developmental; week 2; identifying atypical development

Multiple reasons for atypical development

• Pre-natal effects (e.g. exposure to teratogen)

  • Fetal Alcohol Spectrum Disorder

• Environmental effects (e.g. complications during birth)

  • Cerebral Palsy

• Genetic effects

  • Hereditary

  • Spontaneous mutations (e.g. Copy Number Variants)

• Unknown (likely multifaceted) effects:

  • Autism Spectrum Conditions

  • ADHD

Genetic reasons for atypical development

DNA > GENES > CHROMSOME > CELL

genetic abnormalities = too many or too few of particular genes resulting from:

• extra chromosome

  • e.g. Down’s Syndrome

• duplication of a certain part of a chromosome

  • e.g. 16p11.2

• deletion of a certain part of a chromosome

  • e.g. William’s syndrome

What is atypical development?

• Difficult to define “atypical development” in the context of

  a)Individual differences in the rate of development

  b)Individual differences in people’s traits, strengths and weaknesses

• Textbook Definition = “The extremes of individual differences in development”

• Can include: advanced development and delayed development

• Generally associated with neurodevelopmental conditions (e.g. autism spectrum conditions, ADHD, William’s syndrome, intellectual disability, etc).

There are numerous ways that development can be atypical

Developmental regression

• Regression is typically seen in children with  autism spectrum conditions and / or intellectual disability.

• Regression is defined as a period where a particular skill is developing along a typical trajectory, but then a child loses aspects of this skill, e.g. stops speaking in two word phrases.

• Most often seen in language and in motor skills.

• 

Development occurs across multiple domains

• adaptive behaviour

  • daily living skills

  • independence

  • personal responsibility

  • managing money

  • personal safety

  • ability to work

  • functional decision making

• social

  • gestures

  • emotional IQ

  • turn-taking

  • non-verbal communication

  • social interactions

  • verbal communication

  • empathy

  • reciprocal eye contact

• cognitive

  • memory

  • IQ

  • attention

  • language

  • executive function

  • numerical ability

• physical

  • facial dysmorphism

  • macrocephaly

  • physical features

    • e.g. heart

  • microcephaly

• motor skills

  • fine motor skills

  • balance

  • gross motor skills

  • activity level

  • coordination

How can we identify if development is atypical?

• the normal distribution

  • obtained by testing many (100+s) participants

  • for many variables (e.g. height, weight, IQ, other cognitive abilities) samples from the population generate a normal distribution

  • ‘normal distribution’ - ‘normal curve’ = ‘bell shaped curve’

Identifying (and measuring) atypical development

• Group comparisons against a representative (or ‘normative’) sample.

• It is important to choose an appropriate control group

Example:

• Claire is 10. She obtained 45 / 160 on a test which measures verbal reasoning. A representative sample of 500 other 10 year olds generated an average score of 100/160.

• Is Claire developing atypically? does she have advanced or delayed verbal reasoning?

• 

• Claire is 10. Claire has an intellectual disability and a mental age of 5. She obtained 45 / 160 on a test which measures verbal reasoning. A sample of 500 5 year olds generated a normal distribution with an average score of 50.

• Claire has average verbal reasoning abilities compared with her mental-age matched peers

Important point:

• We can’t build a full profile and understanding of a child’s abilities unless we identify their strengths and weaknesses

• ‘Strengths’ can be subjective or ‘relative’

  • Something they’re good at compared to their other skills

  • Not necessarily a strength compared to other people

important point: example

Relative strength:

• Claire shows a strength in verbal reasoning relative to her overall IQ profile

• 

Whta to bear in mind when investigating atypical development

• it is important to compare performance against appropriate control groups and also to consider the child’s overall ability and profile of strengths and weaknesses

  • for example, is it usual to compare performance against two ‘control’ groups:

  • one group matched on chronological age, another group matched on mental age

• the examples only consider Claire’s ability at one point in time, but what about her developmental trajectory?

  • investigating skills and abilities over time provides insights into what we can expect of an individual’s development

What tools do we have to measure development?

Measuring development: cognition

specific experimental designs

• Designed to investigate a specific research question or hypothesis; target specific behaviours

• The format can vary widely depending on the research question and methodology.

• Can compare participants’ results with a matched control group (e.g. age, gender, IQ)

• Examples include face recognition tasks, theory of mind tasks & executive function tasks

standardised tests

• Designed to measure knowledge or skills in a consistent and comparable way across a large population

• Follow a fixed format with specific instructions, questions, and scoring procedures

• Participants’ scores can be standardised i.e. assigned a value that indicates how well they performed compared to every other person who has taken the test (regardless of individual differences)

• Examples include generalised intelligence tests e.g.

• Weschler Adult Intelligence Scale (WAIS)

• Weschler Intelligence Scales for Children (WISC)

• British Ability Scales

Specific Vs standardised

• Goal: Standardised tests aim to measure broader knowledge or skills, while experiments aim to test a specific skill or test hypotheses.

• Scope: Standardised tests are broad in scope, covering a range of topics or skills, while experiments are focused on a specific research question.

