ASSESSING BIAS IN PERSONAL EXPOSURE ESTIMATES WHEN INDOOR AIR QUALITY IS IGNORED: A CASE STUDY OF EASTERN NORTH CAROLINA AND RALEIGH/DURHAM

Abstract

Exposure to outdoor air pollution can cause adverse health issues. Indoor air pollution can be more dangerous due to indoor activities such as cooking, painting, smoking, and heating systems, and can worsen the burden associated with air pollution. Neglecting indoor air quality in an exposure assessment may generate bias and erroneous conclusions regarding the association between exposure and health outcomes. Accurately assessing levels of exposure to air pollution helps advance our understanding of the health effects associated with air pollution and identify effective strategies to address public health concerns. Previous exposure studies have utilized outdoor air quality to assess individuals’ exposure to air pollution in their residential locations. These studies neglect the significant amount of time people spend indoors and other activity places. Some other studies assessed exposure to air pollution by measured only indoor air quality inside homes, but they overlook outdoor and other indoor environments where people are exposed to air pollution. Therefore, it remains unknown whether neglecting indoor air quality would generate negligible or non-negligible bias. Likewise, environmental justice studies have also focused only on outdoor air quality to understand exposure disparities among social groups. It remains unknown whether the degree of disparities in exposure becomes larger when indoor environments are considered compared to when only outdoor air quality is considered in exposure assessments. This study aims to assess bias in personal exposure by comparing estimates derived from two different air pollution datasets: outdoor air quality data and data collected in both indoor and outdoor environments. It also investigates whether exposure disparity between different social groups would be more prominent when indoor air quality is considered in exposure assessments. For data that contains both indoor and outdoor pollution measurements, this study used 3-day mobile sensing air quality data of 100 research participants living in eastern North Carolina (ENC), which was collected using GeoAir2.0 portable monitors for a previous pilot project conducted by Park’s research team. The GeoAir2.0 data includes GPS location data as well as particulate matter (PM) measurements in a 1-minute interval. For outdoor air quality data, publicly available PM2.5 data measured every minute were downloaded from 213 PurpleAir monitors across ENC for the corresponding times the GeoAir data were collected. Using a geostatistical method, a Python algorithm was developed to estimate PM2.5 from PurpleAir data for 129,600 minutes over 90 unique days of the entire data collection period. Participants’ PM2.5 values were averaged for each dataset to obtain GeoAir2.0- and PurpleAir-based exposure estimates. Next, paired-sample t-tests were performed to examine whether there was a significant difference between the two exposure estimates. Lastly, to assess the degree of gender and economic disparities in exposure among different income groups, Welch’s t-tests and analyses of variance (ANOVA) were performed. Game-Howell post hoc tests were used to assess whether the difference in exposure estimates among pairs of the income groups was significant. This study identified that the PurpleAir-based exposure estimates were consistently higher than the GeoAir-based estimates in overall exposure, activity places, and days of the week. The paired-sample t-test result revealed a statistically significant difference between the two estimates (t = 5.94; p < 0.001). This indicates that PurpleAir estimates overestimate the actual exposure to air pollution. Exposure levels in different microenvironments also showed significant variations between the two estimates, with home exposure generating the most significant difference (t = 5.74; p < 0.001), revealing that PurpleAir data overestimated the level of exposure in indoors when outdoor air pollution concentration is high and underestimates indoor air quality when outdoor air quality is low. Also, the socioeconomic disparities in exposure estimates were evident among the three income groups when indoor air quality was integrated into exposure assessment compared to when only outdoor air quality was utilized. The study emphasizes the significance of integrating indoor air quality data in exposure assessments to mitigate bias. It also underscores the importance of personal air sampling data in environmental justice studies to provide a more accurate representation of exposure disparities among different social groups.