Clinical Application and Validation of the Autism Detection in Early Childhood (ADEC) in Referred Children Aged 14-36 Months in a US Pediatric Hospital

Background: In the USA early diagnosis of Autism Spectrum Disorder (ASD), and hence access to timely intervention, is often hampered by long wait lists for diagnostic services.  Improved screening may reduce wait times by limiting the number of referrals of children who do not go on to receive an ASD diagnosis.  Children suspected of ASD typically present at clinics that also serve children with suspected language, intellectual, developmental, and/ or behavior problems.  Differentiating ASD from other diagnoses can, however, prove extremely challenging.  For example, a study conducted by the Cincinnati Children’s Hospital Medical Center found the Autism Diagnostic Observation Schedule (ADOS), considered to be the current “gold-standard” test for autism diagnosis, returned specificity of only 29%, although sensitivity was better at 76-99% (Murray et al, 2010).  Furthermore, poor specificity of the ADOS in this population could potentially contribute to over diagnosis.  The challenge is to develop effective instruments for use in “real life” clinical settings.  In the present study we evaluated the Autism Detection in Early Childhood (ADEC), a brief, play-based screening tool designed to identify at-risk children aged 12-36 months which can be implemented with minimal training.  The ADEC has yet to be evaluated in the USA.

Objectives: To evaluate the clinical utility of the ADEC as an ASD screening tool in children at risk of ASD who were referred to a hospital developmental clinic.

Methods: Children (n = 57) aged 14-36 months presenting at the developmental disabilities clinic of a large Midwestern pediatric hospital for diagnostic evaluation were assessed with the ADEC.  ADEC administrators trained to maintain coding reliability were blind to the outcome of the child's intake interview and all subsequent developmental assessments, and all clinical personnel were blind to ADEC scores.  Children suspected of ASD (n = 31) by the diagnostic clinician were assessed by an interdisciplinary team consisting of a speech language pathologist, psychologist, and developmental behavioral pediatrician or pediatric neurologist using DSM-5 diagnostic criteria.  The appropriate ADOS-2 module was administered by research-trained psychometricians.  For the remaining children, ASD had been ruled out by the initial clinician and they were typically referred for non-ASD assessments (n = 26).  Twenty-two children received a diagnosis of ASD.  Other diagnoses included developmental delay, coordination, language and behavior disorders.

Results: Sensitivity for the ADEC was 100% and specificity was 71%, both for the full sample and when limiting analysis to children referred for an interdisciplinary ASD assessment.  PPV ranged from 69-86% and NPV was 100%.  Inter-rater reliability for 30 dual-coded (video or live) administrations was ICC = .96.  Cohen’s kappa for individual items ranged from .56 - .90.  Internal consistency was Cronbach’s alpha = .78.

Conclusions: The ADEC was found to be psychometrically sound and compared favorably with the ADOS-2 in predicting final diagnosis using current diagnostic criteria.  Our preliminary findings suggest that the ADEC has the potential to contribute valuable information during initial screening, which may help to reduce wait lists for targeted diagnostic assessments and facilitate earlier access to appropriate services.