Final Report: Intra-Urban Variation of Air Pollution and Cardiovascular Health Effects

EPA Grant Number: R834898
Title: Intra-Urban Variation of Air Pollution and Cardiovascular Health Effects
Investigators: Ito, Kazuhiko , Clougherty, Jane E. , Matte, Thomas , Ross, Zev
Institution: New York University School of Medicine , Hunter College , New York University , University of Pittsburgh
EPA Project Officer: Ilacqua, Vito
Project Period: April 1, 2011 through March 31, 2013 (Extended to December 31, 2015)
Project Amount: $299,998
RFA: Exploring New Air Pollution Health Effects Links in Existing Datasets (2010) RFA Text |  Recipients Lists
Research Category: Air Quality and Air Toxics , Health Effects , Air


The objectives of this project were to determine: (1) the impacts of air pollution and weather effects on the cardiovascular health outcomes available at NYCDOHMH, including cardiovascular emergency department syndrome data, hospitalizations, and mortality; (2) the effect modification of the effects on cardiovascular disease (CVD) outcomes by intra-urban variation of combustion sources as measured at up to 150 locations by the New York City Community Air Survey (NYCCAS); and (3) the effect modification of the cardiovascular effects of air pollution by socio-economic status. We recently developed a cardiovascular emergency department (ED) visits syndrome indicator (available at Zip-code level) that is useful in determining near-real time impacts of weather and air pollution. New York City (NYC) residents are exposed to multiple air pollutants coming from a variety of combustion sources, including transported secondary aerosols, local sources including traffic, building space-heating, and oil burning from ships in nearby ports. NYC residents also reflect a wide range of health and socio-economic status, and therefore likely present a range of susceptibility indicators associated with neighborhood characteristics. Thus, this study takes advantage of the unique databases that have been developed to determine the cardiovascular effects of air pollution in unique environmental and population settings of NYC to answer the relevant research questions.

Summary/Accomplishments (Outputs/Outcomes):

Intra-urban variation of PM2.5 elemental concentrations in New York City: We developed land-use regression (LUR) models to characterize spatial variation of fifteen PM2.5 elements collected at 150 street-level locations in New York City during December 2008 - November 2009: aluminum, bromine, calcium, copper, iron, potassium, manganese, sodium, nickel, lead, sulfur, silicon, titanium, vanadium and zinc. Summer- and winter-only data available at 94 locations in the subsequent 3 years, up to November 2012, were analyzed to examine variation of LUR results across years. Temporally adjusted spatial variation of each element was modeled in LUR including six major emission indicators: boilers burning residual oil; traffic density; industrial structures; construction/demolition (these four indicators in buffers of 50 to 1,000 m), commercial cooking based on a dispersion model; and ship traffic based on inverse distance to navigation path weighted by associated port berth volume. All the elements except sodium were associated with at least one source, with R-squared ranging from 0.2 to 0.8. Strong source-element associations, persistent across years, were found for: residual oil burning (nickel, zinc); near-road non-exhaust traffic (copper, iron); and, ship traffic (vanadium). These emission source indicators were also significant and consistent predictors of PM2.5 concentrations across years. We submitted this manuscript to a journal, received comments (revision requested), and currently (March 2016) revising the manuscript for the April 2016 deadline for a revision.

Estimation of annual average PM2.5 elemental concentrations: Using the LUR models mentioned above, we estimated spatial variation of 14 PM2.5 elements at 100 m x 100 m lattice cells (there are 82,746 lattice cells in New York City) where emission indicators (i.e., predictors of the LUR models) were procured. Based on these estimated PM2.5 element concentrations, we computed: (1) Zip-code level annual average PM2.5 elemental concentrations for the analysis to examine potential effect modification of PM2.5 associations with CVD morbidity; and (2) maternal address PM2.5 elements exposures for a birth cohort (see below) for the analysis of PM2.5 elemental concentrations and birth weight, using the nearest lattice cells. The estimated concentrations at these lattice cells also allow computing average concentrations at other geographic boundaries (e.g., census tract).   

