Symptoms of autism spectrum disorder and developmental delay in children with low mental age

Structured Abstract

Objective:

Autism Spectrum Disorder (ASD) in very young children with significant cognitive impairment is difficult to diagnose, depriving them of the earliest opportunities for autism-specific intervention. This study delineated specific symptoms in this group, compared to symptoms in children with Global Developmental Delay (GDD) and in ASD with milder developmental delays.

Method:

Autism Diagnostic Observation Schedule, 2^nd Edition, Toddler Module revealed symptoms in three groups of toddlers, with mean ages of 17–20 months: (1) ASD and cognitive/language functioning below the 12-month level (ASD-MA<12 mos; n=28), (2) GDD (n=27), and (3) ASD and cognitive/language functioning at or above the 12-month level (ASD–MA≥12 mos; n=29). Logistic regressionmodels were fit tocontrol for developmental level.

Results:

Items in all domains (social interaction, communication, repetitive movements) discriminated ASD-MA<12 mos from GDD. The two ASD groups, matched for age but differing on developmental level, showed strikingly similar ASD symptomatology.

Conclusion:

ADOS-2 symptoms differentiated ASD-MA<12 mos from GDD, after controlling for cognitive impairment. Symptoms in the two ASD groups were minimally related todevelopmental level.The ADOS-2 Toddler Module successfully captured ASD symptomatology even in children whose developmental level was below the recommended ADOS-2 cutoff of12 months, which may increase their access to early ASD-specific intervention.

Autism spectrum disorder (ASD) occurs across the intellectual spectrum(Charman et al., 2011), with stable diagnoses reported as young as 14–15 months (Pierce et al., 2019). The most severe presentation of ASD may occur in very young, very cognitively impaired children (Maenner et al., 2020; Mayes and Calhoun, 2011).These children may also be difficult to distinguish from children withother neurodevelopmental disorders, such as global developmental delay (GDD;Gardner et al., 2018; Thurm et al., 2019).

Two challengesin particular present when attempting to diagnose ASD in children who are both very young and cognitively impaired. The first is that children who are developmentally functioning below 12 months cannot be expected to show behaviors that onlyappear in typical development in the second year of lifeor later. For example, if a child does not display pretend play, which typically develops during the second year, this may be attributable to their developmental level and not to ASD. This suggests that social impairment in this group should be measured against social behaviors typically shown in the first year of life, such as eye contact, social smiling, simple imitation, and response to their name.

The second challenge in diagnosing ASD in children who are very young and significantly cognitively delayed is that many individuals with GDD without ASD show sensory and motor behaviors that can be difficult to differentiate from those in ASD, especially very early in development. Behaviors that were found to be of equal or higher frequency in ASD and other developmental disorders (DD) included stereotyped play, unusual posturing, hypotonia, long visual fixations, lack of animation in facial expressions, delayed motor milestones, sensory aversions, and sensory seeking (Baranek, 1999; Baranek et al., 2007; Desombre et al., 2006; Donkers et al., 2020; Provost et al., 2007).

In addition to overall ASD severity, specific ASD symptoms may be associated with lowercognitive functioning. For example, repetitive hand and finger movements, and moving, spinning or fidgeting with objectsappear more prevalent in ASD children with lower IQ, whereas circumscribed interests appear associated with higher IQ (Bishop et al., 2006). Such findings are well summarized by the dichotomy between “lower” order repetitive behaviors, such as stereotyped movements and repetitive manipulation of objects, and “higher” order repetitive behavior, including circumscribed interests and resistance to changes in routines and environments (Turner, 1999). Higher order symptoms presume awareness of routines and details of their surroundings as well as understanding some phenomena that can be the focus of unusual, circumscribed interests (train timetables, electric blanket controls) and that require more abstract cognitive ability than the lower order symptoms.

Young children with ASD and significant cognitive impairment present particularly daunting diagnostic challenges because their developmental ages are especially low. The most commonly used observational instrument for ASD in toddlers, the Toddler Module of the Autism Diagnostic Observation Schedule, 2^nd Edition(ADOS-2; Lord et al., 2012)is not validated for children with developmental levels, or mental ages (MA), below 12 months. This group of severely delayed children with signs of ASD has not been systematically described, and the stability of an ASD diagnosis for children this delayed group remains in doubt.

We recently studied a group of such children (Hinnebusch et al., 2017; Miller et al., 2019)who wereon average chronologically 25 months of age(Hinnebusch et al.: M=25.5 months, range: 15.7–36.7; Miller et al.: M= 25.92 months, range not provided in paper),were diagnosed with ASD,and who hadno cognitive or language skills at or above a 12-month level (ASD-MA<12 mos).Three important findings emerged. First, children in the ASD-MA<12 mosmainly differed from typically developing (TD) children in social communication. Notably, the ASD children with MA below 12 months, as well as a group of ASD children with higher levels of cognitive and language skills,displayed impaired social behaviors that usually appear in the first year of typical development, such as eye contact, social smiling, and response to name (Miller et al., 2019). Second, the diagnosis of ASD wasstable for children in thisASD-MA<12 mos group; after two years, 96% of these children continued to fully meet criteria for ASD, suggesting that their early ASD symptoms were not due solely to low cognitive functioning(Hinnebusch et al., 2017). Finally, the ASD-MA<12 mos group showed the least amount of progress on cognitive, social, and language skills after two years of intervention, suggesting that this group may require specific accommodations in the type and intensity of intervention that they receive (Hinnebusch et al., 2017).

The American Academy of Pediatrics recommends that children with identified early delays receive intensive early intervention (Hyman et al., 2020; Lipkin et al., 2020); the majority of children diagnosed with GDD go on to receive a diagnosis of intellectual disability (ID) when they are school aged (Guralnick, 2017; Moeschler & Shevell, 2014). This suggests that these children are either not receiving the recommended intensive intervention, that they need a different form of intervention, or that gains in overall cognitive functioning in response to early intervention are somewhat limited in this group.If diagnoses of ASD in children who are developmentally functioning below 12 months are both clinically possible and stable, then those children should be able to receive more timely access to early intervention.Evidence for how best to differentiate ASD with cognitive impairment from GDD in very young children will allow further studies on the most effective intervention for each group, andis one goal of the current study.

We are cognizant of the precision lost when the MA variable is dichotomized (as done here), rather than treated as a continuous measure that could, for example, be correlated with presence of individual symptoms. However, the general exclusion of children with MA below 12 months from a valid administration of the most commonly used diagnostic measure (ADOS-2 Toddler Module) should be challenged, if these children have ASD symptomatology which is stable and severe, as shown by Hinnebusch et al.(2017) and Miller et al. (2019).

