Utility of the Modified Anxiety Dimensional Observation Scale in Autistic Preschoolers with Varying Intellectual Functioning

Abstract

Objective:

Co-occurring anxiety affects 40–80% of autistic individuals; however, little is understood about how anxiety manifests in young autistic children, especially those with intellectual disability (ID), partly due to the paucity of measures designed to assess anxiety symptoms in this population. The present study examined the utility of the Modified Anxiety Dimensional Observation Scale (M-Anx-DOS), an observational measure of anxiety-related behaviors, in preschool-aged autistic children with and without ID.

Method:

This study included 48 autistic children (M age = 43.96 months; 81.3% with ID) and 30 non-autistic (NA) controls (M age = 43.66 months). Anxiety-related behaviors were measured during the M-Anx-DOS. Parent-reported anxiety symptoms were assessed via the Preschool Anxiety Scale-Revised (PAS-R).

Results:

Groups exhibited comparable scores on both the M-Anx-DOS and PAS-R. Within the autism group, a subset of M-Anx-DOS scores were related to age, autistic features, or IQ. The M-Anx-DOS exhibited excellent inter-rater reliability and acceptable internal consistency. Convergent validity was promising, with specific M-Anx-DOS scores correlated with parent-reported social, separation, and overall anxiety symptoms. M-Anx-DOS scores were not correlated with parent-reported ADHD or externalizing symptoms, suggesting strong discriminant validity.

Conclusions:

This study provides preliminary evidence of the reliability and validity of the M-Anx-DOS. These findings are promising given the importance of observational measurement of anxiety and lack of existing measures for this critical developmental period. Given the sample size and the complexity of identifying prodromal signs of anxiety in young autistic preschoolers with ID, future longitudinal work is essential to replicate and extend this work.

Autism spectrum disorder is a neurodevelopmental condition present in 1 in 36 children (Maenner et al., 2023). Co-occurring psychiatric conditions are common in autism, with anxiety among the most prevalent, affecting ~40% of autistic youth (van Steensel et al., 2011). Left untreated, anxiety symptoms contribute to increased behavioral challenges and disruption to family life as well as reduced independence and quality of life (B. Y. Lau et al., 2019; van Steensel et al., 2012). Thus, scientists and stakeholders (e.g., autistic self-advocates) argue that research advancing the assessment and treatment of anxiety is a critical priority (Benevides et al., 2020; Vasa et al., 2018).

Mounting evidence indicates that anxiety symptoms emerge early in childhood in autism. A scoping review of 44 studies on young autistic children reported rates ranging from 1.6% to 62% (Vasa et al., 2020), and more recent studies found that 40–65% of autistic preschoolers met diagnostic criteria for an anxiety disorder (Chan et al., 2023; Vasa et al., 2023). Research in non-autistic (NA) children has shown that targeted intervention during early childhood is effective at mitigating both short- and long-term anxiety symptoms (Kennedy et al., 2009; Rapee, 2013), and there is preliminary evidence of effectiveness in autistic preschoolers as well (Bischof et al., 2018). However, to intervene early in childhood, we must be able to accurately identify prodromal anxiety symptoms early in childhood, which can be especially challenging in autistic children with co-occurring intellectual disability (ID). Thus, this study aimed to examine the utility of an observational measure of anxiety-related behaviors in preschool-aged autistic children with varied intellectual functioning.

Measurement of Anxiety in Young Autistic Children

Historically, the most widely used measures of anxiety symptoms in young autistic children have relied upon parent report (i.e., questionnaires, clinical interviews). For example, of the 44 studies included in Vasa and colleagues’ scoping review (Vasa et al., 2020) only one utilized a direct observational measure (Scherr et al., 2017), whereas all other studies relied upon parent report. Parent report provides valuable information about anxiety-related behaviors and symptoms and should routinely be considered; however, there are several limitations associated with an over-reliance upon parent-report measures, especially questionnaires, when assessing anxiety in young autistic children. First, most parent-report questionnaires are “trait-based,” using a scale that broadly measures average symptom frequency (e.g., never, sometimes, often) during a specific preceding window of time (e.g., one week, three months). This type of scale can obscure the impact of infrequent but severe symptoms and may not take into account the extent to which the family accommodates the child’s anxiety by avoiding anxiety-eliciting situations. Second, with the exception of the Parent-Rated Anxiety Scale for Youth with Autism Spectrum Disorder (PRAS-ASD; Scahill et al., 2019) and Anxiety Scale for Children – Autism Spectrum Disorder (ASC-ASD; Rodgers et al., 2016), most parent-report questionnaires were not developed for use with autistic children, and few have been validated post-hoc in autistic samples. As a result, most parent-report questionnaires measure only prototypical anxiety symptoms aligned with the Diagnostic and Statistical Manual of Mental Disorders (DSM-5; American Psychiatric Association, 2013), ignoring nuanced or “distinct” autism-specific anxiety symptoms that are common in autistic children, particularly those with co-occurring ID (Kerns et al., 2021; Klein et al., 2023). In fact, recent evidence suggests that commonly used parent-report anxiety questionnaires designed for NA children have poor sensitivity in autistic youth with co-occurring ID (Kerns et al., 2021). Third, many questionnaires are not designed for preschool-aged children or include only a broad anxiety symptom scale that does not provide information about specific anxiety disorders, with one exception being the Preschool Anxiety Scale in its original and revised formats (PAS-R; Edwards et al., 2010; PAS; Spence et al., 2001). Distinguishing between anxiety disorders is especially important in preschool-aged children, as research on NA children has established that certain anxiety disorders (e.g., specific phobia) are relatively common during early childhood and may either resolve or develop into other anxiety subtypes later in development (De Vries et al., 2019; Paulus et al., 2015; Steinsbekk et al., 2022). Furthermore, different anxiety disorders may represent at least partially distinct etiological influences (Dougherty et al., 2013; Eley et al., 2003, 2008; Gouze et al., 2022; Lau et al., 2007; Sandstrom et al., 2020) and may require disorder-specific targeted intervention for optimal treatment response.

