Journal of Autism and Developmental Disorders

https://doi.org/10.1007/s10803-018-3620-z

ORIGINAL PAPER

Correlates of DSM-5 Autism Spectrum Disorder Levels of Support

Ratings in a Clinical Sample
Lauren M. Gardner1   · Jonathan M. Campbell2 · Bruce Keisling3 · Laura Murphy3

© Springer Science+Business Media, LLC, part of Springer Nature 2018

Abstract
The DSM-5 features level of support ratings for social communication (SC) and restrictive and repetitive behaviors (RRB)
for individuals with autism spectrum disorder (ASD). We contrasted cognitive, adaptive, and autism severity scores across
SC and RRB groups for 158 individuals with ASD diagnosed in a developmental disabilities clinic. Roughly 46% of indi-
viduals were identified by licensed psychologists’ clinical judgement as needing Level 2 SC support and 49% were identified
as needing Level 2 RRB support. No individuals were rated as needing a combination of Level 1/Level 3 supports across
domains. MANOVA and direct discriminant analysis revealed that both SC and RRB groups showed a graded pattern of
higher adaptation/lower autism severity to lower adaptation/higher autism severity from Level 1 to Level 3.

Keywords  Autism spectrum disorder · DSM-5 · Levels of support · Severity · Diagnosis

Introduction spectrum disorder, the presentation of symptoms are diverse.

Prognoses vary, in part, based on the severity of the symp-
The diagnostic criteria for Autism Spectrum Disorder toms displayed.
(ASD), as presented in the Diagnostic and Statistical Man- The DSM-5 features specifiers to describe individu-
ual of Mental Disorders, Fifth Edition (DSM-5; American als’ symptom severity and needs for support in two areas,
Psychiatric Association 2013), requires that during the early Social Communication (SC), and Restricted and Repetitive
developmental period a child demonstrates impairments in Behaviors (RRB). The specifiers require clinicians to use
social communication and interaction, and restricted and their clinical judgement to delineate between three classifi-
repetitive patterns of behaviors. Examples of social com- cations: Level 1 (“Requiring support”), Level 2 (“Requiring
munication deficits that may be present include difficulties substantial support”), and Level 3 (“Requiring very substan-
in social-emotional reciprocity, impaired nonverbal com- tial support”) in both SC and RRB (APA 2013). Research
munication skills, and difficulties building and maintaining supports that the determination and specification of ASD
relationships with others. Restricted and repetitive behav- severity at time of diagnosis is valuable and may convey
iors that may be observed in individuals with ASD include important information about symptom course and prognosis
stereotyped repetitive movements, object use, or speech; (Gotham et al. 2012). The DSM-5 SC and RRB levels of
rigidity; highly fixated interests; and over or under reaction severity and support may aide in the identification of areas
to sensory input. Although these core symptoms are com- of relative strengths and weaknesses as they relate to ASD
mon behaviors among individuals diagnosed with ASD, as a core symptomology and facilitate individualized interven-
tion planning. However, the utility of the SC and RRB levels
* Lauren M. Gardner of severity in determining levels of support is limited given
[email protected] the current lack of quantitative methods or practice recom-
mendations for differentiating between the levels based on
1
Autism Center, Johns Hopkins All Children’s Hospital, 880 an individual’s impairments across cognitive, adaptive,
6th Street South, Suite 410, St. Petersburg, FL 33701, USA
behavioral, and ASD-specific symptom domains. Cur-
2
Department of Educational, School, and Counseling rently, the descriptions of SC and RRB levels of severity
Psychology, University of Kentucky, Lexington, USA
provided within the DSM-5 remain conceptual in nature,
3
Boling Center for Developmental Disabilities, The University
of Tennessee Health Science Center, Memphis, TN, USA

