Presentation is loading. Please wait.

Presentation is loading. Please wait.

JEREMY SARNAT, STEFANIE SARNAT, W. DANA FLANDERS, HOWARD CHANG, JAMES MULHOLLAND, LISA BAXTER, VLAD ISAKOV, HALÛK ÖZKAYNAK Annual Conference of ISES/ISEE/ISIAQ.

Similar presentations


Presentation on theme: "JEREMY SARNAT, STEFANIE SARNAT, W. DANA FLANDERS, HOWARD CHANG, JAMES MULHOLLAND, LISA BAXTER, VLAD ISAKOV, HALÛK ÖZKAYNAK Annual Conference of ISES/ISEE/ISIAQ."— Presentation transcript:

1 JEREMY SARNAT, STEFANIE SARNAT, W. DANA FLANDERS, HOWARD CHANG, JAMES MULHOLLAND, LISA BAXTER, VLAD ISAKOV, HALÛK ÖZKAYNAK Annual Conference of ISES/ISEE/ISIAQ Basel, Switzerland – August 21st Spatiotemporally-Resolved Air Exchange Rate as a Modifier of Acute Air Pollution-Related Morbidity

2 Motivating Question Do spatiotemporally varying air exchange rates (AERs) modify observed short-term estimates of health risk associated with ambient air pollution? Premise: True population exposure to ambient pollution is greater when AER is higher (  infiltration indoors) Measured or modeled ambient pollution provide more accurate surrogates of true exposures when AERs are higher (  exposure error or misclassification) Variability in AERs may explain heterogeneity in short-term observed health risk  Evidence from cross-sectional studies, long-term risk (Chen et al, EHP 2012; Hodas et al., JESEE 2012; Levy et al., Epi 2005)

3 Approach Do spatiotemporally varying air exchange rates (AERs) modify observed short-term estimates of health risk associated with ambient air pollution?  Assess effect measure modification in Atlanta timeseries of daily pollution and emergency department (ED) visits (SOPHIA)  Outcome: Daily ED visits for Asthma/Wheeze (ICD-9: 493, 786.07)  4-year subset: January 1999 – December 2002  Total visits: 85,157 (Mean: 58 counts/day)  Exposure: Spatially interpolated hybrid model for pollutants  Nitrogen oxide (NOx), carbon monoxide (CO), fine PM (PM 2.5 ), ozone  Monitoring-based background concentrations and local AERMOD output for 186 ZIP code centroids (see Ozkaynak poster)  Exposure modifier: Daily AERs modeled for each of the 186 ZIP codes

4 Estimating AER  Hard to quantify  Function of numerous physical, chemical and meteorological factors  Varies geographically (city-to-city, home-to-home) and over time  Data required to quantify not readily accessible  Previous studies have used broad AER indicators  Air conditioning prevalence (Janssen et al., 2002; Levy et al., 2005)  I/O sulfate ratios (Sarnat et al., 2000)  Here, we use relatively simple estimates with publicly-available data

5 Estimating AER*  AER *Chan et al. Distribution of residential air leakage: Implications for health outcome of an outdoor toxic release. Indoor Air 2005: Proceedings of the 10th International Conference on Indoor Air Quality and Climate 2005; 1– 5: 1729–1733

6 Estimating AER*  AER *Chan et al. Distribution of residential air leakage: Implications for health outcome of an outdoor toxic release. Indoor Air 2005: Proceedings of the 10th International Conference on Indoor Air Quality and Climate 2005; 1– 5: 1729–1733

7 Estimating AER*  AER *Chan et al. Distribution of residential air leakage: Implications for health outcome of an outdoor toxic release. Indoor Air 2005: Proceedings of the 10th International Conference on Indoor Air Quality and Climate 2005; 1– 5: 1729–1733

8 Estimating AER*  AER *Chan et al. Distribution of residential air leakage: Implications for health outcome of an outdoor toxic release. Indoor Air 2005: Proceedings of the 10th International Conference on Indoor Air Quality and Climate 2005; 1– 5: 1729–1733 f W = wind coefficient f S = stack coefficient empirical and assumed constants across domain ** Method based solely on data from US Census, local met stations, and several simplifying assumptions

