Service Use Among Transition-Age Youth with Autism Spectrum Disorder

Although this score is typically subtracted from 100 to yield an index that becomes more negative with greater levels of maladaptive behavior, the original summary score was used to simplify interpretation.

Although number of services is technically a discrete count, OLS regression was justified on the basis that: (a) Values in the sample extended the full range (0-15) and were approximately normally distributed (M = 6.13, SD = 3.23, skewness = 0.27, kurtosis = -0.53); (b) predicted values generated using the final model were approximately normally distributed (M = 6.13, SD = 2.05, skewness = 0.29, kurtosis = 0.63), did not extend beyond the range of possible values, and did not include negative values (range = 1.96 to 11.72); and (c) model residuals were normally distributed and no influential outliers were detected (leverage values all < .20). A sensitivity analysis was conducted to evaluate the potential for bias resulting from the use of OLS. A logistic regression model with aggregated data was determined to be the most appropriate alternative model. In such a model, each of the 15 services is regarded as a trial, the number of trials is fixed, the trials are independent, and the probability of “success” (i.e., use) is the same for each service. The model estimates the effect of each predictor on the odds of using an additional service.