situations. Children in warm, sensitive, responsive, and consistent family environments are more likely to have their emotional needs met, feel more secure and supported, and are able to express their emotions without repercussion (Morris et al., 2007). When the family environment is negative, conflictual, or unpredictable, children may observe dysregulated emotional and behavioral responses in their parents, witness or participate in dysregulated displays themselves, and are less emotionally secure within the family system (Morris et al., 2007) and this may undermine the development of children’s self-regulation (Thompson, 2019).

As is true for all children (regardless of developmental status), the family environment and child self-regulation abilities are posited to interact in transactional ways across time (Morris et al., 2007). In this way, the family environment may be a modifiable set of intervention targets that can be used to promote optimal child development in the context of ASD. Research has shown that raising a child with ASD can be physically, emotionally, and financially challenging (Karst & Hecke, 2012; Sim et al., 2018). Families of children with ASD are at risk for a negative family environment, reflected by poorer parent psychological well-being (e.g., higher depressive symptoms and stress), more parent marital dysfunction (e.g., more conflict and less effective dyadic coping) and more distant and critical parent-child relationships relative to families of neurotypical children (Karst & Hecke, 2012). A substantial body of research has explored how these aspects of the family environment are influenced by parenting challenges including the child with ASD’s emotional and behavioral regulation problems (e.g., Costa, Steffgen, & Ferring, 2017; Davis & Carter, 2008; Karst & Hecke, 2012; Tervo, 2012).

While research suggests that child dysregulation contributes to a negative family emotional climate, there is growing empirical evidence that children with ASD respond to their family environment in addition to acting upon it (e.g., Baker et al., 2011). For example, a more negative family emotional climate has been associated with less warmth and more criticism within the parent-child relationship (Hickey, Hartley, & Papp, 2020), higher child internalizing symptoms (Rodriguez, Hartley, & Bolt, 2019), and concurrent child emotional and behavioral challenges in families of a child with ASD (Yorke et al., 2018). Given that (1) children with ASD, on average, experience a more negative family emotional climate, including increased parent depression, more negative parent-child relationship, and greater parent marital discord, than do their peers without disabilities (Bitsika & Sharpley, 2004; Gau et al., 2012; Hayes & Watson, 2013), and (2) a negative family emotional climate is associated with poor child outcomes, it is important to delineate whether the family emotional climate is associated with trajectories of dysregulation during childhood in ASD.

Current Study

The present study provides the first exploration of dysregulation trajectories in children with ASD across middle childhood and early adolescence, and the individual and family environment factors associated with these trajectories. The study aims were to: 1) evaluate the average trajectory of dysregulation in children with ASD and identify between-person variability in this trajectory; and 2) test whether individual- and family-level factors are associated with changes in the trajectory of dysregulation in children with ASD. Previous research highlights a link between ASD symptom severity and dysregulation (Cibralic et al., 2019); thus, we hypothesized that children with ASD who have less (vs. more) severe social-communication impairments and restricted behaviors/repetitive interests RRBs would show declines in dysregulation over time. Guided by models of the impact of the family context on self-regulation in non-ASD populations (e.g., Morris et al., 2007), we also hypothesized that the emotional climate of the family would be associated with changes in dysregulation trajectories such that children in less stressed family environments would show a steeper decline in dysregulation over time. We chose three indicators of the emotional climate of the family that correspond to the intrafamilial processes identified in the theoretical model proposed by Morris and colleagues (2007): parental couple relationship (dyadic coping), parent-child relationship (closeness in the parent-child relationship), and parent mental health (depression symptoms). In addition, previous research has shown these intrafamily processes to be associated with child self-regulation in both ASD samples (e.g., Hickey, Hartley, & Papp, 2019; Yorke et al., 2018) and neurotypical children (e.g., Bridgett et al., 2015).

Method

Participants

The current research draws on data from a longitudinal study (four time points across three years –T1, T2, T3, and T4 – each spaced 12 months apart) of 188 children with ASD (5-12 years at baseline). To be included in the study, families had to include a child aged 5 to 12 years with a documented diagnosis of ASD from a medical or education professional that included receiving a score above the Autism Spectrum cutoff (69% reported scores above the Autism cutoff) on the Autism Diagnostic Observation Schedule (ADOS-2; Lord et al., 2012). Most children were assessed using Module 2 of the ADOS (53%) and the average age of initial diagnosis was 48.17 months (range: 18-140 months). In addition, parents completed the Social Responsiveness Scale (SRS-2; Constantino & Gruber, 2012) to verify the child’s current ASD symptoms. Five children received an SRS-2 Total t-score ≤ 60; however, after reviewing all available information (i.e., medical and educational records, ADOS-2 scores, available teacher report of SRS-2), these families were included in the study as the child was deemed to meet criteria for ASD. In cases where multiple children had been diagnosed with ASD, the oldest child with ASD was selected as the target child. In line with the original study aims, families also had to include cohabitating, heterosexual couples in a relationship for at least three years at T1. Couples did not have to be married nor did they have to the biological parent of the child to participate. Family, parent, and child sociodemographic information can be found in Table 1.

