Emotion dysregulation in autism: Severity and correlates in early childhood

Abstract

Emotion dysregulation (ED) is common and severe in older autistic youth, but is rarely the focus of early autism screening or intervention. Moreover, research characterizing ED in the preschool years (when autism is typically diagnosed) is limited. This study aimed to characterize ED in autistic children by examining (1) prevalence and severity of ED as compared to children without an autism diagnosis; and (2) correlates of ED in autistic children. A sample of 1864 parents (Mean child age = 4.21 years, SD = 1.16 years; 37% female) of 2–5 year-old children with (1) autism; (2) developmental concerns, but no autism; and (3) no developmental concerns or autism completed measures via an online questionnaire. ED was measured using the Emotion Dysregulation Inventory-Young Child, a parent report measure characterizing ED across two dimensions: Reactivity (fast, intense emotional reactions) and dysphoria (low positive affect, sadness, unease). Autistic preschoolers, compared to peers without developmental concerns, had more severe ED (+1.12 SD for reactivity; +0.60 SD for dysphoria) and were nearly four and three times more likely to have clinically significant reactivity and dysphoria, respectively. Autistic traits, sleep problems, speaking ability, and parent depression were the strongest correlates of ED in the autism sample. While more work is needed to establish the prevalence, severity, and correlates of ED in young autistic children, this study represents an important first step. Results highlight a critical need for more high-quality research in this area as well as the potential value of screening and intervention for ED in young autistic children.

Lay Summary

Emotion dysregulation is common for older autistic youth; however, it is unknown the prevalence and severity of emotion dysregulation in young, preschool aged autistic children compared with peers. Using a large sample of children with and without autism and/or developmental concerns, we found an increased prevalence and severity of emotion dysregulation in young autistic children, both in reactivity (fast, intense emotional reactions) and dysphoria (low positive affect, sadness, and unease) domains. We also identified family and child characteristics that are associated with emotion dysregulation, namely child sleep concerns, autism traits, speaking ability, and parent depression.

INTRODUCTION

Autism spectrum disorder (ASD) is a neurodevelopmental condition characterized by differences in social interaction and communication, preferences for repetition and routine, intense and nonconventional interests, and sensory differences (American Psychiatric Association, 2013). Although not part of the core criteria, accumulating evidence indicates that emotion dysregulation (ED) is common and disruptive to the lives of autistic individuals and their families (Cai et al., 2018; Cibralic et al., 2019; Mazefsky et al., 2013). ED in autism has been linked to higher rates of mental health disorders, use of psychotropic medications, psychiatric hospitalizations, and suicidality in adolescence and adulthood (Conner et al., 2020, 2021). Despite this, understanding of ED in early childhood when autism diagnoses are commonly made, remains limited. Clinically, ED is not typically assessed or considered when making a diagnosis or beginning early intervention. In fact, there is currently no empirically supported treatment targeting ED in young autistic children. Understanding the prevalence, severity, and correlates of ED in autism early in childhood is an essential first step towards directing and informing prevention and intervention efforts.

Emotion dysregulation: Typical development

Emotion regulation has been broadly defined as the capacity to monitor, evaluate, and modify one’s emotional state in order to achieve one’s goals (Gross, 2013; Thompson, 1994). The ability to regulate one’s emotions is a core developmental task which begins in infancy and continues through young adulthood. In infancy and early childhood, emotion regulation develops and occurs largely through co-regulation with parents or other important adults (Eisenberg et al., 1998; Morris et al., 2017). Over time, children gradually gain independence and begin to internalize these regulatory processes. Emotion dysregulation is a term generally used to describe the behavioral and emotional manifestations of difficulty regulating one’s emotions− emotional responses (or lack of emotional response) that get in the way of one’s goals. ED is a transdiagnostic characteristic implicated in nearly all psychiatric disorders and maladaptive behaviors (e.g., aggression, self-injury; McLaughlin et al., 2011). In the general developmental literature, ED has been associated with poorer language abilities (Eisenberg et al., 2005), social difficulties (Berkovits & Baker, 2014), poorer cognitive functioning (Jahromi & Stifter, 2008; Norona & Baker, 2017), greater internalizing and externalizing problems (Calkins & Dedmon, 2000), sleep problems (Williams et al., 2017), and features of the parent–child relationship (Morris et al., 2017).

Emotion dysregulation: Autism

Several reviews (Cai et al., 2018; Mazefsky et al., 2013) have highlighted that ED challenges are common and significant amongst autistic individuals. Recent research suggests that rates of clinically elevated reactivity (fast, intense reactions and difficulty calming down) and dysphoria (low positive affect, sadness, unease) are four times and two times higher, respectively, in school-aged autistic youth (age 6–17) than in the general population (Conner et al., 2021). Importantly, the majority of research to date has focused on older children and adolescents, and high-quality research characterizing ED in young autistic children is scarce. A few studies indicate that young autistic children have more severe emotion dysregulation relative to neurotypical peers, however this research is limited by small and homogenous samples (see Cibralic et al., 2019 for a review; Jahromi et al., 2013; Samson et al., 2014).

There are both theoretical and empirical reasons why we might expect autistic children to have more severe emotion dysregulation than neurotypical peers. Core characteristics of autism, such as differences and challenges with communication, difficulty with flexibility and change, and sensory dysregulation would all be expected to increase the likelihood of ED (Mazefsky et al., 2013). Consistent with this, research on correlates of ED in autistic children has found that higher levels of autism traits are associated with more severe ED (Berkovits et al., 2017; Goldsmith & Kelley, 2018; Konstantareas & Stewart, 2006; Samson et al., 2014).