13
Vol.:(0123456789)
 Journal of Autism and Developmental Disorders

subjective, and determined by the clinical judgment of the studies utilized two or more measures to quantify severity.
diagnostician. Due to recent changes in DSM-5 diagnostic criteria and
The concept of determining ASD severity has the poten- introduction to levels of support ratings, further investiga-
tial to yield valuable information related to symptomology, tion of quantifying and reporting severity is warranted.
prognosis, and treatment. Ideally, determining ASD severity
accounts for information regarding an individual’s cogni-
tive functioning, language deficits, adaptive behavior impair- Investigations of DSM‑5 Levels of Support
ment, and severity of behavior problems (Weitlauf et al.
2014). Thus, levels of support for SC and RRB may not be The changes to the diagnostic criteria presented in DSM-5
adequately assessed utilizing only ASD-specific measures. are not without controversy. Initial research has dem-
Although ASD-specific diagnostic measures assess for the onstrated that the DSM-5 diagnostic criterion for ASD
core deficits of ASD, they do not assess functional skills or results in increased specificity when compared with
limitations, language level, cognitive functioning, behavior DSM-IV-TR, which may reduce the number of children
problems, comorbid psychopathology, or other health con- who are diagnosed as having ASD when they do not (Fra-
ditions. Determining the level of ASD severity requires a zier et al. 2012). However, there has also been concern
comprehensive diagnostic battery of measures to more thor- that the DSM-5 criteria may not identify individuals pre-
oughly capture the individual’s level of functioning across a viously diagnosed with ASD (Mazurek et al. 2017). In
variety of domains (Mehling and Tasse 2016). addition, the use of severity specifiers was proposed to
Further research has indicated that even with standardized convey important information about symptom course and
assessments, there are inconsistencies between severity clas- prognosis. This shift emphasizes “disability in context”
sifications based on autism symptoms, cognitive skills, and and allows for the provision of practical information for
adaptive functioning (Weitlauf et al. 2014). These findings practitioners regarding intervention planning and improv-
highlight the need for further clarification within the diag- ing quality of life (Mehling and Tasse 2016). However,
nostic criteria to classify the level of support needed in SC current use of severity specifiers for SC and RRB relies
and RRB for individuals with ASD. If we cannot quantify heavily on clinical judgement. Currently, there is a need
methods for accurately and reliably determining levels of to improve upon the operationalization of the specifiers to
severity, we risk repeating our history of discrepancies in assure high agreement in the application of severity ratings
the application of ASD diagnostic categories from one site between professionals across sites.
to another. A few recent investigations have examined reliability
Hus et al. (2014) examined calibrated raw totals from and clinical correlates of level of support ratings. Taylor
the autism diagnostic observation schedule (ADOS) social et al. (2017) found poor agreement in clinicians’ ratings of
affective (SA) and restricted repetitive behavior (RRB) level of support. For 27 clinicians viewing seven videotapes,
domains to provide a metric of symptom severity by domain. agreement for SC ratings ranged from 10 to 100% and agree-
Results indicated a small but significant association between ment for RRB ratings ranged from 0 to 100%. Craig et al.
calibrated SA scores and verbal IQ, accounting for just under (2017) compared SC level of support groups with 3- to
11% of variance in the calibrated SA score. Verbal IQ, non- 6-year-olds diagnosed with ASD and found significant cor-
verbal IQ, and race emerged as small but significant predic- relations between higher level of support ratings and lower
tors of calibrated RRB scores, explaining less than 5% of adaptive communication, motor functioning, and maladap-
variance. Domain calibrations provided a measure of ASD tive behavior. When group differences were tested, children
severity that was less influenced by child characteristics in the Level 3 SC support group (n = 6; 12%) demonstrated
(e.g., age, language skills) than use of domain specific raw greater motor impairments and more maladaptive behavior
scores. Authors proposed that these calibrated scores may when compared to Level 1 (n = 25; 50%) and 2 (n = 19; 38%)
provide a better measure of ASD symptom severity which is groups, which did not significantly differ from each other.
relatively independent from child characteristics.
Although metrics assessing core ASD symptomology
are an important component in the determination of level Purpose of the Study
of severity, these metrics alone do not account for the
impact of cognitive, language, behavioral and adaptive The purpose of the current study is to determine how indi-
functioning in determining ASD severity levels. Mehling viduals with mixed levels of impairments across cognitive,
and Tasse (2016) found considerable variability in prac- adaptive, behavioral, and ASD specific-symptom domains
tices utilized to assess and quantify ASD severity. A sig- were classified in terms of DSM-5 levels of support in both
nificant number of investigations (n = 72) operationalized SC and RRB as a result of a comprehensive diagnostic
autism severity using one measure, while the remaining evaluation.

13
Journal of Autism and Developmental Disorders

Methods diagnosis was determined by one of four licensed psy-

chologists. Participants in this study were diagnosed with
Participants ASD based on current DSM-5 diagnostic criteria, and no
participants received a diagnosis based on a previously
Participants were 165 children and adolescents who well-established DSM-IV diagnosis. Participants were
received a primary diagnosis of ASD following a com- selected from a group of consecutive referrals for children
prehensive psychological evaluation completed by a with noted delays in development whose families sought
licensed psychologist, or a multidisciplinary team includ- comprehensive psychological evaluations at a center for
ing a licensed psychologist. Each comprehensive psycho- developmental disabilities in a large city in the southern
logical evaluation included a structured diagnostic inter- United States. The study was reviewed and approved by
view, behavioral observations, caregiver report on various the center’s Institutional Review Board. Participant char-
behavior rating forms, standardized testing of intelligence acteristics are presented in Table 1.
and adaptive skills, as well as autism-specific testing. Final

Table 1  Participant Variable Min Max M SD Skew Kurtosis n %

characteristics (N = 165)
Age (months) 20 185 52.15 25.09 1.805 5.36
Full scale IQ 40 122 65.23 17.27 .50 − .25
Adaptive ­behaviora 12 104 65.98 16.39 − .59 .68
CARS-2 total raw ­scoreb 24 54 37.04 5.80 .39 .08
ADOS-2 total raw score 6 28 19.10 4.91 − .41 − .44
 Module T (n = 25) 12 28 22.04 4.38 − .80 − .45
 Module 1 (n = 78) 10 28 20.38 4.03 − .46 − .08
 Module 2 (n = 49) 6 27 16.24 4.89 − .10 − .39
 Module 3 (n = 13) 9 24 16.54 4.68 .05 − 1.15
ADOS-2 comp score 3 10 7.97 1.69 − .44 − .54
SACSS 3 10 7.64 1.82 − .36 − .48
RRBCSS 1 10 8.03 1.80 − 1.25 2.51
Gender
 Male 137 83.0
 Female 28 17.0
Race/ethnicity
 Black/African-American 76 46.1
 White/Caucasian 73 44.2
 Asian-American 6 3.6
 Other 6 3.6
 Hispanic 3 1.8
 Missing 1 0.6
Social communication levels of support
 Level 1 (“requiring support”) 32 19.4
 Level 2 (“requiring substantial support”) 75 45.5
 Level 3 (“requiring very substantial support”) 58 35.2
Restrictive repetitive behavior level of support
 Level 1 (“requiring support”) 43 26.1
 Level 2 (“requiring substantial support”) 81 49.1
 Level 3 (“requiring very substantial support”) 41 24.8

