The Diagnostic Value of English Language Sample Measures for Identifying Developmental Language Disorder in Bilingual...

This study examines the diagnostic accuracy of four measures derived from English narrative language samples for identifying developmental language disorder (DLD) in 5- and 6-year-old Spanish-English bilinguals.

Ninety-two children aged 5;0-6;11 (years;months) produced narrative retells in English using Mercer Mayer's frog stories. Forty-six matched pairs of children with DLD and typical language were identified based on age and English language exposure. Percent grammatical utterances (PGU), errors per C-unit (EPC), mean length of utterance (MLU), and subordination index (SI) were calculated. Pooled, covariate-adjusted, and multivariate receiver operating characteristic curves were generated to evaluate the measures' diagnostic performance at optimal cutoffs. Analyses were also performed at different levels of language exposure (i.e., high English, balanced Spanish/English, high Spanish).

None of the measures reached 80% for both sensitivity and specificity for the full participant group. Covariate-adjusted curves based on exposure were not significantly different from pooled analyses. When exposure groups were analyzed separately, PGU and EPC each yielded good diagnostic accuracy for the high English group (> 70% exposure) at 94% sensitivity and specificity. Combining these measures with MLU and/or SI yielded similar results. For the balanced exposure group, all measures fell below the 80% threshold.

Measures of grammaticality (PGU and EPC) were the best indicators of DLD when analyzing English narrative language samples of young bilinguals with at least 70% exposure to English. None of the measures evaluated were adequate for identifying DLD in bilinguals with less exposure to English. Clinical implications are discussed.

https://doi.org/10.23641/asha.31449370.