The limitations of existing parent-report anxiety measures are especially problematic for autistic children with ID, who comprise ~30% of autistic children (Shenouda et al., 2023). These children are often unable to describe their internal experiences and anxiety symptoms to caregivers or clinicians (Gilmore et al., 2022). Thus, they are particularly vulnerable to diagnostic over-shadowing, which results from parents and clinicians misattributing anxiety-related behaviors to the child’s primary diagnoses of autism and ID instead of recognizing them as behavioral manifestations of anxiety (Tarver et al., 2021). Due to these inherent challenges, many prior studies have excluded autistic children with ID. Thus, there is little understanding of how, when, and why anxiety manifests or how to accurately identify anxiety in young autistic children with ID (Appleton et al., 2019).

Given these challenges, experts have begun to emphasize that a multi-method, multi-source approach that integrates direct observation with other sources (e.g., parent report, clinical diagnostic interview) is essential when assessing anxiety in autistic children with ID (Winder-Patel et al., 2022). Observational measures are excellent complements to parent and clinician ratings because they: 1) allow researchers to determine symptom presence and severity rather than relying solely on informant report, 2) increase ecological validity with naturalistic or semi-naturalistic contexts, 3) reduce measurement bias, as raters can be kept naïve to study hypotheses and subject information (e.g., diagnostic history), 4) are amenable to different situational presses, which can be altered based on the constructs of interest, and 5) can serve as sensitive indicators of treatment responsiveness. Furthermore, observational measures can be used to assess subtler, prodromal signs of anxiety in young children who have not yet developed “full-blown” anxiety symptoms that exceed a diagnostic threshold. It is well established that in very young children, fearful reactions to ordinary, developmentally appropriate experiences (e.g., brief separation from a caregiver, approach by an unfamiliar adult, introduction to a novel object) are robust predictors of later anxiety disorder diagnoses (Sandstrom et al., 2020). In other words, observational measures designed to mimic everyday experiences can provide evidence of elevated early risk for anxiety disorders, even when anxiety symptoms have not yet reached a diagnostic threshold that would be captured by parent-report measures. Such evidence enables earlier identification of high-risk children and the implementation of preventative intervention before anxiety symptoms become fully entrenched. To our knowledge, the only direct observational measure for anxiety in autism is the Observation Schedule for Children with autism – Anxiety, Behaviour, and Parenting (OSCA-ABP; Palmer et al., 2021). This measure is not anxiety-specific, however, as it assesses additional child behaviors (e.g., non-compliance, aggression) as well as parent behaviors. Thus, there remains a critical need for an observational measure that can assess anxiety-related behaviors in autistic preschoolers with varying intellectual functioning.

The Anxiety Dimensional Observation Schedule (Anx-DOS)

The Anxiety Dimensional Observation Schedule (Anx-DOS) is a multidimensional observational measure of anxiety in young children (Mian et al., 2015). The Anx-DOS includes several semi-naturalistic “presses” that are designed to elicit different dimensions of anxiety-related behaviors, common in early childhood, including fear, avoidance, separation distress, and caregiver proximity seeking. These presses were designed to mimic developmentally appropriate experiences and daily interactions that might elicit mild-to-moderate fear responses, which signal elevated risk for later anxiety diagnoses. Specific presses include (a) a remote-controlled spider, which elicits fear in response to novelty; (b) an auditory startle, which elicits the startle reflex; (c) a mystery jar, which elicits fear of novelty and uncertainty; (d) parental separation, which elicits fear related to separation from a caregiver. The Anx-DOS coding system captures a range of anxiety-related behaviors as well as coping responses (e.g., seeking proximity to caregiver). This breadth captures greater symptom dimensionality, in comparison to other existing measures, such as the OSCA-ABP (Palmer et al., 2021) which only contains two anxiety-related behaviors – avoidance and reassurance seeking. This is especially important because young children express fear across different contexts using a wide and varying repertoire of behaviors (e.g., verbal behaviors, physical behaviors, facial expressions). Thus, the multidimensionality of the Anx-DOS enables direct observation of the nuanced expression of anxiety-related behaviors most common in young children. The Anx-DOS garnered initial evidence of reliability and validity in a 3-to-6-year-old sample enriched for trauma exposure (Mian et al., 2015). Preliminary convergent validity was established with parent-reported separation anxiety and specific phobia symptoms. Inter-rater reliability was also acceptable, with intraclass correlations ≥ 0.65 for all dimensions except hypervigilance.

Present Study: The Modified Anxiety Dimensional Observation Schedule

Though the development of the Anx-DOS addressed the need for a multidimensional, observational measure of anxiety-related behaviors in preschoolers, this measure was not designed or validated for use in autistic children. It is critically important that observational measures are independently evaluated or even modified as needed for use in autistic children (Palmer et al., 2021). In collaboration with the authors of the original Anx-DOS (Mian et al., 2015), we modified the measure to make it more appropriate for autistic children with and without ID. We chose the Anx-DOS because of its dimensionality, its continuous assessment of anxiety-related behaviors, and its variety of developmentally appropriate contexts. The modifications included (a) simplifying verbal instructions, (b) emphasizing interactions with an examiner instead of the caregiver, and (c) including a social fear press given the prevalence of social fear and anxiety in autistic children. This press was designed to capture fear in response to the approach of an unfamiliar adult, which is an early behavioral indicator of social anxiety. Quantifying social fear is especially important in young autistic children and children with ID who may not understand the concept of social evaluation and therefore may never meet DSM-5 diagnostic criteria for social anxiety despite experiencing impairing levels of social fear (Kerns et al., 2014). The objectives of the present study were to (a) determine whether autistic preschoolers with varying intellectual abilities exhibit more anxiety-related behaviors than their NA peers; (b) investigate whether anxiety-related behaviors are associated with individual differences that represent common sources of heterogeneity in autism (e.g., age, IQ, autistic features); and (c) examine the reliability and validity of the M-Anx-DOS in autistic preschoolers with varying intellectual abilities.