Findingsfrom these first two studies of children in the ASD-MA<12 mos group (Hinnebusch et al, 2017; Miller et al, 2019) have somewhat limited generalizability for three reasons. First, diagnostic criteria in the current DSM-5 (American Psychiatric Association, 2013)are substantially different from DSM-IV, which was used in these earlier studies(Hinnebusch et al., 2017; Miller et al., 2019). Second,Hinnebusch et al. (2017) and Miller et al. (2019) used an earlier version of the ADOS (Lord et al., 2000), which has since been updated tothe ADOS-2and now includes a Toddler Module(Gotham et al., 2007;Luyster et al., 2009). Although the Toddler Module still excludes children with developmental levels below 12 months, it better discriminates ASD from other developmental delays in very young children. Third, the average chronological ageof children in the earlier studies, as mentioned above,was 25 months. Given the emphasis on beginning intervention before the second birthday if possible(MacDonald et al., 2014; Rogers et al., 2012), it is desirable to study theASD-MA<12 mosgroup at an even younger ageto see if ASD symptoms are sufficiently clear to warrant a diagnosis. Additionally, the ADOS-2 is accepted by many states, and required in some such as Connecticut, to support an ASD diagnosis before a child receives services.If ASD can be diagnosed in children with mental ages below 12 months using the ADOS-2 Toddler Module, then these children are more likely to receive earlier access to evidence-based interventions at younger ages, critical for learning and development(Clark et al., 2018; Klintwall et al., 2015; MacDonald et al., 2014; Vivanti et al., 2016).

The current project addresses these gaps in the literature by investigating whether the ADOS-2 Toddler Module (Lord et al., 2012) identifies symptoms of ASD in very young children (average chronological age: 19.1 months) in three groups (ASD-MA<12 mos, GDD, andASD-MA≥12 mos).

We had three goals. First, we aimed to identify which DSM-5 symptoms are shown in both ASD-MA<12 mos and GDD children, and therefore presumably relate more to developmental delay than to ASD.We hypothesized that items that differed between the GDD and ASD-MA<12 mos groups would reflect core deficits in ASD, primarily those in social communication.Second, wesought to identify which DSM-5 ASD symptoms are shown in both ASD groups. We hypothesized that items that differed between the two ASD groups would reflect the severity of their developmental delays, with items such as initiating and responding to joint attention, which require the triangulation of attention, and pretend play, which requires the representational capacity generally developing in the second year, more impaired in the ASD-MA<12 mos group. Finally, we investigated whether the ADOS-2 Toddler Module can classify children with mental agesbelow 12 months as having ASD, even though their developmental level is lower than the recommended nonverbal mental age of 12 months(Esler et al., 2015; Luyster et al., 2009).We alsocompared levels of ASD concern on the ADOS-2 Toddler Module for the three groups. We hypothesized that ASD severity would be higher in children with more impaired cognitive functioning (Maenner et al., 2020; Mayes and Calhoun, 2011), and that the ADOS-2 Toddler Module would be able to adequately detect symptoms of ASD in children with developmental ages lower than 12 months.

Methods

Participants and Procedure

Pediatric providers at three sites in the United States (Georgia, Pennsylvania, Connecticut) screened children for ASD starting at 12-, 15-, or 18-months with an age-appropriate screening instrument: First Year Inventory, Lite version (FYI-Lite; Baranek et al., 2014; Baranek, personal communication)and Infant Toddler Checklist (ITC; Wetherby & Prizant, 2002) at 12 months; Modified Checklist for Autism in Toddlers, Revised, with Follow-Up (M-CHAT-R/F; Robins et al., 2014) and FYI-Lite at 15 months; M-CHAT-R/F at 18 months. The sample (N=84) includes children who screened positive or were flagged bytheir pediatric provider with an ASD concern, were between 12.9 and 34 months at the time of evaluation,and completed the ADOS-2 Toddler Module.All children included in the sample, therefore, had elevated risk for ASD, regardless of their final diagnosis. Children were excluded from participation if they had a prior ASD diagnosis or significant motor or sensory impairments, such as cerebral palsy or blindness, that would interfere with testing. In line with the ADOS-2 Toddler Module criteria of being able to navigate independently, all children were able to walk or crawl independently.

The ASD-MA<12 mos group (n=28) had age equivalents below 12 months on Mullen Scales of Early Learning (MSEL) Expressive Language, Receptive Language, and Visual Reception scales, and met CBE criteria for Autism Spectrum Disorderas per DSM-5 (American Psychiatric Association, 2013), using a combination of parent report, developmental history, ADOS-2 scores, and clinical observation. The criteria for mental age based on age equivalents below 12 months on MSEL language and visual reception domains were consistent across the current study, Hinnebusch et al. (2017), and Miller et al. (2019). The diagnostic criteria for the ASD groups reflected the current DSM version (DSM-IV in the two prior studies; DSM-5 in the current study).

The Global Developmental Delay (GDD) group (n=27) consisted of children who met DSM-5 criteria for Global Developmental Delay but not for ASD. The GDD criteria were aligned with the DSM-5 and operationalized as scores 1.5 standard deviations or more below the mean on a minimum of one verbal and one non-verbal measure. A combination of parent report of adaptive behavior, developmental history, and clinical observation was used to inform diagnoses. The ADOS-2 was used to observe possible ASD behaviors.

The ASD-MA≥12 mosgroup (n=29) met the same CBE diagnostic criteria for ASDand had one or more cognitive or language scores at or above 12 months.

There were 28 children who met criteria for the ASD-MA<12 mos group; participants in each of the other two groups (GDD, ASD-MA≥12 mos) were matched to children in the ASD-MA<12 mos group. Participant characteristics are shown in Table 1. The children with ASD-MA<12 moswere matched in pairs to children with a GDD diagnosis as closely as possible on one verbal (Expressive Language) and one nonverbal (Visual Reception) score from the MSEL age equivalents. However, despite attempts at close matching, the ASD-MA<12 mos and GDD groups were still significantly different on all domains of cognitive functioning except Fine Motor. One child in the GDD group was dropped from subsequent analysis after being identified as an extremely low outlier in cognitive functioning (age equivalents of four months on visual reception, one month on all other domains of the MSEL). Developmentally, the ASD-MA<12 mos and GDD groups were significantly different on Expressive Language age equivalents (GDD M= 9.3 months, ASD-MA<12 mos M=7.6 months), Receptive Language (GDD M=10.1 months, ASD-MA<12 mos M=6.5 months) and Visual Reception (GDD M=10.9 months, ASD-MA<12 mos M=9.3 months), but not on Fine Motor skill (GDD M = 14.7 months, ASD-MA<12 mos M= 12.6 months). The groups were also significantly different on chronological age (GDD M=17.5 months, ASD-MA<12 mos M=19.5 months). The groups were not significantly different on ethnicity, maternal education, or sex distribution. The two groups were significantly different on race, with more families identifying as White in the ASD-MA<12 mosgroup than the GDD group(χ²(1, N=51)=7.07p=.008).

