Separate Scoring Algorithms Optimize the Screening Properties of the Screening Tool for Autism in Toddlers for Different Screening Priorities

Abstract

Detecting autism in young children allows for timely access to specialized early intervention services. The Screening Tool for Autism in Toddlers (STAT) is a validated stage-2 Autism Spectrum Disorders (ASD) screening measure that takes 20 minutes to administer and comprises 12 play-based items that are scored according to specific criteria. An expanded version (STAT-E) includes the examiner’s subjective ratings of children’s social engagement and atypical behaviors. This study examines the screening properties of STAT-E using the original STAT scoring algorithm and the extent to which an algorithm that includes the subjective ratings of social engagement and atypical behaviors improves the screening properties of the STAT-E relative to the original STAT scoring algorithm. Two-hundred and thirty-eight (238) families of children between 24 and 35 months old participated. The STAT-E was administered by assessors with limited experience who were trained using a scalable web-based platform and children received a comprehensive evaluation from a separate team of ASD research or clinical experts who were blind to the STAT-E results. Logistic regression, ROC curves, and classification matrices and metrics (Youden’s J and F1 score) were used to determine the screening properties of the STAT-E using the original STAT scoring algorithm and the extent to which an algorithm that included the subjective ratings of social engagement and atypical behaviors improved the screening properties of the STAT-E relative to the original STAT scoring algorithm. The concurrent validity of the STAT-E using the original STAT scoring algorithm in this sample was fair (sensitivity = .67, specificity = .66). Inclusion of the examiner ratings of social engagement and atypical behaviors on the STAT-E improved positive risk classification appreciably (F1 score = .80-.85 versus .74), while the specificity declined (specificity = .62). Results suggest that the STAT-E using the original STAT scoring algorithm optimizes specificity, while the STAT-E scoring algorithm with two new ratings optimizes the positive risk classification. Using multiple scoring algorithms on the STAT may provide improved scoring accuracy for diverse contexts and children. A fast and scalable web-based tutorial may be a pathway for increasing the number of community providers who can administer the STAT and contribute toward increased rates of autism screening.