Individuals with autism spectrum disorder (ASD) tend to have significant delays in adaptive functioning. In this study, the relationship between adaptive behavior and ASD symptomatology was investigated in minimally verbal, school-aged children with ASD (n = 333). Both the social affect (SA) and restricted and repetitive behavior (RRB) domains from the Autism Diagnostic Observation Schedule (ADOS) were analyzed in relation to adaptive skills. ADOS SA scores contributed unique variance to scores in each Vineland domain, though cognitive ability and age accounted for considerably more variance across domains. Results indicate that there is a significant, but small, association between social affect deficits and adaptive skills, challenging clinicians, educators, and caregivers to target adaptive skills in addition to more specific features of ASD.
Among individuals with developmental disabilities, including autism spectrum disorders (ASD), measurements of adaptive skills indicate the degree to which a person is able to function independently, or needs support, in daily life. Cross-sectional research has shown that, in childhood, standard scores of adaptive behavior tend to decrease with increased age for individuals with ASD, suggesting that this population does not develop adaptive skills at a rate expected from standardized norms (Bal, Kim, Cheong, & Lord, 2015; Kanne et al., 2011; Pugliese et al., 2015). This has also been demonstrated in longitudinal research, showing that, over time, children with ASD show decreases in developmental quotient scores of adaptive behavior (Fisch, Simensen & Schroer, 2002), although standard scores may remain more stable in more-able children (Pugliese et al., 2016). Although most individuals with ASD have clear deficits in adaptive skills, the relationship between ASD symptomatology and adaptive functioning remains unclear. In order to understand how to support the independent functioning of individuals with ASD, it is important to identify the factors associated with development of adaptive skills.
The current body of literature in this area suggests that adaptive functioning deficits in ASD can be related to several factors, including cognitive ability and core ASD symptoms. Within a cognitively able (i.e., higher-functioning) sample of individuals with ASD, an individual's level of independent functioning may be more directly associated with his or her diagnosis of ASD than in less cognitively able children, because he or she has stronger cognitive abilities to support the development of adaptive skills. However, about one-third of people with ASD have comorbid diagnoses of intellectual disability (Centers for Disease Control and Prevention, 2014). Clear associations between cognitive level and adaptive functioning in ASD have been demonstrated (Kanne et al., 2011; Liss et al., 2001; Ray-Subramanian, Huai, & Weismer, 2011). Perry, Flanagan, Dunn Geier, and Freeman (2009) found that children with ASD demonstrate more delayed adaptive behavior profiles than children with intellectual disability alone when matched by IQ. Children with ASD who had IQs above 70 had significantly higher IQ scores relative to their adaptive behavior scores, whereas children with IQs below 70 had significantly lower IQ scores than adaptive behavior scores (Perry et al., 2009). When taken together, these findings indicate that cognitive ability alone does not predict adaptive functioning within ASD, suggesting that ASD symptoms may play a unique role. Furthermore, in a heterogeneous sample of people with both intellectual disabilities and other disabilities, including ASD, Harries, Guscia, Nettelback, and Kirby (2009) found that, while controlling for severity of intellectual disability (ID), the number of disabilities predicted adaptive skill, with more disabilities predicting fewer adaptive skills. For children with ASD and comorbid intellectual and language impairments, it is particularly important to understand if and how ASD symptoms uniquely relate to the everyday ability to function in daily life. Research about the relationship between ASD and adaptive functioning has implications not only for clinical practice and research in the ASD community, but also for other populations with intellectual disabilities.
