Atypical communication characteristics (ACC), such as speech delay, odd prosody, neologisms, and pragmatic difficulties, are recognized as important phenotypic characteristics of autism spectrum disorder (ASD; Kanner, Reference Kanner1944) and are common reasons for clinical referral for young children with ASD (Tager-Flusberg et al., Reference Tager-Flusberg, Paul, Yoder, Rogers, Cooper, Landa and Yoder2009). ACCs are associated with more pronounced early ASD symptom onset (Miodovnik, Harstad, Sideridis, & Huntington, Reference Miodovnik, Harstad, Sideridis and Huntington2015); poorer response to treatment and poorer long-term outcome (Howlin, Reference Owlin, Volkmar, Paul, Klin and Cohen2005; Perry et al., Reference Perry, Cummings, Geier, Freeman, Hughes, Managhan and Williams2011; Sallows & Graupner, Reference Sallows and Graupner2005); and concurrent psychiatric symptom severity and impairment (Kang, Gadow, & Lerner, Reference Kang, Gadow and Lerner2020). ACCs in childhood and other speech, language, and communication problems are associated with a range of emotional, behavioral, and academic outcomes in youth (Yew & O'Kearney, Reference Yew and O'Kearney2013), and prior research suggests stable associations between these types of problems and psychopathology (Beitchman, Nair, Clegg, Ferguson, & Patel, Reference Beitchman, Nair, Clegg, Ferguson and Patel1986; Cantwell & Baker, Reference Cantwell and Baker1991; Miniscalco, Nygren, Hagberg, Kadesjö, & Gillberg, Reference Miniscalco, Nygren, Hagberg, Kadesjö and Gillberg2006). Rates of ASD-associated ACCs are significantly higher in youth with ASD than typically developing peers (Baltaxe & D'Angiola, Reference Baltaxe and D'Angiola1992; Gadow, Devincent, Pomeroy, & Azizian, Reference Gadow, DeVincent, Pomeroy and Azizian2004a, Reference Gadow, DeVincent, Pomeroy and Azizian2005; Geurts & Embrechts, Reference Geurts and Embrechts2008; Volden & Lord, Reference Volden and Lord1991) and non-ASD youth psychiatric referrals (Gadow et al., Reference Gadow, DeVincent, Pomeroy and Azizian2004a; Gadow, Devincent, Pomeroy, & Azizian, Reference Gadow, DeVincent, Pomeroy and Azizian2005; Kang et al., Reference Kang, Gadow and Lerner2020), and both ASD and non-ASD clinic referrals evidence higher rates than typically developing peers (Gadow et al., Reference Gadow, DeVincent, Pomeroy and Azizian2004a; Gadow et al., Reference Gadow, DeVincent, Pomeroy and Azizian2005). ACCs are an important feature of ASD, as they have demonstrated diagnostic value for identifying youth with ASD, and specific configurations of ACCs can be used to identify clinically useful subgroups (Kang et al., Reference Kang, Gadow and Lerner2020). Importantly, ACCs, like all aspects of language development, are susceptible to change over the course of development. However, most studies of ACCs in ASD are cross-sectional snapshots of symptom profiles and do not address the pattern of stability or change over time. Hence, there remains a need to understand whether there are clinically relevant subgroups of individuals with ASD based on stability of ACCs across time, and the degree to which such subgroups differ in terms of psychiatric symptoms and functional correlates.
Much less is known about the relation between stability or change in communication difficulties and psychiatric symptoms in older children and teens (e.g., youth between 6 and 18 years) relative to younger children. Traditional theories of ASD conceptualize its core pathology as being relatively stable over time (e.g., Matson & Horovitz, Reference Matson and Horovitz2010), and there is some evidence to suggest this is true for ASD-related communication deficits (Charman et al., Reference Charman, Taylor, Drew, Cockerill, Brown and Baird2005; Sigman & McGovern, Reference Sigman and McGovern2005). That said, there is substantial heterogeneity in speech, language, and communication abilities across individuals with ASD (Kjelgaard & Tager-Flusberg, Reference Kjelgaard and Tager-Flusberg2001) and their developmental trajectories (Anderson et al., Reference Anderson, Lord, Risi, DiLavore, Shulman, Thurm and Pickles2007; Lombardo et al., Reference Lombardo, Pierce, Eyler, Carter Barnes, Ahrens-Barbeau, Solso and Courchesne2015; Pickles, Anderson, & Lord, Reference Pickles, Anderson and Lord2014). Theories of developmental plasticity suggest that children interact with and respond differently to diverse early environments and experiences during critical periods of development (e.g., Mundy & Neal, Reference Mundy and Neal2001), which over time likely results in intra- and inter-individual variability in core ASD symptoms and co-occurring psychopathology that in turn impact functional outcomes.
Developmental models of ASD (e.g., Klin, Jones, Schultz, & Volkmar, Reference Klin, Jones, Schultz and Volkmar2003) suggest that repeated failure to enact typical functional processes (e.g., language) disrupts normative social brain development, and this may in turn result in characteristic symptoms of ASD and maintain or alter the presentation of core ASD and co-occurring psychiatric symptoms and impairment. Children with ASD show dramatic inter-individual phenotypic heterogeneity in trajectories of speech and language development, and ontogenetic change in these symptoms is a predictor of later social and communicative outcomes (e.g., Anderson et al., Reference Anderson, Lord, Risi, DiLavore, Shulman, Thurm and Pickles2007; Lord, Luyster, Guthrie, & Pickles, Reference Lord, Luyster, Guthrie and Pickles2012a). Therefore, examination of stability of ACCs may provide unique predictive information about severity of ASD and co-occurring psychiatric symptoms; however, these relations have not previously been explored.
