2 - (IOP18) Autism Assessment from Home: A Psychometric Evaluation of a Brief, Accessible, Remote Autism Observation for Children

Tele-based methods of autism assessment are vital to overcoming longstanding barriers (e.g., geography, cost, limited providers) to obtaining an autism diagnosis, which is often necessary to access and inform evidence-based interventions. Observational measures are central in face-to-face autism diagnostic evaluations, but few validated observations exist for remotely assessing autism across childhood, particularly for older children and adolescents. Sanchez and Constantino (2020) validated a brief, face-to-face, clinician-facilitated observation coded according to the first eight items of the Childhood Autism Rating Scale, Second Edition (CARS-2). During the recent pandemic, this measure was adapted as a remote observation conducted from families’ homes, but has yet to be validated in this remote format. The current psychometric study validated this new remote observation in relation to the “gold-standard,” in-person Autism Diagnostic Observation Schedule, Second Edition (ADOS-2) and evaluated its utility in predicting an ADOS-2 classification and research diagnosis. 30 children ages 1.97-16.66 years (M = 7.35, SD = 4.00) completed the brief remote observation and in-person ADOS-2. Convergent validity between total scores on the remote observation and ADOS-2 was assessed via Spearman’s rank correlation. Exploratory analyses evaluated whether correlations differed by gender or by which CARS-2 form (Standard or High Functioning) was used to code the remote observation. Receiver operating characteristic (ROC) curves were generated to determine optimal remote observation cutoff scores to maximize sensitivity and specificity in predicting ADOS-2 classification and research diagnosis of autism. At optimal cutoff scores, agreement between diagnostic classification (i.e., autism or not) on the remote observation and ADOS-2 was calculated via percent agreement, sensitivity, and specificity; analyses were then repeated to determine agreement between the remote observation and a research diagnosis for each child. Total scores on this novel remote observation and the ADOS-2 were highly correlated (ρ = .644, p < .001). Exploratory analyses demonstrated that high correlations maintain across males (ρ = .729, p < .001) and females (ρ = .801, p = .002); however, correlations with ADOS-2 were lower when the High Functioning CARS-2 form was used to code the remote observation (ρ = .328, p = .214) than when the Standard Form was used (ρ = .867, p < .001). When evaluating diagnostic agreement, ROC analysis confirmed that 16 was the optimal cutoff score for predicting a diagnosis (88.9% sensitivity, 80% specificity), whereas 15 was optimal for predicting ADOS-2 classification (83.3% sensitivity, 70.6% specificity). At optimal cutoff scores, the remote observation predicted the correct diagnosis in 82.8% of cases and same classification as the ADOS-2 in 75.9% of cases. Validation of this instrument provides support for an accessible and efficient remote autism observation and supports the potential utility of optimized cutoffs in predicting an autism diagnosis. Adding remote observation to extant diagnostic toolkits may help mitigate long-standing barriers to autism diagnosis and service recommendations and access.