Disentangling Socioeconomic Status and Race in Infant Outcomes: A Neural Network Analysis

Abstract

AbstractRace is commonly used as a proxy for multiple features including socioeconomic status. It is critical to dissociate these factors, identify mechanisms that impact infant outcomes, such as birthweight, and direct appropriate interventions and shape public policy. Demographic, socioeconomic, and clinical variables were used to model infant birthweight. Non-linear neural networks better model infant birthweight than linear models (R2 = 0.172 vs. R2 = 0.145, p-value=0.005). In contrast to linear models, non-linear models ranked income, neighborhood disadvantage, and experiences of discrimination higher in importance while modeling birthweight than race. Consistent with extant social science literature, findings suggest race is a linear proxy for non-linear factors. The ability to disentangle and identify the source of effects for socioeconomic status and other social factors that often correlate with race is critical for the ability to appropriately target interventions and public policies designed to improve infant outcomes as well as point out the disparities in these outcomes.