CARS-2 childhood autism rating scale-2 raw scores, ADOS-2 autism diagnostic observation schedule-2,
ADOS-2 Comp autism diagnostic observation schedule-2 comparison score, SACSS social affect calibrated
severity score, RRBCSS restrictive and repetitive behavior calibrated severity score
a
 Scores missing for four participants
b
 Scores missing for nine participants

13
 Journal of Autism and Developmental Disorders

Measures reporting, and the child’s performance on other testing

measures. The rating form includes 15 areas of behaviors
Cognitive Functioning that are defined by a rating system developed to identify
symptoms of autism. The CARS-2 yields a standard score
Standardized testing of intelligence was completed using based on a clinical sample of those with ASD, which
either the cognitive and language scales of the Bayley reflects the level of autism-related behaviors present in
Scales of Infant and Toddler Development, Third Edition an individual at the time of evaluation. The CARS-2 was
(Bayley-III; Bayley 2005) or the Stanford Binet Intelligence completed by the licensed psychologist completing the
Scales, Fifth Edition (SB-5; Roid 2003). The Bayley-III was diagnostic evaluation to identify the presence of behav-
administered to younger participants, when the initial start- iors related to ASD.
ing point on the SB-5 subtests were too difficult for the child
and basal rules were not met. For typically developing pre-
schoolers, the Bayley-III and SB-5 yield summary scores Data Analysis
that consistently fall within 95% confidence intervals of each
scale (Kamppi and Gilmore 2010). Our apriori data analytic plan consisted of a 3 (SC Level
of Support) × 3 (RRB Level of Support) multivariate
Adaptive Functioning analysis of variance (MANOVA) to analyze six depend-
ent variables: (a) cognitive functioning, (b) adaptive
Standardized testing of adaptive skills was completed by functioning, (c) CARS-2 scores, (d) ADOS-2 Comp (e)
structured caregiver interview using the Vineland Adaptive SACSS scores, and (f) RRBCSS scores. In the presence
Behavior Scales, Second Edition (Vineland-II; Sparrow et al. of significant MANOVA omnibus findings, we planned
2005), and the Early Development Form of the Scales of direct discriminant analysis procedures (i.e., all variables
Independent Behavior-Revised (SIB-Bruininks et al. 1996). considered simultaneously; Tabachnick and Fidell 2007) to
identify variables that significantly discriminated between
Autism Diagnostic Observation Schedule, Second Edition groups. Four participants were missing adaptive data and
(ADOS‑2; Lord et al. 2012) nine were missing CARS-2 data; i.e., 152 of 165 partici-
pants with complete data, 13 (7.9%) missing data. Little’s
The ADOS-2 was used to assess for symptoms of ASD. The Missing Completely at Random (MCAR) test was non-sig-
ADOS-2 is a play-based assessment consisting of standard nificant, χ2 (14) = 7.91, p = .89, indicating that data were
social interactions and activities that allow examiners to not missing in a systematic manner. We handled miss-
observe behaviors that have been identified as important ing data using the Markov chain Monte Carlo imputation
to the diagnosis of ASD. This instrument consists of semi- method (IBM Corporation 2013) for estimating adaptive
structured and unstructured situations that allow the indi- and CARS-2 scores resulting in a single imputed data set
vidual to both initiate social communicative behaviors and used in the analysis.
respond to social cues or overtures from the examiner. The
ADOS-2 was administered by a licensed psychologist as part
of the diagnostic evaluation.
Results
Calibrated Comparison and Severity Scores  ADOS-2 com-
parison scores (ADOS-2 Comp) were calculated using Roughly 46% of the sample was rated as needing Level 2
the ADOS-2 manual for modules 1, 2, and 3 (Lord et  al. support in the area of SC; approximately 49% of the sam-
2012). Social affect calibrated severity scores (SACSS), ple was rated as needing Level 2 support in the area of
and restrictive and repetitive behavior calibrated severity RRB. A Chi square test of independence revealed a sig-
scores (RRBCSS) were calculated using Hus et  al. (2014, nificant relationship between SC and RRB coding, χ2 (4,
p. 2404) algorithms. For the Toddler Module, we calculated N = 165) = 106.85, p < .001, Spearman’s ρ = .71, p < .001;
ADOS-2 Comp, SACSS, and RRBCSS scores using Esler no individuals with ASD were identified as needing Level
et al. (2015, p. 2710) algorithms. 1 support in one category and Level 3 support in another
category. Severity groups did not differ according to gen-
Childhood Autism Rating Scale, Second Edition (CARS‑2; der, [SC: χ 2 (2, N = 165) = 4.78, p = .09; RRB: χ 2 (2,
Schopler et al. 2010) N = 165) = 2.65, p = .27], race, [SC: χ2 (8, N = 164) = 4.31,
p = .83; RRB: χ2 (8, N = 164) = 7.57, p = .48], or age, [F val-
The CARS-2 is a rating form completed by the clini- ues ranged from 0.38 to 1.85, p = .16–.67, for SC and RRB
cian that is informed by clinical observation, caregiver main effects and their interaction].

