USF Libraries
USF Digital Collections

Association between area socioeconomic status and hospital admissions for childhood and adult asthma

MISSING IMAGE

Material Information

Title:
Association between area socioeconomic status and hospital admissions for childhood and adult asthma
Physical Description:
Book
Language:
English
Creator:
Tamulis, Tomas
Publisher:
University of South Florida
Place of Publication:
Tampa, Fla.
Publication Date:

Subjects

Subjects / Keywords:
Asthma epidemiology
Environmental triggers
Socioeconomic deprivation index
Spatial analysis
Non-linear multiple regression model
Dissertations, Academic -- Public Health -- Doctoral -- USF   ( lcsh )
Genre:
government publication (state, provincial, terriorial, dependent)   ( marcgt )
bibliography   ( marcgt )
theses   ( marcgt )
non-fiction   ( marcgt )

Notes

Summary:
ABSTRACT: Despite an improved understanding of the disease, the prevalence of asthma and asthma-related morbidity continue to rise, particularly among minority and inner-city populations. Despite the growing epidemic of asthma, the surveillance of disease at the state or even local levels is very limited. Such information is very important to identify high-risk population groups and to design more effective community-based preventive interventions or risk management programs that may modify these trends. The study provided important information about spatial differences by the geographical area of residence and changes in asthma hospital admissions over time in the selected area. Environmental exposure to ambient air pollution by ambient particles, sulfur dioxide and ozone was a significant factor to explain the increase in asthma hospitalizations in simple regression analysis, but was not significant after the adjustment to area socioeconomic status characteristics.Sulfur dioxide was the only significant independent variable in a multiple adjusted regression model of hospitalizations for childhood asthma, however, more detailed environmental exposure assessment by calendar quarter suggested that ambient air pollution by sulfur dioxide is not significant variable in the multiple regression model. Future asthma prevention interventions and risk management programs should address population groups described by such socioeconomic status characteristics as poverty, unskilled workers, single parent families with children, families having no vehicle available, people living in less crowded households or socially excluded conditions without adequate family members or relatives support, and also people residing in houses heated by fuel. Developed complex area socioeconomic deprivation index was shown to be a significant predictor of hospital admissions for childhood and adult asthma by zip code area of residence.
Thesis:
Thesis (Ph.D.)--University of South Florida, 2005.
Bibliography:
Includes bibliographical references.
System Details:
System requirements: World Wide Web browser and PDF reader.
System Details:
Mode of access: World Wide Web.
Statement of Responsibility:
by Tomas Tamulis.
General Note:
Title from PDF of title page.
General Note:
Document formatted into pages; contains 274 pages.
General Note:
Includes vita.

Record Information

Source Institution:
University of South Florida Library
Holding Location:
University of South Florida
Rights Management:
All applicable rights reserved by the source institution and holding location.
Resource Identifier:
aleph - 001680995
oclc - 62500458
usfldc doi - E14-SFE0001134
usfldc handle - e14.1134
System ID:
SFS0025455:00001


This item is only available as the following downloads:


Full Text
xml version 1.0 encoding UTF-8 standalone no
record xmlns http:www.loc.govMARC21slim xmlns:xsi http:www.w3.org2001XMLSchema-instance xsi:schemaLocation http:www.loc.govstandardsmarcxmlschemaMARC21slim.xsd
leader nam Ka
controlfield tag 001 001680995
003 fts
005 20060215071104.0
006 m||||e|||d||||||||
007 cr mnu|||uuuuu
008 051207s2005 flu sbm s000 0 eng d
datafield ind1 8 ind2 024
subfield code a E14-SFE0001134
035
(OCoLC)62500458
SFE0001134
040
FHM
c FHM
049
FHMM
090
RA425 (Online)
1 100
Tamulis, Tomas.
0 245
Association between area socioeconomic status and hospital admissions for childhood and adult asthma
h [electronic resource] /
by Tomas Tamulis.
260
[Tampa, Fla.] :
b University of South Florida,
2005.
502
Thesis (Ph.D.)--University of South Florida, 2005.
504
Includes bibliographical references.
516
Text (Electronic thesis) in PDF format.
538
System requirements: World Wide Web browser and PDF reader.
Mode of access: World Wide Web.
500
Title from PDF of title page.
Document formatted into pages; contains 274 pages.
Includes vita.
520
ABSTRACT: Despite an improved understanding of the disease, the prevalence of asthma and asthma-related morbidity continue to rise, particularly among minority and inner-city populations. Despite the growing epidemic of asthma, the surveillance of disease at the state or even local levels is very limited. Such information is very important to identify high-risk population groups and to design more effective community-based preventive interventions or risk management programs that may modify these trends. The study provided important information about spatial differences by the geographical area of residence and changes in asthma hospital admissions over time in the selected area. Environmental exposure to ambient air pollution by ambient particles, sulfur dioxide and ozone was a significant factor to explain the increase in asthma hospitalizations in simple regression analysis, but was not significant after the adjustment to area socioeconomic status characteristics.Sulfur dioxide was the only significant independent variable in a multiple adjusted regression model of hospitalizations for childhood asthma, however, more detailed environmental exposure assessment by calendar quarter suggested that ambient air pollution by sulfur dioxide is not significant variable in the multiple regression model. Future asthma prevention interventions and risk management programs should address population groups described by such socioeconomic status characteristics as poverty, unskilled workers, single parent families with children, families having no vehicle available, people living in less crowded households or socially excluded conditions without adequate family members or relatives support, and also people residing in houses heated by fuel. Developed complex area socioeconomic deprivation index was shown to be a significant predictor of hospital admissions for childhood and adult asthma by zip code area of residence.
590
Adviser: Dr. Yehia Y. Hammad.
653
Asthma epidemiology.
Environmental triggers.
Socioeconomic deprivation index.
Spatial analysis.
Non-linear multiple regression model.
690
Dissertations, Academic
z USF
x Public Health
Doctoral.
773
t USF Electronic Theses and Dissertations.
4 856
u http://digital.lib.usf.edu/?e14.1134



PAGE 1

Association between Area Socioeconomi c Status and Hospital Admissions for Childhood and Adult Asthma by Tomas Tamulis A dissertation submitted in partial fulfillment of the requirements for the degree of Doctor of Philosophy Department of Environmental and Occupational Health College of Public Health University of South Florida Major Professor: Yehia Y. Hammad, Sc.D. Noreen D. Poor, Ph.D. Raymond D. Harbison, Ph.D. Thomas J. Mason, Ph.D. Getachew A. Dagne, Ph.D. Date of Approval April 8, 2005 Keywords: asthma epidemiology, environmen tal triggers, socioeconomic deprivation index, spatial analysis, non-linear multiple regression model Copyright 2005, Tomas Tamulis

PAGE 2

i Table of Contents List of Tables ii List of Figures x List of Pictures xii Abstract xiii Chapter 1 Association between Environmental Asthma Triggers and Hospitalizations for Childhood and Adult Asthma 1 1.1. Literature Review 1 1.1.1. Etiology, Pathology and Epidemiology of Asthma 1 1.1.2. Environmental Triggers of Asthma 6 1.2. Methodology 17 1.2.1. Study Population 17 1.2.2. Study Design 19 1.2.3. Data Analysis 27 1.3. Results 31 1.3.1. Hospital Admissions for Childhood and Adult Asthma 31 1.3.2. Environmental Asthma Triggers 38 1.3.3. Multiple log-linear Regression Analysis 58 Chapter 2 Association between Area Soci oeconomic Status and Hospitalizations for Childhood and Adult Asthma 63 2.1. Literature Review 63 2.2. Methodology 72 2.2.1. Study Design 72 2.2.2. Environmental Exposure Assessment 73 2.2.3. Socioeconomic Status Indicators 79 2.2.4. Socioeconomic Deprivation Index 83 2.2.5. Data Analysis 85 2.3. Results 90 2.3.1. Spatial Interpolation 90 2.3.2. Socioeconomic Status In dicators 109 2.3.3. Socioeconomic Deprivation Index 131 2.3.4. Multiple Regression Modeling and Analysis 137 2.3.5. Validation of Predictive Regression Model 161 Chapter 3 Summary Discussion 166 Chapter 4 Final Conclusions 183 References 194 Appendices 203 About the Author End Page

PAGE 3

ii List of Tables Table 1. Prevalence rates of lifetime and current asthma, severe asthma attacks, emergency room visits and hospital admissions for asthma, and asthma mortality in the US in year 2000 3 Table 2. State of Florida and federal national ambient air quality standards (NAAQS) 23 Table 3. Total number of hospital admi ssions for asthma and average length of stay (ALOS) in days by calenda r quarter in Florida, 1999 31 Table 4. Children 14 years and younger hospital admissions for asthma and average length of stay (ALOS) in days by calendar quarter in Florida, 1999 32 Table 5. Ambient Air Quality Monito ring Network in Hillsborough County, FL, 1997-1999 39 Table 6. Number of peak concentra tions above air quality standards for SO2, PM10 and O3, and a number of childhood and adult asthma hospital admissions by calendar month and quarter, in Hillsborough County, FL, 1997-1999 40 Table 6. Number of peak concentra tions above air quality standards for SO2, PM10 and O3, and a number of childhood and adult asthma hospital admissions by calendar month and quarter, in Hillsborough County, FL, 1997-1999 (cont.) 41 Table 7. Average monthly and qu arterly particulate matter (PM10) concentrations, g/m3, by separate ambient air quality monitoring site, Hillsborough County, FL, 1997 43 Table 7. Average monthly and quarterly PM10, g/m3, by separate ambient air quality monitoring site, Hills borough County, FL, 1997 (cont.) 43 Table 8. Average monthly and quarterly PM10, g/m3, by separate ambient air quality monitoring site in Hillsborough County, FL, 1998 43 Table 8. Average monthly and quarterly PM10, g/m3, by separate ambient air quality monitoring site, Hills borough County, FL, 1998 (cont.) 44 Table 9. Average monthly and quarterly PM10, g/m3, by separate ambient air quality monitoring site, Hillsborough County, FL, 1999 44 Table 9. Average monthly and quarterly PM10, g/m3, by separate ambient air quality monitoring site, Hills borough County, FL, 1999 (cont.) 44 Table 10. Average monthly and quarterly ozone concentration, ppb, by separate monitoring site, in Hillsborough County, FL, in 1997 46 Table 11. Average monthly and quarterly ozone concentration, ppb, by separate monitoring site, in Hillsborough County, FL, in 1998 47 Table 12. Average monthly and quarterly ozone concentration, ppb, by separate monitoring site, in Hillsborough County, FL, in 1999 47

PAGE 4

iii Table 13. Average monthly and quarterly su lfur dioxide concentration, ppb, by separate monitoring site, in H illsborough County, FL, in 1997 49 Table 14. Average monthly and quarterly su lfur dioxide concentration, ppb, by separate monitoring site, in H illsborough County, FL, in 1998 49 Table 15. Average monthly and quarterly su lfur dioxide concentration, ppb, by separate monitoring site, in Hillsborough County, FL, in 1999 50 Table 16. Average monthly and quarterly ambient temperature, F, Tampa Bay, 1997-1999 52 Table 17. Average monthly and seasonal total tree pollen counts in Hillsborough County, FL, 1997-1999 54 Table 18. Average monthly and seasonal total weed pollen counts in Hillsborough County, FL, 1997-1999 55 Table 19. Average monthly and seasonal total grass pollen counts in Hillsborough County, FL, 1997-1999 57 Table 20. Census-derived variables with corresponding standardized relative weights contributing to the Unde rprivileged Area (UPA) score 84 Table 21. Dimensions of deprivation and variables contributing to the Urban Deprivation Index (UDI) 85 Table 22. List of ambient air quality m onitoring stations and average annual values for PM10, g/m3, in 1997, 1998, and 1999 93 Table 23. Interpolated PM10 concentrations, g/m3, by zip code area of residence in Hillsborough County, FL, in 1997, 1998, and 1999 94 Table 23. Interpolated PM10 concentrations, g/m3, by zip code area of residence in Hillsborough County, FL, in 1997, 1998, and 1999 (cont.) 94 Table 24. High and low environmental e xposure to coarse particulate matter categories (strata) and adult asthma hospital admissions within separate category (stratum) 99 Table 25. High and low environmental e xposure to coarse particulate matter categories (strata) and children as thma hospital admissions within separate category (stratum) 99 Table 26. List of ambient air quality monitoring stations for SO2, ppb, and average annual values in 1997, 1998, and 1999 100 Table 27. Interpolated sulfur dioxide concentration values, ppb, by zip code area of residence in Hillsborough County, FL, 1997, 1998, and 1999 101 Table 28. Separate high and low envir onmental exposure to sulfur dioxide categories (strata) and adult asthma hospital admissions within different stratum 103 Table 29. Separate high and low envir onmental exposure to sulfur dioxide categories (strata) and adult asthma hospital admissions within different stratum 104 Table 30. List of ozone ambient air qua lity monitoring stations and average annual values in 1999 105 Table 31. Interpolated environmental exposure to ozone, ppb, by separate zip

PAGE 5

iv code area of residence in Hillsborough County, FL, 1997, 1998, and 1999 106 Table 32. Separate high and low envir onmental exposure to ozone categories (strata) and adult asthma hospital admissions within each stratum 108 Table 33. Separate high and low envir onmental exposure to ozone categories (strata) and children asthma hospital admissions within each stratum 108 Table 34. Calculated UPA and UDI index values and crude rates of hospital admissions for adult and childhood asthma per 10,000 population by zip code area of residence in Hillsborough County, FL, 1999 110 Table 35. List of significant socioec onomic status variables along with corresponding correlation coefficient and p-values 111 Table 36. Rate ratio (RR) and corre sponding 95% confidence intervals (95% CI) of adult asthma hospita l admissions in low and high socioeconomic status categories by poverty (SES – socioeconomic status; high socioeconomic status by poverty used as a reference) 115 Table 37. Rate ratio (RR) and corre sponding 95% confidence intervals (95% CI) of childhood asthma hospita l admissions in low and high socioeconomic status categories by poverty (SES – socioeconomic status; high socioeconomic status by poverty used as a reference) 116 Table 38. Rate ratio and correspondi ng 95% confidence intervals (95% CI) of adult asthma hospital admissions in low and high socioeconomic status categorie s by family income (SES – socioeconomic status; high socioeconomic status by the percentage of persons with a nnual family income of $15,000 and less used as a reference) 118 Table 39. Rate ratio (RR) and corre sponding 95% confidence intervals (95% CI) of childhood asthma hospita l admissions in low and high socioeconomic status categorie s by family income (SES – socioeconomic status; -high socioeconomic status by the percentage of persons with a nnual family income of $15,000 and less used as a reference) 118 Table 40. Rate ratio (RR) and corre sponding 95% confidence intervals (95% CI) of adult asthma hospita l admissions in low and high socioeconomic status categories by white-collar occupation (SES – socioeconomic status; -high socioeconomic status by the percentage of population employe d in white-collar occupations used as a reference) 121 Table 41. Rate ratio (RR) and corre sponding 95% confidence intervals (95% CI) of childhood asthma hospita l admissions in low and high socioeconomic status categories by white-collar occupation (SES – socioeconomic status; high socioeconomic status by the percentage of population employe d in white-collar occupations used as a reference) 121

PAGE 6

v Table 42. Rate ratio (RR) and corre sponding 95% confidence intervals (95% CI) of adult asthma hospita l admissions in low and high socioeconomic status categories by single parent with children status (SES – socioeconomic stat us; -high socioeconomic status by the percentage of single parent status used as a reference) 124 Table 43. Rate ratio (RR) and corre sponding 95% confidence intervals (95% CI) of childhood asthma hospita l admissions in low and high socioeconomic status categories by single parent with children status (SES – socioeconomic stat us; high socioeconomic status by the percentage of single pare nts used as a reference) 124 Table 44. Rate ratio (RR) and corre sponding 95% confidence intervals (95% CI) of adult asthma hospita l admissions in low and high socioeconomic status categorie s by overcrowding housing status (SES – socioeconomic status; high socioeconomic status by the percentage of population living in overcrowding housing status used as a reference) 126 Table 45. Rate ratio (RR) and corre sponding 95% confidence intervals (95% CI) of childhood asthma hospita l admissions in low and high socioeconomic status categorie s by overcrowding housing (SES – socioeconomic status; RRrate ratio; 95% CI – 95% Confidence Intervals; -high socioeconomi c status by the percentage of population living in overcrowding hous ing used as a reference) 127 Table 46. Rate ratio (RR) and corre sponding 95% confidence intervals (95% CI) of adult asthma hospita l admissions in low and high socioeconomic status categories by Urban Deprivation Index (SES – socioeconomic status; -high socioeconomic status by Urban Deprivation Index used as a reference) 130 Table 47. Rate ratio (RR) and corre sponding 95% confidence intervals (95% CI) of childhood asthma hospita l admissions in low and high socioeconomic status categories by Urban Deprivation Index (SES – socioeconomic status; high socioeconomic status by Urban Deprivation Index used as a reference) 130 Table 48. Selected reside ntial area socioeconomic st atus indicators with standardized relative weights used to calculate the Socioeconomic Deprivation Index (SDI) 133 Table 49. Socioeconomic Deprivation Inde x (SDI) values and crude rates of hospital admissions for adult and childhood asthma by geographical area of residence in Hillsborough County, FL, in 1997, 1998, and 1999 135 Table 50. Calculated DFBETAS values for selected extreme cases (outliers) in the childhood asthma re gression model in 1997 141 Table 51. Transformed best-fit childhood asthma hospital admissions regression model paramete r estimates (1997) 141 Table 52. Calculated DFBETAS values for selected extreme cases (outliers) in the childhood asthma regr ession model in 1998 143

PAGE 7

vi Table 53. Transformed best-fit childhood asthma hospital admissions regression model paramete r estimates (1998) 143 Table 54. Transformed best-fit childhood asthma hospital admissions regression model paramete r estimates (1999) 144 Table 55. Calculated DFBETAS values for selected extreme cases (outliers) in the adult asthma regression model in 1997 147 Table 56. Transformed best-fit adult asthma hospital admissions regression model parameter estim ates (1997) 147 Table 57. Calculated DFBETAS values for selected extreme cases (outliers) in the adult asthma regression model in 1998 148 Table 58. Transformed best-fit adult asthma hospital admissions regression model parameter estimates (1998) 149 Table 59. Calculated DFBETAS values for selected extreme cases (outliers) in the adult asthma regression model in 1999 151 Table 60. Transformed best-fit adult asthma hospital admissions regression model parameter estimates (1999) 151 Table 61. Transformed best-fit childhood asthma hospital admissions regression model para meter estimates 155 Table 62. Calculated DFBETAS values for selected extreme cases (outliers) in the childhood asthma regr ession model in 1999 157 Table 63. Calculated DFBETAS values for selected extreme cases in the adult asthma regression model 160 Table 64. Actual and predicted es timates of childhood asthma hospital admissions by zip code area of re sidence in Pinellas County, FL, 1999 162 Table 65. Actual and predicted estimate s of adult asthma hospital admissions per 10,000 population by zip code ar ea of residence in Pinellas County, FL, 1999 164 Table B-1. Total number of asthma hos pital admissions within different age strata by zip code area of resi dence in Hillsborough County, FL, in 1999 205 Table B-2. Total population distribution w ithin separate age strata by zip code area of residence in Hillsborough County, FL, in 1999 206 Table B-3. Total number of asthma hosp ital admissions and standard weight (Standard 1,000,000 population) for se parate age category (strata) group used to calculate age-adjust ed rates by direct adjustment technique 207 Table B-3. Total number of asthma hos pital admissions and standard weight for separate age category group used to calculate ageadjusted rates by direct adjustment technique (cont.) 207 Table B-4. Total number of asthma hospita l admissions within different strata by race (White, Black, Hispanic and Other) and by gender (Male and Female) by zip code area of residence in Hillsborough County, FL, in 1999 208 Table B-5. Total population distribution w ithin different strata by race (White,

PAGE 8

vii Black, Hispanic and Other) and by gender (Male and Female) by zip code area of residence in Hillsborough County, FL, in 1999 208 Table B-6. Total number of asthma hosp ital admissions and standard weight (Standard 1,000,000 population) within separate race and gender category (stratum) group used to calculate r aceand genderadjusted rates by direct adjustment technique 209 Table C-1. Distribution of socioeconomic status indicators by zip code areas of residence, Hillsborough County, FL 210 Table C-1. Distribution of socioeconomic status indicators by zip code areas of residence, Hillsborough County, FL. (cont.) 211 Table C-2. Distribution of socioeconomic status indicators by zip code areas of residence, Hillsborough County, FL 211 Table C-2 Distribution of so cioeconomic status indicators by zip code areas of residence, Hillsborough County, FL. (cont.) 212 Table C-3 Distribution of socioeconomic status indicators by zip code areas of residence, Hillsborough County, FL. 212 Table C-3 Distribution of socioeconomic status indicators by zip code areas of residence, Hillsborough County, FL. (cont.) 213 Table D-1. Association between environm ental exposure to particulate matters with a diameter of 10 microns or less (PM10) and children asthma hospital admissions stratifie d by poverty status (SES – socioeconomic status; RRrate ratio; 95% CI – 95% Confidence Intervals) 214 Table D-2. Association between environm ental exposure to particulate matters with a diameter of 10 microns or less (PM10) and adult hospital admissions stratified by poverty status (SES – socioeconomic status; RRrate ratio; 95% CI – 95% Confidence Intervals) 214 Table D-3. Association between envir onmental exposure to sulfur dioxide (SO2) and children asthma hospital admissions stratified by poverty status (SES – socioeconomic status; RRrate ratio; 95% CI – 95% Confidence Intervals) 215 Table D-4. Association between envir onmental exposure to sulfur dioxide (SO2) and adult asthma hospital ad missions stratified by poverty status (SES – socioeconomic status ; RRrate ratio; 95% CI – 95% Confidence Intervals) 215 Table D-5. Association between environm ental exposure to particulate matters with a diameter of 10 microns or less (PM10) and children asthma hospital admissions stratified by family income (SES – socioeconomic status; RRrate ratio; 95% CI – 95% Confidence Intervals) 216 Table D-6. Association between environm ental exposure to particulate matters with a diameter of 10 microns or less (PM10) and adult asthma hospital admissions stratified by family income (SES – socioeconomic status; RRrate ratio; 95% CI – 95% Confidence Intervals) 216

PAGE 9

viii Table D-7. Association between envir onmental exposure to sulfur dioxide (SO2) and children asthma hospital admissions stratified by family income (SES – socioeconomic st atus; RRrate ratio; 95% CI – 95% Confidence Intervals) 217 Table D-8. Association between envir onmental exposure to sulfur dioxide (SO2) and adult asthma hospital ad missions stratified by family income (SES – socioeconomic st atus; RRrate ratio; 95% CI – 95% Confidence Intervals) 217 Table D-9. Association between environm ental exposure to particulate matters with a diameter of 10 microns or less (PM10) and children asthma hospital admissions stratified by th e percentage of white collar employees (SES – socioeconomic status; RRrate ratio; 95% CI – 95% Confidence Intervals) 218 Table D-10. Association between environm ental exposure to particulate matters with a diameter of 10 microns or less (PM10) and children asthma hospital admissions stratified by th e percentage of white collar employees (SES – socioeconomic status; RRrate ratio; 95% CI – 95% Confidence Intervals) 218 Table D-11. Association between envir onmental exposure to sulfur dioxide (SO2) and children asthma hospital admissions stratified by the percentage of white collar em ployees (SES – socioeconomic status; RRrate ratio; 95% CI – 95% Confidence Intervals) 219 Table D-12. Association between envir onmental exposure to sulfur dioxide (SO2) and adult asthma hospital admissions stratified by the percentage of white collar em ployees (SES – socioeconomic status; RRrate ratio; 95% CI – 95% Confidence Intervals) 219 Table D-13. Association between environm ental exposure to particulate matters with a diameter of 10 microns or less (PM10) and children asthma hospital admissions stratified by singl e parent with children status (SES – socioeconomic status; RRrate ratio; 95% CI – 95% Confidence Intervals) 220 Table D-14. Association between environm ental exposure to particulate matters with a diameter of 10 microns or less (PM10) and adult asthma hospital admissions stratified by singl e parent with children status (SES – socioeconomic status; RRrate ratio; 95% CI – 95% Confidence Intervals) 220 Table D-15. Association between envir onmental exposure to sulfur dioxide (SO2) and children asthma hospital admissions stratified by the percentage of single parent living with children (SES – socioeconomic status; RRrate ratio; 95% CI – 95% Confidence Intervals) 221 Table D-16. Association between envir onmental exposure to sulfur dioxide (SO2) and adult asthma hospital admissions stratified by the percentage of single parent living with children (SES – socioeconomic status; RRrate ratio; 95% CI – 95% Confidence Intervals) 221

PAGE 10

ix Table D-17. Association between environm ental exposure to particulate matters with a diameter of 10 microns or less (PM10) and children asthma hospital admissions stratified by overcrowded housing conditions (SES – socioeconomic status; RRrate ratio; 95% CI – 95% Confidence Intervals) 222 Table D-18. Association between environm ental exposure to particulate matters with a diameter of 10 microns or less (PM10) and adult asthma hospital admissions stratified by overcrowded housing conditions (SES – socioeconomic status; RRrate ratio; 95% CI – 95% Confidence Intervals) 222 Table D-19. Association between envir onmental exposure to sulfur dioxide (SO2) and children asthma hospital admissions stratified by overcrowded housing conditions (SES – socioeconomic status; RRrate ratio; 95% CI – 95% Confidence Intervals) 223 Table D-20. Association between envir onmental exposure to sulfur dioxide (SO2) and adult asthma hospita l admissions stratified by overcrowded housing conditions (SES – socioeconomic status; RRrate ratio; 95% CI – 95% Confidence Intervals) 223 Table D-21. Association between environm ental exposure to particulate matters with a diameter of 10 microns or less (PM10) and children asthma hospital admissions stratified by UDI index (SES – socioeconomic status; RRrate ratio; 95% CI – 95% Confidence Intervals) 224 Table D-22. Association between environm ental exposure to particulate matters with a diameter of 10 microns or less (PM10) and children asthma hospital admissions stratified by UDI index (SES – socioeconomic status; RRrate ratio; 95% CI – 95% Confidence Intervals) 224 Table D-23. Association between envir onmental exposure to sulfur dioxide (SO2) and children asthma hospital admissions stratified by UDI index (SES – socioeconomic status ; RRrate ratio; 95% CI – 95% Confidence Intervals) 225 Table D-24. Association between envir onmental exposure to sulfur dioxide (SO2) and adult asthma hospital ad missions stratified by UDI index (SES – socioeconomic status; RRrate ratio; 95% CI – 95% Confidence Intervals) 225

PAGE 11

x List of Figures Figure 1 The total number of hosp ital admissions for childhood and adult asthma and average ambient particle concentration, g/m3, in Hillsborough County, FL, by calendar quarter during the period of 1997-1999 45 Figure 2 The total number of hosp ital admissions for childhood and adult asthma and average ozone concentration, ppb, in Hillsborough County, FL, by calendar quarter dur ing the period of 1997-1999 48 Figure 3 The total number of hospital admissions for childhood and adult asthma and average sulfur dioxide concentration, ppb, in Hillsborough County, FL, by calendar quarter during the period of 1997-1999 50 Figure 4 The total number of hosp ital admissions for childhood and adult asthma and average number of peak concentrations for sulfur dioxide in Hillsborough County, FL, by calendar quarter during 1997-1999 51 Figure 5 Average ambient temperatur e, F, and number of hospital admissions for childhood and adult asthma by calendar quarter in Hillsborough County, FL, 1997-1999 53 Figure 6 Average number of total tree pollen counts and number of hospital admissions for childhood and adult asthma by calendar quarter during the period of 1997-1999, in Hillsborough County, FL 55 Figure 7 Average number of total weed pollen counts and number of hospital admissions for childhood and adult asthma by calendar quarter during the period of 1997-1999, in Hillsborough County, FL 56 Figure 8 Average number of total gr ass pollen counts and number of hospital admissions for childhood and adult asthma by calendar quarter during the period of 1997-1999, in Hillsborough County, FL 57 Figure 9 Scatter plot of ambient temp erature and childhood asthma hospital admissions in Hillsborough County, FL, by calendar quarter 19971999 59 Figure 10 Scatter plot of ambient te mperature and adult asthma hospital admissions in Hillsborough County, FL, by calendar quarter 19971999 59 Figure 11 Standard residual deviance dist ribution against predicted values for childhood asthma hospital admissions 61 Figure 12 Standard residual deviance dist ribution against predicted values for adult asthma hospital admissions 61 Figure 13 Description of various sect ors of environmental exposure 74

PAGE 12

xi Figure 14 Distribution of residual deviances against predicted values for the fitted regression model with childhood asthma hospital admissions as a dependent variable in 1997 140 Figure 15 Distribution of residual deviances against predicted values for the fitted regression model with childhood asthma hospital admissions as a dependent variable in 1998 142 Figure 16 Distribution of residual deviances against predicted values for the fitted regression model with childhood asthma hospital admissions as a dependent variable in 1999 144 Figure 17 Distribution of residual deviances against predicted values for the fitted regression model with adult asthma hospital admissions as a dependent variable in 1997 146 Figure 18 Distribution of residual deviances against predicted values for the fitted regression model with adult asthma hospital admissions as a dependent variable in 1998 148 Figure 19 Distribution of residual deviances against predicted values for the fitted regression model with adult asthma hospital admissions as a dependent variable in 1999 150 Figure 20 Distribution of residual deviances against predicted values for the fitted socioeconomic deprivation index and sulfur dioxide regression model with childhood as thma hospital admissions as a dependent variable 154 Figure 21 Scatter plot of area so cioeconomic deprivation index and childhood asthma hospitalizations 155 Figure 22 Distribution of residual deviances against predicted values for the fitted socioeconomic deprivati on index regression model with childhood asthma hospital admissions as a dependent variable 156 Figure 23 Scatter plot of area socio economic deprivation index and adult asthma hospitalizations in 1999 159 Figure 24 Distribution of residual deviances against predicted values for the fitted socioeconomic deprivati on index regression model with adult asthma hospital admissions as a dependent variable in 1999 159

PAGE 13

xii List of Pictures Picture 1 Crude childhood asthma hos pitalization rates per 10,000 by zip code area of residence in Hillsborough County, FL, in 1999 35 Picture 2 Crude adult asthma hospita lization rates per 10,000 by zip code area of residence in Hillsborough County, FL, in 1999 36 Picture 3 Ambient air quality moni toring network in Tampa Bay, FL, 1997-1999 92 Picture 4 PM10 concentrations, microg/m3 in Tampa Bay, FL, in 1997 95 Picture 5 PM10 concentrations, microg/m3, in Tampa Bay, FL, in 1998 96 Picture 6 Interpolated PM10 concentrations, microg/m3, in Tampa Bay, FL, in 1999 96 Picture 7 PM10 concentrations by postal zi p code area of residence, microg/m3, in Tampa Bay, FL, in 1999 97 Picture 8 Relatively high and low averag e annual environmental exposure to particulate matter category (str atum) areas in Hillsborough County, in 1999 98 Picture 9 Interpolated SO2 concentrations, ppb, in Hillsborough County, in 1999 102 Picture 10 Relatively high and low aver age annual environmental exposure to sulfur dioxide category (stratum ) areas in Hillsborough County, in 1999 103 Picture 11 Interpolated ambient ozone concentrations in Tampa Bay, FL, in 1999 107 Picture 12 Distribution of ambient air pollution by zip code area of residence in Hillsborough County, FL, in 1999 107 Picture 13 Percentage of people livi ng below poverty level by zip code area of residence in Hillsborough County, FL, in 1999 114 Picture 14 Percentage of people empl oyed in white-collar occupations by zip code area of residence 120 Picture 15 UDI score by zip code ar ea of residence in Hillsborough County, FL, 1999 129 Picture 16 SDI score by separate zip code area of residence in Hillsborough County, FL, 1999 136 Picture 17 Distribution of SDI score by zip code area of residence in Hillsborough County, FL, 1999 136

PAGE 14

xiii Association between Area Socioeconomi c Status and Hospital Admissions for Childhood and Adult Asthma Tomas Tamulis ABSTRACT Despite an improved understanding of the di sease, the prevalence of asthma and asthma-related morbidity continue to rise, particularly among mi nority and inner-city populations. Despite the growing epidemic of as thma, the surveillance of disease at the state or even local levels is very limited. Such information is very important to identify high-risk population groups and to design mo re effective community-based preventive interventions or risk management progr ams that may modify these trends. The study provided important informati on about spatial differences by the geographical area of residence and changes in asthma hospital admissions over time in the selected area. Envi ronmental exposure to ambient ai r pollution by ambient particles, sulfur dioxide and ozone was a significant factor to expl ain the increase in asthma hospitalizations in simple regr ession analysis, but was not sign ificant after the adjustment to area socioeconomic status characteristic s. Sulfur dioxide was the only significant independent variable in a multiple adjusted regression model of hospitalizations for childhood asthma, however, more detailed environmental exposure assessment by calendar quarter suggested that ambient air pollution by sulfur dioxide is not significant variable in the multiple regression model. Future asthma prevention interventions and risk management programs should addre ss population groups described by such socioeconomic status characteristics as poverty, unskilled workers, single parent families with children, families having no vehicle available, people living in less crowded households or socially excluded conditions with out adequate family members or relatives support, and also people residing in hous es heated by fuel. Developed complex area socioeconomic deprivation index was shown to be a significant predictor of hospital

PAGE 15

xiv admissions for childhood and adult asthma by zi p code area of residence. Predictive loglinear regression model for asthma hospita lizations was further validated by using standard statistical model validation techni ques to estimate the accuracy of prediction with new independent dataset outside of our study area. Increase in complex area socioeconomic deprivation index by 1 extr a unit could explain th e increase by 7.9% in childhood and 7.5% in adult asthma hospita lization in 1997, 8.3% in childhood and 7.2% in adult asthma hospitalizations in 1998, a nd 7.7% in childhood and 6.7% in adult asthma hospitalizations in 1999 respec tively. Predictive log-linear regression model could be successfully applied to develop more effectiv e asthma prevention interventions and risk management programs and to address more sensitive population groups within specific high risk geographical areas.