Table 1.

Family, parent, and child socio-demographic characteristics

Demographic Items
Family
Annual Household Income, n (%)
$19,999 or less	2(1.1)
$20,001 – 39,999	13(7.4)
$40,000 – 59,999	27(14.9)
$60,000 – 79,999	33(18.2)
$80,000 – 99,999	34(18.8)
$100,001 and greater	72(39.8)
Missing (n=8,4.2%)
Additional child with SHCN (yes), n (%)	67(35.6)
Length of relationship, M years (SD, range)	14.55(5.59, 1-31)
Child
Age, M years (SD,range)	7.92 (2.25, 5-12)
Sex (male), n (%)	160 (85.6)
ID status (yes), n (%)	65(34.8)
Parent	Mother	Father
Age, M years (SD, range)	38.73(5.61, 24-54)	40.76(6.19, 22-60)
Relationship to child, n (%)
Biological parent	167(88.4)	167(88.4)
Step-parent	5(2.6)	10(5.3)
Adoptive parent	5(2.6)	5(2.6)
Previously married (yes), n (%)	25(13.2)	25(13.2)
Race/Ethnicity, n (%)
White, non-Hispanic	158(84)	154(82)
Other	30(16)	34(18)
Education, n (%)
Less than high school	0(0.0)	4(2.1)
Some high school	2(1.1)	4(2.1)
High school degree or GED	9(4.8)	22(11.8)
Associates or technical school	19(10.3)	18(9.6)
Some college	30(16.3)	29(15.5)
College degree	68(37.0)	62(33.2)
Some graduate school	10(5.4)	11(5.9)
Master’s degree	32(17.4)	29(15.5)
Doctoral or Law degree	6(3.3)	5(2.7)

Note: SHCN = special healthcare needs; M = mean; SD = standard deviation

Procedure

The longitudinal study was approved by the Institutional Review Board at a university in the Midwestern U.S. Participants were recruited from local schools, ASD clinics, community centers, and ASD research registries. All participants provided informed consent. At each time point, approximately one year apart, families completed a 2-hour in-home or laboratory visit that included questionnaires about the parent couple relationship, parent mental health, and the child’s emotional and behavioral functioning. Parents each received $50 at each time point for this portion of the study.

Baseline (T1) Measures

Family, Parent, and Child Socio-demographics.

Each parent reported on socio-demographics at Time 1. Parents jointly reported on the child with ASD’s age (coded in years) and biological sex (female =1, male = 2). Children with ASD were considered to have intellectual disability [ID status; coded no (0) and yes (1)] if they had a medical diagnosis of intellectual disability and/or met criteria based on review of medical and/or educational records reporting IQ and adaptive behavior testing. Parents reported on household income starting at ≤$9,999 (1) and increasing by $10,000 to $20,000 intervals to ≥$160,000 (14).

ASD Symptom Severity.

Parents independently completed the Social Responsiveness Scale-2 (SRS-2; Constantino & Gruber, 2012), a 65-item measure assessing their child’s level of social-communication impairment and severity of restricted and repetitive behaviors. Items are rated on a scale of 1 (‘not true’) to 4 (‘almost always true’) regarding their child’s behavior in the past six months. Items were summed to create two subscale scores: Social Communication Index (SCI; a measure of the severity of social-communication impairment) and RRBs (a measure of the severity of restricted interests and repetitive behaviors associated with ASD). Higher scores reflect more impairment. Composite variables were created by averaging mother (Cronbach’s α SCI = .93; RRB = .83) and father (α SCI = .92; RRB = .80) subscale scores to create a composite SCI and composite RRB score used in present analysis given the strong correlation between the two reports (r SCI = .55; RRB= .53).

Parent-Child Relationship.

Each parent independently rated the positive affect of their relationship with the target child using the Positive Affect Index (PAI; Bengtson & Allen, 1993). Parents responded to ten questions regarding the understanding, trust, respect, fairness, and affection between the parent and child on a 6-point scale (1 ‘not at all’ to 6 ‘extremely’). The PAI has been used reliably as a measure of positive affect in the parent-child relationship in an ASD context (e.g., Taylor & Seltzer, 2011). Mother (Cronbach’s α = .85) and father (Cronbach’s α = .86) total scores were both included in the present analysis as unique scores.

Positive Dyadic Coping.

The Dyadic Coping Inventory (DCI; Bodenmann, 2008) is a 37-item self-report measure of the coping behaviors and communication patterns of couples. Mothers and fathers separately rated (1 ‘very rarely’ to 5 ‘very often’) statements related to supportive coping by oneself (e.g., “I show empathy and understanding to my partner”), supportive coping by one’s partner (e.g., “My partner expresses that he/she is on my side”), delegated coping by oneself (e.g., “I take on things that my partner would normally do in order to help him/her out”), delegated coping by one’s partner (e.g., “When I am too busy, my partner helps me out”), and common coping (e.g., “We try to cope with the problem together and search for ascertained solutions”) to create an overall Positive Dyadic Coping score for each parent (mother Cronbach’s α = .91, father Cronbach’s α = .89) that were both included in the current analysis as independent scores. The DCI has shown acceptable psychometric properties in previous research of couples caring for a child with ASD (Sim et al., 2017).