13
Journal of Autism and Developmental Disorders

MANOVA Results and RRBCSS scores were variably related and ranged from
.83 (ADOS-2 Comp/SACSS; p < .001) to .11 (SACSS/
Prior to conducting the MANOVA, we examined whether RRBSS, ns).
the data met the assumption of multivariate normality Given the absence of individuals with Level 1 × Level 3
(MVN) by using Mardia’s test for multivariate skew and classifications and the small numbers of participants in SC
kurtosis via DeCarlo’s (1997) SPSS macro. Mardia’s test Level 1 × RRB Level 2 (n = 6) and SC Level 2 × RRB Level
revealed that the assumption of MVN was violated: skew, 3 (n = 7) cells, we elected to evaluate only main effects of
b1,p = 5.21, χ2 = 143. 31, p < .001, and kurtosis, b2,p = 52.96, SC and RRB using Type III sum of squares. The MANOVA
p = .001. We identified a single multivariate outlier using resulted in a main effect for SC severity, Wilks’s Λ = .63,
Mahalanobis’ distance (p < .001); after removing this case, F(12, 296) = 6.33, p < .000, 𝜂p2 = .20, and main effect for
the MVN assumption remained untenable: skew, b1,p = 4.11, RRB severity, Wilks’s Λ = .85, F(12, 296) = 2.14, p = .015,
χ2 = 112. 46, p < .001, and kurtosis, b2,p = 49.51, p = .32. 𝜂p2 = .08. The findings indicated that, across severity groups,
Due to continued violation of MVN, we examined uni- there were significant mean differences on the combination
variate distributions and identified six outliers across four of six clinical variables we entered into the analysis.
variables: cognitive, adaptive, CARS-2 and RRBCSS. After
removing the six additional cases, the dataset resulted in Direct Discriminant Analysis Results
158 participants and met MVN assumptions: skew, b1,p =
2.72, χ2 = 71. 67, p = .08, and kurtosis, b2,p = 46.05, p = .21. Social Communication
Next, we examined whether the data met the assumption
of homogeneity of covariance matrices using Box’s test of To interpret the patterns of difference across the SC severity
equality of covariance matrices. Based upon Tabachnick and groups, we performed direct discriminant analysis (DDA)
Fidell’s (2007) guideline (i.e., p < .001 indicating significant using SPSS DISCRIMINANT (IBM Corporation 2013). The
heterogeneity), the assumption of homogeneity of covari- six clinical variables were entered as predictors of mem-
ance matrices was met, Box’s M = 178.08, p = .006. bership in SC severity level. Two discriminant functions
Descriptive data for all dependent variables are presented were calculated, resulting in a combined Wilks’s Λ = .46, χ2
in Tables 2, 3 and 4. Correlations between study variables (12) = 119.55 p < .001, 𝜂p2 = .23. After the first function was
are presented in Table 5. The magnitude of correlations were removed from the analysis, significant association remained
deemed acceptable for the use of MANOVA. Correlations between SC support levels and predictors, Wilks’s Λ = .92,
between ADOS-2 scores and cognitive and adaptive scores χ2 (5) = 13.57, p = .018, 𝜂p2 = .03. The first discriminant
ranged from − .14 to − .05 (all ns); ADOS-2 Comp, SACSS, function accounted for 50.13% of between-group variance,

Table 2  Descriptive statistics for full scale IQ and adaptive composite scores across social communication and restrictive, repetitive behavior
levels of support (N = 158)
SC level of support Full scale IQ
RRB level of support
Level 1 Level 2 Level 3 Row total
M (SD) M (SD) M (SD)

 Level 1 83.8 (10.9) 76.0 (21.1) – 82.3 (13.7)

 Level 2 66.2 (13.4) 64.0 (15.4) 64.4 (12.2) 64.5 (14.6)
 Level 3 – 58.4 (11.7) 52.0 (9.8) 54.6 (11.1)
 Column total 77.2 (14.6) 63.2 (15.4) 54.2 (11.2) 64.5 (16.4)
SC level of support Adaptive composite
RRB level of support
Level 1 Level 2 Level 3 Row total
M (SD) M (SD) M (SD)

 Level 1 78.9 (10.7) 69.7 (23.8) – 77.1 (14.2)

 Level 2 71.2 (14.8) 68.2 (14.7) 61.6 (14.1) 68.3 (14.7)
 Level 3 – 58.9 (11.6) 57.7 (12.6) 58.2 (12.1)
 Column total 76.73 (12.7) 65.5 (15.1) 58.4 (12.8) 66.5 (15.3)

SC social communication, RRB restrictive and repetitive behavior

13
 Journal of Autism and Developmental Disorders

Table 3  Descriptive statistics for childhood autism rating scale-2 (CARS-2) raw scores and autism diagnostic observation schedule-2 (ADOS-2)
comparison scores across social communication and restrictive, repetitive behavior levels of support (N = 158)
SC level of support CARS-2 total raw scores
RRB level of support
Level 1 Level 2 Level 3 Row total
M (SD) M (SD) M (SD)

 Level 1 32.1 (3.9) 33.4 (4.5) – 32.4 (4.0)