PAGE 16

1 Chapter 1 ASSOCIATION BETWEEN ENVI RONMENTAL TRIGGERS AND HOSPITALIZATIONS FOR CHIL DHOOD AND ADULT ASTHMA 1.1. Literature Review 1.1.1. Etiology, pathology and epidemiology of asthma Asthma is characterized by reversible airflow obstruction and airway hyperresponsiveness with prominent clinical manifestations of wheezing, coughing, chest tightness and shortness of breath.1 Asthma is a multifactorial disease described by immunological, inflammatory and neurogenic m echanisms and associated with genetic, environmental, socioeconomic, and psychosocial factors .2 The underlying pathophysiology of asthma is airway inflammation of the large airways of the tracheo-bronchial tree th at may be initiated by environmental or occupational exposure to an alle rgen, respiratory tract infectio n, occupational aerosol, or such environmental stimulus as exer cise, cigarette smoke, or cold air.3 Asthma has been divided into “extrinsic” type (due to a specifi c allergen), or “intrinsic” type (when the sensitizing agent is unknown). The underlyi ng process driving and maintaining the asthmatic inflammatory process appears to be an abnormal or inadequately regulated CD+4 T-cell immune response. In asthma, the T-helper 2 (Th2) immune response is overactive, while cell-mediated Th1 activity is decreased. The Th2 subset produces cytokines including interleukin-4 (IL-4), IL-5, IL-6, IL-9, IL-10, and IL-13, which stimulate the growth, differentiation, a nd recruitment of mast cells, basophils, eosinophils, and B-cells, all of which are involved in inflammation, and allergic response.3 In the classical example of the allergic (atopic) asthma, the individual is

PAGE 17

2 sensitized by a specific environmental allerge n, such as a dust, pollen, or animal protein. A latent period with the excessive generati on of IgE antibodies is followed by allergen binding to IgE-coated mast cells, which induces mast cell degranula tion and release of a variety of active mediators, such as histam ine, prostaglandins, and leukotrienes. During the periods between asthmatic attacks, thes e atopic individuals continue to have hyperreactive airways. The rel ease of histamine and production of prostaglandins and leukotrienes induces both immediate a nd prolonged bronchoconstriction. Both the mucosal swelling and smooth muscle contractio n cause constriction of the airway lumen, and subsequent mucus secretion within the ai rway lumen further obstructs airway flow. Effective asthma management requires a long-term and multifaceted approach, including patient education, fr equent medical follow-up, beha vioral changes, effective drug therapy, and avoidance of such envir onmental asthma triggers as irritants and inhaled allergens .4 Despite major advances in ou r knowledge and understanding of its pathogenesis and medical treatment, asthma remains a common chronic disease that causes substantial morbidity and mortality. Data collected by the Center for Disease Control and Prevention (CDC) revealed la rge and unexplained increases in asthma prevalence, hospitalization rates, and mortality over the last decades in the United States.2 Asthma has been estimated to affect 17 milli on people or over 5% of the total population in the United States, with 10.4 million asth ma-related office visits to medical care providers, 1.9 million emergency room visits, and 466,000 hospital admissions annually.5,6 9 million children under 18 years of age (or 13 % of total children population) had asthma and more than 4 m illion children (6 %) had an asthma attack during the last year in 2000.7 The prevalence rate of asthma increased by 75 % and rose from 3.1% to 5.4% from 1980 to 1994.2 The most substantial increases oc curred among children aged 0-4 years (160 %, from 22.2 to 57.8 per 1,000, p<0.05) and 5-14 years of age (74 %, from 42.8 to 74.4 per 1,000, p<0.05).2 In 1994, asthma affected an estimated 4.8 million children of an estimated 68 million children under 18 years of age.5 Among children and young adults, 5-24 years old, the asthma death rate nearly doubled from 1980 to 1993. State-specific asthma prevalence data was estimated from the Behavioral Risk Factor Surveillance

PAGE 18

3 System (BRFSS) survey conducted by the Ce nter for Disease Control and Prevention (CDC), Atlanta, Georgia. The survey indica tes that approximately 7.2 % of adults aged 18 years or older reported having as thma in the United States in 2000.8 The National Center for Health Statistics (NCHS) annually conducts the Nationa l Health Interview Survey (NHIS), which registers self-reporte d asthma prevalence, asthma office visits, asthma emergency room visits, asthma hospita lization, and asthma deaths in subsets of the sample.2,9 Crude rates of lifetime and curr ent asthma, acute asthma attacks prevalence, emergency department room visits as an indicator of h ealth care utilization, asthma hospital admissions and asthma mo rtality by gender in the US in 2000 are presented in Table 1.7 Table 1 Prevalence rates of lifetime and curr ent asthma, severe asthma attacks, emergency room visits and hospital admissions for asthma, and asthma mortality in the US in year 2000 Gender Lifetime Asthma per 10,000 Current Asthma per 10,000 Asthma Attack Prevalence per 10,000 Emergency Room Visits per 10,000 Asthma Hospitalizations per 10,000 Asthma Mortality per 100,000 Male 1,080 640 360 60 15 1.2 Female 1,190 830 500 74 19 2.0 Average 1,135 735 430 67 17 1.6 Questions on lifetime and current asthma prevalence in the BRFSS are comparable to the NHIS, however, prevalence estimate s vary due to sampling design and chance .1 The BRFSS estimated 25.2 million (11.8%) with lifetime asthma and 16 million adults (7.5%) with current asthma, as compared to 21.9 million (10.7%) and 14.0 million adults (6.8%) who were diagnosed with lifetime a nd current asthma, respectively, by the NHIS in 2002.1 The difference between these estimates was shown to be statistically valid based on confidence intervals for the preval ence rates. The measurement of asthma prevalence was recently changed because of redesign of the NHIS survey in 1997. As a result, most of the decrease in asthma prevalence could be explained by definitional changes in asthma surveillance. Because th e newer estimates cannot be compared with pre-redesign estimates, it will be necessary to follow the trend for several more years to determine whether asthma prevalence will declin e, plateau, or continue to increase as it

PAGE 19

4 did through the 1980’s and 1990’s. However, the increasing trend in asthma prevalence and hospital admission rates during 1980-1994 was evident in every region of the United States.2 The National Hospital Discharge Surv ey disclosed significant geographical differences in the rate of hospitalizat ions for asthma in the United States.1 State-to-state differences also occur, with state-specifi c estimates for current asthma prevalence ranging from 4.8% in Oklahoma to 7.2% in Nevada.11 Asthma hospital admission rates by geographical region were 24.5 in the Nort heast, 18.4 in the Midwest, 15.8 in the South, and 14.2 per 10,000 persons in the West. Asthma morbidity and mortality have increased disproportionately among innercity poor ethnic minority children an d young adults in the United States.12 Urban areas have a higher burden of asthma compared with less populous areas, with asthma prevalence growing at a rapid rate.3 Inner-city children are a specific sensitive group of urban population and have the highest preval ence of asthma and the highest asthmaassociated hospitalization rates. Asthma is the leading chronic illness in children affecting almost five million children, and the fourth leading cause of disability in children.13 Population-based surveys of childhood asthma in inner-city areas of New York City and Chicago estimated the prevalence of physic ian-diagnosed asthma at between 8.6% and 14.3% respectively, which is two to thr ee times higher than the country rate.3 Among some inner-city elementary school childre n populations, prevalence of asthma was reported at more than 20%.3,14 Asthma is the most common cause for hospital admissions in children, accounting for 159,000 annual hospitaliz ations with an average length of stay of 3.4 days in the US.2 The average rate of asthma hospitalizations among children 18 years and younger in Boston was over twi ce the rate in Massa chusetts in 2000.3 Childhood asthma hospitalization rates increased threefold in the St Louis metropolitan area from 1983 to 1995, much higher than th e increase of hospita lizations across the entire state of Missouri.3 Previous studies have also indicated that asthma mortality is higher among poor ethnic minority inner-c ity children than other children.3,6 Asthma significantly affects the quality of life of many people and is responsible for a large social and economical burden in the United States, causing 10 million lost school days and 3 million lost work days.13 Asthma epidemic also caused heavy increases in

PAGE 20

5 health care expenditures. Dir ect and indirect financial impact of the disease was estimated to be $12.7 billion in 1998 and $14.0 billion in 2002, with asthma hospital admissions accounting for the majority of the cost.1, 10 Understanding the disproportionate increase of as thma in inner cities may provide insight into the roots of the asthma epidemics and effective diseas e management. A number of hypotheses and theories have been proposed to explain the increasing trend in the prevalence and severity of asthma.3,15 Specific factors suggested to explain differences in asthma prevalence in time and place include an increase in allergen exposure (particularly to house dust mites and cockroaches), tobacco smoking by women of child bearing age, dietary differences (particularly salt intake but possibly also cha nges in antioxidant constituents in diet), occupational agents encountered at work, su ch as isocyanates, flour dust, wood dust, laboratory animal urine, and solder flux, indoor and outdoor air pollution, and respiratory viral infection. Complex factors, which could a ccount for the dispropor tionate increase in the prevalence and severity of asthma include changi ng patterns of medication, inadequate preventive health care for asthma management, limited disease knowledge and management skills, and psychosocial stress of living in poor inner-city neighborhoods.12, 16 Previous inner-city asthma epidemio logical studies su ggest that the contributions of ethnicity, poverty and reside nce area cannot be sepa rated and should be viewed together when trying to understand risk factors for asthma. Eggleston et al. proposed the model of environmental exposur e and societal susceptibility factors influencing airway inflammation and obstruc tion that subsequently leads to severe asthma attack.12 In this model, environmental exposur e to such allergens as dust mites, cockroaches and home pets; such ambient air pollutants and in door irritants as particulate matter (PM), sulfur dioxide (SO2), ozone (O3), environmental tobacco smoke (ETS); such susceptibility factors as psychosocial stress, limited knowledge about disease management, inappropriate medication use, re stricted health care resources and poor access to quality health care services; and ge netic background, race, gender and age as individual susceptibility factors are the ma in factors responsible for the increase in asthma morbidity and mortality in the inner city areas.12

PAGE 21

6 1.1.2. Environmental triggers of asthma Several factors have an important influe nce on an individual’s susceptibility to environmental exposure to ambient air pollution Genetic background, age, nutrition, and lifestyle factors are among these and are mutual ly inclusive and influenced by each other, and could describe each i ndividual’s susceptibility.17 Wide genetic variability modulates predisposition to perturbation of envi ronmental contaminants and allergens.18 Genetic polymorphism is closely associated with suscep tibility to a wide variety of environmental contaminants, including indoor and outdoor pollutants and allergens.18, 19 The likelihood of an adverse response to an inhaled envi ronmental pollutant depends on the degree of exposure, the site of deposition, the rate of remo val or clearance, and the susceptibility of the exposed person.18,19 The intake of pollutants into th e lungs and reten tion at potential sites of injury depend on the physical and chem ical properties of the pollutant as well as the extent of activity of the subject exposed.18 Gases that are highly water soluble, such as sulfur dioxide and formaldehyde, are almost completely removed in the upper airways. Less-soluble gases, such as nitrogen dioxide and ozone, pene trate to the small airways and alveoli. Environmental exposures, pathwa ys of absorption, tissue distribution, ability to bio-transform and eliminate environmenta l irritants and allergens are different in children as compare to adults,17 and aforementioned differences must be taken into account when considering the health im pact of environmental exposure. The association between air pollution and human health has been identified most clearly in episodic situations Several major episodes esta blished the fact that high concentrations of air pollu tion increased chronic respiratory disease morbidity and mortality.18 Recent epidemiological studies demonstrated that relatively low levels of air pollution exposure to respirable particles, black smoke, sulfur dioxide, and ozone (i.e. levels below current ambient air quality guide lines) may also be associated with daily hospital admissions for respirat ory conditions in cluding asthma.20,21,22,23 The Inner-City Asthma Study (ICAS) and the Inner-City Air Pollution study (ICAP), conducted by the National Institute of Environmental Health Sc iences and the National Institute of Allergy and Infectious Diseases, concluded that by using only ambient air measurements, simple

PAGE 22

7 models could predict indoor air particle concentrations.24 Previous longitudinal study of a children panel group concluded that personal particulate matter exposure was strongly correlated and could be predicted by ambient particle concentrations.25 Both experimental and epidemiological studies have shown that commonly measured air pollutants have th e potential to aggravate asthma by either direct irritation or by enhancing the effects of allergens to which the individual is sensitive.26,27 The main advantage of epidemiologic studies of air pollution is that they reflect the real environment, while the main disadvantages are existing difficulties and limitations in environmental exposure and health effect assessment. Conducting an air pollution epidemiological study does not require extrapolation from la boratory to real environment conditions, extrapolation from laboratory an imals to humans as main subjects of study, extrapolation from selected volunteers in e xperimental study to the population at large, extrapolation from short-term exposure to l ong-term exposure, or extrapolation from high to low doses. However, experimental studi es permit the control of other important factors, e.g. pollutant exposure, and allow physiological measurements that often cannot be employed in epidemiological studies. The understanding of associations in epidemiological studies benefits from labor atory work, while epidemiological findings generate hypotheses for experimental research. During the last few decades significant impr ovement in ambient air quality has been achieved nationwide. The improvement in air quality resulted mainly from reduction of pollution from industrial sites and coal-burning facilities.23, 28 On the other hand, air emissions from mobile sources, such as tr ansportation, has been increasing in recent years.23,28 The most common ambient air polluta nts in urban areas nationwide are respirable particulate matter (PM10), nitrogen dioxide (NO2), sulfur dioxide (SO2), and ozone (O3).18,27 Respirable particulate matter with a diameter of less than 10 microns (PM10) was shown to be associated with increased requirements for asthma medication, exacerbations of asthma, and hospital admissions for severe asthma.27 The biological effect of particles is determined by their physical and chemical nature, as well as th eir distribution in the respiratory tract. However, the identification of specific factors that may mediate asthma

PAGE 23

8 exacerbation is a very difficult task.27 Active agents in particulate matter include silica, such metal ions as iron, vanadium, nickel, a nd copper, organic residues, acid aerosols, and biological contaminants.27 Most inhaled particles with an aerodynamic diameter of more than 5 m deposit in the upper airways or larger lower airways, while smaller particles are more likely to deposit in the pe riphery of the lungs. Deposition also depends on the pattern of breathing, and deep breat hing increases the upper size cut-off for particles penetrating the lower airways from about 10 to 15 m.29 Particles are cleared from the trachea, bronchi and bronchioles by traveling upwards on the mucociliary escalator and then expelled by coughing or swallowing. Particles deposited in the alveolar region are largely cleared by l ung macrophages and dissolution. Experimental studies have shown that laboratory animal s develope inflammation associated with particle deposition.29 Ultrafine particles (about 2 m or less diameter) are able to penetrate the epithelium and vasc ular walls and are then tran sported by the blood to distal organs where proinflammatory events occur. In experimental and cl inical studies, high concentrations of acid aerosol particles have been shown to have such deleterious effects on the lungs as bronchoconstriction, hyperrespon siveness, and impaired mucocilliliary clearance.29 Diesel exhaust particles (DEP) appear also to act as an adjuvant for the production of IgE.27,29 Previous epidemiological studies shown th at hospital admissions for asthma during both winter and summer we re significantly associated with daily particle concentrations.27, 29 Daily frequencies of asthma medi cation (bronchodilator use), lung function impairment, respiratory symptoms and hospital/emergency room admissions were associated with short-term increase s in respirable particle concentrations.29 A European expert panel meeting reviewed availa ble literature and repor ted that an increase in the 24-hour mean of PM10 concentration by 10 g/m3 would increase respiratory admissions by 0.5 %, use of bronchodilators in asthmatics by 2 %, and exacerbation of symptoms among asthmatics by 5 %.29 Both increased emergency department room visits and increased hospital admission rates for asthma have been associated with ambient exposure to sulfur dioxide (SO2) in previous experimental a nd epidemiological studies.27 SO2 is a highly water-soluble gas

PAGE 24

9 which can be almost completely absorbed during nasal breathing, forming sulfuric acids and bi-sulfates. These irritants are thought to act directly on smooth muscle or via sensory nerve fibers causing reflex bronchoc onstrictions. The experimental work on the effect of sulfur dioxide has been review ed several times in the last five years .29 In controlled exposure experimental st udies, a high concentration of SO2 induced bronchoconstriction, hyperactivity and airway inflammation, while more relevant outdoor concentrations resulted only in mild respir atory effects. Repeated pre-exposure to high levels of SO2 followed by respiratory antigen chal lenge in guinea pigs resulted in increased sensitization rates. Controlled e xposure studies in huma ns have shown that lower concentrations of SO2 increased airway resist ance in asthmatic subjects.29 The response to SO2 varies greatly among individuals. A few studies on humans have shown an increase in inflammatory cells in bronchoa lveolar lavage after exposure to high levels of SO2.29 Variations in peak flow measurem ents, wheezing and bronchodilator use in a panel group of children with chronic respirator y symptoms have been associated with the average 24-h mean value of 105 g/m3 for SO2. A similar relationship was demonstrated in another study of changes in peak flow, asthma medication, symptoms and school absence in a panel group of asthma tic children where the 24-h mean SO2 concentration ranged from 3 to 383 g/m3.29 A significant association was also found between peak flow, respiratory symptoms and medication in an adult group with asthma and increased 24-h SO2 concentrations with a maximum concentration of 117 g/m3.29 While morbidity in asthmatic children was significantly predicte d by high concentrations of sulfur dioxide and sulfate, only weak and inconsistent eff ects on respiratory symptoms and peak flow measurements were seen in anot her panel group of adult asthmatics.29 Nitrogen dioxide (NO2) is a poorly water-soluble gas a nd is therefore deposited far more peripherally in the lungs than the highly water-soluble SO2. NO2 can act as a potent oxidant, and may affect the stru cture of proteins and lipids.29 A number of reviews of such experimental studies have recently been published.29 Nitrogen dioxide at high doses has been shown to cause several biochemical alterations in animal studies. The host defense systems are affected, there is in creased airway reac tivity and decreased pulmonary function and some findings have also indicated an enhanced allergic

PAGE 25

10 sensitization. In controlled expos ure chamber studies, exposure to NO2 in the lower concentration range has been le ss consistent in affecting re spiratory lung function, and its main effect has been to increase airway re sistance. Asthmatic patients and patients with other chronic pulmonary diseases are more susceptible. Studies on humans have also demonstrated inflammatory effects and a dverse effects on the immune system, e.g. impaired inactivation of influenza virus, as a result of exposure to NO2.29 The NO2 concentrations used in these experimental st udies have, with a few exceptions, been much higher than typical outdoor levels.29 In contrast to the effect of ‘traditional’ criteria ambient air pollutants such as sulfur dioxi de and particulates, nitrogen dioxide has generally not been demonstrated to have acute effects on respiratory health in previous epidemiological studies.30 Due to insufficient and inconsistent epidemiological results it has not been possible to es tablish exposure-response rela tionships for ambient air pollution by NO2. NO2 may be more important as an ambient air pollutant playing a crucial role in production of ozone rather than considering its effect on the airways. 27 Recent epidemiological studies have also provided some additional evidence that nitrogen dioxide may enhance the asth matic response to inhaled allergens.29 Ozone (O3) may damage lung tissue, reduce l ung function, and also sensitize the lungs to other environmental irritants.29 O3 induces inflammatory effect in lower airways favoring the migration into nasal and bronc hial mucosa of eosinophils, neutrophils, eosinophil peroxidase, myeloperoxidase, eosinophil cationic protein and other inflammatory mediators.29 It was clinically shown that th e number of neutrophils and the level of some prostaglandins increased 3-h after termination of a 2-hour exposure to 0.4 or 0.6 ppm O3.29 Studies on allergic asthmatics revealed an increased sensitivity to inhaled allergens in subjects pre-expose d as compare to non-exposed to ozone.27,29 Previous epidemiological studies have shown that environmental exposure to elevated levels of ambient ozone has been associated with in creased frequency of emergency department visits and hospital admissions for asthma and hospital admissions for other respiratory causes.31 Higher levels of O3 have been also found associated with school absenteeism among asthmatic children.31 Given previous experi mental findings of the responses to controlled exposures to ozone, one would expect asthmatics to be

PAGE 26

11 particularly sensitive to ozone.31 On one hand, most contro lled-exposure studies have failed to document greater response by asthmatics.31 On the other hand, several epidemiological studies have demonstrated an increase in exacerbation of asthma in asthmatic panel studies, and in hospital admissions and emergency room visits for asthma .31 This discrepancy in identifying asthmatics as a sensitive population group could be the result of the exclusion of se vere asthmatics from the controlled exposure studies, or a failure to simulate the comp lex mix of air pollutants to which such individuals may be exposed. Howe ver, it could also be argu ed that the effects in the epidemiologic studies are due to other environm ental irritants or allergens rather than to ozone.31 Some air pollution related incidents w ith asthma aggravation do not depend only on the increased level of air pollution but rather on climatic factors that favor the accumulation of air pollutants at ground level.29 Previous studies have revealed that cold air may induce a bronchoconstr iction process in asthmatics.32 Exacerbation of asthma symptoms also was linked with atmospheric relative humidity and barometric pressure.32 Previous studies have demons trated that urbanization a nd high levels of vehicle emissions along with changes in westernized lifestyle all are correlated with the increasing frequency of pollen-induced resp iratory allergy. People who live in urban areas are more affected by pollen-induced al lergy than those from rural areas. Pollen allergy has been one of the most frequent models used to study the interrelationship between air pollution and respiratory allergic diseases. There is also evidence that air pollutants may promote airway sensitiza tion by modulating the allergenicity of airborne allergens .29 Furthermore, airway mucosal damage and impaired mucociliary clearance induced by air pollution may facilitate the access of inhaled allergens to the cells of the immune system. Several factors influence this interaction, in cluding type of air pollutants, plant species, nutrient balance, clim atic factors, airway sensitization level and hyperresponsiveness of exposed s ubjects. Experimental eviden ce suggests that the effects of aeroallergens may be increas ed by exposure to air pollution.29 Exposure to aeroallergens (pollen, ragweed, gr ass, and fungal spores) is re lated to weather conditions

PAGE 27

12 and is a potential confounder in analyses of the effects of air pollution on daily hospital admissions for asthma. It is widely believed that current leve ls of ambient air po llution may provoke or exacerbate asthma. Analysis of emergency room attendance and hospital admissions for asthma offer a convenient method of testi ng the hypothesis that outdoor air pollution provokes asthma.29 Several recent studies suggested that even relatively low levels of air pollution – levels below recommended air po llution guidelines – are also significantly associated with daily hosp ital admissions for asthma.20-23 However some studies have observed only small changes in lung functi on, which were similar to those experienced by non-asthmatic subjects, while some other pr evious studies have shown no association between ambient air pollution and asthma. A nu mber of studies have shown that gaseous and particulate air pollution has been associat ed with the acute exacerbation and possibly the onset of asthma.28 In several recent studies, relatively low concentrations of ambient air pollution (levels below current ambient air pollution gui delines) were found to be associated with adverse health effects, in cluding daily hospital admissions for respiratory and heart conditions.21 Diminished lung function is anothe r possible adverse outcome of poorly managed asthma. Previous cohort studies also confirmed significant reduction of forced expiratory volume (FEV1) and forced vital capacity (FVC) in relation to increasing levels of ambient nitrogen dioxide (NO2), particular matter PM10 and PM2.5, ozone (O3), sulfur dioxide (SO2), and black smoke as an indicator of soot and diesel exhaust.28 With the exception of a few studies, most of these st udies have been based on single locations selected without a defined standardized methodology.33,34 Very often there are different data analysis strategies, which make subseque nt interpretation and ge neralization of final study results very limited. Due to significan t inter-correlation among various ambient air pollutants, these studies very of ten were not able to discern individual effects of specific pollutant versus general mixed effects of ambient air pollution. Inconsistencies among separate air pollution studies could also be explained by differences in exposure levels and differences in susceptibility to adve rse health effects among population subgroups examined by age, gender, race, and socioec onomic status. Because the generality of

PAGE 28

13 findings is uncertain and analysis strategies are very often different among separate studies, critics have questioned whether the st udy results indicate an effect of particles specifically or air pollution ge nerally.33,34 Advanced analytical methods were developed recently to address limitations of single-city time-series analysis by combining evidence across multiple geogr aphical locations.35, 36 John Samet with his colleagues from John Hopkins University developed and proposed two-stage Bayesian semi-parametric hierarchical models to estimate health eff ect over time within selected cities and to compare temporal patterns acro ss cities and geographical regions.34, 35, 36 Semi-parametric hierarchical approach was upda ted to address the existing limitations in air pollution studies and was incorporated into the Na tional Morbidity, Mortal ity, and Air Pollution Study (NMMAPS) for the period 1987-2000.36 Information on mortality, morbidity, meteorological conditions, socioeconomic characteristics, and air pollution representing numerous metropolitan areas was assembled and analyzed by using aforementioned loglinear regression models. A two-stage analytical approach by using hi erarchical bivariate time-series models described a statistically significant associa tion between daily part iculate and both daily cardio-respiratory and cardiovascular hospitalizations and mortality.33-36 In the first stage, city-specific analyses were conducted and relative mortality rates associated with particulate matter pollution for a selected c ity were estimated by using semi-parametric log-linear models. In the second stage, mu lti-city analyses were conducted to analyze geographical patterns in the pollution-relative rates to pr oduce overall estimates of the pollution effects by using Bayesi an hierarchical models. Firs t, a separate log-linear regression model of air polluti on and other confounders as rega rds daily total and cardiorespiratory mortality were devel oped to estimate relative rate values along with statistical significance for selected urban ar eas, and, in the next step, obt ained relative rate estimates were combined across urban areas to adjust fo r the different levels of uncertainty and to obtain an overall estimate. The aforementione d research group found particulate matter less than 10 microns in aerodynamic diameter to be consistently a ssociated with both total and cardio-respiratory mortality after adjusting to potential confounding by other pollutants. Combined analysis of the eff ect of particulate matter on mortality and

PAGE 29

14 morbidity in 10 metropolitan areas during th e period 1986-1993 concluded that increase in particulate exposure by 10 g/m3 was associated with 0.26% increase in mortality and 0.71% increase in hospital admissi ons for cardiovascular disease.34 The 20 U.S. cities study for the period 1987-94 revealed that tota l and cardio-respirator y mortality increased by approximately 0.5% and 0.68% per 10 g/m3 increase in PM10 respectively.35 National Morbidity and Mortality Air Pollution Study da tabase covered 100 U.S. cities for the period 1987-2000.36 The national average estimate of th e effect of particulate matter on mortality was shown to be largest at 1-day la g and was equal to an increase in mortality by 0.19% per 10 g/m3 increase in PM10. The estimate for summer (0.36) was shown to be more than twice as large as fo r winter, spring and fall (0.15, 0.14, and 0.14 respectively). However, the aforementi oned studies could not avoid exposure measurement bias because of the limited numbe r of ambient particle exposure monitoring stations in selected relatively large urban areas represented by c ounties. The interaction and interrelationship with other ambient air pollutants was not evaluated. Evaluation of socioeconomic confounding did not include individual data a nd did not represent existing differences and variations in socioeconomic status by geographical ar ea within selected counties. Differences in particle emission sources and chemical particle composition were not evaluated in the ove rall final hierarchical mode l. Finally, correlation in the overall hierarchical model (but not within single city) makes interp retation and further generalization of results very questionable. Taking overall evidence, curre ntly existing studies lack consistency as to the presence of different exposures and effects, the type of po llutant or where effects have been observed. In addition, a few studies ha ve adequately addr essed the issue of confounding or effect modification by other en vironmental and socioeconomic factors or examined the association in par ticular sensitive population groups.29 There are also very common problems encountered trying to disc ern the complex effect of a mixture of ambient air pollutants and a single pollutant effect, and to account for multicolinearity because of high intercorrelation among various ambient air pollutants. Despite the evidence of a correlation between the increasi ng frequency of respiratory allergy and an increase in ambient air pollution, the causal link and possible intera ction are still very

PAGE 30

15 speculative. Interpretation of studies is usually confounded by the effect of active and passive cigarette smoke and exposur e to indoor and outdoor allergens.3,29 Therefore the estimated crude association between environm ental exposure to ambient air pollution and asthma could be misleading and seriously biased by different confounding factors and possible interaction effect.