• Generalisability: Standardised tests aim to generalise results to a larger population, while experiments may have limited generalisability depending on the sample and conditions.

Wechsler Intelligence Scale for Children (WISC)

• the Weschler Intelligence Scale for children is a widely used intelligence test for children aged 6 to 16

• it includes:

  • Working Memory Index

  • Verbal Comprehension Index

  • Processing Speed Index

  • Visual Spatial index

  • Fluid reasoning Index

Outline the indexes of the WISC

• Verbal Comprehension Index (VCI)

  • This index measures a child's ability to understand and use language, as well as their verbal reasoning skills.  

• Visual Spatial Index (VSI)

  • This index measures a child's ability to perceive, analyse, and manipulate visual information.  

• Fluid Reasoning Index (FRI)

  • This index measures a child's ability to solve novel problems and think flexibly.  

• Working Memory Index (WMI)

  • This index measures a child's ability to hold information in mind

• Processing Speed Index (PSI)

  • This index measures a child's ability to quickly and accurately process information.

Tasks of the WISC (full scale)

verbal comprehension

• similarities

  • i.e. how are a lion and a rabbit similar?

  • how are happiness and anger similar?

• vocabulary

  • what does the word ‘coat’ mean?

  • what does ‘tradition’ mean?

• information

• comprehension

visual spatial

• block design

• visual puzzles

  • 

Fluid reasoning

• matrix reasoning

  • 

• figure weights

  • 

• picture concepts

• arithmetic

Working Memory

• digit span

  • i.e. repeat these numbers in the same order I say them to you

  • “3,7,14,9,10”

  • repeat these numbers in the reverse order

  • “7,9,12,19,8”

• picture span

• letter-number sequencing

processing speed

• coding

• symbol search

• cancellation

Measuring development: Adaptive behaviour

Vineland Adaptive Behaviour Scale (VABS)- semi structures interview carried out with parent/ caregiver/ teacher

• Communication: Receptive: what he or she understands; Expressive: What the individual says; Written: What he or she read and writes.

• Daily Living Skills: Personal: How the individual eats, dresses; Domestic: What household tasks the individual performs; Community: How the individual uses time, money, etc.

• Socialization: Interpersonal Relationships: How the individual interacts with others. Play and Leisure Time: How the individual plays. Coping skills: How the individual demonstrates responsibility and sensitivity to others.

• Motor Skills: Gross Motor: How the individual uses arms and legs for movement and coordination. Fine Motor: How the individual uses hands and fingers to manipulate objects.

• Maladaptive Behaviour Internalizing, Externalizing and other types of undesirable behaviour that may interfere with the individual’s adaptive functioning

What about tests for non-verbal participants?

Wechsler Nonverbal Scale for Ability (WNV):

• Assesses non-verbal reasoning and problem-solving skills in individuals aged 4 to 21.

• It uses visual stimuli and requires minimal verbal instruction, making it suitable for children with language difficulties or those who are non-verbal.

• Consists of subtests such as Object Assembly, Block Design, and Picture Arrangement.

Leiter International Performance Scale- Revised (Leiter-R)

• Assesses cognitive abilities in individuals aged 3 to 75.

• It uses a variety of tasks, such as matching pictures, completing patterns, and solving mazes, to assess different aspects of intelligence.

• Particularly useful for assessing individuals with autism, language impairments, or hearing impairments.

What about tests for younger patients?

Bayley Scales of Infant and Toddler Development (Bayley-III):

• Infants and toddlers aged 1 to 42 months.  

• It evaluates cognitive, motor, language, social-emotional, and adaptive behavior.  

• Tests include observation of motor skills (e.g. rolling), tests of cognition (e.g. attention span) and social interaction

Infant-Toddler Developmental Assessment (IDA):

• From birth to 36 months who are at risk of developmental delays or conditions

• It evaluates cognitive, motor, language, social-emotional, and adaptive behavior through observation, parent report, and standardised tasks.

• The IDA is often used in early intervention programs to identify children who need additional support.  

Scoring standardised tests

• After completing the test add the scores together to create a raw score

• How useful is this score?

• Can we compare it to older individuals, for example?

• Do you think it’s fair to compare a 10 year old to a 30 year old?

• No, it’s not comparable!

• So what do we do?

• We convert the raw score to something called a ‘standardized score’

• Standardising a score converts the raw score to a value that represents how a participant has performed compared to others of the same age/gender

• This allows us to remove individual differences and generate a score that we can compare across participants

Why do we standardise scores?

For example:

We want to investigate if Johnny and Vera are a typical height. But we know that boys & girls have different height expectations. So how can we check if their heights are typical? What we expect for Johnny is not what we expect for Vera. Answer: we standardise their scores. We compare Vera’s height to a sample of girls the same age and give it a value (e.g. a t-score) to indicate where it falls in the bell curve. Then we do the same for Johnny. This lets us know how Johnny and Vera compare to an age and gender matched sample. We can then compare this value between Johnny and Vera to see how their development compares. T-scores are scaled such that 50 represents the mean and 10 represents 1 standard deviation (e.g. a t-score of 60 = 1 sd above the mean and a t-score of 40 – 1 sd below the mean). Vera’s score is a few SD above the mean. Johnny’s score is a few SD below the mean. So Johnny might be taller then Vera. But the t-scores show that Johnny is short for his age, whereas Vera is tall for her age

After completing a test and obtaining a raw score, the experimenter / clinician uses a “look-up table” based on appropriate representative sample to identify a scaled score.