 Level 2 35.7 (5.0) 35.8 (4.6) 38.9 (3.7) 36.1 (4.7)
 Level 3 – 40.3 (4.3) 40.7 (5.6) 40.5 (5.0)
 Column total 33.5 (4.6) 37.0 (5.0) 40.3 (5.3) 36.9 (5.5)
SC level of support ADOS-2 comparison scores
RRB level of support
Level 1 Level 2 Level 3 Row total
M (SD) M (SD) M (SD)

 Level 1 7.0 (1.8) 6.8 (1.2) – 7.0 (1.7)

 Level 2 8.1 (1.5) 8.1 (1.6) 7.7 (1.5) 8.0 (1.5)
 Level 3 – 8.3 (1.5) 8.8 (1.4) 8.6 (1.5)
 Column total 7.4 (1.8) 8.0 (1.6) 8.6 (1.4) 8.0 (1.6)

SC social communication, RRB restrictive and repetitive behavior

Table 4  Descriptive statistics for ADOS-2 SACSS and RRBCSS across social communication and restrictive, repetitive behavior levels of sup-
port (N = 158)
SC level of support ADOS-2 SACSS scores
RRB level of support
Level 1 Level 2 Level 3 Row total
M (SD) M (SD) M (SD)

 Level 1 6.7 (1.9) 6.3 (0.8) – 6.6 (1.8)

 Level 2 7.9 (1.9) 7.5 (1.7) 6.3 (2.4) 7.5 (1.9)
 Level 3 – 8.4 (1.5) 8.4 (1.4) 8.4 (1.5)
 Column total 7.2 (1.9) 7.7 (1.7) 8.0 (1.8) 7.6 (1.8)
SC level of support ADOS-2 RRBCSS scores
RRB level of support
Level 1 Level 2 Level 3 Row total
M (SD) M (SD) M (SD)

 Level 1 7.2 (1.5) 7.8 (1.9) – 7.4 (1.6)

 Level 2 7.7 (1.5) 8.7 (1.3) 9.1 (0.7) 8.5 (1.4)
 Level 3 – 7.5 (1.4) 8.6 (1.5) 8.1 (1.5)
 Column total 7.4 (1.5) 8.3 (1.5) 8.7 (1.4) 8.2 (1.5)

SACSS social affect calibrated severity score, RRBCSS restrictive and repetitive behavior calibrated severity score

the second discriminant function accounted for 8.53% of adaptive functioning, SACSS, and ADOS-2 Comparison
between-group variance. scores. For the second discriminant function, the single
The structure matrix of correlations between predic- significant predictor was RRBCSS. We contrasted centroid
tors and discriminant functions is presented in Table 6. means for both discriminant functions via separate uni-
Using the convention of |.33| (Tabachnick and Fidell 2007, variate analysis of variance and Bonferroni adjustment for
p. 400), significant predictors for the first discriminant follow-up contrasts (see Table 7). For the first discriminant
function were: cognitive functioning, CARS-2 scores,

13
Journal of Autism and Developmental Disorders

Table 5  Correlations between 1 2 3 4 5 6
study variables (N = 158)
1. IQ –
2. Adaptive .52** –
3. CARS-2 − .38** − .46** –
4. ADOS-2 comp − .11 − .05 .44** –
5. SACSS − .12 − .10 .41** .83** –
6. RRBCSS − .14 − .02 .21* .50** .11 –

N = 158, ADOS-2 Comp autism diagnostic observation schedule, 2nd edition comparison score, SACSS
social affect calibrated severity scores, RRBCSS restrictive and repetitive behavior calibrated severity
scores
*p < .01; **p < .001

Table 6  Results of direct discriminant analysis of clinical variables (N = 158)

Variable SC support level
Correlations of variables with discriminant functions
1 2

Cognitive .74 − .29

CARS-2 − .64 − .31
Adaptive .51 .22
SACSS − .38 − .18
ADOS-2 comparison − .37 .17
RRBCSS − .16 .81
 Canonical R .71 .29
 Eigenvalue 1.00 .09
Variable RRB support level
Correlations of variables with discriminant functions
1 2

Cognitive .75 − .10

CARS-2 − .62 − .48
Adaptive .59 − .05
RRBCSS − .41 .34
ADOS-2 comparison − .34 − .18
SACSS − .22 − .02
 Canonical R .62 .12
 Eigenvalue .61 .02

N = 158, ADOS-2 Comp autism diagnostic observation schedule, 2nd edition comparison score, SACSS social affect calibrated severity scores,
RRBCSS restrictive and repetitive behavior calibrated severity scores

function, group centroids differed significantly, F(2, Restrictive Repetitive Behavior

155) = 77.76, p < .001; the Level 1 group scored signifi-
cantly higher than the Level 2 and Level 3 groups. Level We used identical DDA procedures for the RRB levels of
2 and Level 3 groups were also significantly different. For support. Two discriminant functions were calculated, with a
the second discriminant function, group centroids differed combined Wilks’s Λ = .61, χ2 (12) = 75.23 p < .001, 𝜂p2 = .15.
significantly, F(2, 155) = 7.23, p = .001; the Level 2 group After the first function was removed from the analysis, there
scored significantly higher than the Level 1 and Level 3 was no significant association between RRB severity groups
groups, which did not differ. and predictors, Wilks’s Λ = .99, χ2 (5) = 2.27, p = .810, 𝜂p2
= .01. The first discriminant function accounted for 38.1%

13
 Journal of Autism and Developmental Disorders

Table 7  Group means for clinical variables and discriminant functions (N = 158)