Method

Participants

This study included 48 autistic children (41 male; M(SD) age = 43.96(7.12) months). A comparison group of 30 NA children (23 male; (M(SD) age = 43.66(11.84) months) was included in a subset of analyses. Subjects were enrolled in a larger longitudinal study of anxiety in preschool-aged children with ID (R01MH107573; PI: Roberts). For the present study, data were included from the first timepoint at which children completed the M-Anx-DOS. Groups were age-matched, t(76) = −0.14, p = .888, with an age range from 33.94 to 71.94 months, (M(SD) = 48.32(9.09) months). The autism group was enriched for co-occurring ID through targeted recruitment to achieve the aims of the larger longitudinal study. Thus, in the present sample, 81.3% (n = 39) of autistic children had an estimated IQ <70, and the groups differed significantly on estimated IQ, t(76) = 14.47, p < .001. Despite the preponderance of ID, the range of estimated IQ in the autism group (49–108) was representative of the range commonly observed in young autistic children. Table 1 depicts participant characteristics.

Table 1.

Participant characteristics

Variable	Autism (n=48)	NA (n=30)	Group Comparison
Age (months)
M(SD)	43.96 (7.12)	43.66 (11.84)	t(76) = −0.14, p = .888
Range	33.94–68.79	34.69–71.94
IQ
M(SD)	57.69 (13.59)	105.50 (15.14)	t(76) = 14.47, p < .001
Min-Max	49–108	76–133
ADOS-2 CSS
M(SD)	7.60 (1.63)	2.03 (1.47)	t(76) = 15.20, p < .001
Min-Max	4–10	1–6
Race, %			X2(5) = 3.22, p = .665
White	77.08%	86.67%
Black or African American	10.42%	6.67%
Hispanic or Latino	2.08%	0%
Asian	2.08%	0%
More than one race	4.17%	6.67%
Not reported	4.17%	0%
Ethnicity, %			X2(2) = 2.04, p = .361
Hispanic or Latino	2.08%	3.33%
Not Hispanic or Latino	91.67%	96.67%
Not reported	6.25%	0%
Maternal Education, %			X2(6) = 11.09, p = .086
Less than high school diploma	2.08%	0%
High school diploma	10.42%	0%
Some college, technical, or associates degree	15.3%	16.67%
Bachelors degree	39.58%	26.67%
Some graduate	4.16%	13.33%
Graduate or advanced degree	22.92%	43.33%
Not reported	6.25%	0%
Annual Household Income, %			X2(3) = 5.28, p = .152
<$50,000	25.00%	20.00%
$50,000-$90,000	16.67%	33.33%
>$90,000	41.67%	43.33%
Not reported	16.67%	3.33%

Participants were recruited statewide through local service providers, social media, and pediatricians. Autistic children were required to have a prior clinical diagnosis of autism, documented via clinical report shared by their parents and confirmed during study participation via well-established clinical best estimate (CBE) diagnostic procedures (Hogan et al., 2017; Lord, Petkova, et al., 2012; Roberts et al., 2020). This CBE procedure incorporated information from the Autism Diagnostic Observation Schedule – Second Edition (ADOS-2; Lord, Rutter, et al., 2012), Autism Diagnostic Interview-Revised (ADI-R; Rutter et al., 2003), developmental and adaptive assessments, behavioral observations, and clinician judgment to determine whether children met DSM-5 criteria for autism spectrum disorder (ASD). Non-autistic controls were required to have no clinical diagnosis of ASD or developmental delay, also confirmed through the study via CBE diagnostic procedures, as well as no first- or second-degree relatives with ASD or related disorders. Presence of ID, defined as an estimated IQ <70, was an additional exclusionary criterion for the NA children. Even with this criterion a representative range of IQ scores (76–133) was observed within the NA group. Exclusionary criteria for all children included <37 weeks gestation, seizure disorder, or confirmed genetic disorder.

Procedures

Participants were assessed in their home or at the research laboratory by two trained examiners. Child behaviors did not significantly differ by location (i.e., home or research laboratory) for any behaviors coded (ts(77) ≤ 0.62, ps ≥ .536). Children completed observational assessments of developmental ability, autistic features, and the M-Anx-DOS; parents completed clinical interviews and rating forms regarding their child’s behavior. The M-Anx-DOS was coded from video. The Institutional Review Board approved the procedures, and written informed consent was obtained from parents prior to data collection. Study participants received monetary compensation for their participation.

Measures

Modified Anxiety Dimensional Observation Schedule (M-Anx-DOS).

The four presses from the original Anx-DOS were administered: Spider, Auditory Startle, Mystery Jar, and Maternal Separation. A Stranger Approach press was added to provide an opportunity to observe behavioral responses associated with social fear. The Stranger Approach press has been used in prior studies to characterize social fear in young autistic children and children with ID (Macari et al., 2021; Nuske et al., 2017; Scherr et al., 2017). The Spider and Stranger Approach presses were adapted to mimic procedures from the Laboratory Temperament Assessment Battery (Lab-TAB; Goldsmith & Rothbart, 1988), enabling continuity of data collection between this study and a previous longitudinal study on infant temperament in neurodevelopmental disorders (R01MH090194, PI: Roberts). The child’s primary caregiver was present for all presses of the M-Anx-DOS.

M-Anx-DOS Administration.

During the Spider press, an examiner presented the child with a rubber toy spider that was six inches in diameter and prompted the child to pet the spider. If the child’s hand came within approximately two inches of the spider, the examiner activated the spider to vibrate by pulling a hidden cord attached to the spider. The examiner repeated this process a total of four times. The Spider press lasted approximately 1–1.5 min.