Parent Depression Symptoms.

Parents rated their depression symptoms using the 20-item Center for Epidemiologic Studies Depression (CESD-R; Eaton et al. 2004) Scale. The CESD-R is a widely used measure and has been shown to be a valid measure of depression symptoms in the general population (Van Dam & Earleywine, 2011). A total score is created by summing all responses for a potential overall symptom score ranging from 0-60, with higher scores indicating the presence of more depression symptomatology (scores > 16 indicating significant risk for clinical depression). Mother (Cronbach’s α = .92) and father (Cronbach’s α = .89) depression scores were both included separately in the current study.

Repeated Measures

Child Dysregulation.

A dysregulation composite score (average of mother and father report) was created at each time point using the Child Behavior Checklist Dysregulation Profile (CBCL-DP; Althoff et al., 2010) by summing the t-scores on the anxiety/depression, aggression, and attention subscales (Mother Cronbach’s α range = .90-.91, Father Cronbach’s α range = .89-.91; correlation amongst mother and father CBCL-DP scores across time: r = .49-.56, all ps < .001). Scores below 180 are considered to fall into non-clinical ranges, scores between 180 and 210 [between 1 (t-score = 60) and 2 (t-score = 70) SDs] are considered to be “borderline” clinical, also known as deficient emotional self-regulation, and scores above 210 are considered to be clinically significant dysregulated emotion.

The CBCL-DP is a widely used measure to assess symptoms of child emotional and behavioral challenges over the past six months. Studies in both neurotypical children as well as clinical populations (e.g., ADHD) have shown elevated CBCL-DP score to be associated with more severe psychopathology (e.g., Holtmann et al., 2011; Peyre et al., 2015). In addition, studies on the psychometric properties of the CBCL-DP suggests the presence of a broad dysregulation factor that is correlated with but independent from the specific underlying subdomains that compromise the CBCL-DP (i.e., anxiety/depression, aggression, attention; Deutz et al., 2016; Geeraerts et al., 2015; Keefer et al., 2020).

Analytic Approach

We used longitudinal multilevel modeling (MLM), defining children as the Level 2 unit of analysis and repeated measures (time) nested within children as the Level 1 unit, to examine trajectories of dysregulation in children with ASD and to test the effects of individual and family factors on dysregulation across the four time points that spanned three years. This approach allows for modeling of both fixed and random effects, to enable us to identify variability in change in dysregulation over time. Analyses were conducted in SPSS 26 (IBM, 2019) and R software (lmerTest package; R Core Team, 2013) using maximum likelihood estimation. The current study used a model building process, testing four models aimed at analyzing between-person, within-person, and cross-level effects on children’s dysregulation of ASD symptom severity, parent-child relationship quality, parental depressive symptoms, and parent couple dyadic coping, and the interactions of both individual- and family-level variables with time. First, two unconditional models were analyzed to evaluate the relative magnitude of within-person and between-person variance in the CBCL-DP scores and to determine the average trajectory of dysregulation across the sample. Second, two conditional models were analyzed to examine predictors of initial levels and change in dysregulation.

Unconditional Models.

First, the unconditional (null) model in which no predictors were included at either level 1 or level 2 was tested to estimate sources of variance in CBCL-DP scores at the occasion level (Level 1; within participants) and the participant level (Level 2; between participants). This model acts as a baseline model to test whether the addition of fixed effects at level-1 and level-2 improves the fit of the model. Model 2 (unconditional growth model) estimated the trajectory of dysregulation (aim 1 fixed effects) using the time of measure (units in years) as a predictor and captures the shape of the within-individual growth trends by predicting dysregulation (Y_ti) as a function of time (Time_ij):

$$\mathrm{Level}1:Y_{ti}={\pi }_{0i}+{\pi }_{1i}({Time}_{ij})+e_{ti}$$

We first tested a linear, then a quadratic model of individual-level change in dysregulation. The linear growth parameters are represented by the intercept (π_0i) and the slope (π_1i) and includes residuals (e_ti) that represent deviations of the observed data from the predicted trajectory. The time variable was the visit number minus 1 (i.e., 0, 1, 2, 3) that set the baseline visit as the reference category and permitted interpretation of the intercept as the baseline CBCL-DP score. The linear model was run twice, first including a random intercept (i.e., allowed CBCL-DP scores to vary across children at the first time point) and a fixed slope (i.e., forced CBCL-DP to change at the same rate between children). This model was then compared to a random intercept and random slope model (i.e., CBCL-DP scores were allowed to vary at both the intercept and in the rate of change between children) using the chi-square difference test, which was significant [χ²(2) = 17.68, p <.001], suggesting the use of random linear slopes.

