Identifying the Best Discriminating Items for Three Autism Screening Methods in a High-Likelihood Sibling Population.

The questions used in caregiver-completed autism screening tools vary in how well they differentiate between children with autism and those without, depending on the age group screened, the target population, and language or cultural contexts. This study examined the best discriminating items for the Modified Checklist for Autism in Toddlers, Revised with Follow-up (M-CHAT-R/F), Social Attention and Communication Surveillance, Revised (SACS-R), and ASDetect in a high-likelihood younger sibling population in a multicultural setting. Alternative scoring methods to improve the utility of these screening tools were also examined. Item response data were extracted from a published study on 141 siblings who underwent autism screening at 12, 18 and 30 months old, followed by a diagnostic assessment at 36 to 48 months old. 32 (22.7%) were diagnosed with autism. The best discriminating items from the M-CHAT-R/F were 'follows a point', 'pretend play', 'protoimperative pointing', 'protodeclarative pointing', 'shows objects', 'responds to name', 'follows gaze' and 'understands instructions'. An alternative scoring method ('Sing-CHAT') was established using these items and performed better than the full M-CHAT-R/F when using failing ≥1 item as a cutoff. In contrast, almost all items on the SACS-R and ASDetect were found to be significantly discriminating at 18 and 30 months old. For these, however, using all items instead of just key items, and using a lower cut-off threshold, resulted in improvements in psychometric properties. Best discriminating items in screening tools vary by age, population and culture, and alternative scoring methods could potentially improve autism detection.