Joint Modeling of Birth Weight and Gestational Age

• Activity Number: 142
• Type: Contributed
• Date/Time: Monday, August 4, 2008 : 10:30 AM to 12:20 PM
• Sponsor: Section on Bayesian Statistical Science
• Author(s): Scott L. Schwartz*+ and Alan Gelfand
• Companies: Duke University and Duke University
• Keywords: Mixture Model ; Bivariate Mixture Model ; Interval Censoring ; Bayesian Modeling ; Birthweight ; Gestational Age

Abstract

Customary modeling for birth weight (BW) is conditional on gestational age (GA). However, joint modeling of GA and BW offers increased flexibility and interpretation in studying relevant birth outcomes (e.g., small for GA). First, we model the bivariate distribution of BW and GA using mixtures of bivariate normals. Second, due to the interval censoring associated with GA, we introduce this mixture model as a latent specification. Finally, we introduce well-established risk factors in the form of a bivariate regression to facilitate interpretation of the joint model. The mixture modeling enables us to capture the clearly non-Gaussian distributional shapes observed in histograms. The model is applied to a portion of the 1990-2006 NC Detailed Birth Record (N=1862405) through CEHI at Duke University and is fitted using Bayesian methodology and MCMC approaches.