2009 Progress Report: Development of Exposure and Health Outcome Indicators for Those with Asthma or Other Respiratory Problems

EPA Grant Number: R833629
Title: Development of Exposure and Health Outcome Indicators for Those with Asthma or Other Respiratory Problems
Investigators: Meng, Ying-Ying , Jerrett, Michael , Ritz, Beate R. , Wilhelm, Michelle
Institution: University of California - Los Angeles , University of California - Berkeley
Current Institution: University of California - Los Angeles
EPA Project Officer: Nolt-Helms, Cynthia
Project Period: September 1, 2007 through August 31, 2010 (Extended to February 29, 2012)
Project Period Covered by this Report: November 7, 2008 through November 6,2009
Project Amount: $500,000
RFA: Development of Environmental Health Outcome Indicators (2006) RFA Text |  Recipients Lists
Research Category: Health Effects , Health

Objective:

This research will investigate the feasibility of combining existing environmental monitoring and health survey data (California Health Interview Survey [CHIS]) to develop health outcome indicators, such as asthma-related emergency department (ED) visits, doctor’s visits, absences from school/work, medication use, and frequency of asthma symptoms for those with asthma, as well as asthma-like symptoms, doctor’s visits and absences from work/school due to breathing problems for those without asthma diagnoses. 

We have focused our project activities to date on obtaining data access approvals, preparing data for exposure measures and data analyses, creating subject-specific air pollution exposure estimates for O3, NO2, PM2.5 and PM10 (12-, 24- and 36-month averages) and creating traffic density and proximity to roadways estimates using GIS software for all CHIS 2003 respondents and 2005 asthmatic respondents, and performing geostatistical modeling of O3, PM2.5, PM10, NO2, NO, and NOx for CHIS 2005 respondents in Los Angeles County. The project objectives are:

Objective 1. To develop long-term (12-, 24- and 36-month) criteria air pollutant exposure indicators for all CHIS 2003 respondents and 2005 asthmatic respondents using existing measurement data for O3, NO2, PM10, and PM2. 5 from the California Air Resources Board (CARB) air monitors.

Objective 2. To explore additional exposure indicators using geostatistical modeling (e.g., kriging for O3, and land use regression (LUR) for PM10, PM2.5,NOx, NO and NO2) for CHIS 2005 respondents in Los Angeles and traffic-related exposure indicators based on residential TD and proximity to roadways for all CHIS 2003 respondents and 2005 asthmatic respondents.

Objective 3. To examine whether health outcome indicators (e.g., ED visits or frequent symptoms) are associated with spatial and/or temporal variations in pollutant exposures while taking into account other risk factors, such as secondhand smoking.

We have almost accomplished Objectives 1 and 2 and have started to work on Objective 3 of the project. The project has been delayed due to an Institutional Review Board (IRB) issue related to the use of CHIS confidential data. In the beginning of 2009, we encountered an IRB protocol violation involving the use of CHIS confidential data, which has been satisfactorily resolved with the UCLA IRB; however, further delays in the access to CHIS confidential data have occurred due to the implementation of new security policies and procedures within the CHIS data access center (DAC). The CHIS DAC received IRB approval for their new policies and procedures at the end of 2009. We are working on a new DAC application to resume our project based on their new policies. None of the project objectives have been changed from the original application; the details of activities under each objective are in the next section.

Progress Summary:

We have developed several long-term (12-, 24- and 36-month) criteria air pollutant exposure indicators for all CHIS 2003 respondents and 2005 asthmatic respondents using existing measurement data for O3, NO2, PM10, and PM2. 5 from CARB ambient air monitors. Additionally, we have explored exposure indicators using geostatistical modeling (e.g., kriging for O3, and land use regression (LUR) for PM10, PM2.5, NOx, NO and NO2) for CHIS 2005 respondents in Los Angeles. We have found that ordinary kriging for O3 and universal kriging for PM2.5 and PM10 produced reasonable predictions on cross validation, while the scarcity of monitors precluded generating land use regressions for PM10 and PM2.5. We also have developed traffic-related exposure indicators based on residential traffic density and proximity to roadways for all CHIS 2003 respondents and 2005 asthmatic respondents. Lastly, we have updated our literature review on asthma, air pollution, and related susceptibility factors such as comorbidities, neighborhood environment, and socioeconomic status.

Work Status

Objective 1. To develop long-term (12-, 24- and 36-month) criteria air pollutant exposure indicators for all CHIS 2003 respondents and 2005 asthmatic respondents using existing measurement data for O3, NO2, PM10, and PM2. 5 from the CARB air monitors.

