Abstract Alaskan wildfires have major ecological, social, and economic consequences, but associated health impacts remain unexplored. We estimated cardiorespiratory morbidity associated with wildfire smoke (WFS) fine particulate matter with a diameter less than 2.5 μm (PM2.5) in three major population centers (Anchorage, Fairbanks, and the Matanuska‐Susitna Valley) during the 2015–2019 wildfire seasons. To estimate WFS PM2.5, we utilized data from ground‐based monitors and satellite‐based smoke plume estimates. We implemented time‐stratified case‐crossover analyses with single and distributed lag models to estimate the effect of WFS PM2.5 on cardiorespiratory emergency department (ED) visits. On the day of exposure to WFS PM2.5, there was an increased odds of asthma‐related ED visits among 15–65 year olds (OR = 1.12, 95% CI = 1.08, 1.16), people >65 years (OR = 1.15, 95% CI = 1.01, 1.31), among Alaska Native people (OR = 1.16, 95% CI = 1.09, 1.23), and in Anchorage (OR = 1.10, 95% CI = 1.05, 1.15) and Fairbanks (OR = 1.12, 95% CI = 1.07, 1.17). There was an increased risk of heart failure related ED visits for Alaska Native people (Lag Day 5 OR = 1.13, 95% CI = 1.02, 1.25). We found evidence that rural populations may delay seeking care. As the frequency and magnitude of Alaskan wildfires continue to increase due to climate change, understanding the health impacts will be imperative. A nuanced understanding of the effects of WFS on specific demographic and geographic groups facilitates data‐driven public health interventions and fire management protocols that address these adverse health effects.

Background Organophosphate (OP) pesticide exposure is associated with various cancers, neurodegenerative diseases, and respiratory health outcomes. Proximity to agricultural operations and direct occupational contact are hypothesized to be important routes of exposure. Understanding these routes and exploring methods to estimate exposure will improve epidemiological studies, especially among agricultural communities where pesticide exposure due to drift and indirect contact is disproportionally high. To understand factors that influence household environmental exposures to OPs, we collected dust samples from homes in the Central Valley of California. We hypothesized that OP levels would be higher among samples collected during the agricultural spraying season and from homes in which a household member worked in agriculture. Methods Household dust samples were collected using a high-volume small surface sampler during the agricultural spraying (June) and non-spraying (January) seasons from 28 households located within 200 feet of agricultural fields. T-tests and paired t-tests were conducted to assess differences in total OP levels by occupational status and spraying season. Results A total of 50 samples were analyzed for the presence of OPs. Homes in which a household member worked in agriculture had significantly higher OP levels (130.0 parts per billion (ppb), SD 168.5), compared to homes without anyone working in agriculture (28.7 ppb, SD 24.2; p-value=0.001). No statistically significant differences were detected by spraying season (t=0.41, p-value = 0.69). Conclusions Our results demonstrate that detectable levels of OPs are prevalent in households within 200 feet of agricultural fields. Additionally, OPs may persist indoors for extended periods of time. Although the California Pesticide Use Registry indicates that OP application in our study area is seasonal, dust samples had similar levels across seasons. Our work will inform future research by revealing important factors related to routes of exposure to harmful pesticides experienced by agricultural workers and their families.

BackgroundExposure to agricultural pesticides, specifically organophosphates (OP) has been linked to adverse respiratory outcomes in agricultural settings. However, these studies have been based on childhood outcomes, with limited information in relation to respiratory endpoints in adults. Urinary leukotriene E4 (uLTE4) is a cysteinyl leukotriene indicative of respiratory inflammation, and is associated with several respiratory diseases, including asthma. Levels of uLTE4 are known to increase during severe asthma attacks. Though OP pesticides are an inhalation hazard, there is little known about their effect on respiratory inflammation in a population without asthma. We evaluated the relation between OP pesticides found in household dust and urinary LTE4 in a cohort of adults residentially exposed to agricultural pesticides.Methods Dust and urine samples were collected during the agricultural spraying (June) and non-spraying seasons (January) from 28 households located within 200 feet of agricultural fields in the Central Valley of California. We implemented linear regression models to test the association between uLTE4 and OP concentrations, as well as t-tests comparing mean uLTE4 concentrations by occupational status (agricultural v. non-agricultural).ResultsA total of 103 urine samples were analyzed for LTE4 and 50 dust samples for OPs. We did not see an association between OP dust concentration (97.4 ppb, sd = 147.0 ppb) and uLTE4 level, nor was there a difference by season. We did observe a difference in mean uLTE4 level by occupational status (non-agriculture = 1067 pg/ml, agriculture = 1345 pg/ml) but it was not statistically significant (p-value = 0.12).ConclusionOrganophosphates found in house dust were not associated with markers of respiratory inflammation. However, respiratory inflammation may be increased in community members with occupational exposures to pesticides.