In the Auditory Startle press, the children watched a silent, animated film. The child’s distance from the computer and speaker were standardized. One minute into the film, a single 200 ms broadband burst at 100 dB was presented. The Auditory Startle press lasted 2.5 min total. For a subset of participants tested at the beginning of the study (autism: n = 5; TD: n = 14), an initial version of the Auditory Startle press was administered with the film’s soundtrack audible throughout the press but otherwise identical parameters. Review of these initial data indicated no behavioral response in most participants; therefore, the soundtrack was silenced to elevate the likelihood of eliciting a startle response. The Auditory Startle press score was significantly higher for the revised version than the initial version, t(73) = 3.46, p < .001. For the participants given the initial version at Time 1, their Auditory Startle data were imputed through the next observation carried backwards (NOCB) using their second M-Anx-DOS assessment, collected one year later (i.e., Time 2). The NOCB imputation is an appropriate method for minimizing selection bias and loss of representation and power in studies with time series data (Engels & Diehr, 2003; Oluwaseye et al., 2022). In the children who received the initial version at Time 1 and revised version at Time 2, paired samples t-tests indicate that for most codes, the Auditory Startle press score was greater at Time 2, t(16) = 4.66, p < .001. In contrast, for children who received the revised version at both Time 1 and Time 2, scores were not different across exposures, t(41) = 0.48, p = .632. These preliminary analyses confirmed that (a) the revised version was more effective at eliciting anxiety-related behaviors than the initial version, and (b) scores were stable from Time 1 to Time 2 in children who received the revised version at both timepoints. Once imputed, the Auditory Startle press score did not differ between the children with imputed data and those without, t(71) = 1.02, p = .311. These findings combined with evidence from existing work confirming one-year stability in auditory startle responses for both autistic and NA children (Takahashi et al., 2017), support the use of NOCB imputation in this circumstance.

During the Mystery Jar press, the examiner presented the child with an empty, opaque, black jar that was held at or above the child’s eye level to ensure that the child could not see inside. The examiner delivered up to three prompts at 15 second intervals for the child to put their hand inside the jar. Once the child either put their hand in the jar or if the child did not stick their hand into the jar after the third prompt, the jar was turned upside down and revealed to be empty. The Mystery Jar press was approximately 10 seconds to one minute in duration.

For the Maternal Separation press, the mother and child played together with a standard set of toys (sticker pad, book, ball, peg game, puppets, etc.) for three minutes. Following a prompt from the examiner directing the mother to step outside, the mother stood up and instructed the child to stay where they were and continue playing. If the child protested, the mother was allowed to repeat “Stay here and play, I’ll be right back.” The examiner and the mother exited the room and remained out of the child’s sight for 30 seconds while the child was left with a second examiner. Once the mother returned, the mother and child were encouraged to continue playing for another 30 seconds. The Maternal Separation press lasted approximately five minutes.

During the Stranger Approach press, a female examiner dressed in a baseball cap, dark sunglasses, an oversized gray sweatshirt, and a long black skirt entered the room and slowly approached the child. The examiner’s outfit was selected to disguise the examiner, standardize the press, and allow for a variety of examiner body shapes and sizes. The examiner knelt approximately two feet away from the child for two minutes while maintaining a neutral expression, then stood and exited the room. The Stranger press lasted approximately three minutes.

M-Anx-DOS Coding.

The M-Anx-DOS was videotaped and coded by trained research assistants, naïve to the diagnostic status of the children and study hypotheses. All coders established initial reliability with a master coder (KS) by obtaining ≥80% item-level agreement on three consecutive videos. Training was fairly efficient, requiring approximately 10–15 hours per coder. To assess ongoing reliability in the present sample, 20% of coded videos were randomly selected to be coded by the master coder. Inter-rater reliability is reported in Results. The M-Anx-DOS coding scheme consisted of six codes (see Table 2): Fear Arousal, Physical Avoidance, Exaggerated Startle, Proximity Seeking, Separation Distress, and Global Rating. Codes take behavior severity and duration into account and are generally rated as: 0-No evidence or no concern, 1-Mild or normative, 2-Of concern, and 3-Atypical. Given the complexities associated with coding anxiety-related behaviors in autistic children, coders were instructed to only code behaviors that appeared to be in direct response to the press. Furthermore, coders were asked to establish the child’s baseline behavioral pattern by watching neutral periods of the M-Anx-DOS (e.g., playtime during the Maternal Separation press, transition periods in between presses). Establishing the child’s baseline pattern of behavior helped to minimize the misattribution of certain autistic behaviors as anxiety-related behaviors. Potentially idiosyncratic behaviors of anxiety in autistic children, such as repetitive behaviors or self-stimulatory vocalizations, were only coded as anxiety-related behaviors if these behaviors increased in frequency or intensity in response to a specific press.

Table 2.

Behavioral code descriptions for M-Anx-DOS

Behavior	Presses	Code	Behavioral Description
Fear Arousal	SP, AS, MJ, ST	0	No signs of facial fear; no signs of postural fear; vocalizations and/or gestures do not indicate fear
		1	Brief mild facial fear (limited to one region of the face); mild postural fear (decrease in activity level, holding breath); mild expressions of fear (whimpering/protesting, occasional hesitant or inquisitive vocalizations, shake head no or sign “All done”)
		2	Moderate facial fear (limited to two regions of the face) or prolonged facial fear in one region of the face; moderate postural fear (visible and sustained tensing of the muscles); moderate expressions of verbal fear (tone sounds more fearful, pitch of voice increases, voice may sound more tense, gestures may appear more frequent and vehement)
		3	Facial fear in three regions of the face or prolonged facial fear in two regions of the face; strong postural fear (entire body tenses, trembling due to muscle strain); verbal expressions of fear are severe (yelling, crying, screaming, fearful interjections such as “I’m scared”)
Physical Avoidance	SP, AS, MJ, ST	0	May be slow to approach, but no indication of avoidance
		1	Turns or leans away, distracts themselves with nearby objects, momentarily backing away
		2	Moves away, arches back, pushes stimulus away, actively twists/turns away, takes a few steps back without reproaching, hides face/body but occasionally peeks
		3	Bolting/running away, jumps away, hides without peeking, maximum effort child can put into escaping
Exaggerated Startle	SP, AS	0	Does not touch spider; Does not react to startle
		1	Blinks, slight jump or twitch
		2	Closes eyes tightly, lowers brows, stretches lips tensely, moves head back, mild jump, gasps, short and often non-fearful vocalizations such as “That scared me”
		3	Increase in rate of eye blinks, eyes widen, brows are raised, mouth opens, may ask “What was that?” or make other short verbalizations
		4	Longer or louder gasp or scream, thrusts head forward, covers ears, jumps back in chair, flexes shoulders, jumps out of chair, moves away from spider/speaker. loud and fearful vocalizations, cling to mom
Proximity Seeking	SP, MS, ST	0	No indication or attempt to get closer to mother
		1	Looks to or touches mother, leans toward mother, uses mother to hide, hold/manipulates mother’s hands
		2	Moves to sit next to mother, gets up to look for mother, gets up to use mother to hide, buries face in mother
		3	Holding or clinging to mother, runs to find mother
Separation Distress	MS	0	No indication of distress
		1	Asking mother not to leave, whimpering or protesting, nervous or hesitant tone, brief facial fear as mother leaves, asks about mother’s whereabouts
		2	Reports physical ailment, self-soothing behaviors, tone is fearful and tense, facial fear extends beyond initial departure
		3	Crying, screaming, facial fear lasts most of press
Global Rating	SP, AS, MJ, MS, ST	0	No evidence or concern of elevated fear or anxiety
		1	Mild or normative behaviors indication fear or anxiety
		2	Fearful or anxious behaviors are clinically concerning
		3	Fearful or anxious behaviors or atypical and interfere with performance in press

