Models of Birthweight and Low Birthweight

Rand Corporation, Santa Monica CA

Investigators

Abstract

DESCRIPTION (provided by applicant): This project will apply a new statistical approach to analyzing an important topic in health research: identifying the role that demographic, socioeconomic, and behavioral factors play in explaining differences in birth weight. Methods in currently used in linear regression for mean birth weight and logit or probit regression for low or very low birth weight status--suffer from serious drawbacks. Our solution is to use quantile regression to study covariate effects across the entire birth weight distribution. This approach promises to provide new and important insights into the determinants of birth weight. The results will help inform future research as well as public policy--in particular, efforts to reduce substantial disparities in birth weight by race/ethnicity and socioeconomic status, which have important consequences for health disparities across the life course. The specific aims of the project are: 1) To estimate models of birth weight using quantile regression and assess the benefits of this approach compared to ordinary least squares and logit or probit regression; and 2) To investigate the determinants of birth weight using quantile regression, focusing on new substantive findings that emerge and their implications for public policies to improve birth weight, and comparing these results to those based on traditional approaches. Data from two complementary sources will be used for this project. The first is the 1997 Natality Data Set (NDS), compiled from birth certificates by the National Center for Health Statistics. The second data source is the 1988 National Maternal and Infant Health Survey (NMIHS).