Behavioral Predictors of Improved Speech Output in Minimally Verbal Children with Autism
Objectives: We wished to understand what factors predict the amount of improvement in % Syllables Approximately Correct after 25 sessions of an intonation-based spoken-language therapy in 38 children with MV ASD.
Methods: Participants were 38 children receiving one of two spoken-language treatments (7 female, aged 3;5-10;8). Treatments were Auditory Motor Mapping Training (AMMT), a novel intonation-based intervention; and Speech Repetition Treatment (SRT), a non-intonation-based control intervention. Inclusion criteria were: diagnosis of ASD, expressive vocabulary ≤20 words, no productive syntax, and the ability to repeat ≥2 phonemes. Seven predictor variables were chosen based on the literature: Sex; chronological age; and baseline measures of expressive language, phonetic inventory, autism severity, and nonverbal IQ. First, bivariate correlations were performed to eliminate predictors not significantly correlated with the outcome variable. Then, regression models were fit to assess the effect of the remaining predictors. Both complete case and multiple-imputation analyses were employed because of missing predictor data. Finally, because some children in each group improved, subgroup analyses were performed to investigate whether predictors differed according to treatment group and improvement status.
Results: Over the course of 25 treatment sessions, AMMT participants (n = 27) improved by a mean of 17.8% Syllables Approximately Correct (SD 18.8), while SRT participants (n = 11) improved by a mean of 0.5% Syllables Approximately Correct (SD 12.0). Bivariate correlations eliminated all predictors save autism severity (ADOS score) and phonetic inventory (the number of phonemes a child could correctly repeat). Together, the two variables accounted for 50-60% of the variance in change in % Syllables Approximately Correct. Across groups and analyses, every additional phoneme a child could repeat at Baseline resulted in a 1.4-2% increase in the amount of improvement after 25 therapy sessions. A one-point increase in ADOS score resulted in a 2-3% decrease in the amount of improvement. Complete-case analyses resulted in lower standard errors than multiple-imputation analyses, despite the smaller number of participants (24 complete cases out of 38 participants total). Phonetic inventory significantly predicted improvement only in children who received AMMT.
Conclusions: ADOS score emerged as a significant predictor when SRT participants were included. These results are consistent with previous findings that children with MV ASD benefit more, on average, from intonation-based therapies and suggest that intonation may be more efficient at enabling these children to harness their speech production abilities to learn new phonetic patterns. The fact that age did not significantly predict improvement is optimistic, in that it means that older children may still show some improvement in spoken language.