In everyday practice, clinicians routinely try to establish whether a child's current problem(s) represents a change from previous level of functioning with an eye toward identifying the emergence of a new disorder and/or its environmental precipitants. However, there is relatively little systematic study of this practice or its clinical utility, particularly in early-onset neurodevelopmental disorders such as ASD. Much of the research in this area pertains to the documentation of change in premorbid symptomatology in adolescent-onset schizophrenia (Lyngberg et al., Reference Lyngberg, Buchy, Liu, Perkins, Woods and Addington2015; Marchesi et al., Reference Marchesi, Affaticati, Monici, De Panfilis, Ossola, Ottoni and Tonna2015) and mood disorders (e.g., Wilson, Vaidyanathan, Miller, McGue, & Iacono, Reference Wilson, Vaidyanathan, Miller, McGue and Iacono2014) or childhood problems in adult-diagnosed attention-deficit/hyperactivity disorder (ADHD) (e.g., Moffitt et al., Reference Moffitt, Houts, Asherson, Belsky, Corcoran, Hammerle and Caspi2015). Although retrospective methods to evaluate developmental history associated with particular disorders (e.g., Moffitt et al., Reference Moffitt, Houts, Asherson, Belsky, Corcoran, Hammerle and Caspi2015; Pereira, Pasman, van der Heide, van Delden, & Onwuteaka-Philipsen, Reference Pereira, Pasman, van der Heide, van Delden and Onwuteaka-Philipsen2015) do not provide information on developmental trajectories of symptoms and clinical correlates per se, they are commonly used, straightforward, and cost-effective procedures for assessing change in symptom status. As such, assessing changes in clinical status of ACCs using retrospective methods may be a useful proxy for longitudinal assessment of stability of these symptoms in clinic-referred youth.
Current study
This study aimed to elucidate the relation between developmental changes in ACC and a broad range of psychiatric symptoms and associated clinical correlates among children and adolescents with ASD, as well as non-ASD psychiatry outpatient referrals using a simple retrospective method that is common in clinical practice. To examine whether ASD diathesis increases the risk for a change in ACCs over the course of development, we first compared the stability of past and current ACCs in the ASD and non-ASD diagnostic groups. We hypothesized that the ASD group would evince greater variability in number of ACCs, whereas the non-ASD group would demonstrate a mostly stable pattern of ACCs. We then investigated whether stability of ACCs offers unique information about psychopathology above and beyond current levels, and whether this information yields meaningful subgroups of youth with ASD to assess phenotypic heterogeneity in stability of ACCs in ASD. Toward this aim, we used latent class analysis (LCA) to determine whether the ASD diagnostic group could be meaningfully partitioned into subgroups based on ACC features. Given that the contemporaneous ACCs have been shown to characterize youth with ASD but not non-ASD psychiatry referrals (Kang et al., Reference Kang, Gadow and Lerner2020), we confined the LCA analysis to the ASD diagnostic group. We then compared these subgroups with respect to ASD symptomatology, co-occurring psychiatric symptoms (ADHD, anxiety, major depressive, and manic symptoms), and functional (academic and social) correlates.
Method
Participants
Case records for consecutive referrals to one of two university hospital clinics located on Long Island, New York were screened for youth between 6 and 18 years old who were assessed with prerequisite questionnaires completed by parent and teacher at the time of their initial diagnostic evaluation. The ASD diagnostic group (n = 283) was composed of youth who were evaluated in a developmental disabilities specialty clinic and diagnosed as having an ASD according to the fourth edition of Diagnostic and Statistical Manual of Mental Disorders (DSM-IV) (American Psychiatric Association, 2000) diagnostic criteria (see Procedure). Non-ASD psychiatry referrals (n = 724) were assessed in a child and adolescent psychiatry outpatient clinic, excluding youth with an ASD or psychosis diagnosis. Psychiatric diagnoses for the non-ASD diagnostic group were learning disorder, language disorders, ADHD, oppositional defiant disorder, conduct disorder, generalized anxiety disorder, social anxiety disorder, separation anxiety disorder, specific phobia, obsessive-compulsive disorder, tic disorders, major depressive disorder, dysthymic disorder, bipolar disorder, substance abuse disorder, adjustment disorder, post-traumatic stress disorder, eating disorder, elimination disorder, reactive attachment disorder, and developmental coordination disorder, in the order of decreasing prevalence. Study samples are described in prior publications (e.g., Gadow, Perlman, Ramdhany, & de Ruiter, Reference Gadow, Perlman, Ramdhany and de Ruiter2016, for ASD diagnostic group; Gadow, Kaat, & Lecavalier, Reference Gadow, Kaat and Lecavalier2013, for non-ASD diagnostic group). Collectively, participants (Mage = 11.41, SDage=3.67) were largely male (71.9%) and Caucasian (83.9%). Sample characteristics are provided in Table 1. This study was approved by the university Institutional Review Board, and appropriate measures were taken to protect child and caregiver confidentiality.
Table 1. Demographic characteristics of study samples
a t test for continuous variables and chi-squared statistic testing association between sample and categorical variables. ***p < .001.
Measures
Psychiatric symptoms and impairment
Parents and teachers rated youth's ASD and co-occurring psychiatric symptoms with either the parent or teacher version of the Child and Adolescent Symptom Inventory-4R (CASI-4R; Gadow & Sprafkin, Reference Gadow and Sprafkin2005). Individual items bear one-to-one correspondence with DSM-IV symptoms and are rated on a scale from 0 (never) to 3 (very often). The CASI-4R covers a range of disorders, including the following symptom subscales which are considered in this study: ADHD, generalized anxiety disorder, major depressive episode, and manic episode. Numerous studies indicate CASI-4R subscales demonstrate satisfactory psychometric properties in community-based normative, clinic-referred non-ASD (Gadow, Reference Gadow2016; Gadow & Sprafkin, Reference Gadow and Sprafkin2002, Reference Gadow and Sprafkin2008) and ASD (e.g., Gadow, Reference Gadow2016; Kaat, Gadow, & Lecavalier, Reference Kaat, Gadow and Lecavalier2013; Lecavalier, Gadow, DeVincent, Houts, & Edwards, Reference Lecavalier, Gadow, DeVincent, Houts and Edwards2009; Sprafkin, Steinberg, Gadow, & Drabick, Reference Sprafkin, Steinberg, Gadow and Drabick2016) diagnostic groups.
Parent questionnaire
The parent questionnaire (Gadow, Devincent, & Schneider, Reference Gadow, DeVincent and Schneider2008a) obtains information about child, family, developmental, medical, and treatment characteristics, including ACCs. For the ACCs, the checklist includes the following 13 ACCs: (a) being nonverbal, (b) speech delay, (c) babbling without intent to communicate, (d) repeating words/phrases out of context, (e) echolalia (echoing what others say/echoing questions asked rather than answer them), (f) neologism (using made-up words or language), (g) pronoun reversals (“you” instead of “I”, etc.), (h) excessive stammering/stuttering, (i) perseverating (cannot stop talking about certain topics), (j) odd voice (monotone, odd pitch, odd pitch or “sing-song” voice), (k) lecturing (speaking as if lecturing others), (l) pragmatic difficulties (cannot maintain social conversation), and (m) limited facial affect (does not use a range of facial expressions, e.g., those that communicate guilt, surprise, sadness, etc.).
