Grantee Research Project Results
2010 Progress Report: Measuring the Impact of Particulate Matter Reductions by Environmental Health Outcome Indicators
EPA Grant Number: R833627Title: Measuring the Impact of Particulate Matter Reductions by Environmental Health Outcome Indicators
Investigators: Johnson, Jean , Yawn, Barbara , Pratt, Greg
Institution: Minnesota Department of Health , Olmsted Medical Center , Minnesota Pollution Control Agency
Current Institution: Minnesota Department of Health , Minnesota Pollution Control Agency , Olmsted Medical Center
EPA Project Officer: Hahn, Intaek
Project Period: January 1, 2007 through December 1, 2011 (Extended to May 31, 2012)
Project Period Covered by this Report: January 1, 2009 through December 31,2009
Project Amount: $488,650
RFA: Development of Environmental Health Outcome Indicators (2006) RFA Text | Recipients Lists
Research Category: Air Quality and Air Toxics , Airborne Particulate Matter Health Effects , Air Toxics , Particulate Matter
Objective:
This report summarizes progress made on a collaborative research project to develop and evaluate a set of environmental health outcome indicators that can be used to measure and track the public health impacts of particulate matter (PM) reduction policies. Collaborators include the Minnesota Department of Health (MDH), the Minnesota Pollution Control Agency (MPCA), and the Olmsted Medical Center, Rochester Epidemiology Project (REP).
In Minnesota, a wide range of national and local air pollution reduction strategies have been implemented, starting in 2005 and continuing through 2009, including several strategies that were expected to reduce population exposure to PM. Data collected and analyzed to measure the public health impact of these strategies covers the study years 2000 through 2009. The study area is the Minneapolis-St. Paul seven county metropolitan area and Olmsted County, two areas in the state that have the most air quality alerts, influenced largely by regional air patterns from the southeast as well as local point and mobile sources.
PM reduction strategies that have been implemented and their current status include the following:
Summary of PM Reduction Strategy Implementation in Minnesota, 2005-2009
PM Reduction Strategy | Implementation Status |
---|---|
Minnesota Emissions Reduction Project (MERP): Coal-to-natural gas conversions of 2 energy plants; cleaner technology at 1 plant in Minneapolis-St. Paul metro area. | Alan S. King Plant, operational; Early 2008 High Bridge Plant, operational; May 2008 Riverside Plant, operational; 2009 |
Project Green Fleet | Diesel emission reduction retrofits completed for 1,100 metro area school buses; 2005-ongoing. |
The Metropolitan Council's Go Green Initiative | Ultra low sulfur fuel started mid-2005 for all metro area transit buses; new hybrid buses; diesel emission reduction retrofits started 2002-ongoing; fuel composition changes and anti-idling ordinance in effect (2007). |
Congestion Mitigation and Air Quality (CMAQ) grants | Emission reduction retrofits of public, heavy duty vehicles in 7-county metro area; 287 completed by late 2009. |
National Initiatives:
|
In effect, late 2006 In effect, late 2007 In effect, 2008 |
Progress Summary:
Work Progress and Accomplishments
The objectives of the research as originally proposed have not changed. Significant progress had been made with the collection and evaluation of available environmental and health outcome data from 2000 through 2007 and with developing the analytical methods for developing environmental outcome indicators.
1. Indicator Measures of Population Exposure to PM
Project investigators have collected the available PM2.5 data for study years 2000-2008 and identified the strengths and limitations with each data set for development of indicator measures. The available data sets included the daily BAM continuous PM mass monitoring data for 2002-2008, FRM monitoring data for 2000-2006 (1 in 3 day and 1 in 6 day), ambient emissions modeled data (CAMx) for model year 2005, and Hierarchical Bayesian (HB) modeled data provided by the EPA, for model years 2001 through 2006. The CAMx emissions model produces a 4 km grid and zipcode level surface for assigning population exposure. The HB modeled data, which combines monitored and modeled data, produces a 12km and 36km grid surface for estimating PM2.5 population exposure.
Criteria used to evaluate each data set included: temporal and spatial variability, missing data, data collection frequency, potential for measurement errors and misclassification of exposure, continuity of the method over time, availability and translation of the method to other locations, and ease of communication with stakeholders. We determined that the FRM monitoring data was the least suitable for the purposes of this project due to the large number of missing days of data. Daily measurements are necessary to local analyses to reduce exposure misclassification and allow the use of all health events, thus maintaining sample size necessary for precise risk estimates. Imputation or averaging would likely attenuate peak exposures which are important to this analysis, reducing daily variability in PM2.5 concentration and statistical power.
A close comparison of the remaining three data sources, all of which provide a daily population exposure estimate (PM2.5 continuous monitor, CAMx emissions model, and HB model data) in a time series display for 2005 revealed that on average the three data sources track well together, and identify similar daily peak exposures. The HB modeled data and the CAMx modeled data both tended to underestimate the daily peak concentrations, thus reducing daily variability, while the PM2.5 continuous monitor data provided the greatest daily variability. Two periods of missing continuous PM mass concentration data were observed over from 2002 -2007 in Olmsted County where a single central monitor is located due to periodic maintenance. Using the measured correlation between daily levels in the seven county metro area and Olmsted County, we were able to develop a regression model that used average daily PM2.5 mass concentration data from the seven county metro area to predict daily PM2.5 mass concentrations levels for the missing days of data in Olmsted County.
