A neighborhood and spatial measure of racial isolation applied to birthweight

Research has shown that racial residential segregation is associated with poor birth outcomes among blacks in the United States. As the most conceptually sound dimension of segregation to capture where and under what conditions mothers live, racial isolation has been widely applied in these studies. Traditional measures of racial isolation aggregate population counts at the neighborhood level over a large geographical area without accounting for population composition in adjacent neighborhoods. This approach, however, is limiting because it precludes studying the effects of racial isolation within a given region, assumes mothers within a region face the same risk, and over- or underestimates racial isolation by ignoring spatial relationships among neighborhoods. To better characterize racial isolation within the framework of modeling birthweight, we developed a measure that allows within-region variation, reduces dependence on neighborhood boundaries, and is applicable at refined geographical scales. Using the North Carolina Detailed Birth Record (1999-2001), we use multiple linear regression to model the effect of racial isolation on birthweight, adjusting for maternal age, education, and parity. To explore whether birthweight is spatially patterned, we employ a Conditional Autoregressive (CAR) model to adjust for residual spatial autocorrelation. We show that racial isolation at the neighborhood level is associated with a decrease in birthweight of similar magnitude as the presence of maternal risk factors, including whether a mother has completed high school or is over 40 years of age. There is evidence that birthweight is a spatial process, which suggests the importance of accounting for spatial dependence in estimation.