Table 1.

Participant Characteristics by Diagnostic Group

Variable	GDD (n=27)	ASD-MA<12 mos(n=28)	ASD-MA≧12 mos(n=29)	p (ASD-MA<12 mos vs GDD)	p (ASD-MA<12 mos vs ASD-MA≥12 mos)
Age in months at time of evaluation [M(SD), range]	17.5(2.7), 13.1–21.8	19.5 (4.3), 12.9–28.2	20.1 (4.7), 13.1–32.9	.049	.602
Sex [n(%)]				.080	.017
Male	19(71.4%)	25 (89.3%)	18 (62.1%)
Female	8 (28.6%)	3 (10.7%)	11 (37.9%)
Race [n(%)]				.008	.115
White	8 (30.8%)	17 (68.0%)	12 (46.2%)
People of color	18 (69.2%)	8 (32.0%)	14 (53.8%)
Racial identities within people of color variable
Black or African-American	12 (44.4%)	5 (19.2%)	8 (30.8%)
Asian	1 (3.7%)	0 (0.0%)	2 (7.7%)
Biracial or Multiracial	4 (14.8%)	2 (7.7%)	3 (11.5%)
American Indian or Alaska Native	1 (3.7%)	1 (3.8%)	1 (3.8%)
Other	1 (3.7%)	1 (3.8%)	0 (0.0%)
Ethnicity [n(%)]				.174	.825
Hispanic or Latinx	5 (20.8%)	10 (38.5%)	12 (41.4%)
Non Hispanic or Latinx	19 (79.2%)	16 (61.5%)	17 (58.6%)
Maternal education [n(%)]				.200	.563
Less than high school diploma	3 (11.5%)	2 (7.1%)	0 (0.0%)
High school diploma/GED	5 (19.2%)	11 (39.3%)	13 (44.8%)
Vocational or technical degree	7 (26.9%)	2 (7.1%)	3 (10.3%)
Associate degree	2 (7.7%)	5 (17.9%)	3 (10.3%)
Bachelor’s degree	4 (15.4%)	5 (17.9%)	5 (17.2%)
Advanced degree	5 (19.2%)	3 (10.7%)	5 (17.2%)
VABS-II Standard Score [M(SD), range]
Communication	81.6 (9.9), 62–97	68.5 (10.8), 44–90	79.1 (14.9), 46–117	<.001	.003
Daily Living Skills	88.7(14.1), 52–111	77.5 (11.9), 57–107	87.6 (11.1), 65–109	.002	.002
Socialization	90.3 (10.6), 76–110	78.7 (10.9), 56–96	84.2 (10.7), 65–104	<.001	.064
ADOS-2 Toddler Module Level of Concern [n(%)]				.006	.043
Little-to-No Concern	12 (44.4%)	3 (10.7%)	3 (10.3%)
Mild-to-Moderate Concern	3 (11.1%)	1 (3.6%)	8 (27.6%)
Moderate-to-Severe Concern	12 (44.4%)	24 (85.7%)	18 (62.1%)
ADOS-2 Toddler Module Calibrated Severity Scores [M(SD), range]
Social Affect (SA)	4.4 (1.9), 1–8	8.9 (1.1), 7–10	7.2 (1.9), 4–10	<.001	<.001
Repetitive Behavior (RRB)	3.2(2.3), 1–8	6.8 (2.2), 1–10	6.2 (1.7), 1–9	<.001	.271
MSEL Developmental Quotient a [M(SD), range]
Overall	66.5 (11.9), 43.9–86.0	48.3 (13.4), 18.3–71.7	75.3 (15.9), 43.3–102.9	<.001	<.001
Nonverbal	75.3 (13.5), 48.4–101.3	59.0 (18.2), 20.6–93.1	86.9 (16.7), 49.9–117.3	<.001	<.001
Verbal	57.8 (13.8), 30.8–84.6	37.6 (13.2), 16.0–75.0	63.6 (20.2), 28.1–106.3	<.001	<.001
MSEL age equivalent in months [M(SD), range] b
Expressive language	9.6 (3.0), 4–14	7.6 (2.3), 4–11	12.4 (4.6), 6–28	.007	<.001
Visual reception	11.1 (1.3), 9–13	9.3 (2.2), 5–11	16.1 (3.4), 10–23	<.001	<.001
Receptive language	10.4 (3.0), 5–18	6.5 (2.3), 2–11	12.2 (4.2), 6–25	<.001	<.001
Fine motor	14.7 (2.8), 9–21	12.6 (3.5), 3–17	17.7 (3.3), 12–24	.017	<.001

GDD Global Developmental Delay, ASD-MA<12 mosAutism Spectrum DisorderWithMental Age Below 12 Months, ASD-MA≥12 mosAutism Spectrum DisorderWithMental Age At or Above 12 Months,VABS-II Vineland Adaptive Behavior Scales, 2^nd Edition, ADOS-2 Autism Diagnostic Observation Schedule, 2^nd Edition, MSEL Mullen Scales of Early Learning

^aDevelopmental quotient was defined as ((Mental Age/Chronological Age) × 100))

^bMSEL age equivalent scores are a metric of mental age and calculated based on the raw scores of each domain, as provided by the MSEL manual (Mullen, 1990)

Children in the ASD-MA<12 mosgroup were matched on chronological age (CA) to children in the ASD-MA≥12 mosgroup. On sociodemographic variables, the two ASD groups were not significantly different on ethnicity, race, or maternal education. There weresignificantly more boys in the ASD-MA<12 mosgroup (25/28) than the ASD-MA≥12 mosgroup (18/29)(χ²(1, N=57)=5.70p=.017).

Because of the floor effect at the low end of the MSEL, with a lowest possible score of 49 on the ELC, we calculated a Developmental Quotient (DQ; sometimes called Ratio IQ), defined as Mental Age/Chronological Age × 100 (Bishop et al., 2015), combining Fine Motor and Visual Reception for Nonverbal DQ and combining Expressive and Receptive Language for Verbal DQ. The overall DQ was an average of the four subtest DQs. The DQs of all three groups differed significantly from each other (Table 1). While all groups had mean DQs below average, the ASD-MA≥12 mos group included 10 children with an overall DQ within 1 SD of the mean (≥85). Neither of the other groups included children whose developmental score was within the average range.

Measures

Mullen Scales of Early Learning (MSEL; Mullen, 1995)

The MSEL is a norm-referenced, standardized instrument for assessing cognitive and motor abilities in infants and young children from birth through 68 months. Four MSEL subscales were collected: Visual Reception, Fine Motor, Expressive Language, and Receptive Language. Age equivalents are provided in the MSEL manual and are based on the total raw score for each domain that is expected at each age. An overall Early Learning Composite (ELC) is a standard scorebased on all four MSEL subscales.To avoid floor effects at the low end of the ELC, with a lowest possible score of 49, DQs were calculated for each child.