Given the relative deficits in adaptive skills within the ASD population, there has been some research on the relationship between ASD symptomatology and adaptive functioning. In these studies, adaptive functioning is often assessed using the Vineland Adaptive Behavior Scales (Vineland; Sparrow, Balla, & Cicchetti, 1984; Sparrow, Cicchetti, & Balla, 2005). The Vineland is a semistructured parent interview that provides information about a child's adaptive functioning in the domains of Communication, Socialization, Daily Living, and Motor Skills (up to age 7). In contrast, instruments used to measure ASD symptomatology have been much more varied across studies. In two studies that examined adaptive behavior in less cognitively able individuals with ASD, relationships were found between ASD symptomatology and domains of adaptive skills. Perry et al. (2009) examined adaptive behavior in a sample of children of varying cognitive levels who were younger than 6 years old. They found ASD severity, as measured by the Childhood Autism Rating Scale (CARS; Schopler, Reichler, & Renner, 1988), to be significantly correlated with all domains of adaptive behavior from the Vineland, with associations between standard scores ranging from r = −0.33 (Motor Skills) to r = −0.57 (Socialization). The CARS is an instrument used to characterize individuals with ASD and yields a single severity score that combines item scores for ASD-specific symptoms, as well as language level and intelligence (Schopler et al., 1988). In another paper, Liss et al. (2001) separated domains of symptomatology using the Wing Autism Diagnostic Interview Checklist (Rapin, 1996), which collects information regarding Socialization, Communication, and Restrictive/Repetitive Behaviors that reflect impairments often seen in a population with ASD. In a group of children with nonverbal IQs below 80, they found significant correlations between social impairments on the Wing Checklist and the Vineland Socialization domain (r = −0.36), while other domains were unrelated. Interestingly, however, in a group of children with ASD with nonverbal IQs above 80, they did find significant relationships between all domains of ASD symptomatology and adaptive behavior. This finding suggests that separating different domains of ASD symptomatology may yield unique findings, with the recognition that these relationships may vary across cognitive levels within the ASD population.
Considerable evidence supports the Diagnostic and Statistical Manual of Mental Disorders—Fifth edition (DSM-5; American Psychiatric Association, 2013) concept of separating dimensions of ASD symptoms (Frazier et al., 2012; Mandy, Charman, & Skuse, 2012; Snow, Lecavalier, & Houts, 2008). This is further supported by investigations of the relationship between adaptive functioning and ASD symptoms in verbal samples. When Kanne et al. (2011) explored the relationship between ASD severity, using the Autism Diagnostic Observation Schedule (ADOS; Lord et al., 2000; Rutter, DiLavore, Risi, Gotham, & Bishop, 2012) overall calibrated severity score, and adaptive functioning in a large sample of verbal children with ASD between the ages of 4–17, they found no significant correlations between totals. However, when Klin et al. (2007) specifically investigated how social and communication behaviors related to adaptive skills in a sample of verbal 7- to 18-year-olds, they found significant but weak correlations (ranging from r = −0.20 to r = −0.28) between adaptive socialization and communication skills and impairments.
The ADOS is a well-established instrument for diagnosing ASD. It is administered by an objective rater, and provides algorithm totals for a Social Affect subscale and a Restricted/Repetitive Behaviors subscale (Gotham, Risi, Pickles, & Lord, 2007). To our knowledge, there is currently no literature available that investigates these separate domains of ASD symptomatology as they relate to domains of adaptive functioning, in a group of minimally verbal school-aged children with ASD. Thus, for the present study, we sought to elucidate the unique contribution of ASD symptomatology in predicting adaptive skills in a sample of children impacted by significant language delays and ASD. Due to the nature and course of ASD, we expected to find a negative correlation between ASD-specific deficits in social affect and adaptive socialization and communication skills. We did not expect to find analogous relationships between impairments in restricted and repetitive behaviors and adaptive skills. The present study investigated whether ASD-specific impairments contributed unique variance to adaptive behavior scores, above and beyond what could be accounted for by chronological age and developmental level.
The sample included children with ASD whose parents provided consent to contribute information to a research database. All children received a best estimate clinical diagnosis (BEC) of ASD from, in most cases, two clinicians (e.g., clinical psychologist, psychiatrist). BECs were based on all available information from a comprehensive evaluation, including cognitive and diagnostic testing and parent interviews. ASD includes those who had a BEC of Asperger syndrome, pervasive developmental disorder—not otherwise specified, and autistic disorder.