Further, research on neurotypical children has highlighted the role of language and developmental ability in the development of emotion regulation (Beck et al., 2012; Norona & Baker, 2017), thus the common co-occurrence of language, developmental delay, and intellectual disability in autism might also be expected to influence ED. Thus far, however, empirical research has not supported this hypothesis within autism studies. Intellectual and language ability have generally not been found to be associated with ED in young autistic children (Berkovits et al., 2017; Samson et al., 2014) although both may be associated with the types of emotion regulation strategies children use (e.g., Nuske et al., 2017; Zantinge et al., 2017). It is important to note that most of the studies looking at IQ and language ability in young autistic children have had small samples of mostly speaking participants with relatively high IQs (e.g., Berkovits et al., 2017; Samson et al., 2014), thus limiting the ability to examine how ED might be related to these features across the full spectrum of ability in these areas.

Autistic children and their families are at increased risk for a number of other challenges and stressors that may be associated with ED, but that have been less well studied in the empirical literature. An estimated 60%–86% of autistic children have significant sleep problems (Souders et al., 2017). Sleep disruption is known to have a negative impact on emotional and behavioral functioning (Williams et al., 2017), and has been shown to correlate with these challenges in both typical (Han et al., 2022) and autistic samples (Favole et al., 2023).

Families of autistic children are also under increased emotional and economic pressure compared to families of non-autistic children (Estes et al., 2009; Montes & Halterman, 2008b, 2008a). It is well established that parents and the larger family context play a crucial role in typical emotion regulation development (Morris et al., 2007), and parent mental health challenges are a known risk factor for ED in neurotypical samples (Silk et al., 2006; Zimmer-Gembeck et al., 2021). There is some evidence for a connection between parent stress and child ED in families of autistic children (Mills et al., 2022; Davis & Carter, 2008), which would suggest the importance of examining both socioeconomic and parent mental health as potential correlates.

One important limitation of prior research on ED in autistic children has been the use of measures that were not developed specifically to measure emotion dysregulation (e.g., the CBCL; Samson et al., 2014) and/or measures that were developed for and with children who do not have autism (e.g., the Emotion Regulation Checklist; Jahromi et al., 2013). The Emotion Dysregulation Inventory-Young Child (EDI-YC; Day, Northrup, & Mazefsky, 2023b) was developed using Patient-Reported Outcomes Measurement Information System (PROMIS^®) methodology to capture emotion dysregulation in both autistic and non-autistic young children. The EDI has demonstrated excellent internal and external validity in samples with and without autism and other neurodevelopmental disorders (Day, Mazefsky, et al., 2023). The measure captures two dimensions of ED: Reactivity, which includes rapidly escalating, intense, and labile negative affect and difficulty calming down; and dysphoria, which includes sadness, unease, and anhedonia. These two factors capture challenges with down-regulating negative emotion (reactivity) and with upregulating positive emotion (dysphoria), thereby giving a fuller picture of emotion dysregulation difficulties that have not typically been capture by other measures.

No prior work has examined reactivity and dysphoria in autistic preschoolers, however research using the original EDI (parent report measure developed for youth aged 6–17) suggests that autistic youth have higher severity for both dimensions compared to non-autistic peers, though effect sizes were larger for reactivity (Conner et al., 2021). A study examining correlates of dysphoria and reactivity in autistic youth (Age 6–17; Northrup et al., 2021) found that younger participants, female participants, participants whose parents had lower educational attainment, participants with more restricted and repetitive behaviors, and individuals with more fluent speaking ability (as compared to non/minimally speaking individuals) had higher reactivity severity. With regards to dysphoria, older participants, those without intellectual disability, and those with more restricted and repetitive behaviors had more severe dysphoria. Overall, these results suggest somewhat different patterns of results for reactivity and dysphoria, emphasizing the importance of examining correlates of these two dimensions separately.

The present study

In summary, research on the prevalence, severity, and correlates of ED in early childhood in autism remains limited by small, relatively homogenous samples that have not allowed researchers to capture the full spectrum of autistic individuals. The present study represents the first large-scale characterization of ED in autistic preschoolers (aged 2–5 years). This secondary data analysis uses data collected as part of the psychometric validation of the Emotion Dysregulation Inventory – Young Child (EDI-YC; Day, Mazefsky, et al., 2023). It builds on existing research in several important ways. First, the large sample is representative of the full spectrum of autistic traits, language, and cognitive abilities. Second, the study utilizes the newly developed EDI-YC which is validated for both autistic and non-autistic samples and measures multiple dimensions of dysregulation (i.e., reactivity and dysphoria) that have not been differentiated in previous studies (Day, Mazefsky, et al., 2023). Finally, it allows for a broad examination of potential correlates of ED in autistic children, including a number of factors that have been strongly linked to ED in neurotypical children but remain underexplored in autism research. The study had the following two aims:

1. Examine differences in the prevalence and severity of ED (reactivity and dysphoria) between autistic children, children with developmental concerns but not autism, and children with neither developmental concerns nor autism. Based on prior work, we expected the autistic sample to have more severe ED than children without an autism diagnosis.

