' {Hi-‘88 r A ct alw- This is to certify that the thesis entitled Effect Of Outdoor Air Pollution On Hospital Admissions For Asthma In Detroit, Michigan presented by Alireza Sadeghnejad has been accepted towards fulfillment of the requirements for the MS. degree in Epidemiolgy (M605 L/J Major Professor’s Signature Z/L‘f/O 5‘ Date MSU is an Affirmative Action/Equal Opportunity Institution .-._.-----.-.-u—o----.--—.—---o-n-o-o-n-o-o-.-o—-.-u-u----—o—o—-o—o-o--u-u-c-—n—o--o-—c—-—o-o-¢—-—¢-o-c---o-a--- LIBRARY Michigan State University PLACE IN RETURN BOX to remove this checkout from your record. TO AVOID FINES return on or before date due. MAY BE RECALLED with earlier due date if requested. DATE DUE DATE DUE DATE DUE 6/01 c:/CIRC/DateDue.p65-p. 15 Effect Of Outdoor Air Pollution On Hospital Admissions For Asthma In Detroit, Michigan By Alireza Sadeghnejad A THESIS Submitted to Michigan State University In partial fulfillment of the requirements for the degree of MASTER OF SCIENCE DEPARTMENT OF EPIDEMIOLOGY 2004 ABSTRACT EFFECT OF OUTDOOR AIR POLLUTION ON HOSPITAL ADMISSIONS FOR ASTHMA IN DETROIT, MICHIGAN By Alireza Sadeghnejad We investigated the spatio-temporal relationship between number of asthma hospital admissions and levels of air pollutants ozone, particulate matter, nitrogen dioxide and sulfur dioxide. For the period 1999-2000, data were obtained on daily asthma hospital admissions for a contiguous region covering 23 zip codes in the East Seven Mile and Linwood areas of Detroit, Michigan. Each zip code falls within a 4-kilometer radius of an air quality monitoring station that provided detailed data on the air pollutants and meteorological assessments. Exposure to a pollutant was assessed based on its mean daily level in the 4-day period preceding a hospital admission. Linwood and East Seven Mile areas were predominantly African-American (about 66%). Mean daily admission rates per 100,000 for asthma were 1.4 in Linwood, and 1.1 in East Seven Mile. The month of September showed a very sharp increase in admissions. In a negative binomial model, we estimated an average of 8% increase in the number of daily asthma hospital admissions by 6 ppb increase in nitrogen oxide levels. We observed a significant protective effect for ozone. The levels of nitrogen dioxide and ozone were negatively correlated. Higher levels of nitrogen dioxide might increase asthma hospital admissions. Individual level data are needed to verify air pollutant effects on asthma. DEDICATIONS This work is specially dedicated to my parents and my wife Negin. iii ACKNOWLEDGEMENTS The planning, designing, implementation and completion of this work is the result of a team of dedicated individuals. Sincere gratitude to my thesis advisor, Dr. Wilfried Karmaus, for his insight, supervision and moral support in the process of completing this thesis. I would also like to thank the other members of my thesis committee, Drs. Joseph Gardiner and Rober L. Wahl for taking time from their busy schedule to provide support during this project. iv TABLE OF CONTENTS LIST OF TABLES ........................................................................................... vi LIST OF FIGURES ........................................................................................ vii LIST OF ABBREVIATIONS ........................................................................... ix BACKGROUND .............................................................................................. 1 METHODS ...................................................................................................... 8 RESULTS ..................................................................................................... 16 DISCUSSION ........................................................................................................... 34 APPENDIX .................................................................................................... 4O BIBLIOGRAPHY ........................................................................................... 43 LIST OF TABLES TABLE 1. Zip codes contained in each of the two sites ................................. 9 TABLE 2. Correlation among various pollutants in the two sites during specified periods ........................................................................................... 20 TABLE 3. Means and percentiles for daily average levels of pollutants during the study period ............................................................................................ 21 TABLE 4. Means and percentiles for daily meteorological variables during the study period .................................................................................................. 22 TABLE 5. Statistics for daily hospital admissions in East 7 Mile ................... 26 TABLE 6. Statistics for daily hospital admissions in Linwood ....................... 26 TABLE 7. Demographic characteristics, site of residence and year of admission for census data along with hospital admissions ........................... 27 TABLE 8. Mono-pollutant models'r (802, N02 or 03) controlling for year and/ or month of admission ................................................................................... 31 TABLE 9. Tri-pollutant model (802, N02 or 03) controlling for year and month of admission .................................................................................................. 32 TABLE 10. Bi-pollutant modelT (N02 and 03) controlling for year and month of admission ...................................................................................................... 33 vi LIST OF FIGURES FIGURE 1. N02 in the pathway of 03 production ........................................... 9 FIGURE 2. The two stations and a 4-kilometer radius around them ............... 8 FIGURE 3. Different steps for merging the four data sets to get working data set ................................................................................................................. 13 FIGURE 4. Distributions of pollutants during the study period ...................... 18 FIGURE 5. Mean monthly levels of pollutants in East 7 Mile during study penod ............................................................................................................ 19 FIGURE 6. Mean monthly levels of pollutants in Linwood during the study pefiod ............................................................................................................ 19 FIGURE 7. Variation of mean daily levels of pollutants during week in East 7 Mile in the two years of study ........................................................................ 23 FIGURE 8. Variation of mean daily levels of pollutants during week in Linwood in the two years of study ................................................................. 23 FIGURE 9. Monthly averages for the mean daily temperature and the mean daily relative humidity in Detroit area during study period ............................. 24 FIGURE 10. Distribution of total daily asthma admissions during 1999-2000 in Detroit, East 7 Mile .................................................................................... 26 FIGURE 11. Distribution of total daily asthma admissions during 1999-2000 in Detroit, Linwood ........................................................................................ 26 FIGURE 12. Total number of asthma hospital admissions by age group and gender during the study period ..................................................................... 28 vii FIGURE 13. Monthly number of asthma hospital admissions in the two sites during the study period ................................................................................. 28 FIGURE 14. Asthma hospital admissions by day of the week during the study pefiod .................................................................................................. i .......... 