We then fit two quadratic growth models with fixed and random slopes by adding π_2i(Time_ij²) to the Level 1 equation above. There was a statistically significant quadratic effect of time on dysregulation scores (γ₂₀ = 1.049, ρ = .002); however, the chi-square difference test indicated that allowing the quadratic slopes to vary did not significantly improve the model compared to a fixed quadratic slope, χ²(2) = 1.594, p = .451. We then examined the fit of the random linear slope and the fixed quadratic slope models to determine which model best fit the data. The random linear slope model had the lowest AIC (5824.2) and BIC (5851.9) values and was retained for all remaining analyses. Model 2 was also used to identify variability between individuals in the intercept and the slope coefficients (aim 1 random effects), therefore justifying the addition of other variables to predict individual differences in intercepts and/or slopes.

Conditional Models.

All predictors were standardized (i.e., mean centered and divided by the sample standard deviation; separately for mothers and fathers when applicable) prior to analysis to aid in the interpretation of results and avoid multicollinearity with interaction terms. The general Level-2 equations below model interindividual differences in growth parameters:

$$\begin{array}{l}\mathrm{Level}2:{\pi }_{0i}={\gamma }_{00}+{\gamma }_{01}(SCI)+{\gamma }_{02}(RRB)+{\gamma }_{03}(PCrelmom)\dots {\gamma }_{08}(Depdad)+{\varsigma }_{0i}\\ {\pi }_{1i}={\gamma }_{10}++{\gamma }_{11}(SCI)+{\gamma }_{12}(RRB)+{\gamma }_{13}(PCrelmom)\dots {\gamma }_{18}(Depdad)+{\varsigma }_{1i}\end{array}$$

Coefficients in Models 3 and 4, therefore, represent differences in the outcome (CBCL-DP or dysregulation) per unit of change in the predictor in standard deviation (SD) units. Using SCI as an example, γ₀₁ reflects the difference in the mean value of dysregulation at baseline (intercept) for 1 SD change in the SCI score, and γ₁₁ represents the differences in slope of dysregulation for 1 SD change in the SCI score. In Model 3, ASD symptoms, parent depression, quality of parent-child relationship, and positive dyadic coping measured at baseline were included as level-2 (between-participant) predictors of initial CBCL-DP scores (main effect on intercept). Finally, cross-level interaction terms between the level-2 predictors in Model 3 and time (level-1 predictor) were included in a final model (Model 4; main effect on slopes). Follow-up analyses on statistically significant slope main effects were conducted and plotted using the Johnson-Neyman technique (Johnson & Fay, 1950), which provides regions of significance for the conditional effect rather than relying on splitting the data into “low” and “high” categories.

Results

Missing data and attrition

Missing data analysis via Little’s MCAR test revealed no significant pattern to missing data. There were no significant (p >.05) sociodemographic differences (i.e., child age, sex, ID status, parent education, race, household income) between families who completed all four study visits and those who did not. From 188 families with baseline data, 28.7% (n = 54) did not complete all four time points, with three-quarters of families (75.6%) having at least three timepoints of data. Two families were removed due to incomplete baseline data for a final sample of 186 children with ASD. Multiple imputation was used to account for missing data.

Descriptive Data

All study variables had a normal distribution without outliers. The average CBCL-DP score remained relatively stable across time, although the mean at T3 (183.55) was significantly less than the mean at T1 (188.713, p = .02). Across timepoints most children fell into the average (34-44%; scores > 180) and borderline clinically significant ranges (45-52%; scores between 1 and 2 SDs above clinical cut-off; ≤ 180 < 210) of the CBCL-DP. Correlations between CBCL-DP scores at all time points and child and family socio-demographic variables (child age, parent age, race, education, and household income) were non-significant (p > .05). Boys and girls did not differ in their average dysregulation score at any time point nor did intellectual disability status (yes/no) alter the mean CBCL-DP score at any time.

Table 2 provides descriptive statistics for main study variables as well as correlations amongst these variables. Dysregulation was significantly positively correlated with SRS-2 scores and negatively correlated with mother-child and father-child relationship scores at all time points. Paternal depression symptoms did not significantly correlate with dysregulation at any time point while maternal depression symptoms significantly positively correlated with dysregulation at T2, T3, and T4. Maternal-rated parent dyadic coping was only significantly associated with dysregulation at T4. Paternal-rated parent dyadic coping was significantly negatively associated with dysregulation at all time points.

Table 2.