Using a retrospective method, parents indicate whether the youth has each of the 13 ACCs (a) currently at the time of evaluation, and if so, (b) whether the youth had the same ACC in the past (“stable”) or if this is something new that the youth did not have in the past (“change”). These were summed to reflect the total number of past and current ACCs, respectively. Number of ACCs in the past and in the present demonstrated acceptable to good internal consistency (Cronbach alpha = .81 and .78, respectively; Kang et al., Reference Kang, Gadow and Lerner2020). To assess stability of current ACCs, each ACC was scored as a categorical variable (0 = not at current; 1 = current only, “change”; 2 = past and current, “stable”).
ASD symptom severity
The Social Communication Questionnaire, Lifetime (SCQ; Rutter, Bailey, & Lord, Reference Rutter, Bailey and Lord2003) is a 40-item, caregiver-report that assesses ASD symptoms and generates three subscale scores: reciprocal social interaction; communication; and restricted, repetitive, and stereotyped behaviors. Higher scores indicate greater impairment. The SCQ demonstrates good psychometric properties in samples with and without ASD (e.g., Ung et al., Reference Ung, Johnco, McBride, Howie, Scalli and Storch2016; Wei, Chesnut, Barnard-Brak, & Richman, Reference Wei, Chesnut, Barnard-Brak and Richman2015).
The Autism Diagnostic Observation Schedule (ADOS; Lord, Rutter, Dilavore, & Risi, Reference Lord, Rutter, Dilavore and Risi2008) is a widely used standardized diagnostic assessment for ASD involving social presses to obtain and rate normative social responses and is administered by a trained clinician. ADOS scores have the following subscales: communication; reciprocal social interaction; social affect; and restricted and repetitive behaviors. Study participants were administered the original ADOS, but ADOS-2 scores were derived via revised ADOS-2 algorithm to ascertain diagnostic cutoffs for ASD, as well as ASD severity (calibrated severity score [CSS]; Hus Bal & Lord, Reference Hus Bal and Lord2015; Lord et al., Reference Lord, Rutter, DiLavore, Risi, Gotham and Bishop2012b). ADOS-2 CSS is shown to have good psychometric properties in similar populations with ASD (De Bildt et al., Reference De Bildt, Oosterling, Van Lang, Sytema, Minderaa, Van Engeland and De Jonge2011; Shumway et al., Reference Shumway, Farmer, Thurm, Joseph, Black and Golden2012).
Procedure
Prior to their initial diagnostic evaluation, parents of potential patients in each hospital outpatient clinic completed an intake assessment battery that included the CASI-4R, parent questionnaire, and permission for release of school reports. The school provided copies of teacher-rated CASI-4R, psycho-educational evaluations, special education evaluation records, and IQ testing results directly to the clinic by mail.
For the ASD group, DSM-IV-based ASD diagnoses were confirmed by an expert diagnostician and based on five sources of information: (a) comprehensive developmental history, (b) clinician interview with child and caregiver(s), (c) direct observations of the child, (d) review of validated ASD rating scales including the CASI-4R (Gadow et al., Reference Gadow, Schwartz, DeVincent, Strong and Cuva2008b), and (e) in most cases (73%) the ADOS (Lord et al., Reference Lord, Rutter, Dilavore and Risi2008) administered by a certified examiner. The only exceptions were children with a prior well-documented diagnosis of ASD (e.g., prior clinician or school evaluations) who received all of the aforementioned assessments but the ADOS.
Statistical analyses
We first conducted three two-tailed, paired-samples t tests to compare difference in the number of past and current ACC in the overall sample, as well as in the ASD and non-ASD diagnostic groups separately. We then conducted moderation analysis to determine if the association between number of past and current ACC varied as a function of ASD status.
To better understand the heterogeneity of stability in ACC and how these patterns may result in substantively meaningful groups within ASD, multicategorical LCA (Lazarsfeld & Henry, Reference Lazarsfeld and Henry1968) was conducted for the ASD diagnostic group using Mplus Version 7 (Muthén & Muthén, Reference Muthén and Muthén2012) to identify distinct classes with similar patterns of stability of ACC.
The following criteria for fit evaluation were considered in the current analyses to determine the number of classes best represented by the data (Berlin, Williams, & Parra, Reference Berlin, Williams and Parra2014; Nylund, Asparouhov, & Muthén, Reference Nylund, Asparouhov and Muthén2007): (a) the Bayesian information criteria (BIC; Schwarz, Reference Schwarz1978), the Akaike information criteria (AIC; Akaike, Reference Akaike1987), and the sample-size-adjusted BIC (SSABIC; Sclove, Reference Sclove1987), with BIC considered as comparatively better for which a lower value suggests a more optimal balance between model fit and parsimony, and (b) the bootstrapped likelihood ratio test (BLRT; McLachlan & Peel, Reference McLachlan and Peel2004), which assesses whether a model provides a significant improvement over a less restrictive form of the model, such that significant p value indicates whether the k − 1 class model is rejected in favor of the k class model, (c) entropy as an index of accuracy, with values closer to 1 indicating better classification in conjunction with other model fit indices; and (d) substantive interpretation of the results.
After specifying the latent classes, one-way analyses of variance (ANOVAs) (for continuous variables) and chi-squared tests (for categorical variables) were conducted to determine if the classes derived from the LCA demonstrated meaningful differences on following variables: demographic characteristics (age, full-scale IQ, sex, ethnicity [white vs. non-white], and household income level); ASD symptomatology on ASD diagnostic measures (ADOS, and SCQ) and ASD symptoms from the CASI-4R; parent- and teacher-rated psychiatric symptoms on the CASI-4R (ADHD, anxiety, depression, and mania); as well as academic functioning (being in an inclusion classroom) and social/play behaviors (repetitive play, preoccupation with certain objects, imitating others’ behaviors, having any friend, difficulty relating to peers). For these analyses, participants were assigned to their most probable class on the basis of the LCA results as previously described. In multiple comparisons for co-occurring psychiatric symptoms and social/play behaviors, familywise Bonferroni corrections were used to reduce the risk of Type I error (adjusted α-level was 0.0125 and 0.01, respectively). The pattern of significant differences between pairs of means was examined by post hoc comparisons by least significant difference (LSD) method.
