Journal of Biometrics & Biostatistics

Characterizing Components in a Mixture Model for Birthweight Distribution

Abstract

Wei Deng, Richard Charnigo, Hongying Dai and Russell S. Kirby

Low birthweight (LBW) is a well-known risk factor for infant mortality worldwide. Although infant mortality has decreased in the United States during the past 20 years, the incidence of LBW has increased, suggesting that further reductions in infant mortality may be possible if the incidence of LBW can be reduced. In the present work, we introduce a new analytic framework for revealing the relationships between latent variables representing components in a mixture model for birthweight distribution and various other risk factors. More specifically, we show how to estimate the probability that a risk factor is present within one of the mixture components as well as the probability of mixture component membership among infants for whom a risk factor is present, both at a fixed birthweight and averaged across birthweights. We illustrate our analytic framework using publicly available data for white singletons born in the United States between 1998 and 2002. This framework provides a quantitative approach for the prediction of how addressing a modifiable risk factor may affect both the incidence of LBW and infant mortality, thereby facilitating decision making regarding resource allocation toward addressing that risk factor.