29 FIGURE 15. Average number of daily asthma hospital admissions for different levels of N02 and 03 ..................................................................................... 39 viii LIST OF ABBREVIATIONS CO ........................................................................................ Carbon Monoxide Cl ....................................................................................... Confidence Interval ED ............................................................................... Emergency Department MDCH ............................... ' ........... M ichigan Department of Community Health MDEQ ..................................... Michigan Department of Environmental Quality NA ................................................................................................ Not Available N02 ................................................................................................... Nitrogen Dioxide NOx .................................................................................................... Nitrogen Oxides 03 ........................................................................................................... Ozone P5 ............................................................................................................ 5th Percentile P50 ......................................................................................................... Median P95 .............................................................................................. 95th Percentile PM2,5 .................... Particulate Matter with a diameter less than 2.5 micrometer PMm ..................... Particulate Matter with a diameter less than 10 micrometer ppb .......................................................................................... Parts Per Billion ppm ......................................................................................... Parts Per Million SAS ...................................................................... Statistical Analysis Software 802 ............................................................................................. Sulfur dioxide VOCs .................................................................. Volatile Organic Compounds u ............................................................................................................... Mean pg/m3 ..................................................................... Microgram per cubic meter ix BACKGROUND Asthma prevalence has been increasing since the mid 19703 1 and has emerged as a major public health problem over the past 20 years in the United States 2. The overall rate of hospitalization for asthma increased during the late 19805 and has since plateaued. However, the rate among African Americans remained 2-3 times higher than for white Americans 2. Air pollution is considered as a risk factor for asthma hospital admission. Asthmatics appear to be more susceptible to short-term peak concentration of air pollutants 3. Research has strongly shown that air pollution increases asthma hospitalization through exacerbation of attacks in asthmatics. We tabulated previous published studies (Appendix). The following studies are from the North America. In 1993, a study conducted over a 13-month period in Seattle reported that the relative risk of asthma emergency room visits for a. 30pg/m3 increase in particulate matter PM10 was 1.12 (95% confidence interval: 1.04, 1.20) 4. They noted that the mean of the previous 4 days’ PM10 was a better predictor than shorter lag periods. In addition for the number of asthma emergency room visits, an evident peak was observed during September. In New Brunswick during 1984-1992, for the period May-September, Stieb et al. examined the relationship between asthma emergency department visits and air pollutants nitrogen dioxide (N02), ozone (03), and sulfur dioxide (802) Daily emergency department (ED) visit frequencies were filtered to remove day of the week and long wave trends. Filtered values were regressed on air pollution and weather variables for the same day and the 3 days previous to the ED visits. They found a positive relationship for higher levels of 03 and the increased number of ED visits, but not for other pollutants 5. For the period 1987-1994 in Seattle, Washington, Sheppard et al. reported the effects of ambient air pollution on non-elderly asthma hospital admissions 6. In a Poisson regression model controlling for time trends, seasonal variations, and temperature-related weather effects, they regressed daily hospital admissions on levels of 03, particulate matter with a diameter, both less than 2.5pm and less than 10pm (PM2_5 and PM1o) and $02. An estimated 4-5% increase in the rate of asthma hospital admissions associated with an interquartile range increase PM“) and PM2,5 levels with one-day lag (19.0 ug/m3 for PM1o and 11.8 jig/m3 for PM2,5). Similar findings for carbon monoxide (CO) and 03 but not 802 were observed. Correlations between levels of pollutants were: +0.8 for (PM10, PM2.5)-CO, -0.23 for O3-PM25, +0.34 for 03-802 and +0.22 for PMzs-sozs. In 2000, Tolbert et al. reported Pediatric emergency room visits for asthma in relation to 03, PM10, N02 in Atlanta, Georgia during the summers of 1993-1995 7. The estimated relative risk per 20 parts per billion (ppb) increase in the maximum 8-hour 03 level was 1.04 (p < 0.05). The estimated relative risk per 15 jig/m3 increase in PM“) was 1.04 (p < 0.05). Exposure-response trends (p < 0.01) were observed for ozone (>100 ppb vs. <50 ppb: odds ratio = 1.23, p = 0.003) and PM1o (>60 ug/m3 vs. <20 jig/m3: odds ratio = 1.26, p = 0.004). In models with ozone and PM10, both terms became nonsignificant because of collinearity of the variables (r=0.75). Correlation between pairs of pollutants were: +0.51 for O3- Nox and +0.44 for PM1o-NOX 7. In 2001, Ritchie et al. studied 1 to 17 year old children admitted to one of 20 hospitals within nine counties in the Indianapolis metropolitan area. For warmer months (May-September), during 1997-1999, they found that as O3 concentrations increased, asthma hospitalization probability decreased. During the study period, the mean for 24-dain 03 concentration was 0.038 ppm and the daily one-hour maximum was 0.066 ppm 8. Although study design varies in previous studies, but in almost all of them, exposure was allocated by aggregative method. The period that has been used to assess exposure prior to admission (lag period) varied between 1-5 days. For all pollutants these studies found an adverse or no effect on asthma hospital admissions. The only exception was 03. Some studies, including Ritchie et al. reported a protective effect of 03 on the number of asthma hospital admissions 8' 11. Correlation coefficients between pairs of pollutants were reported as positive in previous studies. The only exception to this was 03 that showed both positive and negative correlation with other pollutants 6'10. Most of the above-mentioned studies have demonstrated seasonal patterns in hospitalizations associated with asthma. A Canadian study examined the seasonal patterns of asthma hospitalizations for a 15—34 year age group, and found that hospitalizations peaked in the autumn season 12. Marked differences between the number of asthma hospitalization for males and females have been reported in the literature, with admissions for young males being higher than for young females 13. The only previous study conducted in Michigan to investigate the relationship between the daily air pollution levels and asthma hospital admissions was by Thorell. He examined the relationship between the daily air pollution levels and occurrences of asthma hospitalization as well as emergency department visits at Hurley Medical Center, Flint, MI. The study was limited to children under age 16 residing in ten zip codes in Flint. He found increases in emergency department visits and hospitalization when 03 levels increased by 135.1 jig/m3 above the mean daily maximum. There was also an increase in emergency visits when 802 levels increased by 21.8 jig/m3 above the mean daily maximum 14. In 1998 the Michigan Department of Community Health (MDCH) reviewed inpatient hospital records and death certificates for children less than age 15 with asthma diagnoses for the period 1985-94 15. The overall annual state childhood asthma hospitalization rate for this period was 34.3 cases per 10,000 children, with much higher rates for African American children. Wayne County (with an annual rate of 53.7/10,000) was among a group of counties with rates above the overall annual state rate for childhood asthma. The highest Detroit rates were for those children residing in zip codes. 