Autism Diagnostic Observation Schedule – 2^nd Edition (ADOS-2; Lord et al., 2012)

The ADOS-2 is a semi-structured, standardized assessment of communication, social interaction, and play skills designed to evaluate symptoms of ASD. The ADOS-2 is not a standalone measure for the diagnosis of ASD, and is used in conjunction with developmental history and parent report of ASD symptoms (Coulter et al., 2021) to inform a clinical best estimate (CBE) diagnosis of ASD. The Toddler module is designed for children 12–30 months of age who have minimal speech, ranging from no spoken words to simple two-word phrases, have a nonverbal mental age of at least 12 months, and are walking independently. Items on the ADOS-2 are scored as 0 to indicate behavior that is not symptomatic of ASD. Scores of 1, 2, or 3 are used to indicate the presence of the symptom, with scores of 1 indicating the mildest behavior. The ADOS-2 generates scores for Social Affect (SA) and Restricted and Repetitive Behavior (RRB), as well as an overall Total. The algorithm items used to generate these scores are based on the child’s language level. All but three participants in this sample received the scoring algorithm for children between 12 and 20 months who have few to no words. Level of concern (0–9: Little-to-No Concern, 10–13: Mild-to-Moderate Concern, 14 or higher: Moderate-to-Severe Concern) is based on cutoffs of the overall Total score. The percent of each group meeting each level of ASD concern is shown in Table 1.

Vineland Adaptive Behavior Scales, 2^nd Edition, Parent Interview Form (VABS-II; Sparrow et al., 2005)

The VABS-II is a semi-structured parent interview form administered by a clinician to identify adaptive behavior in the domains of Communication, Daily Living, Socialization, and Motor Skills. An overall Adaptive Behavior Composite is a standard score derived from the domain-level standard scores. Domain-level standard scores for Communication, Daily Living, and Socialization were used to characterize the sample but were not used in analyses in the current study. The ASD-MA<12 mos group scored significantly below the other two groups on Communication and Daily Living, and below the GDD group on Socialization (Table 1).

Data Analysis Plan

Scores from the ADOS-2 Toddler Module were dichotomized such that scores of 0 indicated lack of the ASD symptom, and scores equal to or greater than 1 were collapsed into a single category to indicate the presence of the ASD symptom. We examined group differences on ADOS-2 Toddler Module items using chi-square or Fisher’s exact test. For these comparisons, we used a Bonferroni correction (.05/41) so that only group differences with p values of .001 or smaller were considered significant.

For ADOS-2variables that differed in frequency between groups and met the Bonferroni threshold, we then fit several logistic regressions models to covary significantly different demographic and developmental variables. Because theASD-MA<12 mosgroup and GDD groups were significantly different on DQand chronological age, these variables wereincluded ascovariates in logistic regression models. The ASD-MA<12 mosand ASD-MA≥12 mosgroupsdiffered on DQ, as expected, and on sex distribution. Therefore, sex and DQwere included as covariates in logistic regressionmodels ofADOS-2items that differed in frequencybetween the ASD groups.For these items, adjusted odds ratios and confidence intervals are shown in Table 2. For items that did not differ in frequency between the groups, we compared the severity of items using chi squares. For these analyses, severity ranged from 1 (mild) to 2 (moderate to severe; scores 2 and 3 are collapsed into scores of 2 on the ADOS-2).

Tabel 2.

Number of Children by Group with Endorsed ADOS-2 Toddler Module Symptoms

ADOS-2 Toddler Module Item	GDD n (%)	ASD-MA<12 mos n (%)	ASD-MA≥12 mos n (%)	Chi-squarebp (ASD-MA<12 mosvs. GDD)	AORc (95% CI) ASD-MA<12 mos vs. GDD	Chi-squarebp (ASD-MA<12 mos vs. ASD-MA≥12 mos)	AORd (95% CI) ASD-MA<12 mos vs. ASD-MA≥12 mos
A1 (overall level of non-echoed spoken language)	27(100%)	28 (100%)	29 (100%)	N/A		N/A
A1a (frequency of babbling)	23 (85.2%)	26 (92.9%)	24 (100%)	.422		.493
A2 (frequency of spontaneous vocalizations directed to others)	19 (70.4%)	28 (100%)	27 (96.4%)	.002		1.000a
A3 (intonation of vocalizations and verbalizations)	3 (18.8%)	9 (52.9%)	11 (64.7%)	.071		.728
A4 (immediate echolalia)	N/A	N/A	N/A	N/A		N/A
A5 (stereotyped/idiosyncratic use of words or phrases)	N/A	N/A	N/A	N/A		N/A
A6 (use of another’s body)	9 (39.1%)	16 (69.6%)	19 (67.9%)	.075		1.000a
A7 (pointing)	22 (84.6%)	28 (100%)	25 (89.3%)	.047		.236
A8 (gestures)	22 (84.6%)	26 (100%)	26 (96.3%)	.110		1.000a
A9 (frequency of undirected vocalization)	7 (31.8%)	22 (91.7%)	18 (72.0%)	<.001	8.83(1.23–63.70), p=.031	.138
B1 (unusual eye contact)	11 (40.7%)	27 (96.4%)	27 (96.4%)	<.001	26.17(2.31–296.39), p=.008	1.000a
B2 (teasing toy play)	14 (51.9%)	19 (79.2%)	19 (73.1%)	.077		.745
B3 (unable toy play)	8 (30.8%)	23 (88.5%)	12 (42.9%)	<.001	11.97 (1.96–73.04), p=.007	.001	n.s.
B4 (facial expressions directed to others)	13 (48.1%)	28 (100%)	23 (82.1%)	<.001	n.s.	.051
B5 (integration of gaze and other behaviors)	16 (59.3%)	28 (100%)	25 (89.3%)	<.001	n.s.	.236
B6 (shared enjoyment)	5 (18.5%)	19 (67.9%)	16 (57.1%)	<.001	5.43 (1.20–24.58), p=.028	.582
B7 (response to name)	8 (30.8%)	25 (89.3%)	19 (67.9%)	<.001	6.63 (1.17–37.56), p=.032	.101
B8 (ignore)	18 (69.2%)	24 (88.9%)	26 (92.9%)	.099		.669
B9 (requesting)	12 (44.4%)	29 (92.9%)	26 (92.9%)	<.001	14.90 (1.97–112.79), p=.002	1.000a
B10 (amount of requesting)	22 (81.5%)	28 (100%)	22 (81.5%)	.024		.023
B11 (giving)	16 (61.5%)	28 (100%)	24 (85.7%)	<.001	n.s.	.111
B12 (showing)	23 (85.2%)	28 (100%)	27 (96.4%)	.051		1.000a
B13 (spontaneous IJA)	14 (51.9%)	27 (96.4%)	16 (57.1%)	<.001	40.70 (3.27–507.35), p=.004	.001	n.s.
B14 (response to JA)	17 (65.4%)	26 (96.3%)	12 (42.9%)	.005		<.001	n.s.
B15 (quality of social overtures)	15 (57.7%)	28 (100%)	28 (100%)	<.001	n.s.	N/A
B16a (amount of social overtures to examiner)	20 (74.1%)	28 (100%)	27 (96.4%)	.004		1.000a
B16b (amount of social overtures to parent)	19 (73.1%)	26 (92.9%)	25 (92.6%)	.072		1.000a
B17 (level of engagement)	17 (65.4%)	26 (96.3%)	26 (92.9%)	.005		1.000a
B18 (quality of rapport)	14 (53.8%)	25 (92.6%)	25 (89.3%)	.002		1.000a
C1 (functional play with objects)	24 (88.9%)	28 (100%)	24 (85.7%)	.111		.111
C2 (imagination/creativity)	25 (92.6%)	28 (100%)	27 (96.4%)	.236		1.000a
C3 (functional and symbolic imitation)	22 (82.5%)	28 (100%)	23 (82.1%)	.023		.051
D1 (unusual sensory interests)	7 (25.9%)	21 (75.0%)	15 (53.6%)	<.001	8.934 (1.85–43.06), p=.006	.162
D2 (hand/finger movements)	3 (11.5%)	15 (53.6%)	12 (42.9%)	.001	8.25 (1.43–47.47), p=.018	.593
D3 (other complex mannerisms)	8 (30.8%)	21 (75.0%)	14 (50.0%)	.002		.097
D4 (SIB)	2 (7.4%)	3 (10.7%)	4 (14.3%)	1.000		1.000a
D5 (repetitive interests/stereotyped behaviors)	7 (25.9%)	20 (71.4%)	23 (82.1%)	.001	18.26 (3.09–107.97), p=.001	.528
E1 (overactivity)	9 (33.3%)	21 (75.0%)	14 (50.0%)	.003		.097
E2 (fussiness/irritability)	9 (33.3%)	19 (67.9%)	16 (57.1%)	.015		.582
E3 (aggression and disruptive behavior)	2 (7.4%)	1 (3.6%)	4 (14.3%)	.611		.352a
E4 (anxiety)	8 (29.6%)	6 (21.4%)	7 (25.0%)	.547		1.000

