Socioexposomics of COVID-19 across New Jersey: a comparison of geostatistical and machine learning approaches.

BACKGROUND: Disparities in adverse COVID-19 health outcomes have been associated with multiple social and environmental stressors. However, research is needed to evaluate the consistency and efficiency of methods for studying these associations at local scales. OBJECTIVE: To assess socioexposomic associations with COVID-19 outcomes across New Jersey and evaluate consistency of findings from multiple modeling approaches. METHODS: We retrieved data for COVID-19 cases and deaths for the 565 municipalities of New Jersey up to the end of the first phase of the pandemic, and calculated mortality rates with and without long-term-care (LTC) facility deaths. We considered 84 spatially heterogeneous environmental, demographic and socioeconomic factors from publicly available databases, including air pollution, proximity to industrial sites/facilities, transportation-related noise, occupation and commuting, neighborhood and housing characteristics, age structure, racial/ethnic composition, poverty, etc. Six geostatistical models (Poisson/Negative-Binomial regression, Poison/Negative-Binomial mixed effect model, Poisson/Negative-Binomial Bersag-York-Mollie spatial model) and two Machine Learning (ML) methods (Random Forest, Extreme Gradient Boosting) were implemented to assess association patterns. The Shapley effects plot was established for explainable ML and change of support validation was introduced to compare performances of different approaches. RESULTS: We found robust positive associations of COVID-19 mortality with historic exposures to NO2, population density, percentage of minority and below high school education, and other social and environmental factors. Exclusion of LTC deaths does not significantly affect correlations for most factors but findings can be substantially influenced by model structures and assumptions. The best performing geostatistical models involved flexible structures representing data variations. ML methods captured association patterns consistent with the best performing geostatistical models, and furthermore detected consistent nonlinear associations not captured by geostatistical models. SIGNIFICANCE: The findings of this work improve the understanding of how social and environmental disparities impacted COVID-19 outcomes across New Jersey.

Modeling past and future spatiotemporal distributions of airborne allergenic pollen across the contiguous United States.

Exposures to airborne allergenic pollen have been increasing under the influence of changing climate. A modeling system incorporating pollen emissions and atmospheric transport and fate processes has been developed and applied to simulate spatiotemporal distributions of two major aeroallergens, oak and ragweed pollens, across the contiguous United States (CONUS) for both historical (year 2004) and future (year 2047) conditions. The transport and fate of pollen presented here is simulated using our adapted version of the Community Multiscale Air Quality (CMAQ) model. Model performance was evaluated using observed pollen counts at monitor stations across the CONUS for 2004. Our analysis shows that there is encouraging consistency between observed seasonal mean concentrations and corresponding simulated seasonal mean concentrations (oak: Pearson = 0.35, ragweed: Pearson = 0.40), and that the model was able to capture the statistical patterns of observed pollen concentration distributions in 2004 for most of the pollen monitoring stations. Simulation of pollen levels for a future year (2047) considered conditions corresponding to the RCP8.5 scenario. Modeling results show substantial regional variability both in the magnitude and directionality of changes in pollen metrics. Ragweed pollen season is estimated to start earlier and last longer for all nine climate regions of the CONUS, with increasing average pollen concentrations in most regions. The timing and magnitude of oak pollen season vary across the nine climate regions, with the largest increases in pollen concentrations expected in the Northeast region.

Flexible Bayesian Ensemble Machine Learning Framework for Predicting Local Ozone Concentrations.

