Background: The current project was a prospective feasibility study using a dynamic assessment of statistical learning of a novel morpheme for the identification of developmental language disorder (DLD) in Spanish-English–speaking preschoolers. The aims were to (a) create a reliable dynamic assessment and (b) determine which elements of scoring yielded the highest classification accuracy. Method: Spanish–English bilinguals, aged 4;0–5;11 (years;months), who had no history of neurologic or other impairments were eligible ( N = 18; 12 typically developing TD, six with DLD). The data were collected from a convenience sample in Arizona. All participants were administered the dynamic assessment in English followed by the reference standard, the Bilingual English-Spanish Oral Screener, which determined the language status as DLD or TD. Results: The dynamic assessment had high fidelity ( M = 98%), and the child accuracy scores had high interrater reliability ( M = 99%). However, the modifiability rating scale had moderate interrater reliability ( ρ = .62). Receiver operator characteristic curves using Bayesian logistic regression and random forests were used to determine the area under the curve (AUC). The classification accuracy was compared for each scoring element. Combining the modifiability rating scales and the teaching phase score yielded the highest AUC (Bayesian AUC = .94, 95% CI .81, 1.00; random forest AUC = .87, 95% CI .61, 1.00) with fair sensitivity (83%) and good specificity (100%) for this sample. Discussion: The feasibility of the procedure and its potential for diagnostic accuracy were demonstrated. Conclusions are not broadly generalizable given the small sample size and case–control design, but the results merit further investigation. Modifiability rating scales should be adjusted, and both languages should be assessed in future studies.
Neiling et al. (Tue,) studied this question.