Three hundred and forty-one children were selected from an existing research database of consecutive referrals to six different clinics whose families consented to participate. Clinics, which we later refer to as “sites,” were located in Ann Arbor, MI; North Carolina; Cincinnati, OH; White Plains, NY; and two in Chicago, IL. Children were between 4 and 10 years of age (49 to 119 months) and were minimally verbal, as determined by having no or single-word speech during Module 1 of the ADOS. Two participants were excluded based on having a nonverbal mental age (NVMA) below 10 months, in which case the ADOS was unlikely to be an appropriate evaluation. Additionally, participants who did not meet autism spectrum cutoffs on the ADOS algorithm totals were eliminated from our dataset (n = 6). This left us with a total of 333 children (79.3% male), ranging in age from 49 to 119 months with a mean age of 71.35 months (SD = 18.35). The majority of the sample identified as White (73.3%), but included African American (12.9%) and Asian, Pacific Islander, Native American, Biracial, or Other (9%) racial identities as well. Additionally, 4.5% of the sample identified as Hispanic and 4.8% of the sample chose not to report race. Further characterization of the sample can be found in Table 1.
The current study aimed to compare scores across multiple instruments. Despite concerns about using age equivalents (Maloney & Larrivee, 2007), we chose to use age equivalents from the Vineland and cognitive instruments and raw scores from the ADOS. Standardized scores are calculated in several different ways across instruments (e.g., deviation IQs, T-scores, clinical global impressions), however, most take into account age, while some control for language level and/or general level of functioning. In an effort to take these differences into account in a fair way, we chose raw and age equivalents scores so as not to control for language level or age in any instrument. Raw scores from the Vineland and cognitive instruments, instead of age equivalents, yielded very similar results, but because age equivalents were more interpretable, we report them here. We covaried chronological age throughout analyses. Table 1 includes ADOS raw scores and Vineland age equivalents, as well as Vineland, IQ, and ADOS (severity) standard scores for reference.
Adaptive behavior was measured using the first or second edition of the Vineland Adaptive Behavior Scales—Survey form (Vineland; Sparrow et al., 1984; Sparrow et al., 2005). The measure yields age equivalents for each subdomain, which were averaged to create a domain age equivalent. The majority of our sample (n = 208) received a raw score of 0 on the written subscale of the communication domain, therefore an estimated communication domain age equivalent was calculated based on the expressive and receptive subdomains only. An Adaptive Behavior Composite (ABC) age equivalent score was calculated by averaging the domain scores (all four domains for individuals under 7 years old, and the Communication, Socialization, and Daily Living domains for older individuals).
To measure ASD symptomatology, research-reliable staff administered versions of the ADOS - Module 1. The ADOS is a semistructured assessment consisting of various tasks meant to elicit different social-communicative and play behaviors. A subset of items (“algorithm items”) are divided into two subscales: Social Affect (SA) and Restricted and Repetitive Behaviors (RRB). The present study used raw totals from both subdomains, as well as an overall total score, all from Module 1. All children met algorithm cut-offs for an ADOS classification of ASD, with children who used fewer than five words during the ADOS having an overall raw score of at least 11 and children who used at least five words having a raw score of at least 9. In most cases, the same examiner administered both the ADOS and the Vineland, usually with the Vineland several days before the ADOS; however, each measure was scored independently of the other.
Verbal and nonverbal intelligence were measured using a hierarchy of developmentally appropriate cognitive assessments. For the majority of our sample, nonverbal intelligence was measured using the Mullen Scales of Early Learning (n = 198; Mullen, 1995), the Differential Ability Scales (n = 98; Elliot, 1990), or the Merrill-Palmer-Revised (n = 25; Roid & Samper, 2004). A minority of participants received the Leiter International Performance Scale – Revised (n = 6; Roid & Miller, 1997) or the Raven's Progressive Matrices Test (n = 6; Raven & Court, 1998), in which case verbal intelligence was ascertained from another measure. In cases where a child's score fell outside the range of standardized norms, or the instrument did not provide standard scores, IQ was estimated using a Ratio IQ ([MA/CA]x100) for descriptive purposes. Ratio IQs were calculated for 13.5% of nonverbal scores and 12% of verbal scores. NVMA was ascertained from the instrument from published norms if available; otherwise, NVMA was calculated in a similar fashion ([CA*IQ]/100; n = 46).