Descriptive statistics and bivariate correlations amongst study variables

Study Variable	1	2	3	4	5	6	7	8	9	10	11	12
1. T1 Dysregulation	-
2. T2 Dysregulation	.791**	-
3. T3 Dysregulation	.682**	.816**	-
4. T4 Dysregulation	.738**	.856**	.865**	-
5. SCI	.612**	.480**	.480**	.489**	-
6. RRB	.636**	.521**	.510**	.492**	.843**	-
7. P-C Relationship, Mother	−.374**	−.707**	−.411**	−.395**	−.380**	−.326**	-
8. P-C Relationship, Father	−.382**	−.364**	−.362**	−.328**	−.314**	−.288**	.358**	-
9. Positive Dyadic Coping, Mother	−.021	−.070	−.100	−.123	.000	−.019	.034	−.035	-
10. Positive Dyadic Coping, Father	−.141	−.133	−.156*	−.131	−.018	−.054	.130	.254**	.121	-
11. Depression Sx, Mother	.075	.198*	.273**	.210**	.112	.077	−.130	−.081	.049	−.072	-
12. Depression Sx, Father	.096	.118	.138	.122	−.002	−.014	−.039	−.159*	.072	.029	.231**	-
Mean (SD)	188.901 (20.29)	186.562 (19.75)	184.371 (18.23)	186.188 (18.90)	80.761 (19.63)	21.306 (6.04)	45.822 (6.74)	45.310 (7.00)	63.931 (11.64)	63.638 (10.17)	18.385 (7.11)	16.271 (6.08)
Range	151-263	151-264	141-231	150-238	21.5-134.5	6-34	27-60	22-60	41-94	42-90	6-46	0-38

Note:

^**p < .01,

^*p < .05;

T1 = time 1 or baseline; T2 = time 2 or study visit 2; T3 = time 3 or study visit 3; T4 = time 4 or study visit 4; SCI = social communication index; RRB = restricted and repetitive behaviors; P-C = parent-child; Sx = symptoms; SD = standard deviation.

Unconditional Models

Estimated from the unconditional means model (Model 1; Table 3), the Intra-Class Correlation coefficient was .74, suggesting that variance in CBCL-DP scores was largely due to differences between children (74%). From the unconditional linear growth model (Model 2; Table 3), the fixed effects indicated that both the intercept and the slope were significantly different from zero. The average CBCL-DP score at baseline was 188.055 with a linear decline of 1.033 units per year. The random effects indicate significant individual variability in CBCL-DP score in the intercept (i.e., differences at baseline; 328.180), and slope (i.e., difference in trajectories; 9.045). The covariance between intercept and slope (−17.579) was also significant; thus, a child’s initial CBCL-DP score was associated with their change over time. Specifically, a higher initial CBCL-DP score was associated with less decline in dysregulation over time.

Table 3.

Summary of longitudinal multi-level models of dysregulation in children with ASD

Parameters	Model 1: Null Model Intercept only or B (S.E)	Model 2: Linear Growth Model Change over timeB (S.E)	Model 3: Random-Intercepts Predictors of the level of Dys. B (S.E)	Model 4: Random Intercepts & Slopes Predictors of change in Dys. B (S.E)
Fixed Effects
Intercept (γ00)	186.505 (1.29) ***	188.055 (1.41) ***	187.177 (1.05) ***	187.163 (1.04) ***
Time (γ10)	----	−1.033 (0.34) **	−1.130 (0.39) **	−1.117 (0.36) **
SCI score (γ01)	----	----	0.249 (1.92)	0.139 (2.05)
RRB score (γ02)			7.992 (1.81) ***	9.415 (1.932) ***
P-C relationship, Mother (γ03)	----	----	−3.812 (1.11) **	−3.366 (1.19) **
P-C relationship, Father (γ04)	----	----	−2.919 (1.13) **	−3.123 (1.21) **
Positive Dyadic Coping, Mother (γ05)	----	----	−0.640 (1.05)	0.546 (1.13)
Positive Dyadic Coping, Father (γ06)	----	----	−1.156 (1.04)	−1.240 (1.11)
Depression Symptoms, Mother (γ07)	----	----	1.542 (1.01)	0.677 (1.08)
Depression Symptoms, Father (γ08)	----	----	1.109 (1.02)	1.023 (1.09)
Time*SCI score (γ11)	----	----	----	0.106 (0.71)
Time*RRB score (γ12)				−1.376 (0.67) *
Time*P-C relationship, Mother (γ13)	----	----	----	-0.432 (0.41)
Time*P-C relationship, Father (γ14)	----	----	----	0.198 (0.42)
Time*Positive DC, Mother (γ15)	----	----	----	−1.147 (0.39) **
Time*Positive DC, Father (γ16)	----	----	----	0.082 (0.39)
Time*Depression, Mother (γ17)	----	----	----	0.836 (0.37) *
Time*Depression, Father (γ18)	----	----	----	0.083 (0.37)
Random Effects
Level 1 Variance   (occasions within individuals)	82.007 (4.88)***	65.153 (4.75) ***	71.752 (5.80) ***	71.752 (5.80) ***
Level 2 Variance   (intercepts b/w individuals)	291.582 (32.21)***	328.180 (38.70) ***	118.62 (20.10) ***	115.240 (19.35) ***
Slope Variance	----	9.045 (2.57) ***	8.528 (2.86) **	5.364 (2.54) *
Covariance of Intercept and slope	----	-17.579 (7.30) **	−2.135 (9.50)	1.138 (5.20)
Model Fit
−2 log likelihood	5959.88	5928.79	4742.44	4719.66
Akaike’s Information Criterion (AIC)	5965.88	5940.79	4770.44	4763.66
Schwarz’s Bayesian Criterion (BIC)	5979.75	5968.53	4832.28	4860.83

Note:

^***p < .001,

^**p < .01,

^*p < .05;

S.E. = standard error; Dys. = dysregulation; SCI = social communication index; RRB = restricted and repetitive behaviors; P-C = parent-child; Sx = symptoms; SD = standard deviation.