Results
Past and current rates of atypical communication characteristics
Paired-samples t tests revealed that the mean number of ACCs decreased in the overall sample (t = −2.44, p = .015) and in psychiatry referrals (t = −2.19, p = .029; Figure 1A), but not the ASD group (t = −1.32, p = .19; Figure 1A). However, in the ASD group, 26% evinced no change in number of ACCs (i.e., parent-endorsed same number of past and current ACC), whereas 45% experienced a decrease (i.e., parent-endorsed fewer current than past ACC), and 29% experienced an increase (i.e., parent-endorsed more current than past ACC). In the non-ASD group, 84% had no change in number of ACCs, whereas 23% experienced a decrease and 7.1% experienced an increase in ACC (Figure 1B). The ASD group experienced greater change in number of ACCs than non-ASD group (ΔM = −.28 ± 3.15 vs. ΔM = −.13 ± 1.54).
Figure 1. Prevalence of atypical communication characteristics (ACCs) in autism spectrum disorder (ASD) and non-ASD samples in the past and in the present. (A) Mean number of ACCs in past and currently at the time of evaluation. *p < .05; ***p < .001. (B) Percentages of no change, decrease, or increase in number of ACCs.
Moderation analyses examining within-subject variations in change in number of ACCs from past to current revealed a significant interaction between diagnostic group and number of past ACCs in predicting number of current ACCs (B = −.23, p = .007). Post hoc analyses revealed a stronger association between past and current ACCs in non-ASD (B = .53, p < .001) than ASD (B = .31, p < .001) groups. In other words, within-subject variation in ACCs (i.e., change from past to current) was greater for youth with ASD than non-ASD psychiatry referrals.
Latent class analyses for the ASD group
Models positing two to five classes were evaluated in relation to the various measures of fit (see Table 2). The BIC was smallest for the three-class model. The AIC and the sample-size-adjusted BIC continued to decrease across the models considered, although the values did not decrease significantly after the three-class model. The results of the BLRT tests were significant for the two- and three-class solutions but not for four-, five-, or six-class solutions, suggesting that three classes fit better than a two-class model (p < .0001). The entropy values were greater than .85 for all solutions considered, indicating a good separation of the identified groups (Ramaswamy, Desarbo, Reibstein, & Robinson, Reference Ramaswamy, Desarbo, Reibstein and Robinson1993). Overall, the three-class solution emerged as the optimal fit for the data. Mean posterior probability of a person's likelihood of being assigned to the group was good (0.95).
Table 2. Criteria for assessing model fit for different number of classes for the autism spectrum disorder (ASD) sample
Note. AIC = Akaike information criterion; BIC = Bayesian information criterion; BLRT = bootstrapped likelihood ratio test. Bold indicates best fit according to each index.
a Indicates the optimal class selection based on overall fit.
As depicted in Figure 2, Class 1 (n = 69; identified as Stable ACCs) had the greatest number of current ACCs both present and past. Almost all participants (91%) exhibited stable pragmatic difficulties, and the majority exhibited stable speech delay (54%), repeating words (55%), and perseveration (54%), and some exhibited stable pattern of difficulties with facial expressions (29%), lecturing (25%), and being nonverbal (23%). Class 2 (n = 51; identified as Mostly Current-Only ACCs) evidenced high rates of current-only pragmatic difficulties (78%) and perseveration (63%), and moderate rates of current-only speech delay (45%), echolalia (45%), and high rates of stable symptoms of repeating words (63%) and neologism (26%). Class 3 (n = 102; identified as “Little Professors”) exhibited few current ACCs, but had moderate rates of stable perseveration (30%) and pragmatic difficulty (38%) and slightly elevated rates of current-only perseveration (9%) and pragmatic difficulties (16%). Figure 2 illustrates the frequency of exhibiting each ACC in the present only or stably (both in the past and in the present) given class membership.
Figure 2. Within the autism spectrum disorder (ASD) sample, probability of each atypical communication characteristic (ACC), given class membership, reflecting the percentage of individuals assigned to each class exhibiting each ACC at current only or at both past and current timepoints.
Latent class distinctiveness
Demographics
Differences were found between the three classes for age, such that Class 3 (Little Professors) was older than Class 2 (Mostly Current-Only ACCs; F(3,219) = 3.11, p = .046). Classes 1 (Stable ACCs) and 2 (Mostly Current-Only ACCs) evinced lower IQ than Class 3 (Little Professors; F[3,168] = 4.61, p = .011; Table 3), with the same pattern of effect seen in verbal (F[2,88] = .3.87, p = .025), but not in performance (p = .55) IQ. There were no differences between classes in terms of sex, ethnicity, or household income (all ps > .48).
Table 3. Demographic variables by latent class in the autism spectrum disorder (ASD) sample
Note. Class 1: “Stable Atypical Communication Characteristics (ACCs),” Class 2: “Mostly Current-Only ACCs,” Class 3: “Little Professors.”
*p < .05; **p < .01; ***p < .001.
ASD symptomatology
Classes differed in terms of ASD symptom severity across multiple measures (see Table 4).
Table 4. Differences in severity of autism spectrum disorder (ASD) symptoms and psychiatric symptoms across latent classes in the ASD sample
Note. SCQ = Social Communication Questionnaire. ADOS = Autism Diagnostic Observation Schedule Calibrated Severity Scores. CASI-4R = Child and Adolescent Symptom Inventory-4R. ASD = autism spectrum disorder symptoms. ADHD-H/I = attention deficit/hyperactivity disorder-predominantly hyperactive/impulsive symptoms. GAD = generalized anxiety disorder symptoms.
SCQ
Classes 1 (Stable ACCs) and 2 (Mostly Current-Only ACCs) had higher total scores than Class 3 (Little Professors; p = .003). With regard to specific subscales, Class 2 (Mostly Current-Only ACCs) had higher (i.e., more severe impairments in) reciprocal social interaction scores than Class 3 (Little Professors; p = .013). Classes 1 (Stable ACCs) and 2 (Mostly Current-Only ACCs) had higher repetitive and stereotyped behavior scores than Class 3 (Little Professors; p = .003).
