A multilevel modeling analysis of predictors of neonatal hypoglycemia among diabetic and non-diabetic mothers

Neonatal hypoglycemia (NH) is the most frequently encountered metabolic disorder of newborn infants and has been linked to various adverse health outcomes. The risk of developing hypoglycemia among infants of diabetic mothers (IDMs) is even higher. The purpose of this chapter is to construct multilevel models that include individual-level and contextual-level characteristics to predict NH in diabetic and non-diabetic mothers. We analyzed a stratified probability sampling of 3.7 million pediatric hospital discharges using the 2012 Kids' Inpatient Database. Neonates (first 28 days after birth) with a diagnosis of hypoglycemia were included in the study. Groups were compared using the Rao-Scott Chi-Square test for categorical variables. Hierarchical models were constructed to simultaneously examine individual and contextual predictors of NH. As 20% of the variability in the rate of NH is accounted by hospitals, our specification of hierarchical modeling was appropriate to account for the variability among hospitals. We estimated that 67,124 pediatric discharges were associated with NH. Among hypoglycemic neonates, 30% were IDMs. Race, morbidity, small for gestational age, large for gestational age, prematurity, delivery by caesarian section, diabetes status, history of substance/alcohol abuse, elective admission, emergency service use, teaching hospitals, large hospitals, Medicaid insurance, privately owned hospitals, and rural hospitals were significantly associated with NH. IDMs had more than a 5-fold increased risk of developing NH compared to infants of non-diabetic mothers. The Hispanic health paradox was also observed in our study. Future research should focus on the long term clinical significance of NH by including more robust individual and environmental risk factors.