Modeling community-level effects on preterm birth.

We demonstrate modeling of community-level socioeconomic influences on risk of preterm birth (< 37 weeks gestation) in the Pregnancy, Infection, and Nutrition (PIN) Study.Community-level information from the US Census was linked to 930 White and 817 African-American (Black) participants from a prospective cohort in central North Carolina through geocoded addresses, providing 123 census tracts with community-level and individual-level data for multi-level statistical analyses.Preterm delivery was experienced by 12.1% of Black and 10.4% of White participants. No appreciable aggregation of risk by community was discernable for White women. For Black women, random-coefficient logistic regression tract-specific preterm prevalence estimates ranged from 10.1% to 14.5%, "shrunk" from observed prevalences of 0% to 100%. Adding tract-level variables to the model representing median splits for household income and percent of single women heads of households with dependents, adjusting for individual-level maternal age and household income, accounted for much of the remaining between-tracts variation.Residing in a wealthier tract (> $30,000/year median income) was associated with reduced risk for Black women, adjusted OR = 0.59 (95% CI: 0.36, 0.96). The estimated conditional effect of lower community prevalence of female headed households was OR = 0.71 (95% CI: 0.43, 1.17).

Duke Scholars

Author Amy H Herring Statistical Science