Effect modification of the short-term associations between PM2.5 and CVD morbidity by Zip-code level: Using the estimated Zip-code level average PM2.5 elements' concentrations described above, we evaluated, through meta-regression, if they explained the heterogeneity in PM2.5 risk estimates for CVD syndrome ED visits computed for each of 172 zip-codes for the years 2002-2012.  While several elements (e.g., S, V), as well as census variables (e.g., percent non-Hispanic black and percent poverty) and traffic density indicator at Zip-code level were positive predictors of larger PM2.5 CVD ED risks (up to ~25% larger the citywide risk estimate without the modifiers), none of these associations (i.e., effect modification) were statistically significant. The smallest geographic boundary available for ED visits syndrome data from the Bureau of Communicable Diseases is zip-code. However, with an approval from our IRB, we geocoded ED visits (years 2005-2013) and hospitalizations data (years 2000-2013) from New York State Department of Health as well as mortality data (years 1997-2013) from the Bureau of Vital Statistics. Geocoding was completed in 2015. Therefore, we will complete this component of the project using the PM2.5 elemental concentration estimates at finer geographic boundary, census tract (as well as Zip-code level), and using a case-crossover design with the citywide PM2.5 mass as the main effect and an interaction term for the census tract level PM2.5 elemental concentrations and census variables as potential effect modifiers. While this additional analysis is beyond what was proposed (Zip-code level analysis), an examination of the influence of geographic scale of exposure estimation in the analysis will be an important contribution. This analysis of multiple health outcomes is expected to be completed by summer 2016.   

PM2.5 elemental concentrations, emission indicators, and term birth weight in New York City: In collaboration with Brown University and other institutions, we examined the relationship between spatial variation of PM2.5 elements, emission indicators, and birth weight reduction in New York City, where the association was found for PM2.5 mass. The PM2.5 elemental concentrations were estimated at maternal address for the 252,967 singleton term births to non-smokers in NYC born 2008-2010 using the LUR model. Adjusting for individual and neighborhood characteristics, the estimated Ni and Cu exposures as well as their associated emission indicators, residual oil combustion and traffic density, were associated with birth weight reductions. When the constituents' exposures were assigned using the nearest monitors' (of the 150 monitors), as was often done in the past studies, the associations were weaker and not significant for near-road associated elements, suggesting that, for the air pollutants whose spatial extent is relatively short (e.g., near-road), the scale of analysis matters. The result was presented at the 2014 Annual Conference of International Society for Environmental Epidemiology (ISEE), but we updated the LUR model since then and re-computed PM2.5 elemental concentrations at maternal address. The result remained the same; a manuscript is being prepared.

Changing air pollution levels and their short-term associations with mortality in New York City: The short-term mortality risks for the daily citywide average values of PM2.5, NO2, SO2, CO, and O3 from the regulatory monitors were estimated for: (1) 2000-2003; (2) 2004-2007; and (3) 2008-2012, to determine if the risk estimates changed over the period. All the pollutants, with the exception of O3, exhibited a monotonic decline in the mean levels across the three periods, with the magnitude of reduction ranging from ~20% (NO2) to over 50% (SO2 and CO). The significance of association with all-cause mortality was diminished in the most recent period for CO, NO2, SO2, and O3, but PM2.5's risk estimates and their significance have not declined. The result was presented at the 2014 ISEE Annual conference. Since then, mortality data for 2013 and 2014 became available. We also updated the citywide daily air pollution data up to 2014. Thus, we are updating the analysis with these recent data. Parallel analyses for CVD and respiratory hospitalizations are being conducted and will be part of a manuscript in preparation that includes all of these health outcomes.

Seasonal pattern of associations between PM2.5 and CVD hospitalizations and mortality: To investigate the difference in seasonal pattern of associations between PM2.5 and CVD mortality and PM2.5 and CVD hospitalizations reported in the past multi-city studies, we stratified CVD deaths into those who died in vs. outside hospital and CVD hospitalizations into those who were discharged alive vs. died using the 2000-2012 data. Among CVD deaths, the association with PM2.5 was stronger among the outside-hospital deaths and stronger in the warm season. Among the CVD hospitalizations, the association with PM2.5 was stronger among those who were discharged alive, and while their associations with PM2.5 were seen in both seasons, they were stronger in the cold season. We are currently analyzing the linked hospitalization-death records (this data set became available in 2015) to further examine these seasonal patterns of associations with PM2.5.

