Childhood cancer constitutes a major cause of death in children. In a recent study of the Georgia Cancer Registry, joint exposures to environmental and social/behavioral stressors were associated with spatial clustering of lymphomas and reticuloendothelial neoplasms among the 159 counties in Georgia, USA. The present study aims to further investigate these associations on a more granular level. Bayesian Poisson and zero-inflated Poisson regression models with spatial and non-spatial variance structures were utilized to investigate whether county-specific cancer patterns may be explained by single or combinations of social stressors and ambient air pollution while adjusting for confounding and accounting for overfitting using differences in expected log predictive densities. While we did not find associations between lymphoma rates and social variables, air pollution, or their interactions, our proposed analysis workflow can serve as a blueprint for future studies investigating dependencies in regression models that feature combinations of unobserved and observed dependency structures.
Keywords: Air pollution; Bayesian regression; Cross-validation; Pediatric cancer; Socioeconomic factors; Spatial correlation.
Copyright © 2024 The Authors. Published by Elsevier Ltd.. All rights reserved.