Exploration of Diagnostic Validity of Autism Diagnostic Interview-Revised (ADI-R) in Korean Population.

Background: Autism Diagnostic Interview-Revised (ADI-R) is a semi-structured investigator-based interview, which has been characterized as one of the gold-standards in the assessment of children and adults with autism spectrum disorder (ASD). Research on the use of the ADI-R and its applicability for accurate identification of individuals with ASD among non-English speaking populations is significantly limited.

Objectives: The objectives of the current study is to examine the validity of the Korean version of ADI-R to determine its efficacy in identifying individuals with ASD aged 24 months to 34 years.

Methods: We used pooled data from an ASD genetic study, clinical trial of social skills training, research on the development of early screening instruments, and clinical record of Seoul National University Bundang Hospital, from 2008 to 2017. The subjects were composed of individuals with ASD, their unaffected siblings, children referred as developmental delay, and typically developing children volunteered in community. The ADI-R were administered and scored by professionals having research reliability for the instrument. ADOS (Autism Diagnostic Observation Schedule) was also completed for the subjects referred for diagnosis. Other measurements for ASD symptoms were also administrated, including Social Responsiveness Scale (SRS), Social Communication Questionnaire (SCQ) and Korean version of the Childhood Autism Rating Scale (K-CARS). We applied the cut-off scores for K-CARS (cutoff=24) and SCQ (cutoff=12 for 48 months and over, 10 for be­low 47 months) based on the validity studies performed in Korea (Kwon et al., 2017; Kim et al., 2015). Clinical best estimate diagnosis of ASD was confirmed by two board-certified psychiatrists based on DSM-IV-TR and DSM-5^TM criteria. We classified the clinical diagnosis into ASD (including pervasive developmental disorder and Asperger’s syndrome) versus non-ASD in the validity analyses. Statistical analyses were used t-test and Cohen’s kappa for comparison of two groups.

Results: Total 1,073 subjects (age 93.48 ± 62.42 months, male = 760, female = 313)participated. The sensitivity and specificity of ADI-R were 99.39% and 84.65%, respectively. Positive Predictive Value was 92.50 (between 84.21 to 95.31) and Negative Predictive Value was 98.64 (between 97.96 to 100) in all age group (below 48 months and 48 months and over) and language levels (with or without verbal language). In Individual items within diagnostic algorithms of ADI-R showed significant differences in scores for all algorithm items between ASD and non-ASD group in all age and language subgroups (p<0.05). The agreement between ADI-R and other measurements in diagnosing ASD was good to excellent level (Cohen's kappa of K-CARS, SCQ and SRS with ADI-R = 0.78, 0.66, 0.65, respectively).

Conclusions: This study examines the diagnostic validity of the Korean version of the ADI-R. The Korean version of ADI-R showed high levels of sensitivity, specificity, positive predictive value, and negative predictive value in all age groups and verbal language levels. This study suggests that ADI-R might be valuable diagnostic instrument for individuals with ASD, across countries with different languages, cultural backgrounds and levels of awareness for ASD.