ADOS-2
The three classes did not differ in terms of their CSS or restricted and repetitive behaviors scores (p > .093). However, Class 1 (Stable ACCs) had a higher (i.e., more severe impairment in) social affect score than Class 3 (Little Professors; p = .023), particularly in the communication subdomain (p = .038).
CASI-4R
Class differences were found in total CASI-4R ASD symptom severity. According to parent-report, Classes 1 (Stable ACCs) and 2 (Mostly Current-Only ACCs) had the higher total ASD as well as language and repetitive behaviors subdomain symptom severity scores than Class 3 (Little Professors; all p < .001), and Class 1 (Stable ACCs) had more social deficit symptoms than Class 3 (Little Professors; p = .042; Figure 3). According to teacher-report, Class 2 (Mostly Current-Only ACCs) had the higher total ASD as well as social deficits and repetitive behaviors subdomain symptom severity scores than Classes 1 (Stable ACCs) and 3 (all p < .01), and Classes 1 (Stable ACCs) and 2 (Mostly Current-Only ACCs) had more language symptoms than Class 3 (Little Professors; p < .001; Figure 3).
Figure 3. Mean scores on the Child and Adolescent Symptom Inventory, 4th edition (CASI-4R) for the autism spectrum disorder (ASD) symptom severity rated by parent and teacher.
Psychiatric symptoms and impairment
According to parent-report, Classes 1 (Stable ACCs) and 2 (Mostly Current-Only ACCs) had higher ADHD total and manic episode symptoms than Class 3 (Little Professors; p < .002). The ADHD effect was driven by ADHD hyperactivity/impulsivity subdomain (p < .001). According to teacher-report, Class 2 (Mostly Current-Only ACCs) had the highest ADHD total symptom scores (p = .012; Figure 4), driven by ADHD hyperactivity/impulsivity subdomain (p = .011; Table 4). There were no class differences in terms of anxiety or major depressive symptoms or symptom-induced impairment in either parent- or teacher-report (all ps > .038).
Figure 4. Mean subscale scores on the Child and Adolescent Symptom Inventory, 4th edition (CASI-4R) for the Attention-Deficit/Hyperactivity Disorder (ADHD) Combined symptom severity rated by parent and teacher.
Functional correlates
Classes differed in percentage receiving special education services in inclusion classroom: Class 2 (Mostly Current-Only ACCs; 88.2%) was the highest followed by Class 1 (Stable ACCs; 82.6%) and Class 3 (Little Professors; 64.7%). Class 2 (Mostly Current-Only ACCs) had the highest percentage of exhibiting preoccupation with objects (80%), followed by Class 1 (Stable ACCs; 73%) and Class 3 (Little Professors; 50%).
Discussion
It is de rigueur for clinicians to inquire about changes in clinical course in both diagnostic and follow-up evaluations in an effort to inform their decision making; nevertheless, the value of this strategy is rarely examined, particularly with respect to what it reveals about characterizing the phenotypic heterogeneity of ASD. Previously we reported that ACCs are associated with ASD symptomatology, comorbid psychiatric symptoms, and functional outcomes, suggesting ACCs may be a useful consideration for more precisely defining the ASD clinical phenotype (Kang et al., Reference Kang, Gadow and Lerner2020). In the present study we sought to determine whether an evaluation completed by the youth's parent indicating present and past ACCs was useful in addressing changes in the clinical course of ACCs and constructing data-driven, clinically informative subgroups. We compared stability of ACCs in clinic-referred children with and without ASD and ascertained the relation of stability of ACCs in youth with ASD with a broad range of psychiatric symptoms and clinical correlates. Youth with ASD exhibited higher rates of ACCs and a more variable pattern of stability (i.e., 29% increase and 45% decrease) than non-ASD psychiatry referrals (7% increase and 23% decrease). Above and beyond the associations with current ACCs (Kang et al., Reference Kang, Gadow and Lerner2020), our findings revealed that stability of these ACCs differentiated youth with ASD into three subgroups (Stable ACCs, Mostly Current-Only ACCs, and Little Professors). These subgroups evidenced differences in parent- and teacher-rated psychiatric symptom severity and functional outcomes. These findings suggest that ACC stability may reveal unique information regarding the relation between ASD pathogenic processes and psychiatric symptoms.
Patterns of stability of ACCs in ASD vs. non-ASD diagnostic groups
Results indicated those with ASD, on average, did not show decreases in mean number of ACCs. However, the ASD group demonstrated more fluid profiles (i.e., greater instability) of ACCs at an individual level than non-ASD psychiatry referrals (Stable: ASD = 26%; non-ASD = 84%). Our findings can be seen as a specific instantiation of the broader literature on atypical language development in ASD. Studies in this area have found language development in children with ASD to show substantial variability before age 6 years, but a relatively stable pattern beyond age 6 years (Pickles et al., Reference Pickles, Anderson and Lord2014), or from mid-childhood to adolescence (Sigman & McGovern, Reference Sigman and McGovern2005). More specifically, though, despite gaining receptive language ability as they grow older, ASD youth may have significantly more problems in broader communication skills than the non-ASD group with language disorder (Mawhood, Howlin, & Rutter, Reference Mawhood, Howlin and Rutter2000). Our findings support the contention that variability in ACCs is particularly common in those with ASD, and suggest that this pattern extends to late childhood and adolescence.
The present findings are more consonant with developmental plasticity models of ASD (e.g., Klin et al., Reference Klin, Jones, Schultz and Volkmar2003) that suggest the way individuals with ASD interact with their social world yields variations in symptom presentations across development. Although many youth with ASD experience persistent and elevated ACCs, we found greater variability in patterns of change in ACCs among the ASD group (Anderson et al., Reference Anderson, Lord, Risi, DiLavore, Shulman, Thurm and Pickles2007; Howlin, Mawhood, & Rutter, Reference Howlin, Mawhood and Rutter2000). These between-group differences in stability of ACCs suggest that the ASD diathesis increases risk for more variability in ACCs (i.e., perturbs the canonical pattern of increase in communicative capacity that is emblematic of normative development), further supporting the notion that change in ACCs is an important, unique clinical feature of ASD.
