key: cord-1002529-iwc7jerp
authors: Lipsitt, Jonah; Chan-Golston, Alec M.; Liu, Jonathan; Su, Jason; Zhu, Yifang; Jerrett, Michael
title: Spatial Analysis of COVID-19 and Traffic-related Air Pollution in Los Angeles()
date: 2021-03-22
journal: Environ Int
DOI: 10.1016/j.envint.2021.106531
sha: f8cd7d1f88a796d9f0aca1efc091bfdcecc95865
doc_id: 1002529
cord_uid: iwc7jerp

nan

As of February 23 rd , 2021, more than 113 million people worldwide have been diagnosed with COVID-19, resulting in more than 2.5 million deaths (World Health Organization, 2021) . Extensive investigation has been conducted on the etiology of COVID-19, yet researchers are still determining how exposure risk factors may influence COVID-19 incidence and mortality. Recent evidence from China, Italy, England, and the United States suggest that exposure to air pollution may play a role in incidence and deaths (Brandt et al., 2020; Coker et al., 2020; Li et al., 2020; Lippi et al., 2020; Travaglio et al., 2021; B. Wang et al., 2020; X. Wu et al., 2020; Zhang et al., 2020; Zhu et al., 2020) . These findings are consistent with prior research suggesting that air pollution, including traffic-related air pollution (TRAP), is associated with many respiratory morbidities (e.g., asthma, chronic pulmonary disease, lung cancer, and respiratory tract infections) Dales et al., 2008; Franklin et al., 2015; Sydbom et al., 2001) , hospitalizations (Neupane et al., 2010) , all-cause mortality (Beelen et al., likely to live in areas where there is greater air pollution (Ailshire and García, 2018; Collaco et al., 2020; Gaither et al., 2019) .

Building on past research demonstrating an association between Severe Acute Respiratory Syndrome (SARS) and air pollution (Cui et al., 2003) , reported associations between county-level COVID-19 mortality rates in 3,089 counties through June 2020 and long-term average (from years 2000 to 2016) PM 2.5 concentration across the United States. They reported that each 1 µg/m -3 increase in PM 2.5 concentration was associated with an 11% increase in COVID-19 mortality rate (95% CI: 6% -17%). In another recent study of COVID-19 in 3,122 U.S. counties through July 2020, researchers found an increase in interquartile range (IQR) of 4.6 ppb of NO 2 to be associated with a 16.2% (95% CI: 8.7%, 24.0%) increase in mortality rate and an 11.3% (95% CI: 4.9%, 18.2%) increase in case-fatality rate (Liang et al., 2020) . In both studies, exposure estimates were based on concentrations at the county level and, therefore, could not account for variation in air pollution observable on a smaller scale within cities (Kulhánová et al., 2018; Wu et al., 2019) . In addition, the quality and comparability of COVID-19 health outcome information may vary considerably across U.S. counties as reporting protocols may differ among jurisdictions at the local, state, and national levels (Bergman et al., 2020; Bialek et al., 2020) , which may lead to case ascertainment bias.

Los Angeles became one of the only metropolitan cities globally to publicly report neighborhoodlevel COVID-19 cases in March 2020 and mortality in June 2020 (LACDPH, 2020) . These data afforded the opportunity to conduct spatial modeling for a large population with a smaller geographical area neighborhood unit of analysis. These smaller geographic areas allow for more accurate pollution exposure estimates than the county-level studies above. Los Angeles has a wide range of air pollution exposure levels with which to investigate intra-urban relationships with COVID-19. Furthermore, because the Los Angeles County Department of Public Health (LACDPH) governs all health statistics, Los Angeles

County is likely to have consistent health reporting practices. This diminishes the possibility of case ascertainment bias that may have been present in the national studies comparing among more than 3,000 counties.

Here we aim to analyze the relationship of air pollution and COVID-19 case incidence, mortality, and case-fatality rates in neighborhoods of Los Angeles County, using high-resolution exposure models.

We focus on NO 2 because this gaseous pollutant serves as a marker for traffic pollution, which displays substantial intra-urban variation over small areas in Los Angeles and elsewhere (Su et al., 2009 (Su et al., , 2020 Zeldovich, 2015) . In California, where this study is situated, 62% of NO x emissions come from mobile sources such as vehicle traffic (Almaraz et al., 2018) .