The Influence of seasonal Influenza Infections on CVD Mortality: CVD deaths have seasonal trends that appear to be influenced by influenza epidemics. In air pollution epidemiological studies, such associations are usually treated as unmeasured seasonal confounding and modeled with a smooth function of study days in the regression models. A postdoctoral researcher, Jennifer Nguyen, joined our IRB protocol and analyzed the database at NYCDOHMH to quantify the temporal association between population increases in seasonal influenza infections, using NYCDOHMH’s daily syndromic surveillance of influenza-like-illness emergency department visits as an indicator, and CVD mortality for the study period 2006-2012, and to test if influenza incidence indicators are predictive of cardiovascular mortality during the influenza season. Among adults age ≥ 65 years who accounted for 83% of the CVD mortality during influenza seasons, seasonal average influenza incidence was strongly correlated year-to-year to excess cardiovascular mortality. Associations were most acute and strongest for myocardial infarction mortality. The manuscript of these results has been accepted (Nguyen et al., 2016).   

Spring tree pollen, allergy medication sales, and asthma emergency department visits: Using the daily weather and air pollution data set developed for this project, and collaborating with Columbia and Fordham University researchers who collected daily pollen counts, we analyzed associations between daily variations of nine clinically relevant pollen types and over-the-counter allergy medication sales, and asthma syndrome ED visits in the spring period for years 2002-2012. We found that mid-spring pollen types (e.g., ash, sycamore, oak) had substantive impacts on both outcomes, and for asthma syndrome ED visits, the associations were strongest for children (age 5-17). The result was published in 2015. A parallel analysis was conducted for CVD syndrome ED visits (because this outcome also showed variations in the spring), but we found no association. In addition, using the spatially-stratified time-series analysis method developed for this project, Kate Weinberger (Columbia University at the time of analysis and now at Brown University) joined our IRB protocol and examined if the heterogeneity in asthma syndrome ED risks for spring pollen estimated at Zip-code level could be explained by neighborhood characteristics. She found that a higher tree canopy density at Zip-code level was associated with a larger risk of mid-spring pollen impacts on asthma syndrome ED visits. This manuscript is in preparation.

Excess mortality attributable to extreme heat in New York City: Heat-stroke (i.e., hyperthermia, an external cause of deaths) is often used as an indicator to measure the impact of heat-waves on mortality, but no study has compared its magnitude with that of an increase in non-external cause of deaths associated with heat-waves. Using the daily weather and air pollution data set (years 1997-2013) developed for this project, we conducted a time-series analysis of the impacts of extreme heat event (EHE) as defined as days meeting current National Weather Service forecast criteria for issuing heat advisories in NYC based on observed maximum daily heat index values. The cumulative relative risk (CRR) of death associated with EHE was estimated in a Poisson time-series model for each year using an unconstrained distributed lag for days 0-3. A random-effects combined relative risk accounting for year-to-year variation was 1.11 (95%CI 1.08-1.14) per extreme heat event day. Average excess non-external cause deaths associated with EHE (median of 121 deaths per summer) were nearly eleven-fold greater than hyperthermia deaths. A manuscript describing these results has been accepted (Matte et al., 2016). 

Comparing associations between extreme heat events and mortality in NYC using observed and forecast weather products: Using the database developed for this project, we have also investigated the relationships between forecast weather, observed weather, National Weather Service-issued heat alerts, and non-external cause mortality for May through September, 2004-2011.  Mortality CRRs over lag 0 through 3 days for EHE days based on the same-day morning forecast (RR = 1.092, 95% CI: 1.047, 1.138) and NWS-issued heat alerts (RR = 1.083, 95% CI: 1.040, 1.127) were close to that for the observed EHE days (RR = 1.080, 95% CI: 1.036, 1.127) while the RRs for extreme heat events based on 1- and 2-day ahead forecasts were smaller (RR = 1.06 and 1.05, respectively). Meeting criteria for extreme heat based on the same-morning forecast predicts mortality risk similarly to observed weather. A manuscript describing these results has been written (Chen, lead author) and is currently being reviewed by the NYCDOHMH review system.