Subgroups in youth with ASD based on stability of ACCs
The results of LCA overall indicated three divergent patterns in stability of ACCs among youth with ASD: Class 1 (Stable ACCs) accounted for 31% of the sample and showed more stable patterns in all ACCs considered, with rates ranging from 10% to 91% of the participants, with pragmatic difficulty being the most prevalent stable ACC. Youth with Stable ACCs had higher rates of persistent speech delays, as well as tendency to repeat words and perseverate. To a lesser degree, they exhibit echolalia, as well as speaking with a notably unusual voice, as well as use of odd facial expressions. Consistency of these features appears to be a hallmark of this subgroup, with all of these features being straightforwardly identifiable via a speech-language evaluation or repeated ADOS administration. When such features co-occur earlier in development, they do not appear to often decrease independently—that is, sustained levels of any one or two of these features are likely to co-occur with sustainment of all of them. It may be that a common source of developmental language canalization (e.g., difficulty with joint attention can lead to less opportunities to learn various communicative functions in the context of a shared focus of attention; Luyster, Kadlec, Carter, & Tager-Flusberg, Reference Luyster, Kadlec, Carter and Tager-Flusberg2008) manifests itself in these ACCs concurrently. Such entwinement has implications for intervention, suggesting that an intervention that might affect one of these domains may have knock-on effects for the others. Future study of this subgroup may seek to identify such a diathesis common for this set of ACCs.
Class 2 (Mostly Current-Only ACCs), which accounted for 23% of the ASD sample, exhibited mostly current-only ACCs, except stable pattern of repeating words and neologism. This group most frequently experienced current pragmatic difficulties, perseveration, speech delays, and echolalia as well as some odd voice. This subgroup is notable for several reasons. First, it may be that later onset of these ACCs is associated with increased social demands, even in children who do not exhibit such characteristics earlier on; that is, perhaps they are a marker for “compensation”-type behaviors (Livingston, Shah, & Happé, Reference Livingston, Shah and Happé2019). Second, the co-occurrence of echolalia with ACCs associated with more complex language calls into question an older presumption that echolalia is an ACC mostly seen in individuals limited language ability (Fay & Schuler, Reference Fay and Schuler1980; Roberts, Rice, & Tager-Flusberg, Reference Roberts, Rice and Tager-Flusberg2004). While echolalia is observed clinically across the ASD spectrum, it is less frequently described in the empirical literature about more verbally-able individuals. Third, it appears that early and unitary emergence of repeating words (and to a lesser extent, neologisms) may be seen as a developmental precursor to the later emergence of the broader pattern of ACCs seen in this group. Future work should examine whether clinical observation of such a specific pattern early on can predict later onset of the full complement of features of this subgroup.
Class 3 (Little Professors) comprised 46% of the ASD sample. This subgroup had fewer ACCs, with some elevations in primarily stable pattern of pragmatic difficulties and perseveration, resembling Little Professors. Such a subgroup has, of course, been well-described in the literature for decades (Asperger, Reference Asperger1944; Kang et al., Reference Kang, Gadow and Lerner2020; Wing, Reference Wing1981). A prior examination of ACCs in youth with ASD revealed that, while most (91%) exhibit one or more current ACCs, the rates of ACCs varied greatly by each ACC, with the majority of youth with ASD displaying pragmatic difficulties (72%) and perseveration (52%; Kang et al., Reference Kang, Gadow and Lerner2020). Therefore, it was not too surprising that a large proportion of the sample would show a stable pattern in these ACCs. It remains notable, though, how discrete the ACCs of pragmatics and preservation are in this group—infrequently co-occurring with any other ACC. Future research may seek to examine how and why these specific ACCs should present in this way, and whether individuals in this subgroup have early experiences (such as interventions) that may act preventatively against emergence of other ACCs.
It is interesting that the stability of ACCs seemed to hang together within these subgroups across the range of measured behaviors. This suggests that, regardless of specific ACCs, those who experience greater variability in ACCs are more likely to exhibit changes across most ACCs. That said, even for these youth with fewer early issues but with current ACCs (Mostly Current-Only ACCs; Class 2), they were more likely to be described by their parents as repeating words in both the past and the present. In comparison to those who currently exhibit few ACCs (Little Professors; Class 3), rates of repeating words were consistently elevated in youth who currently exhibit ACCs. In fact, having this ACC in isolation may be an indicator of increase in ACCs.
These subgroups differed in age and IQ (specifically in verbal IQ). On average, Little Professors (Class 3) were older than Mostly Current-Only ACCs (Class 2), whereas they did not differ in age from Stable ACCs (Class 1). That there was no age difference between those who present with elevated rates of current ACCs (Stable ACCs and Mostly Current-Only ACCs) suggests that stability or change may not be related to different developmental periods or dependent on the age of the child. That said, we acknowledge that our metric of ACC stability was not assessed longitudinally; therefore, we cannot assess whether this metric is or is not an age-dependent construct in any definitive way. A longer developmental period than the age range considered in this study (6–18 years) or consideration of specific timepoints may be required to identify and differentiate more developmentally stable patterns of ACCs among Stable ACCs or Little Professors. Little Professors (Class 3) also had higher IQ, specifically verbal IQ, than the other two classes that are more severely affected with ACCs. This is consistent with clinical descriptions of those who are described as little professors by Asperger (Hippler & Klicpera, Reference Hippler and Klicpera2003) and with diagnostic criteria for Asperger's disorder in the DSM-IV (i.e., a normal cognitive functioning and the absence of significant general delay in language). Therefore, those who present with perseveration and pragmatic difficulties in the absence of other ACCs (even early on) may require different intervention targets and strategies that capitalize on their strengths more than in the other two classes (Mottron, Reference Mottron2017). Notably, the three classes did not differ in terms of performance IQ, despite association in the literature between nonverbal skills and language acquisition (Wodka, Mathy, & Kalb, Reference Wodka, Mathy and Kalb2013). Our findings support the notion that many individuals with ASD with substantial verbal delay—or even those lacking communicative speech—have higher nonverbal IQ than verbal IQ (Siegel, Minshew, & Goldstein, Reference Siegel, Minshew and Goldstein1996), and that this may reflect a reorganization of brain allocation and perceptual function (Samson, Mottron, Soulières, & Zeffiro, Reference Samson, Mottron, Soulières and Zeffiro2012). The three classes did not differ by other demographic characteristics, such as sex, ethnicity, medication status or household income, suggesting that these variables are not contributory to development of ACCs.