First, we ran correlations to yield a simple indication of how the variables of interest related to one another. Subsequently, partial correlations were run to examine the relationship between ASD symptomatology and adaptive skills, controlling for the effects of age, NVMA, and site. We also ran a series of planned hierarchical regression analyses to elucidate the unique contribution of ADOS SA and RRB scores in predicting Vineland age equivalents. Separate analyses were run for each Vineland domain, with NVMA, age, and site entered at Step 1 and ADOS SA or ADOS RRB scores entered at Step 2.
Our preliminary correlation analysis, shown in Table 2, revealed that the ADOS SA domain raw total was negatively correlated with all Vineland domain age equivalents; the ADOS RRB domain was negatively correlated with the Vineland Socialization domain, Daily Living Skills domain, and Adaptive Behavior Composite (ABC); and the ADOS Total was negatively correlated with all Vineland domains. NVMA was negatively related to all domains of the ADOS, with the ADOS Total raw scores demonstrating the strongest relationship with NVMA. NVMA was also significantly related to all domains of the Vineland, having the highest correlations with the Vineland Daily Living Skills domain age equivalents. Age was also significantly related to ADOS SA scores, ADOS Total scores, and the Vineland Daily Living age equivalents.
Note. NVMA = Nonverbal mental age. ADOS = Autism Diagnostic Observation Schedule. SA = Social affect. RRB = Restricted/Repetitive behavior. ABC = Adaptive Behavior Composite.
p < .05, two-tailed. **p < .01, two-tailed.
An analysis of variance (ANOVA) of key variables by clinic site revealed significant differences in age (F[5,327] = .18, p < .001) and overall adaptive functioning (F[5,313] = 15.20, p < .001). Mean age across sites ranged from 67 to 108 months, while mean Vineland ABC age equivalents ranged from 21 to 36 months. Therefore, site was included as a covariate in analyses.
The results of our partial correlation analyses are displayed in Table 3. As shown, ADOS RRB scores were no longer a significant predictor of any Vineland domains when controlling for covariates, while the ADOS Total remained significant.
Note. ADOS = Autism Diagnostic Observation Schedule. SA = Social affect. RRB = Restricted/Repetitive behavior. ABC = Adaptive Behavior Composite.
p < .01, two-tailed.
A series of planned hierarchical regression models predicting Vineland domains, with NVMA, age, and site entered at Step 1 and ADOS SA entered at Step 2, revealed that ADOS SA scores contributed unique variance in every model, though, in each of them, NVMA and age accounted for the most substantial amount of the variance (Figure 1). The total explained variance of the Vineland Socialization domain was 16%, with ADOS SA scores accounting for 4% unique variance, after controlling for NVMA and age. For the Vineland Communication domain, the total variance explained was 26%, with ADOS SA scores accounting for 2% unique variance, after controlling for NVMA and age. For the Vineland Daily Living Skills domain, the total explained variance was 37%, with ADOS SA scores accounting for 2% unique variance, after controlling for NVMA and age. Across Vineland domains, the effect size of ADOS SA was small, while the effect size of NVMA and age was medium to large (Cohen, 1988). The results of the ADOS SA regression models are shown in Table 4. As expected, ADOS RRB scores did not account for a significant amount of unique variance in adaptive skills, explaining between 0.2% and 0.4% of the variance in Vineland domain scores. Site did not contribute unique variance in adaptive skills.
The present study investigated the relationship between adaptive functioning and different domains of ASD symptomatology, as measured by the Vineland and the ADOS respectively, in a sample of minimally verbal, school-aged children with ASD. Because social communication deficits are a core feature of ASD, we were particularly interested in the relationship between the Vineland Socialization domain and the ADOS Social Affect subscale. Additionally, we examined whether severity of ASD-specific impairments in social affect contributed unique variance to other adaptive behavior scores above and beyond what could be accounted for by chronological age and developmental level.
As expected, NVMA was positively correlated with adaptive behavior age equivalents, and was significantly (but less strongly) negatively associated with social communication and repetitive symptom severity. Chronological age was largely unrelated to adaptive functioning, with the exception of a modest correlation with daily living skills. Previous studies have found negative correlations between chronological age and standard scores of adaptive behavior, suggesting children with ASD do not tend to develop adaptive skills at a typical rate as they grow older (Klin et al., 2007; Perry et al., 2009). However, in this sample, a positive correlation was found between chronological age and age equivalents for daily living skills, which is encouraging. Although minimally verbal children are characterized by significant delays across domains of adaptive functioning, they are continuing to develop daily living skills as they grow older, albeit at a slower rate. This sample also showed a relative strength in daily living skills, as compared to socialization and communication skills, replicating a characteristic “autism profile” found in other studies (Carter et al., 1998; Kraijer, 2000).