48208, 48201, 48202, 48206, 48226, and 48238. lngham County had an annual hospitalization rate lower than the state as a whole (22.5/10,000). ' They were included in this thesis. The purpose of this aggregative study is to investigate potential associations between different air pollutants (N02, 03, PM2,5 and 802, as described below 16) and asthma hospital admissions. In order to conduct this investigation, .we used four data series (air pollution, meteorological, hospitalization and census data) on two geographical areas in Detroit, Michigan. The units of analysis were sites of residence stratified by age, gender, and race. Institutional Review Board at Michigan State University approved us to work on these existing data sets. Nitrogen dioxide (NO;) 16 Nitrogen oxides (NOx) are byproducts of fuel burned at high temperatures, as in a combustion process. The primary sources of NOx are motor vehicles (49%), electric utilities (27%), and other industrial, commercial, and residential sources that burn fuels (24%). In a complex reaction, NO reacts with volatile organic compounds WOCs) to produce N02 (Figure 1). Then N02 and oxygen reaction is catalyzed by sunlight to produce NO and O3. NO + VOCS —> N02 Sunlight I N02 + OzaNO + 03 Figure 1. NO; in the pathway of 0; production Ozone (03) Within the scope of this thesis, ozone refers to ground level (tropospheric) ozone, such as smog, and not to stratospheric ozone. Stratospheric ozone is the layer of ozone gas in the upper atmosphere that screens out harmful ultraviolet radiation from the sun. Ground level ozone is not emitted directly into the air, but is formed through complex chemical reactions between precursor emissions VOCs and nitrogen oxides (NOx) in the presence of sunlight (Figure 1). VOCs are emitted from sources such as automobiles, dry cleaners, and paint shops. NOx, as stated earlier come from sources including coal-fired power plants and motor vehicles. The chemical reactions that produce 03 are activated by sunlight and high temperature; therefore peak O3 levels occur mostly during the summer when the weather is warmer and in the middle of the day as emissions build up and the temperature rises 16. Typically, the length of the 03 season is May through October, coinciding with the warmer months of the year. Since varying meteorological conditions influence ambient levels and year-to-year trends, 0;; monitoring seasons may vary from one area of the country to another. Particulate Matter (PM) Particulate matter includes dust, dirt, soot, smoke and liquid droplets that are directly emitted into the air from sources such as windblown dust, automobiles, construction sites, factories, and fires. PM is also formed in the atmosphere by condensation or by the transformation of emitted gases such as SOz, NOx, and VOCs. Particulate matter is distinguished by it diameter. Fugitive sources such as agricultural tilling, construction, fires, and unpaved roads contribute much more PM10 (PM with a diameter less than 10pm) emissions in specific regions than others (this include dry forested areas susceptible to fire and agricultural areas). Sulfur dioxide ($02) $02 is a gaseous product from stationary and mobile sources burning coal and oil-containing sulfur. Processes found in pulp and paper mills and in nonferrous metal smelters also contribute to $02. The largest contributors to $02 emissions are coal-burning power plants. Once released, $02 and other oxides of sulfur combine with oxygen to form sulfates, and with water vapor to form aerosols of sulfurous and sulfuric acid. This mixture is a precursor of acid rain. Many emissions originate from tall stacks enabling them to be dispersed according to the pattern of the wind and variable wind speed. For example, Vermont’s air quality is partly affected by emissions carried from more industrialized areas both close by and far away. Sulfur compounds also contribute to visibility impairment in areas other than the primary source area. METHODS First, we will define the two geographical sites included in this study. Then population and the four data sets of study will be addressed. After defining exposure, outcome and potential confounders, we will describe methods proposed for descriptive and regression analysis in “Statistical analysis”. Information from Linwood and East Seven Mile, two geographical areas in Detroit, Michigan was used in this study. Each site was defined by an air pollution monitoring station and zip codes that were wholly or partially contained within a 4-kilometer radius of the air pollution monitoring station (Figure 2). The 4- kilometer radius around the East Seven Mile and Linwood sites wholly or partially contained 10 zip codes and 13 zip codes respectively (Table 1). Linwood Figure 2. Linwood and East 7 Mile air monitoring stations at center and a 4- kilometer radius around them (circles) Numbers represent zip codes .‘fl : Inter-state highway --— : Zip code boundaries Table 1. Zip codes contained in each of the two sites East 7 Mile Linwood 48015 48214 48201 48209 48021 48224 48202 48210 48089 48234 48203 4821 1 48091 48204 48216 48205 48206 48226 48212 48207 48238 48213 48208 Studv population and data Residents of the two sites at who aged one to 45 years’ people comprised the study population. We used data on census, air pollution, meteorological indices and hospital admissions in this study. Census data was obtained from the United States Census Bureau, the later three data sets were provided by MDCH and Michigan Department of Environmental Quality (MDEQ). Census data By assuming that the population was steady during the two years of study period, we used year 2000 census data as the reference. Data was available online through the United States Census Bureau 17. We downloaded census data for the desired zip codes by race, gender and age in years and converted the data to SAS (Statistical Analysis Software) format for the two sites. Air pollution data The Michigan Air Sampling Network measures air quality throughout the state. The pollutants that we used were nitrogen dioxide (N02), ozone (03), particulate matters with a diameter less than 2.5 micrometer (PM2.5) and sulfur dioxide (80;). Pollutant concentrations were monitored using a direct reading instrument by standardized methods. These methods and their units were: . Gas-phase chemiluninescence.Ppb for N02 0 Ultra violet analysis, ppb for 03 . R&P 2025 sequential sampler for PM2.5 lgravimetric method, pg/m . Ultra violet fluorescence, ppb for $02 Hourly measurements were recorded and electronically sent to the Air Quality Division of MDEQ. In the two air monitoring stations, air pollution data has been incompletely collected from January 1, 1999 through December 31, 2001. For each site, MDCH calculated daily mean and maximum levels of 802, N02 and 03 as well as daily PM2,5 values. We used mean daily levels of the pollutants