3D-grid-based chemical transport models, such as the Community Multiscale Air Quality (CMAQ) modeling system, have been widely used for predicting concentrations of ambient air pollutants. However, typical horizontal resolutions of nationwide CMAQ simulations (12 × 12 km2) cannot capture local-scale gradients for accurately assessing human exposures and environmental justice disparities. In this study, a Bayesian ensemble machine learning (BEML) framework, which integrates 13 learning algorithms, was developed for downscaling CMAQ estimates of ozone daily maximum 8 h averages to the census tract level, across the contiguous US, and was demonstrated for 2011. Three-stage hyperparameter tuning and targeted validations were designed to ensure the ensemble model's ability to interpolate, extrapolate, and capture concentration peaks. The Shapley value metric from coalitional game theory was applied to interpret the drivers of subgrid gradients. The flexibility (transferability) of the 2011-trained BEML model was further tested by evaluating its ability to estimate fine-scale concentrations for other years (2012-2017) without retraining. To demonstrate the feasibility of using the BEML approach to strictly "data-limited" situations, the model was applied to downscale CMAQ outputs for a future-year scenario-based simulation that considers effects of variations in meteorology associated with climate change.

A hybrid approach to predict daily NO2 concentrations at city block scale.

Estimating the ambient concentration of nitrogen dioxide (NO2) is challenging because NO2 generated by local fossil fuel combustion varies greatly in concentration across space and time. This study demonstrates an integrated hybrid approach combining dispersion modeling and land use regression (LUR) to predict daily NO2 concentrations at a high spatial resolution (e.g., 50 m) in the New York tri-state area. The daily concentration of traffic-related NO2 was estimated at the Environmental Protection Agency's NO2 monitoring sites in the study area for the years 2015-2017, using the Research LINE source (R-LINE) model with inputs of traffic data provided by the Highway Performance and Management System and meteorological data provided by the NOAA Integrated Surface Database. We used the R-LINE-predicted daily concentrations of NO2 to build mixed-effects regression models, including additional variables representing land use features, geographic characteristics, weather, and other predictors. The mixed model was selected by the Elastic Net method. Each model's performance was evaluated using the out-of-sample coefficient of determination (R2) and the square root of mean squared error (RMSE) from ten-fold cross-validation (CV). The mixed model showed a good prediction performance (CV R2: 0.75-0.79, RMSE: 3.9-4.0 ppb). R-LINE outputs improved the overall, spatial, and temporal CV R2 by 10.0%, 18.9% and 7.7% respectively. Given the output of R-LINE is point-based and has a flexible spatial resolution, this hybrid approach allows prediction of daily NO2 at an extremely high spatial resolution such as city blocks.

Comparison of Machine Learning and Land Use Regression for fine scale spatiotemporal estimation of ambient air pollution: Modeling ozone concentrations across the contiguous United States.

BACKGROUND: Spatial linear Land-Use Regression (LUR) is commonly used for long-term modeling of air pollution in support of exposure and epidemiological assessments. Machine Learning (ML) methods in conjunction with spatiotemporal modeling can provide more flexible exposure-relevant metrics and have been studied using different model structures. There is however a lack of comparisons of methods available within these two modeling frameworks, that can guide model/algorithm selection in air quality epidemiology. OBJECTIVE: The present study compares thirteen algorithms for spatial/spatiotemporal modeling applied for daily maxima of 8-hour running averages of ambient ozone concentrations at spatial resolutions corresponding to census tracts, to support estimation of annual ozone design values across the contiguous US. These algorithms were selected from nine representative categories and trained using predictors that included chemistry-transport model predictions, meteorological factors, land use and land cover, and stationary and mobile emissions. METHODS: To obtain the best predictive performance, model structures were optimized through a repeated coarse/fine grid search with expert knowledge. Six target-oriented validation strategies were used to prevent overfitting and avoid over-optimistic model evaluation results. In order to take full advantage of the power of different algorithms, we introduced tuning sample weights in spatiotemporal modeling to ensure predictive accuracy of peak concentrations, that is crucial for exposure assessments. In spatial modeling, four interpretation and visualization tools were introduced to explain predictions from different algorithms. RESULTS: Nonlinear ML methods achieved higher prediction accuracy than linear LUR, and the improvements were more significant for spatiotemporal modeling (nearly 10%-40% decrease of predicted RMSE). By tuning the sample weights, spatiotemporal models can predict concentrations used to calculate ozone design values that are comparable or even better than spatial models (nearly 30% decrease of cross-validated RMSE). We visualized the underlying nonlinear relationships, heterogeneous associations and complex interactions from the two best performing ML algorithms, i.e., Random Forest and Extreme Gradient Boosting, and found that the complex patterns were relatively less significant with respect to model accuracy for spatial modeling. CONCLUSION: Machine Learning can provide estimates that are actually more interpretable and practical than linear regression to improve accuracy in modeling human exposures. A careful design of hyperparameter tuning and flexible data splitting and validations is crucial to obtain reliable and stable results. Desirable/successful nonlinear models are expected to capture similar nonlinear patterns and interactions using different ML algorithms.