2. Examine demographic, child-related, and parent-related correlates of ED (reactivity and dysphoria) in an autism sample. Based on prior research (outlined above), we expected that greater autism characteristics, more severe sleep problems, parent depression, and lower family income would be associated with greater ED. We did not have a hypothesis as to whether child speaking ability and intellectual disability would be associated with ED, given that prior research on young autistic children has not found an association but has been limited in the amount of variability represented in the samples. We also did not have specific hypotheses regarding how these variables may be differentially associated with reactivity verses dysphoria, as this has not been examined in prior work with young children.

METHODS

Participants

All participants were parents or guardians of 2- to 5-year-old children. Multiple recruitment methods were utilized to increase sample representation. The majority of autism participants were recruited through Simons Powering Autism Research (SPARK), a national research registry of individuals with a professional ASD diagnosis (participants in the SPARK registry also had Social Communication Questionnaire [SCQ; a well-validated autism screening tool; Rutter et al., 1993] scores ≥12 at the time they registered with SPARK). Additional parents of 2- to 5-year-old autistic and non-autistic children were recruited through all 14 pediatrics practices that are part of Pediatric PittNet at the University of Pittsburgh (urban, suburban, and rural locations) as well as 25 early intervention programs, and over 100 preschools and community daycare programs in the Mid-Atlantic region and New York (n = 1053; see Day, Mazefsky, et al., 2023 for details).

Participants (total n = 1864) were separated into three groups for analysis: (1) children with an ASD diagnosis (“autism group”; N = 853), which included all participants from SPARK (n = 811) and those in the regional sample whose parents endorsed an ASD diagnosis (i.e., answered “yes” in response to the question, “Has your child been diagnosed with autism spectrum disorder?”; n = 42); (2) children whose parents reported having developmental concerns (i.e., answered “yes” in response to the question, “Have you ever had any concerns about your child’s development?”), but who did not report an ASD diagnosis (“parent concerns group”; N = 296), and (3) children whose parents reported no developmental concerns or ASD (“no concerns/autism group”; N = 714).

Procedures

All participants provided their consent to participate in accordance with University of Pittsburgh ethical standards and the 1964 Helsinki declaration and its later amendments. Cross-sectional data were collected between January 13, 2021 and February 12, 2021. All participants in both samples underwent the same protocol concurrently and completed the questionnaires online.

Measures

Data were originally collected as part of a psychometric study of the EDI-YC (see Day, Mazefsky, et al., 2023), and thus measures were not originally selected to fulfill the aims of the present study. A full list of measures can be found in the National Institutes of Mental Health Data Archive (NDA) under R01 HD079512. Measures included in the present study were selected a priori from the available measures based on existing empirical and theoretical research (as outlined in the introduction).

Emotion dysregulation

ED was measured using the Emotion Dysregulation Inventory – Young Child (EDI-YC), a 22-item parent-report measure that assesses two domains of ED in preschool-aged children: reactivity (fast, intense emotional reactions; 15 items) and dysphoria (low positive affect, unease, sadness; 7 items; Day, Northrup, & Mazefsky, 2023b). Parents rate each item on a 5-point Likert scale (0 = not at all; 4 = very severe). The EDI-YC has demonstrated excellent reliability and strong psychometric properties (Day, Mazefsky, et al., 2023). EDI-YC items do not show differential item functioning based on age, sex, or developmental status (i.e., presence of autism versus other developmental disorder/concern versus no concern). Raw scores were converted to standardized theta scores (mean of 0, SD of 1) based on a general population sample. In the present paper, the EDI-YC’s established standardized clinical cut-offs were utilized, which are based on a score greater than one standard deviation above the general population mean as an indicator of significant emotion dysregulation. This approach is similar to other standardized measures of children, such as the ASEBA measures (e.g., CBCL; Achenbach & Rescorla, 2000) that base determination of whether a score is elevated based on what would be expected from a ‘normative’ sample. Standards for measure development include having a general/normative sample to determine norms as best practice as this aids in interpretation (Hunsley & Mash, 2008, PROMIS Instrument Development Standards Version 2.0; Youngstrom et al., 2017). The same norms and cutoff (>1 SD above the general population mean) was used for all participants, regardless of group.

Autism traits

Autism traits were assessed using the lifetime version of the SCQ, a well validated, 40-item parent-report autism screening questionnaire (Rutter et al., 1993). Parents indicate the presence or absence of social, communication, and behavioral characteristics common in autism.

Child sleep problems

Sleep problems were assessed using the PROMIS Early Childhood Sleep Problems scale, a 16-item parent-report scale that measures sleep disturbances and sleep-related impairment in young children (Lai et al., 2022). Parents rate the frequency of sleep problems in the past 7 days using a 5-point Likert scale (1 = never; 5 = always). Higher scores indicate more severe sleep problems.

Parent depression

Parent depression was measured using two items (i.e., “having little interest or pleasure in doing things”; “feeling down, depressed, or hopeless”) from the Survey of Well Being of Young Children (SWBYC; Sheldrick & Perrin, 2013), rated on a 4-point Likert scale (1 = Not at all; 4 = Nearly every day). These two items were highly correlated (r = 0.74) and were collapsed into a single variable for analysis using their mean.