Raw scores are scaled by converting to standard scores. There are different ways to create standardised scores. One example is t-scores.

Benefits of using standardised scores

• enables researchers/ clinicians to standardised performance across different groups, different tests, etc

• they provide a common language for discussing test performance regardless of how the actual test is designed

• easily interpretable for clinicians/ researchers

• there isn’t one set way of standardising, although the all end up allowing the same comparison

Reading

Wechsler Intelligence Scale for Children, Fifth Edition Profiles of Children With Autism Spectrum Disorder Using a Classification and Regression Trees Analysis

Abstract

The Wechsler Intelligence Scales for Children (WISC) are the most widely used instrument in assessing cognitive ability, especially with children with autism spectrum disorder (ASD). Previous literature on the WISC has demonstrated a divergent pattern of performance on the WISC for children ASD compared to their typically developing peers; however, there is a lack of research concerning the most recent iteration, the Wechsler Intelligence Scale for Children, Fifth Edition (WISC V). Due to the distinctive changes made to the WISC-V, we sought to identify the pattern of performance of children with ASD on the WISC-V using a classification and regression (CART) analysis. The current study used the standardization sample data of the WISC-V obtained from NCS Pearson, Inc. Sixty-two children diagnosed with ASD, along with their demographically matched controls, comprised the sample. Results revealed the Comprehension and Letter-Number Se quencing subtests were the most important factors in predicting group membership for children withASDwithanaccompanyinglanguageimpairment.ChildrenwithASDwithoutanaccompanying language impairment, however, were difficult to distinguish from matched controls through the CART analysis. Results suggest school psychologists and other clinicians should administer all primary and supplemental subtests of the WISC-V as part of a comprehensive assessment of ASD diagnosis of autism spectrum disorder is best accomplished through evidence-based assessment practices, which includes the use of psychometrically sound standardized measures (Lord et al., 2011). Although cognitive assessment is only a piece of a more comprehensive assessment, psychologists must understand how children with ASD perform on these measures to ensure an evidence-based assessment occurs and the most diagnostic clarity is possible. Although guidelines have been suggested in choosing assessments for cognitive and adaptive functioning of individuals with ASD (Klinger et al., 2018), the Wechsler Scales of Intelligence are the most commonlyutilized tests of cognitive ability in an evaluation of ASD (Saulnier & Ventola, 2012). Aiello et al. (2017) found the majority of school psychologists choose the Wechsler Intelligence Scale for Children, Fourth Edition (WISC-IV; Wechsler, 2003) when assessing the cognitive functioning of children with ASD. Little research, however, has been conducted using the Wechsler Intelligence Scale for Children, Fifth Edition (WISC-V; Wechsler, 2014) with children with ASD. Cognitive profile analysis has been studied to help understand the cognitive abilities of individuals with ASD. Various patterns of performance have been identified using the past versions of the Wechsler Intelligence Scale for Children (WISC), and these patterns can differentiate children with ASD from their peers with or without other disabilities (Kuriakose, 2014; Mayes & Calhoun, 2003a, 2003b; Nader et al., 2015; Oliveras-Rentas et al., 2012). The following sections review the literature on the performance of children with ASD on cognitive abilities measures. Then, the results of a classification and regression trees (CART) analysis are presented and discussed.

cognitive profiles in ASD

Children with ASD have been found to exhibit variability in cognitive skills across measures and time, which generates concern with using IQ as an outcome measure (Lord et al., 2005). Although cognitive profiles may vary across children with ASD, there is some evidence of a cognitive phenotype of ASD (Kuriakose, 2014). Previous research has demonstrated a large number of individuals with ASD fall two to three SDs below the mean on tests of intellectual functioning, representing an intellectual disability (Baird et al., 2000; Kielinen et al., 2000). More recent literature, however, suggests one-third of individuals with ASD perform two SDs below the mean on measures ofcognitive ability, with approximately half of the population performing at least one SD below the mean (Christensen et al., 2016). Additionally, an abundance of support shows average to superior performance on fluid reasoning, visuospatial skills, and working memory in some children with ASD (Dawson et al., 2007; Muth et al., 2014; Scheuffgen et al., 2000; Shah & Frith, 1983, 1993). These peaks may be accompanied by deficits in processing speed and verbal reasoning. Due to the divergent profiles observed across individuals with ASD, the DSM-5 includes the specifiers “with or without accompanying intellectual impairment” (APA, 2013, p. 51; Miller et al., 2016). For children with ASD, IQ has been found to be the most highly correlated variable to symptom severity and demonstrates a negative relationship, such that as IQ increases, the severity of symptomsdecreases, and viceversa (see, e.g., Matson& Shoemaker,2009; Matsonetal., 2010; Ozonoff et al., 2005). Given the relationship between IQ and symptom severity, coupled with the importance of diagnostic specifiers, understanding the performance of children with ASD on popular intelligence tests is essential (Miller et al., 2016).