Development of a semi-mechanistic allergenic pollen emission model.

Modeling pollen emission processes is crucial for studying the spatiotemporal distributions of airborne allergenic pollen. A semi-mechanistic emission model was developed based on mass balance of pollen grain fluxes in the surroundings of allergenic plants. The emission model considers direct emission and resuspension and accounts for influences of temperature, wind velocity, and relative humidity. Modules of this emission model have been developed and parameterized with multiple years of pollen count observations to provide pollen season onset and duration, hourly flowering likelihood, and vegetation coverage for oak and ragweed, as two examples. The simulated spatiotemporal pattern of pollen emissions generally follows the corresponding pattern of area coverage of allergenic plants and diurnal pattern of hourly flowering likelihood. It is found that oak pollen emissions start from the Southern part of the Contiguous United States (CONUS) in March and then shift gradually toward the Northern CONUS, with a maximum emission flux of 5.8 × 106 pollen/(m2 h). On the other hand, ragweed pollen emissions start from the Northern CONUS in August and then shift gradually toward the Southern CONUS. The mean ragweed emission flux during August to September can increase up to 2.4 × 106 pollen/(m2 h). This emission model is robust with respect to the input parameters for oak and ragweed. Qualitative evaluations of the model performance indicated that the simulated pollen emission is strongly correlated with the plant coverages and observed pollen counts. This model could also be applied to other pollen species given the relevant parameters.

Enhanced exposure assessment and genome-wide DNA methylation in World Trade Center disaster responders.

DNA methylation has emerged as a promising target linking environmental exposures and cancer. The World Trade Center (WTC) responders sustained exposures to potential carcinogens, resulting in an increased risk of cancer. Previous studies of cancer risk in WTC-exposed responders were limited by the deficiency in quantitative and individual information on exposure to carcinogens. The current study introduces a new exposure-ranking index (ERI) for estimating cancer-related acute and chronic exposures, which aimed to improve the ability of future analyses to estimate cancer risk. An epigenome-wide association study based on DNA methylation and a weighted gene co-expression network analysis were carried out to identify cytosine-phosphate-guanosine (CpG) sites, modules of correlated CpG sites, and biological pathways associated with the new ERI. Methylation was profiled on blood samples using Illumina 450K Beadchip. No significant epigenome-wide association was found for ERI at a false discovery rate of 0.05. Several cancer-related pathways emerged in pathway analyses for the top ranking genes from epigenome-wide association study as well as enriched module from the weighted gene co-expression network analysis. The current study was the first DNA methylation study that aimed to identify methylation signature for cancer-related exposure in the WTC population. No CpG sites survived multiple testings adjustment. However, enriched gene sets involved in cancer, were identified in both acute and chronic ERIs, supporting the view that multiple genes play a role in this complex exposure.

Per and polyfluoroalkyl substances (PFAS) blood levels after contamination of a community water supply and comparison with 2013-2014 NHANES.

