Evaluation of confounding by socioeconomic status (SES) on the association between PM2.5 and mortality in the U.S Medicare population

Abstract

Background/Aim:Socioeconomic status (SES) is a complex construct that is influenced by many individual- and area-level factors encompassing material, human, and social capital. SES has been suggested to be an important confounder to consider in environmental epidemiological studies. However, many previous studies were limited in their ability to adequately capture the complexity of SES and likely have resulted in residual confounding. The objective of our study was to assess the potential for utilizing publicly available SES data to more accurately describe confounding by SES on the association between PM2.5 and mortality in the Medicare population in the U.S. Methods: Monthly mortality counts and PM2.5 concentrations in the contiguous U.S. for 2000-2012 were obtained at the ZIP code level. We also searched for publicly available SES data beyond what is typically used (such as income, education, and race). The base model was only adjusted for age and sex, while the multivariable models were adjusted for multiple SES indicators by controlling for individual SES variables as well as using Principal Component Analysis (PCA) approach. Results:29 publicly available SES data were collected from the Census Bureau and the Internal Revenue Service (IRS). Some examples of the SES indicators used in our study include income inequality (Gini index), non-salary wealth, % occupation in management/business/science, % moved from abroad, and % household with > 1 person/room. We found that the increased risks of mortality associated with PM2.5 in our Medicare population were slightly confounded by SES with different methods of controlling for SES resulting in different magnitudes of confounding effect. Conclusion:These results warrant further investigation examining the best approach for using multiple SES data to accurately capture the confounding effect of SES. In addition, our study demonstrated the possibility of utilizing publicly available data to examine the complex confounding effect of SES in environmental epidemiological studies.