Assessing Agreement in Exposure Classification between Proximity-Based Metrics and Air Monitoring Data in Epidemiology Studies of Unconventional Resource Development.

Recent studies of unconventional resource development (URD) and adverse health effects have been limited by distance-based exposure surrogates. Our study compared exposure classifications between air pollutant concentrations and "well activity" (WA) metrics, which are distance-based exposure proxies used in Marcellus-area studies to reflect variation in time and space of residential URD activity. We compiled Pennsylvania air monitoring data for benzene, carbon monoxide, nitrogen dioxide, ozone, fine particulates and sulfur dioxide, and combined this with data on nearly 9000 Pennsylvania wells. We replicated WA calculations using geo-coordinates of monitors to represent residences and compared exposure categories from air measurements and WA at the site of each monitor. There was little agreement between the two methods for the pollutants included in the analysis, with most weighted kappa coefficients between -0.1 and 0.1. The exposure categories agreed for about 25% of the observations and assigned inverse categories 16%-29% of the time, depending on the pollutant. Our results indicate that WA measures did not adequately distinguish categories of air pollutant exposures and employing them in epidemiology studies can result in misclassification of exposure. This underscores the need for more robust exposure assessment in future analyses and cautious interpretation of these existing studies.