Parent-reported and clinician-observed autism spectrum disorder (ASD) symptoms in children with attention deficit/hyperactivity disorder (ADHD): implications for practice under DSM-5

Data from the current study were extracted from an existing database of children whose parents had provided written informed consent for their child’s clinical information to be included in a research database which was approved by Institutional Review Boards (IRB) at three different ASD-specialty clinics (sites): the University of Chicago Developmental Disorders Clinic (UCDDC; 45 % of the sample), the University of Michigan Autism and Communication Disorders Center (UMACC; 47 % of the sample), or the New York Presbyterian Center for Autism and the Developing Brain (CADB) at Weill Cornell Medicine (8 % of the sample). Children were only eligible for the current study if they had been referred for a clinical ASD diagnostic assessment (i.e., not seen as part of a specific research study) and had received an overall best estimate clinical diagnosis (BEC) of ASD or ADHD following the comprehensive assessment (see below).

Research has shown that distinguishing between ASD and ADHD in children who are verbally and intellectually able is a particular challenge [18, 36, 37]. Therefore, in this initial exploration, we limited our sample to children who were between the ages of 4 and 18 years old, with IQs in the borderline range or above (full scale IQs ≥70), and received module 3 or 4 of the ADOS [38], indicating verbal fluency. Forty-eight participants (38 males) who received clinical assessments and met the above criteria received a BEC of ADHD (no diagnosis of ASD). Children with a BEC of ASD were chosen as a comparison group, based on the same criteria outlined above. Due to the DSM-IV-TR prohibition against diagnosis of comorbid ADHD in individuals with ASD, ADHD was not systematically ruled in/out in children with ASD prior to 2013 (when DSM-5 criteria were implemented); however, children who were given a comorbid diagnosis of ASD and ADHD (post-DSM-5 implementation; n = 66) were not eligible for inclusion in this study. The final sample of children with a BEC of ASD included 164 children (138 males). Information about current psychotropic medication treatment was only available for 66 % (n = 139) of the sample, as this information was not systematically gathered for children seen at UCDDC. There was no statistically significant difference between the ADHD and ASD groups with regard to the proportion of children receiving psychotropic medication (χ ²(1, 139) = 3.36, p = 0.07). The children with ADHD did not differ from children with ASD in verbal IQ (VIQ), non-verbal IQ (NVIQ), age, gender, or ethnicity. However, given the large age (4 to 18 years old) and NVIQ (72–153) ranges included in the study, these variables were included as covariates in secondary analyses of dimensional data despite no significant overall group differences in means. In addition, because significantly more children with ADHD were seen in Michigan than New York and Chicago, site was included as a covariate in these secondary analyses (see Table 1).

For the majority of participants, a standard hierarchy was followed to select appropriate cognitive tests: the Differential Ability Scales (DAS) [43] was chosen first, followed by an age-appropriate Wechsler test (Wechsler Adult Intelligence Scale (WAIS), Wechsler Abbreviated Scale of Intelligence (WASI), Wechsler Intelligence Scale for Children (WISC), Wechsler Preschool and Primary Scale of Intelligence (WPPSI)) [44‐47]. These tests provided verbal IQ (VIQ) and non-verbal IQ (NVIQ) for 93 % and 85 % of the ADHD and ASD groups, respectively. These measures use chronological age standardization with a mean of 100 and a standard deviation of 15 and have adequate convergent validity [45]. A number of other tests were used to estimate IQ for the remaining participants (7 % of the ADHD sample and 15 % of the ASD sample). Specifically, approximations of VIQ were gathered from the Clinical Evaluation of Language Fundamentals (CELF) [48], Peabody Picture Vocabulary Test (PPVT) [49], Vineland Adaptive Behavior Scales (VABS) [50], and NVIQ from the Raven’s Progressive Matrices [51]. There was no difference between the diagnostic groups with regards to which cognitive test was used.

