Phenotypic Heterogeneity of Autism Spectrum Disorders and Its Association with Early Identification In a US Population-Based Study

Background: Public health campaigns promote early detection and diagnosis of autism spectrum disorders (ASD), often by raising awareness of certain behavioral ‘signs’ that may be indicative of ASD. Several population-based studies examined autism prevalence and correlates of receiving a diagnosis, but did not consider differences in the presentation of core behavioral features.  Other studies suggested that the phenotypic or behavioral presentation of ASD may differ by sex, race, socioeconomic status, and co-occurring intellectual disability.

Objectives: To describe the frequency of behavioral impairments among children meeting surveillance criteria for ASD in a population-based US study and to determine whether the behavioral impairments of children meeting surveillance criteria for ASD differ depending on whether they have been previously identified as having ASD, after adjusting for race, gender and other factors.

Methods: The Autism and Developmental Disabilities Monitoring (ADDM) Network performed population-based ASD surveillance of 8-year-olds in areas of 11 US states in 2006, using documented information in medical and/or educational records. Case ascertainment was performed by expert clinician reviewers and based on DSM-IV-TR criteria for autism.  Of primary interest are 12 impairments (in domains of social interaction, communication, and repetitive/restricted behaviors and interests) which comprise the DSM-IV-TR diagnostic criteria for Autism and ASD.  Children classified as ASD cases by ADDM were not all previously identified as having an ASD.  Of the 2,757 children that met the ADDM case definition, 75.4% had previously received a clinical diagnosis and/or a special education classification of autism/ASD. Logistic regression was used to determine whether the probability of previous identification was associated with specific impairments, and random forests were used as a confirmatory analysis to evaluate the potential for complex interactions. 

Results: Among 2,757 children meeting ADDM criteria for ASD, the frequency of impairments corresponding to the 12 diagnostic criteria ranged widely from 49.1% (DSM-1c – joint attention) to 90.4% (DSM-1d – emotional reciprocity). 351 (12.7%) of the children were described as having all 12 impairments.  Among the 2,136 children meeting ADDM criteria for Autistic Disorder, inflexible routines or rituals (DSM-3b, OR=3.1 (95%CI: 2.2-4.3)), emotional reciprocity (DSM-1d, OR= 2.8 (95%CI: 1.7-4.5)), nonverbal communication (DSM-1a, OR=2.7 (95%CI: 1.9-3.8)), and stereotyped motor behaviors (DSM-3c, OR=2.1, (95%CI: 1.6-2.7)), were most strongly associated with previous ASD identification in the adjusted regression model. Gender, race, and intellectual disability were not significantly associated with the likelihood of previous identification in the adjusted model. The random forest analysis produced similar findings, and ranked repetitive motor behaviors (DSM-3c) and inflexible routines or rituals (DSM-3b) as the most “important” predictors of previous ASD identification.

Conclusions: Findings from this study have implications for efforts to improve early identification of ASD, suggesting that impairments related to social interaction, repetitive behaviors and restricted interests appear more likely to trigger a diagnosis or special education classification of ASD than impairments related to verbal communication.   Consideration of possible combinations of impairments that meet diagnostic criteria for ASD and the relative frequency of each combination in the population provides insight into disparities in the prevalence and identification of ASD.