Note: some individual ADOS-2 items were missing for individual children; the values in the table reflect the percent out of those with data for that item

GDD Global Developmental Delay, ASD-MA<12 mos Autism Spectrum Disorder With Mental Age Below 12 Months, ASD-MA≥12 mosAutism Spectrum Disorder With Mental Age At or Above 12 Months, ADOS-2 Autism Diagnostic Observation Schedule, 2^nd Edition, Toddler Module (endorsed items indicate impairment on item)

^bFisher’s exact two-tailed test run when expected cell count <5 in chi-square; Bonferroni correction to adjust for multiple comparisons

^cAOR reported when chi-square indicated significant group difference in endorsement of item. Bold indicates significant p-values and AORs after controlling for chronological age and DQ.

^dAOR reported when chi-square indicated significant difference in endorsement of item. Bold indicates significant p-values and AORs after controlling for sex and DQ.

Results

Autism vs. Non-autism Comparison (ASD-MA<12 mosversus GDD)

As expected, the ASD-MA<12 mosgroup had significantly higher (more impaired) scores on both SA(t(52)=−10.66, p<.001) and RRB (t(52)=−5.90, p<.001) ADOS-2 total scores (Table 1). The two groups were also significantly different(χ²(2, N=55)=10.39p=.006) on the proportion at each level of ASD concern (Table 1). Significantly more (85.7%) of the ASD-MA<12 mos group were classified in the Moderate-to-Severe Concern range, compared to 44.4% of the GDD group. Of the GDD group, 44.4%were classified in the Little-to-No Concern range, compared to only 11.1% of the ASD-MA<12 mos group.

In Table 2, we present the ADOS-2 items, with bold type used to indicate group differences that survived the Bonferroni correction and were still significant after fitting logistic regressionmodels. On the ADOS-2 Toddler Module, items in the A group relate to communication, in the B group relate to social interaction, in the C group play and imitation, in the D group RRBs, and in the E group mood symptoms not specifically related to ASD.

Of the 41 ADOS-2 items, four could not be compared because scores were identical or because the item did not apply to those children. In the logistic regression model controlling for DQ and CA, 10 of the remaining 37 itemsdiffered between groups.Toddlers in the ASD-MA<12 mosgroup were significantly more likely than toddlers in the GDD group to be endorsed as symptomatic on: frequency of undirected vocalizations, unusual eye contact, lack of shared enjoyment, lack of response to name, less requesting behaviors, inadequate responding to a socially ambiguous situation during toy play, lack of spontaneous initiation of joint attention (IJA),unusual sensory interests, unusual hand/finger movements, and repetitive interests/stereotyped behaviors.

Although 27 ADOS-2 Toddler Module items did not differ in frequency (scored as 0 or 1) between the ASD-MA<12 mos and GDD groups, it was still possible that the average severity of these symptoms was different in the two groups(Table 3). After correcting for multiple comparisons with a significant p-value set to <.001, toddlers in the ASD-MA<12 mos group received more severe ratings than toddlers in the GDD group onfrequency of spontaneous vocalizations to others,integration of gaze and other behaviors, quality of social overtures,amount of social overtures to examiner, amount of social overtures to parent, level of engagement, and overall quality of rapport.Thus 10 items differed in frequency of endorsement, and seven additional items differed in severity. Key features that should raise suspicion of ASD in toddlers with global developmental delays are shown in Table 4.

Table 3.