This study is situated in Los Angeles (LA) County. In 2019, LA County had a population size of 10,039,107 and was diverse in its racial, ethnic, and socioeconomic composition (U.S. Census Bureau, 2020). For example, 51% are White, 48% are Latinx, 15% are Asian, and 8.3% are Black. The median income of LA County is $64,251 USD. LA County is spread across a large geographic area of 4,057 square miles or 10,508 square kilometers. The sprawling landscape induces high levels of travel by automobile and attendant traffic-related air pollution (TRAP) (Su et al., 2009 ). In addition, the presence of two major seaports and associated goods movement infrastructure creates additional emissions from diesel vehicles (Kozawa et al., 2009; Su et al., 2016 Su et al., , 2020 . The first case of COVID-19 in California was identified on January 26 th , 2020 (LACDPH, 2020; Los Angeles Times, 2020) , and the first communityacquired case in the United States was confirmed in California on February 26 th , 2020 (CADPH, 2020;

Heinzerling et al., 2020). Department of Public Health (LACDPH) COVID-19 dashboard website. These outcome data were split into two time periods: a main study period from March 16 th to September 8 th , 2020; and a secondary period for sensitivity analyses from September 8 th , 2020 to February 23 rd , 2021. These data are reported at a neighborhood statistical area unit geography. LACDPH reports infectious disease data for 'Countywide Statistical Areas (CSAs)', used by many LA County agencies to report data to the County Board of Supervisors, and which include mixed areal classifications such as 'city', 'community', 'neighborhood', or 'unincorporated area' (Harris, 2020; LACDPH, 2021) This study refers to CSAs as 'neighborhoods'.

Based on prior research (Su et al., 2020) , a land-use regression model was used to produce an annual pollution surface of NO 2 across California at a spatial resolution of 30 m using data from 2016. This surface was used previously for another recent health study in Los Angeles (Wing et al., 2020) . The landuse regression model had an out-of-sample cross-validation R 2 of 0.76. This average annual NO 2 surface was used to define the main exposure metric by neighborhood.

Potential covariates were identified a priori based on existing literature on risk factors for disease or severity of disease (including death) for COVID-19 (Myers et al., 2020) , other pneumonic infectious diseases (Neupane et al., 2010) , and previous studies on air pollution and COVID-19 (Liang et al., 2020; . Demographic covariates, including age, race/ethnicity, median household income, and household owner occupancy, were downloaded from the U.S. Census Bureau's American Community Survey (ACS) 5-year moving estimate for 2018 (U.S. Census Bureau, 2018) . Population counts at the neighborhood level, and smoking and obesity prevalence at the census-tract level were downloaded from the LACDPH website (LACDPH, 2018) . Population counts from LACDPH were compared to counts from the ACS to assess variable aggregation methods. Hospital and testing facility locations were acquired from the LA County's GeoHub website for their potential association with case ascertainment (LAC, 2014) . Residential building footprints were also downloaded from the LA County Geohub website to facilitate these demographic data aggregations (LAC, 2014) .

We also considered hypertension and diabetes as health outcomes potentially associated with COVID-19 severity (Myers et al., 2020) . These health outcomes were modeled in the U.S. Centers for Disease Control and Prevention's (CDC) 500 Cities Project health dataset (Centers for Disease Control and Prevention, 2019). These covariates were included as sensitivity analyses due to incomplete spatial coverage, Uncertainties in the testing regime raise questions about potential case-ascertainment bias. As testing became more widely available, rates of testing likely changed from testing only suspected cases to people potentially exposed as a result of occupational or social interactions. This could have affected the case rate and subsequently the case-fatality rates. It is also possible that in the earlier stages of the pandemic, there was more undercounting of the deaths, which would have diminished over time as medical professionals learned how to identify more accurately deaths resulting from COVID-19.

To address the potential shifts in case-ascertainment, case-fatality, and mortality rates that could have occurred over time, we conducted further sensitivity analyses. Specifically, we extended our original study period which captured approximately the first six months of the pandemic (March 16 th to September 8 th , 2020), to the subsequent six months (September 8 th , 2020 to February 23 rd , 2021). Thus, we replicated the analysis for the subsequent 6-month period, which had nearly four times the incident cases (875,368 cases) as the first period (230,621 cases). In the latter period, the County changed the neighborhood definitions to exclude or combine about 13 neighborhoods, so the count of neighborhoods was less than in the original period (348 vs. 335). Consequently, the two data sets are not uniformly constructed, but they are quite similar.