To determine the best choice for modeling random effects we compared three models: random intercept only (parallel random slopes), independent random intercepts and slopes, and covarying random intercepts and slopes (unstructured variance-covariance). The AIC and log-likelihood indices were lowest for the covarying intercepts and slopes model and the log-likelihood ratio test was significant (p <.001), indicating a significantly better goodness-of-fit for the covarying compared to the independent intercepts and slopes model. Therefore, we allowed the intercept and slope to covary in remaining models.

Conditional Models

The first conditional model (Model 3; random-intercepts) examined the main effects of ASD symptom severity and family-level variables on initial CBCL-DP score. There was a significant positive association between RRBs, but not social-communication impairments, and initial CBCL-DP score. For every 1-SD increase in RRBs, children had, on average, a 7.99-unit higher score on the CBCL-DP [t(153) = 4.427, p < .001, 95% C.I. = (4.425, 11.558)]. In addition, children with more positive parent-child relationships (both maternal and paternal) had a significantly lower initial CBCL-DP score than those with less positive parent-child relationships. For every 1-SD increase in maternal-child and paternal-child relationship scores, children had a 3.812-unit [t(153) = −3.428, p =.001, 95% C.I. = (−6.009, −1.615)] and 2.919-unit [t(153) = −2.577, p = .01, 95% C.I. = (−5.157, −0.681)] decrease in initial CBCL-DP scores respectively. Child sex, age at baseline, and ID status were not significant predictors of CBCL-DP intercept or slope and thus dropped from final model testing for parsimony and to preserve statistical power.

The random intercepts and slopes model (Model 4; Table 3) tested potential interactions between level-1 (time) and level-2 (ASD and family) predictor variables (or main effect of level-2 predictors on change in CBCL-DP scores over time). Results indicated a significant interaction between time and RRBs, maternal depression, and maternal positive dyadic coping. For each 1-SD unit increase in RRB symptom scores, the slope decreased by an additional 1.376 per year [t(153) = −2.063, p = .04, 95% C.I. = (−2.694, −0.059)]. As can be seen in Figure 1, children with more severe RRB symptoms (i.e., higher SRS RRB subscale scores) at baseline showed a significant decline in dysregulation over time. Regions of significance analysis indicated that children with (centered) RRB mean scores above 0.110 (or raw score of 21.42) had significant declines in dysregulation over time. Similarly, a 1-SD unit increase in maternal positive dyadic coping indicated a 1.147 decrease in the slope of dysregulation [t(153) = −2.951, p = .004, 95% C.I. = (−1.915, −0.379)]. Follow-up analysis revealed that mean (centered) scores greater than −1.13 (or 62.80 raw score) on maternal perceptions of positive dyadic coping within the parent couple relationship also predicted declines in dysregulation across time (Figure 2). Finally, for every 1-SD unit increase in maternal depression symptoms, the slope of dysregulation increased 0.836 per year [t(153) = 2.243, p = .03, 95% C.I. = (0.099, 1.574)]. In other words, children of mothers with mean (centered) depression symptom scores above −1.87 (or a 16.52 raw score) did not show significant change in dysregulation over time (Figure 3).

Figure 1.

Conditional effect of time on emotion dysregulation as a function of mean centered child RRB symptom severity measured by the SRS. The dashed vertical line (RRB = 0.11) represents the point where the change in emotion dysregulation over time (slope) transitions from statistically significant to nonsignificant and is determined using the Johnson-Neyman technique. Children with more severe RRBs (shaded in dark grey) exhibited statistically significant declines in emotion dysregulation over time while children with less severe RRBs (shaded in light grey) did not show a significant change in emotion dysregulation over time.

Figure 2.

Conditional effect of time on emotion dysregulation as a function of mean centered maternal perceptions of positive dyadic coping. The dashed vertical line (Positive Dyadic Coping = −1.13) represents the point where the change in emotion dysregulation over time (slope) transitions from statistically significant to nonsignificant and is determined using the Johnson-Neyman technique. Children of parents reporting the use of more positive dyadic coping (shaded in dark grey) exhibited statistically significant declines in emotion dysregulation over time while children of parents who report using positive dyadic coping less (shaded in light grey) did not show a significant change in emotion dysregulation over time.

Figure 3.

