You are here:
Evaluation of Consumer Product Co-occurrence to Inform Chemical Exposure
Citation:
Tornero-Velez, R., K. Isaacs, M. Nye, AND T. Buckley. Evaluation of Consumer Product Co-occurrence to Inform Chemical Exposure. ISES Annual Meeting, Henderson, NV, Henderson, NV, October 18 - 23, 2015.
Impact/Purpose:
The National Exposure Research Laboratory (NERL) Human Exposure and Atmospheric Sciences Division (HEASD) conducts research in support of EPA mission to protect human health and the environment. HEASD research program supports Goal 1 (Clean Air) and Goal 4 (Healthy People) of EPA strategic plan. More specifically, our division conducts research to characterize the movement of pollutants from the source to contact with humans. Our multidisciplinary research program produces Methods, Measurements, and Models to identify relationships between and characterize processes that link source emissions, environmental concentrations, human exposures, and target-tissue dose. The impact of these tools is improved regulatory programs and policies for EPA.
Description:
Consumer products are an important target of chemical innovation. Used daily for personal hygiene, home care, disinfection and cleaning, consumer products provide a host of benefits, and also an efficient delivery vehicle for a variety of chemicals into our homes and bodies. Although efforts have been made to enumerate the chemicals in consumer products and understand their health risks, little is known of how chemical co-occurrence and risk may be influenced by population purchasing patterns. To address this need, we employed association rules mining, a method with origins in market basket analysis. We applied this method to an anonymized database of consumer product transactions for 60476 households and 200 product categories collected over one year, provided by a major market research firm. To identify co-occurring sets of products, we applied the Apriori algorithm which proceeds by finding higher order combinations as long as all satisfy a minimum ‘support’ (occurrence) threshold. Association rules were then developed and sorted by the ‘lift’ measure, the ratio of observed support to that expected if the individual items occurred independently. Setting minimum support to 20% of homes we observed 10803 rules with lift scores < 1.55. The implication was that these were common purchases with loose associations between goods. Lowering the minimum support to 5% (3024 homes), generated 6.6 million rules with highest lift scores approaching 3.85. Example parings were denture adhesive/denture cleaner, pet accessories/pet treatments, and combined purchases of cosmetics. The strongest associations were of lower order combinations and suggested logical and deliberate parings of consumer goods. In future analysis, structured assemblages of products could be combed for high risk chemicals. We demonstrate an efficient framework to consider consumer product co-occurrence to inform chemical exposure and risk assessment.
