STAT Behavioral Domains As Predictors of ASD Severity and Cognitive Outcomes

Background: The Screening Tool for Autism in Toddlers (STAT) is a brief, interactive, validated screening assessment for autism comprised of 12 activity-based items that provide 4 behavioral domain scores—play, imitation, requesting, and directing attention—as well as a total score indicating autism risk. The item-domain structure of the STAT has never been validated via confirmatory factor analysis (CFA). The STAT domains represent discrete, developmentally appropriate social-cognitive skills, and each domain may differentially predict both children’s overall autism severity and cognitive ability at 2-3 years. Each domain of the STAT has been found to augment the STAT’s overall high predictive validity with autism diagnosis. However, it is not known how differentially predictive the STAT domains are of autism symptomatology and cognitive ability in a large sample of children recruited from the general population with elevated autism concern.

Objectives:  To (1) perform a CFA to confirm that items of the STAT theoretically proposed to form domains do indeed form the 4 STAT domains, and (2) perform a latent path model to assess the predictive validity of the STAT domains to gold-standard autism diagnostic instruments, the ADOS and Mullen.

Methods: Participants (n=380) were 24-39 months of age and recruited from eight sites funded as part of a National Children’s Study formative research project. 79% of participants had prior autism concerns; 21% had concerns about developmental delay. Participants received diagnostic evaluations from a qualified clinician, including the ADOS and Mullen, in the community or as part of the study. The STAT was administered as part of a larger battery of autism screening tools. A CFA was conducted to assess how well the items comprising the 4 domains of the STAT—play, imitation, requesting, and directing attention—mapped on to each domain. A latent path model assessed how participants’ latent STAT domain scores related to their ADOS severity score and verbal and nonverbal Mullen scores. State-of-the-art model-fitting procedures were employed for evaluation.

Results:  The theorized 4 latent factors for the play, requesting, directing attention, and imitation STAT domains fit the data well (e.g., CFI=.98; RMSEA=.03), supporting the item allocation into each domain as originally designed. The 4 resulting latent STAT domains were highly correlated. The final latent path model fit the data well (e.g., CFI=.97; RMSEA=.03). Three of the 4 STAT domain scores (imitation, requesting, and directing attention) were found to significantly predict ADOS severity. STAT play and imitation domains were found to significantly predict Mullen nonverbal subscale score. The STAT directing attention domain was found to be the best predictor of Mullen verbal subscale score (Figure 1).

Conclusions: The STAT is both a theoretically and empirically coherent measure of 4 social-communicative behavioral domains. Three of these domains could be sufficient to serve as a predictor of autism severity. The STAT domains are also useful individually; they are brief measures of developmental ability and can provide specific clinical information on social-cognitive challenges for young children. Implications of the different STAT domains as indicators of general autism symptomatology and cognitive ability will be discussed.