In this study, we found substantive within-city variations in PM2.5 elemental concentrations, and these variations were significantly explained by major emission source indicators for residual oil burning, near-road traffic, and ship emissions. While the concentrations of some of these elements declined significantly over the 2008-2013 period, especially those associated with residual oil burning, the spatial pattern of these concentrations were correlated across years. These spatial variations of PM2.5 elements did not significantly explain the within-city heterogeneity in the short-term PM2.5 risk for CVD ED visits, at least at the Zip-code level (the originally proposed geographic boundary). Analyses using the exposure estimation at census tract level is being conducted for both CVD hospitalizations and mortality. In the birth cohort for 2008-2010, PM2.5 elemental concentrations, particularly those best predicted from the residual oil burning indicator, were associated with a reduction in term birth weight. The estimated PM2.5 elemental concentrations based on the land-use regression models were more significantly associated with birth weight reductions than those based on the nearest monitors, highlighting the importance of spatial resolution in these analyses. Using the databases we developed for this project, we conducted a number of other analyses examining other health outcomes (e.g., asthma ED visits) and different environmental risk factors (e.g., pollen, extreme heat event) and yielded policy-relevant results. Overall, this EPA project helped us create an extremely useful framework with which we can routinely evaluate the impacts of air pollution (and weather) that continue to change in the future. The routine NYCCAS monitoring and the annual reporting became mandated by the city law in 2015. Therefore, we will be able to characterize within-city variation of air pollutants, their trends, and the impacts on health outcomes using the framework developed in this project.


Journal Articles on this Report : 4 Displayed | Download in RIS Format

Other project views: All 10 publications 4 publications in selected types All 4 journal articles
Type Citation Project Document Sources
Journal Article Ito K, Weinberger KR, Robinson GS, Sheffield PE, Lall R, Mathes R, Ross Z, Kinney PL, Matte TD. The associations between daily spring pollen counts, over-the-counter allergy medication sales, and asthma syndrome emergency department visits in New York City, 2002-2012. Environmental Health 2015;14(1):71. R834898 (2014)
R834898 (Final)
  • Full-text from PubMed
  • Abstract from PubMed
  • Associated PubMed link
  • Full-text: Environmental Health-Full Text HTML
  • Abstract: Environmental Health-Abstract
  • Other: Environmental Health-Full Text PDF
  • Journal Article Ito K, Johnson S, Kheirbek I, Clougherty J, Pezeshki G, Ross Z, Eisl H, Matte TD. Intraurban Variation of Fine Particle Elemental Concentrations in New York City. Environmental Science & Technology 2016;19;50(14):7517-7526. R834898 (Final)
  • Abstract from PubMed
  • Full-text: ES&T-Full Text HTML
  • Abstract: ES&T-Abstract
  • Other: ES&T-Full Text PDF
  • Journal Article Matte TD, Lane K, Ito K. Excess mortality attributable to extreme heat in New York City, 1997-2013. Health Security 2016;14(2):64-70. R834898 (Final)
  • Abstract from PubMed
  • Abstract: Liebert-Abstract
  • Other: ResearchGate-Abstract
  • Journal Article Nguyen JL, Yang W, Ito K, Matte TD, Shaman J, Kinney PL. Seasonal influenza infections and cardiovascular disease mortality. JAMA Cardiology 2016;1(3):274-281. R834898 (Final)
  • Abstract from PubMed
  • Full-text: JAMA-Full Text HTML
  • Abstract: JAMA-Abstract
  • Other: JAMA-Full Text PDF Link
  • Supplemental Keywords:

    Weather, air pollution, cardiovascular disease, asthma, hospitalizations, mortality, fine particulate matter, nickel, copper, vanadium

    Progress and Final Reports:

    Original Abstract
    2011 Progress Report
    2012 Progress Report
    2013 Progress Report
    2014 Progress Report