Predicting Early Language Gains in Young Children On the Autism Spectrum

Background: There is a scarcity of information regarding predictors of language outcomes in young children on the autism spectrum. Some research has examined predictors of language and communication within investigations of the predictive validity of early autism diagnostic assessments more generally (Charman et al., 2005), whereas other studies have focused specifically on predicting language outcomes (Charman et al., 2003; Paul et al., 2008). Prior investigations have comprised small samples and have yielded conflicting findings. Additional research is needed to provide a clearer characterization of early predictors of language development in young children on the autism spectrum.
Objectives: The goal of this investigation was to examine predictors of early language gains from 2 ½ to 3 ½ years of age in a large and well-specified group of young children on the autism spectrum.
Methods: A total of 102 children on the autism spectrum who are enrolled in an ongoing longitudinal investigation of language development participated in this study. Autism spectrum diagnoses were determined using comprehensive evaluations including the ADI-R and ADOS/ADOS-T. Participants had a mean chronological age of 31 months (25-36 mo) at Visit 1 (V1). Language and communication abilities at V1 were measured using the Communicative Development Inventory: Words & Gestures (CDI), Vineland Receptive and Expressive Communication scales, and Preschool Language Scale (PLS-4). Cognitive abilities were assessed by the Bayley Scales of Infant and Toddler Development – III, and joint attention was evaluated using the Early Social Communication Scales. SES was indexed by years of maternal education. Visit 2 occurred approximately one year after the initial evaluation, at a mean age of 44 months (37-53 mo). Language outcomes at Visit 2 (V2) were measured using the Auditory Comprehension and Expressive Communication Scales of the PLS-4.
Results: Pearson correlation coefficients indicated that V1 and V2 performance on the PLS-4 was significantly correlated, both for Auditory Comprehension (r=.71, p=.000, n=89) and Expressive Communication (r=.77, p=.000, n=91). Regression models (n=78 by excluding cases that did not have all measures at both visits) were constructed to conduct a residualized gain analysis such that V2 outcome was modeled both as a function of the same variable at V1 as well as other variables suspected to relate to growth in performance on the PLS-4. Gains in PLS-4 Auditory Comprehension scores were significantly predicted by Response to Joint Attention (p=.013); CDI words understood approached but did not reach statistical significance (p=.066). Gains in PLS-4 Expressive Communication scores were significantly predicted by SES (p=.003), Response to Joint Attention (p=.014), and Bayley cognition (p=.026). Variables that were non-significant were calibrated ADOS scores, Initiating Joint Attention, Vineland and CDI comprehension and production.
Conclusions: Language abilities at 2 ½ years were strongly predictive of language skills at 3 ½ years. Several non-linguistic measures predicted gains in early language (as measured by the PLS-4). Cognitive skills predicted gains in language production but not comprehension. Higher SES was associated with more gains in productive language. Response to joint attention was the only significant predictor of gains in both language comprehension and production.