Parent reported demographics and characteristics

Several variables were generated using responses on the study demographic form, including child age and sex assigned at birth, parent education, family income, child race and ethnicity, presence of intellectual disability in the child, and child speaking ability. Child race was collected as mandated by the US National Institutes of Health (NIH). Parents selected their child’s race using a “check all that apply” question with the following options: American Indian or Alaskan Native; Asian; Native Hawaiian or other Pacific Islander; White; Black or African American; Other. Nine percent of the sample (N = 173) selected more than one racial identity. For analysis, we used a deterministic bridging method to assign multi-racial participants to the single race they identified with that had the highest prevalence in the sample other than White (thus prioritizing their minoritized racial identity). This resulted in four racial categories: Asian (N = 87; 4.7%), Black (N = 284; 15.2%), White (N = 1411; 75.7%), and Another Race (N = 70; 3.8%). Amongst participants identifying only as White, 150 (10.6%) also identified as Hispanic or Latinx. Thus, approximately 68% of the sample was White, non-Hispanic, while 32% identified with a minoritized racial or ethnic identity.

Intellectual disability (ID) was measured using parent response to the question, “Has your child ever been diagnosed with intellectual disability?” (Yes/No). Speaking ability was measured with a single item asking parents: “How would you best describe │child name│’s current verbal ability?” (1 = meaningful, fluent speech; 2 = meaningful, phrase speech; 3 = meaningful, single words; 4 = non-meaningful words or phrases only; 5 = nonverbal). Parents who rated their child’s speaking ability as non-meaningful words or phrases only (N = 107, 5.8%) or nonverbal (N = 191, 10.4%) were collapsed into a single category to represent children who were minimally or non-speaking.

Data analysis

Data analytic plan

All analyses were conducted in R version 4.3.1 (R Core Team, 2023). Two separate multivariable linear regressions, one predicting reactivity and one predicting dysphoria, were used to evaluate the effect of group (autism, parent concerns, no concerns/autism) on ED (Aim 1). In these models, we controlled for demographic features that differed between the groups. We also controlled for speaking ability and intellectual disability to ensure that group differences were not due solely to differences in developmental abilities between the groups. Models were first run with the no concerns/autism group as the reference group, and then re-run with the autism group as the reference group to assess differences between the autism and parent concerns groups.

Multivariable linear regressions were also used to examine correlates of reactivity and dysphoria within the autism group (Aim 2). Our primary interest was in understanding what features predict ED within autistic children, thus this second set of analyses included only the autistic sample. The samples of children without an autism diagnosis were not included in these analyses as we were interested in understanding the association between ED and individual characteristics that are common in autistic children, but uncommon in the general population (e.g., autism characteristics, minimal language, and intellectual disability). Thus, inclusion of the non-autistic samples in these analyses would lead to confounds between autism diagnosis and our variables of interest (e.g., 92% of the non-speaking children also have an autism diagnosis) and make it difficult to examine interactions between group and these other variables.

Chosen predictors in these models were based on (a) prior research on correlates of ED in both autistic and neurotypical samples; and (b) measures available in the EDI-YC psychometric validation study. Specifically, we included child sex, age, parent education, household income, speaking ability, ID, autism traits, parent depression, and child sleep problems as predictors. In addition, because we were interested in identifying unique correlates of reactivity and dysphoria, in each model we controlled for the other EDI scale (i.e., in the model predicting reactivity, we controlled for dysphoria, and vice versa).

Missing data description and imputation

Table 1 displays available data for each variable in each group (in the “N(total)” column). Overall <1% (n = 16) of the sample was missing data for at least one variable in the Aim 1 analyses, but 45% (n = 384) of the autism sample was missing data for at least one variable in the Aim 2 analyses (see Supporting Information for detailed description of missingness). All participants had complete EDI-YC data, and no individual variable was missing for more than 17% of the sample.

TABLE 1.

Demographics and characteristics of the study sample.