and built up a SAS file. Meteorological data Both sites shared the same daily measurement for minimum and maximum temperatures as well as mean daily relative humidity during 1999 and 2000. This information was gathered in Detroit-Linwood station. We made a SAS formatted data from this information. Hospital admission data Information from almost all Michigan hospitals is included in the Michigan Inpatient Database. We received a file with the data on the number of patients with a discharge diagnosis of asthma. Data structure was based on the date of admission, zip code of residence, age in years, gender and race. We used data from January 1, 1999 till December 31, 2000 based on the date of admission. We organized final hospital admission data set as a SAS-formatted data set by date of admission, site (versus zip code in the original data set) of residence, age in years, gender and race. 10 OutcomeLexposgre and confounders Outcome We defined outcome as the number of patients admitted to Michigan hospitals, per day and by the site of residence, age, gender and race. Outcome quantification was based on the following criteria: . Patient age, between one year and 45 years . Diagnosis of asthma at discharge (code 493 in the International Classification of Diseases, version 9) 0 Use of date of admission for calculating number of daily asthma hospitalization Exposure We assessed exposure to N02, 802, PM2,5 and 03. Using an aggregative method we allocated exposure to individuals. Residents in zip codes that completely or incompletely felt into a 4-kilometer radius of each air monitoring station were assumed to have the same amount of exposure to the pollutants. There were two zip codes that felt in both sites. We assigned each of these zip codes to the site that predominantly encountered it. In previous studies exposure to pollutants was assessed using different lag periods between daily levels of pollutants and subsequent hospital admissions. Previously lag periods with a range between one day and 5 days were used, however there is no universally agreed standard. In this study we used mean level of pollutants during preceding four days of admissions 4'5'9'18. Confounders Variables that could be related to the level of air pollutants as well as daily number of hospital admissions assumed to be year, month and the weekday of 11 admission; race, age and gender of admitted patients; and meteorological indices 5,7,9 For each of the meteorological variables (minimum, maximum and mean temperature as well as mean relative humidity), we calculated an average in the preceding 4—day period of admission and controlled for that in the models. Statistical analysis We used SAS software, Release 8.2 ‘9 to conduct statistical analysis. In this part, we present the way to get a working data set from the four data sets (Figure 3). Then methods for descriptive analysis and regression will be explained. Finally, we will discuss modeling strategies. Working data set We created a working data set (Figure 3) by merging hospital admission, census, meteorological and air pollution data sets. Following are the steps of getting final data set 1. A data frame (linkage file) made to resemble all potential combinations of sites and dates of admission as well as age groups, gender and racial groups of admittees. We used this data frame as a base for merging the four data sets. 2. Hospital data set was merged with the linkage file by site, date, age group, gender and racial group. 3. Census data set was merged with the data from step 2 by site, age group, gender and racial group. 4. As both sites shared the same meteorological data, we assigned information to both sites and then merged it with the data set from step 3 by site and date. 5. Air pollution data set was merged by site and date with the data set derived from step 4. 6. At the last step, we calculated average level for each of the pollutants and meteorological indices in the 4 preceding days of admission in the final data set (Figure 3). 12 Linkage file: Containing all values for variables site, date, age, gender and race: 2 sites x [(2years x 365)] days x 5 age groups x 2 genders x 3 racial groups Variable 'Site’ included ‘ _ the hospital data. ing the two datasets by site, date, Census data: . .e, gender and race Numbers of residents in each zip code by age, gender and race .. ariable ‘Site’ included in -censu (merging by site and date).j_fa,,d m Air pollution data: _ _ _ Organized by site and Adding an pollutIon data date (merging by site and date) [Calculating previous four days’ averages | Figure 3. Different steps for merging the four data sets to get working data set Descriptive analysis We investigated distribution as well as spatial and temporal trends of air pollutants, meteorological indices and hospital admissions. Spearman correlations were calculated between daily levels of pairs of pollutants. The two sites were compared on their pollutant levels and demographic characteristics. Regression analfiis The outcome variable was the count of hospital admissions that was calculated by site, day, age group, gender and racial group. We modeled the outcome variable as a Poisson variable (and negative binomial) and used log linear regression analysis to assess relative risk of hospital admission according to the pollutants level. Formally let Y= number of hospital admissions and x =(x1,...., xm) being set of relevant covariates (exposure + potential confounders). The Poisson assumption is: P(Y=k|x) = 6““) my)“ / kl, k=0, 1, where p(x)=E(le) The log-linear model would be: log p(x) = (30+ 81x1 +...+ Bme In our analysis we should include an offset term which represents the population at risk (=N(x)). So, the modification to the above equation is: '09 ”(1) = '09 Ml) + 130+ BIXI +---+ I3me Parameter interpretation We modeled E(Y/N(x) Ix) = u(x) /N(x). The analysis of the incidence index with a Poisson or negative binomial regression allows the estimation of Relative 14 Risks (RR), which are equal to the ratio y; /yo where y,- is the incidence of admissions in level j, and yo corresponds to the reference (first) level 2°. Assessingthe fitted model GENMOD procedure in SAS was used to fit log linear models. We applied deviance and scaled Pearson X2 statistics in order to gauge adequacy of fit. The scaled values should be close to one 21. The reason for the inadequacy of fit could be due to over-dispersion (a condition that occurs when variance of the distribution is larger than its mean). If the fit seems inadequate one should use the negative binomial method instead of Poisson to incorporate over-dispersion 20. In order to choose the appropriate method of analysis we looked at the ' distribution of the outcome variable and criteria for ‘Goodness of fit’. (Y~ NB E(Y)=u Variance(Y)= u+ku2, k>0). This also provides an approach to test Ho : k=0( the alternative hypothesis, Ha : k>0). Statistically this test must be performed carefully because the null hypothesis places the dispersion parameter on the boundary of the parameter space 2°. Modeling strategie_s Average of pollutants and meteorological indices over a 4-day period preceding hospital admissions were ranked by their quartiles. Fully saturated models were run with only one pollutant in each along with all the potential confounders. If the risk ratio was significant for the pollutant, backward elimination was used to get the most parsimonious model. While rate ratios of pollutants in several reduced single-pollutant models show significance, models with combinations of pollutants were run. 15 RESULTS The first two part of this section will focus on descriptive analysis of air quality, meteorological, hospital admission and census data. In the third part, results of regression analyses will be provided. Descriptive analyses of air quality and meteorological data In this part, we present distribution of mean daily levels of N02, 302, PM2,5 and 03 along with their monthly variations and their variations upon weekdays. Correlations among daily levels of pollutants are; also provided in this part. For meteorological data, statistics in terms of mean and percentiles along with monthly variations of minimum, mean and maximum temperature as well as relative humidity investigated. Distributions of daily mean levels of pollutants were generally skewed to the left (Figure 4). This was more prominent for $02 and PM2,5. O3 possessed the most symmetrical distribution in comparison to the other pollutants. N02 and 802 levels were available for all months of the study period in the two sites (Figure 5 and Figure 6). The yearly measurement period for 03 was April through September in both sites. In East Seven Mile, PM2,5 was not measured during first three months of each year. Monthly variation was present for all the four pollutants. While 03 levels showed obvious peaks in June and July, concurrent drop of mean monthly N02 levels in these two months was evident. Mean monthly levels of $02 and PM2,5 did not resemble a pattern. 