Variable SC support level
Level 1 Level 2 Level 3 UnivariateA p

Cognitive 82.29a 64.48b 54.75c 14.54 < .001

Adaptive 77.13a 68.32b 58.19c 4.95 .008
CARS-2 32.39a 36.08b 40.51c 11.69 < .001
ADOS-2 comparison 6.97a 8.03b 8.57b 5.05 .008
SACSS 6.64a 7.45a 8.41b 10.20 < .001
RRBCSS 7.35a 8.53b 8.14b 6.12 .003
Discriminant function 1 1.69a 0.12b − 1.09c 77.76 < .001
Discriminant function 2 − 0.33a 0.33b − 0.24a 7.23 .001
Variable RRB support level
Level 1 Level 2 Level 3 UnivariateA p

Cognitive 77.20 63.20 54.23 1.70 .187

Adaptive 76.30 65.46 58.38 1.98 .414
CARS-2 33.49 36.98 40.35 0.81 .445
ADOS-2 comparison 7.40 8.03 8.59 0.25 .782
SACSS 7.18 7.67 8.03 1.33 .267
RRBCSS 7.40a 8.29b 8.69b 7.46 .001
Discriminant function 1 1.18a − 0.11b − 0.99c 47.54 < .001
Discriminant function 2 − 0.10 0.12 − 0.14 1.16 .315
A
 F values from MANOVA; for primary variables, F(2, 153), for centroid contrasts, F(2, 155). Within rows, means with different superscripts
differ at p < .05

of between-group variance, the second discriminant func- functioning, adaptive behavior, and indicators of symptom
tion accounted for 1.5% of between-group variance. The severity.
structure matrix of correlations between predictors and
discriminant functions is presented in Table 6. Significant Review of Main Findings
predictors for the first discriminant function were: cogni-
tive functioning, CARS-2 scores, adaptive functioning, Roughly half of individuals fell within the Level 2 support
RRBCSS, and ADOS-2 Comparison scores. For the first rating for both SC (45.5%) and RRB (49%); and level of
discriminant function, group centroids differed significantly, support ratings were significantly correlated. In no case were
F(2, 155) = 47.54, p < .001; the Level 1 group scored signifi- individuals identified with a Level 1 rating in one domain
cantly higher than the Level 2 and Level 3 groups. Level 2 and a Level 3 rating in the other domain. Therefore, approxi-
and Level 3 groups were also significantly different. mately half of the sample was determined to need “substan-
tial support” in both SC and RRB.
Discriminant function analysis for SC ratings shows
a general progression from greater cognitive ability and
Discussion adaptive functioning between Level 1 and Level 3 groups
(i.e., Level 1 > Level 2 > Level 3). The reverse pattern was
The introduction of levels of support ratings in DSM-5 has observed for ASD severity ratings across several measures,
raised questions regarding their use and meaning in applied including the ADOS-2 and CARS-2 (i.e., Level 1 < Level
clinical settings. We sought to contribute to the knowledge 2 < Level 3). The general pattern is of clinical importance
base regarding the use of these severity ratings by compar- as the cognitive and adaptive variables were unrelated to
ing clinical measures across different severity groups. In one ADOS-2 symptom scores. For the first discriminant func-
respect, this study provides insight into the type of informa- tion, SC groups showed an expected pattern for the SACSS
tion clinicians may use to derive severity ratings. On the and RRBCSS scores with SACSS scores contributing sig-
other hand, this study aims to examine the usefulness of the nificantly while RRBCSS scores did not. The second dis-
severity ratings by evaluating clinical differences between criminant function showed an unexpected pattern with the
severity groups on meaningful indicators, such as cognitive Level 2 group demonstrating higher scores on a composite