		Whole sample (N = 1864)	Autism (N = 853)	Parent concerns (N = 296)	No concerns/autism (N = 714)
Variable	N (total)	M (SD)/N (%)	M (SD)/N (%)	M (SD)/N (%)	M (SD)/N (%)	F or X2	Contrasts
Child sex at birth (% Male)	1856	1170 (63%)	651 (76%)	188 (64%)	331 (47%)	Χ 2 = 146.00***
Child age (months)	1863	50.46 (13.97)	55.13 (13.21)	47.38 (13.14)	46.17 (13.48)	F = 97.70***	Aut>PC &NC
Child ethnicity (%Hisp/Lat)	1852	219 (12%)	166 (20%)	17 (6%)	36 (5%)	Χ 2 = 90.20***
Child race1	1857
American Indian or Alaska Native		92 (5%)	44 (5%)	9 (3%)	39 (6%)	Χ 2 = 2.84
Asian		14 (1%)	12 (1%)	2 (1%)	0 (0%)	Χ 2 = 10.26**
Black		284 (15%)	110 (13%)	52 (18%)	122 (17%)	Χ 2 = 7.02*
Native Hawaiian or Other Pacific Islander		36 (2%)	21 (2%)	2 (1%)	13 (2%)	Χ 2 = 3.75
White		1578 (85%)	735 (86%)	252 (85%)	591 (83%)	Χ 2 = 2.20
Another race		45 (2%)	27 (3%)	7 (2%)	11 (2%)	Χ 2 = 4.25
Multiple racial identities		173 (9%)	87 (10%)	25 (8%)	61 (9%)	Χ 2 = 1.46
Household income	1773
Less than 20,999		203 (11%)	113 (14%)	39 (14%)	51 (8%)	Χ 2 = 54.33***
21,000 to 35,999		212 (12%)	101 (12%)	28 (10%)	83 (12%)
36.000 to 50,999		223 (13%)	128 (16%)	30 (11%)	65 (10%)
51,000 to 65,999		164 (9%)	82 (10%)	30 (11%)	52 (8%)
66.000 to 80,999		195 (11%)	95 (12%)	26 (9%)	74 (11%)
81,000 to 100,999		221 (12%)	91 (11%)	36 (13%)	94 (14%)
101,000 to 130,999		224 (13%)	89 (11%)	33 (12%)	102 (15%)
131,000 to 160,999		150 (8%)	56 (7%)	26 (9%)	68 (10%)
Over 160,000		181 (10%)	62 (8%)	32 (11%)	87 (13%)
Parent education	1849
Less than 8th grade		3 (<1%)	3 (<1%)	0 (0%)	0 (0%)	Χ 2 = 71.24***
Some high school		36 (2%)	19 (2%)	8 (3%)	9 (1%)
High School Degree		211 (11%)	104 (12%)	30 (10%)	77 (11%)
Some college/AA/tech school		579 (31%)	326 (38%)	78 (27%)	175 (25%)
Bachelor’s degree.		500 (27%)	230 (27%)	81 (28%)	189 (27%)
Post graduate degree		520 (28%)	169 (20%)	97 (33%)	254 (36%)
Child ID (% yes)	1748	294 (17%)	273 (37%)	17 (6%)	4 (1%)	Χ 2 = 369.71***
Child speaking ability	1845
Minimally or nonspeaking		298 (16%)	276 (33%)	16 (5%)	6 (1%)	Χ 2 = 691.23***
Single words		225 (12%)	158 (19%)	36 (12%)	31 (4%)
Phrase speech		468 (25%)	270 (32%)	83 (28%)	115 (16%)
Fluent		854 (46%)	141 (17%)	158 (54%)	555 (79%)
Autism traits (SCQ)	1838	13.33 (9.06)	19.84 (7.19)	9.43 (6.96)	7.06 (5.98)	F = 754.11
Parent depression score	1550	0.49 (0.68)	0.63 (0.76)	0.50 (0.67)	0.37 (0.59)	F = 23.77***	Aut>PC>NC
Child sleep problems	1769	0.44 (1.17)	0.90 (1.13)	0.38 (1.16)	−0.06 (1.00)	F = 143.58***	Aut>PC>NC
Child Reactivity (EDI)	1864	0.66 (1.21)	1.23 (1.15)	0.57 (1.15)	0.03 (0.95)	F = 240.84***	Aut>PC>NC
Child Dysphoria (EDI)	1864	0.39 (0.93)	0.75 (0.89)	0.28 (0.96)	0.01 (0.79)	F = 143.48***	Aut>PC>NC

Note:

^1-Race: participants were able to “check all that apply” for US Census race categories; thus, values represented in this table do not total to 100% and include individuals with multiple racial identities.

^***p < 0.001;

^**p < 0.01;

^*p < 0.05.

Abbreviations: Aut, Autism; ID, Intellectual Disability; NC, No concerns/autism; PC, Parent Concerns; SD, standard deviation.

Multiple imputation with chained equations (using the Mice package in R; Zhang, 2016) was used to account for missing data in analyses. Fifteen individuals (8 autism; 6 no concerns/autism; 1 unknown group) were missing key demographic data (race, ethnicity, diagnostic status) and were excluded from imputation and analysis, as we felt it would not be appropriate to impute these variables. The remaining 1850 participants were included in the imputation and study analyses. Multiple imputations were conducted using the random forest technique, which involves constructing multiple decision trees using random subsets of available variables and observations, and then averaging the predictions from these trees to impute the missing values. The random forest approach is suitable for datasets with complex relationships and many variables, as it can capture non-linear dependencies and interactions between them. An important assumption of multiple imputation is that the missing data is missing at random (MAR). The MAR assumption does not require that data be missing completely at random, but rather that the missingness can be explained by other variables in the dataset. Our data fit this assumption well. To have the most complete information for data imputation, we included all available variables from the original dataset (which included variables not included in the present manuscript analyses, but that we deemed would be useful for imputing missing data). A complete list of all variables can be found in the National Institutes of Mental Health Data Archive (NDA) under R01 HD079512. Following recommendations by Bodner (2008), Graham et al. (2007), and White et al. (2011) we performed 20 imputations. Pooled regression analyses are reported below.

RESULTS

Table 1 displays descriptive statistics and group comparisons for demographic characteristics and measures of interest. There were several significant differences between the autism, parent concerns, and no concerns/autism groups with regards to sociodemographic characteristics, including race, ethnicity, sex, income, and parent education (though effects sizes were negligible to small). The groups also differed significantly in speaking ability, presence of ID, EDI reactivity and dysphoria, autism characteristics, sleep problems, and parent depression (see Table 1).

Aim 1: Characterizing ED in preschoolers with and without an autism diagnosis

Using a clinical cutoff of one standard deviation (SD) above general population norms, 58% (n = 492) and 46% (n = 396) of children in the autism group (total n = 853) exceeded cutoffs for reactivity and dysphoria respectively. In comparison, 34% (n = 101) and 24% (n = 70) of children in the parent concerns group (total n = 296) exceeded cutoffs for reactivity and dysphoria respectively, and 15% (n = 108) and 14% (n = 103) of children in the no concerns/autism group (total n = 714) exceeded cutoffs for reactivity and dysphoria respectively (Figure 1).

