15433
Early Predictors of Expressive and Receptive Vocabulary in Initially Nonverbal Preschoolers with ASD
Between 24 – 30% of children with autism spectrum disorders (ASD) do not use spoken communication by 5 years old. Spoken communication at 5 years predicts later adaptive outcomes in individuals with ASD. Identifying parsimonious predictive models of theory-guided predictors of vocabulary growth moves us towards understanding the variability in learning to communicate via speech within the ASD population. Parsimonious models include only predictors that account for significant variance in the outcome after statistically controlling for other predictors.
Objectives:
The incremental validity of eight putative predictors of growth curves of parent-reported expressive and receptive vocabulary was tested. Each putative predictor had empirical and theoretical grounds for selection.
Methods:
Eight-six initially nonverbal preschoolers with ASD were assessed 5 times in 4-month intervals over 16 months. Receptive and expressive vocabulary sizes were estimated using a parent report (i.e., the McArthur-Bates Communicative Development Inventory, Words and Gestures). The Communication and Symbolic Behavior Scale, Early Social Communication Scales, Motor Imitation Scale, Mullen Early Learning Scale, 2 parent-child interaction sessions, the Developmental Play Scale, and an oral motor assessment were used to measure putative predictors. Mixed level modeling was used to quantify individual vocabulary growth curves. The Time 5-centered intercept was selected as the parameter of interest because it is arguably the most interpretable parameter when quadratic or cubic models are needed to model growth (i.e., the best estimate of vocabulary size at Time 5). Putative predictors were measured at Times 1 or Time 2. To afford interpretable effect sizes, ordinary least square estimates of the intercepts for the growth curves were analyzed as the criterion variables in multiple regressions used to identify unique predictors (i.e., after controlling all other predictors in the model).
Results:
Quadratic models fit the data better than simple linear models. Except for IQ, all putative predictors predicted either expressive or receptive vocabulary. After controlling for other variables and after reducing the model to predictors with incremental validity, 3 predictors remained in each model. The number of parental linguistic responses to child leads at Time 2 (R2 change = .12), number of intentional communication acts at Time 1 and 2 (R2 change = .11), and number of different object play actions at Time 1 (R2 change = .04) added to account for 29% of the variance (adjusted R square) in expressive vocabulary. The number of words understood at Time 1 (R2 change= .27), number of object play actions at Time 1 (R2 change = .10), and number of parental linguistic responses to child leads at Time 2 (R2 change= .04) added to account for 49% of the variance in receptive vocabulary. Putative predictors without incremental validity were oral motor functioning, IQ, motor imitation, responding to joint attention, and consonant inventory.
Conclusions:
The results support selecting the unique predictors as goals for nonverbal children with ASD. The number of predictors, number of measurement periods, use of growth curves, long interval between predictor measurement and end-point of study, and large sample size make this study particularly important to the field.