PAGE 31

16 Hypothesis Ambient air pollution and other probable environmental triggers of asthma are associated with severe exacerbations and hospi talizations for childhood and adult asthma. The aim of the study was (1) to describe th e geographical distribution and changes over time in childhood and adult asthma hosp ital admissions, and (2) to evaluate the association between various local environmen tal asthma triggers and hospital admissions for childhood and adult asthma in a given area. Study objectives 1. Describe the geographical distributi on and changes over the study time in hospital admissions for childhood and adult asthma in Hillsborough County, FL; 2. Evaluate the association between local environmental triggers of asthma and hospital admissions for childhood and adult as thma in the selected study area over the period of 1997-1999.

PAGE 32

17 1.2. Methodology 1.2.1. Study Population Severity of asthma may be measured (1) by the proportion of individuals with active asthma symptoms at a given time ; (2) by the number of exacerbations of respiratory symptoms resulti ng in emergency rooms visits ; or (3) by the number of hospital admissions for asthma during a given pe riod of time. Severity of asthma can be also studied as increase in th e duration of symptoms, in th e progression and loss of lung function, and in the frequency of use of medication or heal th care services. Registers containing data on asthma mortality, hosp ital visits, sick-leav es, and physicians’ diagnoses have often been used in studi es of adverse eff ects of air pollution.18 Such studies may be relatively inexpensive and valid if other possible risk factors and determinants can be taken into account. Se vere exacerbation of asthma symptoms, which resulted in the hospital admission for asthma as principal diagnosis, was the main health outcome of interest in our study. Asthma diagnosis and disease coding were based upon the International Classification of Disease, 9th Revision, Clinical Modification (ICD-9-CM).55 The principal codes of ICD-9-CM for asthma were 493.0-493.9. The State of Florida Hospital Inpatient Discharge Data was obtained from the Agency for Health Care Administration (AHCA). Since the fourth quarter of 1987, all hospitals have been required to submit detailed patient discharge data in the State of Florida. Asthma hospital admission event was defined based on primary diagnosis of asth ma in patient’s hosp ital discharge record. Only hospital admissions for asthma as a prin cipal diagnosis of hos pital discharge (493.0493.9 ICD-9-CM) were used in the study. The main study area covered Hillsborough County in Florida. St. Joseph, Tampa Gene ral, Brandon Medical Center, University Community, South Florida Ba ptist, Community-Carrolwood, Memorial, and Town and Country were the main leading hospitals, which admitted 94% of total study subjects living in Hillsborough County, FL, in 1999. St Joseph’s (Tampa) and Tampa General have the highest health care resources u tilization for asthma hospital admissions, and

PAGE 33

18 majority of patients admitted to the hosp ital (90.5% and 80% of total patients) were permanent residents of Hillsborough Count y, FL. The major sources of hospital admissions for asthma were emergency department rooms and physicians. Other reference sources of hospita l admission included clinics, other hospitals, health maintenance organizations (HMO), and fede ral or state court/l aw institutions. Personal information on age, gender, r ace, time of hospital admission, source of hospital admission, admitting hospital, and postal zip code area of residence was available. Because of existing etiological and pathological differences, asthma hospital admissions in children younger than 15 years of age and adults of 15 years of age and older population groups were analyzed separa tely. There were a total of 1712 hospital admissions for asthma as a principal dia gnosis reported in Hillsborough County, FL, in 1999. 781 children younger than 15 years and 931 adults of 15 years of age and older were admitted to the hospital with seve re asthma in 1999. There were 729 hospital admissions for males and 983 hospital admissions for females. There were a total number of 1677 and 1406 hospital admissions, incl uding 774 and 549 childhood asthma hospital admissions, in 1997 and 1998 respectively. The race was defined as white (Caucasian), black (African American), Hispanic a nd others. The others group included Native American, Asian, multiracial persons and pe ople with unknown race. In 1999, whites had the highest (828) and others had the lowest number of hospitalizations (75), while Blacks and Hispanics shared 447 and 362 hospital admi ssions of asthma as a primary diagnosis by race respectively. The standard direct adjustment technique was used to calculate adjusted (standardized) hospitalization rates by personal demographic characteristics. Denominators for computing crude and sta ndardized rates repres ented population of specific zip code areas of residence and were obtained from the U.S. Census 2000.56 The standard US 2000 population was used as a re ference population to obtain the standard weights within different categories (strata) of af orementioned demographic characteristics. The reference study p opulation included all persons who were permanently residing in Hillsborough County, FL, in 1999 (n=1,009,855). There were 794,358 adults and n=215,497 children living in Hillsborough county, FL, in 1999. Age-,

PAGE 34

19 gender-, and race-adjusted asth ma hospital admission rates were standardized to the US 2000 standard population obtained from the US Census 2000. To calculate age-adjusted rates, the study population and all persons admitted for asthma were divided into 10 separate categories (strata): less than 5 years (0-4 or 4 group); equal or more than 5 and less than 15 years (5-14 group) ; equal or more than 15 and less than 25 years (15-24 group); equal or more than 25 and less than 35 years (25-34 group); equal or more than 35 and less than 45 years (35-44 group); equal or more than 45 and less than 55 years (45-54 group), equal or more than 55 and le ss than 65 years (55-64 gr oup); equal or more than 65 and less than 75 years (65-74 group); equal or more than 75 but less than 85 years (75-84 group); and equal or more than 85 years ( 85 group). Race was divided into white (Caucasian), black (African American ), Hispanic, and Others population groups, respectively. There was a much higher proportion of white residents (n=644,503) compared with other racial groups liv ing in Hillsborough County, FL, in 1999. 178,495 Hispanic, 149,105 black, and 37,752 Others repres ented other major groups of permanent resident population in our study area. There were slightly more females (n=515,317) than males (n=494,538) living in Hillsborough County, FL, in 1999. The total children younger than 15 years a nd adults of 15 year s of age and older population groups by zip code area of residence were used to calcula te crude rates, and the standard US 2000 population was used to calculated age, gende r, and race-adjusted asthma hospital admission rates. Standardiz ed rates reflected selected reference population and were used to evaluate possibl e confounding effect by specific age, gender or race differences within the study area and ge neral population characte ristics in the US. 1.2.2. Study Design Environmental asthma epidemiological st udies have been categorized based on different dimensions, such as the unit of data analysis, the directi onality of time in the selected study design and types of available data sources. The special characteristics of air pollution exposure and availa bility of measurement data have led to the development of special study designs with most often employed aggregated data.18, 38 In air pollution

PAGE 35

20 epidemiology, most studies are partly ecological as air pollution exposure is very seldom measured at the individual level. Studies can also be either longitudi nal or cross-sectional and they are either based on population sampli ng or some type of incomplete sampling (case-control studies). The main objectives in environmental asthma studies could be defined as to :29 (1) Evaluate differences between populati on subgroups on exposure to ambient air pollution over time and space and on the hea lth effect of interest related to those exposures; (2) Ascertain whether there is an association between exposure to a specific air pollutant or to a mixture of pollutants and to the specific health outcome of interest, and, whenever possible, to describe expo sure-response relationship quantitatively. Exposure to air pollutants can be monito red at different levels. The different measures or indicators can be listed in order of increasing accuracy as follows:38 Qualitative (categorical) assessment of exposure that distinguishes between relatively high and low exposure; Fixed geographical location concentr ation measurements and continuous monitoring; Multi-microenvironment quality assessm ent and concentration measurements including time activity pattern information; Personal exposure monitoring; and Biological monitoring using biol ogical exposure indicators. Ideal studies of air pollution exposure s hould include biological dose markers or personal exposure measurements. Time-activity patterns can dramatically affect personal exposure and individual variation can be more accurately asse ssed by biological dose markers and personal samplers. Since such m easurements are very expensive in a fullscale epidemiological study, fixed site measur ements, modeled concentrations, or even qualitative categorical classifications are more frequently used as exposure information in environmental epidemiology. The utilization of data from ambient air quality monitoring network sites provides the uni form methodology used for site selection procedures, measurements techniques and qua lity control. Environmenta l exposure was measured at

PAGE 36

21 fixed geographical location continuous m onitoring stations. Comparable standard measurements become widely available fr om many different areas towns and cities, which makes it possible to develop predicti ons for other populations based on available environmental exposure data. A population-based longitudinal ecological study was conducted to evaluate the association between local environmental tri ggers of asthma and hospital admissions for childhood and adult asthma. The study area covered Hillsborough County in Florida. Hillsborough County is located in the west coast of Florida and has 1,048 square miles of land and 24 square miles of inland water area for a total of 1,072 square miles. Incorporated major cities are Tampa, Te mple Terrace, and Plant City. The total population within the se lected study area was 1,009,855 residents Study data was collected and analyzed during th e selected period of study 1997-1999. The postal zip code areas of residence were used to calcula te adjusted (standardi zed) asthma hospital admission rates by age, gender and race and to evaluate spatial dist ribution of asthma hospitalization rates within the county. The zip code areas were coded by five-digit number according to standard definitions and codes developed by the U.S. Bureau of the Census. Each single hospital admission record was linked to the patient’s postal zip code area of residence. There were a total number of 44 zip code areas used where the number of annual hospital admission varied from 2 hos pitalizations in the postal zip code area 33572 to 118 hospital admissions for asthma in the postal zip code area 33610 in Hillsborough County, FL, accordingly. The ambient air pollution data were obt ained from the Aerometric Information System (AIRS) database maintained by th e US Environmental Protection Agency (US EPA). The environmental exposure to such crit eria ambient air pollu tants as respirable particulate matters (PM10), sulfur dioxide (SO2), and ozone (03) in Hillsborough County, FL, in 1999, was evaluated from the data set of the AirData air quality monitoring program coordinated by the Division of Ai r Resources Management of the Florida Department of Environmental Protection as part of the US EPA AIRS database ( www.dep.state.fl.us/Air/pub lications/techrpt/amr.htm ).58 The data on ambient temperature was retrieved from the daily data base of the Tampa International Airport.

PAGE 37

22 The changes in average weekly total tree, w eeds and grass pollen c ounts over the total area of study were also explored duri ng the period of study 1997-1999. Monitoring of outdoor pollen counts during the study period was conducted by the Division of Allergy and Immunology, University of South Florid a College of Medicine, Tampa, FL. The monitoring site conducting c ontinuous pollen count measurements in the study was located at the James A. Haley VA Hospital, Tampa, FL. The ambient air quality mon itoring program measures pollu tant concentrations in the ambient air defined as the portion of th e atmosphere near ground level and external to buildings or other structur es. Ambient air quality standa rds have been established by the U.S. Environmental Protection Agency (EPA) and the Florida Department of Environmental Protection (DEP) for six crite ria pollutants: particulate matters (10 microns or less in diameter, PM10, and 2.5 microns or less in diameter, PM2.5), sulfur dioxide (SO2), nitrogen dioxide (NO2), carbon monoxide (CO), and lead (Pb). Since health-based criteria have been used to establish the standards, these six pollutants are referred to as criteria ambient air pollutan ts. Two types of nationa l ambient air quality standards (NAAQS) have been established by the EPA for the six criteria air pollutants. Primary ambient air quality standards are desi gned to protect public health with an adequate margin of safety. Secondary ambi ent air quality standards are designed to protect public welfare-related values, includ ing property, construction materials, plant and animal life. In Florida, state ambient air quality standards have been adopted at least as stringent as the national s econdary standards. Federal a nd Florida ambient air quality standards for criteria air pollu tants are presented in Table 2. Ambient air quality monitoring network c onsisted of 193 monito ring stations in 34 counties throughout the state of Florida. Two types of monitoring networks are used to collect the ambient air data. State/Local Air Monitoring Station (SLAMS) and National Air Monitoring Station (NAMS) network are de signed to meet a minimum of four basic objectives:58 (1) To determine the highest concentrations expected to occur in the area covered by the network; (2) To determine representative concentra tions in areas of high population density;

PAGE 38

23 (3) To determine the impact on ambient pollu tion levels of significant sources or source categories; (4) To determine general bac kground concentration levels. Table 2. State of Florida and federal nationa l ambient air quality standards (NAAQS)58 Air Pollutant Averaging Time Primary NAAQS Secondary NAAQS Florida Standard Carbon Monoxide 8-hour 1-hour 9 ppm 35 ppm 9 ppm 35 ppm Lead Quarterly 1.5 g/m3 1.5 g/m3 1.5 g/m3 Nitrogen Dioxide Annual 100 g/m3 (53 ppb) 100 g/m3 (53 ppb) 100 g/m3 (53 ppb) Ozone 8-hour 1-hour ** 0.08 ppm 0.12 ppm 0.08 ppm 0.12 ppm 0.08 ppm 0.12 ppm Particulate Matter (PM10) Annual 24-hour 50 g/m3 150 g/m3 50 g/m3 150 g/m3 Particulate Matter (PM2.5) Annual 24-hour 15.0 g/m3 65 g/m3 15.0 g/m3 65 g/m3 Sulfur Dioxide Annual 24-hour 3-hour 80 g/m3 (30 ppb) 365 g/m3 (145 ppb) 1300 g/m3 (500 ppb) 60 g/m3 (20 ppb) 260 g/m3 (100 ppb) not to be exceeded by the three-year average of the 4th highest daily maximum; ** not to be exceeded on more than an average of one day per year over a three-year period. An essential component of air quality mana gement in the state is the identification of (1) areas where the ambient air quality standards are being violated and plans are needed to reduce pollutant con centration levels to be in at tainment with the standards, and (2) areas where the ambient standards are being met but plans are needed to ensure maintenance of acceptable levels of air quality in the face of anticipated population or industrial growth.58 The end-result of this attainment/maintenance analysis is the development of local and stat ewide strategies for controll ing emissions of criteria air pollutants from stationary and mobile sources. Th e first step in this process is the annual compilation of the ambient air monitoring results. The second step is the analysis of the monitoring data for general air quality and pollutant trends. Data from the SLAM/NAMS network provide an overall view of the state’s air quality and are used in the development of statewide control strategies. The Spatia l Purpose Monitoring (SPM) network, which is distinct from the SLAMS/NAMS network, me ets the local and sometimes temporary monitoring objectives, and supplements the SL AM/NAMS network in data-sparse areas. Particulate matters with a diam eter of 10 microns and less (PM10), particulate matters

PAGE 39

24 with a diameter of 2.5 microns and less (PM2.5), sulfur dioxide (SO2), nitrogen dioxide (NO2), ozone (O3), carbon monoxide (CO), and lead (P b) are main criteria pollutants measured on continuous basis by SLAM/ NAMS and SPM monitoring networks in Hillsborough County, FL. Particulate matters (PM10), sulfur dioxide (SO2), and ozone (O3) were used to define and evaluate envir onmental exposure to ambient air pollution in the study area. Because of limited number of air pollution monitoring stations for NO2 and PM2.5 concentration measurements over the study area, environmental exposure to ambient nitrogen dioxide and fine particles were not included in the study. There were only 2 continuous ambient air pollution stat ions for nitrogen dioxide monitoring in Hillsborough County, FL, in 1999. Only 2 ambient air quality monitoring stations for PM2.5 concentration measurements were located in Hillsborough County, FL. Due to limited number of monitoring sites and pollution concentr ation measurements for nitrogen dioxide and particulate matters w ith a diameter of 2.5 microns and less (PM2.5) over the study area, the spatial interpolation re sults and predicted values of environmental exposure to these pollutants by zip code area of residence could be biased and inaccurate. Environmental exposure to PM2.5 as a fraction of total resp irable particles could be indicated by estimated PM10 concentrations in the geographical area of residence over the study area. Various standard referen ce concentration measuremen t instruments and methods developed by the US EPA were used for the determination of particulate matter concentration. The high volume air sampler (hivol) is one of the reference methods for determining PM10 concentration. The hi-vol consists of an electric motor, blower, filter holder, enclosure and a size se lective inlet which allows onl y particles of 10 microns or less to flow through and impact on the filte r. Ambient air is pulled through a preweighted quartz filter at a rate of about 1.5 cubic meters per minute for a 24-hour period. The mass of particles captured is the difference in weight of the filter before and after sampling. Dividing the particles weight by th e volume of air that passes through the device gives the average particulate concen tration for the sampling period. The TEOM (tapered element oscillating microbalance) is another continuous instrument used for real time PM10 concentration measurements in th e continuous air pollution monitoring

PAGE 40

25 network in the State of Flor ida. The TEOM has a size selec tive inlet of 10 microns and uses a 3 liter per minute flow. The air pa sses through a heated column to remove moisture, when impacts a dime-sized filter which fits onto a glass yoke. The yoke vibrates at a frequency, whic h decreases as the mass of the particles on the filter increases. This mass is reported hourly. The other instrument being used to monitor continuous PM10 concentration levels is the Beta ga uge. The Beta gauge is equipped with a size selective inlet of 10 micr ons and employs a reeled tape on which the particulate is collected. A radiation so urce sits below the tape and as particles accumulate the amount of radiation able to pass through the tape to the detector atte nuates. The level of attenuation relates to the am ount of particles measured. The National Ambient Air Quality Standard (NAAQS) for PM10 is 50 g/m3 annual arithmetic mean, and 150 g/m3 24-hour average, not to be exceeded more than once per year. The values for PM10 generally are 40-60 percent of the total suspended particulate (TSP) concentrations for the same air mass in the state of Florida.58 The ambient air quality monitoring network for PM10 consisted of 61 monitoring stations in 27 counties in the state of Florida. There were 9 monito ring stations located in Hillsborough County, FL. The average seasonal environmental qua lity changes by cale ndar quarter were evaluated over the study area based on availa ble daily concentration measurements at stationary ambient particles monitoring st ations. Additional part icles concentration measurements representing 1997 and 1998 years we re used to evaluate and determine the effect of environmental quality changes by ambient particles on th e number of hospital admissions over longer period of ti me during the period of 1997-1999. Ozone was sampled by measuring adsorption of ultraviolet light by ozone. By this method, an ultraviolet photometer measures the transmittance of light through a sample of ambient air. The amount of ultraviolet lig ht transmitted is inversely proportional to the concentration of ozone in the air. A ne w National Ambient Air Quality Standard (NAAQS) for ozone based on 8-hour average wa s promulgated and became effective in 1997. Under the new rule, attainment is reached by having a 3-year average of the annual fourth-highest daily maximum 8-hour averag e not exceeding 0.08 ppm. This adjustment to the 8-hour standard represents the transi tion from an exceedance based standard to a

PAGE 41

26 concentration based one. The new standard is designed to provide in creased protection to the public, particularly those predisposed a nd more sensitive to respiratory problems.58 The 1-hour 0.12 ppm standard was used instead of 8-hour average standard to calculate average monthly and quarterly concentrations for environmental exposure to ambient ozone. The 8-hour standard was promulgated in the end of 1997, and the previous 1-hour standard was dropped in the end of 1998. We us ed the 1-hour standard to evaluate the attainment of standard over the total period of study fr om 1997 to 1999. Under the new rule, attainment to the 8-hour standard is satisfied by having 3-year average of the annual four highest daily maximum av erage of less than 0.08 ppm. The ambient air quality m onitoring network for ozone consisted of 45 monitoring stations strategically locate d in major urban areas and ce rtain rural areas chosen for background comparisons in the State of Fl orida. Over the 3-year period from 1997-1999 for Hillsborough and Escambia Counties, FL, showed concentrations above the standard of the average of the annual fourth highest daily 8-hour maximum of 0.08 ppm. Four ozone monitoring stations were located in Hillsborough County, FL. The average monthly environmental quality changes were evaluated for each county area based on available daily concentration measurements. The average seasonal environmental quality changes by calendar quarter we re evaluated based on avai lable daily concentration measurements at stationary ozone monitoring stations. Additional ozone concentration measurements in 1997 and 1998 were used to evaluate and determine the effect of environmental quality changes by ozone on th e number of hospital admissions over the period of 1997-1999. During sulfur dioxide (SO2) continuous monitoring, the sample air passes through a pulsed ultraviolet light chamber, where SO2 molecules receive energy and become “excited” by increased frequency of vibrati on. The energy release provides a measure of the sulfur dioxide concentration in th e ambient air. Compliance with the SO2 standards in Florida is defined as the 3-hour aver ages less than 45 percent of the 1300 g/m3 standard and the 24-hour averages less than 50 percent of the 260 g/m3 standard.58 Sulfur dioxide monitoring network consis ted of 30 ambient air quality monitoring stations in 14 counties in the State of Florid a. There were 7 monitoring stations located in

PAGE 42

27 Hillsborough County, FL. The average seas onal environmental quality changes by calendar quarter were evaluate d based on available daily sulfur dioxide concentration measurements at stationary conti nuous monitoring stations. Additional SO2 concentration measurements in 1997 and 1998 were used to evaluate and determine the effect of environmental quality changes by sulfur di oxide on the number of hospital admissions over the period of 1997-1999. 1.2.3. Data analysis SAS System V8.2 version60 and the CDC EPI-INFO 3.2 version61 statistical data analysis software programs were applied a nd simple descriptive, correlation, and nonlinear regression analysis techniques were us ed for statistical data analysis. Crude association analysis results were also e xplored in the multiple regression analysis. Stepwise backward selection model building pr ocedures were used to estimate possible interaction among various environmental asth ma triggers and individual effects of independent variables after controlling for (adjusting to) other environmental asthma triggers in the log-linear regression model. Separate regression mode ls were designed for asthmatic children and adults. Possible confounding and intera ction effects in the multiple regression equation were evaluated by using standard epidemiology methods.38,63,64,65 Socioeconomic confounding by individual demographic char acteristics was evaluated by comparing crude asthma hospitalization rates with age-, gender-, and race-adjusted hospital admission rates. Several different standard ep idemiological approaches are used to assess the presence of possible confounding and eff ect modification by interaction, which are related primarily to the questions as whether:63,64,65 (1) Confounding variable is related to bot h the exposure and the outcome in the study; (2) Exposure-outcome association seen in the crude analysis have the same direction as and similar magnitude to the associations seen within specified strata of the confounding variable; and

PAGE 43

28 (3) Exposure-outcome association seen in th e crude analysis have the same as and similar magnitude to the a ssociation after controlling (adjusting) for the confounding variable in multiple regression analysis. Multivariate regression analysis (modeling) analytical techniques were used to control for and evaluate po ssible confounding effect by soci oeconomic status, to assess possible effect modification and interacti on, and to summarize the association between asthma hospitalizations and adjusted indepe ndent predictor variable in the multiple regression model. The term ef fect modification means that the effect of the exposure on the outcome differs depending on whether a nother variable (effect modifier) is present.64,65 Poisson non-linear regressi on stepwise model building techniques were used to develop the best-fit log-li near regression models and to predict asthma hospital admissions rates for childhood and adult asth ma by retained significant explanatory variables. Poisson regression model is used mo stly in analysis of counts or rates as the main outcomes of interest, and is particular suitable for studying rare health outcome of interest developed in re latively large populations.66 The simple log-linear regression model with one predictor vari able Xi and counts of hospita l admissions by zip code area of residence as our independent variable Yi could be explained as function: log (E(Yi))=log ti + *Xi, where E-expectation or probability of event Yi to happen, is a regression coefficient; logti is so called offset variable used to account for refe rence population and to use rates instead of counts for asthma hospital admissions. Final regression model deviance estimates we re used as a goodness of fit criterion to evaluate the validity and adequacy of the model. The deviance of a fitted model compares the log-likelihood of the best-fit model to the l og-likelihood of a model with n parameters that fits n observations perfectly. Such a perfectly fitting model is called a saturated model The log-likelihood value for the fitted model cannot be larger than the log-likelihood value for the saturated m odel because the fitted model has fewer

PAGE 44

29 parameters. The smaller the difference in the two log-likelihood values – the smaller is the deviance and the closer is the fitted m odel to the saturated model. The larger the model deviance, the poorer is the fit. Therefor e, large values of the model deviance will indicate that the fitted nonlinear regression model is not co rrect. In addition, the Chisquare value and its fraction with corresponding degrees of freedom (df) were also used to evaluate the adequacy of the model fit. If the best-fit regression model is correct then both model deviance and Pearson chi-square statistics quantities are asymptotically distributed as with n-p degrees of freedom (df), where n is the number of all study units (geographical areas of residence) and p is the number of fitted parameters. Therefore, the regression model is adequate if the expected value of th e model deviance and Pearson Chi-square estimates are equal or close to th e value of degrees of freedom (df) in the particular model. If both the scaled model deviance Deviance/df >>1 and/or the scaled Pearson Chi-square 2/df >>1, there may be serious doubt about the validity and adequacy of the model. The inadequate model suggests that there is a greater variability among counts than would be expect ed for log-linear distribution. The term overdispersion is used to describe such extra-variability in the multivariate log-linear regression model. The most common reason for data being overdispersed is that experimental conditi ons are not perfectly under control and the unknown parameters may vary not only with meas ured covariates but also with latent and uncontrolled factors in the model. The parameter as a ratio of the model deviance or the Pearson Chi-square to its associated degr ees of freedom was proposed to estimate the overdispersion.63,64 The SAS statistical data analysis software introduced options of DSCALE (or SCALE= DEVIANCE) or PSC ALE (or SCALE= PEARSON) in the general linear models statement of PROC GENMOD to control the overdispersion.60 The main outcome was the number of hospital admissions for children and adult populations by zip code area of residen ce. DIST=POISSON or DIST=P indicated that Poisson loglinear regression analysis model was used in the generalized linea r regression models’ option specified by SAS. Wald 95% Confidence Interval values (95% CI) were used to evaluate model parameters. Maximum likelihood estimates were used to predict the effect of socioeconomic status and environmental exposure descriptive variables in the model.