FIGURE 1.

Percentage of children in each group exceeding clinical cutoffs (scores greater than one standard deviation above general population mean) for reactivity and dysphoria on the Emotion Dysregulation Inventory-Young Child.

Due to significant differences between groups in child sex, age, race, ethnicity, family income, and parent education, we controlled for these sociodemographic variables as well as speaking ability and intellectual disability in Aim 1 regression analyses. Pooled regression results (Table 2) indicated that group was a significant predictor of EDI-YC reactivity and dysphoria scores, controlling for demographic variables, speaking ability, and ID, with all three groups differing from one another. Reactivity in the autism group was higher than in the no concerns/autism group (+1.12 SD) and the parent concerns group (+0.63 SD). Similarly, the autism group had higher dysphoria than the no concerns/autism group (+0.60 SD) and the parent concerns group (+0.34 SD). The parent concerns group was also higher than the no concerns group in reactivity (+0.49 SD) and dysphoria (+0.26 SD).

TABLE 2.

Aim 1 pooled regression results.

	Reactivity		Dysphoria
	B (SE)	β	B (SE)	β
Group: Autism	1.36 (0.08)	1.12 ***	0.56 (0.07)	0.6 ***
Group: Concerns	0.59 (0.08)	0.49 ***	0.24 (0.06)	0.26 ***
Sex: Female	0.13 (0.05)	0.1*	0.15 (0.04)	0.16***
Age (months)	0 (0)	−0.01	0 (0)	0.04
Race: Asian	−0.19 (0.12)	−0.16	0.13 (0.09)	0.14
Race: Black	−0.07 (0.07)	−0.06	0 (0.06)	0
Race: Another Race	−0.19 (0.14)	−0.16	−0.03 (0.11)	−0.03
Hispanic/Latinx	−0.07 (0.08)	−0.06	0.03 (0.06)	0.04
Parent Education	−0.06 (0.03)	−0.06*	−0.05 (0.02)	−0.06*
Family Income	−0.04 (0.01)	−0.09***	−0.03 (0.01)	−0.1***
Speaking ability: Single words	0.08 (0.1)	0.07	−0.19 (0.08)	−0.21*
Speaking ability: Phrase speech	0.15 (0.08)	0.12	−0.38 (0.07)	−0.41***
Speaking ability: Fluent speech	0.26 (0.09)	0.21**	−0.44 (0.07)	−0.47***
Intellectual Disability	−0.09 (0.08)	−0.07	−0.09 (0.06)	−0.09
Constant Pooled R2 [95% CI]	0.33 (0.16) 0.23 [0.195–0.262]	−0.72***	0.66 (0.13) 0.19 [0.155–0.219]	−0.02

Note:

^*p < 0.05;

^**p < 0.01;

^***p < 0.001.

Given that autism diagnosis was almost entirely confounded with recruitment source (i.e., SPARK verses regional recruitment, see methods), we also ran a sensitivity analysis to examine differences in EDI scores between the small autism group (N = 42) in the regionally recruited sample and the non-autism group in the regionally recruited sample. These analyses show extremely similar results to those reported above with the SPARK sample. We have included the full results of these sensitivity analyses in the Supporting Information (see Table S2).

Aim 2: Correlates of ED in autistic sample

Reactivity

Lower family income, more fluent speaking ability, more autism traits, higher levels of parent depression, and more child sleep problems were all significantly associated with higher reactivity scores (Figure 2; Table 3). Examination of standardized coefficients suggest that sleep problems and speaking ability were the two strongest predictors of reactivity. A one SD change in sleep problems was associated with a 0.30 SD change in reactivity. A one-step increase in speaking ability (e.g., from minimally speaking to single word speech to phrase speech, etc.) was associated with approximately a 0.25 SD increase in reactivity. Thus, children with fluent speaking ability had reactivity scores that were 0.75 SD higher on average than non/minimally speaking children. Altogether the model explained about 41% of the variability in reactivity scores, however about 10% of this variability was explained by dysphoria scores (included in the model as a control).

FIGURE 2.

Aim 2 pooled regression results, standardized Beta (β) coefficients for predictors of reactivity and dysphoria.

TABLE 3.

Aim 2 pooled regression results, predictors of reactivity and dysphoria.

	Reactivity		Dysphoria
Independent variables	B (SE)	β	B (SE)	β
Sex: Female	0.04 (0.07)	0.04	0.05 (0.06)	0.06
Age(months)	0.00 (0.00)	0.00	0.00 (0.00)	0.03
Parent Education	−0.04 (0.04)	−0.04	−0.07 (0.03)	−0.08*
Family Income	−0.03 (0.02)	−0.07*	0.01 (0.01)	0.01
Speaking ability: Single words	0.25 (0.09)	0.22**	−0.19 (0.07)	−0.22**
Speaking ability: Phrase speech	0.52 (0.08)	0.45***	−0.28 (0.07)	−0.32***
Speaking ability: Fluent speech	0.71 (0.1)	0.62***	−0.27 (0.08)	−0.31**
Intellectual Disability	0.01 (0.07)	0.01	−0.04 (0.06)	−0.05
Autism Traits	0.01 (0.01)	0.07*	0.03 (0.00)	0.21***
Parent Depression	0.13 (0.05)	0.08*	0.10 (0.04)	0.08*
Sleep Problems	0.30 (0.03)	0.3***	0.08 (0.03)	0.10**
Dysphoria control	0.50 (0.04)	0.39***
Reactivity control			0.31 (0.03)	0.40***
Constant Pooled R2 [95% CI]	0.32 (0.23) 0.41 [0.352–0.456]	−0.30***	0.09 (0.18) 0.39 [0.334–0.437]	0.20***

Note:

^*p < 0.05;

^**p < 0.01;

^***p < 0.001.

