Cognitive-linguistic processing markers for the identification of European Portuguese speaking school-age children with specific language impairment

Abstract

The purpose of this dissertation was to extend to European Portuguese three processing-dependent measures identified as being effective in differentiating children with Specific Language Impairment (SLI) from typically developing peers: (1) Children’s Non-word Repetition Test (CNRep; Gathercole & Baddeley, 1996), (2) Competing Language Processing Task (CLPT; Gaulin & Campbell, 1994), and (3) Recalling Sentences- subtest from the Clinical Evaluation of Language Fundamentals (CELF-3; Semel, Wiig, & Secord, 1995). Given that processing-dependent measures have shown promise as nonculturally biased tasks that could be used in the identification of children with SLI, it was hypothesized that these European Portuguese versions would prove valuable in the identification of school-age children with SLI in Portugal. Three measures were developed: (1) European Portuguese Children’s Non-word Repetition Test (EP-CNRep), (2) European Portuguese Competing Language Processing Task (EPCLPT), and (3) European Portuguese Recalling Sentences Task (EP- RecSent). Participants in this research study were 150 children (75 children with SLI; 75 typically developing age-gender-matched peers), aged 7 to 11 years, from 2nd to 5th grades, in 17 state-funded schools (rural, suburban and urban contexts) in the North of Portugal (i.e., Districts of Braga and Viana do Castelo). Group differences indicated that performance on EP-CNRep, EP-CLPT, and EP-RecSent tasks was significantly lower for children with SLI when compared with their typically developing peers. In this study, analysis was performed through a strong graphic method, which studies the variation of sensitivity, specificity and likelihood ratios to different cut points. The analysis for each of the measures is described, as well as the relationship with the index area under the ROC (Receiver Operating Characteristic) curve. Findings demonstrate that all measures differentiated children with SLI from their typically developing peers with accuracy: EP-CNRep (.971), EP-CLPT-T/F (.703), EP-CLPT-Words Recalled (.826), and EP-CLPT-Span level (.726), and EP-RecSent (.855). ROC curve analysis revealed an excellent value for the European Portuguese Children’s Nonword Repetition Test (.971) and was therefore considered the most accurate marker to differentiate European Portuguese-speaking school-age children with SLI, followed by EP-RecSent (.855), and EP-CLPT-Words Recalled (.826). Results from this study show these processing-dependent measures (EPCNRep, EP-CLPT, EP-RecSent) to be appropriate to distinguish European Portuguese school-age children with SLI and their typically developing peers. Since all three measures had good sensitivity/specificity values at different cut points, further clinical and educational feasibility can be determined by combining results from these measures in identifying children with SLI in Portugal.

O objectivo desta dissertação consistiu em desenvolver quatro provas no âmbito do processamento cognitivo-linguístico, previamente identificadas como sendo eficazes na diferenciação entre crianças com e sem Perturbações da Linguagem (PL) em Idade Escolar. O Teste de Não-palavras, uma versão adaptada do Children´s Non-word Repetition Test (CNRep; Gathercole & Baddeley, 1996), a Tarefa de Processamento da Linguagem, uma versão adaptada de Competing Language Processing Task (CLPT; Gaulin & Campbell, 1994) e a Tarefa de Rechamada Frásica adaptada do Teste Clinical Evaluation of Language Fundamentals-3 (CELF-3; Semel, Wiig, & Secord, 1995). As provas foram adaptadas à realidade portuguesa e analisadas em termos psicométricos. A sua aplicação decorreu em 17 escolas da rede pública em contextos urbanos, suburbanos e rurais nos Distritos de Braga e Viana do Castelo. A amostra consistiu em 150 crianças, dos 7 aos 11 anos de idade: 75 crianças identificadas com PL e 75 pares aleatórios da mesma idade cronológica e género. Os resultados do estudo comparativo indicam que existem diferenças significativas entre os grupos para as quatro provas, havendo também evidência de que o factor idade apresenta uma correlação positiva e forte no desempenho dos grupos. O desempenho das crianças com e sem PL nas quatro provas aumenta à medida que a idade cronológica também aumenta. Para todas as provas estudadas, foi realizada uma análise criteriosa de sensibilidade, especificidade, ratio de máxima verossimilhança (Likelihood Ratio) e curva ROC (Receiver Operating Characteristic) de modo a conseguir investigar o poder discriminante para diferentes valores de corte relativamente aos dois grupos de crianças. Os resultados indicaram que as provas mostram valores considerados bastante razoáveis para os parâmetros de análise ao teste e com valores preditivos elevados na discriminação das crianças com e sem Perturbações da Linguagem. O Teste de Não-palavras apresenta a curva ROC com o valor mais elevado (.971) sendo considerada a prova com maior poder discriminativo e, consequentemente com maior potencial de identificação de crianças com Perturbações da Linguagem em Idade Escolar.