Percent Endorsed with Score of 2 (More Severe) on Symptomatic ADOS-2 Toddler Module Items^e

ADOS-2 Toddler Module Itemse	GDD n (%)	ASD-MA<12 mos n (%)	ASD-MA≥12 mos n (%)	Chi-square p (ASD-MA<12 mos vs. GDD)f	Chi-square p (ASD-MA<12 mos vs. ASD-MA≥12 mos)f
A1a (frequency of babbling)	16 (69.6%)	21 (80.8%)	19 (79.2%)	.508	.999
A2 (frequency of spontaneous vocalizations directed to others)	9 (47.4%)	26 (92.9%)	20 (74.1%)	<.001	.078
A3 (intonation of vocalizations and verbalizations)	1 (33.3%)	6 (66.7%)	7 (63.6%)	.523	.999
A6 (use of another’s body)	3 (33.3%)	8 (28.6%)	12 (63.2%)	.677	.433
A7 (pointing)	15 (68.2%)	28 (100%)	21 (84.0%)	.002	.043
A8 (gestures)	16 (72.7%)	22 (78.6%)	17 (65.4%)	.478	.109
A9 (frequency of undirected vocalization)	2 (28.6%)	17 (60.7%)	8 (44.4%)	.030	.033
B1 (unusual eye contact)	5 (45.5%)	22 (81.5%)	16 (59.3%)	.047	.074
B2 (teasing toy play)	7 (50.0%)	15 (78.9%)	14 (73.7%)	.136	.999
B3 (unable toy play)	6 (75.0%)	20 (87.0%)	7 (58.3%)	.583	.091
B4 (facial expressions directed to others)	3 (23.1%)	22 (78.6%)	12 (52.2%)	.001	.047
B5 (integration of gaze and other behaviors)	2 (12.5%)	22 (78.6%)	10 (40.0%)	<.001	.004
B6 (shared enjoyment)	2 (40.0%)	10 (52.6%)	9 (56.3%)	1.000	.830
B7 (response to name)	3 (37.5%)	15 (60.0%)	7 (36.8%)	.417	.128
B8 (ignore)	9 (50.0%)	21 (87.5%)	19 (73.1%)	.014	.294
B9 (requesting)	8 (66.7%)	23 (88.5%)	19 (73.1%)	.176	.159
B10 (amount of requesting)	6 (27.3%)	17 (60.7%)	11 (50.0%)	.024	.449
B11 (giving)	12 (75.0%)	25 (89.3%)	16 (66.7%)	.236	.046
B12 (showing)	19 (82.6%)	28 (100.0%)	23 (85.2%)	.035	.051
B13 (spontaneous IJA)	9 (64.3%)	23 (85.2%)	14 (87.5%)	.231	.999
B14 (response to JA)	7 (41.2%)	22 (84.6%)	N/A	.007	N/A
B15 (quality of social overtures)	2 (13.3%)	21 (75.0%)	10 (35.7%)	<.001	.003
B16a (amount of social overtures to examiner)	7 (35.0%)	27 (96.4%)	22 (81.5%)	<.001	.101
B16b (amount of social overtures to parent)	7 (38.9%)	25 (96.2%)	19 (79.2%)	<.001	.093
B17 (level of engagement)	2 (11.8%)	23 (88.5%)	12 (46.2%)	<.001	<.001
B18 (quality of rapport)	3 (21.4%)	22 (88.0%)	11 (44.0%)	<.001	<.001
C1 (functional play with objects)	15 (62.5%)	26 (92.9%)	15 (62.5%)	.015	.008
C2 (imagination/creativity)	23 (92.0%)	28 (100.0%)	23 (85.2%)	.218	.051
C3 (functional and symbolic imitation)	14 (63.6%)	27 (96.4%)	18 (78.3%)	.007	.079
D1 (unusual sensory interests)	3 (42.9%)	16 (76.2%)	7 (46.7%)	.165	.069
D2 (hand/finger movements)	3 (100.0%)	10 (66.7%)	7 (58.3%)	.522	.706
D3 (other complex mannerisms)	5 (62.5%)	15 (71.4%)	12 (85.7%)	.675	.431
D4 (SIB)	1 (50.0%)	2 (66.7%)	2 (50.0%)	1.000	.999
D5 (repetitive interests/stereotyped behaviors)	1 (14.3%)	14 (70.0%)	8 (34.8%)	.024	.021
E1 (overactivity)	0 (0.0%)	12 (63.2%)	5 (38.5%)	.006	.169
E2 (fussiness/irritability)	2 (22.2%)	11 (57.9%)	6 (37.5%)	.114	.229
E4 (anxiety)	3 (37.5%)	2 (33.3%)	1 (14.3%)	1.000	.559

Italicized p-values indicate significant chi-squares after correcting for multiple comparisons with p<.001

GDD Global Developmental Delay, ASD-MA<12 mos Autism Spectrum Disorder With Mental Age Below 12 Months, ASD-MA≥12 mosAutism Spectrum Disorder With Mental Age At or Above 12 Months, ADOS-2 Autism Diagnostic Observation Schedule, 2^nd Edition (endorsed items indicate impairment on item)

^eAccording to the ADOS-2 Toddler Module, assigned ratings of 3 are converted to 2 (therefore the range of possible scores for endorsed items1 or 2; percent is out of total with endorsed item)

^fFisher’s exact two-tailed test run when expected cell count <5 in chi-square; Bonferroni correction to adjust for multiple comparisons

Table 4.

Key Features to Alert Providers to Assess for ASD in Globally Developmentally Delayed Toddlers

Social Affect	Restricted and Repetitive Behaviors
Infrequently directs vocalizations to provider or parent	Clear demonstrations of interests in sensory aspects of objects or environment (e.g., sniffing, licking, or repeatedly touching a specific texture; bringing objects up to eye level and peering at them)
Absence of consistent eye gaze during social interaction to modulate interaction (e.g., looking at provider when provider speaks to child, looks and points to item to request)	Repetitive movements of hands or fingers (e.g., tensing, finger flicking)
Lack of response to name when called by provider or parent	Restricted interests (e.g., only wants to play with one part of toy or object, such as spinning wheels of car) or stereotyped behaviors (e.g., lining up toys; insistence on other people performing a ritual) that cause distress if interrupted
Diminished shared enjoyment during play (e.g., not smiling, laughing. Or looking at provider/parent when tickled or playing peekaboo)
Absence of eye contact and another behavior (e.g., vocalization, handing object to provider/parent, facial expression) to request specific object (e.g., snack or toy) from provider or parent
Does not make eye contact with provider or parent if provider or parent pretends to be unable to open toy or snack container
Absence of behaviors to draw another person’s attention to an object that is out of reach (e.g., a toy across the room)

ASD Comparison by Developmental Level (ASD-MA<12 mosversus ASD-MA≥12 mos)

The ASD-MA<12 mos group had more impaired functioning on the ADOS-2SAscore than the ASD-MA≥12 mos group (t(51)=−4.15, p<.001)but the two groups did not differ on the RRB score (t(52)=−1.11, p=.271). The percent of the two ASD groups who met the varied levels of ASD concern on the ADOS-2 (Table 1) was significantly different(χ²(2, N=57)=6.29, p=.043). The majority of each ASD group was classified in the Moderate-to-Severe Concern range (85.7% of the ASD-MA<12 mos group, 62.1% of the ASD-MA≥12 mos group). Only 3.6% of the ASD-MA<12 mos group and 27.6% of the ASD-MA≥12 mos group were classified as having Mild-to-Moderate Concern, and 10% of each group were classified as Little-to-No Concern.