Dysphoria

Lower parental educational attainment, less fluent speaking ability, more autism traits, higher levels of parent depression, and more child sleep problems were all associated with higher dysphoria (Figure 2; Table 3). A comparison of standardized coefficients suggests that speaking ability and autism traits were the strongest predictors of dysphoria. With regards to speaking ability, non/minimally speaking children had the highest dysphoria scores and were about 0.19 SD higher on dysphoria than children with single words and about 0.27 SD higher than children with phrase or fluent speech. A one SD increase in autism traits was associated with a 0.21 SD increase in dysphoria. Altogether, the model explained about 39% of the variability in dysphoria. Again, about 10% of the variance was explained by the reactivity control.

DISCUSSION

This study establishes that clinically significant ED, which has been shown to be highly prevalent and disruptive in the lives of older autistic children and adults (Conner et al., 2021; Northrup et al., 2021), is also highly prevalent in autistic children early in life. Although some dysregulation is developmentally expected in the preschool years, our research clarifies that the severity of ED for many autistic children is well outside the range of what is common in similarly aged non-autistic children. Additionally, this study highlights that child sleep problems, child speaking ability (albeit in complex ways, discussed below), parental depression, and autism traits are all meaningfully associated with severity of ED in young autistic children.

Using the EDI-YC—a newly developed measure of ED validated for use with both autistic and non-autistic 2–5-year-old children—we found that 60% of autistic children exceeded clinical cutoffs (>1 SD above general population mean) for reactivity and 46% exceed cutoffs for dysphoria. Autistic children were four times more likely to exceed clinical cutoffs for reactivity compared to peers without autism or parental concerns, and more than three times as likely to exceed clinical cutoffs for dysphoria. These findings suggest that the majority of families with young autistic children are struggling with ED, likely from the time they receive the diagnosis or even earlier.

Drawing on a large and relatively heterogenous sample, the present study corroborates prior research reporting poorer emotion regulation in young autistic children compared to neurotypical peers (see Cibralic et al., 2019 for a review), including some research suggesting heightened ED starting in infancy and early toddlerhood (Mallise et al., 2020; Northrup et al., 2024; Raza et al., 2020). Our findings also lend perspective to the literature showing increased internalizing and externalizing challenges and co-occurring mental health conditions in autistic children as compared to peers (Guerrera et al., 2019; Salazar et al., 2015; Vaillancourt et al., 2017). Emotion dysregulation is a transdiagnostic process that underlies numerous mental health challenges and conditions (Cole et al., 1994; Keenan, 2000), thus identifying and intervening on clinically significant emotion dysregulation early in development may be key to preventing more serious challenges later in childhood.

It is notable that our “parent concerns” group, made up of young children whom do not (yet) have an autism diagnosis, but about whom parents reported some developmental concern, fell between the no concerns/autism and the autism groups in the severity and prevalence of ED challenges. This heterogenous group is likely made up of some children who will ultimately be diagnosed with autism, some with various other forms of neurodiversity, and some whose development will ultimately be typical. The results suggest that struggles with ED are more common amongst children with developmental concerns and thus screening and intervention could benefit a broader range of young children, not just those with an autism diagnosis. Further research on this group, particularly work that follows up on both developmental and emotional outcomes, will be important to further understand these results.

In the second aim, we identified autism traits, sleep problems, speaking ability, parental depression, parent education, and income as correlates of ED in autistic preschoolers. The finding that autism traits are strongly associated with ED corroborates mounting evidence in this area across the lifespan (Goldsmith & Kelley, 2018; Northrup et al., 2021; Samson et al., 2014). Autistic traits likely impact children’s use and the effectiveness of coping strategies (Hirschler-Guttenberg et al., 2015; Jahromi et al., 2012) and increase their vulnerability to experiencing ED (e.g., sensory dysregulation, unaccommodating environments, difficulty with change). Understanding the unique needs of autistic children to both prevent and help intervene on ED is an essential clinical need that is currently unmet. Future work with richer measurement of autism traits (e.g., separately measuring social communication challenges, restricted and repetitive behaviors, and sensory dysregulation) would provide important additional insight into this association.

Speaking ability was strongly associated with both reactivity and dysphoria, but in opposite directions− more fluent speech was associated with higher levels of reactivity but lower levels of dysphoria. These findings are in line with research from Northrup et al., which found that more fluent speaking ability was associated with higher EDI reactivity scores (though speaking ability was not associated with dysphoria) in a sample of 6–17-year-old autistic youth (Northrup et al., 2021). More research is needed to better understand why more fluent speaking ability may be associated with higher reactivity but lower dysphoria at this age. Perhaps preschool aged autistic children who speak more fluently are more likely to be in more inclusive, and less structured environments (e.g., community preschool settings rather than therapeutic schools) and have higher expectations placed on them (e.g., for participating in chores and routines independently). These environments and expectations might be more challenging (e.g., more sensory overload, more transitions, unexpected events), leading to more reactivity, but may also afford more opportunities for independence and positive interaction that are protective against dysphoria. Regardless of the explanation, this finding highlights the importance of understanding ED as consisting of both reactivity and dysphoria. While strongly correlated with one another, differences in how these features of ED relate to other characteristics (like speaking ability) suggest they may require different clinical considerations.

