Selection of Preschool Language Tests

Abstract

Clinicians are confronted with a wide range of norm-referenced tests designed to evaluate preschool language skills. This article approaches test selection from a data-based perspective. Twenty-one tests of language skills that included norms for children ages 4 and 5 years were reviewed for information on 10 psychometric criteria. Only 38% of these tests met half or more of the 10 psychometric criteria employed by McCauley and Swisher (1984a) in their review. Four tests that met a relatively high number of psychometric criteria (6 or more) were administered to 20 preschool children with specific language impairment and 20 age-matched controls. High interexaminer reliability was obtained for all tests. All tests had low correlations with a measure of nonverbal skills. Only one of the four tests provided acceptable accuracy in discriminating between the children with normal and the children with impaired language in our sample. Our empirical examination of four tests revealed that even tests that pass relatively high numbers of psychometric criteria may not be precise discriminators of normal and impaired language in 4- and 5-year-old children, indicating the need to complement psychometric review with data-based validation procedures.

Telehealth Measures Screening for Developmental Language Disorders in Spanish-Speaking Toddlers

“…36,37 In screening for DLDs, diagnostic accuracy values greater than 0.80 are considered desirable, values between 0.70 and 0.79 are considered suggestive, and values less than 0.7 are undesirable. [37][38][39] …”

The Composition of Normative Groups and Diagnostic Decision Making: Shooting Ourselves in the Foot

“…Specificity is the correct classification of typically developing children as having normal language (NL). Sensitivity refers to the correct classification of children with LI as having LI (Dollaghan, 2004;Plante & Vance, 1994). Here we ask an independent question: whether sensitivity and specificity might also be affected by the inclusion of children with LI in norms.…”

“…As seen here, characteristics of the normative sample directly affect diagnostic accuracy. Plante and Vance (1994) suggest that tests with 80% specificity and sensitivity have " fair " classification accuracy and those with 90% or more have "good " classification accuracy. In this simulation, sensitivity changes from good to fair at a cut score of -1.14 by including individuals with LI in the normative sample.…”

“…Here we demonstrate that the composition of the normative group can either support the use of a test for absolute score interpretations or undermine it for relative interpretations. Elsewhere, we (Merrell & Plante, 1997;Plante & Vance, 1994) and others (McCauley & Swisher, 1984b) have criticized the practice of drawing therapy targets from norm-referenced test items from both psychometric and evidence-based perspectives. Therefore, it is typically not the case that the psychometric characteristics of a test will be optimal for multiple diagnostic purposes.…”

“…This evolution is driven, in part, by clinical training that emphasizes psychometric review of test properties (e.g., McCauley & Swisher, 1984a, 1984b and the importance of the diagnostic purpose and the related clinical decisions in test validity ( Merrell & Plante, 1997;Plante & Vance, 1994, 1995 and evidence-based practice (Dollaghan, 2004). These trends have decreased the emphasis on general psychometric checklists in favor of evidence of strong sensitivity and specificity for tests intended for use in identifying cases of child LI ( Plante & Vance, 1994). More recently, researchers have placed an emphasis on the content representation of tests intended for the diagnosis of particular forms of LI.…”