You are here:
Market Basket Data as a Tool to Sharpen Aggregate and Cumulative Chemical Risk Assessment
Citation:
Tornero-Velez, R. Market Basket Data as a Tool to Sharpen Aggregate and Cumulative Chemical Risk Assessment. 2019 SOT Annual Meeting, Baltimore, MD, March 10 - 14, 2019.
Impact/Purpose:
While current chemical testing tends to focus on individual chemicals, the exposures that people actually experience involve mixtures of chemicals. The number of mixtures that can be formed from the thousands of environmental chemicals is enormous, and testing all of them would not be realistic. In recent years, the ongoing revolution in exposure science and analytic chemistry (e.g., non-targeted analysis) is permitting better assessment of exposures to more and more chemicals at lower cost. It appears likely that we will be facing the big data challenge in exposure science in the not very distant future and novel statistical methods are needed for analyzing these data. Collaboration between mixtures toxicologists and exposure scientists has great promise. Exposure science has a very important role to play by 1) determining the combinations of chemicals to which people are actually exposed, reducing the combinatoric problem facing toxicologists; 2) identifying highly correlated exposures that might be better studied using whole mixtures methods than component-based methods; 3) providing information needed by epidemiologists studying exposure to mixtures. A critical problem is understanding the patterns of exposure: for example, which exposures tend to occur together and how does this tendency depend on demographics and other factors? This symposium will bring together exposure scientists and mixtures toxicologists to examine methods for analyzing patterns of co-exposures, apply them to large data sets—such as National Health and Nutrition Examination Survey (NHANES) biomonitoring data and personal care product purchasing databases—and discuss their implications for research on the health effects of exposure to mixtures in toxicology and epidemiology.
Description:
Surveys to acquire information on consumer product use, including use of household cleaning products and personal care products (PCP), can provide valuable information on potential chemical exposures and population habits and practices associated with these products. However, surveys take considerable effort to conduct and may become dated over time as new products emerge and markets change. Market basket data offers a potentially efficient mechanism to capture current product use patterns. Frequent itemset mining and Association Rule Mining were used to analyze a database of PCP purchases for sixty thousand households over a one-year period in 2012. Robust co-occurrence patterns and associations were found for several PCP product categories and were consistent with use surveys. Furthermore, PCP purchase patterns varied by demographics explored (race, education, income, and family composition). We conclude that purchase data can fill a critical data gap when use survey data is absent; and, when coupled with product ingredient databases can inform aggregate and cumulative risk assessment. The views expressed in this abstract are those of the authors and do not necessarily reflect the views or policies of the U.S. EPA.
