A Comparison of Three Methods for Estimating Percentile Curves for Language Development in Down Syndrome Babies (303059)

Keywords: curve estimation, quantile regression, moving average, splines

The Boston Children's Hospital Down Syndrome (DS) clinic sees babies up to three years of age, with visits every 6--9 months beginning at 7--9 months of age. Language skills are tested at each visit using a standardized test. Currently, clinicians report to families only the raw scores so the child is compared against itself and progress over time can be assessed. There are no norms for typically developing DS babies. Parents often question how their child is doing compared to other DS babies, rather than against only themselves or norms from typically developing peers (non-DS ‘normals’). In this talk, I discuss the development of percentile curves for language development in a large sample of DS babies. Results from three approaches are presented and compared: quantile regression, a moving-percentile approach, and mixed model splines. Some SAS code will be presented as part of the talk, but software is not a central feature. Clinical implications and limitations are briefly considered with the conclusions.