Capturing the Moment: Using the Autism Observational Scale for Infants (AOSI) to Assess the Developing Phenotype in a Prospective Sample of High Risk Children for ASD

Background:  Following the development of high-risk (HR) sibling cohorts allows for the identification of earlier differences in those who later develop ASD, and can serve as an indicator of early life development. While certain markers have been documented, the cascade of symptom development with environmental influences is not reported and is important for early intervention.  The Autism Observational Scale for Infants (AOSI) is an experimenter-led assessment that measures early behavioral markers of ASD between 6 -18 months. The measure captures atypical social communication, non-social behaviors, temperament and repetitive behavior markers, yielding a Total Score (TS).  

Objectives:  To assess the predictive accuracy of AOSI scores for ASD classification at 36 months in a HR infant cohort and to examine accuracy across variable AOSI cutoff scores. 

Methods: Participants represented a subset of HR children in the Early Autism Risk Longitudinal Investigation (EARLI).  Infants (n=132) were given the AOSI at 6 and/or 12 months of age and an ASD diagnostic evaluation at 36 months. Diagnostic outcomes were determined by two clinicians using clinical best estimate, utilizing information from the Mullen Scales of Early Learning and Autism Diagnostic Observation Schedule (ADOS-2).  ANOVAs were conducted to assess the difference in AOSI TS at 6 and 12 months, receiver operatic characteristic (ROC) analyses examined how well AOSI TS predicted diagnostic outcome when comparing ASD/NON-ASD samples combined, and then comparing NTD/TD separately.  Spearman Correlations compared 36 month ADOS Comparison Scores to the AOSI TS at both 6 and 12 months. 

  Results:  At 36 months, 30 children (23%; 26 males) met DSM-5 ASD criteria; 38 had  other delays (non-typically developing; NTD); and 64 were developing typically (TD).Twelve month AOSI TS predicted 36 month ASD status better than 6 month AOSI TS. There was a difference in AOSI TS across the three outcome groups at 12 months (Kruskal-Wallis, p=0.004), but not at 6 months (ANOVA, p=0.12).  AOSI 12 month scores (cutoff point = 6 using distance closest to (0,1) and Youden criteria) had a sensitivity of 0.63 and specificity of 0.68 (Area under Curve (AUC) =0.68; PPV=0.37, NPV=0.83) when comparing children with ASD to Non-ASD samples overall.  At 12 months, when comparing ASD vs. TD and ASD vs. NTD, the AOSI had a higher sensitivity (vs. TD=0.61, vs. NTD= 0.57) and specificity (vs. TD=0.77, vs. NTD=0.62).  The AOSI score had little predictability at 6 months when comparing ASD to Non-ASD (AUC=0.59).  36 month ADOS Comparison Scores were significantly correlated to the AOSI TS at 12 months (r=0.24; p=0.008), but not 6 months (r=0.08; p=0.33).

Conclusions:  Our findings are consistent with previous reports of emergence of behavioral symptoms over time as the AOSI differentiated children with ASD from NTD/TD better at 12 months than 6 months. Additionally, the 12-month AOSI better differentiated ASD from TD than NTD, which is not surprising given the overlap in behavioral presentation between ASD and NTD groups.  Along with overall diagnostic classification, the 36-month ADOS was significantly correlated with the AOSI at 12 months, supporting the emergence of the ASD behavioral phenotype over time.