Note. Facial fear was coded based on fear displayed in any of the following three regions: forehead/brows, eyes/nose/cheek, and mouth/lips/chin. SP=Spider; AS=Auditory Startle; MJ=Mystery Jar; MS=Maternal Separation; ST=Stranger

Each press score was calculated by summing the individual code scores for that press. The M-Anx-DOS total score was computed by averaging the five press scores. The M-Anx-DOS total score can range from 0 to 11.80, with higher scores reflecting more severe anxiety-related behaviors.

Convergent Validity Measures.

Parent reported anxiety symptoms were measured using the Preschool Anxiety Scale-Revised (PAS-R; Edwards et al., 2010), a 28-item questionnaire designed for use in preschool-aged children. The PAS-R subscale and total raw scores were used to examine convergent validity of M-Anx-DOS scores. The PAS-R was selected because it focuses on early childhood and demonstrates good construct validity, satisfactory internal consistency, and good test-retest reliability in NA preschoolers (Edwards et al., 2010). While prior studies of young autistic children have utilized the PAS-R (Bischof et al., 2018; Charman et al., 2021; Palmer et al., 2023; Thurman & Dimachkie Nunnally, 2022), its psychometric properties in this population remain unknown, particularly for those with ID. Autism-specific anxiety measures, such as the PRAS-ASD (Scahill et al., 2019) and ASC-ASD (Rodgers et al., 2016), were developed for and validated in older children (PRAS-ASD: 5–17 years, ASC-ASD: 8–16 years), rendering them inappropriate for our preschool-aged sample. In the present sample, internal consistency was strong for the PAS-R total score and satisfactory for all subscales except Specific Fears (Table 3). The low internal consistency on Specific Fears is expected because each item in the subscale refers to a distinct stimulus (e.g., dogs, thunderstorms, heights), and children are unlikely to exhibit consistent fear responses across widely varying stimuli.

Table 3.

Internal consistency of PAS-R Subscale and Total Raw Scores

		Cronbach’s α
Scale	Number of items	Autism (n=42)	NA (n=29)	Combined (n=71)
Social Anxiety	7	.69	.73	.71
Separation Anxiety	5	.57	.71	.63
Specific Fears	9	.51	.41	.45
Generalized Anxiety	7	.72	.65	.71
Total	28	.82	.78	.81

Discriminant Validity Measure.

The Child Behavior Checklist 1.5–5 years (CBCL; (Achenbach & Rescorla, 2000) is a 99-item parent-report questionnaire that rates the emotional and behavioral functioning of 1.5- to 5-year-old children. Good reliability and validity are reported for the CBCL (Achenbach & Rescorla, 2000). The factor structure of the CBCL 1.5–5 has been confirmed in young autistic children (Pandolfi et al., 2009). For discriminant validity analyses, the Externalizing Problems and ADHD Problems raw scores were used. Both subscales have been used and validated in studies of autistic children (Hoffmann et al., 2016; Hong et al., 2021; Pandolfi et al., 2009; Vaillancourt et al., 2017).

Other Descriptive Measures.

The Early Learning Composite Score (ELC) of the Mullen Scales of Early Learning (MSEL; Mullen, 1995), a standardized measure of developmental abilities, was used to estimate IQ for children 68 months and younger (n = 75). The ELC comprises visual reception, receptive language, expressive language, and fine motor domains. For the three participants older than 68 months, the General Conceptual Ability (GCA) score of the Differential Abilities Scale – Second Edition (DAS-II; Elliott, 2007) was used to estimate IQ. Both measures have mean scores of 100 and standard deviations of 15. The Calibrated Severity Score from the ADOS-2 (Lord, Rutter, et al., 2012), a gold-standard assessment tool for autism, was utilized as a continuous measure of autistic features.

Analysis Plan

In preliminary analyses, we contrasted PAS-R anxiety symptoms between groups (autism v. NA). Using nonparametic Spearman correlations, PAS-R scores were not correlated with IQ or maternal education, r_ss ≤ ± .115, ps ≥ .349, despite group differences on those variables. Thus, independent samples t-tests were run to compare PAS-R scores between group with no covariates included.

To investigate whether autistic preschoolers exhibited more anxiety-related behaviors than their NA peers, we ran independent samples t-tests and report those for descriptive purposes. IQ and maternal education were evaluated as potential covariates. In the full sample, IQ was correlated with the Mystery Jar press score, r_s = −.48, p < .001, and maternal education was correlated with the M-Anx-DOS total score, r_s = .24, p = .036, and the Maternal Separation press score, r_s = .26, p = .022. Therefore, we utilized analyses of covariance (ANCOVAs) to investigate whether groups differed on M-Anx-DOS scores with IQ and maternal education entered as covariates in all models. To determine whether individual differences that represent common sources of autistic heterogeneity (e.g., age, IQ, autistic features) were related to observed anxiety-related behaviors, Spearman correlations were calculated within the autism group.

