Use of the Systematic Observation of Red Flags (SORF) for Autism Spectrum Disorder in a Naturalistic Home Setting

Background: While routine universal screening for autism spectrum disorder (ASD) is critical to allow timely initiation of intervention, some have expressed skepticism due to the lack of effective screening tools available (Al-Qabandi et al., 2007; Campos-Outcalt, 2011). Improved screeners are needed that not only effectively identify children at risk, but also offer increased feasibility for widespread use across community settings. The Systematic Observation of Red Flags (SORF) is an observational measure that has been shown to effectively detect risk for ASD when used with the Communication and Behavior Scales (CSBS) Behavior Sample (Dow et al., 2016). Because many children with developmental concerns do not receive a clinic evaluation, adapting this tool for use in the home could improve community-viable screening options accessible to more families, decreasing the age of diagnosis.

Objectives: To examine use of the SORF in the home for 16-24 month olds, including (1) diagnostic group differences and item-level performance of SORF items, and (2) sensitivity, specificity, and appropriate cutoff scores. 

Methods: Preliminary analyses were conducted on the current sample of 32 participants (10 with ASD, 15 developmental delayed, 7 typically developing) recruited by the FIRST WORDS® Project at Florida State University. A sample size of 194 is estimated at the time of presentation, sufficient to achieve a medium effect size and power of 0.80. Children will be included if they have (1) a completed SORF based on behavior during a video-recorded home observation between 16-24 months, and (2) a concurrent diagnostic evaluation to confirm or rule out ASD.

Receiver operating characteristic (ROC) curves were used to evaluate individual items and summary scores, including the Total score of all summed items, Number of Red Flags (RF) score of the number of items indicating clear symptom presence, Social Communication (SC) score of SC items summed, and Restricted and Repetitive Behavior (RRB) score of RRB items summed. Area under the curve (AUC), sensitivity, specificity, and cutoffs were examined. Further analyses utilizing analysis of variance (ANOVA) will also be completed for the full sample to examine diagnostic group differences in items and summary scores, and a Composite score comprised of the best performing items will be created for optimal performance.

Results: Preliminary results suggest excellent discrimination between ASD and nonspectrum groups for the Total (AUC=.97), RF (AUC=.93), SC (AUC=.93), and RRB (AUC=.91) scores. High sensitivity and specificity was found for the Total, RF, and SC scores (sensitivity=.90, specificity≥.96) with appropriate cutoffs. The best performing items were poor directed eye gaze (AUC=.89), limited conventional gestures (AUC=.86), greater interest in objects than people (AUC=.86), repetitive speech/intonation (AUC=.83), limited coordinated nonverbal communication (AUC=.81), and limited consonant sounds (AUC=.80).

Conclusions: Preliminary results are promising indicating that the SORF may provide an effective observational screening tool in the home context, though the full sample is needed to support these findings. The Total and RF scores provided good discrimination, sensitivity, and specificity, and the SC domain may offer an abbreviated measure when used independently with similar effectiveness.