Very few spatial variables were available at the neighborhood geographies, as most environmental, health, and demographic areal data are published by postal ZIP code or census tract.

Environmental Systems Research Institute's (ESRI) ArcGIS 10.7 (ESRI, 2020) was used to summarize NO 2 zonal mean by each of the N=348 neighborhoods ('neighborhood' statistical area geographies, as delineated by the LACDPH) in LA County. To account for misalignment in areal boundaries between COVID-19 case/mortality and selected covariates, all areal covariates were first reaggregated to residential building footprints (acquired from LA County's Geohub website (LAC, 2014) ) and then reaggregated to neighborhoods by using counts per area density-based raster surfaces. This intermediate step was taken to minimize the effect of geographies with highly variable population densities (e.g., a large neighborhood with few total residents). Hospital and testing facility areal densities were calculated using a 10-mile radius kernel density process -generating a raster surface describing the number of hospitals or facilities per sq km within 10 miles of each raster grid cell. History of hypertension and diabetes drawn from the 500 Cities Project covered only 61% (212 of 348) of neighborhoods; therefore, we imputed the global mean for the remaining 136 neighborhoods and report these results as sensitivity analyses.

Extracted neighborhood NO 2 concentrations were modeled in relation to incident case rate (cases/population), mortality rate (deaths/population), and case-fatality rate (deaths/cases). NO 2 concentrations from neighborhoods were scaled to the interquartile range to aid in interpretation of model results (Liang et al., 2020; Xiao Wu et al., 2020) . We used three different statistical models to assess sensitivity of our results to model specification: (1) zero-inflated Poisson, (2) zero-inflated negative binomial, and (3) Bayesian conditional autoregressive (CAR) zero-inflated Poisson models. To assess incident case rate, all models treated the count of COVID-19 cases in the neighborhoods as the dependent variable and the total population as the offset. These models were also run for mortality counts with the total population as the offset (mortality rate) and for mortality counts with the total number of COVID-19 cases as the offset (case-fatality rate). Zero-inflated models were selected to account for a high number of neighborhoods with zero counts of cases or deaths and often low total populations; without zero-inflation, these low-count areas could disproportionately influence the model results. We also employed a Bayesian zero-inflated Poisson model whose spatial random effects were assigned a CAR prior distribution to account for potential spatial autocorrelation between neighborhoods. This model incorporated a spatial adjacency matrix of first-order neighbors and employed flat priors.

All three models were run with and without adjustment for covariates. The final model included the following covariates: mean percent owner occupancy, mean median income, mean percent above 65 years old, mean percent nonwhite; mean smoking prevalence, mean obesity prevalence, and mean hospital density per square mile within a 10-mile radius. Covariates selected for final model were identified a priori; however, in the event of highly correlated or colinear covariate pairs, the covariate with the highest bivariate association with the outcome was included. In all models, no covariates were found to be significant predictors of the zero-inflation component. We used R version 3.6.3 to run all statistical analyses (R Core Team, 2020).

The average area of the 348 LA County neighborhoods was 44.7 sq km (SD=171.8 sq km), with the largest being 1,144 sq km (Antelope Valley) in the northern exurban areas of the county and the smallest being 0.67 sq km (San Pasqual) in a more densely populated area near Pasadena, north of downtown LA. The annual mean NO 2 across the study region was 11.7 ppb (SD=7.3 ppb; range of 1.6 ppb to 31.3 ppb). Concentrations of NO 2 derived from the 2016 land-use regression surface are depicted in Figure 1A . The mean aggregated NO 2 across neighborhoods was 15.6 ppb (SD=6.0 ppb) with an interquartile range of 8.7 ppb. Between March 16 th and September 8 th , 2020, the LACDPH recorded 230,621 confirmed cases of COVID-19, and 5,653 deaths due to COVID-19 were observed. In a population of 10.0 million, this translated into a case rate of 2.2% ( Figure 1B) , a mortality rate of 0.054% ( Figure 1C ), and a case-fatality rate of 2.5% ( Figure 1D ). The period from September 8 th , 2020 to February 23 rd , 2021 included 875,368 cases and 13,344 deaths and was used as sensitivity analysis.