cognitive profiles in ASD- Weschler intelligence scales

Cognitive profiling has been studied to help understand the cognitive abilities of children with ASD. Previous literature has suggested children with ASD exhibit atypical cognitive profiles across broad and narrow domains of various versions of the WISC, with more variability than typically developing children. Although previous research has demonstrated some variation in strengths and weaknesses ob served in children with ASD, similar patterns have emerged from multiple studies across various versions of the WISC. Specifically, the Wechsler profile among school-age children with ASD is characterized by lower scores on the Freedom from Distractibility Index (FDI) from the Wechsler Intelligence Scale for Children, Third Edition (WISC-III; Wechsler, 1991) and the Working Memory Index (WMI) on the WISC-IV, and lower performance on the Processing Speed Index (PSI) in comparison to the Verbal Comprehension Index (VCI) and the Perceptual Reasoning Index (PRI) WISC-IV(Mayes&Calhoun,2003a,2003b;Oliveras-Rentasetal.,2012).Regarding subtest Performance, most studies find relatively high scores on Block Design and low scores on Comprehension and Coding/Digit Symbol (Oliveras-Rentas et al., 2012). Mayes and Calhoun (2003a, 2003b) found no significant differences between the WISC-III Index scores for children with autistic disorder who obtained IQ scores less than 80, whereas children with autism who obtained IQ scores greater than or equal to 80 exhibited significantly higher scores on VCI and POI over FDIandPSI.MayesandCalhounfoundthatalthoughchildrendiagnosed with autistic disorder who obtained IQs less than 80 did not demonstrate significant variability in subtests scaled scores, those who had IQs of 80 or higher demonstrated significantly more scatter than the norm. Children with high-functioning autism tend to demonstrate attention, graphomotor, and processing speed weaknesses in contrast to strengths in verbal and visual reasoning (Mayes et al., 2008). Historical research suggests children with ASD can be distinguished from children in other diagnostic groups by their cognitive profiles. When utilized within a more comprehensive evaluation that includes ASD-specific measures, interview data, and observations, understanding these distinctions can help school psychologists make more accurate eligibility decisions. Al though the names of the index scores have changed with the various revisions of the WISC, many of the subtests and constructs measured remain the same. Furthermore, it is important to un derstand the performance of children with ASD on previous versions when evaluating the patterns of performance on the WISC-V. For instance, utilizing the WISC-III, Mayes and Calhoun (2004) found three distinct patterns of performance supporting the use of profile analysis to discriminate neuropsychological conditions in children. Specifically, they found low Coding and low Com prehension scores characterized children with ASD; low Coding without low Comprehension scores identified children with ADHD and learning disability; and low Performance without low Coding characterized children who had a traumatic brain injury. Therefore, Mayes and Calhoun (2004) were able to identify high-functioning autism with 73% accuracy based on low FDI, PSI, and Comprehension subtest scores. In addition, Ehlers et al. (1997) conducted a study aimedat evaluating the discriminating ability of the WISC-R (Swedish version; Wechsler, 1977) for autistic disorder, Asperger syndrome, and attention disorders through identifying characteristic WISC profiles within each of the three diagnostic groups. The findings from their study demonstrated an overall rate of correct diagnostic classification of the three groups of 63%. They showed that discriminating across groups was suitable for 49% of the cases when separating autism from Asperger’s and autism from attention disorders; however, only 16% of cases were able to be distinguished between Asperger’s from attention disorders. An intragroup analysis showed Asperger syndrome was associated with good verbal ability but poor perceptual ability, whereas autism was associated with relatively superior ability in visuospatial skills. The findings showed that the Asperger syndrome and autistic disorder groups differed on subtests thought to measure crystallized intelligence. Subtest results of the WISCmayprovideausefulbasisforcomparisonofcognitive peaks and troughswithin andacross different clinically defined groups; however, it is not perfect at discriminating among other disorders (Ehlers et al., 1997). To date, much of the updated literature on cognitive profiles in children with ASD utilize the WISC-IV, and similar patterns of performance occur. For example, children with high-functioning ASD performed significantly weaker on subtests making up the Verbal Comprehension Index, specifically Vocabulary and Comprehension as well as the Processing Speed Index consisting of Symbol Search and Coding.

Purpose of the present study

Assessment of intellectual functioning is considered an important component of an educational evaluation of ASD. Tests of cognitive ability can help guide intervention planning (Volkmar et al., 2014), which results in more appropriate IEP goals for students with ASD. Given the Wechsler tests are the most preferred cognitive ability measure for psychologists conducting autism evaluations (Aiello et al., 2017; Saulnier & Ventola, 2012), understanding how this new edition performs with the autism population will aid in its utility in the school and clinic settings. Current knowledge of the WISC-V with children with ASD is limited. The present study explored the cognitive abilities of the individuals with ASD who participated in the standardization of the WISC-V through a classification and regression trees (CART) analysis.

