Clustering the Predictors of Breast Cancer Mortality

Abstract

Abstract Background Over the last few decades advances have been made in the diagnosis and treatment of breast cancer (BC). Despite progress in diagnosis and treatment, no study has utilized a multidisciplinary data set outside of clinical and laboratory settings to analyze BC mortality. Methods Using U.S. county-level data, we find the environmental, behavioral, and demographic predictors of age-adjusted BC mortality rates, and cluster them into groups by their internal correlation. Principal components were derived to reduce data dimension, and various functional forms were utilized to predict BC mortality. Results We find evidence that environmental contaminants and one’s surrounding living conditions are correlated and significantly associated with BC mortality. Factors associated with poverty (e.g., low literacy, income, and female Medicaid eligibility) and lacking access to mammogram facilities also relate to BC mortality rate. Conclusion This methodology and data set can be used to investigate other chronic diseases, e.g., diabetes and cardiovascular disease. Mortality rates associated with these other diseases can be analyzed just as BC mortality rates were in this paper, and valuable insights discovered.