16 Mean daily levels of pollutants varied during week (Figure 7 and Figure 8). On weekends (Saturday and Sunday), there was a decrease in N02 levels and an increase in 03 levels. 802 and PM2,5 were almost stable during the week in Linwood, but not East Seven Mile. The highest correlations among pollutants were observed in Linwood between N02 and 802 during cold months (Table 2). Generally, N02, PM2,5 and 802 were significantly correlated. Significant correlations between 03 and PM2_5 were found. 03 was not correlated significantly with N02 or $02. All pollutants were significantly correlated with PM2,5. Mean daily levels of N02 were higher in colder periods of the two years in the two sites (Table 3), a situation that was true in the case of 802 just in East Seven Mile. PM2,5 also had higher mean daily levels in colder period of year. In Linwood, mean daily levels of 03 were lower and of N02 were higher than their levels in East Seven Mile. The year 2000 was colder and more humid than 1999 in the Detroit area (Table 4). The warmest month in the study period was July and the coldest, January (Figure 9). 17 5°l 10‘ Percent N02 % PPb 5O 30 2O 1O 010203040506070— Percent 03 REEL ‘Lijighn‘ 010 "20130 40 50 60 7o 50 2O 1O 50 30 20 10 O f Percent W so2 Ppb 010203040506070 Percent PM2.5 119/ m3 l II I‘r‘fl: 010203040506070 Figure 4. Distributions of pollutants during the study period Abbreviations; ppb, part per billion; ug/ m3, microgram per cubic meter; N02, nitrogen dioxide; 03, ozone; PM”, particulate matter with a diameter less than 2.5 micrometer; SOZ, sulfur dioxide 18 IN02 (ppb) IPM2.5 microgram/ m3 ISOZ (ppb) EIO3 (ppb) ppb- microgram/m3 Jan Feb Mar Apr May Jun Jul Aug Sep Oct Nov Dec Figure 5. Mean monthly levels of pollutants in East 7 Mile during the study period Abbreviations; ppb, part per billion; ug/ m3, microgram per cubic meter; N02, nitrogen dioxide; 03, ozone; PM”, particulate matter with a diameter less than 2.5 micrometer; SOz, sulfur dioxide 40-———_ IN02 (ppb) IPM2.5 microgram/ m3 35_ I802 (ppb) C103 (ppb) 30 I: I U 20 15 10 ppb- microgram/m3 Jan Feb Mar Apr May Jun Jul Aug Sep Oct Nov Dec Figure 6. Mean monthly levels of pollutants in Linwood during the study period Abbreviations; ppb, paIt per billion; 119/ m3, microgram per cubic meter; N02, nitrogen dioxide; 03, ozone; PM”, particulate matter with a diameter less than 2.5 micrometer; SOz, sulfur dioxide Table 2. Correlation among various pollutants in the two sites during specified periods Spearman correlation coefficient (P value) Number of days Site name Period Pollutant N02 PM2,5 $02 03 East7Mile 1.00 N02 344 0.14 (.36) 1.00 PM2-5 40 41 Apr- Sep 80 0.41(<. 001) 0.31(.07) 1.00 2 315 35 319 O —0.10(.05) 0.47 (.001) 0.15(.006) 1.00 3 337 41 314 356 1.00 N02 208 Oct- Dec 0.63 (<. 001) 1.00 Jan- Mar PM” 27 29 0.55 (<. 001) 0.30 (.10) 1.00 802 204 28 239 Linwood 1.00 N02 249 0.54 (<. 001) 1.00 A r_ Se PM2-5 211 259 p 9 SO 0.57(<.001) 0.51(<.001) 1.00 2 249 259 366 O —0.10(.10) 0.52 (<. 001) —o.02 (.64) 1.00 3 249 259 360 360 1.00 NO?- 267 Oct- Dec 0.51(<. 001) 1.00 Jan- Mar PM2-5 248 255 0.69 (<. 001) 0.70 (<. 001 ) 1.00 SO?- 267 255 365 Abbreviations: N02, nitrogen dioxide; 03, ozone; PM”, particulate matter with a diameter< 2.5 pm 20 o_nm__m>m .0: .<2 Eo____n Log tma dag 62x06 5:8 .Nom ”E1 mN vLoCoEmB m 5;) 5sz $932th .32.”. ”3:03 50 62x06 comet: .NOZ ”2:583 53 noon. €662). .Sn. 65:30.3 5m 5n. Enos. .1 chozm_>oonn< . . . . . . . . . . . . 80 .50 <2 <2 <2 <2 mm; Nv mo Nm oom owe wv mt now ©VN vow va :22 rcmfi BR 582:: now m.mN md SN 98 N.m v.0 Wm ANN ON? 0v 52 NS wéN v.9 QNN amm Lad. . . . . . . . . . . . own -80 <2 <2 <2 <2 :2 mm vo mm mm m? we Nor m I. NVN o3 va .65. ram... 89 582:: m5. m.mN o. 3 mKN v.5 o6 m.o Qm ode Q? N6 63 Qmm N. N ozm N. N 9.5 na< <2 <2 <2 <2 9: 0m Yo Nb adv 3 wd 0.9 5mm O; Na m; omD -80 N N 55. -cmw mar; m.©N fin— NNN v.3 Em Né me m.mN Q: On Q? MEN 5: No 09 amw I._Q< OOON E52 5 .w . . . . . . . . own. -80 <2 <2 <2 <2 m m? w n v o m m <2 <2 <2 <2 N mm m m: N m N 0N Ems— ICE. 3.: 8.5 k .m 9% Wow Q: va Om o.m 20 EN <2 <2 <2 <2 o.mN NSF EN 5.9 Qmw 1.94 mm on m 1 mm on m 1 mm on m 5‘ mm on m l m m d Q Q m a Q Q a Q o. oozed 53> \95 see mo 3%: 4.0m 8553 32o 39: N02 9:3 292.8 noted >033 05 acts—u 3532.3 .0 £05. 3235 >=nu .5. 3:23.83 new acne—2 .n anh 21 35:30:03 5mm “83 £06.22 .83 ”35:00:33 Em 53 Ewes. .1 ”m:o_0m_>o:00< . . . . . . . . . . . . . . . . 030-000 090 0N5 000 N2 000 000 000 004 000 000 N2 000 0N0 000 00 00N :m 57:0... 0.00 0.: 0.00 0.0K 00 0.0L 0.5 0.8. 0.2 0.00 0.00 0.00 0.00 0.50 0.00 0.00 33w.:3< QOON . . . . . . . . . . . . . . . . 030-60 0A0 0E 0F. 00» 000 one 00m New NR 000 .03 000 000 000 03 000 5.2 :2. mm? :00 0.00 0N0 0.00 0.00 0.0“ 0.5 A: :2 :0 0.04 0.00 0.: 0.30 0.00 0.00 30m¢Q< mom oml ml 1 mom oml mn‘ 1 mml cml ml 1 me omQ ml 1 UCt¢Q Lm®> bfiEsI 939395... BBEoQEoP 929.3%th 3203.33 :35. 82:08.2 coo—2 E:E_:__>_ 03.09:; “3:03 2.3m 2: 9.2.5 moi—mtg 30329023.: >=eu .0. 35:39.33 ace acne—2 .v 03¢... 22 IN02, ppb IPM2.5,microgram/m3 I802, ppb 003. ppb ppb, microgram/m3 Figure 7. Variation of mean daily levels of pollutants during week in East 7 Mile in the two years of study Abbreviations; ppb, part per billion; N02, nitrogen dioxide; 03, ozone; PM“, particulate matter with a diameter less than 2.5 micrometer; SOZ, sulfur dioxide IN02, ppb IPM2.5,microgram/m3 ISO2, ppb E103. ppb ppb. microgram/m3 Figure 8. Variation of mean daily levels of pollutants during week in Linwood in the two years of study Abbreviations; ppb, part per billion; N02, nitrogen dioxide; 03, ozone; PM”, particulate matter with a diameter less than 2.5 micrometer; SOz, sulfur dioxide 23 I Mean Terrperaue(Fd119rl'Ieit) oReIaIIveI-ummpeoet) FahrenheIt/ Percent 9 8 B 8 8 if 83 <3 93 8 8 | HIIIIII 1 MarAprMame .IuI AugtSep'OctNovDec JmFeb Figure 9. Monthly averages for the mean daily temperature and the mean daily relative humidity in the two sites during the study period. Descriptive analyses of hospital admissions and censgs data Hospital admissions along with census data are presented by demographic characteristic, site of residence and year of admission (Table 7). During 1999 and 2000, a total number of 4,847 hospital admissions with a discharge diagnosis of asthma were recorded in the Michigan inpatient database for patients who were one to 45 years of age and resided in the 23 zip codes used in this study. Distribution of the number of daily hospitalizations per site (Figures 10 and 11) represents an asymmetric left skewed curve. The average of the above- 24 mentioned hospital admissions was 3.3, with a variance of 6.02, median of 3 and mode of 2 (Table 5 and 6). Asthma hospital admission ratios in African-Americans compared to Caucasians were about 5:1 in Linwood and 3:1 in East 7 Mile. Residents up to 18 years old showed a higher rate of hospital admissions in both sites. Linwood had more admissions than East Seven Mile. In 2000 more admissions were recorded in comparison to the previous year, 1999 (Table 7). For ages less than 19 years, male had greater number of asthma admissions. On the other hand, for ages 19 years and older, females were more often admitted due to asthma (Figure 12). In warmer months, the total monthly admissions dropped in June and July, then increased in August and peaked in September (Figure 13). Mondays and Tuesdays had the maximum number of admissions among weekdays (Figure 14). African-Americans had a higher proportion in Linwood in comparison to East 7 Mile (68.6% versus 59.6%, Table 7). Mean daily admissions rate was higher in Linwood than East 7 Mile (1.4 versus 1.1 per 100,000). 