An overall best estimate clinical diagnosis (BEC) [52] was determined based on all available information, including questionnaires, information from teachers, results of the ADOS and ADI-R, and the diagnostic impression of a clinical team, which included a psychologist and/or psychiatrist.

First, we examined the proportion of participants who met standard ADI-R and ADOS algorithm cut-offs in the ADHD group. We expanded this to also explore less strict ADI-R cut-offs for ASD as defined by Lainhart et al. (Collaborative Program of Excellence in Autism (CPEA) ADI-R Criteria) [53] since original ADI-R cut-offs were intended to capture autistic disorder, with less sensitivity to broader ASD presentations now captured under DSM-5 criteria. Second, using analysis of variance (ANOVA), the groups were compared on ADOS CSS (Overall, SA, and RRB), ADOS factors (basic SOC, interaction quality, and RRBs), ADI-R algorithm domains (social, communication, RRB-ADI), and ADI-R current symptom domains (SC, RSM, and IS). Secondary analyses of covariance (ANCOVA) co-varying for age, NVIQ, and site were conducted. The ADI-R age of onset algorithm domain is intended for categorical use; therefore, a chi-square analysis was conducted comparing groups on the proportion of children who met the domain algorithm cut-off (≥1). Third, since current diagnostic practice (DSM-5) focuses primarily on the presence/absence of specific symptoms and since ADI-R and ADOS items represent ordinal rather than interval data, we compared the two groups with regard to the proportion of symptoms endorsed (1, 2, or 3) versus not endorsed (0). Specifically, χ ² analyses were conducted to compare proportions of endorsed versus non-endorsed items that have been mapped to each symptom domain [30]. Fourth, using methodology previously employed for the purpose of identifying items for diagnostic algorithms and other discrimination purposes [39, 54, 55], we identified specific social communication items that discriminated best between the groups. This methodology aims to identify items that show both a shifted distribution by diagnostic group and also items that are most likely to differentiate between ASD and non-ASD. An item was determined to discriminate “adequately” between groups if it was endorsed (code of 1–3) in >66 % of the ASD group and <33 % of the ADHD group. This criterion was not applied to RRB items because endorsement was not expected in a large proportion (>66 %) of individuals with ASD [39], especially given that participants in this study were primarily school-aged and adolescent children with average or above IQ. To account for multiple comparisons, a significance threshold was set at α = 0.01 across analyses.

As Table 2 shows, 21 % of the ADHD sample and 85 % of the ASD sample met ADOS ASD algorithm cut-offs (cut-off for ASD ≥7 for module 3 and ≥8 for module 4) [39, 40]. On the ADI-R, 30 % of the ADHD sample and 67 % of the ASD sample met algorithm criteria across all four domains resulting in an ADI-R classification of autism. When using CPEA criteria, which are less strict than standard ADI-R autism cut-offs [53], 67 % of the children with ADHD and 84 % of the ASD sample received an ADI-R classification of ASD. When the measures were employed in combination, most children (60–75 %) with ASD met criteria for ASD on both the ADOS and the ADI-R, while few children with ADHD met criteria for ASD on both the ADOS and the ADI-R (11 % when using traditional ADI-R cut-offs and 15 % when using CPEA ADI-R cut-offs).

Our relatively small sample of children with ADHD represents a particular subset of children with ADHD: children whose parents had sufficient concern about ASD to bring them to an ASD-specialty clinic to determine whether a diagnosis of ASD was warranted. As such, these children are not representative of all children with ADHD. In addition, our group of children with ASD was deliberately selected from a much larger sample to be equivalent to the ADHD group in language level, age, and IQ and is therefore not representative of the whole ASD population. Future work should extend these results to samples that include individuals with lower language and intellectual abilities. In addition, this sample included relatively few females. Although this is consistent with the gender discrepancy seen in ASD samples, results of this study may not be characteristic of females with ASD and/or ADHD. Future studies should explore the unique symptom presentations in females with ASD and/or ADHD. Last, we could not rule out ADHD in our sample of children with ASD, although children with formal comorbid diagnoses of ADHD were excluded.