Three of the 41 ADOS-2 items (Initiating Joint Attention [IJA], Responding to Joint Attention [RJA], and responding to socially ambiguous play) differed in frequency of endorsement, before controlling for sex and DQ differences. After controlling for sex and DQ in logistic regression models, noneof the items remained significantly different between the two ASD groups (Table 2).

To confirm that differences in the twoJA items (IJA and RJA) were due to mental age differences, point-biserial correlations were run between DQ and the ADOS-2 JA items for all children with ASD. Both items were significantly negatively correlated with DQ (IJA: r_pb(56) = −.38, p = .004; RJA: r_pb(55) = −.62, p< .001); the negative direction indicates that higher DQis associated with less impaired joint attention. As a comparison, correlations were also run between DQ and two ADOS-2 items (pointing, shared enjoyment) that were not different between the groups and can generally be seen in the first year of life. Correlations between these items and DQ were smaller and did not reach significance (pointing: r_pb(56) = −.24, p = .073; shared enjoyment: r_pb(56) = −.26, p = .057).

Although noneof the ADOS-2 Toddler Module items differed in frequency between the ASD-MA<12 mos and ASD-MA≥12 mos groups, it was still possible that the severity of each symptom was different in the two groups.After correcting for multiple comparisons with a significant p-value set to <.001, toddlers in the ASD-MA<12 mos group received more severe ratings than toddlers in the ASD-MA≥12 mosgroup on level of engagement and overall quality of rapport(Table 3). Thus, none of the items differed in frequency of endorsement, and only two additional items differed in severity between the two groups.

Discussion

The goals of the current study were to elucidate symptoms of ASD in children with mental ages below 12 months, to determine which of these symptoms are most clearly related to their cognitive delays and which to their ASD, and to examine whether the ADOS-2 Toddler Module could identify ASD in children with lower mental ages than those for which the test is designed.

Althoughcognitive impairment in very young children can make separating symptoms of ASD from those of non-ASD delaysdifficult, many symptoms observable on the ADOS-2 Toddler Module appeared in children with ASD who were functioning below a one-year level in verbal and nonverbal domains. In addition, there were many differences in symptoms between the ASD-MA<12 mos and GDD groups even aftercontrolling for CAand DQ. This suggests that the ADOS-2 Toddler Module distinguishes ASD from global developmental delay in young children, and that ASD symptoms in this group are not attributable primarily to low developmental age but to core ASD symptomatology; this is consistent with the finding that ASD is stable in this group (Hinnebusch et al., 2017; Miller et al., 2019). The symptoms seen more frequently in the ASD-MA<12 mosgroup than in GDDgroup were in communication (more undirected vocalizations, less requesting, and lack of spontaneous IJA), social responsiveness (unusual eye contact, lack of shared enjoyment, poor responding to a socially ambiguous situation during toy play),sensory interests, stereotyped movements (unusual hand/finger movements) and repetitive interests.

Contrary to some literature (Barton et al., 2013; Stone et al., 1999; Troyb et al., 2016)that suggested that repetitive behaviors and interests may not be apparent in very young children, behaviors in this domain strongly differentiated toddlers in the GDD and ASD-MA<12 mosgroups and occurred at similarly high rates in the two ASD groups. These behaviors seem to be a useful diagnostic marker of ASD even in toddlers with mental agesbelow 12 months. This finding is consistent with the presence of severe repetitive and restricted interests and behaviors in toddlers aged one to four years diagnosed with ASD on the Toddler Autism Symptom Inventory(Coulter et al., 2021).

There were a number of items on the ADOS-2 that were not significantly different in frequencyof endorsementbetween children with GDD and those with ASD-MA<12 mos. These results replicate prior findings that some children with GDD demonstrate impairments in pointing and imagination/creative play (Miller et al., 2019), which may require more advanced cognitive skills than children in either the ASD or the GDD group had. In addition to pointing and creative play, the current results suggest that some children with GDD demonstrate impaired gestures, showing objects, imitation, babbling. These items occurred at similar rates in GDD and ASD-MA<12 mos groups, suggesting that they may not be reliable behaviors to differentiate ASD from GDD in children who are very developmentally delayed.

Despite the fact that a number of children with GDD did display symptoms often associated with ASD, the severity of some of these behaviors was lower in the GDD than in the ASD-MA<12 mos groups. These symptoms included the integration of gaze and other behaviors, and level of engagement. Of particular relevance to the clinical implications of these findings is that the ASD-MA<12 mos group received significantly more severe ratings on all measures of social overtures and rapport. This suggests that while children in the GDD group demonstrated some difficulties in sustaining the interaction with the examiner, these skills were more noticeably impaired in the ASD-MA<12 mos group. Key features that should alert providers to further evaluate toddlers with global developmental delays for the possibility of ASD are summarized in Table 4.

Of clinical relevance, the ADOS-2 Toddler Module classified only about 10% of the ASD-MA<12 mos as having Little-to-No ASD concern, and classified a high percent (86%) as having Moderate-to-Severe concern. These results are consistent with the findings of Hinnebusch et al. (2017) that this group of children have severe autism and diagnoses are stable even when an initial diagnosis of ASD is made in children with mental ages below 12 months. The ADOS-2 thus appears to be quite sensitive to ASD even when developmental functioning is lower than previously validated for this instrument. On the other hand,44% of the GDD group was classified in the Moderate-to-Severe Concern range, suggesting that ADOS-2 Toddler Module maylack specificityfor ASD in children presenting with global developmental delays(Risi et al., 2006; Zander et al., 2015).This finding is clinically noteworthy, as distinguishing ASD from GDD is likely to be most challenging if the child with GDD has symptoms that overlap with those of ASD.

The lack of specificity of the ADOS-2 under some conditions should be explored in larger samples of a wider age range oflower functioning children with a variety of neurodevelopmental disorders, such as ADHD, temperamental differences, language disorders, and defined genetic syndromes.The comparison of the ASD-MA<12 mos group to the age-matched group of children with ASD and less severe cognitive impairment was also revealing. No items on the ADOS-2 differed between these groups,after controlling for sex and DQ, consistent with the Mayes and Calhoun (2011) findings of no effects of age or IQ on the total count ofcore ASD symptoms. When examining the severity of items, only level of engagement and quality of rapport differed between the ASD groups, with the low MA group showing more severe symptoms and more severe overall level of concern on the ADOS-2. Additionally, the ADOS-2 characterizedonly 10% of each ASD group as having Little-to-NoASDconcern, confirming that the measure was as sensitive for the children functioning below a 12-month level as for the higher-functioning children, in our sample (but with suboptimal specificity, as mentioned above).