PDS 65: Exposure assessment: implications for epidemiology, Exhibition Hall (PDS), Ground floor, August 27, 2019, 1:30 PM - 3:00 PM Background: Though individuals spend ~90% of their time indoors, ambient air pollution data are frequently used in exposure assessment for epidemiologic studies. We sought to understand the relation between indoor and outdoor concentrations of black carbon (BC) and additional housing characteristics associated with indoor BC concentrations among cohort study participants in Denver, CO. Methods: Households from the Healthy Start cohort were selected for monitoring based on diversity in housing type and built environmental characteristics. For one week in spring and summer 2018, households were provided two low-cost air samplers, one for outdoor and one for indoor measurement. Participants completed questionnaires addressing housing specifics such as building type, flooring, and use of heating and cooling systems. BC was measured using transmissometry and weekly mean concentrations were log transformed prior to model fitting. A linear model was fit using all available predictors obtained from surveys as well as outdoor BC concentrations. Model selection was performed using stepwise backwards elimination; model fit was evaluated using AIC. Results: A total of 25 households participated in the study: 11 filters had BC measurements below LOD, and data were available for 39 filters. The average (SD) indoor and outdoor concentrations were 1.1 (0.8) and 1.2 (0.6) μg/m3, respectively. Outdoor BC concentrations were significant predictors of log(indoor BC) in a single-predictor model (β = 0.49, p = 0.02, R2=0.12). The final model included outdoor BC, single family home, tile flooring, and use of electric heater and had an adjusted R2 of 0.5. Conclusion: Outdoor BC and housing characteristics were able to account for ~50% of the variability in indoor BC concentrations measured in Denver, CO homes. In the absence of personal monitoring, household characteristics and time-activity patterns may be used in calibrating ambient air pollution concentrations for personal exposure estimation.

TPS 652: Air pollution exposure modeling 2, Exhibition Hall, Ground floor, August 28, 2019, 3:00 PM - 4:30 PM Background: Denver, Colorado, a large metropolitan area at the base of the Rocky Mountains (elevation: 1609 m), suffers from poor air quality. Traffic is thought to be the predominant pollution source, but agriculture and industry also contribute. Our goal was to develop a land use regression (LUR) model to predict black carbon (BC) concentrations across the study area. Methods: Two campaigns were conducted to capture filter-based integrated black carbon (BC) measurements during spring and summer of 2018. BC was measured using transmissometry; weekly concentrations were calculated for each filter and averaged to monthly means. Land-use characteristics were identified and summarized for buffer distances ranging 100-2500 m. We also included BC data from the local monitoring network and variables indicating the presence of wildfire smoke. BC concentrations were log-transformed prior to model fitting. Covariates were selected using a two-step process. First, a generalized linear model was fit using the least absolute shrinkage and selection operator (LASSO). Second, a linear model was fit using LASSO-selected variables and used to predict monthly concentrations for a 250 m grid across the study area. We validated our LUR using leave-one-out cross-validation (LOOCV). Results: We collected n=512 filters from >50 locations across the study area. Minimum, mean (SD), and max BC concentrations were 0.4, 1.5 (0.8), and 5.6 μg/m3, respectively. As expected, BC concentrations were highest near major roads. For BC, the three most important predictors (based on t-statistics) were average elevation (500 m buffer), annual average daily traffic (100 m buffer), and development intensity (1000 m buffer). The LOOCV R2 value was 0.67. Conclusions: Our LUR reasonably predicted BC concentrations for the two seasons sampled. Additional measurements collected during the fall and winter seasons should improve model fit and allow us to better estimate traffic-related air pollution exposure in a high elevation region.