INTRODUCTION: Per and polyfluoroalkyl substances (PFAS), including perfluorononanoic acid (PFNA) and perfluorooctanoic acid (PFOA), were detected in the community water supply of Paulsboro New Jersey in 2009. METHODS: A cross-sectional study enrolled 192 claimants from a class-action lawsuit, not affiliated with this study, who had been awarded a blood test for 13 PFAS. Study participants provided their blood test results and completed a survey about demographics; 105 participants also completed a health survey. Geometric means, 25th, 50th, 75th, and 95th percentiles of exposure of PFNA blood serum concentrations were compared to that of the 2013-2014 NHANES, adjusted for reporting level. Associations between PFNA, PFOA, PFOS, and PFHxS and self-reported health outcomes were assessed using logistic regression. RESULTS: PFNA serum levels were 285% higher in Paulsboro compared with U.S. residents. PFNA serum levels were higher among older compared with younger, and male compared to female, Paulsboro residents. After adjustment for potential confounding, there was a significant association between increased serum PFNA levels and self-reported high cholesterol (OR: 1.15, 95% CI: 1.02, 1.29). DISCUSSION/CONCLUSION: Further investigation into possible health effects of PFAS exposure in Paulsboro and other community settings is warranted. Since exposure has ceased, toxicokinetics of PFAS elimination should be explored.

Predicting Onset and Duration of Airborne Allergenic Pollen Season in the United States.

Allergenic pollen is one of the main triggers of Allergic Airway Disease (AAD) affecting 5% to 30% of the population in industrialized countries. A modeling framework has been developed using correlation and collinearity analyses, simulated annealing, and stepwise regression based on nationwide observations of airborne pollen counts and climatic factors to predict the onsets and durations of allergenic pollen seasons of representative trees, weeds and grass in the contiguous United States. Main factors considered are monthly, seasonal and annual mean temperatures and accumulative precipitations, latitude, elevation, Growing Degree Day (GDD), Frost Free Day (FFD), Start Date (SD) and Season Length (SL) in the previous year. The estimated mean SD and SL for birch (Betula), oak (Quercus), ragweed (Ambrosia), mugwort (Artemisia) and grass (Poaceae) pollen season in 1994-2010 are mostly within 0 to 6 days of the corresponding observations for the majority of the National Allergy Bureau (NAB) monitoring stations across the contiguous US. The simulated spatially resolved maps for onset and duration of allergenic pollen season in the contiguous US are consistent with the long term observations.

Allergenic pollen season variations in the past two decades under changing climate in the United States.

Many diseases are linked with climate trends and variations. In particular, climate change is expected to alter the spatiotemporal dynamics of allergenic airborne pollen and potentially increase occurrence of allergic airway disease. Understanding the spatiotemporal patterns of changes in pollen season timing and levels is thus important in assessing climate impacts on aerobiology and allergy caused by allergenic airborne pollen. Here, we describe the spatiotemporal patterns of changes in the seasonal timing and levels of allergenic airborne pollen for multiple taxa in different climate regions at a continental scale. The allergenic pollen seasons of representative trees, weeds and grass during the past decade (2001-2010) across the contiguous United States have been observed to start 3.0 [95% Confidence Interval (CI), 1.1-4.9] days earlier on average than in the 1990s (1994-2000). The average peak value and annual total of daily counted airborne pollen have increased by 42.4% (95% CI, 21.9-62.9%) and 46.0% (95% CI, 21.5-70.5%), respectively. Changes of pollen season timing and airborne levels depend on latitude, and are associated with changes of growing degree days, frost free days, and precipitation. These changes are likely due to recent climate change and particularly the enhanced warming and precipitation at higher latitudes in the contiguous United States.

Manganese concentrations in soil and settled dust in an area with historic ferroalloy production.

Ferroalloy production can release a number of metals into the environment, of which manganese (Mn) is of major concern. Other elements include lead, iron, zinc, copper, chromium, and cadmium. Mn exposure derived from settled dust and suspended aerosols can cause a variety of adverse neurological effects to chronically exposed individuals. To better estimate the current levels of exposure, this study quantified the metal levels in dust collected inside homes (n=85), outside homes (n=81), in attics (n=6), and in surface soil (n=252) in an area with historic ferroalloy production. Metals contained in indoor and outdoor dust samples were quantified using inductively coupled plasma optical emission spectroscopy, whereas attic and soil measurements were made with a X-ray fluorescence instrument. Mean Mn concentrations in soil (4600 µg/g) and indoor dust (870 µg/g) collected within 0.5 km of a plant exceeded levels previously found in suburban and urban areas, but did decrease outside 1.0 km to the upper end of background concentrations. Mn concentrations in attic dust were ~120 times larger than other indoor dust levels, consistent with historical emissions that yielded high airborne concentrations in the region. Considering the potential health effects that are associated with chronic Mn inhalation and ingestion exposure, remediation of soil near the plants and frequent, on-going hygiene indoors may decrease residential exposure and the likelihood of adverse health effects.