Impairment on both IJA and RJA were endorsed for only about half of the ASD-MA≥12 mos group, but almost all of the ASD-MA<12 mos group, although these differences did not remain significant after controlling for DQ and sex.Additionally, JA items, but not several simpler social behaviors,such as pointing or shared enjoyment, were significantly correlated with developmental age in the ASD sample. This suggeststhat both cognitive delay and ASD contribute separately to JA, as suggested by others (Mundy et al., 1994; Paparella et al., 2011). Developmental delays, especially when severe, undoubtedly exacerbate social cognition difficulties in children with ASD.

Although the literature suggests that greater ASD severity and cognitive delays tend to even out the sex ratio in ASD prevalence (Lai et al., 2015), our results suggested the opposite: there were proportionately more boys than girls in the ASD-MA<12 mos group than in the ASD-MA≥12 mos group. This may be a serendipitous finding, due to the relatively small samplewhich prevents us from drawing firm conclusions about sex differences between children with ASD and mental ages below versus above 12 months. In addition, controlling for sex in logistic regression models did not change any differences in frequency of symptom endorsement between the two ASD groups.In order to match the groups on cognitive functioning as closely as possible, the sample was not big enough to also match on gender, SES, or other relevant demographic factors. In a larger sample, planned comparisons on gender would certainly be in order to examine.

On average, the ASD-MA≥12 mos group as a whole was not functioning within the average range of development on any measure. Comparing ASD children who are severely delayed to those within or close to a typical range of development would doubtless produce increased differences between groups. Few studies identify children with ASD who are in the average or above average range on cognitive and language measures when they are younger than three years old. One possible explanation for this is that children with milder ASD or cognitive impairment are less likely to be detected when they are very young,before social and language demands increase with age and symptoms are more readily apparent.

Although there is a wide range of MA and CA within individual groups, only the ASD-MA≥12 mos group had some children (n=10) with DQs in the average range (defined as within one standard deviation of the mean). Therefore, although our grouping strategy does mean that children with a range of DQs are included in the same group, none of the groups had a mean DQ score at the average level.

An additional finding is that race was unequally distributed in our groups; significantly more children of color had a GDD diagnosis. The significant differences in representation of children of color in the ASD-MA<12 mos versus GDD group highlights the historical nature of people of color being under- or over-represented in certain diagnostic groups, an inequity that is especially salient in pediatric populations. Prior research(Mandell et al., 2009) indicates that children of color are significantly less likely to have a documented ASD diagnosis than White children, and that providers are less likely to assess for ASD in children of color who have a documented ID or GDD. However, this bias may be partially mitigated by the use of standardized screening and evaluation procedures (Herlihy et al., 2014), as used in the current study. Future research with large samples should explore factors that may underlie bias on the part of evaluators to diagnose GDD more readily in children of color, or whether GDD is truly more prevalent in children of color.Although pediatricians were asked to adhere to specific study procedures for all children, it is also possible that differences in practice may be attributable to previous bias. One such factor, maternal education, was not different across groups in the current study, but many other relevant factors need to be examined, such as income, access to prenatal and pediatric health care, and nutrition. We used maternal education as a proxy for socioeconomic status (SES), but recognize the limitations of using this metric to account for the impact of SES on health outcomes (Williams et al., 2016). Additionally, the ADOS-2 Toddler Module does not have normative data stratified by racial or ethnic categories, thus limiting its interpretability in non-White samples.

There are several limitations to the current study. The sample size was relatively small; given the corrections for multiple analyses, some real group differences might have remained significant in a larger sample. Although we covaried DQ in these analyses to account for differences in cognitive functioning between groups, ideally, we would have a GDD group that was matched very tightly to the ASD-MA<12 months group on cognitive functioning. Given our small sample sizes, the groups are relatively heterogenous in terms of DQ. However, we addressed this by covarying all variables that differed between the groups, including DQ as a continuous variable. With a larger sample size, one potential future direction might be to split the sample by median DQ and covary CA or MA, or split the sample by CA and covary MA or DQ. With our current small sample size, such grouping procedures would yield groups with fewer than 15 children, making it difficult to interpret the results of such analyses.

An additional limitation of the current study is thatall children included in the study were identified as at risk for ASD based on screening or pediatrician concern. Therefore, the GDD sample may not be representative of the broader group of children with GDD. Replicating these analyses in a larger sample with children withGDDwhose mental ages were below 12 months, as well as children with GDD who were not referred for an ASD evaluation, might help illuminate whether specific items should be weighed more heavily in considering a diagnosis of ASD or GDD in very impaired toddlers. These analyses might also provide more quantitative data on the specificity of the ADOS-2 Toddler Module in samples with GDD.

Additionally, the groups were not perfectly matched on chronological age or cognitive functioning. Even standard deviations of 2–3 months are meaningful in children who are functioning below a 12-month level. The ASD-MA<12 mos group and the GDD group differed by about two monthson visual reception and expressive language skills, but differed by about four months on receptive language abilities. The particularly low receptive language score in the ASD-MA<12 mos group may be due to language disorder or to inattention to others’ language, either during the evaluation or in general.

The current study also has limitations based on methodology. In particular, some items had extremely wide confidence intervals. For example, unusual eye contact was endorsed for less than half of the GDD group, compared to over 90% of the ASD-MA<12 mos group, resulting in a very wide confidence interval, which limits the interpretation of these estimates. An additional limitation is that our analyses used age equivalents from the MSEL to calculate mental age for group assignment, and also covaried a DQ calculated from the MSEL to control for the wide variability in mental age. Ideally, we would have two separate measures of cognitive functioning to calculate group assignment and to covary in analyses.

Another limitation is that thesame instrument (ADOS-2 Toddler Module) contributed to the CBE diagnosis and was used to identify individualbehaviors that were attributable to low mental age versus ASD. Ideally, our analyses would have includedtwo separate measures of ASD symptomatology. In addition,the symptoms were drawn from an observational instrument (the ADOS-2 Toddler Module), which precludes evaluatingsome symptoms, such as interest in peers, which cannot be observed during a child’s evaluation.

Conclusion: The ADOS-2 Toddler Module identified ASD symptomatology even in children whose mental age is below 12 months, with good sensitivity but questionable specificity. Symptomatology in the ASD-MA<12 mos and ASD-MA≥12 mos groups was strikingly similar; developmental level affected the clinical presentation of ASD less than was expected. Contrary to some prior literature, unusual interests and repetitive behaviors were seen in a majority of the very young children with ASD-MA<12 mos. Many behaviors distinguished children with ASD-MA<12 mos from children with GDD in all autism domains (communication, social interaction, repetitive behaviors), supporting the conclusion that ADOS-2 Toddler Module is valid to inform ASD diagnosisin children with mental ages below 12 months. However, ASD diagnosis in this group will need to rely largely on absent or reduced social and communication behaviors typically seen in the first year of life, such as eye contact and shared enjoyment, to rule out developmental delay as the main factor in these deficits. In addition, however, presence of RRBs can add to impaired social communication to support diagnosis of ASD, even in children functioning below 12 months.