Reliability and validity analyses were run within the autism group. Average measures intraclass correlations (ICCs) were utilized to investigate inter-rater reliability for M-Anx-DOS scores. Cronbach’s alpha (unstandardized) was computed to examine the internal consistency of each M-Anx-DOS press. Convergent and discriminant validity were investigated within the autism group via correlations with measures we hypothesized would be related (i.e., anxiety symptoms) and unrelated (e.g., ADHD, externalizing symptoms). Because age, IQ, and autistic features were correlated with specific M-Anx-DOS scores, we used partial correlations to control for the influence of these variables in convergent and discriminant validity analyses.

Results

Preliminary Analyses: Parent-Reported Anxiety Symptoms

Table 4 depicts parent-reported anxiety symptoms in both groups. Independent samples t-tests indicated that scores were similar between groups for the Social Anxiety, Separation Anxiety, and Specific Fears subscales of the PAS-R, ps ≥ .186. Generalized anxiety subscale scores, t(67) = 1.96, p = .054, d = .48, and the PAS-R total score, t(66) = 1.77, p = .082, d = .44, were slightly elevated in the autism group.

Table 4.

Parent-Rated Anxiety Symptoms and M-Anx-DOS scores.

Variable	Autism (n=48)	NA (n=30)	Group Comparison
PAS-R Raw Scores, M(SD)
Total Score	21.07 (12.39)	16.16 (9.07)	t(66) = 1.77, p = .082
Subscale Scores
Social Anxiety	4.26 (3.92)	3.05 (3.40)	t(68) = 1.34, p = .186
Separation Anxiety	3.38 (3.15)	3.40 (3.01)	t(69) = 0.21, p = .983
Specific Fears	7.95 (4.73)	6.77 (3.59)	t(69) = 1.12, p = .267
Generalized Anxiety	5.22 (4.84)	3.32 (2.96)	t(67) = 1.96, p = .054
M-Anx-DOS Scores, M(SD)
Total Score	3.33 (1.42)	2.93 (1.55)	t(76) = 1.17, p = .244
Press Scores
Spider	4.88 (4.05)	5.24 (4.93)	t(75) = 0.35, p = .724
Auditory Startle	2.65 (1.98)	2.57 (1.46)	t(71) = 0.20, p = .842
Mystery Jar	2.49 (2.12)	0.90 (1.58)	t(75) = 3.53, p < .001
Maternal Separation	2.21 (2.90)	1.50 (2.10)	t(76) = 1.16, p = .249
Stranger Approach	4.75 (2.26)	4.60 (2.50)	t(76) = 0.27, p = .785

Anxiety-Related Behaviors in Autistic v. NA Children

Unadjusted M-Anx-DOS scores are reported in Table 4. In separate ANCOVAs, the effect of diagnostic group was investigated, with IQ and maternal education included as covariates. For the M-Anx-DOS total score, the omnibus model was significant, F(3,71) = 3.63, p = .017, ƞ² = .13. Group was not a significant predictor, F(1,71) = 0.00, p = .992, nor was IQ, F(1,71) = 1.88, p = .175, but maternal education significantly predicted M-Anx-DOS total score, F(1,71) = 6.49, p = .013, ƞ² = .08. B = .33. This was likely driven by the Maternal Separation press score, for which the omnibus model was significant, F(3,71) = 3.36, p = .023, ƞ² = .12, and maternal education was the only significant predictor, F(1,71) = 6.00, p = .017, ƞ² = .08, B = 0.56. Post-hoc partial correlations were run to determine whether the relationship between maternal education and the Maternal Separation press score (controlling for IQ) was similar in the two groups. The correlation was significant in the autism group, r = .37, p = .014, but not the NA group, r = .01, p = .954. However, when the ANCOVA was rerun with the group*maternal education interaction included in the model, the interaction was non-significant, F(1,71) = 2.34, p = .131. The omnibus model was also significant for the Mystery Jar press score, F(3,70) = 4.80, p = .004, ƞ² = .17, but no individual predictors were significant, ps ≥ .197. The omnibus models were non-significant for the following press scores: Spider, F(3,70) = 0.30, p = .829; Auditory Startle, F(3,67) = 0.30, p = .826; and Stranger Approach, F(3,71) = 1.61, p = .195.

Correlations Between Anxiety-Related Behaviors and Individual Differences Associated with Autistic Heterogeneity

Correlations between individual differences and M-Anx-DOS scores within the autism group are presented in full in Table 5. Older chronological age was associated with lower Mystery Jar press score, r_s = −.34, p = .020, and lower Maternal Separation press score, r_s = −.32, p = .027. Higher IQ was correlated with lower Mystery Jar press score, r_s = −.37, p = .011. Higher ADOS-2 Calibrated Severity Score (i.e., more autistic features) was correlated with lower Spider press score, r_s = −.31, p = .031.

Table 5.

Correlations between M-Anx-DOS scores and individual differences, parent-rated anxiety symptoms, and parent-rated ADHD/externalizing symptoms within the autism group.

	Individual Differences				PAS-R (Convergent)a				CBCL (Discriminant)a
	Age	IQ	ADOS-2 CSS	Social Anxiety	Separation Anxiety	Specific Fears	Generalized Anxiety	Total Score	Externalizing	ADHD
Total Score	-.20	-.09	-.01	.32 *	.22	.31	.23	.34 *	-.02	-.06
Press Scores
Spider	.16	.11	-.31 *	.45 **	.23	.32 *	.27	.41 *	.03	-.02
Auditory Startle	-.01	.06	.25	.03	-.03	.13	-.11	-.01	-.18	-.13
Mystery Jar	-.34 *	-.37 *	.27	.14	-.09	.17	.17	.13	.12	.12
Maternal Separation	-.32 *	-.15	.14	.24	.36 *	.36 *	.32	.40 *	.06	-.02
Stranger Approach	-.23	-.09	.10	-.19	-.10	-.23	-.18	-.22	-.14	-.11

Note:

^aPartial correlations controlling for chronological age, IQ, and ADOS-2 CSS;

^*p < .05;

^**p < .01

M-Anx-DOS Reliability and Validity

Within the autism group, inter-rater reliability was excellent for the M-Anx-DOS total score (ICC = .99) and press scores (ICCs = .96–1.00). All presses showed acceptable or good internal consistency, with Cronbach’s alphas ranging from .65-.90. Table 6 details the reliability results for the autism group, with the NA group and combined sample included for reference.