Methods

procedures

The present study employed the standardization sample data obtained from the publishers of the WISC-V. Data from this sample were obtained through the procedures specified in the Technical and Interpretative Manual of the WISC-V. The current study was approved by the researchers’ university’s Institutional Review Board.

Participants

Sixty-two children diagnosed with ASD and their demographically matched controls (mean age = 11.5 years.; n = 124) were included in the sample. The WISC-V manual did not specify how diagnosis was determined. The sample was further described according to DSM-5 language specifiers, with accompanying language impairment and without accompanying language im pairment. Thirty children with accompanying language impairment (ASD-LI) and thirty-two children without accompanying language impairment (ASD-NLI) were included. Children in all groups were initially excluded from participation if they had existing cognitive ability scores below a specified range indicating intellectual disability (<60 for ASD-LI and <70 for ASD-NLI), and the former group was required to have adequate language skills to participate in test activities. According to the manual, control participants were randomly selected from the standardization sample and matched on the following demographic variables: age, sex, race/ethnicity, parent education, and geographic region. All demographic variables, except for parental education, were provided to the researchers, and characteristics of the sample are presented in Table 1.

Measure

The WISC-V is the most recent version of the popular Wechsler Intelligence Scale for Children (Wechsler, 1949), originally published in 1949. The WISC-V is an individually administered intelligence test for use with children ages 6 years, 0 months through 16 years, 11 months. Twenty one subtests make up the WISC-V (see the technical manual for a thorough description of all 21 subtests) with 10 considered primary subtests and required to obtain the five Primary Index scores (Verbal Comprehension, Visual Spatial, Fluid Reasoning, Working Memory, and Processing Speed). Of the 10 primary subtests, seven are used to calculate the Full Scale IQ (FSIQ). Ac cording to the manual, the five Primary Index scores and the FSIQ provide a comprehensive evaluationofachild’sintellectualfunctioning.SeeTable2forabreakdownoftheFullScaleand PrimaryIndexScales.

Data Analysis

Toidentifythetestscorecombinationsthatdifferentiatethegroups,classificationandregression trees(CART)analysiswasused.Comparedtootherformsofregressionanalyses,CARTmore easily identifies the interactionsof test scores to identify if certaincombinationsclassify the groups.Specifically,aconditionalapproachtofittingCART(Hothorn&HornikZeileis,2006)was employed,usingtheRsoftwarepackagefunctionctreeinthepartykit library.Thefamily-wise TypeIerror ratefor thestatisticdeterminingthenumberandlocationofsplits inthetreewas controlled using the Bonferroni correction, and a Monte Carlo approach was employed to de termine the level of statistical significance. A total of four CARTanalyses were conducted. Two of these analyses focused on differentiating children with ASD from the matched controls using either the WISC-V subtests or the indexes, and the other two analyses were used to differentiate ASD-LI, ASD-NLI, andthe matched sample using either the subtests or the indexes. To assess the overall performance of the models, both prediction accuracy rates using the sample and Cohen’s kappa statistic were calculated for each CART model.

Results

ASD Vs matched controls

The first CART analysis focused on differentiating ASD and the matched controls based on the WISC-V subtests. The resulting tree appears in Figure 1, and the terminal node information appears in Table 3.

The results demonstrate that 100% of the individuals with Comprehension scores less than or equal to six and Symbol Search scores less than or equal to 10 (n = 24) had previously been identified with ASD. In addition, 65% of individuals with Symbol Search scores greater than six and Letter-Number scores less than or equal to 9 (n = 30) had ASD. In contrast, those with Comprehension scores greater than 6, and Letter-Number scores greater than nine had a relatively low likelihood of having been identified with ASD, ranging between 0% and 40%. The accuracy of the classifications derived from the CART model and the associated Cohen’s kappa appear in Table 8. The model accurately classified group membership for 77.4%% of the individuals in the sample, with somewhat greater accuracy for the ASD group. The value of kappa, 0.548, fell into the weak predictive accuracy category. In addition to the WISC-V subtests, the ASD and matched control groups were also differ entiated based on their Primary Index scores. The resulting CART tree appears in Figure 2, and the terminal node information is presented in Table 4. These results demonstrate that individuals with Working Memory scores of 85 or less, and Verbal Comprehension scores of 95 or less were all previously identified as ASD (100%). Children with Working Memory scores less than or equal to 85 and Verbal Comprehension scores greater than 95 had ASDin62.5%ofcases. Individuals with WorkingMemory scores greater than 85, andProcessingSpeedscoresgreater than86were relativelyunlikely tohavebeenASD (28.4%),whereas44%of thosewithWorkingMemoryscoresgreater than85andProcessing Speedscoreslessthanorequalto86wereASD.Themodelbasedontheindexhadaclassification accuracyrateof73.4%(Table8),withmuchhigheraccuracyfortheASDgroup.Thekappavalue fell intotheweakrange.

