Differential item functioning in the autism behavior checklist in children with autism spectrum disorder based on a machine learning approach

Abstract

Read online

AimOur study utilized the Rasch analysis to examine the psychometric properties of the Autism Behavior Checklist (ABC) in children with autism spectrum disorder (ASD).MethodsA total of 3,319 children (44.77 ± 23.52 months) were included. The Rasch model (RM) was utilized to test the reliability and validity of the ABC. The GPCMlasso model was used to test the differential item functioning (DIF).ResultThe response pattern of this sample showed acceptable fitness to the RM. The analysis supported the unidimensionality assumption of the ABC. Disordered category functions and DIF were found in all items in the ABC. The participants responded to the ABC items differently depending not only on autistic traits but also on age groups, gender, and symptom classifications.ConclusionThe Rasch analysis produces reliable evidence to support that the ABC can precisely depict clinical ASD symptoms. Differences in population characteristics may cause unnecessary assessment bias and lead to overestimated or underestimated symptom severity. Hence, special consideration for population characteristics is needed in making an ASD diagnosis.

Keywords

• ABC
• Rasch model
• differential item functioning
• ASD
• machine learning