Structural and Convergent Validity of the Autism Impact Measure (AIM)

Background: Psychometrically sound outcome tools are needed to evaluate the effects of autism treatments in both clinical and research settings. The Autism Impact Measure (AIM) was designed to enable efficient measurement of incremental changes in both frequency and impact of core ASD symptoms. The initial psychometric study of the AIM indicated strong to moderate test-retest reliability and inter-rater reliability (Kanne et al., 2014). However, a larger sample was needed to further establish structural validity and scoring guidance. An extensive battery of domain-specific measures of communication, social functioning, and repetitive behaviors was also necessary to thoroughly examine convergent validity.

Objectives: The objective of the current study was to examine the structural and convergent validity of the AIM in a large sample of children with ASD.

Methods: Participants included 890 children and adolescents with ASD. The study included a primary sample (n = 450, 82% male) used in both convergent and structural validity analyses and a secondary sample (n = 440, 84% male) used only in structural validity analyses. The following convergent validity measures were administered (Sample 1): Repetitive Behavior Scale – Revised (RBS-R), Aberrant Behavior Checklist (ABC), Social Responsiveness Scale (SRS-2), Autism Diagnostic Observation Schedule – 2nd Edition (ADOS-2), Ohio Autism Clinical Global Impression Severity Scale (OACIS-S), Vineland Adaptive Behavior Scales (VABS-II), MacArthur Communicative Development Inventory (CDI), Pediatric Quality of Life Inventory (PedsQL).

Results:

A series of exploratory (EFA) and confirmatory (CFA) factor analyses was conducted to establish construct and structural validity of the AIM. AIM item scores were calculated by adding the Frequency and Impact scores for each item, yielding a single summative item metric representing overall severity. EFA was conducted first, followed by cross-validation using CFA with a hold-out sample of subjects. To accomplish this, the total sample was randomly assigned to one of two groups (each consisting of half the sample). EFA was conducted on the first randomly established sample, while the CFA was conducted on the second randomly established hold-out sample. The results yielded a final model with five theoretically and empirically meaningful subdomains: Repetitive Behavior, Atypical Behavior, Communication, Social Reciprocity, and Peer Interaction. The final model showed very good fit both overall and for each of the five factors (RMSEA <.06, CFI and TLI >.95), indicating excellent structural validity. Convergent validity analyses indicated that AIM subdomain scores were significantly correlated with measures of similar constructs across all five domains.

Conclusions: The current study examined the construct and structural validity of the AIM in a sample of 890 children with ASD. The study extended prior research on the AIM by including a larger sample and using a more comprehensive measurement battery to test convergent validity. The results indicate that the psychometric properties of the AIM are strong, including excellent structural validity and solid convergent validity. The results suggest that the AIM may be useful for efficient assessment of short-term improvements in core ASD symptoms across five empirically-derived and clinically meaningful symptom domains.