Table 6.

Inter-rater reliability and internal consistency for M-Anx-DOS presses

Press	Variables Included	Intraclass Correlations (ICCs)			Cronbach’s α
		Autism (n=10)	NA (n=7)	Combined (n=17)	Autism (n=48)	NA (n=30)	Combined (n=78)
Spider	FA, PA, ES, PS, GR	.97	1.00	.98	.90	.94	.92
Auditory Startle	FA, PA, ES, GR	.96	.94	.95	.73	.64	.70
Mystery Jar	FA, PA, GR	1.00	.78	.99	.84	.82	.86
Maternal Separation	PS, SD, GR	1.00	.89	.95	.90	.89	.89
Stranger Approach	FA, PA, PS, GR	.99	.89	.95	.65	.76	.68

Note. FA=Fear Arousal, PA=Physical Avoidance, ES=Exaggerated Startle, PS=Proximity Seeking, SD=Separation Distress, GR=Global Rating

Full convergent and discriminant validity results are detailed in Table 5. Within the autism group, the M-Anx-DOS total score was correlated with the PAS-R Social Anxiety subscale score, r_s = .32, p = .049, and the PAS-R total score, r_s = .34, p = .035. The Spider press score was correlated with the PAS-R Social Anxiety subscale score, r_s = .45, p = .004, the PAS-R Specific Fears subscale score, r_s = .32, p = .045, and the PAS-R total score, r_s = .41, p = .010. The Maternal Separation press score was correlated with the PAS-R Separation Anxiety subscale, r_s = .36, p = .025, and PAS-R total score, r_s = .40, p = .013. No correlations with discriminant validity measures were significant, r_ss ≤ ± .18, ps ≥ .317.

Discussion

The paucity of observational measures of anxiety is a significant barrier to understanding the manifestation of anxiety symptoms in young children with autism, especially those with ID. A better understanding of how anxiety symptoms are expressed behaviorally across early development in this population is critical for advancing early prevention and intervention efforts that may mitigate the long-term impacts of anxiety in autistic children. In this study, we modified a multidimensional observational measure of anxiety-related behaviors for use in preschool-aged autistic children and investigated whether autistic preschoolers displayed elevated anxiety-related behaviors on this measure. We also investigated whether individual differences were related to anxiety-related behaviors and examined the reliability and validity of the measure. Findings suggest that the M-Anx-DOS holds promise as an appropriate measure of anxiety-related behaviors in preschool-aged autistic children with and without co-occurring ID.

Our findings suggest that during the preschool developmental period, anxiety-related behaviors are expressed at similar frequency and severity in autistic and NA children, as M-Anx-DOS scores were similar between groups. This finding was consistent with the parent ratings of anxiety symptoms on the PAS-R, which were also largely comparable between the two groups, with margin. This consistency across the M-Anx-DOS and PAS-R is promising, given that experts have called for integrating both informant-report and direct observation in anxiety assessments for autistic children with co-occurring ID (Winder-Patel et al., 2022).

It was a bit surprising that the groups exhibited similar levels of anxiety, given the elevated anxiety symptoms and diagnoses in autistic preschoolers reported in previous studies (Chan et al., 2023; Klein et al., 2023; Vasa et al., 2020). It is possible that our sample of autistic preschoolers is unique in that it does not have elevated anxiety symptoms. Likewise, it is possible that our sample of NA preschoolers has more anxiety symptoms than would be expected in the general population. The NA sample was part of a larger longitudinal study that tracked development prospectively from infancy and provided feedback to parents about their child’s development along the way. The parents of the NA children may have been motivated to participate by their own anxiety about their infant’s development or because their infant was exhibiting temperament features that caused them concern (e.g., elevated negative affect), resulting in a sample inadvertently biased toward elevated risk for anxiety. In a study from our group with a sample that partially overlaps with that of the present study, 33% of the NA group met DSM-5 diagnostic criteria for an anxiety disorder (Klein et al., 2023), which is slightly higher than the rates reported in most studies (see Lawrence et al., 2019 for meta-analysis; 10–20%), but in line with a recent population-based study of NA preschoolers (Hussong et al., 2022: 35%). Longitudinal work is needed to determine whether autistic and NA preschoolers exhibit similar or divergent trajectories of anxiety symptoms and diagnoses across early and middle childhood.

Interestingly, higher maternal education was associated with greater anxiety-related behaviors during the Maternal Separation press. These findings appear to be driven by a positive correlation between maternal education and anxiety-related behaviors during maternal separation in the autism group specifically. These incidental findings are surprising given prior research suggesting that in NA children, higher maternal education is associated with more secure attachment in infancy (Scher & Mayseless, 2000), lower risk for separation anxiety in preschool (Battaglia et al., 2016), and better mental health later in development (Reiss et al., 2019). Given the relatively small sample size, it is possible that these findings would not be replicated and/or do not generalize to the broader population of autistic children. Future research should more systematically investigate the role that sociodemographic factors such as maternal education play in anxiety risk and resilience in young autistic children.