Stability of ACCs and clinical correlates
Our results partially suggest that subgroups with more stable versus temporally dynamic ACCs differ in terms of ASD and co-occurring psychiatric symptom severity. Specifically, while parents and teachers converged in their report of higher language-related ASD symptoms in Stable ACCs and Mostly Current-Only ACCs (Classes 1 and 2) relative to Little Professors (Class 3), severity of other domains of ASD symptoms differed based on the informant between the two classes. That is, although Stable ACCs and Mostly Current-Only ACCs (Classes 1 and 2) did not differ in that both classes presented with more severe parent-reported repetitive behaviors than Little Professors, Mostly Current-Only ACCs demonstrated the most severe teacher-reported social deficits and repetitive behaviors; Stable ACCs displayed more severe examiner-rated social affect, specifically communication, relative to Little Professors. Overall, the finding that the three classes differed across ASD symptoms in non-language domains (e.g., repetitive behaviors) further underscores the finding that ACCs represent a core aspect of the broader ASD clinical phenotype (Kang et al., Reference Kang, Gadow and Lerner2020), and that their stability may reveal insights into other aspects of this phenotype.
It is also notable that there is a discrepancy between informants on severity of symptoms when comparing the stability and change in ACCs (i.e., comparing Stable ACCs with Mostly Current-Only ACCs). It is possible that the association of ACCs with other ASD symptoms is comparable in the home setting, whether these ACCs are stable or they are new symptoms, due to relatively lower social demands in this setting. On the other hand, ASD symptoms may be more evident in school settings where social demands are relatively higher, for those who present with later onset of ACCs as they may not have had the sufficient opportunities to address these symptoms. It may be important to further examine how change in ACCs relates to these informant-related differences in ASD symptoms, which may shed light on who may benefit from increased school-based support for ASD symptoms as newer ACCs emerge. Indeed, researchers have begun to investigate how the degree of parent–teacher discrepancy about ASD symptom severity provides information about patterns of phenotypic presentation in ASD and the impact of these patterns on school-based supportive services (Lerner, De Los Reyes, Drabick, Gerber, & Gadow, Reference Lerner, De Los Reyes, Drabick, Gerber and Gadow2017).
In addition, the three subgroups differed in ADHD symptomatology and group differences varied as a function of who was completing the evaluation. Together with evidence that difficulties in speech and language development are evident in up to two-thirds of youth with ADHD (Gurevitz, Geva, Varon, & Leitner, Reference Gurevitz, Geva, Varon and Leitner2014), our finding supports the notion of difficulties in communication and linguistic skills as a point of convergence between ASD and ADHD (Geurts & Embrechts, Reference Geurts and Embrechts2008), and highlights specific roles that ACCs may play in ASD and ADHD comorbidity. It was not unexpected that patterns of class differences varied as a function of informant (parent vs. teacher): similar to ASD symptoms, whereas parents rated higher symptoms in Stable ACCs and Mostly Current-Only ACCs (Classes 1 and 2) compared to Little Professors (Class 3), teachers rated higher symptoms in Mostly Current-Only ACCs (Class 2) compared to other classes. ADHD symptoms may be more evident in school settings for youth in Mostly Current-Only ACCs (Class 2), as appropriate level of support or school-based services to address these new symptoms may not be in place yet. Taken together with findings on ASD symptoms, these findings suggest that each informant (e.g., parent, teacher, clinician) may add useful information regarding ASD and ADHD symptoms that typically vary across home, school, or clinical settings (De Los Reyes, Reference De Los Reyes2011; Gadow et al., Reference Gadow, Drabick, Loney, Sprafkin, Salisbury, Azizian and Schwartz2004b; Gadow et al., Reference Gadow, DeVincent, Pomeroy and Azizian2005), and these variations may be related to different pattern of stability of ACCs. Overall, results suggest that distinct patterns of stability in ACCs place youth with ASD at a differential risk for psychiatric outcomes related to ADHD. Conversely, despite the widely-held notion of a link between language problems and a broad range of psychiatric symptoms and disorders (e.g., Yew & O'Kearney, Reference Yew and O'Kearney2013), classes did not differ in severity of other psychiatric disorders supporting the notion of symptom-specificity.
In addition to ASD and ADHD symptomatology, patterns of ACC stability were associated with other clinical correlates, such as play behaviors and special education services. Specifically, the Mostly Current-Only ACCs (Class 2) subgroup had the highest rates of preoccupation with certain objects. Also, Mostly Current-Only ACCs (Class 2) had the highest rates of receiving special education, followed by Stable ACCs (Class 1) and Little Professors (Class 3). It is interesting that the three classes differed in rates of ASD and ADHD symptoms especially in the context of different special education outcomes. To date, only a few studies have examined psychiatric symptoms on the delivery of school-based services for youth with ASD. For example, co-occurring ADHD in youth with ASD either does not predict school-based services despite associations with a greater need for services (Narendorf, Shattuck, & Sterzing, Reference Narendorf, Shattuck and Sterzing2011) or higher parent-rated externalizing symptoms predict lower likelihood of receiving school-based services (Rosen, Spaulding, Gates, & Lerner, Reference Rosen, Spaulding, Gates and Lerner2019). This is in contrast to the current study, in which higher teacher-rated ADHD symptoms predicted higher rates of receiving special education.
Clinical implications
The primary aim of this study was to determine if parent-rated changes in ACCs as assessed at time of initial evaluation (i.e., using simple method of assessing whether the current ACC represents a stable presentation or a new emergence of something that the youth did not experience before) has the practical value of revealing differences in clinical symptoms across settings. Our results suggest that a brief parent-completed questionnaire regarding stability of ACCs—even in the absence of an exact time frame of these patterns of stability—provides useful information about clinical presentation, particularly for ASD and ADHD symptoms, in different settings (e.g., home vs. school). In particular, information collected in this manner may provide an indicator that a more comprehensive psychiatric evaluation (i.e., for ADHD symptoms) is needed, and warrants consideration in future research in the development of refined assessment, especially in low-resource settings where access to comprehensive psychiatric evaluations are limited and collection of relevant data in the past did not take place.