Between neighborhoods (N=348), the mean percent owner occupancy was 54.7% (SD=22.0%); mean median income was $47,483 (SD = $68,898); mean percent above 65 years old was 13.9% (SD=0.1%); mean percent nonwhite was 45.2% (SD= 20.1%); mean smoking prevalence was 12.7% (SD=2.6%); mean obesity prevalence was 23.7% (SD=7.4%); and the mean hospital density per square mile within a 10-mile radius was 1.2x10 -4 (SD=1.4x10 -4 ). The mean hypertension and diabetes prevalence were 18.7% (SD=9.6%), and 7.4% (SD=4.2%), respectively. Although median income was highly associated with the outcome in the crude model, it was highly correlated with owner occupancy (r = 0.89); therefore, owner occupancy was selected for inclusion as it demonstrated a larger bivariate association with all outcomes.

Ordinary residuals from the zero-inflated Poisson and negative binomial models demonstrated significant spatial autocorrelation at the global (Moran's I p-value <0.001) and local (Anselin hot-spots) levels; thus, we used a Bayesian zero-inflated Poisson model with a CAR prior on the random effects to account for spatial dependence of the residuals per neighborhood. Crude and adjusted model results for (1) zero-inflated Poisson, (2) zero-inflated negative binomial, and (3) zero-inflated Poisson spatial models are shown in Table 2 . In the adjusted zero-inflated Poisson model, we found that the incidence rate ratio (IRR) of NO 2 was 1.31 (95% CI: 1.29, 1.33) for the case rate. That is, we found that an increase of 8.7 ppb (IQR) in mean annual NO 2 (2016) was associated with a 31% increase in COVID-19 incident case rate. The adjusted zero-inflated negative binomial and spatial models demonstrated a smaller effect of 16% (95% CI: 2%, 32%) and 18% (Credible Interval -CrI: 10%, 32%) increase in case rate, respectively (Table 2A) . Adjusted models reduced residual uncertainty compared to the crude model estimates for case rate. The adjusted Poisson, negative binomial, and spatial models all demonstrated an increase in COVID-19 mortality of 35% (95% CI: 23%, 48%), 44% (95% CI: 11%, 86%), and 60% (CrI: 37%, 88%), respectively, across the IQR exposure increment (Table 2B) . Again, these adjusted models improved residual uncertainty compared to crude model estimates for mortality rate. Finally, adjusted Poisson and negative binomial models showed positive yet non-significant results for the association between NO 2 and COVID-19 case-fatality; however, the spatial CAR model demonstrated that an IQR increase in mean annual NO 2 was associated with a 31% (CrI:

10%, 65%) increase in case-fatality. Sensitivity analyses, which included the addition of history of hypertension and diabetes in the models, had comparable results across all three models. In comparing the two time periods, before and after September 8 th , 2020, we found that the results were largely consistent, despite very different case numbers, testing regimes, and improvements in classifying deaths. While some differences exist in the size of the effects, overall the conclusions remain the same. The results of this sensitivity analysis give some assurance that the changes in testing, case ascertainment, and mortality classification over time are not having a substantial effect on the key conclusion that long-term air pollution exposure likely increases the risk of Covid-19 infection and death. Full model results, including incidence risk ratios for all covariates, are described in Appendix A for the main model, Appendix B for sensitivity analysis including hypertension and diabetes, and Appendix C for sensitivity analysis utilizing the period between September 8 th , 2020 and February 23 rd , 2021.

We found annual NO 2 to be associated with COVID-19 incidence and mortality in Los Angeles County neighborhoods while adjusting for selected confounders. These findings were consistent across statistical model specification, although risk estimates displayed some variation between models. In addition, we found in the CAR an association between NO 2 and COVID-19 case-fatality; other models also showed a positive but insignificant associations. Covariates in the models largely had the expected sign of effect. Furthermore, our sensitivity analyses, which included the addition of hypertension and diabetes prevalence covariates, had a minimal impact on the effect size or interpretation of our model estimates for NO 2 and COVID-19 outcomes. Our sensitivity analysis involving a second, approximately 6-month time period, with slightly different outcome reporting (N=348 vs N=335) also demonstrated comparable results to the main model and study period.