ASD-Language Impaired, ASD-nonlanguage Impaired, and matched controls

InadditiontocomparingtheWISC-Vprofilesfor thoseidentifiedasASDversus thematched controls, asecondsetofanalyseswasconductedinwhichtheASDsamplewasdividedinto languageimpaired(LI)andnonlanguageimpaired(NLI)asdescribedabove.Meanperformance andSDsattheindexlevel,asreportedintheTechnicalandInterpretativeManual,arepresented inTable5. CARTwasthenusedtodifferentiatefromamongthethreegroups.TheresultingCARTtree basedon thesubscalesappears inFigure3,with the terminal node informationpresented in Table6

Basedontheseresults,ASD-LIchildrenwerecharacterizedwithlowerscoresonCompre hension,LetterNumber,andCancellation.ASD-NLImembersofthesampledisplayedmultiple profiles.Forexample, all eight individualswithComprehensionscoresbetween9and11and Similaritiesscoresgreaterthan12wereASD-NLI.Inaddition,thosewithComprehensionscores lessthan/equal to9,Letter-Numberscoreslessthan/equal to8,andCancellationscoresgreater thansevenweremost likelytobeASD-NLI.Likewise, thosewithComprehensionscores less than/equalto9,Letter-Numberscoresgreaterthan8,andVerbalComprehensionscoreslessthan/ equaltoeightwerealsomostlikelytobeASD-NLI.Finally,thenonclinicalmatchedsamplealso displayedavarietyofWISC-Vsubscaleprofiles,ascanbeseeninFigure3andTable6.The predictionaccuracyresultsinTable8demonstratethatthemodelwasabletoaccuratelyclassify groupmembership for 79%of the sample,with thegreatest accuracyoccurring forASD-LI (95.8%),andthelowestforASD-NLI(62.1%).Thekappavalueof0.654fellintothemoderately accuraterange. CARTresultsforthecomparisonofASD-LI,ASD-NLI,andthematchedcontrolsbasedonthe compositescoresappear inFigure4andTable7. IndividualswithscoresonWorkingMemorylessthan/equalto85andVerbalComprehension lessthan/equalto96wereverylikelytohavebeenidentifiedasASD-LI(83.3%),withnonebeing inthematchedcontrolgroupand16.7%beingASD-NLI.MembersofthesamplewithWorking Memorylessthan/equal to85andVerbalComprehensiongreaterthan96weremost likelytobe ASD-NLI,followedbythematchedcontrolgroup.ThosewithWorkingMemoryscoresbetween 86and107,alongwithProcessingSpeedscoreslessthan/equalto89weremostlikelytobeASD NLI(38.5%)ormatchedcontrol(34.6%).Thenonclinicalmatchedsamplewasmostcommonly associatedwithWorkingMemorywasgreaterthan107(node4),orwhenWorkingMemorywas greater than85andProcessingSpeedwasgreater than89(node5).Theoverall classification accuracyratesforthisCARTmodelwas67.7%,withakappaof0.477,whichfell intotheweak category (Table 8). The model classified ASD-LI membership with 83.3% accuracy but was not as accurate for the ASD-NLI group (41.2%)

Discussion

Previous literature indicates that children with ASD have similar patterns of performance on the Wechsler Scales (Miller et al., 2016; Nader et al., 2015; Ozonoff et al., 2005). Given the popularity of the WISC as the first-choice cognitive assessment within schools and clinical settings, school psychologists and clinicians must understand how a child’s per formancefits into the overall assessment. Understanding the performance of children with ASD on cognitive ability tests is vital so that school psychologists can make the most appropriate eligibility recommendations to help inform a case conference committee of the student’s educational needs. This study was the first to look at the cognitive patterns of children with ASD through a CART analysis. Overall, children with ASD with accompanying language impairment were most ac curately classified as ASD based on Working Memory Index (WMI) and Verbal Comprehension Index (VCI) scores, but children with ASD without accompanying language impairment were less accurately classified compared to typically developing peers. The following sections expand on these findings.