Specifically, the present assessment approach can guide the establishment of primary and secondary treatment targets. Variability in patterns of ACCs underscores the importance of flexible intervention strategies that adapt to the changing needs of children with ASD. Identifying stability of ACCs in youth with ASD may help to guide the need for (and approach to) interventions for symptoms in other domains (e.g., repetitive behaviors, ADHD). For example, given that patterns of change in ACC showed unique relations to co-occurring symptoms (i.e., ADHD) in the ASD youth, designing adequate treatments for ACC may further improve effectiveness of interventions for ASD for which psychiatric comorbidities can interfere with treatment outcome (e.g., Antshel et al., Reference Antshel, Polacek, McMahon, Dygert, Spencely, Dygert and Faisal2011). Discrepancy in parent- and teacher-reported ASD and ADHD symptoms depending on whether the youth presents with a more stable pattern of ACC or with mostly newly emerging ACCs underscores the importance of developing empirically supported intervention strategies to address different symptom presentation across settings.
Further considerations on a developmental framework for stability of ACCs
Examination of the pattern of stability of ACCs and its relations to psychiatric symptoms and functional correlates provides further support for developmental models of ASD. Communication development is a complex process influenced by numerous environmental and genetic factors. Whereas shared genetic markers between ASD and other types of language impairment have been reported (Eicher & Gruen, Reference Eicher and Gruen2015), ASD pathogenic processes and their interaction with the environment may alter ACC processes to result in unique clinical presentation, resulting in greater pattern of variability in ASD than non-ASD psychiatry referrals. In addition, it may be the case that, as is the case with Stable ACCs (Class 1), sustained difficulties in developmental domains (such as ACCs) may have a cascading effect by placing these youth at a greater risk for psychiatric symptoms (e.g., ADHD). However, it is possible that common genetic factors that are related to ACCs are activated at different time points in development by other genetic or environmental factors (Johnson, Gliga, Jones, & Charman, Reference Johnson, Gliga, Jones and Charman2015). Future studies should attempt to incorporate measurement across levels of analysis to better understand pathways or processes that underlie the relations of ACCs with ASD morbidity and psychiatric comorbidity.
In addition, ACCs may be considered an example of modifier processes (Mundy, Henderson, Inge, & Coman, Reference Mundy, Henderson, Inge and Coman2007): whereas ACCs do not occur exclusively in ASD, different patterns of ACC stability may uniquely alter the expression of ASD and contribute to fundamental behavioral and psychological differences, including patterns of psychiatric symptoms, in this population. Therefore, a more fine-grained analysis of how stability of each ACC may impact clinical correlates may better understanding of the role that each ACC plays as a modifier process in shaping the ASD clinical phenotype, including co-occurring psychiatric conditions and functional outcomes.
Strengths, limitations and future directions
This study has several notable strengths including a large, well-characterized sample of ASD and non-ASD clinic-referred youth and a well-validated psychiatric symptom rating scale. Nevertheless, there are several limitations that constrain generalizability of the results. First, the study is cross-sectional; therefore, inferences about causality or direction of effects cannot be made. Crucially, although information on the age of participant is available for the current ACCs at the time of assessment, there is no information on the exact age when the ACC was present in the past and thus the exact timeframe of stability of ACCs. Future studies should employ retrospective assessment strategies in conjunction with a longitudinal, prospective design with routine assessment of ACC in order to characterize the relation between ACC stability and the ASD clinical phenotype with more temporal precision. Also, such longitudinal assessments will permit further evaluation of whether examining number of ACCs over time with early ASD symptom presentation may be useful in screening for ASD, or vice versa. Second, the study samples comprise youth referred for outpatient evaluation and obtained results may therefore not generalize to the general population. This having been said, as the purpose of the study is about routine clinic evaluation, the samples are highly suited to the goals of the study. Third, parent report was used for assessing ACC rather than formal speech and language assessments. That said, many ASD assessment instruments used in routine, everyday clinical settings are based on parent report, and research indicates that language profiles based on parent ratings are consistent with behavioral and clinician ratings of language (Hauerwas & Stone, Reference Hauerwas and Stone2000; Thal, O'Hanlon, Clemmons, & Fralin, Reference Thal, O'Hanlon, Clemmons and Fralin1999). Moreover, the current study examined stability in the presence of specific ACCs and therefore did not consider changes in severity of ACC over time. Indeed, little literature exists to quantify the nature of severity of ACCs in general. Future studies may extend our findings to investigate the role of severity within each ACC item and psychiatric functioning. In addition, while we did not find strong evidence of external validation of differences between Stable ACCs and Mostly Current-Only ACCs (Classes 1 and 2) against some clinical correlates we considered, our findings underscore that these two classes are distinct from a classification standpoint. Therefore, it will be important for future work to further examine if and how they differ on other measures that are relevant to the ASD population (e.g., genetic background, load of developmental disability within the family, ancillary services received). Lastly, it is also important to note that the entropy values did not exceed .90, which is considered excellent certainty about class assignment. That said, entropy should always be examined in conjunction with other model fit indices. Also, given the nature of ACCs and possible common subconstructs, residual correlations between indicator variables (ACCs) and multicollinearity are possible. However, when comparing models with uncorrelated residuals to correlated residuals between ACCs, models with uncorrelated residuals appear to demonstrate superior fit in similar datasets (Kang et al., Reference Kang, Gadow and Lerner2020).
Summary
In the present study, we examined patterns of stability of ACCs in clinic-referred youth with and without ASD using a simple retrospective method of assessing stability. Youth with ASD had higher rates and more variable patterns of change in ACCs than non-ASD psychiatry referrals. The patterns of stability of ACCs yielded unique subgroups within youth with ASD, which predicted ASD morbidity and co-occurring psychiatric symptoms, as well as functional correlates. Findings provide support for developmental models of ASD, where different patterns of ACCs may differentially alter symptom expression in youth with ASD and give rise to clinically divergent populations. Results support continued research into the clinical utility of characterizing caregiver-perceived changes in ASD associated features.
Acknowledgements
This study was supported, in part, by the Matt and Debra Cody Center for Autism and Developmental Disorders. The funder had no role in study design; the data collection, analysis, and interpretation; manuscript writing; and the decision to submit the article for publication. The authors wish to thank Dr. John Pomeroy, M.D., for directing the ASD diagnoses and Carla DeVincent, Ph.D., for coordinating data collection. Kenneth D. Gadow is shareholder in Checkmate Plus, publisher of the Child and Adolescent Symptom Inventory. Erin Kang and Matthew D. Lerner report no biomedical financial interests or potential conflicts of interest.