Project investigators conducted an evaluation comparing the three PM2.5 data sets available for the year 2005 in the Minneapolis-St. Paul metro study area in a case-crossover analysis with 2005 asthma hospitalizations to examine differences in the strength and precision of effect estimates. Also for evaluation purposes, we compared two approaches for assigning daily PM2.5 continuous monitor concentrations to asthma hospitalization cases: 1) using a nearest monitor approach, and 2) using a calculation of the average daily PM2.5 concentration across all monitors for the seven county metropolitan area. PM2.5 concentrations in the metro area counties show a homogenous local distribution, thus the use of one daily PM2.5 concentration averaged over the metro area from all monitors was the preferred approach. Relative to the other sources, daily PM2.5 mass concentration data (averaged over the metro area) provides the greatest variability in the exposure estimate, and likely is most representative of ambient concentrations experienced by the population working and living in the area, potentially reducing misclassification error.
For the case-crossover model to test the PM2.5 data sources we used a distributed lag model, the average of the daily lags (PM, Lag0, Lag1, and lag2), and adjusted for maximum temperature and relative humidity. Estimates of daily PM2.5 exposure based on the continuous PM2.5 mass monitoring data and one average value for the Twin Cities produced similar effect estimates with precision similar to the same model using the closest grid cell to zipcode centroid. This model produced a slightly stronger effect estimates at lag0 and lag1 than the CAMx emissions model data. The HB model data also produced a weaker effect estimate.
To examine within-city spatial resolution of PM2.5 exposure from major roadways, we began testing several metrics for traffic exposure in Olmsted County. More spatially resolved estimates of PM2.5 exposure are needed to coincide with greater spatial resolution of the health outcome data from the REP.
2. Indicator Measures of Health Outcomes
In 2009, we added available 2007-2008 health outcome data to the 2000-2006 data previously collected, including: asthma, chronic obstructive pulmonary disease (COPD), chronic lower respiratory disease (CLRD), total respiratory disease, and cardiovascular disease (acute myocardial infarctions) hospitalizations data. Total counts and rates of asthma, COPD, CLRD, and myocardial infarction hospitalizations and emergency department visits for 2000-2007 were determined.
We conducted descriptive analysis of data received from the Rochester Epidemiology Project for Olmsted County including: asthma, COPD and acute myocardial infarctions hospitalizations, as well as asthma exacerbations (measured as clusters of 3 asthma clinic visits in single patient over a 2 week time period) for years 2000-2007.
3. Indicator Measures of Association and Public Health Impact
Significant progress was made on the refinement of the case-crossover analysis using the C-CAT software. Because the health effects of PM exposure on a given day are spread out over several subsequent days, a time-stratified referent selection approach was used to assess the associations between various exposure lags (0-1, 0-2, 0-4 days) of daily 24-hour continuous PM2.5 monitoring data and total respiratory, asthma , and CLRD hospitalizations over a six year period, 2002-2007. Conditional logistic regression models that adjust for temperature, relative humidity, influenza epidemics, and nationally holidays were tested to compare effect estimates. Effect modification by age, sex, and season was also examined. Overall, significant effects were observed at lag averages (0-2) for CLRD hospitalizations, asthma hospitalizations, and total respiratory hospitalizations. We observed that odds ratios have decreased during the early years of reduction strategies implementation (2005-2007) compared with a baseline period of 2002-2004. Complete results of these analyses are under review for presentation and manuscript preparation in 2010.
Future Activities:
Plans for 2010 and 2011
- Collect additional years of available data through the end of the study period (2009) to include 2008-2009 PM2.5 monitoring data, HB modeling data, respiratory disease and MI health outcomes, PM speciation data and co-pollutants.
- Collect and test in C-CAT models added health outcome data sets to include: state ambulance report data for respiratory and MI outcomes, available through 2009, and COPD exacerbations (clinic visit clusters) data from the REP.
- Continue to refine C-CAT models, using continuous monitoring, HB data and covariates to include the addition of speciated PM components, refined covariates for influenza, seasonal data for ozone, and spatially assigned traffic exposure data.
- Conduct a spatially refined analysis in Olmsted County using available geocoded (zip+4) health outcomes and incorporating traffic exposure metrics
- Examine changes in associations and attributable fractions over the study period as a measure of public health impact.
Journal Articles:
No journal articles submitted with this report: View all 7 publications for this projectProgress and Final Reports:
Original AbstractThe perspectives, information and conclusions conveyed in research project abstracts, progress reports, final reports, journal abstracts and journal publications convey the viewpoints of the principal investigator and may not represent the views and policies of ORD and EPA. Conclusions drawn by the principal investigators have not been reviewed by the Agency.
Project Research Results
- Final Report
- 2011 Progress Report
- 2009 Progress Report
- 2008 Progress Report
- 2007 Progress Report
- Original Abstract
2 journal articles for this project