PAGE 45

30 Chi-square statistics and corresponding statis tical significance estimates by p-value were used for the analysis and further in terpretation of model parameters. The data set in regression analysis can contain some cases that are outlying or extreme. Extreme cases may involve large re siduals and may have influential effect on the fitted regression function. Therefore, the evaluation of extreme cases and further decision to retain or eliminate outlying cases is very important.66 The next step is to ascertain the effect of each extreme case a nd to decide whether or not these outlying cases are influential A single case is considered as in fluential if its exclusion from the data set causes major changes in the best-fit regression model. DFBETAS estimates were used to evaluate the influence of each outly ing case on the regre ssion coefficients. The DFBETAS estimate is a value of the significan t influence of the single extreme case i th on each regression coefficient bk (k=0, 1, …, p-1, where pnumber of parameters in the regression analysis) based on the difference be tween the estimated regression coefficient bk representing all cases and the regression coefficient obtained when i th case is omitted ( bk(i)). The DFBETAS estimate is obtained by dividing this difference by the standard deviation of bk:66 MSE b b DFBETASi i k k i k ) ( ) ( ) ( k=0, 1, … p-1 An estimator of the standard deviation is the positive square root of error mean square or residual mean square (MSE).66 p n p n SSE MSEY Ypred i i 2 ) ()( where SSE stands for error sum of squares or residual sum of squares p – number of parameters, n – total number of cases, and Yi(pred) – predicted estimate of dependent variable by the fitted model. The DFBETAS value by its sign indicates whether inclusion of a case leads to an increase or a decrease in the estimated regr ession coefficient. A large absolute value of DFBETAS indicates a large impact of the extrem e case on the regression coefficient. The

PAGE 46

31 influential case was defined as an outlier with the absolute value of DFBETAS more than n 2 .66 Human subject protection was assured by using non-identifia ble coded personal information and by providing adequate security fo r initial datasets and final data analysis results. The study met required federal crite ria to qualify as an exempt study and was approved by the Institutional Review Bo ard, Division of Research Compliance, University of South Florida, on May 28, 2004 (Protocol No. 102536, see Appendix A). 1.3. Results 1.3.1. Hospital Admissions for Childhood and Adult Asthma The State of Florida Agency for Health Administration (AHCA) hospital inpatient discharge dataset included information on such patient demographic characteristics as age, gender and race, primary and secondary diagnosis, surgical procedures performed, hospital of admission, and type of payer or in surer. Preliminary data analysis revealed that a total of 23,958 annual hospital admissions coded by either principal or secondary diagnosis of asthma, were reported by licensed hospitals in the State of Florida in 1999. Analysis of the number of total and childhood hospitalizations for asthma by calendar quarter revealed that more patient s with asthma were admitted in the first quarter (January through March) and the last quarter (October through December). Results are presented in Tables 3 and 4. Table 3. Total number of hospital admissions for asthma and average length of stay (ALOS) in days by calendar quarter in Florida, 1999 Calendar Quarter Asthma Hospita lizations Percent, % ALOS, days January March 7,765 32 4 April June 4,648 19 4 July September 4.717 20 3 October December 6,828 28 3 Total/ Average 23,958 100 4

PAGE 47

32 Table 4. Children 14 years and younger hospital admissions for asthma and average length of stay (ALOS) in days by calendar quarter in Florida, 1999 Calendar Quarter Asthma Hospita lizations Percent, % ALOS, days January March 2,551 28 3 April June 1,754 20 3 July September 1,887 21 2 October December 2,800 31 3 Total/ Average 8,992 100 3 A total of 5,366 hospitalizations for asthma as a principal or secondary diagnosis for all age groups were recorded in H illsborough County, FL, in 1999. St. Joseph, Tampa General, Brandon Medical Center, Univers ity Community, South Florida Baptist, Community Hospital in Carrolwood, Memorial and Town and Country were the main hospitals of admission and accounted for 5142 or 95.8 per cent of total study subjects living in Hillsborough County, FL, in 1999. The number of asthma hospital admissions by county of patient’s permanent residence and principal hospita l of admission is presented in Table B-1 (see Appendix B). Other hospitals which admitted asthma patients during the period of our study included Kindred and South Bay Hospitals, ac counted for only 0.9 per cent of the total number of asthma hospital admissions at all hospitals located in Hillsborough County, FL, in 1999. The number of hospital admissions for asthma as either principal or secondary diagnosis by the reference source is presented in Table B-2 (see Appendix B). Emergency department rooms (ER) and refe rring physicians were the main reference sources for asthma hospital admission. ER referred admissions accounted for 51.5 to 73.7 percent of total asthma hospital admissions in Hillsborough County, FL. The study area covered 44 postal zip code ar eas of residence as main geographical small-area units of data analysis. There were 1,009,855 pe rsons, including 215,497 children of less than 15 year s of age, who were living in Hillsborough County, FL, in 1999. There were a total of 5,366 hospital admissions for asthma as either primary or secondary diagnosis in Hillsborough County, FL, in 1999. The number of hospital admission for asthma as both principal and se condary diagnoses varied in separate zip code areas from 3 as a minimum to 328 as a maximum number of hospitalizations for

PAGE 48

33 asthma. Secondary diagnosis of asthma coul d be biased by another different principal disease or medical condition, and could barely reflect existing chronic but not severe clinical condition which woul d require urgent clinical treatment. Only hospital admissions for childhood and adult asthma as a primary principal diagnosis were used to calculate crude childhood and adult asthma hos pital admission rates and used for further stratified and multiple regression analysis. The total number of hospital admissions for asthma as a principal diagnosis of hosp italization was 1712 persons in 1999. There was significant positive correlation of hospital admissions for asthma as a primary diagnosis with hospitalizations for asthma as either primary or any s econdary diagnoses of asthma by zip code areas of residence in Hills borough County, FL, in 1999. (Pearson correlation coefficient r=0.98 with corresponding p-va lue p<0.0001). The distribution of hospital admissions for asthma as principal diagnosis and hospitalizations for asthma as principal or secondary diagnoses by sepa rate zip code area of resi dence in 1999 is presented in Table B-3 (see Appendix B). The total number of hospital admissions for asthma as a primary diagnosis by zip code area varied within the range of 116 from 2 to 118 hospitalizations by separate zip code area in Hillsborough County, FL, in 1999. The average mean value of number of hospital admissions for asthma by zip code area of residence was 38.9 with a corresponding standa rd error value of SE=2.33. The average mean value of crude hospital admission rates by geographical area of residence was 17.14 per 10,000 of total population hospitalizations for asthma with a relevant standard error of SE=1.65 in 1999. The separate descriptiv e analysis results for asthma hospital admissions and crude rates of hospitalizations fo r asthma by zip code area of residence in 1997 and 1998 are provided in Table B-4 (s ee Appendix B). There were 1649 and 1381 total hospital admissions for asthma in 1997 and 1998 respectively. Calculated hospital admission rates vary from zero to 50.34 in 1997, and from 1.23 to 39.8 in 1998 accordingly. The average mean values of hospital admission rates over the overall study area were 15.9 (SE=1.67) and 13.3 (SE=1.28) per 10,000 of total population in 1997 and 1998 respectively. A comparison of adjusted (standardized) and crude asthma hospital admission rates could provide valuable info rmation on how our study sample population differs from a

PAGE 49

34 representative standard populat ion, and whether existing differe nces specific to our study population could bias interpreta tion of final results and appl ication to other geographical areas or population groups of interest outside our study area. The US Census 2000 population data was used to represent the standard population dist ribution by different demographic characteristics and to calculate relevant standard weights within separate categories (strata) by age, race and gender gr oups (see Appendix B). The total population in the given postal zip code ar ea of residence was used as a standard reference group and corresponding denominator to cal culate crude hospita lizations rates for asthma. The US Standard 1,000,000 population distribution by ag e, gender and race, was calculated and used as a standard reference to calculate standardized hospital admission rates by age, race and gender. Because of significant differe nce between children and adult crude rates of hospitalization for asthma, separate crude rates for children less than 15 years and adults of 15 years of age and older per 10,000 of children or adults population respectively were calculated by zip code area of residence. Crude rates of children and adults asthma hospital admissions in 1999 are presented in Table B5 (see Appendix B). Calculated crude rates of hospitalizations fo r asthma as a principal diagnosis varied by separate zip code area from 2.28 to 56.95 hospital admissions per 10,000 population. Descriptive statistical analys is revealed that the averag e mean value was 17.14 hospital admissions per 10,000 with a correspondi ng standard error value SE=1.65. Hospitalizations of children younger than 15 years of age varied by separate zip code area from zero to 133.9 with an average mean value of 35.62 asthma hospitalizations per 10,000 children population and a standard erro r value of SE=4.11 (see Picture 1). There were no pediatric asthma hospital ad missions from the zip code residential area 33573, where the total children populati on was only 56 in 1999. In addition, the average mean value of the cr ude rate of adult asthma hospital admissions was 12.03 per 10,000 of adult population with a standard error value of SE=1.05, and the corresponding crude rate estimates varied by zip code ar eas over the total study area from zero to 36.7 adult asthma hospital admissions per 10,000 of 15 year and older population in 1999 (see Picture 2).

PAGE 50

35 Picture 1. Crude childhood asthma hospitalization rates per 10,000 by zip code area of residence in Hillsborough County, FL, in 1999 There was a very strong association be tween childhood and adu lt asthma hospital admission rates by zip code area of reside nce (correlation coe fficient r=0.81, p<0.0000). Areas with relatively higher adult asthma ra tes had also higher hospitalization rates for childhood asthma. Both childhood and adult asth ma rates were also highly correlated with total rates by single zip code area of residence (r=0.94, p<0.0000 for both childhood and adult crude asthma hospital admission rates correspondingly). The standard direct adjustment techniques were used to calculate standardized hospitalization rates by pers onal demographic characteri stics. The US 2000 population was used as a reference populat ion to obtain the standardized weights within different categories (strata) of individua l socio-demographic characteris tics. Direct standardization analysis is presented in Appendix B. Because of a limited number or no asthma hospital admissions available within a separate age st ratum, 5 zip code area s of residence (33547, 33572, 33573, 33621, and 33647) were excluded from the analysis and calculation of

PAGE 51

36 age-adjusted rates. The distribution of cr ude and age-adjusted asthma hospitalization rates per 10,000 of population by zip code ar ea in Hillsborough County, FL, in 1999, is presented in Table B-6 (see Appendix B). Picture 2. Crude adult asthma hospitalization rates per 10,000 by zip code area of residence in Hillsborough County, FL, in 1999 The relative ratio of age-ad justed and crude rates for asthma hospital admissions by zip code area of residence was calculated to evaluate the difference between crude and adjusted rates by age by postal zip code area of residence. The average relative ratios over the total study area was 1.03 and varied s lightly from 0.91 to 1.29 by separate zip code area of residence, providing the evid ence that there is no significant difference between age-adjusted and crude hospital ad mission rates within the total area of study (p>0.05). The Pearson correlation coefficien t also suggested a strong statistically significant association between crude and adjusted hospital admission rates by age (correlation coefficient r=0.99, p<0.0001). Howeve r, calculated separate crude asthma hospital admission rates for children (less than 15 years of age) and adults (of 15 years

PAGE 52

37 and older) revealed an existing difference in corresponding rates within these two age groups. Genderand race-adjusted rates were ca lculated for all geog raphical zip code areas of residence in Hillsborough County, FL in 1999, and are represented in Table B-7 (see Appendix B). Gender was defined into male and fe male groups, and corresponding standard population weights were used to adjust for gender differences within each group. Genderadjusted rates varied by different zip code over the study area from 2.29 in the zip code area 33647 to 56.9 in the zip code area 33602, with an average mean value of 17.15 (SE=1.65) hospital admissions for asthma per 10,000 population over the total study area. The Pearson correlation analysis revealed an exact statistical association between crude and gender-adjusted rates for hospital admi ssions (correlation coefficient r=1 (0.999) with corresponding p-value p<0.000). The rela tive ratio values between gender-adjusted and crude rates varied by zip code areas from 0.98 to 1.02, with an average mean value of 1 and a corresponding standard error value of SE=0.001. Race was divided into four main groups: white (Caucasian), black (African American), Hi spanic (white and black Hispanic), and Others. Only 2 Black Hisp anics were referred for hospital admission because of exacerbation of asthma-related resp iratory symptoms during the year of 1999. Both patients were included in the Hispanic group. The Other’s cat egory included Native American, Asian, multiracial, and non-iden tifiable ethnic groups. There was only 1 Native American patient, living in the zip code are of 33610, admitted to the hospital for asthma diagnosis as a principal diagnosis in 1999. There were 7 Asian people admitted for asthma: 1 patient for each of 33510, 33604, 33613, 33614, 33634, and 2 patients in 33611 zip code areas respectivel y. Adjusted rates for asthma hospital admissions by race varied over the total stud y area from 7.2 to 56.3, with an average mean value of 19.01 hospital admissions per 10,000 population (SE= 1.64). The Pearson correlation analysis by zip code areas revealed a slightly lowe r but significant association between raceadjusted and crude hospitalization rates per 10,000 population (correlation coefficient r=0.89, with p<0.001). However, the analysis of the rate ratios betw een adjusted by race and crude rates of hospital admissions for as thma revealed much wider variation from 0.68 to 8.05. The small-area socioeconomic depriv ation status by racial differences was

PAGE 53

38 represented by using the percen tage of ethnic minority househol ders as a social status characteristic of a given geogra phical area of residence. Preliminary descriptive data analysis reveal ed differences in asthma hospitalizations over time and by geographical area of resi dence. The associat ion between hospital admissions for childhood and adult asthma and bo th local environmental triggers and area socioeconomic status indicators, while contro lling for environmental exposure to selected criteria ambient air pollutants over the area of residence, was evaluated to ascertain which environmental and socioeconomic factors co uld explain and predict asthma hospital admission rates. 1.3.2. Environmental Asthma Triggers Ambient air pollution by particulate matter (PM10), sulfur dioxide (SO2), ozone (O3), ambient air temperature, as well as to tal grass, total tree, and total weed pollen counts were used to estimate the possible eff ect of aforementioned environmental triggers on the exacerbation of asthma-related respirat ory symptoms which subsequently resulted in hospital admission for asthma. The study area covered geographical area of Hillsborough County, FL. Average calendar quart er estimates of environmental exposure to possible environmental asthma triggers were calculated to evaluate the association between environmental exposure and number of monthly asthma hospital admissions over time. To explore possible association be tween environmental exposure to asthma triggers and hospital admissions for chil dhood and adult asthma over time, the study duration was extended and included data for the period of 1997, 1998, and 1999. The calendar quarters divided each year of study in to four different periods of time: (1) January-March; (2) April-June; (3) July-September; and (4) October-December. The average monthly or seasonal ambient air pollu tion by selected criteria pollutants has not exceeded the standard guidelines; however, hourly and daily variations in ambient air pollutant concentrations were more extreme with monitored concen trations very often exceeding the air quality standards. In additi on, the single peak concentration values and total number of peaks above the national am bient air quality standards (NAAQS) were

PAGE 54

39 identified for selected criteria ambient air pollutants. The number of daily peak concentrations per calendar quarter during the extended period of study was compared with the number of hospital admissions for ch ildhood and adult asthma over the total area of study to evaluate the association betw een extreme peak ambient air pollution concentrations and exacerbations for asthma Air pollution Daily ambient air pollution by particulate matters, oz one, and sulfur dioxide concentration measurements by separate ambien t air quality monitoring site were used to calculate corresponding average monthly and calendar quarter concentration values over the total study area. The air quality data was compared with the average number of hospital admissions for asthma as a princi pal diagnosis during the study period of 1997 through 1999 in Hillsborough County, FL. Table 5. Air Quality Monitoring Network in Hillsborough County, FL, 1997-1999 Site Site Name Site Location Latitude Longitude Pollutant 0085 Eisenhower 7620 Big Bend Rd. 27N 47' 28" 82W 22' 06" PM10 1069 Harbor Island 900 Harbor Island Blvd. 27N 56' 03" 82W 27' 07" PM10 1068 Gaither 4013 Ragg Rd. 28N 06' 04" 82W 30' 15" PM10 1002 Health Dept. 1105 E. Kennedy Blvd. 27N 56' 48" 82W 27' 06" PM10 0109 East Bay 9851 Hwy 41 S. 27N 51' 16" 82W 23' 01" SO2 0081 Simmons Park 2401 19th Ave NW 27N 44' 34" 82W 28' 09" SO2, O3 0095 Causeway 5012 Causeway Blvd. 27N 55' 16" 82W 24' 05" PM10, SO2 2002 Brandon 2929 S. Kingsway Ave. 27N 58' 01" 82W 16' 43" PM10 0083 Gardinier US 41 27N 51' 44" 82W 22' 57" PM10 0053 Ballast Point 5200 Interbay Blvd. 27N 53' 07" 82W 28' 54" SO2 0030 Palma Ceia 3910 Morrison Ave. 27N 55' 50" 82W 30' 35" PM10 1035 Davis Island Davis Blvd S. 27N 54' 33" 82W 27' 19" SO2, O3 1065 Gandy Bridge 5121 Gandy Blvd. 27N 53' 33" 82W 32' 15" O3 0066 Ruskin Hwy 41 N 27N 53' 35" 82W 24' 07" PM10 4004 Plant City One Raider Place 27N 59' 27" 82W 07' 33" SO2, O3

PAGE 55

40 The description of air quality monitori ng network including the number, name, geographical location of each mon itoring site, and type of pollutant measured is given in Table 5. Association between average seasona l environmental exposure to ambient air pollution by particles, sulfur dioxide and ozone was evaluated by calendar quarter over the period of study from 1997 through 1999. Analys is of daily peak concentrations was based on continuous hourly measurements at stationary ambient air quality monitoring stations located within our study ar ea during the period of 1997 to 1999. Table 6. Number of peak concentrations above air quality standards for SO2, PM10 and O3, and a number of childhood and adult asth ma hospital admissions by calendar month and quarter, in Hillsborough County, FL, 1997-1999 1997 1/97 2/97 3/97 Quarter I 4/97 5/97 5/97 Quarter II SO2 11 6 5 22 4 4 7 15 PM10 1 0 0 1 0 0 1 1 O3 0 0 0 0 0 0 0 0 Children Asthma 207 135 Adult Asthma 276 182 Total Asthma 483 317 1998 1/98 2/98 3/98 Quarter I 4/98 5/98 6/98 Quarter II SO2 1 8 7 16 6 4 4 14 PM10 2 1 0 3 0 1 0 1 O3 0 0 0 0 0 5 2 7 Childhood Asthma 160 108 Adult Asthma 286 190 Total Asthma 446 298 1999 1/99 2/99 3/99 Quarter I 4/99 5/99 6/99 Quarter II SO2 4 4 4 16 10 6 2 18 PM10 0 0 0 0 0 1 0 1 O3 0 0 0 0 0 0 0 0 Childhood Asthma 235 153 Adult Asthma 354 158 Total Asthma 589 311 A daily peak measurement was defined and counted as a concentration equal to or more than 100 ppb for sulfur dioxide, and equal to or more than 150 g/m3 or more for

PAGE 56

41 ambient particles based on established 24-hr state ambient air quality standards for the aforementioned pollutants. Ozone peak con centrations were estimated based on hourly peak values equal to or exceeding establis hed 1-hour state standards (0.12 ppm). Table 6. Number of peak concentrations above air quality standards for SO2, PM10 and O3, and a number of childhood and adult asth ma hospital admissions by calendar month and quarter, in Hillsborough County, FL, 1997-1999 (cont.) 1997 7/97 8/97 9/97 Quarter III 10/97 11/97 12/97 Quarter IV Total SO2 4 5 0 9 3 4 7 14 50 PM10 0 2 0 2 0 0 0 0 4 O3 0 0 0 0 0 0 0 0 0 Childhood Asthma 170 262 774 Adult Asthma 162 283 903 Total Asthma 332 545 1677 1998 7/98 8/98 9/98 Quarter III 10/98 11/98 12/98 Quarter IV Total SO2 13 6 4 23 5 6 5 16 69 PM10 0 0 0 0 0 1 0 1 5 O3 0 0 0 0 0 0 0 0 7 Childhood Asthma 91 190 549 Adult Asthma 165 216 857 Total Asthma 256 406 1406 1999 7/99 8/99 9/99 Quarter III 10/99 11/99 12/99 Quarter IV Total SO2 6 5 6 17 2 7 3 12 63 PM10 0 0 0 0 0 0 0 0 1 O3 0 1 0 1 0 0 0 0 1 Childhood Asthma 163 230 781 Adult Asthma 163 256 931 Total Asthma 326 486 1712 Peak concentrations registered by differe nt air quality monito ring station at the same time were counted as one peak over th e study area. Some days could have more than one peak concentration by a specific pollu tant, however, there should be at least one hour difference or more between two peak c oncentrations at the same or different monitoring stations to count measurements as two different peaks. Continuous peak concentrations lasting for a few hours were also counted as one peak. Ambient air pollution by particulate matters and ozone was below air quality standards. Only

PAGE 57

42 environmental exposure to sulfur dioxide has shown numerous peak concentrations above established ambient quality standards. Sulfur dioxide has also had more noticeable variation of daily peak con centration values compared to ambient particles or ozone (Table 6). Even if one sets a lower peak c oncentration cut-off point value of higher than 50 g/m3, it did not show higher variation and di d not reveal more peak measurements for ambient particles. Because of a limited number of peak concentrations for ambient particles and ozone, only the number of daily sulfur dioxide peak concentrations was used to explore possible association betw een number of peak concentrations and corresponding increase in hospital admissions for asthma by calendar quarter over the period of study 1997-1999. The number of peak concentrations of particulate matter, ozone, and sulfur dioxide ambient concentra tions exceeding ambient air quality standards is presented above in Table 6. Particulate matter The ambient particles monitoring network consisted of 9 continuous ambient air quality monitoring stations in Hillsborough County, FL, (Site 30, Site 66, Site 83, Site 85, Site 95, Site 1002, Site 1068, Site 1069 and Site 2002). Calculated average monthly and annual coarse particulat e matter concentrations by se parate ambient air quality monitoring site and by single year of 1997, 1998, and 1999 are presented below in Tables 7-9. The highest average annual concentrati ons were indicated in Site 66 (36 g/m3 in 1997, 33 g/m3 in 1998 and 34.9 g/m3 in 1999) and in Site 1069 (1997 – 28 g/m3, 1998 – 30.4 g/m3 and 1999 – 28 g/m3) in Hillsborough County, FL. The lowest annual values were seen in Site 1068 (1997 – 21.5 g/m3, 1998 – 21.6 g/m3 and 1999 – 19 g/m3) and Site 2002 (1997 – 22.7 g/m3, 1998 – 23.6 g/m3 and 1999 – 22 g/m3) correspondingly. In 1997, the aver age annual value was 26 g/m3 varying from 24.8 in January to 27.6 g/m3 in May in Hillsborough County, FL. The average yearly values were 26.2 g/m3 and 26.1 g/m3 in 1998 and 1999, respectively. There were relatively higher concentrations of ambient particle s in April-June and July-September in 1997 and 1998 respectively, and the highest concentrations were indicated in January-March calenda r quarter in 1999. (see Table 7-9)

PAGE 58

43 Table 7. Average monthly and quarterly particulate matter (PM10) concentrations, g/m3, by separate ambient air quality moni toring site, Hillsborough County, FL, 1997 1997 Site 30 Site 66 Site 83 Site 85 Site 95 Site 1002 Site 1068 January 26.2 35.8 23.6 17.5 31.0 26.4 18.0 February 26.6 35.8 20.3 17.0 29.8 29.0 21.6 March 26.2 38.0 26.0 23.0 30.2 26.6 21.2 April 27.8 35.6 23.2 22.8 27.8 28.0 24.0 May 24.3 45.0 25.0 22.4 29.2 25.2 23.8 June 32.6 39.2 28.6 24.4 27.7 34.2 28.2 July 30.5 31.4 28.0 24.3 32.0 32.2 24.2 August 25.4 35.2 23.2 20.2 26.3 27.0 19.4 September 24.4 34.4 19.4 19.6 22.4 23.0 18.4 October 27.0 39.3 23.2 23.0 26.8 27.5 21.2 November 20.8 39.4 22.0 19.4 27.6 32.3 19.8 December 20.2 24.0 20.4 17.0 23.2 22.4 19.0 Average 26.0 36.1 23.6 20.9 27.8 27.8 21.6 Table 7. Average monthly and quarterly PM10, g/m3, by separate ambient air quality monitoring site, Hillsbor ough County, FL, 1997 (cont.) 1997 Site 1069 Site 2002 Average Concentration Average Concentration January 25.0 20.6 24.9 25.8 February 24.3 23.6 25.3 March 26.8 27.4 27.3 April 27.6 23.0 26.6 28.3 May 29.6 23.8 27.6 June 35.2 25.8 30.7 July 32.6 24.6 28.9 25.7 August 29.0 22.8 25.4 September 25.2 19.5 22.9 October 29.5 22.8 26.7 24.3 November 28.8 20.3 25.6 December 22.2 18.0 20.7 Average 28.0 22.7 26.1 26.1 Table 8. Average monthly and quarterly PM10, g/m3, by separate ambient air quality monitoring site in H illsborough County, FL, 1998 1998 Site 30 Site 66 Site 83 Site 85 Site 95 Site 1002 Site 1068 January 20.2 23.4 19.0 18.5 21.8 22.0 17.2 February 26.4 29.2 24.0 21.6 28.4 28.0 20.2 March 27.0 42.0 24.0 22.6 29.5 30.2 21.6 April 29.4 32.8 26.8 24.4 32.6 29.0 22.8 May 35.8 33.0 31.8 29.2 33.8 34.6 29.4 June 30.6 32.8 31.5 24.8 35.8 32.2 23.0 July 35.4 37.6 38.8 34.2 42.4 38.8 32.0 August 27.7 25.0 22.8 20.0 25.2 26.6 17.2 September 23.7 28.3 15.7 18.3 23.5 25.0 19.0 October 20.4 50.8 20.3 19.8 25.2 27.0 17.3 November 32.8 32.0 28.0 27.0 31.0 31.3 22.0 December 22.8 31.2 21.3 21.5 27.0 28.3 18.0 Average 27.7 33.2 25.3 23.5 29.7 29.4 21.6

PAGE 59

44 Table 8. Average monthly and quarterly PM10, g/m3, by separate ambient air quality monitoring site, Hillsbor ough County, FL, 1998 (cont.) 1998 Site 1069 Site 2002 Average Monthly Average Quarterly January 22.2 17.6 20.2 24.6 February 30.5 22.2 25.6 March 32.0 23.0 28.0 April 31.6 25.6 28.3 30.5 May 33.4 33.8 32.8 June 34.6 28.6 30.4 July 39.2 29.8 36.5 27.4 August 26.3 24.4 23.9 September 25.3 17.8 21.8 October 30.4 21.0 25.8 26.1 November 30.6 21.0 28.4 December 29.2 18.7 24.2 Average 30.4 23.6 27.2 27.2 Table 9. Average monthly and quarterly PM10, g/m3, by separate ambient air quality monitoring site, Hillsborough County, FL, 1999 1999 Site 30 Site 66 Site 83 Site 85 Site 95 Site 1002 Site 1068 January 22.0 31.0 22.4 19.0 25.4 25.4 19.0 February 27.3 42.8 26.3 27.5 34.0 35.3 23.8 March 27.5 46.2 30.0 26.0 35.2 33.5 24.2 April 25.2 40.4 27.2 23.0 28.4 27.2 14.4 May 21.4 28.8 25.6 20.2 27.6 22.4 19.2 June 24.4 27.2 22.4 21.0 27.4 25.6 21.0 July 31.4 31.2 33.6 24.6 32.6 33.4 32.0 August 20.3 24.4 20.6 14.8 22.0 21.2 16.2 September 20.4 34.8 23.2 18.6 25.6 23.4 17.5 October 17.8 37.8 15.2 13.6 16.6 19.4 13.0 November 18.8 33.0 18.6 15.4 21.6 21.0 14.4 December 24.4 41.2 20.2 17.6 22.4 27.4 17.4 Average 23.6 34.9 23.8 20.1 26.6 26.3 19.5 Table 9 Average monthly and quarterly PM10, g/m3, by separate ambient air quality monitoring site, Hillsbor ough County, FL, 1999 (cont.) 1999 Site 1069 Site 2002 Average Monthly Average Quarterly January 27.5 22.3 23.8 28.8 February 34.3 26.7 30.9 March 34.0 27.8 31.6 April 29.2 26.4 26.8 24.8 May 25.4 20.2 23.4 June 28.4 20.0 24.2 July 33.6 29.8 31.4 24.8 August 23.0 17.2 20.0 September 24.8 20.2 23.2 October 25.0 14.6 19.2 21.3 November 21.6 17.6 20.2 December 28.6 20.6 24.4 Average 27.9 21.9 25.0 25.0

PAGE 60

45 The graphical distribution of asthma hos pital admission and average air pollution by ambient particles by calendar quarter during the period of 1997-1999 in Hillsborough County, FL, is illustrated in Figure 1. Figure 1. The total number of hospital admission s for childhood and adult asthma and average ambient partic le concentration, g/m3, in Hillsborough County, FL, by calendar quarter during the period of 1997-1999 0 50 100 150 200 250 300 350 400 Jan-Mar 97 Apr-Jun 97 Jul-Sep 97 Oct-Dec 97 Jan-Mar 98 Apr-Jun 98 Jul-Sep 98 Oct-Dec 98 Jan-Mar 99 Apr-Jun 99 Jul-Sep 99 Oct-Dec 99Calendar QuarterNumber of asthma hospitalizatio n 0.0 5.0 10.0 15.0 20.0 25.0 30.0 35.0Conc., microgr/m 3 Childhood asthma Adult asthma PM10, g/m3 Ambient air pollution by particulate matter a nd other environmental asthma triggers regression analysis results are presented in Appendix D. Simple log-linear regression analysis concluded that there was no signi ficant association between air pollution by ambient particles and hospital admissions for total (p>0.1), adult (p>0.1) and children (p>0.05) asthma. Ozone (O 3 ) The 1-hour NAAQS standard was 0.12 ppm. Although average annual ozone concentrations were very similar in differe nt continuous ambient air quality monitoring sites, the monthly and seasonal variations were more prominent in the selected study

PAGE 61

46 area. The highest peak values were observed in Spring (April-May), while lowest ozone concentration was at the end of Fall and in Winter (December). The number of daily peaks represented the number of extreme concentrations above 0.12 ppm. However, the established federal standard was exceed ed only once (August 1999) during the overall study period of 1997 through 1999. Ozone monito ring station networ k included Sites 81, 1035, 1065 and 4004 in Hillsborough County, FL. Site 4004 started to operate in April 1998. (Table 11) The average monthly values during th e overall study period of 1997-1999 varied from 19.5 ppb in December to 41 ppb in April in Hillsborough County, FL. The distribution of average monthl y and quarterly ozone concentr ations by different year is given in Tables 10-12 below. Site 4004 was started to operate since 1998 and, therefore, obtained ambient air quality data repres ent the time period of 1998-1999 only. Table 10. Average monthly and quarterly oz one concentration, ppb, by separate monitoring site, in Hill sborough County, FL, in 1997 1997 Site 81 Site 1035 Site 1065 Site 4004 Average Monthly Average Quarterly January 26.4 21.4 23.5 NA 23.8 28.6 February 30.1 24.5 26.4 NA 27.0 March 38.2 32.1 34.8 NA 35.0 April 44.2 39.2 41.2 NA 41.5 36.2 May 38.4 35.5 39.5 NA 37.8 June 28.9 28.4 30.9 NA 29.4 July 30.9 27.6 31.8 NA 30.1 31.1 August 34.3 31.5 36.2 NA 34.0 September 29.8 27.1 30.9 NA 29.3 October 33.4 26.9 32.2 NA 30.8 26.1 November 27.3 22.1 27.3 NA 25.6 December 24.6 18.3 22.7 NA 21.9 Average 32.2 27.9 31.4 NA 30.5 30.5 The highest annual averages were observ ed in Sites 81 and 1065 in Hillsborough County, FL. The lowest relevant average valu es were shown in Site 4004 in Hillsborough County, FL. In Site 81, average yearly valu es varied from 30.1 ppb in 1998 to 32.7 ppb in 1999, while average monthly concentrati ons varies from minimum 21.06 ppb in September 1998 to maximum 45.6 ppb in May 1998. In 4004, average yearly value was 26.6 ppb and average monthly concentrations varied from 14.1 ppb in December 1998 to 40.5 ppb in May 1998.