Our findings are consistent with two previous studies demonstrating a relationship between air pollution and COVID-19 nationally at the county scale in the U.S. One study investigated the association between NO 2 and COVID-19, and they observed remarkably similar findings. Specifically, Liang et al. (2020) reported an increase of 4.6 ppb (IQR across all counties) NO 2 to be associated with a 16.2% (CI: 8.7%, 24.0%) increase in mortality rate and an 11.3% (CI: 4.9%, 18.2%) increase in case-fatality rate (Liang et al., 2020) . When we scaled to Liang et al.'s IQR of NO 2 , our models demonstrated similar results of 17.1% (CI: 11.3%, 23.2%), 21.3% (CI: 5.9%, 38.9%), and 30.1% (CI: 12.4%, 50.6%) increases in COVID-19 mortality rate for our zero-inflated Poisson, negative binomial, and CAR models, respectively. In comparison to the study, which observed that a 1 µg m -3 increase in air pollutant PM 2.5 was associated with an 11% increase in mortality rate , we found an 8.7 ppb increase in another traffic-related air pollutant, NO 2, to be associated with a 35 -60% (range of three models; Table 2 ) increase in mortality rate. The Wu et al. study, however, did not report results scaled to the interquartile range of PM 2.5 , so we scaled their results to the IQR for PM 2.5 from the Liang et al. study

(2.6 µg m -3 ), which uses a similar number of U.S. counties. This resulted in a highly comparable 31.2% increase in mortality for a 2.6 µg m -3 increase in PM 2.5 . Although Wu et al., Liang et al., reinforcing the validity of our results based on a priori expectation. The larger effect size on mortality rate in our study compared to the other two studies could be due to greater spatial variability resulting from using building-footprint covariate aggregation on smaller-area neighborhood geographies rather than using county-level data.

To our knowledge, only one study from England has undertaken small-area neighborhood analysis of the association of COVID-19 and air pollution using spatial modeling techniques (Travaglio et al., 2021) . This study also reported associations with nitrogen dioxides at a sub-regional scale. Smallerarea analyses likely reduce potential exposure measurement error and lead to more consistent ascertainment of cases and deaths than those using the larger county units -both of which likely result in more precise and reliable estimates of health effects from air pollution exposures. By utilizing Bayesian models with CAR priors, we also accounted for spatial autocorrelation or clustering between adjacent administratively defined neighborhoods. This addition is important as transmission for COVID-19 and other infectious diseases is likely to be clustered spatially due to respiratory community spread (see map of incidence in Figure 1 ).

Our study has several limitations. Most importantly, our study is limited by population-level counts of COVID-19 cases and deaths. These aggregate data, made publicly available by LACDPH, have facilitated this research but have also introduced some uncertainties. For example, it is difficult to determine how testing rates or the prevalence of asymptomatic cases, which may show significant neighborhood-level variation, could impact our results. We are unable to determine data accuracy, specifically for the earlier phases of the pandemic, when data collection protocols were still being defined. The number of neighborhoods reported by LACDPH has fluctuated, from 348 (in September 2020) to 335 (in February 2021) distinct areas. Our sensitivity analysis on a second time period (September 2020 to February 2021) including N=335 neighborhoods demonstrated similar results to our initial period (March 2020 to September 202), so these potential data quality issues appear to have minimal effect on the interpretation of our results. Ideally, with data access granted, future research would avoid aggregate-level data in favor of individual-level outcomes. Utilizing individual-level home locations of cases and deaths rather than neighborhood-level aggregate counts would greatly improve air pollution exposure attribution and allow for better ascertainment of potential confounders. In addition, we could not include daily or weekly observations to account for changes in case or mortality rate over time, but rather used cumulative counts for the study period. This was due to inconsistent case reporting and incomplete death reporting due to human subject concerns earlier in the pandemic. Deaths may have been undercounted, as death certificates and coroner reports may incorrectly attribute cause of death (Jewell et al., 2020; Quast and Andel, 2020) . Future research may benefit from using excess mortality for comparison (Banerjee et al., 2020) Indicators of symptom severity from hospital, intensive care unit, and emergency room admittance data are more difficult to acquire, as they are not publicly available, but future analyses on outcomes of severity would allow for better understanding of the effect of NO 2 exposure on the progression of the disease.