13
Journal of Autism and Developmental Disorders

variable comprised largely of the RRBCSS. The Level 1 As RRBs may only occur in particular situations (e.g.,
SC group demonstrated higher cognitive, higher adaptive, hand flapping when frustrated or excited, covering ears in
and lower ASD severity when compared to the Level 2 and response to loud noises in the environment), it is more dif-
Level 3 SC groups. ficult to assess their presence and severity over a short period
Discriminant function analysis for RRB ratings showed a of time. Caregiver reporting on standardized assessment
similar pattern to the SA ratings, with a general gradation of tools may provide more information regarding these behav-
higher adaptation/lower symptom severity with ratings pro- iors than direct observation by the clinician given the time-
gressing from Level 1 to Level 3 supports. Of interest, how- limited nature of diagnostic evaluations. As such, including
ever, is the different contributions of SACSS and RRBCSS a specific measure of restricted and repetitive behaviors as
to the scores. For RRB support ratings, RRBCSS contrib- part of a diagnostic battery would likely provide important
uted to the significant discriminant function while SACSS information when determining appropriate levels of support
scores did not. SACSS and RRBCSS scores were unrelated for RRB.
in our sample (r = .11, ns), which is generally consistent with Additional research has suggested that girls with ASD
others’ findings (e.g., r = .28 for Toddler Module, Esler et al. may present with different types of restricted interests in
2015; r = .25 for Modules 1–3; Hus et al. 2014). comparison to boys with ASD (Hiller et al. 2014). In par-
The MANOVA findings suggest that SC level of support ticular, findings from Hiller, Young, and Weber (2016) sug-
ratings account for more variability (~ 20%) in the clinical gest that the nature of restricted/repetitive behaviors in boys
findings when compared to RRB level of support ratings and girls with ASD differ. More specifically, results sup-
(~ 8%). This could also be a function of the clinical assess- ported that boys were more likely to demonstrate non-func-
ment conducted in this setting as no specific measures of tional use of toy cars (e.g., watching wheels spin, lining up
restrictive and repetitive behavior were included in the diag- toy cars), whereas girls were more likely to show restricted
nostic battery. interests in specific toys or objects. These differences may
make it more difficult to assess for RRB in girls given that
Clinical Implications current assessment measures may not adequately elicit RRBs
that are more typically presented by females with ASD. As
Results of the present study support the utility of assessing such, clinicians’ ability to accurately determine appropriate
autism specific symptomology in combination with stand- levels of support for RRB is limited not only by the short
ardized assessment of cognitive skills and adaptive func- duration of the diagnostic evaluation, but also the RRBs
tioning in the determination of levels of support by domain that the standardized measures assess for, and the potential
for individuals with ASD. As cognitive and adaptive scores differences in presentation of RRBs by gender. In addition
were found to be unrelated to ADOS-2 symptom scores, to potential differences in presentation of RRBs by gender,
the importance of a comprehensive assessment battery that these behaviors may also change in presentation throughout
includes standardized assessment in cognitive and adaptive an individual’s development. As the DSM-5 diagnostic cri-
domains in addition to ASD-specific measures is supported. teria requires that deficits are present currently, or by history,
In general, findings indicated that children who presented it is probable that levels of support may change given the
with higher cognitive and adaptive skills, and lower ratings current intensity and severity of the individual’s presentation
on autism-specific measures, received ASD diagnoses that of RRBs over the course of development.
clinicians’ determined warranted lower levels of support rat-
ings for SC and RRB. Approximately half of the sample was Directions for Future Research
determined to need Level 2 supports in both SC and RRB,
and there was a high correlation between the determined In addition to examining the utility of adding a specific
level of support for SC and RRB across the entire sample. measure of restricted and repetitive behaviors to a diagnostic
battery in determination of RRB level of support, there are a
Limitations number of directions for future research. The current study
documented higher cognitive and adaptive profiles combined
The investigation is limited by the use of clinical data that with lower ASD-specific symptomology resulted in lower
was used to inform support ratings. As such, there is a poten- levels of rating on support for SC and RRB. However, addi-
tial problem with tautology, i.e., the level of support ratings tional studies are needed to determine if these findings are
simply reflect severity data, such as the CARS-2, that were replicated across sites, and the consistency with which clini-
used to derive ratings. The pattern of findings suggests oth- cian’s use relevant assessment information in determining
erwise, however, as cognitive and adaptive functioning were levels of support. Once more research has been completed
among the most significant areas associated with severity regarding the utility of designating levels of support for
level, particularly for SC severity ratings. ASD, a formalized system for determining levels of support