Conditional effect of time on emotion dysregulation as a function of mean centered maternal depression symptoms. The dashed vertical line (Maternal Depression = −1.87) represents the point where the change in emotion dysregulation over time (slope) transitions from statistically significant to nonsignificant and is determined using the Johnson-Neyman technique. Children of mothers reporting fewer depression symptoms (shaded in dark grey) exhibited statistically significant declines in emotion dysregulation over time while children of mothers who reported more depression symptoms (shaded in light grey) did not show a significant change in emotion dysregulation over time.

Discussion

Children with ASD commonly struggle with dysregulation and this is a primary reason parents seek treatment (Mazefsky et al., 2013; Samson, Wells, et al., 2015). The current study is the first to examine trajectories of dysregulation across middle childhood and into early adolescence in ASD and to identify individual and family variables related to variability in these trajectories. We found that, on average, dysregulation declined over a period of three years irrespective of child age, sex, or intellectual disability status. A declining trajectory across middle childhood and into early adolescence in ASD is promising and aligns with developmental research on dysregulation in neurotypical children (Deutz et al., 2018). Thus, it may be important to highlight to families of children with ASD that, on average, children with ASD will increasingly be able to self-regulate with time. We also found marked between-person variability in level of and change in dysregulation among children with ASD, suggesting that group means may not reflect the experiences of all children with ASD.

Our findings indicate that the heterogeneity of ASD itself is associated with variability in dysregulation trajectories in children with ASD. Similar to previous work (Samson et al., 2014), we found that severity of RRBs at baseline was positively associated with initial level of dysregulation as well as change in dysregulation across time. Children with ASD with high RRBs had an initially high level of dysregulation and while this group was reported to experience declines in dysregulation with time, they continued to have a high CBCL-DP score three years later relative to children with ASD with initially low RRBs. It is possible that RRBs represent the behavioral manifestation of (or a coping mechanism for) dysregulated emotional states (Samson et al., 2014). For example, some researchers have postulated that RRBs such as insistence on sameness and self-injurious behavior may be the product of dysregulated responses to emotional experiences (Russell et al., 2019). It could also be that the severity of RRBs is linked to broader child emotional and behavioral regulation skills, such that children with ASD who begin with higher RRB also begin with higher dysregulation, but both improve across development in a related fashion. This would suggest alternative underlying neurobiological mechanisms driving a connection between RRBs and self-regulation in ASD. These possibilities should be examined in future research.

Results presented here support theoretical models in which dysregulation is associated with the social environment (Marroquín & Nolen-Hoeksema, 2015), particularly the emotional climate of the family environment. In non-ASD populations, evidence suggests that individuals that share environments influence one another’s emotions, and efforts to and success in regulating emotions (Butler & Randall, 2013; Marroquín & Nolen-Hoeksema, 2015). The present study is among the first to demonstrate, in a similar way, that the emotional climate of the family environment is a key context for child regulation in ASD. We found that more positive parent-child relationships were concurrently associated with a lower child dysregulation score, and more positive parent dyadic coping and fewer maternal depression symptoms predicted declines in child dysregulation over time. These findings fit well within the model suggested by Morris and colleagues (2007) which highlights the direct impact of family emotional climate on the self-regulation skills of children. For instance, parents’ ability to jointly cope with stressors (i.e., positive dyadic coping) may indicate that parents socialize or model effective regulation strategies (e.g., observe parents reframing, problem solving, and using coping strategies to manage emotional distress) for children with ASD as has been shown in children without ASD (Perry et al., 2020). Moreover, positive parent dyadic coping may allow parents to better handle everyday parenting stressors and promote the use of effective parenting strategies. In a transactional process, it is also possible that having a child with ASD with better self-regulation fosters positive dyadic coping in mothers, as these mothers may have less parenting stress and thus, more internal resources to devote to fostering a positive and supportive partner relationship.

Our finding that baseline maternal depression symptoms was associated with changes in child dysregulation across three years is consistent with previous reports that maternal psychological distress and depression is related to later increased behavioral and emotional challenges in children with ASD (Yorke et al., 2018). It is well-documented that parents of children with ASD are at increased risk for mental health problems, with estimates that upwards of 75% of mothers of a child with ASD report significant depression symptoms (Jose et al., 2017). The current study highlights that maternal depression is a significant area of need in families of a child with ASD (60% of our sample had clinically significant depression symptoms) and suggests that treating these symptoms could have a trickle-down effect on child self-regulation outcomes. Maternal depression symptoms may hinder the development of self-regulation in the child with ASD through a variety of mechanisms. Depression is associated with a restricted range of affective and regulatory behaviors, reduced use of emotion words, and reduced level of emotion-related interactions (see Suveg et al., 2011). Depression is also associated with more negative, harsh, or withdrawn parenting (Rueger et al., 2011). Future research should test the potential time-ordered effect of maternal depression on dysregulation in children with ASD and to identify the specific mechanisms driving these effects. However, it is also possible having a child with ASD who is more dysregulated leads to increases in maternal depressive symptoms. Indeed, child dysregulation may lead to stressful and frustrating parenting experiences, which in turn, could increase risk for depression. This alternative direction of effects, and the possibility of bidirectional pathways, should be tested.