Verbal comprehension Index

The various iterations of the WISC have been criticized as having a high language load (Kuehnel et al., 2019). Specifically, all the subtests that comprise the VCI require the child to comprehend the meaning of individual verbal prompts for each item and provide a verbal response. In a disorder characterized by social communication and potential language impairments, the ex pressive and receptive demands would seem place the child with ASD at a disadvantage compared to their typically developing peers. Researchers have found that children with ASD performed significantly weaker on the VCI compared to the Perceptual Reasoning Index (PRI) on the WISC IV with a similar discrepancy seen on the WISC-III’s comparable indexes (Nader et al., 2015). Atthe index level, our results indicated children’s performance on the VCI and WMI weremost predictive in classifying children as ASD. To further understand those results, one must look at the f indings from three group CART analyses indicating low WMI and lower VCI accurately classified the ASD-LI group but not the ASD-NLI group. Much of the previous research focuses on ASD as conceptualized by the Diagnostic and Statical Manual of Mental Disorder, Fourth Edition, Text Revision (DSM-IV-TR; APA, 2000) diagnoses of either Asperger’s disorder or autistic disorder, and this research suggests varying profiles based on diagnoses within the spectrum (Nader et al., 2015; Ozonoff et al., 2005). Nader et al. (2015) found children with Asperger’s disorder performed better on the VCI compared to the other indexes. Given that these conditions have been collapsed into one spectrum of symptomology, it seems imperative clinicians the use of the language impairment specifier. Perhaps the most consistent area of cognitive weakness seen in children with ASD is language comprehension abilities and social reasoning (Mayes & Calhoun, 2003a, 2003b; Mayes et al., 2008; Nader et al., 2015; Oliveras-Rentas et al., 2012). Historically, the Comprehension subtest has consistently been the lowest subtest score compared to all other subtests for children with ASD. Findings of the current study are consistent with previous literature and indicated accurate identification of ASD based on the Comprehension subtest. Furthermore, all children with high average Comprehension scores were in the non-ASD group. Regardless of the diagnostic category from the autism spectrum, researchers have demonstrated that the Comprehension subtest was significantly weaker than other verbal tests on previous editions of the WISC (Nader et al., 2015; Mayes et al., 2008; Oliveras-Rentas et al., 2012). Oliveras-Rentas et al. found a pattern of higher Similarities, intermediate Vocabulary, and low Comprehension scores for the WISC-IVand concluded children with ASD are able to display their strengths more reliably when expected to provide short verbal responses; longer responses are often necessitated for obtaining full credit on a Comprehension item which helps explain this pattern. Our results suggest that higher Similarities scores are more indicative of children with ASD without a language impairment. Given that this newest version only requires the admin istration of the Similarities and Vocabulary to obtain the FSIQ and VCI, which previous research suggests are strengths of a child with ASD, it may hide their weaknesses and present an inflated conceptualization of their verbal skills (Miller et al., 2016). Our results indicated that the WISC-V does indeed exclude the weakest area of verbal ability from the VCI for individuals with ASD

Working memory Index

Asnoted earlier, findings of the current study indicate children with ASD were often differentiated from the matched controls on the WMI; however, a closer look at the CARTanalysis of the three groups provides further insight into this index and suggests Working Memory skills vary based on language skills. In fact, children with ASD have been observed to exhibit variability in Working Memory abilities (Miller et al., 2016), which suggests working memory may be mediated by another ability such as language skills. Specifically, when considering speech onset delay in an autistic group, Nader et al. (2015) found that WMI was significantly weaker than the FSIQ, and those with a speech delay exhibited lower scores. Conversely, on the WISC-IV, Oliveras-Rentas et al. (2012) found that children with Asperger’s (i.e., those without language delays) demon strated WMI scores within the average range. Kuriakose (2014) also found that children with ASD displayed a mean WMI within the average range. At the subtest level, Letter-Number Sequencing was identified frequently in the CARTanalysis as a significant factor. A high percentage of children with low Letter-Number Sequencing scores were either ASD-LI or ASD-NLI, indicating this subtest is a weakness for many individuals with ASDirrespective of language skills. Overall, our results reflect the variability in working memory abilities of children with ASD, and this variability is most present within the ASD-NLI group

Processing Speed Index

Our findings indicated that the greatest number of children without ASD fell in the terminal node characterized by both a WMI standard score above 85 with a Processing Speed Index (PSI) standard score above 86, scores within one SD of the mean. Previous literature suggests children with ASDhistorically perform poorly on the PSI. Kuriakose (2014) found that children with ASD demonstrated a mean PSI falling in the low range. Oliveras-Rentas et al. (2012) examined the WISC-IV profile of children with ASD without an intellectual disability, which PSI was the greatest area of relative and normative weakness for this population. Nader et al. (2015) indicated that a low Processing Speed Index (PSI) score and its significant discrepancy with the VCI score is a distinctive characteristic of children with Asperger’s based on the WISC-IV profile of scores. Research suggests children with high-functioning ASD display the lowest index scores on either the PSI or the WMI of the WISC-IV. The Coding subtest historically represents a significant weakness for the ASD population (Ehlers et al., 1997; Mayes & Calhoun, 2003a, 2003b, 2004; Mayes et al., 2008; Oliveras-Rentas et al., 2012; Siegel et al., 1996); however, Coding was not found as a meaningful subtest in our decision trees. Results of this study indicated Symbol Search and Cancellation were more helpful in separating the groups. For instance, all children in the terminal node with low Comprehension and low Symbol Search scores were in the ASD group. This finding is consistent with Oliveras Rentas et al. (2012) who indicated Symbol Search was also a weakness of children with ASD.

