Language disorder from infancy to adulthood: predictors and outcomes

Abstract

Developmental language delays are common and are associated with increased risk of adverse educational and psychosocial outcomes later in life. Early intervention services supporting language development may prevent these later problems from occurring. However, with the variability that exists in language development, it is less clear which children with developmental language delays actually require intervention. As such, increased understanding of the trajectory of language impairment and the long-term outcomes in adulthood is needed. The purpose of this thesis was to: (i) determine the early life characteristics that placed children at risk for poor language outcomes, and (ii) profile the long-term psychosocial outcomes in adulthood for language impaired populations. To address these aims prospective longitudinal data from two Australian pregnancy cohort studies; the Mater-University of Queensland Study of Pregnancy (MUSP) and the Western Australian Pregnancy Cohort (Raine) Study, were utilised. The first three studies of this thesis investigated the early life characteristics that predicted later language outcomes in childhood and adulthood. Firstly, Chapter 2 investigated the parental and early childhood factors present at, or prior to 5 years, which were associated with different vocabulary developmental profiles in the MUSP cohort. Receptive vocabulary skills were measured at 5- and 21- years, using the Peabody Picture Vocabulary Test-Revised (PPVT-R). A one standard deviation cut-off was applied to determine average and below average performances at the two assessment points. Four vocabulary developmental profiles were then identified, including: ‘consistently typical’ (average performance at both time points), ‘consistently low’ (below average scores at both time points), ‘improved’ (below average performance at 5 years but average performance at 21 years), and ‘deteriorated’ (average skills at 5 years but below average performance at 21 years). The results revealed that adults with deteriorating vocabulary profiles were more likely to have had internalising behaviours at 5 years, mothers who smoked during pregnancy, and fathers with lower educational attainment, compared to adults with consistently typical vocabulary histories. No significant risk factors were identified for adults showing consistently low vocabulary in comparison to those adults whose early vocabulary delays had improved. Applying the same methodological approach as Chapter 2, Chapter 3 investigated possible predictors of change in language performance from 2 to 10 years in the Raine cohort. At 2 years, expressive vocabulary was measured using the Language Development Survey (LDS) and at 10 years, language ability was measured using the Clinical Evaluation of Language Fundamentals (CELF-3). Children with deteriorating language profiles were more likely to be male and have mothers who smoked during pregnancy, have fathers with lower educational attainment, come from families with lower income, and not be read to at 2 years by their parents. Children with consistently low language skills were also more likely to have mothers who smoked during pregnancy compared to those late talkers whose language skills resolved to within normal limits at 10 years. Chapter 4 investigated the predictive accuracy of a general developmental screening instrument (the Infant Monitoring Questionnaire, IMQ) at 3 years of age for explaining language outcomes at 10 years. The predictive ability of traditional logistic regression models was compared to the predictive statistics obtained when using machine learning algorithms. Furthermore, model accuracies were calculated using only communication subscale data as well as broader IMQ data encompassing fine and gross motor, adaptive and personal-social skills. The findings obtained from the machine learning algorithms were equivalent to logistic regression, and the inclusion of other behaviours of neurodevelopment (motor, social and adaptive skills) did not improve predictive accuracy over and above communication skills alone. Chapters 5 and 6 then investigated the long-term psychosocial outcomes into adulthood. Chapter 5 examined the long-term prognosis for late talkers in adulthood in terms of their levels of autistic-like traits as measured by the Autism-Spectrum Quotient (AQ) at the 20-year follow-up in the Raine Study. The results of this chapter revealed that a smaller vocabulary size at 2 years was associated with higher levels of autistic-like traits in adulthood, particularly in the area of social communication. Finally, Chapter 6 examined the educational, vocational and mental health outcomes at 21 years in the MUSP for the four different vocabulary developmental profiles identified in Chapter 2. Results indicated that those adults with deteriorating vocabulary skills had poorer outcomes across the educational, vocational and mental health domains. Adults with consistently low vocabularies had lower educational and vocational prospects, and adults with improved vocabulary skills remained at increased risk of poorer vocational outcomes compared to adults with consistently typical profiles. The findings of this thesis have implications for the identification of those children who require early intervention language services, by increasing awareness of children at risk of persisting language difficulties. It also emphasises the increased risk of future educational and psychosocial problems in children with early and later-emerging language difficulties, highlighting the need for universal preventative strategies to enhance early language development.