PAGE 62

47 Table 11. Average monthly and quarterly oz one concentration, ppb, by separate monitoring site, in Hill sborough County, FL, in 1998 1998 Site 81 Site 1035 Site 1065 Site 4004 Average Monthly Average Quarterly January 27.8 21.7 26.3 NA 25.3 32.7 February 37.6 30.7 36.4 NA 34.9 March 39.7 34.7 39.4 NA 37.9 April 43.8 42.4 45.0 39.7 42.7 39.7 May 45.5 42.5 46.9 40.5 43.9 June 33.1 30.6 33.4 33.2 32.6 July 29.8 25.6 29.8 25.5 27.7 23.6 August 27.5 22.1 26.9 23.6 25.0 September 21.1 16.8 18.3 16.4 18.2 October 30.5 25.2 30.8 25.9 28.1 23.0 November 27.1 21.4 25.8 20.4 23.7 December 21.3 15.0 18.8 14.1 17.3 Average 32.1 27.4 31.5 26.6 29.4 29.4 Table 12. Average monthly and quarterly oz one concentration, ppb, by separate monitoring site, in Hill sborough County, FL, in 1999 1999 Site 81 Site 1035 Site 1065 Site 4004 Average Monthly Average Quarterly January 17.4 19.6 19.3 18.8 30.5 February 33.7 29.0 32.6 29.7 31.3 March 44.7 38.4 43.1 39.3 41.4 April 40.8 36.8 39.8 37.9 38.8 33.0 May 38.3 34.6 37.7 32.8 35.9 June 27.1 22.8 24.4 22.5 24.2 July 28.0 25.2 27.8 22.2 25.8 27.6 August 29.6 24.5 27.1 25.3 26.6 September 34.0 28.5 32.5 26.0 30.3 October 28.6 23.9 26.7 21.6 25.2 24.3 November 31.1 25.4 30.0 24.1 27.7 December 23.6 17.5 20.5 18.0 19.9 Average 32.7 27.0 30.2 26.6 29.1 29.1 The average exposure to ozone and ho spital admissions for childhood and adult asthma by calendar quarter during the peri od of 1997-1999 is presented in Figure 2. Poisson log-linear regression an alysis results are provided in Appendix D. There was no significant association between environm ental ambient air pollution by ozone and hospital admissions for tota l (p>0.1), childhood (p>0.1), and adult (p>0.1) asthma discovered by conducting simple l og-linear regression analysis.

PAGE 63

48 Figure 2. The total number of hospital admission s for childhood and adult asthma and average ozone concentration, ppb, in Hillsborough County, FL, by calendar quarter during the period of 1997-1999 0 50 100 150 200 250 300 350 400 Jan-Mar 97 Apr-Jun 97 Jul-Sept 97 Oct-Dec 97 Jan-Mar 98 Apr-Jun 98 Jul-Sep 98 Oct-Dec 98 Jan-Mar 99 Apr-Jun 99 Jul-Sept 99 Oct-Dec 99Calendar QuarterNumber of asthma hospitalizatio n 0.0 5.0 10.0 15.0 20.0 25.0 30.0 35.0 40.0 45.0Conc., pp b Childhood asthma Adult asthma O3, ppb Sulfur dioxide (SO 2 ) The average changes in environmental qua lity by ambient ozone were evaluated based on average quarterly concentration m easurements and the number of daily peak concentration values above the NAAQS per calendar quarter during the period of 19971999. There were a total of 7 monitoring sta tions in Hillsborough County, FL. The data on air pollution by SO2 was collected in Sites: 53, 81, 95, 109, 1035 and 4004 in Hillsborough County, FL. Average monthly and quarterly values at different ambient air quality monitoring site by separate year of study are presented in Tables 13-15. The average 1997-1999 value by different site varied from 3.1 ppb (s ite 4004) and 4 ppb (S ite 81) to 8 ppb (Site 1035) and 5.7 ppb (Site 53) in Hillsborough County, FL. Yearly average by separate year vari ed from 4.9 ppb in 1998 to 5.6 ppb in 1997 in Hillsborough County, FL (see Table 13 and 14). Av erage monthly variations were higher and varied from 3.8 ppb in April 1998 to 6.8 ppb in August 1997 and 6.1 ppb in August

PAGE 64

49 1998 in Hillsborough County, FL. In Hillsborough County, FL, there were 3.4 times higher average monthly concentration in Site 1035 as compared to the average annual concentration in Site 109 in September 1997. Table 13. Average monthly and quarterly sulfur dioxide concentration, ppb, by separate monitoring site, in Hill sborough County, FL, in 1997 1997 Site 53 Site 81 Site 95 Site 109 Site 1035 Site 4004 Average Monthly Conc., ppb Average Quarterly Conc., ppb January 4.0 5.8 6.3 8.3 5.0 NA 5.9 5.2 February 3.5 5.3 4.5 6.3 7.2 NA 5.4 March 2.7 3.2 5.2 4.7 6.5 NA 4.5 April 3.6 4.4 4.8 5.0 6.9 NA 4.9 5.6 May 5.7 4.7 4.5 4.6 8.1 NA 5.5 June 6.5 5.0 6.3 5.0 8.9 NA 6.3 July 5.4 3.5 6.6 5.4 12.1 NA 6.6 6.4 August 7.1 4.2 6.0 6.5 10.3 NA 6.8 September 7.2 3.9 4.5 2.9 10.0 NA 5.7 October 6.5 4.1 3.5 3.7 8.4 NA 5.2 5.1 November 4.3 4.8 6.5 5.0 5.7 NA 5.3 December 1.8 4.5 3. 6 9.2 5.2 NA 4.9 Average 4.8 4.5 5.2 5.6 7.9 NA 5.6 5.6 Table 14. Average monthly and quarterly sulfur dioxide concentration, ppb, by separate monitoring site, in Hill sborough County, FL, in 1998 1998 Site 53 Site 81 Site 95 Site 109 Site 1035 Site 4004 Average Monthly Conc., ppb Average Quarterly Conc., ppb January 3.3 3.5 2.8 3.9 6.4 NA 4.0 4.6 February 2.6 3.4 3.8 6.2 6.7 NA 4.5 March 3.1 4.8 2.8 7.7 7.4 NA 5.2 April 2.6 1.7 4.6 4.8 6.5 2.8 3.8 4.6 May 4.6 3.7 6.5 4.7 6.4 4.2 5.0 June 4.3 2.8 6.2 7.0 6.1 4.0 5.1 July 4.9 2.9 6.7 9.2 7.2 4.5 5.9 5.5 August 7.1 4.0 4.9 5.8 11.4 3.6 6.1 September 5.6 1.1 3.8 2.8 10.0 2.8 4.4 October 7.2 5.4 3.7 3.6 8.9 2.0 5.1 5.1 November 6.3 3.8 4.2 4.3 8.9 2.7 5.0 December 5.9 3.2 2.7 5.2 10.8 2.3 5.0 Average 4.8 3.4 4.4 5.4 8.1 3.2 4.9 4.9 The average exposure to sulfur dioxide and distribution of asthma hospital admissions for childhood and adult asthma by calendar quarter during the period of 19971999 is given below in Figure 3.

PAGE 65

50 Table 15. Average monthly and quarterly sulfur dioxide concentration, ppb, by separate monitoring site, in Hill sborough County, FL, in 1999 1999 Site 53 Site 81 Site 95 Site 109 Site 1035 Site 4004 Average Monthly Conc., ppb Average Quarterly Conc., ppb January 5.6 NA 3.6 4.9 8.3 3.6 5.2 5.2 February 4.2 4.3 5.4 6.0 6.5 2.9 4.9 March 6.0 5.6 5.6 6.8 7.7 2.1 5.6 April 4.3 2.5 6.7 6.8 9.7 2.5 5.4 5.5 May 5.1 4.0 5.6 4.6 9.0 3.1 5.2 June 8.0 3.8 4.5 4.8 10.8 3.5 5.9 July 5.1 2.6 5.5 6.0 8.9 3.0 5.2 5.2 August 5.2 3.1 6.3 4.3 8.3 3.1 5.1 September 5.0 3.5 3.1 7.5 7.8 NA 5.4 October 5.8 4.1 1.3 4.9 6.3 NA 4.5 4.6 November 4.6 7.0 2.5 3.1 6.4 NA 4.7 December 3.3 6.7 2. 8 3.7 7.2 NA 4.7 Average 5.2 4.3 4.4 5.3 8.1 3.0 5.1 5.1 Figure 3. The total number of hospital admission s for childhood and adult asthma and average sulfur dioxide concentration, ppb, in Hillsborough County, FL, by calendar quarter during the period of 1997-1999 0 50 100 150 200 250 300 350 400 Jan-Mar 97 Apr-Jun 97 Jul-Sept 97 Oct-Dec 97 Jan-Mar 98 Apr-Jun 98 Jul-Sep 98 Oct-Dec 98 Jan-Mar 99 Apr-Jun 99 Jul-Sept 99 Oct-Dec 99Calendar QuarterNumber of asthma hospitalizatio n 0.0 1.0 2.0 3.0 4.0 5.0 6.0 7.0Conc., pp b Childhood asthma Adult asthma SO2, ppb The number of sulfur dioxide peak concen trations and asthma hospital admissions for childhood and adult asthma by calendar quarter during the period of 1997-1999 are represented in Figure 4.

PAGE 66

51 Figure 4. The total number of hospital admission s for childhood and adult asthma and average number of peak concentrations for sulfur dioxide in Hillsborough County, FL, by calendar quarter during 1997-1999 Number of hospitalizations for childhood and adult asthma and a number of SO2 peak daily concentrations of 100 ppb or more by calendar quarter, Hillsborough County, FL, 1997-1999 0 50 100 150 200 250 300 350 400 JanMar 97 AprJun 97 JulSept 97 OctDec 97 JanMar 98 AprJun 98 JulSep 98 OctDec 98 JanMar 99 AprJun 99 JulSept 99 OctDec 99 Calendar Quater No. of hospital admissions/(peaks)x10 SO2, (number of conc. peaks)x10 Childhood asthma Adult asthma Seasonal variation of sulfur dioxide by calendar quarter revealed no significant association between ambi ent air pollution by SO2 and childhood (p>0.1), adult (p>0.1) or total (p>0.01) asthma hospital admissi ons during the period of study 1997-1999. The same trend by calendar quarter during th e overall period of study was shown for calculated number of sulfur dioxide peak concentrations. Simple log-linear regression analysis revealed that there was no statis tically significant association between the number of sulfur dioxide peak concentra tions and both childhood (p>0.1), adult (p>0.1), and total (p>0.1) asthma hospitalizations by calendar quarter during the period of 19971999 (see Appendix D). Temperature Average daily temperature measurements were used to calculate average values by calendar quarter and to evaluate the a ssociation between the total number of

PAGE 67

52 hospitalizations for asthma and changes in temperature by calendar quarter during the period of time 1997-1999. The peak temperatur e values were observed in the middle of Summer (July) and were lowest in Winter (December-January) (see Table 16) The average monthly temperature values over the entire period of study varied from 60.2 F in January 1999 and 60.4 F in December 1997 to 85.8 F in June 1998 and 85 F in July 1999. Daily values varied from as low as 46 F in January 6, 1997, to 88 F in June 27, 1998. Seasonal variations by calendar quarter rev ealed that the lowest temperature during the overall study period was specific to Quar ter I (January-March ) and Quarter IV (October-November). Average yearly temperat ure varied over time only very slightly from 74.1 F in 1997 to 75.5 F in 1998. Table 16. Average monthly and quarterly ambient temperature, F, Tampa Bay, 19971999 Month 1997 Average Temperature, F, 1997 1998 Average Temperature, F, 1997 1999 Average Temperature, F, 1999 January 63.2 68.9 65.3 66.3 60.3 65.3 February 69.6 67.5 70.0 March 73.8 66.2 65.8 April 74.4 78.1 70.1 78.2 72.4 77.7 May 78.0 78.6 79.0 June 81.8 85.8 81.6 July 81.8 81.9 83.2 82.6 85.0 83.7 August 83.4 83.4 84.0 September 80.4 81.3 82.0 October 76.3 67.6 79.8 74.7 76.8 71.9 November 66.0 74.8 72.0 December 60.4 69.7 67.0 Average 74.1 74.1 75.5 75.5 74.7 74.7 There was a strong statistically signifi cant inverse association observed between average temperature values and hospital admi ssions for total as well as both childhood and adult asthma by calendar quart er during 1997-1999 (see Figure 5). The Pearson correlation analysis provided separately for childhood, adult and total asthma hospitalizations by calendar quarter proved a strong invers e correlation between ambient temperature and both childh ood (r=-0.7, p<0.0001) and adult (r=-0.95, p<0.0001) asthma hospital admissions.

PAGE 68

53 Figure 5. Average ambient temperature, F, a nd number of hospital admissions for childhood and adult asthma by calendar quarter in Hillsborough County, FL, 1997-1999 0 50 100 150 200 250 300 350 400 Jan-Mar 97 Apr-Jun 97 Jul-Sept 97 Oct-Dec 97 Jan-Mar 98 Apr-Jun 98 Jul-Sep 98 Oct-Dec 98 Jan-Mar 99 Apr-Jun 99 Jul-Sept 99 Oct-Dec 99Calendar QuarterNumber of Asthma Hospitalizations0.0 10.0 20.0 30.0 40.0 50.0 60.0 70.0 80.0 90.0Temp., F Childhood asthma Adult asthma Temperature Simple log-linear regres sion analysis supported the conclusion that ambient temperature could explain the increase in hospital admission for childhood and adult in the selected study area. There was a str ong significant inverse association between ambient temperature and hospital admissi ons for childhood (p<0.0001), adult (p<0.0001), and total (p<0.0001) asthma (see Appendix D). Adult asthmatics were more sensitive to th e effect of ambient temperature decrease. Transformed simple log-linear regression model parameter estimates suggest that decrease in ambient air temperature by 10 F could account for up to 32%, 41%, and 37% increase in hospital admissions childhood, adu lt, and total asthma respectively. Simple log-linear regression analys is results suggest that s udden changes and decrease in temperature, F, in the colder period of time could explain the increase in asthma hospitalizations in th e Hillsborough County, FL.

PAGE 69

54 Total pollen counts Average monthly values and also seasona l patterns by calendar quarter along with a maximum peak and lowest values by separa te month were calculated in Hillsborough County, FL, 1997-1999. Each type of pollen ha s had its specific blooming season with corresponding maximum peak values. Calculat ed average monthly and seasonal pollen concentrations are presented for total tree, weeds, and grass counts by separate calendar quarter during the period of 1997-1999 (Tables 17-19) The distribution of total tree, grass and weed average number of counts is co mpared to hospital admissions for childhood and adult asthma by calenda r quarter for the period of 1997-1999 in Figures 6-8. Table 17. Average monthly and seasonal total tr ee pollen counts in Hillsborough County, FL, 1997-1999 Total tree pollen counts 1997 Monthly 1998 Monthly 1999 Monthly 1997 Quarterly 1998 Quarterly 1999 Quarterly January 77.4 358.8 95.3 260.6 581.3 496.3 February 187.5 474.3 368.5 March 516.8 911 1025.2 April 179 379.4 344 67.0 133.9 165.6 May 13 19.3 148.8 June 9 3 4 July 1.2 0 2.75 0.7 0.4 37.2 August 0 0 0.3 September 1 1.3 108.6 October 28.8 80 58.8 22.6 33.2 37.8 November 8.3 13.5 22 December 30.8 6 32.6 Total/ Average 87.7 187.2 184.2 87.7 187.2 184.2 Calculated average monthly and seasonal grass pollen counts by calendar quarter are presented below in Table 19 and Figure 8 respectively. Simple l og-linear regression analysis revealed tree, weed and grass pollen was not associated with increase in childhood asthma hospital admissions over th e period of time 1997-1999. There were two peaks in March-April and July -August noticed for total weed pollen during the period of study from 1997 to 1999 (see Tables 17-19). The peak seasonal va lues of total tree pollen counts were observed in February and Marc h. For total grass pollen, the highest number of pollen counts was in both July and August.

PAGE 70

55 Figure 6. Average number of total tree pollen c ounts and number of hospital admissions for childhood and adult asthma by calendar quarter during the pe riod of 1997-1999, in Hillsborough County, FL 0 50 100 150 200 250 300 350 400 Jan-Mar 97 Apr-Jun 97 Jul-Sept 97 Oct-Dec 97 Jan-Mar 98 Apr-Jun 98 Jul-Sep 98 Oct-Dec 98 Jan-Mar 99 Apr-Jun 99 Jul-Sept 99 Oct-Dec 99 Calendar QuarterNumber of Asthma Hospitalizations0.0 100.0 200.0 300.0 400.0 500.0 600.0 700.0Count Childhood asthma Adult asthma Tree Table 18. Average monthly and seasonal tota l weed pollen counts in Hillsborough County, FL, 1997-1999 Total weed pollen counts 1997 Monthly 1998 Monthly 1999 Monthly 1997 Quarterly 1998 Quarterly 1999 Quarterly January 0.6 0 0.5 20.7 10.2 20.5 February 3.8 12.3 17.3 March 57.8 18.3 43.8 April 15 67.4 33.3 10.2 34.5 23.2 May 11 13.5 17 June 4.5 22.5 19.4 July 39.6 11.2 28.5 28.5 16.9 28.3 August 21.8 13.3 15.3 September 24 26.2 41.2 October 16.5 23 26.8 19.0 17.0 28.6 November 31.5 25.5 56.3 December 9 2.6 2.8 Average 19.6 19.6 25.2 19.6 19.6 25.2

PAGE 71

56 Figure 7. Average number of total weed po llen counts and number of hospital admissions for childhood and adult asthma by calendar quarter during the period of 19971999, in Hillsborough County, FL 0 50 100 150 200 250 300 350 400 Jan-Mar 97 Apr-Jun 97 Jul-Sept 97 Oct-Dec 97 Jan-Mar 98 Apr-Jun 98 Jul-Sep 98 Oct-Dec 98 Jan-Mar 99 Apr-Jun 99 Jul-Sept 99 Oct-Dec 99 Calendar QuarterNumber of asthma hospitalizations0.0 5.0 10.0 15.0 20.0 25.0 30.0 35.0 40.0Count Childhood asthma Adult asthma Weed For asthmatic adults, tree and grass but not weed pollen was shown to be significantly associated with hospitalizati ons by calendar quarter Tree pollen counts were positively associated, while grass polle n had negative (inverse) association with adult hospitalization for asthma during the pe riod of study. Preliminary simple regression analysis results disclosed th at increase by 10 concentration units in tree may account for 1% additional hospitalizations for adult asthma while decrease by 1 unit in grass pollen could explain the increase in adult asthma hospital admissions by 11.9%. The interpretation of crude association based on only simple regression analysis models should be drawn very carefully because the established crude association could be explained by other third f actors, which are not include d in the study, or a possible confounder, which could change the significance and magnitude of a cr ude association if it was included in the multiple regression analysis. Ambient temperature was also strongly positively correlated with envi ronmental exposure to total grass (r=0.76,

PAGE 72

57 p<0.0001) and weed (r=0.35, p<0.05), and inversel y associated with total tree pollen counts (r=-0.69, p<0.0001). Table 19. Average monthly and seasonal tota l grass pollen counts in Hillsborough County, FL, 1997-1999 Total grass pollen counts 1997 Monthly 1998 Monthly 1999 Monthly 1997 Quarterly 1998 Quarterly 1999 Quarterly January 0 0 0.5 1.0 1.4 2.3 February 0 1.3 0.8 March 3 3 5.6 April 3 4.6 6.3 3.5 3.3 4.6 May 4 2.8 4 June NA 2.5 3.6 July 3.6 2.6 5.8 7.1 5.0 5.1 August 6.8 3.8 4.8 September 11 8.6 4.8 October 6 3.8 2.8 4.3 2.9 2.0 November 3.8 3.8 2.3 December 3 1.8 1 Average 4.0 3.2 3.5 4.0 3.2 3.5 Figure 8. Average number of total grass pollen c ounts and number of hospital admissions for childhood and adult asthma by calendar quarter during the pe riod of 1997-1999, in Hillsborough County, FL 0 50 100 150 200 250 300 350 400 Jan-Mar 97 Apr-Jun 97 Jul-Sept 97 Oct-Dec 97 Jan-Mar 98 Apr-Jun 98 Jul-Sep 98 Oct-Dec 98 Jan-Mar 99 Apr-Jun 99 Jul-Sept 99 Oct-Dec 99 Calendar QuarterNumber of Asthma Hospitalizations0.0 1.0 2.0 3.0 4.0 5.0 6.0 7.0 8.0Count Childhood asthma Adult asthma Grass

PAGE 73

58 These associations could explain the signi ficant positive crude association of total tree pollen and inverse (negative) association of grass pollen with adult asthma hospital admissions. To answer these questions, all af orementioned criteria ambient air pollutants and other possible environmental triggers of asthma were included in the multiple loglinear regression model by using stepwise back ward selection best-f it non-linear model techniques. 1.3.3. Multiple log-linear regression analysis Environmental exposure to ambient air pollution by ambient particles, sulfur dioxide, and ozone; number of sulfur dioxide peak concentrations; and such aeroallergen counts as weed, tree, and grass pollen leve ls, while controlling for calendar quarter, represented independent variables in the sepa rate multiple log-linear regression model of hospital admissions for childhood and adult asthma The statistical data analysis outcome is presented in Appendix D. The multiple re gression stepwise backward selection model building procedures revealed that ambien t temperature was the only significant and influential factor in the final best-fit re gression model for both asthmatic children and adults. None of the aforementioned selected criteria ambient air po llutants was shown to be significant in the multiple regression m odel after the adjustment to other known and probable environmental triggers of asthma. Tree and grass pollen, which were shown to be significant predictors for adult asthma hos pital admissions in crude analysis, were not significant variables, while c ontrolling for ambient temperat ure. The scatter plots for ambient temperature and hospital admissions for childhood and adult asthma are given in Figures 9 and 10 respectively. More detailed analysis of standardized residual deviances along with predicted estimates for childhood asthma hospital admissions did not show any extreme cases or outliers, which could influence the preci sion of developed best-fit models and interpretation of fitted model parameter estimates.

PAGE 74

59 Figure 9. Scatter plot of ambient temperature and childhood asthma hospital admissions in Hillsborough County, FL, by calendar quarter 1997-1999 50.0 55.0 60.0 65.0 70.0 75.0 80.0 85.0 90.0 050100150200250300 Childhood asthma hospital admissionsAmbient temperature, F Figure 10. Scatter plot of ambient temperature and adult asthma hospital admissions in Hillsborough County, FL, by calendar quarter 1997-1999 50.0 55.0 60.0 65.0 70.0 75.0 80.0 85.0 90.0 0100200300400 Adult asthma hospital admissionsAmbient temperature

PAGE 75

60 Standardized residual deviances analysis against predicte d values for adult asthma revealed two extreme cases above the cut-off value of 2 or below the cut-off value of minus 2 (2.4 and -2.2). Additional analysis of the best-fit regression model without extreme cases concluded that these two outliers are not influential and could be left in the final model. DFBETAS were used to estimate the in fluence of outliers on the model parameter estimates and statistical data anal ysis provided without selected extreme cases is presented in Appendix D. A large absolute value of DFBETAS indicates a large impact of the extreme case on the regression coeffici ent. The influential cases were defined as outliers with the absolute value of DFBETAS more than n 2 or 0.57, where n is a number of sample points (n=12). Calculated square root of mean square errors was 16.89 and DFBETAS value was 0.00018 ( DFBETAS << 0.57). The distribution of standardized model residual deviance estimates against pred icted asthma hospitalizations is illustrated in Figures 11 and 12 for childhood and a dult hospital admissions separately. Significant final best-fit log-linear regression model parameter estimates were transformed by using natural antilog transforma tion to provide quantit ative interpretation of multiple regression analysis results. Mult iple regression analysis outcomes supported previous crude association analysis results a nd suggested that a decrease in ambient air temperature of 10 F could account for up to 32% and 41% of an increase in hospital admissions of childhood and adult asthma respectively. Possible interaction between ambient temper ature and selected criteria ambient air pollutants as well as temperature and ambi ent aeroallergen counts for tree and grass pollen was estimated by including corresponding interaction terms in the multiple regression model to estimate its significan t effect on hospitaliza tions for both childhood and adult asthma (see Appendix D). Interaction term analysis suggest ed that none of the created interaction terms was significant and co uld modify the effect of variables left in the final best-fit re gression model.

PAGE 76

61 Figure 11. Standard residual deviance distributi on against predicted values for childhood asthma hospital admissions -3 -2.5 -2 -1.5 -1 -0.5 0 0.5 1 1.5 2 2.5 3 0100200300 Predicted valueStandard residul deviance Figure 12. Standard residual deviance distributi on against predicted values for adult asthma hospital admissions -3 -2.5 -2 -1.5 -1 -0.5 0 0.5 1 1.5 2 2.5 3 050100150200250300350 Predicted valueStandard residual deviance

PAGE 77

62 Unavailable daily data on asthma hospitaliz ations did not allow us to explore the daily or time-lag association between envir onmental exposure to selected ambient air pollutants and hospitalizations for chil dhood adult asthma. Aggregated data on environmental asthma triggers and asthma hos pitalizations were used to evaluate the association in crude and adjusted analyses. Further studies by using daily or selected time-lag data are required to support or oppose our study results before fi nal conclusions about the eff ect of ambient air pollution could be drawn. However, our preliminary descriptive data analysis and final study results strongly suggest that th ere are other possible risk fact ors, which were not included in our study and could explain the differences in hospitalization for asthma in the study area. Preliminary descriptive analysis revealed wide geographical vari ation in the spatial distribution of asthma hospita lization by area of re sidence and a cons istency of these variations over the overall period of study from 1997-1999. Therefore, the next step was undertaken to explore local spa tial characteristics which coul d be specific to a selected study area or to population groups living in these areas, and could explain geographical differences in hospitalization for asthma within the select ed area of study.

