Agreement Between Multiple Autism Diagnostic Instruments and Clinical Judgment Among Taiwanese Children in the Community

Background: Several diagnostic measures for Autism Spectrum Disorder (ASD) are currently available and widely used in Western countries.  Some of these tools are based on information obtained through caregiver/parent interview (e.g., Autism Diagnostic Interview-Revised (ADI-R)), while others are direct observations of the child (e.g., Autism Diagnostic Observation Schedule (ADOS)).  While these tools are well-validated for Western populations, their utility and validity are unknown for populations in other international settings.  Understanding the agreement between diagnoses derived from these instruments in non-Western populations may provide guidance for diagnosing ASD. 

Objectives: This study aims to examine the agreement among multiple diagnostic instruments for ASD (ADOS, ADI-R, and clinical judgment based on DSM-IV criteria) in a community-based study in Taiwan.  A secondary purpose is to examine whether combining parent-reported Social Communication Questionnaire (SCQ) and ADOS results improves diagnostic agreement with DSM-IV-based clinical judgment for ASD.

Methods: A community-based epidemiologic study of autism in school children aged 6-8 involving a multi-stage case identification design was conducted in PingTung, Taiwan.  The study first screened first and second graders using the SCQ, then invited children who have SCQ>=7 for a clinical visit.  At the clinical visit, 27 boys and 19 girls and their primary caregivers completed the ADOS, ADI-R, and were evaluated by local child psychiatrists to determine their DSM-IV-based diagnosis.  All study instruments were translated and adapted from English into Chinese Mandarin.  DSM-IV-based clinical judgment served as the referent standard.  In addition to recommended cut-offs from the ADOS and ADI-R manuals, we defined ADOS-subclinical as having one domain score one point below the ASD cut-off for that domain; whereas ADI-R-subclinical refers to when one domain score does not meet the autism cut-off.  Consequently, seven diagnostic criteria were defined to compare against the referent standard: 1) SCQ>=11 and ADOS-ASD, 2) SCQ>=11 and ADOS-ASD or subclinical, 3) SCQ>=15 and ADOS-ASD, 4) SCQ>=15 and ADOS-ASD or subclinical, 5) ADOS-ASD, 6) ADOS-ASD or subclinical, and 7) ADI-R autism or subclinical. 

Results: Based on the child psychiatrists’ diagnoses, there were 7 children with ASD, 24 with other developmental disabilities (DD), and 15 unaffected with ASD or other DD.  The criteria of SCQ>=15 and ADOS-ASD yielded the highest agreement with the gold standard (84%), with 4.3% false positive and 10.9% false negative, followed by SCQ>=15 and ADOS-ASD/subclinical (80% agreement, 13.3% false positive, and 6.7% false negative).  ADOS-ASD/subclinical bore the lowest agreement, 60.9%, due to a high percentage of false positives (34.8%). 

Conclusions: These findings support the inclusion of both parent-reported information (e.g. SCQ) and direct observation of the child (e.g. ADOS) in determining ASD diagnoses.  The ADOS intends to provide an objective observation of the child, however, when used alone with its standard criteria, it is prone to false positives.  The ADI-R is time consuming and sensitive to caregivers’ cultural backgrounds and subjective interpretation.  In settings where diagnosticians and expert clinicians are not available, or where cultural backgrounds are substantially different from where the tool was developed, combining the ADOS with parent-reported SCQ can be considered as a way to diagnose ASD.