13
 Journal of Autism and Developmental Disorders

in a systematic manner would be highly beneficial in aid- Author Contributions  LMG conceived of the study, participated in its
ing clinician’s in quantifying recommendations for levels design and coordination and drafted the manuscript; JMC participated
in the design, interpretation of the data, and helped to draft the manu-
of support. script; LM participated in the design and coordination of the study and
Furthermore, future investigations should assess if clini- performed the measurement; BK participated in the performed data
cians are making domain specific designations related to collection. All authors read and approved the final manuscript.
levels of support in SC and RRB as written in DSM-5. Col-
lecting information about clinical practice with the DSM-5 Funding  The funding was provided by Maternal and Child Health
Bureau (Grant No. MC00038-25) and Administration on Intellectual
criteria will provide valuable information in determining and Developmental Disabilities (Grant No. 90DD0003-04-01).
whether these designations are being used at all by diagnos-
tic clinicians, and if designating levels of support in SC and
RRB is perceived to have value and utility in the diagnosis
of ASD. Finally, future research should assess how service References
providers use the designated SC and RRB levels of support
to inform individualized services and long term intervention American Psychiatric Association. (2013). Diagnostic and statistical
manual of mental disorders. (5th edn.).Washington, DC: Ameri-
planning.
can Psychiatric Association.
If future research indicates that levels of support in SC Bayley, N. (2005). Bayley scales of infant and toddler development
and RRB can be reliably determined across clinical sites, third edition, San Antonio, TX: The Psychological Corporation.
further research will be needed to inform whether these lev- Bruininks, R. H., Woodcock, R. W., Weatherman, R. F., & Hill, B.
K. (1996). Scales of independent behavior—revised. Itasca, IL:
els of support remain stable across the course of an indi-
Riverside Publishing.
vidual’s development. It may be that age of diagnosis, and Craig, F., Fanizza, I., Russo, L., Lucarelli, E., Lorenzo, A., Pasca, M.
access to intervention services may impact the trajectory of G., et al. (2017). Social communication in children with autism
a child’s development, and subsequently impact the levels spectrum disorder (ASD): Correlation between DSM-5 and autism
classification system of functioning-social communication. Autism
of support needed over time. As such, future research will
Research. https​://doi.org/10.1002/aur.1772.
likely need to investigate whether a system that determines DeCarlo, L. T. (1997). On the meaning and use of kurtosis. Psychologi-
levels of severity across the population of individuals with cal Methods, 2, 292–307.
ASD, regardless of age, is feasible. Esler, A. N., Bal, V. H., Guthrie, W., Wetherby, A., Weismer, S. E.,
& Lord, C. (2015). The autism diagnostic observation sched-
ule, Toddler Module: Standardized severity scores. Journal of
Summary Autism and Developmental Disorders, 46, 2000–2016. https:​ //doi.
org/10.1007/s1080​3-015-2432-7.
The DSM-5 criteria for diagnosing ASD features level of Frazier, T. W., Youngstrom, E. A., Speer, L., Embacher, R., Law, P.,
Constantino, J., et al. (2012). Validation of proposed DSM-5 cri-
support ratings for SC and RRB. The levels require diagnos-
teria for autism spectrum disorder. Journal of the American Acad-
ticians to use their clinical judgement to delineate between emy of Child & Adolescent Psychiatry, 51(1), 28–40.
three classifications. A clinician’s determination of levels Gotham, K., Pickles, A., & Lord, C. (2012). Trajectories of autism
of support may aide in the identification of areas of rela- severity in children using standardized ADOS scores. Pediatrics,
130(5), e1278–e1284.
tive strengths and weaknesses as they relate to ASD core
Hiller, R. M., Young, R. L., & Weber, N. (2014). Sex differences in
symptomology, and facilitate individualized intervention autism spectrum disorder based on DSM-5 criteria: Evidence
planning. However, the criteria for determining level of sup- from clinician and teacher reporting. Journal of Abnormal Child
port in SC and RRB currently lack clear guidance or prac- Psychology, 42, 1381–1393.
Hiller, R. M., Young, R. L., & Weber, N. (2016). Sex differences in
tice recommendations for distinguishing between the levels
pre-diagnosis concerns for children later diagnosed with autism
based on quantifiable data, and at present rely on the clinical spectrum disorder. Autism, 20(1), 75–84.
judgement of the diagnostician. The present study contrasted Hus, V., Gotham, K., & Lord, C. (2014). Standardizing ADOS domain
cognitive, adaptive, and autism severity scores across SC scores: Separating severity of social affect and restricted and
repetitive behaviors. Journal of Autism and Developmental Disor-
and RRB groups for individuals with ASD diagnosed in a
ders, 44, 2400–2412. https:​ //doi.org/10.1007/s10803​ -012-1719-1.
tertiary developmental disabilities clinic. Findings indicated IBM Corporation. (2013). IBM SPSS advanced statistics version 22.
that roughly half of the individuals were identified as need- Chicago: IBM Corporation.
ing Level 2 support for both SC and RRB. No individuals Kamppi, D., & Gilmore, L. (2010). Assessing cognitive development in
early childhood: A comparison of the Bayley-III and the Stanford-
were rated as needing a combination of Level 1/Level 3 sup-
Binet Fifth Edition. Australian Educational and Developmental
ports across domains. Both SC and RRB groups showed a Psychologist, 27, 70–75.
graded pattern of higher adaptation/lower autism severity Lord, C., Rutter, M., DiLavore, P. C., Risi, S., Gotham, K., & Bishop,
to lower adaptation/higher autism severity from Level 1 to S. L. (2012). Autism diagnostic observation schedule second edi-
tion. Los Angeles: Western Psychological Services.
Level 3. Further research into the determination and use of
Mazurek, M. O., Lu, F., Symecko, H., Butter, E., Bing, N. M., Hund-
levels of support by clinicians in the diagnosis of ASD and ley, R. J., et al. (2017). A prospective study of the concordance
the perceived utility of these designations is warranted. of DSM-IV and DSM-5 diagnostic criteria for autism spectrum

13
Journal of Autism and Developmental Disorders

disorder. Journal of Autism and Developmental Disorders, 44, Tabachnick, B. G., & Fidell, L. S. (2007). Using multivariate statistics
2400–2412. https​://doi.org/10.1007/s1080​3-017-3200-7. 5th ed. New York: Pearson.
Mehling, M. H., & Tasse, M. J. (2016). Severity of autism spectrum Taylor, L. J., Eapen, V., Maybery, M., Midford, S., Paynter, J.,
disorders: Current conceptualization, and transition to DSM-5. Quarmby, L., et al. (2017). Brief report: An exploratory study of
Journal of Autism and Developmental Disorders, 46, 2000–2016. the diagnostic reliability for autism spectrum disorder. Journal of
https​://doi.org/10.1007/s1080​3-016-2731-7. Autism and Developmental Disorders, 47, 1551–1558. https:​ //doi.
Roid, G. H. (2003). Stanford binet intelligence scales: fifth edition. org/10.1007/s1080​3-017-3054-z.
Rolling Meadows, IL: Riverside Publishing. Weitlauf, A. S., Gotham, K. O., Vehorn, A. C., & Warren, Z. E. (2014).
Schopler, E., Van Bourgondien, M. E., Wellman, G. J., & Love, S. R. Brief report: DSM-5 “Levels of Support:” A comment on discrep-
(2010). Childhood autism rating scale second edition. Los Ange- ant conceptualizations of severity in ASD. Journal of Autism and
les: Western Psychological Services. Developmental Disorders, 44, 471–476. https​://doi.org/10.1007/
Sparrow, S., Cicchetti, D., & Balla, D. (2005). Vineland adaptive s1080​3-013-1882-z.
behavior scales second edition. Circle Pines, MN: American
Guidance Service, Inc.

13