PAGE 78

63 Chapter 2 ASSOCIATION BETWEEN AREA SOCIOECONOMIC STATUS AND HOSPITALIZATIONS FOR CHIL DHOOD AND ADULT ASTHMA 2.1. Literature Review Socioeconomic status (SES) is an indicator of an individual’s position in society.37 Because people live in communities or neighborhoods which share the same social milieu, the socioeconomic status can be defi ned and measured at various levels: the individual, the family or household, the ne ighborhood or local community, and at the national level. Measures of the social cl ass of individuals have long been known to predict disease incidence, prevalence and mortality, while similar findings for residential areas have been reported usi ng socioeconomic data to classify small geographical areas by level of socioeconomic deprivation.38 Both individual and small area socioeconomic status indicators were show n to be associated with severe asthma attacks and hospitalization for asthma. The International Study on Asthma and Allergies in Childhood (ISAAC) used individual and area-based socioecono mic status indicators and provided evidence that living in an underp rivileged area is a si gnificant independent predictor of severity and hospital admission for asthma.39 Another study of asthma prevalence in relation to both individual and area socioecono mic status concluded that, irrespective of indivi dual socioeconomic status by edu cation level and social class, persons living in geographical areas with low education leve l had a significantly higher risk of asthma.40 Potential contributing factors could be classified as environmental, lifestyle and behavioral, psychosoc ial (stress-related), and access to health care resources related.3

PAGE 79

64 The study of area socioeconomic deprivati on status and human health has been carried out since the nineteenth century.38 Inequalities of health among different social classes have persisted since observations bega n, and virtually every he alth indicator ever examined proved the association with social class.37 In the United States information about the relationships among health, income, a nd use of health-care se rvices is collected in the ongoing household interview surveys by the National Center for Health Statistics (NCHS). Despite an abundance of cross-se ctional studies of area socioeconomic disparities in health, the temporal monitoring of specific disease prevalence and mortality trends by area socioeconomic characteristics still remains far less common in the US than in the UK and other European Union countries.41 Area-based socioeconomic deprivation indices have been widely used to analyze and monitor health outcome differentials in Europe, Australia, and New Zealand.38,41 The studies that do examine temporal trends in disease prevalence and mortality by area-based socioeconomic depriv ation or inequality measures have usually focused on singl e area-based socioeconomic measures.41 For most diseases displaying a social class gradient the more disadvantaged individuals and households experience higher risk, and this re lationship pattern is consistent over time.38 Socioeconomic status could be classified in many ways and is often expressed on an ordinary scale using such cr iteria as poverty, unemployment, income, education level, and occupation.42 Income differences and, more impor tantly, differences in education level and in social values are generally seen between blue-collar and white-collar workers. The correlation to health status is sometimes even closer with race than socioeconomic status, suggesting that cultura lly determined behavioral factors may be more important than income as determinants of health status. Cultural milieu influences health in several ways. Customs, traditions, religious beliefs and pr actices, health-related values and behavior are all very important. Understanding the dispr oportionate burden of and association between socioeconomic status and asthma may provide insight into the roots of the asthma epidemic and more effective disease management. Asthma differs from region to regi on, from neighborhood to neighborhood, and even from block to block within inner cities. Because asthma is most common among low socioeconomic status inner-city population groups,42 it is important to evaluate and

PAGE 80

65 define the extent to which ra cial and ethnic differences ar e due to social deprivation status.42 The poor are a mixed group, comprising persons who are low income, less educated, unemployed or marginally employed, and persons receiving subsistence from welfare. Previous studies concluded that asthma prevalence, severity of asthma symptoms, and asthma morbidity and mortality have substantially hi gher rates in ethnic minority populations living in geographica l areas with higher proportion of poverty, unemployment, ethnic minorit y, and low education level.43,44 An important group of poor people are single-parent families, elderly and disabled pensioners.42 The support provided for such people from public funds or charities is almost always inadequate to meet their needs for shelter, food, clothi ng, and adequate health care Some are in poor health because they are poor, and others are poor b ecause they are in poor health, but for many, there is a constantly reinforcing vicious closed circle of ‘poverty-poor health-poverty’.37 Poor people usually live in poor-quality hous ing and overcrowded conditions that favor the spread of home allergens and respirator y infections and are also socially more deprived from local providers of health care services. Low socioeconomic status described by poverty, family income, educati on level, and race/ethnicity (black and Hispanic versus white) was also linked to such home environment characteristics as direct or second hand parental smoking, poor ventilation, dampness, and home allergen levels, and may explain increased asthma morbidity and mortality rates for specific population groups.3,45,46,47 Restricted access to health care and reliance only on emergency departments for primary care were found to be associated with poor health outcomes in both rural and innercity poverty areas in the US.3,48 Unfortunately, many previous studies of asthma prevalence and health care use do not include household or family-level indicators of socioeconomic stat us, and usually are base d on clinical samples that are not a representative sa mple of total population at risk.42 Home allergens seem to have a critical ro le in the development and exacerbation of asthma in inner cities. Alle rgen sensitization and exposur e to allergen vary between populations and between socioec onomic strata within population.49 Previous studies clearly indicated that levels of sensitizati on and actual exposure to home allergens could be explained by socioeconomic gradient.49,50 Sensitization to house dust mites,

PAGE 81

66 cockroach, and cat was found to be a significan t risk factor for hospital emergency room visits in several previous epidemiological studies.51 Both allergic to cockroach and exposed to high levels of cockroach a llergens had significan tly higher asthma hospitalization rates, more unscheduled annua l medical visits for asthma, more missed school days, more days with wheezing and mo re nights with disturbed sleep compared with other children. The National Co-ope rative Inner-City Asthma Study (NCICAS) covered eight major inner city areas to ev aluate prevalence of atopy and environmental allergen exposure.51 The NCICAS concluded that most inner-city children with asthma are sensitized and exposed to multiple allergens and highly exposed to environmental tobacco smoke.51,52 In addition, neither increased e xposure to cockroach allergen, nor allergy to cockroach alone, was shown to be associated with greater morbidity in the aforementioned study. The major indoor allergens of relevance ar e dust mite, cockroach, cat, dog, and less commonly molds and other furry pets or rodents.3 Many previous studies have found multiple allergens in homes, and multiple sens itivities to main allergens among residents with asthma.3 Case-control studies of adults in Wilmington, Delaware, and children in Atlanta, Georgia revealed that subjects presenting with acute asthma exacerbations were substantially more likely than non-asthmatic controls to have multiple sensitivity to indoor allergens (cockroach, dust mite, or cat) and were more likely to have significant exposure to that alle rgen in their homes.3 In the latter study, asthmatic patients were exposed to high concentrations of home a llergens: 79% of home dust samples contained excessive mite allergen, and 87% of sample s contained excessive cockroach allergen levels. In the same study, asthmatic children were more sensitive to the allergens – 69% percent had elevated IgE antibody to dust m ite, cockroach, or cat, compared with only 27% of controls. 21 out of 35 asthmatic child ren were sensitive to and exposed to the specific allergen at home, compared with only 3 out of 22 contro l children ( odds ratio was equal to 9.5, p<0.001). Previous studies found that increased sensitization to cockroach, dust mites, and cat is a significan t risk factor for seve rity of asthma and emergency room visits by other studies.3,51

PAGE 82

67 Sensitization to cockroach a nd high cockroach allergen le vels in inner-city ethnic minority children population received particular attention during the last few decades. Based on previous studies, high-poverty ar eas are at lower risk of house dust mite allergen, but are the most important indi cators of high cockroach allergen levels.46,53 Asthma morbidity and mortality in the inne r-city lower socioeconomic status ethnic minority population was significantly associated with higher sensit ization and exposure to cockroach allergen levels.54 In the National Cooperati ve Inner-City Asthma Study,47 children from eight major inne r city areas in the country were assessed for atopy and environmental allergen exposure. Children a llergic to cockroaches and exposed to high levels of cockroach allergens had 3.6-fold higher asthma hospitaliz ation rates, more unscheduled medical visits for asthma per year, significantly more missed school days, more days of wheezing and nights with distur bed sleep compared with other children.51 Because higher cockroach allergen levels ar e more likely to be present in dwellings located in high population density areas and occupied by low-income and low education level ethnic minority families, it is very diffi cult to disentangle the role of individual socioeconomic factors in pr oducing high cockroach levels.53 The two most common dust mite allergens, Der p1 and Der p2, are both ve ry potent. Direct i nhalation causes an immediate fall in forced expiratory volume in 1 second (FEV1) and late-phase response in sensitive asthmatics.3 The prevalence and severity of dust mite allergens differs from region to region, and most homes in regions with high humidity thr oughout the year may have high levels of dust mite allergens. Expos ure to dust mite allergens may be high in some geographical areas where human crowding, older carpeting, bedding, and upholstered furniture, and high indoor humidity caused by poor vent ilation and leaking pipes provide excelle nt conditions for mite proliferation.3 Specific IgE testing has shown that as many as 22% of asthmatic children test positive for home pet allergen.3 Higher household income, higher maternal education le vel, living in a single-family home in a less populated area and less overcrowded housing conditions, and being a white householder were associated with elevated dust mites, cat, and dog, but low cockroach allergen levels.3, 27 Sensitization to certain pet allergen s has a more important role in the ongoing bronchial reactivity with mild-to-moderate asthma.3 An unwanted infestation

PAGE 83

68 with mice and rats describes many low socio economic status inner-city residence areas. Like pet allergens, rodent hair, urine, and f ecal allergens are very difficult to remove even after eradication of the pests. Previous studi es have shown that se nsitization and exposure to rodents can make asthma worse. Epidemiol ogical studies within sc ientific laboratories with rodents showed increase in asthma sy mptoms and decreased respiratory function among workers who handled rats.3 Other studies have also sh own a positive association between rodent allergy and asthmatic symptoms.3 However, little is known about residential exposure to rodent allergens and it s impact on asthma morbidity at the place of residence.3 Excessive humidity and dampness welcom es pests into the home and also causes growth of mold and fungi.3 Mold spores are ubiquitous both in outdoor and indoor environment air. Available evidence suggests that sensitization to molds mainly occurs in people who have a high potential for being se nsitized to other common indoor allergens.3 Although the prevalence of cigarette smoki ng has declined during the past decade, it still remains high in lower socioeconomic status population in inner-city areas in the United States.3 Passive exposure to environmental tobacco smoke (ETS) is more common in low-income, urban communities than in any other demographic population groups.17 Knowledge about the adverse he alth effects of tobacco sm oke is relatively low among poor and less educated people.3 ETS is a risk factor for th e development of asthma in early childhood and an aggravat ing factor that increases morbidity among children who already have asthma. Previous prospectiv e cohort studies have demonstrated a doseresponse relationship between the environmen tal tobacco exposure and the risk for asthma and wheezing symptoms in early childhood.3 The evidence suggests that environmental tobacco smoke may be a risk factor for early-childhood asthma, but it may have a more important role as a trigger to future asthma attacks. Recent comprehensive meta-analysis confirmed that second hand e xposure to parental smoking is associated with increased severity of disease among children with established asthma.3 Viral respiratory infections are another risk fact or known to worsen already existing asthma, but their role in the pathogenesis of as thma and allergy still remains unclear.3 Increased asthma prevalence and severity seen among lo w socioeconomic status inner-city children could be also partially explai ned by restricted access to he alth care which subsequently

PAGE 84

69 results in less antibiotic use and lower vaccination rates.3 Epidemiological studies proved that the low socioeconomic status inner-c ity asthmatic population experiences more difficulty in managing their asthma for a variet y of reasons, including lack of health care insurance, limited access to high-quality prim ary health care, lack of knowledge and understanding of asthma, the psychosocial stress of living in an urba n area, and lack of family and community support.3 In addition, low socioec onomic status inner-city communities with specific high rates of asthma morbidity have relatively low use of antiinflammatory asthma medications.3 There is limited descriptive epidemiologica l understanding of asthma at the state and local levels.14 Despite its importance, there is mi nimal surveillance of asthma across the country, and no comprehensive surveilla nce system has been established that measures asthma trends at the state or local level.2,14 Such information however is needed to identify specific high-risk populations and to design and evaluate further preventive interventions. Asthma is a highly prevalent chro nic disease that affects the quality of life of many people of all age groups, different race, and both genders in the United States. Implementation of better state and local surveillance networks could increase understanding of this disease and contribute to more effe ctive prevention strategies. Despite currently available advanced asthma medications, widely disseminated asthma care guidelines, asthma education and case-management programs, and growing knowledge about environmental and lifestyle factors that aggravate disease, excess asthma morbidity in low-income socioec onomic deprived inner-city communities still remains a major public health problem in the United States. Many scientists are puzzled by the recent increase in asthma morbidity and mortality.48 The major role of genetics in pred isposition to airway hyper-reactivity in people with asthma is supported by previous tw in and genetic linkage studies. However, since changes in the genetic make-up of i ndividuals occurs over generations, the rapid increase in the prevalence of asthma during the last decade suggests that changes at the genetic level are unlikely to be the cause.48 Some of this increase in the prevalence of asthma may be due to increas ed recognition and diagnosis of the disease given greater awareness of the pathophysiology of asthma, and recent changes in guidelines for the

PAGE 85

70 clinical signs and symptoms a ssociated with the disease by the National Heart, Lung, and Blood Institute.48 However, these factors are also unlikely to account for all of the increase in prevalence of asthma. Monitoring of environmental exposures related to asthma is usually restricted to the investiga tion of acute clusters of asthma symptoms or occupational exposures. Few biomarkers of exposure have been developed to indicate exposure to environmental or occupational f actors. However, we are still unable to determine whether the consistent increase in asthma is due to an increase in environmental air pollution levels. Urba nization has increased our exposure to environmental allergens and irritants. Psychos ocial factors that cause us to spend more time indoors and poor overcrowded conditions ma y also lead to increased exposure to indoor pollutants and allergens.48 Socioeconomic status indi cators and social status described by race are powerful predictors of disease at both the individual and ecological small-area level.38,46 In a complex way social factor s and living conditions seem to influence different risk factors and have to be considered as potential confounders or effect mediators in the analysis of diseas e risk factors and risk of atopic disease.46 Sources of industrial pollution tend to be located in relatively more disadvantaged socioeconomic status inner-city areas, ch aracterized by poor housing, low income and high unemployment.46 Local variation in th e socioeconomic status may be large within geographical areas and consistent with the hypothesis of environmental risk. Under these circumstances, there is marked potential for confounding in small-area analyses near local sources of higher environmental exposure.46 Socioeconomic status is likely to confound (usually in a positive direction) the relationship be tween disease and proximity to a point source of environmental exposur e, and may bias the association between environmental exposure to ambient air pollution and asthma.38 Therefore, a critical comparison of the known and possible risk f actors contributing to the exacerbation of asthma is needed to develop more effective preventive interventions that could reduce the disparities by the poor and ethnic minorities and could modify current asthma morbidity and mortality trends.

PAGE 86

71 Hypothesis Area socioeconomic deprivation status is a ssociated with severe exacerbations and hospitalizations for childhood a nd adult asthma, and could be used as an independent predictor variable to predict asthma hospital admission rates in a given area. The aim of the study was (1) to evaluate the association between hospital admissions for childhood and adult asthma and various area socioeconomic status indicators, and (2) to ascertai n which significant area socioec onomic status characteristics could be retained in the complex multidimen sional area socioeconomic deprivation index and could be used to predict asthma hospital admission rates in a given area. Study objectives 1. Explore the association between various area socioeconomic status indicators and hospital admissions for childhood and adult asthma; 2. Ascertain principal significant component s of area socioeconomic status, and to develop a complex multi-dimensional area socioeconomic deprivation index; 3. Evaluate existing opportuni ties and limitations of advan ced spatial data analysis techniques for environmental exposure asse ssment in ambient air pollution and asthma epidemiological studies; 4. Evaluate the association between hos pital admissions for childhood and adult asthma environmental exposure to selected criter ia ambient air polluta nts controlling for area socioeconomic status char acteristics and composite area socioeconomic deprivation index; 5. Develop best-fit predictive regressi on model of hospital admission rates for childhood and adult asthma based on area so cioeconomic deprivation status, and to evaluate the accuracy of mode l prediction through standard statistical regression model validation procedures by using independe nt data outside of the study area.

PAGE 87

72 2.2. Methodology 2.2.1. Study Design A population-based cross-sectional study was conducted to evaluate the association between asthma hospitalizations and local environmental expo sure to selected criteria ambient air pollutants and diverse area socioeco nomic status characteristics. Comparative cross-sectional studies were de signed to evaluate the associ ation in separate consecutive years in 1997, 1998 and 1999. The study area covere d a total of 44 postal zip code areas in Hillsborough County, FL. The zip code areas were coded by five-digit number according to standard definitions and codes de veloped by the U.S. Bureau of the Census. The State of Florida Agency for Health Ca re Administration (AHCA) Hospital Inpatient Discharge Dataset was used to identify a ll asthma hospital admissions of patient by geographical area of residence living in Hillsborough County, FL, during the period of 1997-1999. Asthma diagnosis and disease coding were based upon the International Classification of Disease, 9th Revision, Clinical Modification (ICD-9-CM).55 The principal codes of ICD-9-CM for asthma were 493.0-493.9. There were a total of 1712 hospital admissions for asthma as a principa l diagnosis reported in Hillsborough County, FL, in 1999. A total of 781 children younger than 15 years of age and 931 adults of 15 years of age and older were admitted to th e hospital with severe asthma in 1999. There were a total of 729 hospital admissions for males and 983 hospital admissions for females respectively. There were a total number of 1677 and 1406 hos pital admissions, including 774 and 549 childhood asthma hospital admissions, in 1997 and 1998 respectively. Each single hospital admission record was linked to the patient’s postal zip code area of residence. The number of annual hospital admi ssion varied from 2 hospitalizations in the postal zip code area 33572 to 118 hospital admissions for asth ma in the postal zip code area 33610 in Hillsborough County, FL, accordin gly. The postal zip code area of residence was the main geographical unit of data analysis and was used to calculate crude hospital admission rates for childhood and adul t asthma. Previous comparative crude and adjusted (standardized) asthma hospital admissi on rates by age, gender and race analysis

PAGE 88

73 concluded that calculated crude rates are not confounded by specific area demographic characteristics (see 1.3. Results, Chapter 1, p.31). 2.2.2. Environmental Exposure Assessment Exposure to air pollutants can be monito red at different levels. The different strategies for environmental exposure measur es can be listed in order of increasing accuracy as: qualitative (categorical) assessment of expos ure that distinguishes between relatively high and low exposure; fixe d geographical loca tion concentration measurements and continuous monitoring; multi-microenvironment quality assessment and concentration measurements including tim e activity pattern in formation; personal exposure monitoring; and biological monitori ng using biological exposure indicators.38 Ideal studies of air pollution exposure should in clude biological dose markers or personal exposure measurements. Since such measurem ents are very expensive in a full-scale epidemiological study, fixed site measurem ents, modeled concentrations, or even qualitative categorical classifications are more frequently used as exposure information in environmental epidemiology. The utilization of data from ambient air quality monitoring network sites provides the uni form methodology used for site selection procedures, measurements techniques and qua lity control. Environmenta l exposure was measured at fixed geographical location continuous m onitoring stations. Comparable standard measurements become widely available fr om many different areas towns and cities, which makes it possible to develop predicti ons for other populations based on available environmental exposure data. Environmental exposure to ambient air pollution must be considered in the epidemiology of many diseases, and is also a ve ry important determinan t of the quality of life of specific sensitive population groups. Although the degree of human exposure differs from one location to another, the leve ls of environmental pollution were shown to be geographically distribute d and spatially correlated.18,38 Environmental exposure can be seen as the result of the inte rsection of two separate sector s: (1) environmental pollution and (2) exposed individuals or sp ecific population groups (Figure 13).38

PAGE 89

74 Pollution emissions differ and can be sepa rated into chemical, particulates, and microorganism exposures. Each is derived fr om a wide range of both point and non-point sources and each also can be characterize d by different properties of mobility and persistence. Because of this complexity, mapping of exposure is clearly a challenging task and requires the ability to model the magnitude and distribution of exposure on the basis of other evidence. Figure 13 Description of various sect ors of environmental exposure The modeling of geographically restri cted point-site measurements of environmental exposure can be carried out at various levels of abstraction:38 (1) Integrated modeling – the modeling and mapping of actual levels of exposure by intersecting geographical models of pollutants a nd population distribution; (2) Concentration mapping – the modeling and mapping of the levels of pollutant in the environment (without rega rd to human distribution); (3) Dispersion modeling – the modeling and mapping of pollution dispersion from sources; (4) Emission mapping – the assessment and mapping of levels of emission or release of contaminants at source. Exposure Distribution of environmental pollution Distribution of exposed individuals or population groups Pollution emissions Environment Socioeconomic factors Demographic factors

PAGE 90

75 Geographical Information Systems (GIS) play s a crucial role in community health studies by efficient management of very larg e volumes of information and integration of data from a variety of information sources. By the use of spatial analysis, the original data can be transformed into more useful data sets and data la yers of value-added information. Through map overlay and spatial m odeling one can evaluate the relationship between disease incide nce rates within a gi ven population and sour ces of environmental exposure in order to generate and evaluate etiologic hypothesis. Dire ct linkage of human exposure to certain environmental factors a nd diseases such as cancer, reproductive dysfunction, chronic neurodevelopment disorder, acute myeloid leukemia, multiple myeloma, and dysfunction of the immune and endocrine systems was established previously.38, 57 Several studies conducted in the la st decade have used proximity to traffic flow as a proxy for exposure to tra ffic exhaust. Review of previous studies disclosed existing association be tween more intensive traffic flow with increased risk of childhood hospital admissions for respiratory symptoms, decreased lung function, and childhood asthma.57 GIS has been shown to be an essent ial tool for analysis and modeling of spatial environmental a nd socioeconomic data. Environmental quality measurements t ypically occur either as detailed measurements for specific sampling and mon itoring sites (e.g. ai r or water quality monitoring sites) or as a more general classification for mapped areas (e.g. soil type). It is also apparent that human exposure to envir onmental pollution is often local and specific, and data on human distributions are, in many cases, confined to relatively generalized aggregations (e.g. population census statisti cs). Thus, mapping of environmental exposure for ambient air pollution and asthma epidemiological studies depends crucially upon the ability to model spatial di stribution and relies upon methods of:38 (1) Spatial interpolation – prediction of environmental exposure to ambient air pollution at unmeasured locations from available measured data for surrounded locations; (2) Spatial extrapolation – the estimation of environmental exposure to ambient air pollution at locations be yond the surveyed geographical area.

PAGE 91

76 Techniques of spatial modeling and interpol ation are fundamental to concentration mapping and spatial analysis. To undertake sp atial interpolation an alysis requires a sound knowledge both of the nature of spatial data, a nd of the nature of sp atial variation in the environment. Very often environmental qual ity data refer to point locations: they represent results of routine point measur ements at specific ambient air pollution monitoring network sites, an alyses and observations at randomly or systematically chosen sample points, or data relating to sp ecific point estimate f eatures (e.g. emissions from chimneys). Therefore, to use environmen tal quality data, it is frequently necessary to estimate unknown local environmental conditi ons and more precise spatial distribution by interpolating limited sample measurements of environmental pollution. Many of the point measurements available refer to very precise locations, and very often give only limited geographical area coverage and to be able to use the environmental exposure data there is the need to interpolate or extrapol ate them to wider geogr aphical areas and to predict unknown values be tween limited known sample site measurements. Interpolation predicts values from a limited numb er of sample data points. Visiting every location in a study area to measure the concentration or magnitude of an environmental phenomenon is very often diffi cult and expensive. Instead, strategically dispersed sample input point locations can be selected and predicted values can be assigned to all other locations in the study area. The common spatial analysis techniques used for interpolation in Geographical Information Systems (GIS) ArcView software version 3.1 with Spatial Analyst version 1.1 extension are as follows: (1) Spline Interpolation; and (2) Inverse Distance Interpolation The main assumption that makes interpola tion a viable option is that spatially distributed objects are also spatially correla ted and have similar characteristics in the same geographical area. The va lues of points close to samp led points are more likely to be similar than those that are far apart. A typi cal use for point interpol ation is to create a spatial elevation surface from a se t of sample point measurements. Spline Interpolation estimates values using a mathematical functi on that minimizes overall surface curvature, resulting in a smooth surface th at passes exactly through the input points. The method is

PAGE 92

77 best for geographically varying surfaces, such as surface elevation or slope. Inverse Distance Interpolation uses geostatistical methods that include autocorrelation and measure of the statistical relationshi p among the measured sample points. Inverse Distance Interpolation also weighs surrounding measured point values to derive a prediction for an unmeasured location, and the weights are based not only on the distance between the measured points and the predicti on of location, but also on the overa ll spatial arrangement among the measured points and their values. Hillsborough County, FL, has relatively smooth and without significant ge ographical surface variations. The Inverse Distance Interpolation method was selected for spat ial analysis and modeling of environmental exposure to ambient air polluta nts. The interpolated average selected criteria air pollutant concentrations in each pos tal zip code area were used to stratify the total study area and create relatively high and low categories (strata) areas of environmental exposure to specific ambien t air pollutant in Hillsborough County, FL, 1999. The cut-off point value to define rela tively high and low e nvironmental exposure residency areas was defined by arithmetic average mean value of measured ambient air pollutant concentrations over the total st udy area in Hillsborough County, FL, in 1999. The ambient air pollution data was obtained from the Aerometric Information System (AIRS) database maintained by th e US Environmental Protection Agency (US EPA). The environmental exposure to such crit eria ambient air pollu tants as respirable particulate matters (PM10), sulfur dioxide (SO2), and ozone (03) in Hillsborough County, FL, in 1999, was evaluated from the AirDat a air quality monitoring program dataset coordinated by the Division of Air Resources Management of the Florida Department of Environmental Protection as part of the US EPA AIRS database ( www.dep.state.fl.us/Air/public ations/techrpt/amr.htm).58 Ambient air quality standards and standard measurement procedures used for continuous monitoring of air quality by selected criteria ambient air pollutants we re described previously in the Study Design section (see Methodology, Chapter 1, p. 32). Geographic Information Systems Spatial An alyst 1.1 version software and Inverse Distance Weighted interpolation techniques were used to estimate average annual exposure by ambient particles, sulfur dioxide and ozone by separate zip code area of

PAGE 93

78 residence in Hillsborough County, FL, in 1997, 1998, and 1999 respectively. Additional point-site measurement monitoring stations located in the adjacent counties were included to increase the accuracy of predicted average ambien t particle exposure over the study area. Adjacent Pinellas (4 ), Manatee (1) and Polk (1) Counties, FL, have added additional 6 continuous measurements monitori ng stations. The average annual particle concentrations at total 15 monitoring sites we re used for spatial interpolation to predict the average environmental exposure to PM10 in different zip code areas in Hillsborough County, FL, in 1997, 1998, and 1999 respectively Relatively high and low environmental exposure category areas were defined to calcul ate asthma hospital admission rate ratio attributable to the geographical areas of residence with relatively higher exposure to ambient pollution. The a ssociation between environmental exposure to main criteria ambient air pollutants and asthma hospital admission rates was evaluated by comparing environmental exposure and as thma hospitalization rates in relatively higher and lower exposure to ambient ai r pollution category residency areas. Adjacent Pinellas (4), Manate e (1) and Polk (2) Counties, FL, have added additional 7 continuous measurements monitoring stations. The average annual ozone concentrations at total 14 monitoring sites we re used for spatial interpolation to predict the average environmental exposure to SO2 in separate zip code areas in Hillsborough County, FL, in 1999. Relatively high and low SO2 exposure category (stratum) areas were defined to calculate asthma hospital admission rate ratio attributable to the geographical areas of residence with relatively high er exposure to ambient particles. Additional adjacent to our study area point -site measurement monitoring stations were included in the interpolation model to increase the accuracy of predicted average ambient ozone exposure over the study area. Ad jacent Pinellas (3), Pasco (1), Manatee (4) and Polk (2) Counties, FL, have a dded additional 10 continuous measurements monitoring stations. The average annual oz one concentrations at the total of 14 monitoring sites was used for spatial interpol ation to predict the average environmental exposure to O3 in separate zip code areas in Hi llsborough County, FL, in 1999. Relatively high and low environmental oz one exposure category areas we re defined to calculate

PAGE 94

79 asthma hospital admission rate ratio attributab le to the geographical areas of residence with relatively higher exposur e to ambient particles. 2.2.3. Socioeconomic status indicators There is little disagreement about measures of socioeconomic status represented by educational attainment, occupation, income, wea lth, social class, a nd social status or prestige at the individual level.40,41 However, there is no consensus on the specific variables that compromise the socioeconomic status of small geographical area of residence as a composite and multidimensional descriptive index.38,41 Previous studies suggested that indicator s defining small-area socioeconomic status should directly or indirectly reflect the normative value, so cial good, or social welfare in a given community.41 Previous studies also concluded that, irrespectively of individual socioeconomic status, persons living in th e geographical areas with low socioeconomic status had a significantly higher risk of asthma.40 The socio-economic differentials used to identify the common area socioeconomic status indicator has been studied since the nineteenth century.38 Over the past three decades, a number of standardized complex indices have been developed to measure socio-economic variation across small geogra phical areas. Indicators for constructing an area socioeconomic index may be drawn fr om the broad sub-domains of education, income, employment, housing, and transportation.38,41 Household indicators of socioeconomic status include family income, overcro wding, lack of basic amenities, restricted access to a car, single-parent families, elderly people living alone, mobility and migration of householders, and housing tenure.3,38 Specific individual charac teristics of the head of household such as occupation, ethnicity, or education, are also used to classify the entire household.38 Some socio-economic factors are most naturally measures only at the level of neighborhoods or local communities rather than of individuals: examples include housing stock, educational opportunities, empl oyment, and access to health care. A total number of 20 social and economic i ndicators that may be viewed as roughly approximating both absolute and distribut ive aspects of liv ing conditions and

PAGE 95

80 socioeconomic disadvantage status in a give n community were cons idered in the study. Selected indicators were drawn from the US Bureau of Census 2000 ( www.census.gov 2004) and included: (1) percentage of people living below poverty level; (2) percentage of unemployment ; (3) percentage of persons with education level of ninth grade or less education as the highest degree of school completed; (4) percen tage of householders with the average annual income of or less than $15,000; (5) pe rcentage of persons employed in professional, managerial, administ rative, and clerical positions ( white-collar occupation employees); (6) percentage of unskilled persons ; (7) percentage of householders moved during the last year ; (8) percentage of households with overcrowding conditions (over one person per room); (9) percentage of divorce rate; (10) percentage of single parents with children ; (11) percentage of elderly population of 65 years and older; (12) percentage of children population of 5 years and younger; (13) percentage of black persons; (14) percentage of ethnic minority as a head of household ; (15) percentage of houses heating by fuel oil, kerosene, gaso line and other combustible liquids; (16) percentage of houses heating by wood ; (17) percentage of houses heating by gas ; (18) percentage of houses built in 1960 or before ; (19) percentage of houses lacking kitchen and/or plumbing facilities ; (20) percentage of families with no access to vehicle (no automobile ownership). All of the above mentioned vari ables were selected based primarily on the basis of their theoretical relevance and prior em pirical research study results.38,41 The selected socioeconomic indicators were chosen so as to broadly represent educational opportunities, labor force skills, and housing conditions prevailing in the study area. Taken together, these variab les may be viewed as reflecting key socioeconomic resources and socioeconom ic deprivation status within a study population.41 The zip code areas of permanent reside nce boundaries were used to define and group geographical areas with the same soci oeconomic status category (stratum) into low, medium and high socioeconomic deprivat ion status areas. Diffe rent socioeconomic status areas were defined and stratified into separate categories according to 50% percentiles and tertil es (33% and 66% percentiles) base d on total data distribution and variance as a selected cut-off point values. Socioeconomic status categories were divided

PAGE 96

81 into three (high, medium and lo w) or two (high and low) stra ta by establishing an average cut-off point at 33rd and 66th or 50th percentiles respectively. Selected socioeconomic status indicators and defined so cioeconomic status categories (s trata) were used in further stratified area analysis to evaluate socioeconomic confoundi ng effect on the association between environmental exposure to ambient air pollution and asthma hospital admissions within separate socioeconomic deprivation categories. Crude and adjusted rate of childhood and adult asthma hospital admissi on ratios were compared to evaluated possible confounding effect and e ffect modification (interacti on effect) within selected socioeconomic strata. The household income statistics included the income of the householder and all other individuals 15 years old a nd over in the household, whet her they are related to the householder or not, and represented cale ndar year 1999. Because many households consist of only one person, average household income is usually less than average family income. In compiling statistics on family in come, the incomes of all members 15 years old and over related to the householder were summed and treated as a single amount. Per capita income, the mean income computed for every man, woman, and child in a particular group, was derived by dividing the total income of a particular group by the total population in that group. Higher family in come represents the relative affluence and wealth of communities. The overall poverty st atus of families and unrelated individuals living in a particular household in 1999 wa s determined using 48 thresholds (income cutoffs) arranged in a two dimensional matrix.56 The matrix consists of family size (from 1 person to 9 or more people) cross-classi fied by presence and number of family members under 18 years old (from no children present to 8 or more children present). Unrelated individuals and 2-person families were further differentiated by the age of the reference person (under 65 years old and 65 year s old and over). If the total income of that person's family was less than the threshold appropriate for that particular family, then the person was considered poor, together with every member of his or her family. If a person was not living with anyone related by birth, marriage, or adoption, then the person's own income was compared with his or her poverty threshold. Poverty measures extreme aspects of material deprivation in a given community. Be sides poverty, income

