Using Standardized Diagnostic Instruments to Classify Children with an Autism Spectrum Disorder in the Study to Explore Early Development

Background:  The Study to Explore Early Development (SEED) is a case-control study funded by the Centers for Disease Control and Prevention (CDC) to examine autism spectrum disorder (ASD) phenotypes and etiologies. One of SEED’s greatest strengths is use of gold standard diagnostic instruments – the Autism Diagnostic Interview Revised (ADI-R) and Autism Diagnostic Observation Schedule (ADOS) – as well as a combination of other factors, such as mental age of the child, to classify children with ASD.

Objectives:  The objectives of this analysis were to (1) describe how children were classified in SEED and (2) examine the psychometric properties of SEED ASD criteria, ADOS alone, ADI-R alone, and ADOS plus ADI-R without consideration of other factors (i.e., ADOS/ADI-R standard criteria).

Methods:  Children 2-5 years old were ascertained through birth certificate records and multiple sources that serve children with developmental problems. All children were screened for ASD upon enrollment. Children without a prior ASD diagnosis who had little risk noted on the ASD screen received a cognitive assessment only. Children with a prior ASD diagnosis or who had risk noted on the ASD screen received the ADI-R, ADOS, cognitive assessment, and adaptive assessment. After the developmental evaluation, all children were classified into one of the following groups: ASD, developmental delay or disorder (DD), population comparison (POP, identified from birth certificate records), or Possible ASD. Children classified as DD, POP, or Possible ASD were further divided into one of numerous subgroups. SEED ASD criteria took into account numerous factors, including ascertainment source, results of the ADI-R and ADOS, rules for resolving discordance between the ADI-R and ADOS, mental age of the child, and – in some instances – clinical judgment.

Results:   A total of 2,732 children were enrolled in SEED and completed a clinic visit. SEED final classifications for these children were ASD (N=703), DD (N=999), POP (N=906), and Possible ASD (N=124).  There were 1,063 children who received the ADI-R, ADOS, and cognitive assessment and had clinical judgment noted. Of these, 921 had a mental age of at least 24 months and 142 had a mental age less than 24 months. ADI-R and ADOS results were discrepant in 22.8% of this sample.      

The ADOS yielded the highest sensitivity but the lowest specificity than any other classification scheme. The sensitivity and specificity of other classification schemes were relatively comparable. ADOS/ADI-R standard criteria missed 62 more children with certain ASD (defined by clinical judgment) than SEED ASD criteria.  Likewise, SEED ASD criteria classified 27 more children with uncertain ASD (defined by clinical judgment) than ADOS plus ADI-R standard criteria.        

Conclusions:  These findings support the utility of SEED ASD criteria in population-based research. Three major advantages of SEED ASD criteria are: (1) a method for resolving discordance between the ADOS and ADI-R, (2) a method for classifying children with a mental age less than 24 months, and (3) detailed classifications and sub-classifications that allow phenotypic exploration. Future research should explore whether SEED classifications identify children with different phenotypes that can be used as important outcomes in SEED.