Abstract
PURPOSE: A multimeasure approach was developed to capitalize on the strengths of two screening measures: the Screening Tool for Autism in Toddlers and Young Children (STAT), an observational measure of social communication, and the Systematic Observation of Red Flags (SORF), a checklist including restricted and repetitive behavior (RRB) items. This approach offers a novel method of identifying autism in toddlers. METHOD: This was a retrospective study of data collected from a multidisciplinary diagnostic program for 24- to 36-month-olds with developmental delays. Raters with autism expertise but naïve to diagnoses applied the SORF to STAT videos. Psychometrics were derived for the SORF on STAT observations and a multiple-measure approach that used a Least Absolute Shrinkage and Selection Operator modeling framework to construct a STAT-SORF RRB Hybrid, retaining SORF RRB items based on individual predictive abilities. RESULTS: The SORF alone correctly classified 84% of the sample (84% sensitivity and 86% specificity). The STAT-SORF RRB Hybrid model, which retained four SORF RRB items, correctly classified 90% of a validation sample (95% sensitivity and 75% specificity). CONCLUSION: These findings highlight the potential utility of using multiple autism identification tools and regression-based scoring to establish presumptive eligibility and facilitate early access to autism interventions.