This study is also limited by the use of land-use regression from 2016 to estimate long-term NO 2 exposure. Additional years of land-use regression surfaces could be utilized to better describe long-term trends; however, estimates of NO 2 in Los Angeles County in the years immediately preceding the pandemic are likely similar in their spatial pattern over such a short time span. Other covariate data included in these analyses were spatially misaligned. A strength of this study is our use of residential building footprints as an intermediate step in aggregating areal covariates. Although using these building footprints better accounts for population density patterns than more straightforward aggregation techniques (e.g., census-tract directly to neighborhood), the method may cause misalignment errors due to differences in building characteristics (e.g., height, unit size, etc.). Finally, our spatial models accounted for unexplained spatial variability in the between-neighborhood random effect, which suggests there may be additional covariates with a similar spatial pattern that we have not included in these analyses, and further investigation is necessary. While the CAR model accounts for this dependence in the statistical inference, we cannot rule out important missing confounders.

In summary, our findings imply a potentially large association between exposure to air pollution and population-level rates of COVID-19 cases and deaths. Our findings demonstrate comparable results to other recent literature, especially concerning the association of long-term NO 2 and COVID-19 mortality rate. Our small-area analyses, covariate aggregation methods using building footprints for accounting for population density variability, and utilization of spatial modeling (CAR model with spatial random effect) make novel contributions to the available literature. These findings are especially important for targeting interventions aimed at limiting the impact of COVID-19 in polluted communities.

In the U.S., more polluted communities often have lower incomes and higher proportions of Black and Latinx people. In addition, Black and Latinx people have higher rates of pre-existing conditions, potentially further exacerbating the risk of COVID-19 transmission and death (Clark et al., 2014; O'Neill et al., 2003) . The elevated risk of case incidence and mortality observed in these populations might result from higher exposure to air pollution. As COVID-19 data reporting improves and data access is given more readily to researchers, we will further refine these analyses to the individual-level in a spatial framework.

Word Count: 4,327 (Su et al., 2020) NO 2 (ppb) California (raster; 30 m) 2016 *LACDPH population used for regression modeling; ACS population used to validate aggregation methods. **building footprints used for data aggregation. 

-Environmental factors such as air pollution may contribute to Covid-19 incidence and death -This study investigates whether air pollution relates to Covid-19 incidence and mortality in neighborhoods of Los Angeles -Los Angeles is a global epicenter for the pandemic with 1,105,989 cases to date -Findings suggest that chronic exposure to nitrogen dioxide exerts large effects on Covid-19 disease incidence and mortality -Higher levels of air pollution in neighborhoods with high proportions of Latinx and Black people may partly explain why these population groups have suffered disproportionately from the pandemic

Unequal places: The impacts of socioeconomic and race/ethnic differences in neighborhoods

Agriculture is a major source of NOx pollution in California

Exposure to ambient ultrafine particles and nitrogen dioxide and incident hypertension and diabetes

Estimating excess 1-year mortality associated with the COVID-19 pandemic according to underlying conditions and age: a population-based cohort study

Long-term effects of traffic-related air pollution on mortality in a Dutch cohort (NLCS-AIR study)

Oscillations in U.S. COVID-19 Incidence and Mortality Data Reflect Diagnostic and Reporting Factors. mSystems 5

Geographic Differences in COVID-19 Cases, Deaths, and Incidence -United States

Air pollution, racial disparities, and COVID-19 mortality

Patterns of residential crowding among Hispanics in later life: immigration, assimilation, and housing market factors

CDC Confirms Possible First Instance of COVID-19 Community Transmission in California

COVIDView: A Weekly Surveillance Summary of US

500 Cities Project: Local data for better health | Home page | CDC [WWW Document

Air Pollution and Respiratory Viral Infection

National patterns in environmental injustice and inequality: Outdoor NO2 air pollution in the United States

The Effects of Air Pollution on COVID-19

Impact of road proximity on infants and children with bronchopulmonary dysplasia

Hypertension, overweight/obesity, and diabetes among immigrants in the United States: An analysis of the 2010-2016 National Health Interview Survey

Air pollution and case fatality of SARS in the People's Republic of China: an ecologic study

The Influence of Living Near Roadways on Spirometry and Exhaled Nitric Oxide in Elementary Schoolchildren

Cardiovascular Disease Management in Minority Women: Special Considerations

Association of Body mass index (BMI) with Critical COVID-19 and in-hospital Mortality: a dose-response meta-analysis