This objective is nearly completed. After running quality assurance checks on the exposure indicators we discovered an error requiring some of the annual pollutant averages to be rerun. Also, some exposure averages were missing. Due to IRB issues and new policies and procedures for the use of CHIS confidential data, we need to rerun the SAS code for those corrected and missing averages in the coming project year. The details are as follows:

Task 1: Obtain IRB and CHIS Data Disclosure Review Committee approval for CHIS 2003 and 2005 data access--Ongoing

We applied to the UCLA IRB and received approval to conduct this study. A separate application for access to the CHIS 2003 and 2005 data was submitted to the CHIS Data Disclosure Review Committee and approved. Due to the tightening of security standards for the CHIS data, the residential and school geocodes could no longer be removed from the secure data access center (DAC), which necessitated all exposure work to be conducted in the DAC. Though initially CHIS data access was granted, IRB issues and implementation of new CHIS Data Access Center policies and procedures for working with confidential data (e.g., guest researchers will not be allowed to directly use CHIS geocodes at the DAC) have necessitated a new application to be submitted for approval.

Task 2: Prepare CHIS 2003 and 2005 data file for data linkage and analyses--Completed

We worked with senior statistician (Hongjian Yu) and his staff to construct project data files and create asthma outcome variables (such as symptoms, days of school/work missed, etc). CHIS datasets for adults, teens, and children were merged together in order to look at outcome prevalence across the entire respondent population for CHIS 2003 and 2005. Variables also were constructed for important potential confounders and to define subpopulations who may have greater air pollution exposures, for example, subgroups by age, gender, race, income level, rural/urban residency and insurance status.

Task 3: Obtain air monitoring data for O3, NO2, PM10, and PM2.5 for 2000-2006--Completed

Air monitoring data were obtained from CARB. These data are checked internally by CARB, and we also looked for any systematically missing data.

Task 4: Obtain 2001-2005 traffic count data from Caltrans—Completed

Annual average daily traffic (AADT) count data for all major roads in California were obtained from the Federal Highway Performance Monitoring System (HPMS) program.

Task 5: Map CHIS 2003 and 2005 respondents’ residences, schools (CHIS 2005 only), and air monitors using GIS and check quality—Completed

We mapped residential addresses in GIS and checked for the accuracy of geocodes. We also looked at the proportion of respondents who were mapped based on address, residential cross-street, or zip code.

Task 6: Air pollution exposure measures--Ongoing

Dr. Michelle Wilhelm estimated annual averages for O3, PM10, PM2.5 and NO2 in SAS counting 12 months back from each CHIS 2003 respondent’s interview date for each year, for the 3 years prior to each respondent’s interview date. Based on feedback from our collaborators and other colleagues, the annual average for ozone was calculated using the daily 8 hour max value. In addition to creating an annual average ozone exposure measure, we also generated averages for the hot months (April-September) when ozone levels are highest. The following exposure measures were calculated for all CHIS 2003 and 2005 asthmatic respondents:

  • For ozone (O3), 12, 24 and 36-month averages using all months and using “hot” months (Apr-Sep) of 8 hour daily max values (calculated by CARB);
  • For nitrogen dioxide (NO2) 12, 24 and 36-month averages of daily (24-hour) averages (calculated using hourly data); and
  • For particulate matter (PM10 & PM2.5): 12, 24 and 36-month averages of daily (24-hour) averages.

After performing quality assurance checks on the exposure measures, we found errors in some of the pollutant averages, namely the use of an incorrect daily ozone variable to create the ozone averages and some missing averages. The SAS code already has been updated, and the corrected and missing averages will be run in the coming project year.

Objective 2. To explore additional exposure indicators using geostatistical modeling (e.g., kriging for O3, and land use regression (LUR) for PM10, PM2.5, NOx, NO and NO2) for CHIS 2005 respondents in Los Angeles and traffic-related exposure indicators based on residential TD and proximity to roadways for all CHIS 2003 respondents and 2005 asthmatic respondents.

We have accomplished the major part of this objective and will continue to finish the remaining work as follows:

Task 7: Traffic density and proximity to roadways estimates—Ongoing

  • For all CHIS 2003 and 2005 asthmatic respondents
  • Include school traffic for children (2005)

We have started computing traffic density and proximity to roadways estimates for CHIS 2003 and 2005 asthmatic respondents, but work has not yet been completed. Estimates for traffic near schools for children (CHIS 2005 only) have not been calculated yet. Due to the tightening of security standards for the CHIS data, the residential and school geocodes could no longer be removed from the secure data access center (DAC), which necessitated all exposure work to be conducted in the DAC. This delayed the calculation of exposure variables somewhat because our collaborators were required to travel to the campus to work with the data and to work within the times available at the DAC. In addition to computing traffic density using Federal HPMS data we also are exploring the use of Teleatlas traffic data, which would allow traffic exposures to be calculated for a greater proportion of study subjects.