Within the autism group, some M-Anx-DOS scores were significantly correlated with individual differences that represent common sources of heterogeneity in autism. For example, greater autistic features were moderately associated with lower scores on the Spider press. It is possible that children with more autistic features do not find the spider used in the M-Anx-DOS threatening. There is some evidence that fear of spiders is established early in childhood at least partially through social learning (LoBue et al., 2010). It may be that autistic preschoolers with more autistic features are not picking up on social cues about what stimuli are expected to be perceived as potentially threatening (e.g., spiders), and thus do not develop or display fear of such creatures in early childhood. Future research should investigate whether more autistic features are associated with reduced fear to specific objects (e.g., potentially threatening animals) or just novel objects more broadly, as this knowledge has important implications for the potential mechanisms underlying the development of fear and anxiety in young autistic children. Older age was moderately associated with less anxiety-related behavior during the Maternal Separation press. This suggests that early maternal separation fears may dissipate with age and experience in autistic preschoolers. This trend is associated with lower risk for a later separation anxiety disorder diagnosis in NA children (Battaglia et al., 2016; Halpern et al., 1990). Longitudinal research investigating developmental change in behavioral responses to maternal separation in autistic children is needed to determine whether this is also true for this population. We also found that lower Mystery Jar scores were related to older age and higher IQ. Based on these findings and behaviors observed during the M-Anx-DOS, we believe that some of our participants, particularly those with cognitive and/or language delays, might not have understood what was being asked of them during this press. During the Mystery Jar press, children are asked to place their hand into an opaque jar. The press is meant to represent a novel situation, as fearful responses to novelty early in childhood are known to signal anxiety risk in neurotypical children (Sandstrom et al., 2020). However, it is likely that many young children have not experienced the specific verbal prompt previously, and thus children with greater language or cognitive delays may not have understood the unusual request. This lack of comprehension combined with the novelty of the activity may have contributed to a greater sense of uncertainty and thus more anxiety-related behaviors. In summary, while a small number of correlations between individual differences and M-Anx-DOS scores were evident, most M-Anx-DOS scores were independent of age, IQ and autistic features. This suggests that the M-Anx-DOS would be appropriate for a broad spectrum of autistic preschoolers.

Analyses of internal consistency and inter-rater reliability indicate that the M-Anx-DOS is a reliable measure. Internal consistency was adequate, suggesting that the behavioral codes are consistently measuring anxiety-related behaviors within each press. Inter-rater reliability was excellent for all presses and reliability for the total score was higher than that of the original Anx-DOS (Mian et al., 2015). Inter-rater reliability for the M-Anx-DOS was also higher than the only other known observational measure of anxiety and avoidance in autism, the OSCA-ABP (Palmer et al., 2021).

Preliminary convergent validity was supported by relationships between specific M-Anx-DOS scores and PAS-R scores. First, the Spider press and certain PAS-R scores (Social Anxiety, Specific Fears, and Total score) were moderately related, even when controlling for the effect of age, IQ, and autistic features. This suggests that fearful responses to novel objects early in childhood may be particularly informative as an indicator of multiple anxiety subtypes in autistic preschoolers. Prior research in NA children has consistently shown that greater fear in response to novel stimuli is highly predictive of later social anxiety disorder, generalized anxiety disorder, and specific phobia diagnoses (Sandstrom et al., 2020). In addition, higher Maternal Separation scores were correlated with elevated Separation Anxiety, Specific Fears, and Total scores on the PAS-R. These relationships suggest that anxious behavior during a caregiver separation task may also provide insight into multiple anxiety disorders including separation anxiety. These findings align with decades of research indicating that elevated distress during maternal separation signals insecure or ambivalent attachment which is a risk factor for concurrent and later emerging anxiety (Bar-Haim et al., 2007; Shamir-Essakow et al., 2005). The M-Anx-DOS total score was correlated with the PAS-R Social Anxiety and Separation Anxiety subscales as well as the Total score. These findings suggest that a holistic index of anxiety-related behaviors during the M-Anx-DOS aligns with overall anxiety symptoms but also indicates nuanced relationships among specific anxiety subtypes. Interestingly, the Auditory Startle, Mystery Jar, and Stranger Approach scores were not correlated with any PAS-R scores. It is possible that behaviors observed during these presses represent constructs that are not as directly associated with anxiety symptoms, at least in the manner they are captured by the PAS-R. For example, behaviors exhibited during the Auditory Startle may reflect auditory hyper-responsivity, which is common in autistic children with and without anxiety.

Discriminant validity was also evident, as M-Anx-DOS scores were not correlated with theoretically unrelated symptoms, including ADHD and externalizing symptoms, when controlling for the effects of age, IQ, and autistic features. When interpreted alongside the convergent validity results, these findings suggest that for the most part the M-Anx-DOS is eliciting anxiety-related behaviors and not behaviors associated with other conditions that commonly co-occur with autism (e.g., ADHD).

This study is not without limitations. First, we lack a control sample that has ID without autism, limiting the specificity of our findings to autism. Additionally, given the cross-sectional nature of this study, and because measurement was at an age when most anxiety symptoms have not yet manifested in their most severe form and anxiety-related impairment may not yet be evident, it was not possible to examine the predictive validity of the M-Anx-DOS on later anxiety diagnoses. Our sample was also moderate in size, and this constrained our ability to detect small effects and nuanced interactions. Thus, future directions should include a follow-up with a larger sample and a longitudinal design to determine how early anxiety-related behaviors observed on the M-Anx-DOS relate to later anxiety symptoms or diagnoses. Additionally, though the M-Anx-DOS is designed to capture behavioral responses to everyday experiences, the M-Anx-DOS provides only a momentary glimpse into anxiety-related behaviors. Thus, the measure does not capture behavioral responses across different settings that may elicit significant and impairing anxiety symptoms (e.g., school/daycare, medical appointments). Finally, it is important to acknowledge that the dearth of anxiety measures normed for autistic preschoolers limited the measures available for convergent validity analyses. Thus, while we believe that the convergent validity measures used in the present study represent the best available and most appropriate options, our conclusions should be interpreted with caution given the lack of psychometric evidence. Thus, we view this study as an important advancement of what is known about early and accurate identification of anxiety-related behaviors in autistic preschoolers including those with ID. The field would be significantly advanced by the establishment of psychometric properties for commonly used anxiety measures, and the future directions of this work include such studies in young autistic children with and without ID.

Despite these limitations, it appears that the M-Anx-DOS is a promising complement to the parent-report approach commonly utilized to assess anxiety in young autistic children. This is true even though the autism group did not exhibit elevated parent-reported anxiety symptoms, which highlights the utility of the M-Anx-DOS to capture the subclinical or prodromal anxiety-related behaviors common in this developmental period. Measures that can capture prodromal anxiety in very young children are critical for early identification and preventative intervention. Given the relative ease with which the M-Anx-DOS can be administered and coded, it may be a highly feasible observational measure for use in autistic preschoolers with varying cognitive abilities.

Data Availability Statement.

The data that support the findings of this study are available from the corresponding author (AH) upon reasonable request.