It is worth noting that only maternal depression symptoms and perceptions of positive dyadic coping were associated with change in child dysregulation over time while father report was not. In families of children with ASD, mothers have been found to spend more time, on average, in caregiving compared to fathers (e.g., Hartley et al., 2014) while fathers contribute more indirectly through financial support, family decision making, etc. (Rankin, Paisley, Tomeny, & Eldred, 2019); thus, there may be more opportunity for maternal versus paternal mental health and experienced couple relationship quality to effect child self-regulation. This finding does not diminish the important role of fathers in the self-regulation development of children with ASD. Rather, these results highlight that the emotional climate of the family may impact children with ASD in complex ways. Additional research is needed to identify the specific pathways through which fathers affect child self-regulation, as they appear to differ from those in mothers.

The results of this study should be considered in light of its strengths and limitations. The longitudinal design and use of a nested statistical model given repeated measures was a strength. Both mother and father reports of child dysregulation and family processes provide a robust view of the emotional climate of the family. However, the study was limited in demographics, with generalizability of the findings most relevant to middle-class white and non-Hispanic families of a male child with ASD. Dysregulation is a complex construct and a myriad of additional factors likely contribute to dysregulation in the ASD population (Mazefsky et al., 2013). Parent-reported measures used in the current study offer only one perspective of dysregulation in children with ASD and the family environment. Findings may be biased due to the use of a single-rater; parents with higher depression symptoms or those with less positive dyadic coping may be more likely to endorse child dysregulation. In addition, the CBCL-DP is not a direct measure of dysregulation but of parent perceptions of child behaviors that indicate potential regulatory challenges. In addition, while psychometric studies of the CBCL-DP in children ASD show it to be a reliable and useful measure, more research is needed to refine the measure specifically for use in ASD (Keefer et al., 2020). Research that examines other indicators of the emotional climate of the family (e.g., interparental conflict, parenting behaviors, etc.) and establishes multi-method approaches (e.g., physiological, observational, multi-informant reports) for assessing child dysregulation as well as the family environment will be needed in future studies.

The present study mirrors previous work suggesting that gains in the self-regulation of children with ASD may be delayed (Nuske et al., 2017), and dysregulation may remain higher than what is observed, on average, in non-ASD populations across middle and older childhood. However, this study did not include a comparison group, limiting conclusions regarding differences (or similarities) in the developmental trajectory of dysregulation in children with ASD. It is also worth noting that reductions in dysregulation across childhood may not mirror advancements in the regulation skills of children with ASD and future research should consider trajectories of both dysregulation and regulation skills concurrently to fully understand patterns of development in those with ASD. In addition, while this study identified ASD- and family-related correlates of dysregulation trajectories, additional research aimed at identifying how characteristics such as RRBs and maternal depression are association with dysregulation in children with ASD will be important for the field moving forward. Results presented here suggest more severe RRBs, for instance, is associated with increased risk for problematic dysregulation trajectories across childhood. Understanding the mechanisms through which RRBs and dysregulation are associated will be important for future intervention development and has implications for the prevention of later mental health challenges in this population.

Finally, given the lack of longitudinal research on dysregulation in children with ASD, the primary goal of this study was to identify dysregulation trajectories in children with ASD as well as establish individual- and family-level factors that are associated with these trajectories as potential points of intervention. To meet these goals, the current study treated predictor variables as time-invariant, but these factors (i.e., ASD symptoms, maternal depression symptoms, parent couple coping) also change across time. Therefore, understanding how changes in individual- and family-level factors are associated with changes in dysregulation will be an important next step in moving the field toward targeted intervention. In addition, there is likely a bidirectional association between dysregulation, child ASD symptom severity, and family emotional climate; however, the analyses used in the present study cannot inform causation. This study is an important first step in identifying child and family factors associated with change in dysregulation in children with ASD and provides empirical support for future research that examines the potential bidirectional associations between the family environment and child dysregulation.

The current study has important implications for intervention. Our results indicate that children with ASD become better able to self-regulate, on average, with time. However, there is important variability in dysregulation trajectories that is associated with heterogeneity in ASD symptoms (RRB specifically) and the family environment. These child and family characteristics are amenable to intervention, if their time-ordered effects on child dysregulation are borne out in future studies. For example, RRBs can be reduced using behavioral interventions (Boyd, McDonough, & Bodfish, 2012), and efforts to treat maternal depression and improve the parent couple relationship could be integrated into family centered ASD programs (Gabovitch & Curtin, 2009). There is support in non-ASD groups that family-centered interventions that improve self-regulation during middle childhood may help prevent negative outcomes in adolescence and emerging adulthood (Stormshak et al., 2018). Similar programs that are tailored to the unique needs of children with ASD and their families may be important to consider as a prevention tool moving forward. Finally, our study suggests that parents may play a key role in shaping their child’s self-regulation in negative but also positive ways; this is an important and empowering message that is often lacking in the field of ASD.