9 Mean AER by Zip in Metro Atlanta Mean AER in hr -1

10 Analytical Approach Primary model: Spatially-resolved timeseries: Poisson GLM log(E(Y kt )) = α + β pollution kt +  k γ k ZIP kt + … other covariates Where ‘Y kt ’ is ED visit count for asthma/wheeze in zip ‘k’ on day ‘t’  ZIP modeled as a fixed effect : 186 individual timeseries Single-pollutant time-series models  0-1-2 day moving average pollutant concentrations  Scaling: per IQR Models control for meteorology, holidays, day of week, zip code

11 Analytical Approach Primary model: Spatially-resolved timeseries: Poisson GLM To assess modification by AER: 1)Stratification of primary model by median AER (spatial) 2)Inclusion of daily AER-pollutant product terms (temporal) 3)Tertile analyses of pollutant and AER distributions  Bin each day for each ZIP into 3 x 3 pollutant-AER matrix  Model effect of the 9 indicator terms of each pollutant-AER tertile combination log(E(Y kt )) = α + β pollution kt +  k γ k ZIP kt + … other covariates

12 NOx Results: Stratified Models and Interaction Terms NOx * AER Interaction Estimate95% LCL95% UCLp-value NOx-0.108-0.6200.4030.68 AER-0.035-0.1490.0790.55 NOx * AER1.9060.0803.7310.04

13 High NOx (> 49 ppb) Moderate NOx (28 – 48 ppb) Low NOx (< 27 ppb) NOx- Asthma ED visits by tertile of pollutant concentration and AER Atlanta,1999 -2002

14 High PM 2.5 (> 19.2  g/m 3 ) Moderate PM 2.5 (13.4 – 19.1  g/m 3 ) Low PM 2.5 (< 13.3  g/m 3 ) PM 2.5 - Asthma ED visits by tertile of pollutant concentration and AER Atlanta,1999 -2002

15 Conclusions Indication that short-term indicators of daily AER explain some heterogeneity in observed short-term risk  Supports interpretation that AER may affect intraurban exposure variability  potential exposure error in time series risk estimates AER indicators were relatively easy to obtain and include within an established analytical framework, thereby facilitating replication Limitations:  Estimated, rather than actual AERs used (median AER unusually low)  Simplifying assumptions may not be sufficiently spatially resolved  Uniform meteorology across geographic domain  Other ED outcomes? Other locations?

16 Limitations in Interpretation Estimated, rather than actual AERs used  Median AER for Atlanta domain unusually low  Validation of model required Simplifying assumptions may not be sufficiently spatially resolved  Uniform meteorology across geographic domain Other ED health categories? Other locations?  Contributes to understanding of AER as a factor affecting ambient pollutant infiltration, intraurban exposure variability, and possible exposure misclassification in health risk estimates in time series studies of air pollution

17 Acknowledgments This study was made possible by funding from: the US Environmental Protection Agency (CR 83407301-1 and EPA STAR RD834799) R834799 For complete results: Sarnat JA, Sarnat SE, Flanders WD, Chang HH, Mulholland J, Baxter L, Isakov V, Özkaynak H. “Spatiotemporally-Resolved Air Exchange Rate as a Modifier of Acute Air Pollution-Related Morbidity in Atlanta.” JESEE. 2013 DOI: 10.1038/jes.2013.32.

18

19 Results: O 3, PM 2.5 O 3  Ozone significantly associated with Asthma ED at all AERs  Highest RR’s at highest tertiles of O 3 level  No evidence of AER modification in either direction  Results contradict long-term epi results (Levy et al., 2005)  Possible explanation: Observed range of AERs not sufficient to elicit measurable differences in ozone fate and transport  AERs in epi models are not more parsimonious PM 2.5  On average,  RRs from ZIP codes with high AERs compared to ZIP codes with low (stratified analysis)  PM 2.5 – AER Interaction term negative and significant


Download ppt "JEREMY SARNAT, STEFANIE SARNAT, W. DANA FLANDERS, HOWARD CHANG, JAMES MULHOLLAND, LISA BAXTER, VLAD ISAKOV, HALÛK ÖZKAYNAK Annual Conference of ISES/ISEE/ISIAQ."

Similar presentations


Ads by Google