These findings also provide evidence that ED may not be associated with children’s developmental level in a simple, straightforward manner. In addition to the complex association with speaking ability, ED was not found to be associated with age or with our measure of intellectual disability. Furthermore, differences between groups were still apparent accounting for age, speaking ability, and intellectual disability. While these results need to be corroborated in studies with more rigorous measures of developmental ability, they provide initial evidence that differences in development associated with autism (e.g., reduced or delayed speaking ability, intellectual disability) do not fully account for differences in ED.

We revealed several additional factors, previously well established to be associated with ED outside of autism (Morris et al., 2007; Palmer & Alfano, 2017), as significant correlates of ED in young autistic children. These include sleep, parental depression, parent education, and family income. Sleep problems are at least twice as common for autistic children compared to non-autistic children (Reynolds et al., 2019), and parents of autistic children are at significant increased risk for mental health (Bonis, 2016; Estes et al., 2009) and socioeconomic (Montes & Halterman, 2008a, 2008b) challenges. Importantly, challenges like sleep and parent depression are likely to have bidirectional associations with ED. In other words, it is likely that child ED has negative impacts on sleep and parent mental health, and that sleep problems and parent mental health challenges also influence child ED. While the present study cannot speak to the directionality or causality of these associations, our findings point to the need for more holistic considerations in early intervention for young autistic children.

While there is evidence for successful interventions focused specifically on improving parent mental health (Dykens et al., 2014) and child sleep (Cuomo et al., 2017) in families of autistic children, there is much progress to be made in integrating these evidenced based practices into standard early intervention care for families. Further, there are currently no empirically supported interventions specifically targeting ED in young autistic children (prior to school age), although some are in the early stages of development (Rispoli et al., 2019). Future work should consider how to support families of autistic children across multiple domains simultaneously, taking both parent and child needs into account. There are positive results in the non-autism literature for intervening simultaneously on parent mental health and parent–child social–emotional interactions (e.g., Baggett et al., 2021), that may serve as a model.

Limitations and future directions

Although there were several notable strengths in this study, including a substantial sample size and the use of a measure well-validated to assess ED for autistic and non-autistic children, it is also important to highlight some important limitations and recommendations for future research. First, reliance on parent report comes with inherent limitations, including concerns with bias and shared measure variance. Our measure of intellectual disability, based only on parent report, may be especially flawed given the young age of the participants. Future research using standardized assessments, observational measures and/or multiple reporters would help increase our confidence and understanding of these topics. Further, our selection of potential correlates of ED in this study was limited to what was available in the EDI-YC psychometric validation study. Future research should consider other variables, not available in the current dataset, that may explain additional variability in ED for autistic children and provide a more nuanced understandings of our findings (e.g., sensory processing, executive functioning, family access to resources, experiences of discrimination). Because all data was collected at a single time point, conclusions on the directionality or causality of association cannot be made without further longitudinal research.

Finally, it is important to note that there were numerous sociodemographic differences between our samples with and without an autism diagnosis, and these two samples were largely recruited through different sources. While we controlled statistically for many of these variables, it is likely that the groups differed in ways that were not captured (or not captured adequately) by variables available in the present study. For example, we did not have a high-quality measure of developmental ability, which likely differed between groups. Thus, it will be important for our findings of group differences to be replicated in more closely matched samples. In addition, while about one third of participants identified with a minoritized racial or ethnic identity, the sample was not representative of the United States population. It will be important for future work to strive for more racial and ethnic diversity.

Conclusions and clinical implications

While more work is needed to establish the prevalence and severity of emotion dysregulation challenges in young autistic children, this study represents an important first step. Findings highlight ED challenges in a large, community-based sample of young children with and without autism and developmental concerns and also identify important potential correlates of ED in autism that have previously been underappreciated in clinical practice. In doing so, we draw attention to a critical need for early screening and intervention targeting ED in young autistic children. The EDI-YC (Day, Mazefsky, et al., 2023) is a brief, valid and sensitive measure of ED in young children and represents an excellent candidate for early screening of ED in pediatricians’ offices and as part of autism diagnostic assessments. Identifying autistic children with clinically significant ED at the time of diagnosis could help families identify supports that will be most beneficial to their child. Perhaps more critically, interventions developed specifically to target ED in autistic children, considering the unique needs of this population, are lacking, and urgently needed. Current “best” practice for autistic children recommends intensive early intervention focused on improving skills. Typically, these interventions do not target emotional challenges (or related challenges, like sleep) nor do they take family circumstances or parent mental health into account. Considering accumulating evidence for markedly high rates of mental health challenges and suicidality in older autistic individuals (Hand et al., 2019; Hossain et al., 2020; O’Halloran et al., 2022), addressing ED early should be a top priority for research and clinical practice.

DATA AVAILABILITY STATEMENT

Approved researchers can obtain the SPARK population dataset described in this study by applying at https://base.sfari.org. The complete dataset will also be made publicly available through the NIMH archive.