Purpose: In this study, the authors explored alternative gold standards to validate an innovative, dialect-neutral language assessment.
Method: Participants were 78 African American children, ages 5;0 (years;months) to 6;11. Twenty participants had previously been identified as having language impairment. The Diagnostic Evaluation of Language Variation—Norm Referenced (DELV–NR; Seymour, Roeper, & J. de Villiers, 2005) was administered, and concurrent language samples (LSs) were collected. Using LS profiles as the gold standard, sensitivity, specificity, and other measures of diagnostic accuracy were compared for diagnoses made from the DELV–NR and participants’ clinical status prior to recruitment. In a second analysis, the authors used results from the first analysis to make evidence-based adjustments in the estimates of DELV–NR diagnostic accuracy.
Results: Accuracy of the DELV–NR relative to LS profiles was greater than that of prior diagnoses, indicating that the DELV–NR was an improvement over preexisting diagnoses for this group. Specificity met conventional standards, but sensitivity was somewhat low. Reanalysis using the positive and negative predictive power of the preexisting diagnosis in a discrepant resolution procedure revealed that estimates for sensitivity and specificity for the DELV–NR were .85 and .93, respectively. Conclusion: The authors found that, even after making allowances for the imperfection of available gold standards, clinical decisions made with the DELV–NR achieved high values on conventional measures of diagnostic accuracy.
Journal or Book Title
Journal of Speech, Language, and Hearing Research