Using Latent Class Modeling to Jointly Characterize Economic Stress and Multipollutant Exposure.

Published

Journal Article

BACKGROUND: Work is needed to better understand how joint exposure to environmental and economic factors influence cancer. We hypothesize that environmental exposures vary with socioeconomic status (SES) and urban/rural locations, and areas with minority populations coincide with high economic disadvantage and pollution. METHODS: To model joint exposure to pollution and SES, we develop a latent class mixture model (LCMM) with three latent variables (SES Advantage, SES Disadvantage, and Air Pollution) and compare the LCMM fit with K-means clustering. We ran an ANOVA to test for high exposure levels in non-Hispanic black populations. The analysis is at the census tract level for the state of North Carolina. RESULTS: The LCMM was a better and more nuanced fit to the data than K-means clustering. Our LCMM had two sublevels (low, high) within each latent class. The worst levels of exposure (high SES disadvantage, low SES advantage, high pollution) are found in 22% of census tracts, while the best levels (low SES disadvantage, high SES advantage, low pollution) are found in 5.7%. Overall, 34.1% of the census tracts exhibit high disadvantage, 66.3% have low advantage, and 59.2% have high mixtures of toxic pollutants. Areas with higher SES disadvantage had significantly higher non-Hispanic black population density (NHBPD; P < 0.001), and NHBPD was higher in areas with higher pollution (P < 0.001). CONCLUSIONS: Joint exposure to air toxins and SES varies with rural/urban location and coincides with minority populations. IMPACT: Our model can be extended to provide a holistic modeling framework for estimating disparities in cancer survival.See all articles in this CEBP Focus section, "Environmental Carcinogenesis: Pathways to Prevention."

Full Text

Duke Authors

Cited Authors

  • Larsen, A; Kolpacoff, V; McCormack, K; Seewaldt, V; Hyslop, T

Published Date

  • October 2020

Published In

Volume / Issue

  • 29 / 10

Start / End Page

  • 1940 - 1948

PubMed ID

  • 32856601

Pubmed Central ID

  • 32856601

Electronic International Standard Serial Number (EISSN)

  • 1538-7755

Digital Object Identifier (DOI)

  • 10.1158/1055-9965.EPI-19-1365

Language

  • eng

Conference Location

  • United States