25 Table 5. Statistics for daily hospital admissions in East 7 Mile Variable Estimate 100% Max 20 99% 12 95% 7 50% Median 3 5% 0 0% Min 0 Mode 3 Mean 3.11 Variance 5.76 Table 6. Statistics for daily hospital admissions in Linwood Variable Estimate 100% Max 19 99% 11 95% 8 50% Median 3 5% 0 0% Min 0 Mode 2 Mean 3.51 Variance 6.21 25‘ Percent 20‘ fiT T“ f“ 15‘ 10* 51 [—- 0 2 4 6 8101214161820 Figure 10. Distribution of total daily asthma admissions during 1999-2000 in Detroit, East 7 Mile 25‘ Percent 20‘ A, ./ \ ,/ W L 10‘ , / 1/ 4. 51/ A, 0 \ 1—1 [—1 P‘fi 0 2 4 6 8101214161820 Figure 11 . Distribution of total daily asthma admissions during 1999-2000 in Detroit, Linwood 26 Table 7. Demographic characteristics, site of residence and year of admission for census data along with hospital admissions E. Seven Mile Linwood Year 2000 Yearly admnssuons Year 2000 Yearly admnssnons population Rate pOpulation Rate (Percent) Year N (per 105) (Percent) Year N (per 105) Race African- 165988 1999: 881 530 171337 1999: 1108 647 American (59- 6%) 2000: 1091 657 68. 6% 2000: 1290 723 . 76675 1999: 124 161 54658 1999: 64 117 Caucasnan (27.6%) 2000: 178 232 21.9% 2000: 85 155 Other 35572 1999: 0 0 23590 1999: 10 42 races (12.8%) 2000: 4 1 1 9. 5% 2000: 12 51 Age (years) 1 _5 32143 1999: 257 799 29840 1999: 276 925 (11.6%) 2000: 338 1051 (11.9%) 2000: 378 1267 6- 1 8 89934 1999: 375 417 74545 1999: 343 460 (32. 3%) 2000: 476 529 (29.8%) 2000: 399 535 1 9_22 20319 1999: 31 152 21959 1999: 37 168 (7.3%) 2000: 48 236 (8.8%) 2000: 49 223 23_29 40310 1999: 83 206 41170 1999: 94 228 (14.5%) 2000: 90 223 (16.5%) 2000: 139 338 30-45 955 24 1999: 259 271 32171 1999: 432 526 (34.3%) 2000: 321 336 (32.9%) 2000: 422 513 Gender F | 141449 1999: 523 370 123331 1999: 597 484 ema 9 (50.8%) 2000: 633 447 (49.4%) 2000: 686 556 M ale 136786 1999: 482 352 126354 1999: 585 463 (49.2%) 2000: 640 468 (50.6%) 2000: 701 555 Total 278235 1999.- 1156 415 249685 1999: 1283 514 (100%) 2000: 1122 403 (100%) 2000: 1286 515 27 1000 2 900 800 700 600 500- 400- 300- 200‘ 1001 O ' T 1 1—5 6—1 8 19— 22 23— 29 30—45 IF UM Figure 12. Total number of asthma hospital admissions by age group and gender during the study period 450 400 350 300 250 200 F 150 total number of admissions 100-1 50- 0" l T 1 ’t I I 1 I hI Jan Feb Mar Apr May Jun Jul Aug Sep Oct Nov Dec I East 7 Mile El Linwood Figure 13. Monthly number of asthma hospital admissions in the two sites during the study period 28 total number of admissions Sun Mon Tue Wed Thu Fri Sat Day of the week I East 7 Mile El Linwood Figure 14. Asthma hospital admissions by day of the week during the study period Regression analyses The variance of daily hospital admission (6.02) was greater than its mean (3.3) and the distribution of daily hospital admissions was over-dispersed. We used negative binomial regression to investigate possible associations between air pollution levels and daily hospital admissions. Potential confounders were assumed to be year, month, weekday of admission; with age group, racial group, and gender of admittees; as well as average of relative humidity, minimum, mean and maximum temperature for the 4 days preceding of admission. Among fully saturated models (Table 8) with only one pollutant, models for N02, 802 and 03 were significantly associated with hospital admissions for certain levels of the pollutant. Final (the most parsimonious) models were models with variables that dropping them could have changed RR more than 10%. In other words, year or 29 month or both were the only confounders. In the final model for 802, risk ratio was significant only for the third quartile of 802 level; additionally, the risk ratios did not pursue a trend in this model. In the final model of N02, the first quartile level of the pollutant was not significantly associated with hospital admissions, but higher levels of N02 were associated with the number of hospital admissions. All three levels of 03 showed a significant protective effect on the number of hospital admissions. In the N02 and 03 models, one can notice an obvious trend in RR for different levels of the pollutants. At the next step, a model with year, month, 802, N02 and 03 was run (Tri-pollutant model, Table 9). $02 failed to show significance in the tri-pollutant model. Finally we ran a model with two pollutants, N02 and 03, controlling for year and month of admissions. This model was considered as our final model with having significant RR for both N02 and 03 (Table 10). The interaction term of N02 and 03 did not gain statistical significance. In mono-pollutant models, 802 and N02 were positively and 03 was negatively associated with the daily asthma hospital admissions. Association between PM2.5 and hospital admissions was not evident. In a tri-pollutant model including those pollutants that showed any association in mono-pollutant models (802, N02 and 03), 802 was no more associated with the outcome. N02 had still a positive and ozone a negative association with the outcome. 30 0286 5:3 .NOw ”9.80 80 ”02566 comet: .NOz ”220:5ch 55.820 .3. 60:9 xm: .mm u_o>o_ 8:229 .gom coaoE 053 Log Emu-590:: .58 31 62:5 Log :3 duo HERBS 8:00:50 ._0 ”22639595. 5622.8 30 fine mcfioooa v .8 .05. mom-52¢... so» 8:82.00 9oz, 23:28:00 .mzcouoa 95 __m 2909: 239me 2.2 c. .93 995 A26 92 on. :_ 983 mcozmtomno be 598:: 9: can 66> comm .LonEoEow-_:a< 83 £258 085 E 8295 00:3 9: ._. Ammo .89 Ed 9.85.8 Ammo .39 Ed tom-m9.- aoé3 886.3935 Nam-F.Rao.-mco.v Geodeovfid $1on 8.0 $3 .29 8.0 33.8 8.0 83 .39 Ed nan-m2 58.2 8.? :89 rum-3 8.? 33.3956 389.6 18> +6 63.2.55; mow-mew 83.89%; 31.8 :43 El .5: m3 mew-v.3 Amt-m3 63 .40.: a? own-3N 58.2 +802 36 8: .39 Se mom-QB mod :3 $9 2: mew-<2 8; 25992-: 8.? 25992-: 38962 83 .29 23 $2: $3 .89 :3 33.9 :43 $3 .39 2: 3-3 Amt-8.3 $3 .89 our 3.2 5:22 +N0m So as? .89 Be 515 mod 6: .39 8.9 33.4 03 2.69093 8.? :89 316 396m :0 88,9 29,2 :0 .39 8.86. 9:9 8. so .395. 8.9850 c. 20.89% 299655 292.66 .882 :23.an “.o 5:9: .3 Eco Loo» Lo.— u:___o.zcoo 10 ._o «Oz 509 +3235 unfla=o9oco2 .w aim... 31 0008.0 5.2.00 .0Ow 0:000 .00 0008.0 :000:.: .002 0020:0900 5.90005 .0. ”00.29 x0: .10 ”_0>0_ 00:90.29 .200. 0020:: 0.000 :00 89008.8 .0E .00 60...... :00 :00 .000 ”.0225 00:00.0:00 ..o .0:0=0_>900< 5.00.800 00 9.00 0500090 0 :0: _0>0_ 00903.? .000 0.03 .000 02: 0:. :. 00.000 0:0..020000 00 00:5: 05 0:0 .00.. :000 .00060200w-.:0< 003 .0006 0.5 :_ 00296. 00:00 0:0. ._. .000 .009 000 2000.00 .000 .02-9 00.0 2.00-0.02. .000 .009 000 0002.00 .000 .009 00.0 00.10.00 .002 .009 00.0 0.00-0.00 .000 .009 00.0 0.00-0.2 002 3092.00-00 002 .200. 0.2-0.0 .0096 .00403 802002.22 0.00.000 202009002 0.0.1.00 00.0 .002009002 000-000.00.003 .002.009002 000.000 .202 .009 002 0.00-0.2 00.0 .22 .009 000 0.0.2.2 002 3090.02-02 002 3090.02-02 309002 .002 .009 022 0.2-0.0 .002 .009 202 0.2-2.02 .002 .009 002 00.02. .002 .009 32 02-2.0 .002 .009 :2 021.0 .22 .009 002 0.0.2.2. 002 3090.020 002 3090210 309000 ._0 00.00. 