Consistent with findings from several previous studies using various measures of ASD symptoms, this study confirmed the mild to moderate association between impairments in social affect and skills in adaptive socialization. As expected, we found a significant relationship between ADOS Social Affect raw algorithm totals and the Vineland Socialization domain age equivalents, with more severe ASD symptoms being associated with fewer adaptive socialization skills. The convergence between a “snapshot” of behavior, observed by a clinician during the ADOS, and parent reports of the child's social adaptive functioning elsewhere is very encouraging. Not surprisingly, social affect symptoms were also related to communication and daily living skills.
Regression analyses clarified the amount of unique variance social affect symptoms contributed toward the adaptive functioning domains, above and beyond what could be accounted for by NVMA and chronological age. Social affect symptoms accounted for 4 percent of unique variance in adaptive socialization skills. Although this is a modest contribution, it represents a substantial proportion of the model overall, which explained 16 percent of the variance. Within the other adaptive functioning domains, NVMA in particular contributed to more variance, while social affect symptoms contributed less. The fact that social affect symptoms were related to adaptive functioning while restricted and repetitive behaviors were not suggests that it is indeed theoretically useful to separate out these types of features when investigating adaptive functioning.
Though our findings are statistically significant, they are not necessarily clinically significant. As in other studies, IQ accounts for the majority of variance in adaptive skills for our sample. However, IQ is not as malleable as other characteristics, particularly in a school-aged sample. Additionally, many ASD-specific interventions target skills related to social affect (e.g., joint attention, communication, social skills), which is why it is worth looking into whether impairments in social affect relate to adaptive functioning beyond the known effect of IQ. From this study, we cannot make statements about the effect of targeting social affect symptoms over time, but it is worthwhile for future studies to examine a possible causal relationship.
The current study has several limitations. Adaptive skills were related to ASD symptomatology in this group of children with lower cognitive abilities; however, this relationship may change in individuals with higher levels of cognitive functioning or verbal ability. Further, the relationship between ASD symptomatology and adaptive functioning may change over time within an individual (Kim, Bal, & Lord, 2014). In a longitudinal study, Szatmari and colleagues (2015) found differing trajectories within these two domains, with ASD symptoms remaining more stable over time. We would expect the contribution of social communication skills to be even higher among a population of children with intellectual disability with and without ASD, where the range of ADOS scores would be greater.
The relationship between ADOS and Vineland scores, particularly when considering that measures ranged across contexts (clinicians and parents), supports the importance of social-communicative behaviors in contributing to independence. However, given the small amount of unique variance accounted for by social communication deficits in this and other studies, our findings also challenge clinicians, educators, and caregivers to continue targeting adaptive skills directly, in addition to more specific ASD features. Future research should investigate the relationship between social communication symptoms and adaptive functioning over time. Additionally, it would be beneficial to know whether intervention-related changes in social communication skills would be reflected in changes in adaptive functioning. It will also be useful to understand the degree to which the dimensions that define ASD, social communication deficits, and repetitive behaviors are related to adaptive functioning in other populations (non-ASD) with intellectual disabilities. This may be particularly true in groups where ASD symptoms are common, but not universal, such as those with Angelman syndrome, Fragile X syndrome, Down syndrome, and other populations with intellectual impairments.
This work was supported by the Eunice Kennedy Shriver National Institute of Child Health and Human Development of the National Institutes of Health under award number R01HD081199 to Catherine Lord and R01HD073975 to Connie Kasari.
We are grateful to all of the families who contributed to our research database. In addition, we would like to thank Shanping Qiu for her assistance in data management, as well as New York Presbyterian Hospital and the Center for Autism and the Developing Brain, where our database is housed.
Catherine Lord acknowledges receipt of royalties for the diagnostic instrument used in this research; all royalties related to this project were donated to charity.