Developmental screening tests are widely used for early identification but few are studied for their accuracy—the percentage of children with and without problems correctly detected. The absence of such data makes it difficult for professionals to choose measures wisely and to avoid those that under-detect or over-refer. In this article, the accuracy of three developmental screening tests was assessed. Measures included the Academic Scale of the Developmental Profile-II (DP-II), the Denver II, and the Battelle Developmental Inventory Screening Test (BDIST). Each was administered, along with a criterion battery of intelligence, adaptive, language, and achievement measures, to 89, 7- to 70-month-old children, enrolled in one of five day care centers. Twenty percent of the children tested were found to have developmental disabilities, including language delays, mental retardation, developmental delay, and autism. The Academic Scale identified fewer than one in four of the children with diagnoses, although under-referrals were minimal. The Denver-II detected the majority of children (if questionable/untestable scores were grouped with abnormal) with difficulties, but most children without problems also failed the measure. The BDIST, using the 1.5 cut-offs, was more accurate than the other screening tests and identified correctly 72% of the children with difficulties, and 76% of the children without diagnoses.

Comment: The BDIST has been revised since this article was written. The Developmental Profile is now out in its 3rd edition (but without evidence that it works as a screening measure, i.e., no cutoff scores but is helpful for NICU follow-up and EI intake as is the more recently published PEDS^®: Developmental Milestones. The Denver-II remains as it was in 1992 and continues to under-detect if questionables/untestables are not referred and over-detected if they are referred. Work on this tool is needed!