022 ._0 ..003 222 0. 000.0300 20.000 x 00.0.0300 200000.000 .009-0:05:00 :o_00_E00 00 5:2: 0:0 :00» :2 0:50-3:00 .00 0:0 002 .009 0.0008 0:82.002; .0 030.: 32 mco~o a0 ”09x06 552:: .62 uQmEqu 88.385 .x 80:9 xmt .mm ”.30. 8:882 .901 :29: 033 8a c6592.: .nE 3: 82:5 8a tag .53 ”6225 8:03:00 ._0 8839559.. co_mm_Ecm .8 986 9.6965 v .8 _m>m_ mafia}... N?» 895 A26 92 9: E 963 825238 .6 525: 9: was .63 comm 33528234. mm; .89: £5 E 323:. nozma m5 ._. Ammo .89 5o v8-8.8 88 .89 88 r889 83 .89 8.0 584.8 88 .89 Ed 8.88.8 $3 .59 «ad 33.8 8.0 .89 83 8.8-8.8 8:38 8.? A.._mm:.mm-m.mao8oo.v 8; A.8mvu.8-m.m 30.96 88 a: .8. 3 mm; @838 8 o a? .29 m: 31.8 a: .8. 5 v? 3.3.8 :3 .89 8: 98-98 :3 .89 Be 08-8.9 a: .89 8; 3.3.2 03 898.87: 03 3.89087: 8962 :0 08m 9 296. :o 088 296. v. 5.88am 2980 x :0 881m 298235 c_:9.E£:__oa :o_mm_Euw *0 :22: van .3; .8 u:_._o.=cou 10 can N02V touoE “283.395 .3 23m... 33 DISCUSSION Among residents of 23 zip codes in Detroit, Michigan, we conducted an aggregative study to investigate the relationship between daily asthma hospital admissions and levels of four pollutants nitrogen dioxide (N02), ozone (03), particulate matter with a diameter less than 2.5 pm (PM2,5) and sulfur dioxide (502)- Using negative binomial models and taking into account the average levels of pollutants over the 4-day period preceding daily asthma hospital admissions, regression analysis revealed that number of daily asthma hospital admissions was positively associated (increased) with N02 levels and negatively associated (decreased) with 03 levels (adjusted for year and month of admissions). Asthma hospital admissions showed a positive association with 802 levels in single-pollutant model, but this association disappeared when we controlled for N02 and 03. No association was evident between particulate matter (PM2.5) and asthma hospital admissions. For N02, the only pollutant that consistently showed a positive association with the outcome, an exposure-response trend was suggested. The estimated relative risk per 6 ppb increase in N02 (an increment roughly equal to one standard deviation) was 1.08. It has been shown that both N02 and 03 increase inflammatory response in lung and bronchial tissues 22. Although it is in debate whether or not these pollutants are associated with the incidence of new diagnosis of asthma 23, most of the previous studies support at least a triggering effect for them even at low 34 levels 24. Our finding that increasing N02 levels are positively associated with asthma hospital admissions is consistent with previous studies. A conflicting finding in this study was the protective effect of increasing 03 levels on the number of daily asthma hospital admissions (As 03 levels increased, the number of admissions dropped). However in several previous studies with the same methodology as ours, increasing 03 levels appeared to be negatively associated with the number of asthma hospital admissions 9”". In fact 03 is not emitted to the air directly and it is the product of a complex reaction between volatile organic compounds (VOCs) and nitrogen oxides in the t 16. We did not have data on the pollutants that are related to presence of sunligh 03 levels, except for N02. Our data showed that 03 and N02 were negatively correlated. Additionally, the negative effect of 03 levels on average numbers of daily asthma hospital admissions was best seen when N02 was low in contrast to situations when level of N02 was high (Figure 15). Based on these explanations and the fact that 03 is a byproduct of other pollutants, we think that the negative association between 03 and the daily number of asthma hospital admission, found in this study might be due to the simultaneous lower levels of other pollutants, such as VOCs and not independently due to 03 levels. Additional investigation, in the presence of data on pollutants involved in the pathway of 03 production such as VOCs and various nitrogen oxides (Figure 1), may help to further explain the effect of 03 levels on asthma hospital admission. 35 In the analysis, race emerged as a strong independent predictor of hospital admissions (asthma hospital admission was 3-5 times higher in African- Americans compared to Caucasians). Our results are inconclusive whether ‘race’ can affect asthma hospital admission due to air pollution. The pattern of hospital admission by gender is in agreement with previously reported studies (Figure 12). The effect of seasonality on asthma hospital admissions is a well-known phenomenon. We noticed a sharp rise of asthma hospital admissions in September. Other studies reported a seasonal variation in the asthma hospital ”"25 and Sheppard et al. had identical findings 6. admission It has been argued that parents pay more attention to their children’s preventive measures for asthma during holiday, as well as children being under less stress than when at school 12. Indeed those who travel and become admitted out of the state are no longer being calculated for the number of hospital admissions. In our study, the September peak was evident for all age groups, particularly for ages less than 19 years. Holiday travels facilitate the acquisition of new viral strains by the community. It is possible that children share new strains of viruses and carry them to their families when they first come back to school. Viral upper respiratory infections, especially rhinovirus, are reported to be the most common cause of acute asthma exacerbations. Looking at asthma hospital admissions, Johnston et al. showed that viruses are associated with 80 to 85% of asthma exacerbations in school-age children 26. 36 Limitations and strengms In this study we assumed that residents of zip codes within the 4-kilometer radius of each air-monitoring station were exposed to the same' amount of pollutants. This assumption may evoke information bias, because the monitoring stations were not necessarily at the point that the overall effect of pollution in the site had been exhibited. For example the distance of the Linwood station to the three Interstate highways (I-75, l-94 and l-96) is less than a mile, while distance of most other points at this site is much further from these highways (Figure 2). Ecological fallacy may apply to this investigation as we assumed all residents exposed to the same amount of the pollutants. Data that might affect individual exposures was not available (such as the duration the residents spent outside and individual level factors such as cigarette smoking). Although we did not have data on pollen count and flu episodes (that are related to asthma), however they are not likely to be related to air pollution. The data included information on the daily number of hospital admissions by age, gender and race, so we were unable to account for re-admissions and severity of attacks. Daily hospital admissions may have been biased due to different level of access to hospital among the study population, but the data had the potential to capture nearly all asthma hospital admissions because all hospitals in Michigan are committed to participate in the Michigan Inpatient Database. We attempted to analyze the data using Poisson regression models as similar previous studies explained but the fit was not adequate in Poisson 37 regression models. Using negative binomial models we achieved an excellent goodness of fit for our models of regression analysis. Conclusion This thesis, adds to the body of evidence that supports an adverse effect of air pollution on the increase of asthma hospital admissions and is an example of using regression analysis for the count data in epidemiologic studies. Because of the ecological nature of this study, the results do not necessarily indicate a causal association. Among the four pollutants studied in this thesis, increasing NO; levels seemed to be related to higher daily asthma hospital admissions. Our analyses did not support the adverse effect of PM2_5 and $02 on asthma hospital admissions. The conflicting protective effect of 03 on the number of asthma hospital admissions needs more investigation. We suggest designing future studies that take into account pollutants, which are important in the pathway of 03 production, such as VOCs and nitrogen oxides. Additionally conclusion, the best setting to study the effect of the air pollutants on asthma would be the assessment of exposures, confounders and outcomes at the individual level. 