Abstract
We report integration of the United States Environmental Protection Agency's (USEPA) United States Environmental Justice Screen (EJSCREEN) database with our Public Health Exposome dataset to interrogate 9232 census blocks to model the complexity of relationships among environmental and socio-demographic variables toward estimating adverse pregnancy outcomes [low birth weight (LBW) and pre-term birth (PTB)] in all Ohio counties. Using a hill-climbing algorithm in R software, we derived a Bayesian network that mapped all controlled associations among all variables available by applying a mapping algorithm. The results revealed 17 environmental and socio-demographic variables that were represented by nodes containing 69 links accounting for a network with 32.85% density and average degree of 9.2 showing the most connected nodes in the center of the model. The model predicts that the socio-economic variables low income, minority, and under age five populations are correlated and associated with the environmental variables; particulate matter (PM(2.5)) level in air, proximity to risk management facilities, and proximity to direct discharges in water are linked to PTB and LBW in 88 Ohio counties. The methodology used to derive significant associations of chemical and non-chemical stressors linked to PTB and LBW from indices of geo-coded environmental neighborhood deprivation serves as a proxy for design of an African-American women's cohort to be recruited in Ohio counties from federally qualified community health centers within the 9232 census blocks. The results have implications for the development of severity scores for endo-phenotypes of resilience based on associations and linkages for different chemical and non-chemical stressors that have been shown to moderate cardio-metabolic disease within a population health context.