Autism Symptom Development from 6 to 18 Months of Age in a Prospective High-Risk Cohort: Findings from the Autism Observation Scale for Infants
Objectives: To examine the predictive validity and classification properties of the AOSI in the original cohort of high-risk (HR; infants with an older sibling with confirmed ASD diagnosis) and low-risk (LR; infants with no first- or second-degree relatives with ASD) for which the measure was developed.
Methods: Participants: 501 HR infants, including 136 diagnosed with ASD at age 3 (HR-ASD; 27.1%) and 365 not diagnosed (HR-N), as well as 180 LR infants (one diagnosed with ASD; not included in these analyses). Assessments: The AOSI (Bryson et al., 2008) was administered at 6, 9, 12, 15 and 18 months (9- and 15-month assessments were introduced later in the study). 3-year ASD diagnoses were based on clinical best estimate using the ADOS and ADI-R at 36 months, blind to prior study data. Participants were also assessed at 3 years using the Mullen Scales of Early Learning (MSEL) and Vineland Adaptive Behavior Scales (VABS). Analytic Approach: Groups defined by risk status and ASD outcome (HR-ASD, HR-N, LR) were compared on AOSI total scores at each timepoint using linear mixed modeling, with post-hoc pair-wise group comparisons at each age using Tukey HSD. Next, we used receiver operating characteristics (ROC) curves and Youden’s index to determine the optimal cut-point for AOSI total score at each age (limited to the HR cohort), and estimated sensitivity, specificity, positive predictive value and negative predictive value. Finally, we compared clinical features (ADOS severity, MSEL and VABS scores) at age 3 in children with ASD correctly vs. incorrectly classified by the AOSI, based on these identified cut-points. Results: Findings are summarized in Tables 1 and 2. There were ASD-specific group differences in AOSI total scores starting at 12 months, with higher scores in HR-ASD compared to HR-N and LR groups. AOSI total scores were predictive of ASD within the HR group with a sensitivity and specificity of 52% (95% CI: 48-56%) and 74% (95% CI: 70-78%) at 12 months, and 72% (95% CI: 68-76%) and 65% (95% CI: 61-69%) at 18 months, respectively, based on cut-points identified by ROC analyses. Children with ASD that could be identified based on elevated AOSI score at 12 months, compared to those with lower AOSI scores, had more severely delayed language (MSEL – Expressive Language subscale; p=.03), adaptive functioning (VABS-2 – Social and Communication subscales; p=.04 and .02, respectively), and higher levels of parent-reported ASD symptoms (ADI-R) (p<.001) at age 3.
Conclusions: AOSI total scores discriminate children with ASD ascertained from a HR cohort from non-diagnosed HR and LR peers starting at 12 months, which may assist with making earlier referrals to diagnostic and intervention services. Elevated scores are associated with more severe language and adaptive delays and ASD symptoms at diagnosis.