Combining Nonword and Sentence Repetition to Identify Developmental Language Disorder.

The DLD group was small (28 children), and when the model was tested by holding out one child at a time it gave 50.0% sensitivity compared with 85.7% in the main analysis, so the reported sensitivity may be optimistic. The sample came from two urban schools in northern Vietnam and parents self-selected, so it may not represent all Vietnamese-speaking children. The 46 children classified as Risk were combined with the typically developing group; this broadened the comparison group but could have placed some children with DLD in the typical-language group. The item scores were turned into a predicted probability using weights and a cutoff derived from this sample, and those values need validation before clinical use. The study did not test whether practitioners without phonetics training could score the tasks as easily as trained research assistants.

The authors declare that the first author developed the nonword task, and the first and second authors developed the sentence repetition task; the first, second, and fourth authors have been involved in previous publications using these tools. No funding source is reported in the supplied text.

Do not treat the combined model's 85.7% sensitivity and 81.6% specificity as proof that it will work in your clinic. When the model was tested by holding out one child at a time, it gave 50.0% sensitivity and 76.2% specificity, and the model used sample-specific weights and cutoffs.