Perceptual Reasoning Index, fluid Reasoning Index and Visual Spatial Index

Previous literature on the various performance indexes of the Wechsler Scales indicated children with ASD perform significantly better on the Perceptual Reasoning Index (PRI) compared to the VCI (Mayes et al., 2008; Nader et al., 2015), and PRI is the highest mean index score (Kuriakose, 2014). The PRI has now been separated into two different indexes, the Visual Spatial Index (VSI) and Fluid Reasoning Index (FRI); thus, it is necessary to examine how the cognitive profile may differ in comparison to previous editions of the WISC. The two CARTanalyses at the index level indicated that the VSI and FRI were not helpful in differentiating the three groups (ASD-LI, ASD NL, and non-ASD). This finding is possibly due to the reconceptualization of these abilities and inclusion of two new subtests to the WISC-V (Visual Puzzles and Figure Weights). Additionally, the nonverbal strengths and weaknesses of children with ASD are now included in separate indexes, potentially contributing to these findings. Prior to the WISC-IV, Block Design wasthehighest ofthe nonverbal subtests in children with autism on previous versions of the WISC (e.g., WISC-R and WISC-III; Mayes & Calhoun, 2003a, 2003b, 2004; Siegel et al., 1996). However, it appears that a pattern of peak performance on the Matrix Reasoning subtest emerged for children with autism on the WISC-IV (Mayesetal., 2008;Oliveras-Rentas et al., 2012). Mayes et al. (2008) found that children with high-functioning autism scored the highest on Matrix Reasoning subtest in comparison to the other PRI subtests, and Block Design was found to be the lowest of the nonverbal subtests. Matrix Reasoning and Block Design are now included within different indexes of the WISC-V, which may have impacted the relevance of the index scores within the CART analyses.

IMplications for practice

The results of the present study provide practical guidance for clinicians and school psychologists using the WISC-V in a comprehensive assessment of ASD. Cognitive ability tests are a vital component of an evidence-based assessment for ASD (Aiello et al., 2017); however, diagnosis of ASD should be made through a multi-measure, multi-informant approach by an experienced clinician (Volkmar et al., 2014). Although cognitive ability profiles are not diagnostic, they can provide some insight into whether an individual exhibits a common pattern consistent with a neu rodevelopmental disorder. Overall, the CART analysis that resulted in the highest accuracy rate included all subtests and differentiated the sample into three groups. This model resulted in 95% accuracy for predicting ASD-LI classification, much higher than prediction for ASD-NLI. These results further suggest that greater cognitive differences exist between children with ASD with an accompanying language impairment and typically developing peers compared to those with ASD without a language impairment. When reviewing patterns of performance on the WISC-V, clinicians and school psychologists should understand the greater variability in abilities across the indexes for children with ASD who do not have deficits in language skills. Given the supplemental status of several subtests of the WISC-V that were identified within the CART analysis, clinicians and school psychologists should administer extra subtests in cases where more information on a child suspected of having ASD is necessary. Specifically, these results suggest Comprehension and Letter-Number Sequencing should be included in a WISC-V administration and the child’s language abilities should be strongly considered. Accuracy of the CART models was weak to moderate, with the greatest accuracy at the subtest level for children with ASD-LI, further supporting the administration of all primary and supplemental subtests for an ASD evaluation in schools and clinics.

Limitations and directions for future research

The current study was limited to the sample of children included in the standardization sample (ASD n = 62) of the WISC-V. It is possible that prediction rates of the various CART models would improve with a larger sample. This study should be replicated with an independent sample of children with ASD, especially as CART analysis appears to be useful for this population. Additionally, children with other DSM-5 or educational categorizations should be included in a CARTanalysis to improve our understanding of the differences in cognitive functioning among children with ASD and various other neurodevelopmental and behavioral disorders. Furthermore, children with IQs lower than 60 were excluded from the study’s sample, limiting the ability to generalize these results to individuals with ASD who have low IQs. Children capable of completing the WISC-V, but with IQs lower than 60 should be included in future analysis to determine if the CART results are similar if groups are divided up IQ rather than language impairment. Future research should include various rating scales and assessment of academic skills in addition to intellectual functioning. Research has demonstrated that performance on the PSI is related to autism communication symptoms and adaptive communication abilities (Oliveras Rentas et al., 2012) indicating the importance of utilizing a cognitive ability test within the scope of a comprehensive, multidisciplinary assessment. The significant relationship observed between PSIandcommunicationsymptomsandabilities in children with high-functioning ASD emphasize the importance of assessing and accommodating processing speed deficits for this population (Oliveras-Rentas et al., 2012). The current study was limited to only cognitive ability data and future research should examine the WISC-V profiles in conjunction with adaptive skills, ASD symptomology, executive functioning, and communication skills.

conclusions

The current study was the first to date to examine the cognitive abilities of children with ASD on the WISC-V using a CART analysis. Overall, the best classification accuracy was found in the model that included all primary and supplemental subtests, and it most accurately identified children with ASD with accompanying language impairment. Throughout the CART analyses, children without accompanying language impairment were difficult to differentiate from the matched controls based on WISC-V index and subtest performance. Although cognitive ability tests are not diagnostic, they are essential tools in a best practice assessment of ASD to determine the most specific severity level and additional specifiers. Understanding the cognitive strengths and weaknesses of children with ASD will benefit clinicians and school psychologists as they often treatment and educational intervention recommendations. Proper conceptualization of the cognitive abilities of children with ASD can tailor these intervention recommendations to the individual needs of the child.