Statistical Modeling for Findings the Determinants of Neonates’ Very Low Birth Weight

Abstract

Aim of the epidemiological research is to identify a causal relationship between the risk factors and the disease. The present study aims to identify the determinates of neonates’ very low birth weight which have significant effects on their birth weight using generalized additive probabilistic modeling. In this present study very low birth weight (BWT) is the response variable with heterogeneity and non-normality in nature which can be modeled through either by the Log-normal or by the gamma models. A well-known method is joint modeling of mean and variance (JGLM) to handle this heterogeneity and non-normality but this study introduced the most advanced regression techniques namely generalized additive model (GAM). Materials and Methods: The present article is based on the secondary data on 174 very low birth weighted neonates along with 26 explanatory factors/ variables. The very low birth weight of 174 neonates is heterogeneous, positive, and gamma distributed. Therefore, generalized additive model with gamma distribution and log link function has been introduced to analyze this very low birth weight. The Native American neonates has the smaller birth weight (BWT) than the other racial neonates namely black, white or oriental (P-value = 0.009). The BWT is smaller for those neonates who were outsiders from Duke (P-value = 0.05) and it also decreases to the female births (P- value=0.09). BWT is higher for the pneumothorax infants than non-pneumothorax with P-value <0.001. It is smaller for those infants for whom oxygen supply had been introduced in between 30 days of his/her birth (P-value=0.002). The neonates who were not been survived they also have the smaller birth weight than the alive neonates (P-value= 0.07). Besides these factors, lowest pH in first 4 days of neonates’ life (P-value=0.09), Apgar score at one minute (P-value=0.03) and the interaction effects of lowest pH and Apgar score (0.03) are the significant variables (continuous cofactors) for neonates’ very low birth weight. Hospital stays in number of days, platelet counts and gestational age in weeks are also highly significant factors and these are entered in the GAM model as a non –parametric smoothing term (P-value<0.001). The birth weight of each new born baby is identified as heterogeneous and gamma distributed (possible). Most of the present findings, especially the Apgar score in one minute, occurrence of pneumothorax in neonates, oxygen supply given to new born baby within 30 days of life and smoothing term determinants namely number of days hospital stay, platelet counts and gestational age in weeks (non-parametric smoothing terms) are significant factors for neonate’s birth weight and are completely new in the literature.