PAGE 97

82 disparity measures the uneven distribution of economic resources. Data on educational attainment were derived from answers to long-form questionnaire and was tabulated for the population 25 years old and over. People we re classified according to the highest degree or level of school completed. The pe rcentage of people who completed only 9th grade or less was included in to the analysis of socio economic status indicators. Geographical areas with lower rates of education achievement may have limited economic opportunities in terms of availability of jobs, reduc ed demand for skilled labor, and fewer resources like schools of higher e ducation. All civilians 16 years old and over were classified as unemployed if they were not "at work", were looking for work, and were available to start a job during the last 4 weeks. Also included as unemployed were civilians 16 years old and over who: did not work at all dur ing the reference week, were on temporary layoff from a job, had been inform ed that they would be recalled to work within the next 6 months or had been given a da te to return to work, and were available to return to work during the refe rence week, except for temporar y illness. Examples of job seeking activities were: registering at a publ ic or private employment office; meeting with prospective employers; investigating pos sibilities for starting a professional practice or opening a business; placing or answering advertisements; wr iting letters of application; being on a union or professional register. The un employment rate is an indicator of social disintegration and is associat ed with higher suicide rates. The Census 2000 classification of occupations was based on the 1997 North American Industry Classification System (NAICS) published by the Office of Manageme nt and Budget, Executive Office of the President.56 White-collar occupation employers were persons employed in professional, managerial, administrative, and clerical positions. Such jobs may imply higher wage rates, more stable labor markets, and a greater presence of large, profitable, high technology, and capital intensive industries.41 Unskilled persons included persons with high school education, but without further pr ofessional education or training acquired. Householder movement during the last year usually represents social instability of householder and all individuals related to household. Household overcrowding usually indicated poor environment and lower income Overcrowding was defined as one person or more per room living in the household. Divor ce rates and single parents with children

PAGE 98

83 reflected family status and are associated wi th social deprivation status of households. A higher percentage of single-parent households with young children is generally associated with greater economic deprivation. Similarly, divorce rate, generally an indicator of social disintegration, may reflect social disadvantage to the exte nt that is associated with higher rates of poverty, unemployment, inad equate housing, and declines in social network relationships.41 Children under 5 years and th e elderly population of 65 years and older were additional area socioeconomic status indicators used in the study to represent these specific groups in the study area. Lack of access to a vehicle reflects socioeconomic disadvantaged status of householder and family members. Lack of access to automobiles represents economic depriva tion as well as trans portation difficulties, whereas overcrowding and lack of kitchen and/ or plumbing facilities reflect substandard housing and indoor environment conditions. Black person and ethnic minority head of household were indicators of social status closely related with specific cultural and behavioral factors which could explain espe cially high asthma hospitalization rates in these specific population subgroups. Type of fuel used for house heating indicates environmental exposure to indoor air pollutants especially to volatile organic compounds, which are well-known triggers for asthma a nd are more prevalent in houses that are heated by oil fuel, kerosene and other combustible liquids.27 Cooking and heating fuels are two major sources of indoor volatile or ganic compounds (e.g. formaldehyde, benzene, acetone, etc.). House age was defined as old if was built in 1960 or before; and together with lacking kitchen and plumbing amenities reflects a more deprived and poor indoor environment quality because of ineffective air ventilation systems, higher humidity, and supportive environment for high indoor mold concentrations and higher home allergen and antigen levels. Mold and f ungi proliferate in excessive moisture and in high humidity indoor environment, especially in th e kitchen, laundry or bathroom areas. 2.2.4. Socioeconomic Deprivation Index (SDI) Previous attempts to develop area socio economic indices in the US have included such census variables as median household inco me, percentage of adu lts with at least a

PAGE 99

84 high school or college education, percentage employed in professional or managerial occupations, housing tenure, household crowdi ng, unemployment, poverty, single-parent, income disparity, automobile ownership and migration.41 Initially originated in the datadriven approach of disease cluster analys is, different types of area socioeconomic deprivation status classification schemes were developed for current scientific research.38 The Urban Deprivation Index (UDI) and Unde rprivileged Area (UPA) values were used to describe and represent the complex multidimensional community socioeconomic deprivation status within th e study area. The UDI and UPA sc ores were calculated for each zip code area to evaluate the associati on with asthma hospital admission rates in the area and a possible interaction with environmental exposure to particles, sulfur dioxide, and ozone in multiple regression analysis and modeling. Underprivileged areas could be characte rized by high numbers of patients who are thought to increase the workload of the heal th care providers in a specific geographic area.38,59 The eight census-derived underprivileged status in dicators along with their relative weights are shown in Table 20. Indicators were e xpressed as percentages and were derived for selected zip code areas of residence from the US Bureau of Census 2000.56 Table 20. Census-derived variables with corres ponding standardized relative weights contributing to the Underp rivileged Area (UPA) score Area Socioeconomic Status Indicator Relative Weight Per cent of elderly living alone 6.62 Per cent of children under 5 years old 4.64 Per cent of persons in households with an unskilled head 3.74 Per cent of unemployed 3.74 Per cent of persons in households with single parent 3.01 Per cent of persons living in overcrowded households 2.88 Per cent of persons who moved house in last year 2.68 Per cent of persons in households with ethnic minority head 2.50 The British Inner Cities Directorate published an Urban Deprivation Index to assess the relative levels of deprivation in local communities.38,59 Urban Deprivation Index has three main dimensions of deprivation: so cial, economic, and housing (Table 21). Urban Deprivation Index (UDI) comp rised a weighted sum with all weights set to unity.

PAGE 100

85Table 21. Dimensions of deprivation and variables c ontributing to the Urban Deprivation Index (UDI) Dimension Socioeconomic Variable Social Per cent of households with single parents Per cent of pensioners living alone Economic Per cent of persons unemployed Per cent of persons unskilled Housing Per cent of households overcrowded Per cent of households without amenities Significant socioeconomic status indicat ors were selected for the principal component analysis based on the significance and magnitude of asso ciation between the specific socioeconomic status variable and asthma hospital admission rates. Main goals of principal component analysis were to reduce a large number of socioeconomic variables, select main va riables based on relative wei ght and correlation with the principal components factors, and develop a set of socioeconomic deprivation status indicators which could be treated as unco rrelated variables. The analysis of the significance of developed socio economic deprivation index to predict the risk of asthma hospital admission rates, and to evaluate the effect of environmental exposure to ambient air pollutant while adjusting to community socioeconomic deprivation status were also performed. 2.2.5. Data analysis The statistical data analysis was conducted by using the SAS System V8.2 version60 and the CDC EPI-INFO 3.2 version61 statistical data analysis software programs. Simple descriptive, correlation, stratif ied frequency table, principa l component, and Poisson loglinear regression analysis techniques were used for data analysis. Simple scatter plots and residual deviance analyses were used to e xplore residuals and to identify outlying or extreme cases. DFBETAS values were calculated to co mpare final fitted regression models with and without sele cted extreme cases (outliers) and to identify influential cases. Similar statistical data analyses were used and described in more details in the

PAGE 101

86 previous study of the association between hos pitalizations for asthma and environmental asthma triggers (see 1.2. Methodology, Chapter 1, p. 17). Limited point-source data on main criteria ambient air pollutants were interpolated and spatially analyzed to calculate aver age exposure values by using Geographic Information System (GIS) ArcView 3.2 vers ion and GIS Spatial Analyst 1.1 version62 geospatial analysis softwa re programs. GIS ArcView so ftware was used to map environmental exposure to selected ambient air pollutants and to represent different socioeconomic status areas in relation to asthma hospital admissi ons for children and adults over the overall study area graphical ly. The socioeconomi c deprivation index (SDI) construction was performed by applying principal component analysis (PCA) techniques and SAS System V8.2 stat istical data analysis software.60 PCA was performed to reduce a large number of socioeconomic st atus indicators to a smaller number for modeling purposes, and to develop a co mplex multidimensional area socioeconomic deprivation index based on a de fined subset of significant indicators associated with the principal component factors in the model. De fined set of socioec onomic characteristics could be treated as uncorrelated variable s as one of the approaches to handle multicolinearity and inter-correlation problems in further multiple regression analysis. The main objective of principal component anal ysis is to reduce the original number of explanatory variables into fewer composite variable s The ultimate goal is to account for the maximum portion of the variance present in the original set of variables with a minimum number of composite variables called principal components There are several standard evaluation criteria and most principal component analyses usually use more than one of these criteria to deci de on the number of principal components to be extracted. The Kaiser-Gutmann rule percentage of variance, and the scree test techniques were used as the most common standard techniques in the principal component analysis of selected socioeconomic characteristics to determine th e number of factors (principal components) and to identify significant lo adings of selected variable s within the main component factors. The Kaiser-Guttman evaluation could be e xplained by the “eigenva lues greater than one” rule and is available in various statistica l data analysis computer software packages.

PAGE 102

87 The number of factors (p rincipal components) is extracted based on eigenvalue (variance) of selected factors greater than 1.0. The rati onale for choosing this particular value is based on the assumption that a principal compon ent must have variance at least as large or larger than a single standard ized original variable. Percenta ge of variance explains the percentage or proportion of the common variance (defined by the sum of commonality estimates ) that is shared and e xplained by separate factor s. Factors (principal components) sharing 75 percent or more of the common variance should be retained. The Scree test plots the eigenvalues against the corresponding f actors numbers and gives visual insight into the maxi mum number of factors to ex tract or retain for further interpretation. The most important part of the outcome is a ma trix or factors loadings. By performing the oblique rotation method, the data analysis output includes a factors pattern matrix, which represents a matrix of standardized regression coefficients for each of the original variables on the factors. A rule of thumb used to evaluate factors is that any factor loading described by a standardized regression co efficient of equal to or greater than 0.3 in absolute value are consid ered to be significant and retained. The influence and relative weights of each sele cted socioeconomic indicators in the socioeconomic deprivation index was repr esented by standardized score values. Possible confounding and interaction effect s were evaluated by using standard epidemiology methods.38,63,64,65 Stratified frequency table and multiple regression analyses were performed to evaluate possi ble confounding effect by and interaction of socioeconomic status with environmental e xposure to ambient air pollution. Stratification and multivariate analysis (modeling) analytical tools were used to control for and evaluate the confounding effect of socioecono mic status, to assess effect modification, and to summarize the association of different predictor variables with risk of asthma hospital admissions. Effect modi fication (interaction) was de fined in two different, yet compatible ways: (1) based on homogeneity or heterogeneity of effects in stratified analysis; and (2) based on the comparison between observed and expected joint effects of possible risk factor and third variable in multivariate regression analysis.63,64,65 Based on the previous simple correlati on analysis of socioeconomic status indicators and crude asth ma hospital admission rates, statistically significant

PAGE 103

88 socioeconomic variables were selected for stratified frequency table analysis. Poverty status, family income, white-collar occ upation, single parent with children, and overcrowded housing conditions were main socioeconomic status indicators used to define different homogeneous categories (strata) and to evaluate the association between environmental exposure to ambient air pollution and asthma within defined socioeconomic status category (stratum). B ecause of a limited number of asthma hospital admissions within socioeconomic variable categories we could not evaluate the confounding effect and effect modification (interaction) for such geographical area socioeconomic status indicators as unempl oyment, level of education, ethnic minority householder, age of house, lacking kitchen a nd plumbing facilities, and house heating by fuel. However, all of these aforementioned area socioeconomic status indicators were included in the log-linear multiple regression model to estimate the association between asthma hospitalizations and environmental e xposure to criteria ambient air pollutants controlling for selected area socioeconomic status characteristics. Due to higher intercorrelation and multic olinearity among various area socioeconomic status characteristics, a standardized complex area socioeconomic deprivation index was constructed and used as an independent vari able to evaluate possible interaction with environmental exposure to selected ambien t air pollutants in the multiple regression models and accuracy of fina l fitted regression model prediction to predict childhood and adult asthma hospital admissions in the regression analyses. Poisson log-linear regression modeling techni que was used to develop the best-fit log-linear regression model and to predict as thma hospital admissions rates by significant explanatory variables. The Poisson best-fit log-linear regression model building and analysis techniques were explained in the previous study of the association between hospitalizations for asthma and environm ental asthma triggers (see Methodology, Chapter 1, p. 32). The SAS 8.2 version statis tical data analysis software options DSCALE (or SCALE= DEVIANCE) or PSCALE (or SCALE= PEARSON) were used in the general linear models statement of PR OC GENMOD to control the overdispersion.60 The main outcome variable of interest was the rate of hospital admissions for children and adult populations by zip code area of residence. Therefore, the OFFSET option was

PAGE 104

89 used to represent children and adult reference populati on groups in the selected geographical area of residence in the mu ltiple regression model. DIST=POISSON or DIST=P indicated that the Poisson log-linear regression analysis model was used in the generalized linear regression models’ opti on specified by SAS. Wald 95% Confidence Interval values (95% CI) were used to evaluate mode l parameters. Maximum likelihood estimates were used to predict the effect of socioeconomic status and environmental exposure descriptive variable s in the model. Chi-square statistics and corresponding statistical significance estimates by p-value were used for the analysis and further interpretation of model parameters. DFBETAS estimates were used to evaluate the influence of each outlying case on the regression model parameters (coefficients ). The detailed analysis of residual deviance and selected extreme cases was described in more detail in the previous study of the association between hospitaliza tions for asthma and environmental asthma triggers (see Methodology, Chapter 1, p. 32). The DFBETAS estimate by its value indicated whether inclusion of a case leads to an increase or a decrease in the estimated regression coefficient. A large absolute value of DFBETAS indicates a large impact of the extreme case on the regression coefficient. An extreme case was defined as an influential outlier if the absolute value of DFBETAS was equal to or more than n 2 .66 The final step in the fitted model building procedure is the valida tion of the selected best-fit regression model. Model valida tion usually involves collection of new independent data and checki ng the model against indepe ndent data. The purpose of model validation with new data is to expl ore whether the regression model developed from the previous data set is still applicab le for the new independent data. The method of validation designed to calibrate the predictive ca pability of the select ed best-fit regression model was used.66 Selected best-fit model developed fr om the given data set is chosen, at least in large part, because it fits well the or iginal data. A different model in terms of predictor variables most likely would be de veloped by using a diffe rent set of random data set. During the model development process the error mean square (MSE) will underline the inherent variability for future pr edictions from the selected best-fit model. The actual predictive capa bility of selected model could be evaluated by using the best-fit

PAGE 105

90 regression model to predict each case in the new independent data set and to calculate the mean of the squared prediction errors ( mean squared prediction error estimates MSPR). )2 ) ((n MSPRY Y pred i i Where: Yi – value of the res ponse variable in the i th validation case; Yi(pred) – predicted value of the i th validation case; and n* total number of cases in the validation set; The selected best-fit regression model is not seriously bi ased and gives an appropriate indication of the predic tive ability of the model if the mean squared prediction error (MSPR) is fairly close to error mean square (MSE) based on the original model-building data set. Significance of the predictive model was validated by using the selected model with independent da ta to predict hospital admission rates outside of the study area. Postal zip c ode areas of residence (n=44) in Pinellas County, FL, were used to calculate small-area socioeconomi c deprivation index and to compare both predicted and actual hos pital admission rates by each zip co de area of residence for the final best-fit model valida tion purposes. Both MSPR and MSE were compared to ascertain the predictive ability of selected regression model. Human subject protection was secured by using non-identif iable personal information and by providing adequate security fo r initial datasets and final data analysis results. The study protocol was reviewed and approved by the Institutional Review Board, Division of Research Compliance, Un iversity of South Florida, on May 28, 2004 (Protocol No. 102536, see Appendix A). 2.3. Results 2.3.1. Spatial Interpolation The main study objectives were: (1) to eval uate average environmental exposure to selected criteria ambient air pollutants levels by geographical area of residence; and (2) to

PAGE 106

91 estimate the association betw een hospital admissions for ch ildhood and adult asthma and environmental exposure to ambient air po llution controlling fo r residential area socioeconomic status in Hillsborough County, FL, in 1997, 1998, and 1999. Three separate comparative annual st udies were designed to comp are estimated association and explore the pattern of the association over the period of time. Simple frequency table analysis was provided to support the results of correlation asso ciation analysis by area of residence. Stratified frequency table an alysis was conducted to explore possible interaction and/or modificati on effect of area socioeconomic status on the association between asthma hospital admissions and envi ronmental exposure to selected criteria ambient air pollutants. Only significant select ed criteria ambient air pollutants based on crude association analysis we re used in the stratified an alysis. The period of study for stratified data analysis repr esented the year 1999. The overa ll study area wa s divided by geographical areas of residence into separa te relatively high and low environmental exposure to ambient air pollution categor ies (strata). Rela tively high and low environmental exposure to air pollution areas of residence were defined by estimated 50th percentile value as an arithmetic average cutoff point value, and formed two areas of relatively high and low environmental exposure categories (strata) wi thin the selected study area. The distribution of asthma hospital admissions within separate environmental exposure geographical areas of residence and defined categories (or strata) was used to evaluate the relative rate ratio of asthma hospitalization estimates attributable to relatively high environmental exposure, while adjusting to (controlling for) socioeconomic status effect in the stratified analysis. Finally, simple and multiple nonlinear regression models were developed to s upport previous conclu sions and to explore in more details the influence of area so cioeconomic status on hospitalizations for childhood and adult asthma. Multiple regression analysis provided valuable information to ascertain which local environmental a nd area socioeconomic status indicators are significant variables to explai n and also to predict the increase in asthma hospital admission rates by geographica l area of residence. The Geographic Information Systems (GIS) ArcView 3.2 Spatial Analyst 1.1 extension was used to develop Inverse Distance Weighted interpolation model and to

PAGE 107

92 predict unknown average environmental exposure values of such selected criteria ambient air pollutants as coarse particulate matter, su lfur dioxide and ozone by zip code area of residence in Hillsborough County, FL, in 1997, 1998, and 1999. Continuous ambient air quality by particulate matter, sulfur dioxide and ozone monitoring stat ions are illustrated in Picture 3. Picture 3. Ambient air quality monitoring network in Tampa Bay, FL, 1997-1999 To conduct original data in terpolation and spatial expos ure analysis we created initial database of ambient air monitori ng stations with corresponding average annual concentrations by coarse part iculate matter, sulfur dioxi de and ozone geographicallycoded with decimal degrees coordinates of la titude and longitude transferred from initial UTM system coordinates.

PAGE 108

93 Particulate matters The average annual concentrations by sepa rate ambient air quality monitoring site were used as an initial databa se for spatial interpolation. Table 22. List of ambient air quality monitoring stations and average annual values for PM10, g/m3, in 1997, 1998, and 1999 Site No. Site Name Latitude Longitude PM10, g/m3, 1997 PM10, g/m3, 1998 PM10, g/m3, 1999 057-0030 Tampa City 27.930527 -82.509611 26 27 24 057-0066 Gibsonton 27.893100 -82.401944 36 32 35 057-0083 Gardinier Park 27.862166 -82.382500 24 25 24 057-0085 Eisenhower 27.791055 -82.368305 21 23 20 057-0095 Gannon 27.921000 -82.401416 28 29 27 057-1002 DOH 27.946777 -82.451638 28 29 26 057-1035 Davis Island 27.926472 -82.454833 26 27 25 057-1068 Gaither School 28.101100 -82.504200 20 21 20 057-1069 Harbour Island 27.934200 -82.451900 28 30 28 057-1070 Central Ave. 27.985972 -82.454111 30 30 28 057-2002 Brandon 27.400300 -82.278600 23 23 22 081-0008 Holland 27.621305 -82.539472 22 24 24 103-0012 St. Petersburg 27.783100 -82.659400 25 26 26 103-0018 St. Petersburg 27.784111 -82.740027 21 23 22 103-3004 Largo 27.893900 -82.774700 24 26 25 103-5002 Tarpon Spring 28.088600 -82.701100 21 20 20 105-0010 Mulberry 27.854500 -82.017694 20 24 22 105-2006 Mulberry 28.000000 -82.000000 25 25 22 Initial average values for ambient air pollution by particulate matter PM10 at different ambient air quality monitoring s ites in 1997, 1998, and 1999 are presented in Table 22. The prefix 057 repr esented monitoring sites lo cated in Hillsborough County, FL, and the prefix 103 indicated ambient air quality monitoring sites in Pinellas County, FL, respectively. To increase the environmen tal exposure assessment accuracy adjacent to the study area monitoring sites were also us ed for spatial interpolation. There were 2 monitoring sites (Sites 1 05-0019 and 10-2006) in Polk County, FL, and 1 ambient air quality monitoring site (081-0008) in Manate e County, FL. Interpolated environmental exposure concentrations to am bient air pollution by partic ulate matter by separate zip code area of residence in Hillsborough County, in 1997, 1998, and 1999, are presented in Table 23.

PAGE 109

94 Table 23. Interpolated PM10 concentrations, g/m3, by zip code area of residence in Hillsborough County, FL, in 1997, 1998, and 1999 Zip Code Count No. Area, ft2 Conc., g/m3, in 1997 Conc., g/m3, in 1998 Conc., g/m3, in 1999 33510 43 0.0019 27.5 27.9 26.6 33511 88 0.0039 27.2 27.6 26.4 33527 162 0.0072 26.4 27.1 25.5 33534 70 0.0031 23.9 25.0 23.3 33547 1086 0.0485 24.4 25.9 24.2 33549 312 0.0139 23.6 24.3 23.0 33556 228 0.0102 23.5 23.8 22.8 33565 569 0.0254 25.8 26.5 24.6 33566 118 0.0053 24.9 25.7 23.4 33567 324 0.0145 24.7 26.0 24.1 33569 314 0.014 25.7 26.5 25.0 33570 293 0.0131 24.3 25.6 24.2 33572 29 0.0013 23.8 25.1 22.9 33573 94 0.0042 24.0 25.2 23.2 33584 111 0.005 27.2 27.8 26.2 33592 273 0.0122 26.7 27.3 25.6 33594 119 0.0053 26.5 27.2 25.7 33598 283 0.0127 24.9 25.9 24.3 33602 14 0.0006 28.0 29.1 26.6 33603 21 0.0009 29.1 29.4 27.4 33604 46 0.0021 28.3 28.7 26.8 33605 44 0.002 28.2 29.0 26.9 33606 22 0.001 27.2 28.3 26.2 33607 55 0.0025 27.2 28.0 25.7 33609 25 0.0011 26.6 27.6 25.0 33610 92 0.0041 28.1 28.5 26.9 33611 29 0.0013 27.0 27.8 25.7 33612 54 0.0024 26.4 26.9 25.3 33613 45 0.002 23.7 24.5 23.4 33614 52 0.0023 27.0 27.7 25.8 33615 52 0.0023 25.9 26.5 24.9 33616 23 0.001 26.9 27.7 25.9 33617 49 0.0022 27.4 27.9 26.3 33618 39 0.0017 22.3 23.1 22.0 33619 171 0.0076 28.8 28.8 27.9 33621 46 0.0021 27.1 27.7 26.2 33624 66 0.003 22.5 23.3 22.2 33625 49 0.0022 23.9 24.6 23.3 33626 69 0.0031 24.0 24.2 23.2 33629 26 0.0012 26.4 27.4 24.7 33634 53 0.0024 26.3 27.0 25.2 33635 46 0.0021 24.8 25.2 24.0 33637 43 0.0019 27.3 27.8 26.2 33647 193 0.0086 26.1 26.7 25.1

PAGE 110

95 The total study area was divided into relatively high and low environmental exposure category (stratum) areas. Relatively high and low exposure areas of ambient air pollution by particulate matters with a diam eter of less or equal to 10 microns (PM10), sulfur dioxide (SO2) and ozone (O3) were defined by the 50th percentile of total concentrations distribution as an average cu t-off point value. Di stribution of ambient particles pollution over the study area in 1997, 1998, and 1999 is illustrated below in Pictures 4-6. Picture 4. PM10 concentrations, microg/m3 in Tampa Bay, FL, in 1997

PAGE 111

96 Picture 5. PM10 concentrations, microg/m3, in Tampa Bay, FL, in 1998 Picture 6. Interpolated PM10 concentrations, microg/m3, in Tampa Bay, FL, in 1999

PAGE 112

97 The average annual concentrations for pa rticulate matter by geographical area of residence varied from mini mum concentration of 22 g/m3 to maximum value of 28 g/m3 over the study area in Hillsborough, FL, in 1999 (see Picture 7). Picture 7. PM10 concentrations by postal zip co de area of residence, microg/m3, in Tampa Bay, FL, in 1999 Relatively low average annual environmental exposure to particulate matters with a diameter of less or equal to 10 microns (PM10) areas included postal zip code geographical areas of residence with an aver age annual concentrati on of equal or more than 22 g/m3 but less than 25 g/m3 (see Picture 8).

PAGE 113

98 Picture 8. Relatively high and low average annual environmental expos ure to particulate matter category (stratum) area s in Hillsborough County, in 1999 Relatively high environmental exposure to particulate matters areas included and covered postal zip code areas of residence with average annual concentration of equal to or more than 25 g/m3 and less than or equal to 28 g/m3 respectively. Crude rate ratio of adults and children asthma hospitalizations was respectively 1.5 and 1.8 times higher in the high environmental exposure to ambient particulate matter category (stratum) areas (see Tables 24 and 25). Pearson correlation analysis revealed str ong significant association between average coarse particulate matter pollu tion and crude asthma hospitali zation rates by zip code area of residence for both adults (correla tion coefficient r=0.48, p<0.001) and children (correlation coefficient r=0.53, p<0.001) population groups in 1999.

PAGE 114

99 Table 24 High and low environmental exposure to coarse particulate matter categories (strata) and adult asthma hospital admissi ons within separate category (stratum) Environmental exposure categories Adult asthma hospitalization Total adult population Rate Ratio (95% CI) Crude Rate per 10,000 High environmental exposure to PM10 (25-28 g/m3) 659 486,254 RR=1.5 (1.33-1.77) 13.55 Low environmental exposure to PM10 (22-25 g/m3) 272 308,104 RR=1 8.83 Total 931 794,358 Table 25 High and low environmental exposure to coarse particulate matter categories (strata) and children asthma hospital admissions within separate category (stratum) Environmental exposure categories Children asthma hospitalization Total children population Rate Ratio (95% CI) Crude Rate per 10,000 High environmental exposure to PM10 (25-28 g/m3) 591 136,577 RR=1.8 (1.5-2.1) 43.27 Low environmental exposure to PM10 (22-25 g/m3) 190 78,920 RR=1 24.08 Total 781 215,497 Single non-linear regression analysis results supported our frequency table calculations. Poisson log-linear regression analysis revealed that environmental exposure to ambient particles is a significant factor to explain the increase in hospital admissions for adult (p<0.001) and childhood asthma.( p<0.001). Similar trend of significant association between asthma hospitalizations and ambient particles was shown in the comparative analysis of data by geographi cal area of residence in 1997 and 1998. There was strong statistically significant asso ciation between ambient particles and hospitalizations for child hood (r=0.53, p<0.001) and adults asthma (r=0.59, p<0.001) in 1997. Correspondingly, there was also shown signi ficant relationship between particulate matter and hospital admissions for ch ildhood (r=0.45, p<0.001), and adult asthma (r=0.55, p<0.001).

PAGE 115

100 Sulfur dioxide (SO 2 ) The average annual sulfur dioxide concentr ations by separate ambient air quality monitoring site were used as initial databa se for spatial interpolation. Initial average values for ambient air pollution by sulfur dioxide at separate ambient air quality monitoring sites are presented in Table 26. Table 26. List of ambient air quality monitoring stations for SO2, ppb, and average annual values in 1997, 1998, and 1999 Site No. Site Name Latitude Longitude SO2, ppb, 1997 SO2, ppb, 1998 SO2, ppb, 1999 057-0021 27.791611 -82.360777 4 4 3 057-0053 Interbay 27.885300 -82.481700 5 5 5 057-0081 Simmons Park 27.742800 -82.469200 5 4 5 057-0095 Gannon 27.921100 -82.401400 5 5 5 057-0109 Tampa 27.854400 -82.383600 6 6 6 057-1035 Davis Island 27.909200 -82.455300 8 8 8 057-4004 One Raider 27.990800 -82.125800 3 3 081-3002 Port Manatee 27.631333 -82.546083 5 5 4 103-0023 St. Pete 27.862000 -82.623361 6 6 7 103-3002 Pinellas Park 27.869805 -82.691750 3 3 4 103-5002 Tarpon Springs 28.088600 -82.701100 3 3 3 103-5003 Tarpon Springs 28.140200 -82.740000 2 3 105-0010 Nichols 27.855388 -82.017722 6 6 7 105-2006 High School 27.895527 -81.960166 4 5 4 Adjacent to main study area ambient air pollution monitoring sites located in Pinellas, Polk and Manatee Counties, FL, we re included into the spatial interpolation model. Sites 105-0010 and 105-2006 were locate d in Polk County, FL, and Site 081-3002 was located in Manatee County, FL. Spatial interpolation and information techni ques were used to obtain average annual environmental exposure to ambient air polluti on by sulfur dioxide concentration values by zip code area of residence in Hillsborough County, FL, in 1997, 1998, and 1999. Spatial interpolation results are presented in Table 27.

