Assessing the psychometric properties of the Autism Diagnostic Observation Schedule – Generic (ADOS-G) in a clinical setting in the Chinese mainland

The awareness of autism spectrum condition (ASC) and the estimated prevalence rate are lower in the context of the Chinese mainland, compared to western countries. The Chinese adaptation of the Autism Diagnostic Observation Schedule – Generic (ADOS-G) has been widely used in the diagnostic process of ASC for many years despite its psychometric properties not having been established in a large clinical sample. Thus, the purpose of this study was to evaluate the metrics of the ADOS-G in a well characterized sample from a renowned child developmental-behavioural centre that provides clinical services to a nationwide reach of people in China. The present study analysed a large retrospective record review of individuals who visited the centre’s outpatient clinic between January of 2018 and August of 2021. Participants in this study were divided into ASC (n = 2330) and non-ASC (n = 473) groups according to their clinical diagnosis. All participants completed the ADOS-G Chinese version, with 1333 administered Module 1, 867 Module 2 and 603 Module 3. Psychometric properties of the ADOS-G Chinese version were evaluated in terms of item characteristics, internal consistency, concurrent validity, and convergent validity. The ADOS-G Chinese version demonstrated different levels of internal consistency reliability for different domains, with low Cronbach alpha values for Communication domains (0.40-0.60), but higher for the Communication and Social Interaction combined domains (above 0.75). Evaluation of validity yielded good sensitivity (> 90%) and positive predictive values (> 0.85), but low specificity (0.15-0.60) and negative predictive values (around 0.60). Agreement between ADOS-G diagnosis and clinician’s diagnosis was fair (Cohen’s kappa ranging from 0.20 to 0.35 for different modules), but low between the ADOS-G and ADI-R (< 0.15). Almost all items of the ADOS-G differentiated between ASC and non-ASC. Sample sizes were unevenly distributed between ADOS-G Modules 1 to 3, with Module 1 having the largest sample and none for Module 4. Retest reliability could not be assessed with the retrospective chart data. This study is the first to examine the ADOS-G Chinese adapted version in a large clinical sample in the Chinese mainland. The ADOS-G contributes valuable information to support the clinical diagnosis of ASC, though additional research into its psychometric properties and cultural adaptation in the Chinese mainland context may be necessary. This study highlights the importance of cultural validation and the considerations required when interpreting ADOS results in non-Western settings.