Evaluation of Alternative Methods for Constructing Pediatric Growth Charts Based on the 2000 CDC Growth Charts

• Activity Number: 684
• Type: Contributed
• Date/Time: Thursday, August 5, 2010 : 10:30 AM to 12:20 PM
• Sponsor: Section on Statistics in Epidemiology
• Author(s): Rong Wei*+ and Lester Randolph Curtin
• Companies: National Center for Health Statistics and CDC
• Keywords: CDC growth charts ; percentile smoothing ; LMS transformation ; quantile regression

Abstract

Growth charts are extensively used in epidemiologic studies and in pediatric practice to track the growth of infants, children and adolescents. In year 2000, CDC published a new set of growth charts for use in the United States (CDC, 2000). CDC used a combination of parametric modeling and Box-Cox normalization, which provides parameters to interpolate growth measures at any ages into exact percentiles in the reference population. In this paper, we compare the CDC method to two alternative methods: the LMS method (Cole and Green, 1992) and quantile regression (Wei et al. 2005). The comparison was based on the underlying data used for constructing the CDC growth charts. Strengths and limitations for each alternative method will be discussed. Some specific data complications, including extreme body measures, normalization of data, and effects of data variation will also be discussed.