Task 8: Geo-statistical modeling--Completed

For CHIS 2005 respondents with asthma in Los Angeles County, we proposed to explore additional exposure indicators using geostatistical modeling as follows:

  • O3 (Kriging or best possible exposure surface)
  • PM10, PM2.5, NOx, NO, NO2 (Land Use Regression or best possible exposure surface for PM10, PM2.5)

Our GIS subcontractor Zev Ross evaluated associations between existing air pollutant data (for 2004 and 2005) and predictor variables such as traffic on truck routes and industrial and government land use. To increase pollutant data available, all monitors within the Los Angeles metropolitan statistical area (MSA) were included in the analysis; this provided up to 33 monitors for PM10 and 24 for PM2.5. However, due to the number of monitors available and the relationships between pollutants and predictors reliable land use regression models still could not be developed for PM10 and PM2.5. Because the data did not support LUR models, kriging was performed for PM10, PM2.5 and O3. For the NOx, NO, and NO2 LUR models, our team collaborated with Michael Jerrett and his staff. Pollutant data from 201 samplers set up around Los Angeles as part of a CARB-funded study was used to develop the models. Traffic volumes, truck routes and road network, land use data, satellite-derived vegetation greenness and soil brightness, and truck route slope gradients were used for predicting pollutant concentrations. Cross validation analyses suggested a prediction accuracy of 87-91%. We linked these modeled estimates to CHIS 2005 respondents in Los Angeles.

Task 10: Descriptive and crude association analyses of outcomes and covariate data

After dataset and variable construction, we began descriptive analyses of outcome and covariate data. We have run basic frequencies to examine the prevalence of asthma symptoms, attacks, ED visits, and school absences in asthmatics and non-asthmatic wheezers. We also examined the distribution of these outcomes by various demographic characteristics and potential confounders, such as age, gender, income, health insurance status, urban/rural residence, smoking, poverty level and body mass index.

Objective 3. To examine whether health outcome indicators (e.g., ED visits or frequent symptoms) are associated with spatial and/or temporal variations in pollutant exposures while taking into account other risk factors, such as secondhand smoking.

We updated our literature review on asthma, air pollution, and related susceptibility factors such as comorbidities, neighborhood environment, and socioeconomic status. This review of recent literature will help to inform the statistical analyses for this project and will be cited in the final report and journal publications.

Future Activities:

We will investigate spatial and temporal links between the exposure indicators and health outcome indicators, such as asthma-related ED visits, doctor’s visits, absences from school/work, medication use, and frequency of asthma symptoms for those with asthma, as well as asthma-like symptoms, doctor’s visits and absences from work/school due to breathing problems for those without asthma diagnoses after adjusting for other factors. We also will identify whether the types of health outcomes (e.g., prevalence of persistent asthma or prevalence of asthma-like symptoms) and the magnitude of associations estimated for our air pollution exposure indicators differ by type of area (e.g., rural and urban) and by subpopulation (e.g., children and the elderly, or racial and ethnic groups).

Once our new CHIS DAC application based on their new policies and procedures for confidential data is approved and we are able to access the confidential data again, we will complete the remaining exposure measures for the study, specifically the traffic density measures and some remaining air monitoring averages that need to be rerun. Then, exposure measures will be merged with the CHIS health outcome and covariate data. With the combined data, we will investigate spatial and temporal links between the exposure indicators and health outcome indicators, such as asthma-related ED visits, doctor’s visits, absences from school/work, medication use, and frequency of asthma symptoms for those with asthma, as well as asthma-like symptoms, doctor’s visits and absences from work/school due to breathing problems for those without asthma diagnoses after adjusting for other factors. We also will identify whether the types of health outcomes (e.g., prevalence of persistent asthma or prevalence of asthma-like symptoms) and the magnitude of associations estimated for our air pollution exposure indicators differ by type of area (e.g., rural and urban) and by subpopulations (e.g., children and the elderly, or racial and ethnic groups).

Journal Articles:

No journal articles submitted with this report: View all 3 publications for this project

Supplemental Keywords:

Indoor air, mobile sources, risk, health effects, ecological effects, vulnerability, public policy, decision making, public good, Bayesian, socio-economic,  epidemiology, modeling, analytical, measurement methods, Northwest, California, CA, EPA Region 9

Progress and Final Reports:

Original Abstract
  • 2008 Progress Report
  • 2010 Progress Report
  • 2011
  • Final Report