38 Average number of daily asthma hospital admissions 1 2 3 4 03 quartile levels” Figure 15. Average number of daily asthma hospital admissions for different quartiles of nitrogen dioxide (N02) and ozone (03) '03 quartile levels (parts per billion): 1, 5.9-22. 1; 2, 22. 2-27.0; 3, 27. 1-32. 7; 4, 32. 8-59.1 ”N02 quartile levels (parts per billion): 1, 1.7-16.5; 2, 16.6-20.3; 3, 20.4-24.8; 4, 24.9-46.9 39 EoEtmaoU 889828 .9... 8sz 933228 .SE 82x06 5:8 .NOm 8280 .6 82x06 239:: .52 80> 20:3 .> :35; .>> ”20>:an .w 8_nm__m>m 8: .<2 uootm o: .o uootm gamma: .1 uooto o>£moa .+ 822.9552 68 3 m 95 <2 <2 + <2 88, am 88 __< Em. EWm mhmmaw 88 @5850 8 - 88 88: 888». >882, maso< + <2 + 0 £63 om 28> 2v c0920.. -83 E8225 1 can Ammmsv 0 <2 + <2 8m; 8 28> 8v >5 8:8: -8.. 08? 825m. -882 >38meme <2 <2 + <2 89> om 88> o E 38.2 m 89 $8 3 . . . . EE>> 83 228.88 >25 . conEoEmw E 82888 0mm w 3mm 5 .6 288:: Essie: o o + mov 238m -83 Ntmicow 828.88 mm? 2.8 5 + + + + 558: 88 __< 258: -82 928 296. vow Em Now as, 888 >>.w-> 88: 8888 5 88.8.8 083 8m; 5.2889 new 2.58 0 <2 0 o mums am 88 __< 8>=oocm> 88 82mm .965 wow 8A.: 05 E -Vwmv Now 5E mo Noz >88 :85 8:88:50 EEEBQ oEooSO cozflaon. 20:80.. 8m .6 cocoa 553m . 8E 853» 825.5 no EaEEsm 53:32 40 EoEtmamu >ocm9mEm .Dm 5sz 0632th in 62x06 :55 .NOm 6203 50 62x06 c3022 .62 38> 20:2, .> USE; .>> umEEam .w ”m_nm__m>m 6: .<2 uootm o: .o uomtm gamma: .I uomtm 0238 .+ ”22630.2( 325 20388 86:2me $2 88 s 9032?. 2m 2.2m 58 o + 0 <2 .3381 23> mmv .mEmmw >me 2335 820 .8 co=m_oo$m on new 965 mom 895.; o o o o m 55 38> EA: 983:5 ”Mam” Ahmmwv E 9.52 225083 m>=_moa + + + + a. . Om Em?» NTm .ooEmamm 9.2.6 3238 3mm 3 $3 $6.20 oszémm I - B 963 0m 8 8:865 + o + $7.3 om. 93> E. F Ema. >50 :9; 3.20383 c< a Emwa=om 0 >> Nmmw 058 m m5 5 m :9. I $me mum 9: 390m >ocm§mcoo ho xom. m was 055 Sc 6853 o o o mcwmmmwwfim wmmm =< c0954 Nam .2 mco_mm_Ecm _mgamo: >__mn + + + _ H. I w _. £3, 225083 E8586 0 -Nmm —. 8mm 5 m>mz 9 958 9282.3 __< + cowhmtc< mom _‘ 2.8 9:05 - Cam 5 + + + + .8980 33 __< mtmn. 53 9.622 953‘ 03% $82: 25 280 o as? 5 gas cm. E 6%? 883m 82 Cam : 5328 832.8 Ed I <2 25.; u n 3: . . 9 $22 593$: -82 5&6: . . E + Now 2m md N02 >qu twee 2:08:50 EEEBQ wEooSO cozmfiaoa 20:80.. Em ho norma .wmflum 6.28 228% 41 East—mama >ocwmhmEm .0w 5sz $6.32th .SE ”02x06 5:3 .NOw 6.53 :06 ”09x06 .8ng .62 ”50> 20:3 .> :25; .>> ”.0883 .m um_nm__m>m 6: .<2 uomtm o: .o uootm 02.30: .I uootm m>=_moa .+ ”32639.54 89.3 33> mov 95x 951 WWW” E mco_mm_Eum 0336 0328 + + + + mc0_wm_EUm new no c0923 cozmfiommm _mzamoI wmmw ANOONV 026mm: “585ch Em?» moum _‘ concen— -Nmmw mco>> 98.53 .9636... -233 m5 5 $325 95 22a <2 + <2 <2 mcmmmfim 98> N to 9:98 ”WM" 38% a: Son M_o 399m {an 06 _ . as A v #593 - . - m mam? meow -95.. 65:9: 82:56 + Wu 0 O mu_w_> Om whmm> vm m CEO -rmmw mtmfi I mco_mm_Eum - m mam? Coomv + <2 + .8581 28> t F m__oqmcm_nc_ -32 2:83 E12 m:o_mm_Ecw 39‘ A 503 + + o + .95on 93> mwv 9.3. 0:01 -89 953 no .22.”. SE: I AOOONV c0228 5:29.00 ,6 mud <2 + + o m=m_> Ow m5?» orv 9:32 w mam? tmn_o._. NOm SE «.0 N02 23% Cmmt mEmEEoo ESEEQ mEooSO consign. c0283 .hom “_o normn. .fifim 6.250 2.83% 42 10. BIBLIOGRAPHY Weisel CP. Assessing exposure to air toxics relative to asthma. Environ Health Perspect 2002;110 Suppl 42527-37. Redd SC. Asthma in the United States: burden and current theories. Environ Health Perspect 2002;110 Suppl 4:557-60. Lebowitz MD. Epidemiological studies of the respiratory effects of air pollution. Eur Respir J 1996;9(5):1029-54. Schwartz J, Slater D, Larson TV, Pierson WE, Koenig JQ. Particulate air pollution and hospital emergency room visits for asthma in Seattle. Am Rev Respir Dis 1993;147(4):826-31. Stieb DM, Burnett RT, Beveridge RC, Brook JR. Association between ozone and asthma emergency department visits in Saint John, New Brunswick, Canada. Environ Health Perspect 1996;104(12):1354-60. Sheppard L, Levy D, Norris G, Larson TV, Koenig JQ. Effects of ambient air pollution on nonelderly asthma hospital admissions in Seattle, Washington, 1987-1994. Epidemiology 1999;10(1):23-30. Tolbert PE, Mulholland JA, Maclntosh DL, Xu F, Daniels D, Devine OJ, Carlin BP, Klein M, Dorley J, Butler AJ, Nordenberg DF, Frumkin H, Ryan PB, White MC. Air quality and pediatric emergency room visits for asthma in Atlanta, Georgia, USA. Am J Epidemiol 2000;151(8):798-810. Ritchie l. Health effects of ozone and other environmental measures on children in the lnianapolis metropolitan area. 2001. Buchdahl R, Parker A, Stebbings T, Babiker A. Association between air pollution and acute childhood wheezy episodes: prospective observational study. ij 1996;312(7032):661-5. Anderson HR, Ponce de Leon A, Bland JM, Bower JS, Emberlin J, Strachan DP. Air pollution, pollens, and daily admissions for asthma in London 1987-92. Thorax 1998;53(10):842-8. 43 11. 12. 13. 14. 15. 16. 17. 18. 19. 20. 21. 22. Garty BZ, Kosman E, Ganor E, Berger V, Garty L, Wietzen T, Waisman Y, Mimouni M, Waisel Y. Emergency room visits of asthmatic children, relation to air pollution, weather, and airborne allergens. Ann Allergy Asthma lmmunol 1998;81(6):563-70. Crighton EJ, Mamdani MM, Upshur RE. A population based time series analysis of asthma hospitalisations in Ontario, Canada: 1988 to 2000. BMC Health Serv Res 2001 ;1(1):7. Weiss KB. Seasonal trends in US asthma hospitalizations and mortality. Jama 1990;263(17):2323-8. Thorell D. The effect of ambient air pollution on chilhood asthma hospital admissions and emergency room visits in Flint, Michigan, 1999. MDCH. An Analysis of Childhood Asthma Hospitalization and Deaths in Michigan, 1989-1993. Lansing, MI: Michigan department of community health, 1998. EPA. Criteria air pollutants. http://www.epa.gov. US. Census Bureau. http://www.census.gov/. Atkinson RW, Anderson HR, Strachan DP, Bland JM, Bremner SA, Ponce de Leon A. Short-term associations between outdoor air pollution and visits to accident and emergency departments in London for respiratory complaints. Eur Respir J 1999;13(2):257-65. SAS/STAT Software. Cary, SAS Institute Inc. NC 2001. Cameron A. C, Trivedi, P.K.. Regression Analysis of Count Data Cambridge University Press, 1998. Stokes ME, C. S. Davis and G. G. Koch,. Categorical Data Analysis Using the SAS System SAS Institute, 1995. Bowler RP, Crapo JD. Oxidative stress in allergic respiratory diseases. J Allergy Clin Immunol 2002;110(3):349-56. 44 23. 24. 25. 26. McConnell R, Berhane K, Gilliland F, London SJ, Islam T, Gauderman WJ, Avol E, Margolis HG, Peters JM. Asthma in exercising children exposed to ozone: a cohort study. Lancet 2002;359(9304):386-91. Gent JF, Triche EW, Holford TR, Belanger K, Bracken MB, Beckett WS, Leaderer BP. Association of low-level ozone and fine particles with respiratory symptoms in children with asthma. Jama 2003;290(14):1859- 67. Khot A, Burn R, Evans N, Lenney C, Lenney W. Seasonal variation and time trends in childhood asthma in England and Wales 1975-81. Br Med J (Clin Res Ed) 1984;289(6439):235-7. Johnston SL, Pattemore PK, Sanderson G, Smith S, Campbell MJ, Josephs LK, Cunningham A, Robinson BS, Myint SH, Ward ME, Tyrrell DA, Holgate ST. The relationship between upper respiratory infections and hospital admissions for asthma: a time-trend analysis. Am J Respir Crit Care Med 1996;154(3 Pt 1):654-60. 45 l11111111151111millI