Modeling Population Exposures to Silver Nanoparticles Present in Consumer Products.

Exposures of the general population to manufactured nanoparticles (MNPs) are expected to keep rising due to increasing use of MNPs in common consumer products (PEN 2014). The present study focuses on characterizing ambient and indoor population exposures to silver MNPs (nAg). For situations where detailed, case-specific exposure-related data are not available, as in the present study, a novel tiered modeling system, Prioritization/Ranking of Toxic Exposures with GIS (Geographic Information System) Extension (PRoTEGE), has been developed: it employs a product Life Cycle Analysis (LCA) approach coupled with basic human Life Stage Analysis (LSA) to characterize potential exposures to chemicals of current and emerging concern. The PRoTEGE system has been implemented for ambient and indoor environments, utilizing available MNP production, usage, and properties databases, along with laboratory measurements of potential personal exposures from consumer spray products containing nAg. Modeling of environmental and microenvironmental levels of MNPs employs Probabilistic Material Flow Analysis combined with product LCA to account for releases during manufacturing, transport, usage, disposal, etc. Human exposure and dose characterization further employs screening Microenvironmental Modeling and Intake Fraction methods combined with LSA for potentially exposed populations, to assess differences associated with gender, age, and demographics. Population distributions of intakes, estimated using the PRoTEGE framework, are consistent with published individual-based intake estimates, demonstrating that PRoTEGE is capable of capturing realistic exposure scenarios for the US population. Distributions of intakes are also used to calculate biologically-relevant population distributions of uptakes and target tissue doses through human airway dosimetry modeling that takes into account product MNP size distributions and age-relevant physiological parameters.

Influence of cobalamin on arsenic metabolism in Bangladesh.

BACKGROUND: Arsenic is a carcinogen to which 35 million people in Bangladesh are chronically exposed. The enzymatic transfer of methyl groups to inorganic As (iAs) generates monomethylarsonic (MMA) and dimethylarsinic acids (DMA) and facilitates urinary As (uAs) elimination. This process is dependent on one-carbon metabolism, a pathway in which folate and cobalamin have essential roles in the recruitment and transfer of methyl groups. Although DMA(V) is the least toxic metabolite, increasing evidence suggests that MMA(III) may be the most cytotoxic and genotoxic As intermediary metabolite. OBJECTIVE: We examined the associations between plasma cobalamin and uAs metabolites. METHODS: We conducted a cross-sectional study of 778 Bangladeshi adults in which we over-sampled cobalamin-deficient participants. Participants provided blood samples for the measurement of plasma cobalamin and urine specimens for As measurements. RESULTS: Cobalamin was inversely associated with the proportion of total uAs excreted as iAs (%iAs) [unstandardized regression coefficient (b) = -0.10; 95% confidence interval (CI), -0.17 to -0.02; p = 0.01] and positively associated with %MMA (b = 0.12; 95% CI, 0.05 to 0.20; p = 0.001). Both of these associations were stronger among folate-sufficient participants (%iAs: b = -0.17; 95% CI, -0.30 to -0.03; p = 0.02. %MMA: b = 0.20; 95% CI, 0.11 to 0.30; p < 0.0001), and the differences by folate status were statistically significant. CONCLUSIONS: In this group of Bangladeshi adults, cobalamin appeared to facilitate the first As methylation step among folate-sufficient individuals. Given the toxicity of MMA(III), our findings suggest that in contrast to folate, cobalamin may not favorably influence As metabolism.