OPS 10: Wildfires, Room 210, Floor 2, August 26, 2019, 10:30 AM - 12:00 PM Background: Black carbon (BC) has been used to characterize traffic-related air pollution (TRAP) exposure. However, BC has multiple sources, including wildfire smoke (WS). We examined the potential for WS to bias exposure assessments of BC during wildfire events impacting Denver, Colorado. Methods: Weekly integrated filter-based BC samples were collected during spring and summer of 2018. For each filter we calculated a time-weighted average concentration and assessed the length of major roads in a 300-m buffer around the sample location as a comparative measure of TRAP. A filter was considered impacted by WS if the closest network monitor recorded a weekly mean concentration at least one standard deviation (SD) above the 10-year monthly mean for that monitor and if a smoke plume was present within 50 km. We used the Kruskal-Wallis test and Spearman-Rank correlation to compare BC concentrations across roadway-length quartiles. Results: We collected n=552 filters from >50 locations across the region, 29% of which were smoke-impacted. The mean (SD) BC concentration was 1.5 (0.8) μg/m3. Mean BC concentrations were 38% higher for smoke-impacted filters (2.1 μg/m3) than non-impacted filters (1.3 μg/m3, p<0.001). Similarly, the median absolute deviation for smoke-impacted filters (0.5 μg/m3) was higher than for non-impacted filters (0.3 μg/m3), suggesting increased variability in concentrations during wildfire events. For each roadway-length quartile, mean concentrations for smoke-impacted filters were significantly higher than for non-impacted filters: Q1: 2.0 vs. 1.2; Q2: 1.8 vs 1.2; Q3: 2.0 vs 1.3; and Q4: 2.6 vs 1.5 μg/m3 (all p-values<0.001). Spearman-Rank correlations between BC concentration and roadway-length quartile were similar for smoke-impacted filters (r=0.23) and non-impacted filters (r=0.15). Conclusions: Exposure assessments relying on BC as a proxy for TRAP exposures may be biased by wildfire events. Future work will identify the extent to which this bias may affect studies of traffic-related air pollution.

PDS 65: Exposure assessment: implications for epidemiology, Exhibition Hall (PDS), Ground floor, August 27, 2019, 1:30 PM - 3:00 PM Background/Aim: Protection of proprietary poultry operation information can limit geospatial analyses of zoonotic and livestock-associated infectious diseases and complicate assessment of relevant environmental factors. Because comprehensive location data are not available for the United States poultry industry, a novel "hybrid" method for locating operations was developed that leverages both remote sensing spatial accuracy and simulation modeling feasibility. A validation was conducted to compare spatial and distribution accuracies of the hybrid model and a fully-simulated model (Farm Location and Agricultural Production Simulator, i.e. FLAPS). Methods: For spatial accuracy, buffers generated around each ground truth operation were assigned a "Yes/No" outcome for the inclusion of a hybrid or FLAPS poultry operation. Additionally, root mean square errors (RMSE) of the difference in farm count per grid cell were used to compare clustering patterns of the hybrid and FLAPS models to true distributions of poultry operations. Results: The buffer method demonstrated that, for all radii tested, the proportion of true farm buffers that captures at least one hybrid model operation is greater than the proportion that captures a FLAPS operation. Approximately 57% of the 1,000-meter buffers capture hybrid operations, which is significantly greater than FLAPS (28%, p<0.001). RMSE values suggest that, at all grid sizes, the distribution of hybrid model operations is more similar to the true model compared to the FLAPS data. For example, RMSE is 0.93 for the hybrid model and 1.18 for FLAPS for a 3,000-meter grid. Conclusions: This project implemented an innovative approach to validate a novel, methodologically "hybrid" model. Compared to the FLAPS model, the hybrid model demonstrated greater locational accuracy and more closely resembled true poultry operation distributions. Future adaptations of the hybrid method may mitigate the analytical challenges of evaluating environmental risk factors for infectious disease transmission and support control of high priority livestock-associated animal diseases.