Predicting personal PAH exposure using high dimensional questionnaire and wristband data.

Background: Polycyclic aromatic hydrocarbons (PAHs) are a class of pervasive environmental pollutants with a variety of known health effects. While significant work has been completed to estimate personal exposure to PAHs, less has been done to identify sources of these exposures. Comprehensive characterization of reported sources of personal PAH exposure is a critical step to more easily identify individuals at risk of high levels of exposure and for developing targeted interventions based on source of exposure.

Objective: In this study, we leverage data from a New York (NY)-based birth cohort to identify personal characteristics or behaviors associated with personal PAH exposure and develop models for the prediction of PAH exposure.

Methods: We quantified 61 PAHs measured using silicone wristband samplers in association with 75 questionnaire variables from 177 pregnant individuals. We evaluated univariate associations between each compound and questionnaire variable, conducted regression tree analysis for each PAH compound and completed a principal component analysis of for each participant's entire PAH exposure profile to determine the predictors of PAH levels.

Results: Regression tree analyses of individual compounds and exposure mixture identified income, time spent outdoors, maternal age, country of birth, transportation type, and season as the variables most frequently predictive of exposure.