ArcGIS Desktop: Release 10

Air pollution and cardiovascular disease

African american exposure to prescribed fire smoke in Georgia, USA

Understanding the Los Angeles County coronavirus epidemic: The critical role of intrahousehold transmission

Transmission of COVID-19 to Health Care Personnel During Exposures to a Hospitalized Patient

Spatial Analysis of Air Pollution and Mortality in Los Angeles

Traffic-related air pollution and asthma onset in children: A prospective cohort study with individual exposure measurement

Caution Warranted: Using the Institute for Health Metrics and Evaluation Model for Predicting the Course of the COVID-19 Pandemic

Near-road air pollution impacts of goods movement in communities adjacent to the Ports of Los Angeles and Long Beach

The fraction of lung cancer incidence attributable to fine particulate air pollution in France: Impact of spatial resolution of air pollution models

Countywide Building Outlines (2014) | County Of Los Angeles Enterprise GIS [WWW Document

Los Angeles County COVID-19 Dashboard -Data Dashboard -About

COVID-19 Locations & Demographics -LA County Department of Public Health

Department of Public Health -Health Assessment Unit -Data Topics

Air pollution and temperature are associated with increased COVID-19 incidence: A time series study

Urban Air Pollution May Enhance COVID-19 Case-Fatality and Mortality Rates in the United States

Association between environmental pollution and prevalence of coronavirus disease 2019 (COVID-19) in Italy

California confirms 2 cases of coronavirus in L.A., Orange counties -Los Angeles Times [WWW Document

SARS-CoV-2 Positivity Rate for Latinos in the

Housing tenure, structure, and crowding among Latino households

Characteristics of Hospitalized Adults With COVID-19 in an Integrated Health Care System in California

Long-term exposure to ambient air pollution and risk of hospitalization with community-acquired pneumonia in older adults

Health, wealth, and air pollution: Advancing theory and methods

Excess mortality and potential undercounting of COVID-19 deaths by demographic group in Ohio

Air quality impacts of a scheduled 36-h closure of a major highway

R: A language and environment for statistical computing

Mortality Among Essential Workers in the United States

Predicting traffic-related air pollution in Los Angeles using a distance decay regression selection strategy

Predicting differential improvements in annual pollutant concentrations and exposures for regulatory policy assessment

Identification of Effects of Regulatory Actions on Air Quality in Goods Movement Corridors in California

Health effects of diesel exhaust emissions

Links between air pollution and COVID-19 in England

American Community Survey 5-Year Data

Is there an association between the level of ambient air pollution and COVID-19?

Four-Month Changes in Air Quality during and after the COVID-19 Lockdown in Six Megacities in

Preterm birth among infants exposed to in utero ultrafine particles from aircraft emissions

WHO Coronavirus Disease (COVID-19) Dashboard | WHO Coronavirus Disease (COVID-19) Dashboard [WWW Document

Influence of the Spatial Resolution of the Exposure Estimate in Determining the Association between Heat Waves and Adverse Health Outcomes

Exposure to air pollution and COVID-19 mortality in the United States: A nationwide cross-sectional study

Air pollution and COVID-19 mortality in the United States: Strengths and limitations of an ecological regression analysis

Oxidation of Nitrogen in Combustion and Explosions

Effects of meteorological conditions and air pollution on COVID-19 transmission: Evidence from 219 Chinese cities

Association between short-term exposure to air pollution and COVID-19 infection: Evidence from China

Abstract not included, as per Environment International Author Guidelines for New Developments entries

Spatial Analysis of COVID-19 and Traffic-related Air Pollution in Los Angeles" by Lipsitt et al. Submitted to: Environment International as 'New Development' entry

☒ The authors declare that they have no known competing financial interests or personal relationships that could have

Model conducted for sensitivity analyses -including main model (Appendix A) covariates with the addition of hypertension and diabetes derived from the 500 Cities Project dataset. Hypertension and diabetes data covered 212 of N=348 (61% coverage) neighborhoods. We imputed the global mean for the remaining 136 neighborhoods. This table shows the adjusted association of NO 2 (scaled by interquartile range: 8.7 ppb) and COVID-19 from three models: A. Case rate (cases/population); B. Mortality rate (deaths/population); and C. Case-fatality rate (deaths/cases) for Los Angeles County neighborhoods (N = 348) for the period between March 16 th and September 8 th , 2020. NO 2 was scaled by its interquartile range for interpretation purposes; other covariates were not scaled.