PAGE 116

101 Table 27. Interpolated sulfur dioxide concentr ation values, ppb, by zip code area of residence in Hillsborough County, FL, 1997, 1998, and 1999 Zip Code Count No. Area, ft2 Conc., ppb, 1997 Conc., ppb, 1998 Conc., ppb, 1999 33510 43 0.00195.45.25.2 33511 88 0.00395.45.25.1 33527 162 0.00725.34.44.4 33534 70 0.00315.15.04.7 33547 1086 0.04855.25.15.1 33549 312 0.01395.14.64.8 33556 228 0.01024.23.74 33565 569 0.02545.24.24.2 33566 118 0.00535.13.53.4 33567 324 0.01455.34.14.1 33569 314 0.0145.15.04.7 33570 293 0.01315.04.54.6 33572 29 0.00134.84.64.4 33573 94 0.00424.74.64.2 33584 111 0.0055.45.05 33592 273 0.01225.34.84.9 33594 119 0.00535.34.94.8 33598 283 0.01275.04.84.6 33602 14 0.00066.56.46.3 33603 21 0.00095.95.85.9 33604 46 0.00215.75.55.6 33605 44 0.0026.06.06 33606 22 0.0017.07.07 33607 55 0.00255.85.75.7 33609 25 0.00115.95.85.9 33610 92 0.00415.65.45.4 33611 29 0.00135.55.45.5 33612 54 0.00245.55.25.3 33613 45 0.0025.35.05.2 33614 52 0.00235.65.45.5 33615 52 0.00235.04.75 33616 23 0.0015.55.45.5 33617 49 0.00225.55.35.4 33618 39 0.00175.34.95.1 33619 171 0.00765.45.45.4 33621 46 0.00215.55.35.5 33624 66 0.0035.14.75 33625 49 0.00224.94.54.7 33626 69 0.00314.23.84.1 33629 26 0.00125.95.95.9 33634 53 0.00245.35.15.3 33635 46 0.00214.54.24.5 33637 43 0.00195.55.25.3 33647 193 0.00865.34.95.1

PAGE 117

102 Simple and stratified frequency table analyses included only the information collected over the period of 1999. Distri bution of interpolated sulfur dioxide concentrations over study area is demonstrated in Picture 9. Th e total study area was used to define relatively high and low envir onmental exposure categories (strata) by environmental exposure to sulfur dioxid e (see Picture 10). The average annual concentration for sulfur dioxide (SO2) varied from a minimum concentration of 3.4 ppb in the zip code area 33566 to a maximum va lue of 7 ppb in the zip code area 33606 in Hillsborough, FL, in 1999. Relatively low average annual environmental exposure to sulfur dioxide areas included postal zip code areas with av erage annual concentration of equal to or more than 3.4 ppb but less than 5.2 ppb. Picture 9. Interpolated SO2 concentrations, ppb, in Hillsborough County, in 1999 Relatively high environmental exposure ar eas covered geographical areas with average annual concentrations equal to or more than 5.2 ppb and less or equal to 7 ppb respectfully.

PAGE 118

103 Picture 10. Relatively high and low average annual environmental exposure to sulfur dioxide category (stratum) areas in Hillsborough County, in 1999 Frequency table analysis of separate e nvironmental exposure to sulfur dioxide categories revealed that the crude rate ra tio was 1.4 and 2.4 times higher in the high environmental exposure to sulfur dioxide category areas as compared to the low environmental exposure stratum for adults and children population groups respectively (see Table 28 and Table 29). Table 28. Separate high and low environmental exposure to sulfur dioxide categories (strata) and adult asthma hospital admissions within different stratum Environmental exposure categories Adult asthma hospitalization Total adult population Rate Ratio (95% CI) Crude Rate per 10,000 High exposure to SO2 category (5.2 – 7 ppb) 607 456515 RR=1.39 (1.21-1.59) 15.51 Low exposure to SO2 category (3.4 – 5.2 ppb) 324 337519 RR=1 9.32 Total 931 794,358

PAGE 119

104 Table 29. Separate high and low environmental exposure to sulfur dioxide categories (strata) and adult asthma hospital admissions within different stratum Environmental exposure categories Adult asthma hospitalization Total children population Rate Ratio (95% CI) Crude Rate per 10,000 High exposure to SO2 category (5.2 – 7 ppb) 511 95,070 RR=2.4 (2.08-2.79) 55.5 Low exposure to SO2 category (3.4 – 5.2 ppb) 270 120,427 RR=1 22.42 Total 781 215,497 The simple correlation analysis disclo sed significant correlation between environmental exposure to sulfur dioxide and both adult (correlation coefficient r=0.36 and corresponding p-value p<0.05) and child ren (r=0.6 with p<0.001) hospital admission for asthma by geographical area of reside nce in 1999. In addition, there was a strong statistically significant associ ation between ambient air pollution by sulfur dioxide and hospital admissions for childhood (r=0.59, p< 0.001) and adult asthma (r=0.56, p<0.001) shown in 1997. Correspondingly, there was sign ificant association revealed between sulfur dioxide and childhood asthma (r=0.46, p<0.001) and adult asthma (r=0.37, p<0.001) hospital admission by geographical ar ea of residence in Hillsborough County, FL, in 1998. Ozone (O 3 ) The average annual ozone concentrations measured by separate ambient air quality monitoring sites were used as an initial data base for spatial interpolation. The new 8-hour standard was used to calculate average annual environmental exposure to ozone. The NAAQS 8-hour 0.08 ppm standard is attained by having 3-year average of the annual four highest daily maximum values. Initial average values for ambient air pollution by ozone at different ambient air quality m onitoring sites are presented in Table 30.

PAGE 120

105 Table 30. List of ozone ambient air quality monitoring stations and average annual values in 1999 Site No. Site Name Latitude Longitude O3, ppb, 1997 O3, ppb, 1998 O3, ppb, 1999 057-0081 Simmons Park 27.738555 -82.465583 88 102 95 057-1035 Davis Island 27.926472 -82.454833 85 96 87 057-1065 Gandy Blvd. 27.890833 -82.538583 87 98 91 057-4004 Plant City 27.990944 -82.125750 NA 94 85 081-3002 Palmetto 27.631333 -82.546083 81 99 85 081-4010 Comm. College 27.441055 -82.596638 78 99 52 081-4012 G.T. Bray Site 27.479055 -82.618833 NA NA 94 081-4013 Bradenton 27.447750 -82.522166 NA NA 88 101-2001 Holiday 28.193444 -82.757972 81 94 87 103-0004 Clearwater 27.969166 -82.736222 78 94 94 103-0018 St. Petersburg 27.784111 -82.740027 74 98 94 103-5002 Tarpon Springs 28.088500 -82.700972 77 91 86 105-6005 Lakeland 27.938166 -82.000305 83 92 79 105-6006 Lakeland2 28.027699 -81.972166 82 91 82 In addition to adjacent ambient air qualit y monitoring sites in Polk (Sites 105-6005 and 105-6006) and Manatee (Sites 081-3002, 081-4010, 081-4012 and 081-4013), a monitoring site located in Pasco County, FL, (Site 101-2001) was also used to create the initial dataset of limited point measurements of environmental exposure to ozone for further spatial interpolation. Spatial interp olation results are pr esented in Table 31. The year 1999 was selected to conduct s ubsequent simple and stratified frequency table analyses. Interpolated ozone concentration values ar e presented in Picture 11. The average annual concentration for ozone (O3) varied from minimu m concentration of 84.1 ppb to maximum value of 91.6 ppb, with an average of 88.0 and standard deviation SD=1.7, by separate zip code area of reside nce over the study area in Hillsborough, FL, in 1999. Relatively low average annual environmen tal exposure to ambient ozone category areas included geographical ar eas of residence with an av erage annual concentration of equal or more than 84 ppb but le ss than 88 ppb (see Picture 12).

PAGE 121

106 Table 31. Interpolated environmental exposure to ozone, ppb, by separate zip code area of residence in Hillsborough County, FL, 1997, 1998, and 1999 Zip Code Count No. Area, ft2 Conc., ppb, 1997 Conc., ppb, 1998 Conc., ppb, 1999 33510 43 0.001983.6 95.6 87.2 33511 88 0.003983.7 95.9 87.5 33527 162 0.007283.0 94.4 85.4 33534 70 0.003184.8 97.9 89.5 33547 1086 0.048582.9 95.2 85.0 33549 312 0.013981.7 95.1 88.2 33556 228 0.010279.8 93.7 87.9 33565 569 0.025482.5 93.8 84.6 33566 118 0.005382.6 93.5 84.2 33567 324 0.014582.8 93.7 84.2 33569 314 0.01484.0 96.7 88.0 33570 293 0.013185.4 100.2 90.9 33572 29 0.001386.0 99.8 91.5 33573 94 0.004284.8 98.9 89.5 33584 111 0.00583.3 95.3 86.7 33592 273 0.012282.7 94.9 86.5 33594 119 0.005383.3 95.1 86.4 33598 283 0.012783.8 97.8 87.7 33602 14 0.000684.9 96.1 87.5 33603 21 0.000984.5 96.3 88.1 33604 46 0.002183.9 96.1 88.3 33605 44 0.00284.7 96.2 87.6 33606 22 0.00185.0 96.1 87.3 33607 55 0.002584.5 96.5 89.0 33609 25 0.001185.1 96.8 89.0 33610 92 0.004183.9 96.0 87.8 33611 29 0.001386.1 97.5 90.0 33612 54 0.002483.3 95.9 88.4 33613 45 0.00282.7 95.6 88.4 33614 52 0.002383.7 96.2 88.8 33615 52 0.002381.9 95.5 89.3 33616 23 0.00186.4 97.8 90.5 33617 49 0.002283.5 95.9 88.0 33618 39 0.001782.4 95.5 88.6 33619 171 0.007684.4 96.3 88.0 33621 46 0.002185.6 97.9 90.2 33624 66 0.00382.0 95.3 88.7 33625 49 0.002281.2 94.9 88.8 33626 69 0.003179.6 93.6 88.2 33629 26 0.001285.6 97.1 89.4 33634 53 0.002483.0 96.0 89.2 33635 46 0.002180.2 94.4 89.0 33637 43 0.001983.3 95.7 87.7 33647 193 0.008682.6 95.3 87.5

PAGE 122

107 Picture 11. Interpolated ambient ozone concentrations in Tampa Bay, FL, in 1999 Picture 12. Distribution of ambient air polluti on by zip code area of residence in Hillsborough County, FL, in 1999

PAGE 123

108 Table 32. Separate high and low environmental e xposure to ozone categories (strata) and adult asthma hospital admissions within each stratum Environmental exposure categories Adult asthma hospitalization Total adult population Rate Ratio (95% CI) Crude Rate per 10,000 High environmental exposure to ozone (88-92 ppb) 565 487,247 RR=0.97 (0.85-1.11) 11.6 Low environmental exposure to ozone (84-88 ppb) 366 307,111 RR=1 11.9 Total 931 794,358 Table 33. Separate high and low environmental e xposure to ozone categories (strata) and children asthma hospital admissions within each stratum Environmental exposure categories Children asthma hospitalization Total children population Rate Ratio (95% CI) Crude Rate per 10,000 High environmental exposure to ozone (88-92 ppb) 470 122,962 RR=1.14 (0.99-1.31) 38.2 Low environmental exposure to ozone (84-88 ppb) 311 92,535 RR=1 33.6 Total 781 215,497 Relatively high environmental exposure areas covered postal zip code of residence with average annual con centration of equal to or more than 88 ppb and less or equal to 92 ppb respectively. Asthma hospital admi ssions and total population within each environmental exposure stratum are presented in Table 32 and Table 33. Crude rate ratio (RR) estimates attributable to high expos ure category areas were RR=0.97 for adult and RR=1.14 for children population groups respec tively. There was no association shown between environmental exposure to ozone a nd both adult and child ren asthma hospital admission rates. The Pearson correlation analys is confirmed that there was no association between environmental exposure to ozone and both adult (correlation coefficient r=-0.1, with corresponding p-value p>0.05) and ch ildren (r=0.05, p>0.05) asthma hospital admission rates by zip code areas of reside nce in Hillsborough County, FL, in 1999. This finding was also supported by separate simple log-linear regression analysis. There was also no significant effect of ozone on hos pital admissions for childhood (r=-0.02, p>0.1) and adult asthma (r=-0.04, p>0.1) shown in 1998. In 1997, there was only a slight

PAGE 124

109 significant crude association discovered betw een ambient ozone and hospitalizations for childhood (r=0.32, p<0.05), but not significant with a dult asthma (r=0.28, p>0.05) accordingly. 2.3.2. Socioeconomic status indicators The possible relationship between socioec onomic status and both asthma hospital admission rates and environmental exposure to criteria ambient air pollutants was evaluated by using standard simple corre lation and frequency ta bles for separate socioeconomic category (strata) analysis tech niques. We have considered 20 social and economic indicators, which, based on previous empirical studies, could be viewed as roughly approximating absolute and distributive aspects of living conditions or socioeconomic disadvantage status in a gi ven community. Selected indicators were drawn from the US Census 2000 and included: (1) percentage of people living below poverty level; (2) pe rcentage of unemployment ; (3) percentage of persons with education level of ninth grade or less education as th e highest degree of school completed; (4) percentage of persons with the average annual family income of or less than $15,000; (5) percentage of persons employed in professiona l, managerial, administ rative, and clerical positions ( white-collar occupation employees); (6) percentage of unskilled persons ; (7) percentage of householder moved during the last year ; (8) percentage of households with overcrowded conditions (over one person pe r room); (9) percentage of divorce rate; (10) percentage of single parents with children ; (11) percentage of elderly population of 65 years and older; (12) percentage of children population of 5 years and younger; (13) percentage of black person; (14) percentage of ethnic minority as a head of household ; (15) percentage of houses heating by fuel oil, kerosene, gasoline and other combustible liquids; (16) percentage of houses heating by wood ; (17) percentage of houses heating by gas ; (18) percentage of houses built in 1960 or before ; (19) percentage of houses with lacking kitchen and/or plumbing facilities ; (20) percentage of families with no access to vehicle (no automobile ownership). Distribution of the percenta ge estimates for selected socioeconomic variables by zip code areas of residence is presented in Appendix E.

PAGE 125

110 Table 34. Calculated UPA and UDI index values and crude rates of hospital admissions for adult and childhood asthma per 10,000 populati on by zip code area of residence in Hillsborough County, FL, 1999 Zip code UPA UDI Adult asthma hospitalizations/ 10,000 Childhood asthma hospitalizations/ 10,000 33510 58.7 15.0 9.21 26.03 33511 58.7 14.0 7.72 20.12 33527 58.1 20.6 10.43 24.96 33534 69.4 21.1 18.11 20.25 33547 77.1 16.2 7.77 9.55 33549 59.1 12.7 5.11 16.93 33556 51.2 10.8 6.43 29.00 33565 42.7 18.0 11.31 14.08 33566 62.7 20.7 17.15 32.56 33567 79.9 18.4 9.23 14.01 33569 68.9 15.9 11.32 25.30 33570 63.0 20.8 12.64 23.32 33572 74.7 15.3 1.57 9.18 33573 52.2 17.5 3.69 0.00 33584 67.2 17.4 14.60 10.57 33592 62.4 19.3 20.69 35.75 33594 70.6 12.7 7.45 20.93 33598 49.7 24.0 12.69 35.99 33602 85.1 19.7 36.68 133.90 33603 84.4 20.6 22.19 61.38 33604 80.7 20.6 18.50 59.95 33605 79.8 25.9 22.50 119.22 33606 103.8 10.9 10.06 48.95 33607 50.1 22.4 20.99 80.83 33609 89.5 14.0 11.84 41.23 33610 56.2 25.8 21.72 81.27 33611 107.0 15.2 13.60 47.52 33612 60.3 20.5 18.98 41.99 33613 83.4 17.2 13.54 41.51 33614 74.6 20.9 17.50 54.89 33615 85.2 16.2 13.81 43.48 33616 65.0 18.5 13.78 50.35 33617 73.7 16.1 6.89 35.86 33618 70.8 12.0 5.49 15.15 33619 49.0 22.0 11.48 44.92 33621 84.6 16.0 5.82 10.31 33624 78.7 12.7 7.00 17.11 33625 51.6 14.1 4.44 13.99 33626 54.6 9.8 9.56 21.83 33629 52.4 10.3 3.72 27.28 33634 43.3 17.1 6.57 37.22 33635 65.2 13.5 9.22 37.34 33637 54.8 15.2 16.33 21.93 33647 67.0 7.9 0.00 9.48 The Urban Deprivation Index (UDI) and Underprivileged Area (UPA) scores were calculated for separate zip code area of residence as complex multidimensional

PAGE 126

111 socioeconomic status characteristics of a given area. (see Table 34) The simple correlation analysis was conducted to eval uate the associatio n between selected community socioeconomic status characterist ics and both children and adult crude rates of hospital admissions for asthma as a primary diagnosis by geographical area of residence in 1999 (Table 35). Only statistically significant socioecono mic variables represented socioeconomic status characteristics of primary interest a nd were selected for s ubsequent stratified and multiple regression analyses to evaluate th e confounding effect of socioeconomic status on the association between environmental e xposure to ambient air pollution and hospital admissions for asthma. The percentage values of elderly (65 years and over); children younger than 5 years; house heating by ga s; and house heating by wood were not associated with asthma hospital admissions of children or adult population groups respectively. Table 35. List of significant socioeconomic status variables along with corresponding correlation coefficient and p-values Area socioeconomic deprivation indicators Adult asthma hospitalizations/ 10,000 adult population Children asthma hospitalizations/ 10,000 children population Pearson correlation co efficient (p-value) Poverty, % 0.82 (p<0.001) 0.79 (p<0.001) Education 9th grade or less, % 0.46 (p<0.05) 0.37 (p<0.05) Unemployment, % 0.54 (p<0.05) 0.60 (p<0.001) Unskilled persons, % 0.51 (p<0.05) 0.25 (p>0.05) White-collar occupation, % -0.48 (p<0.05) -0.24 (p>0.05) Usage of public transportation, % 0.76 (p<0.001) 0.91 (p<0.001) No access to vehicle, % 0.77 (p<0.001) 0.87 (p<0.001) Family Income of $15000 or less, % 0.78 (p<0.001) 0.79 (p<0.001) Age of house (40 yr and older) % 0.43 (p<0.05) 0.57 (p<0.001) House heating by fuel, % 0.47 (p<0.05) 0.62 (p<0.001) Lacking facilities, % 0.44 (p<0.05) 0.58 (p<0.001) Overcrowding conditions, % 0.42 (p<0.05) 0.39 (p<0.05) Divorced persons, % 0.66 (p<0.001) 0.63 (p<0.001) Single parent with children, % 0.69 (p<0.001) 0.67 (p<0.001) Ethnic minority householder, % 0.63 (p<0.001) 0.78 (p<0.001) Black, % 0.65 (p<0.001) 0.81 (p<0.001)

PAGE 127

112 The Urban Deprivation Index (UDI) and Underprivileged Area Index (UPA) represented complex standard socioeconomic st atus indices of residential areas and were subsequent used to stratify relatively high and low socioeconomic st atus frequency table analysis to explore possible confounding eff ect and effect modification (interaction effect) between small-area socioeconomic stat us characteristics and local environmental exposure to ambient air pollution on hospita l admissions for childhood and adult asthma. The analysis of complex area socioeconomic i ndices and asthma hosp italizations revealed a strong significant association between the Urban Deprivation Index (UDI) and hospital admissions for childhood and adult asthma. There was strong significant associat ion shown between UDI and hospital admissions for childhood (correlation coe fficient r=0.58, p<0.001) and adult asthma (r=0.64, p<0.001) by zip code area of residence in Hillsborough County, FL, in 1997. In addition, there was strong significant a ssociation with chil dhood (r=0.59, p<0.001) and adult asthma (r=0.69, p<0.001) disclosed in 1998. Significant positive correlation coefficient r=0.68 with p<0.0001 was shown in adult, and r=0.53 with corresponding p<0.001 in children asthmatics by geographical area of residence in 1999. There was only slight non-significant positive association (p>0.05) between Underprivileged Area Index (UPA) and hospital admission rates for both childhood and adult asthma in 1997, 1998, and 1999. Simple correlation analysis revealed signi ficant association between crude asthma hospital admission rates and such socioeconomic status indicators as poverty, education, unemployment, white-collar o ccupation, daily utilization of public transportation, annual family income of $15,000 or less, age of house, house heating by fuel, lacking kitchen and/or plumbing facilities, overcrowding c onditions, single parent with a child or children, ethnic minority householder. Comparat ive analysis revealed that all of the aforementioned small-area socioeconomic status indicators were statistically significantly associated with hospitalizations for bot h childhood and adult asthma in 1997 and 1998. Percentage of unskilled persons, black persons divorced persons and also persons with no access to vehicle were excluded from subse quent analysis because they were highly correlated with and represented by other so cioeconomic status variables. Unskilled

PAGE 128

113 persons were represented by persons with lo w high school education and were inversely related with persons employed in professiona l or managerial occ upations (white collar occupations). Percentage of black persons was represented by the percentage of ethnic minority householders in the given area. Per centage of divorced persons was reflected by the percentage of single parent families with children. Finally, percentage of persons with no access to vehicle was represented by the percentage of people using public transportation on daily basis. Because of th e limited distribution range of variables and limited number of asthma hospital admissions within defined socioeconomic strata, the percentages of unemployed people, ninth grade or less high school education, people daily utilizing public transportation, house heated by fu el, houses with lacking kitchen and/or plumbing amenities, and ethnic minor ity householders were not used in the stratified analysis. However, all of the aforementioned si gnificant area socioeconomic status characteristics of primary re search interest were used in the Principal Component Analysis (PCA) to identify the most influential area socioeconomic status indicators out of the total number of significant variables and to develop a complex and multidimensional small-area socioeconomic deprivat ion index. Percentages of residents living below the poverty level, with annual family income of $15,000 or less, employed in white-collar occupations, livi ng in overcrowding conditions (one person or more per room in the house), or in older houses (age of house 40 years and older), and also the percentage of single parents w ith children were used to de fine different socioeconomic status categories (strata) and to calculate rate ratios of hospital admissions for asthma attributable to lower socioeconomic deprivati on status as compare to the reference total crude association between hos pital admission rates for childhood and adult asthma and environmental exposure to both ambien t particles and sulfur dioxide. Poverty Poverty status indicated th e percent of people living be low poverty level in the given area and was defined into low and hi gh socioeconomic status categories. The 50th percentile value was used as an arithmetic average cut-off point value to define two different socioeconomic strata by poverty. The number of separate zip code areas with a

PAGE 129

114 corresponding number of asthma hospital admission for both childhood and adult asthma and total population data within different socioeconomic and environmental exposure strata in Hillsborough County, FL, in 1999, is presented in Appendix F (see also Picture 13). Picture 13. Percentage of people livi ng below poverty level by zi p code area of residence in Hillsborough County, FL, in 1999 Low socioeconomic status areas also had higher environmental exposure to ambient air pollution. Statistical data analysis revealed that neither 1-hr (NAAQS=0.12 ppm) or 8hr (NAAQS=0.08 ppm) based on highest daily maximum values of ambient ozone pollution by zip code area of residence were not associated with childhood and adult asthma hospitalizations (see Appendix F). Th erefore further stratified analyses were conducted only by adjusting for ambient part icles and sulfur dioxide exposure by zip code area of residence within separate soci oeconomic status strata (or category). 8 of 10 zip code areas of residence with low socioeconomic status (explained by higher

PAGE 130

115 percentage of people living under the federal poverty level) also had relatively higher environmental pollution to ambient particle s (see Appendix F). Only 2 out of 10 low socioeconomic status residential areas had relatively low exposure to ambient particles. For sulfur dioxide, 8 out total 10 areas with higher percenta ge of people living under the poverty level had also rela tively higher con centration of SO2. Low socioeconomic status by poverty was signifi cantly associated with higher ambient air pollution by PM10 (correlation coefficien t r= 0.45, p<0.05) and SO2 (r=0.32, p<0.05) by separate area of residence in Hillsborough County, in 1999. The percentage of persons living below pove rty attributable in the given zip code area as a permanent residence area of study subjects who were admitted to the hospitals because of acute asthma attacks varied from 1.7 to 28.3 percent. The average mean value was 9.86 with standard error SE=1.06. The 50th percent percentile served as a cut-off point value and was equal to 15 percent. Ther efore the low socioeconomic status category (strata) included all postal zip code areas with corresponding poverty st atus of more than or equal to 15 percent and less or equal to 28.3 percent of people living below poverty level in the given area. The high socioeconomic status categ ory (strata) included all zip code areas with relevant povert y status equal to or more th an 1.7 percent but less than 15 percent of population be low poverty level. Table 36. Rate ratio (RR) and corresponding 95% c onfidence intervals (95% CI) of adult asthma hospital admissions in low and high socioeconomic status categories by poverty (SES – socioeconomic status; high so cioeconomic status by poverty used as a reference) SES by poverty Adult asthma hospitalization Total adult population Rate Ratio (95% CI) Low SES by poverty (1528.3 % persons living below poverty) 339 172,824 RR=2.06 (1.8-2.35) High SES by poverty (1.7 – 15 % persons living below poverty) 592 621,534 RR=1 (reference)* Total 931 794,358 Childhood and adult asthma hospital admissi ons were calculated separately along with corresponding total children and adults population subgroups within each

PAGE 131

116 socioeconomic status category by poverty (Table 36 and Table 37). The rate of asthma hospital admission ratio estimates and corresponding 95% confid ence intervals were used to evaluate the magnitude of the difference between two different socioeconomic status strata. Rate ratio of child hood and adult asthma hospitaliza tions was respectively 2.4 and 2.1 times higher in the lower socioeconomic status category as compared to the high socioeconomic deprivation status stratum by percentage of people below poverty level living in the given socioeconom ic status category (stratum). The Pearson simple correlation analysis revealed a strong significant association between poverty level and cr ude asthma hospitalization rates by zip code area of residence for both adults (correlation coefficient r=0.82, p<0.0001) and children (correlation coefficient r=0.8, p<0.0001) groups in 1999. Table 37 Rate ratio (RR) and corresponding 95% confidence intervals (95% CI) of childhood asthma hospital admissions in low and high socioeconomic status categories by poverty (SES – socioeconomic status; high socioeconomic st atus by poverty used as a reference) SES by poverty Childhood asthma hospitalization Total children population Rate Ratio (95% CI) Low SES by poverty (1528.3 % persons living below poverty) 343 52,734 RR=2.4 (2.08-2.76) High SES by poverty (1.7 – 15 % persons living below poverty) 438 161,982 RR=1 (reference)* Total 78 215,497 There was also a strong association betw een asthma hospitalizations and poverty level in 1997 and 1998. In 1997, the percentage of people living below poverty level was significantly associated w ith hospitalizations for ch ildhood (r=0.76, p<0.0001) and adult asthma (r=0.82, p<0.0001). In 1998, poverty was s hown to be strongly correlated with childhood (r=0.75, p<0.0001) and adult asthma (r=0.8, p<0.0001) accordingly. The crude effect represented an unadjusted reference association between e nvironmental exposure and hospital admissions for childhood and a dult asthma. Adjusted to socioeconomic status by poverty rate ratio estimates with corresponding 95 percen t confidence intervals (95% CI) within each socio economic stratum are presented for both children and adult

PAGE 132

117 separately in Appendix E. Slight effect modification and heter ogeneity (interaction effect) by poverty status on the effect of e nvironmental exposure to coarse particulate matters was indicated in the children’s group. A positive confounding effect was observed by poverty status on the association between PM10 and asthma hospital admissions for adults. The same directi on confounding effect by poverty status and overestimated crude risk as compared to ad justed relative risk wa s also observed by the effect of sulfur dioxide on asthma hospitaliz ation within separate socioeconomic status strata by poverty (see Appendix F). Total family income of $15,000 or less per year The percentage of persons with a total annual family income of $15,000 or less in the given area was used to represent socioec onomic deprivation of residential area by family income status. The mean value of fo r the percentage of pe rsons with the total annual family income of $15,000 or less was 14.7 6 percent with a corresponding standard error value of SE=1.19. The socioeconomic st atus by income level varied from 3.1 to 39.7 percent over the study area. Different fam ily income status categories were defined by the 50th percentile estimate as a cut-off point va lue to divide the total study area into low and high socioeconomic status by family income strata. Based on descriptive data analysis of general data distribution, the 50th percentile value wa s 21.4 percent. The low socioeconomic status category included reside nce areas with equal to or higher than 21.4 but less or equal to 39.7 percent of persons w ith annual average family income of less or equal to $15,000 or less. Respectively, the high area socioeconomic status by income level category was defined as equal to or higher than 3.1 but less than 21.4 percent of persons with annual family income of $15,000. Bo th children and adults asthma hospital admissions along with corresponding total ch ildren and adult population subgroups were calculated within each socioeconomic status category by income level (Table 38 and Table 39). Calculated rate ratio of adults and chil dren asthma hospitalizations was 2 and 2.3 times higher in the lower socioeconomic status category area as compared to the stratified

PAGE 133

118 high socioeconomic deprivation status category area by percen tage of persons with the total family income of $15,000 or less per year. Table 38. Rate ratio and corresponding 95% conf idence intervals (95% CI) of adult asthma hospital admissions in low and high socioeconomic status categories by family income (SES – socioeconomic status; high socioeconomic status by the percentage of persons with annual family income of $15,000 and less used as a reference) SES by family income Adult asthma hospitalization Total adult population Rate Ratio (95% CI) Low SES by the percentage of persons with annual family income of $15,000 and less (21.4 – 39.7 %) 292 146296 RR=2.02 (1.76-2.33) High SES by the percentage of persons with annual family income of $15,000 and less (3.1 21.4 %) 639 648089 RR=1* Total 931 794,358 Table 39 Rate ratio (RR) and corresponding 95% confidence intervals (95% CI) of childhood asthma hospital admissions in low and high socioeconomic status categories by family income (SES – socioeconomic st atus; -high socioeconomic status by the percentage of persons with annual family in come of $15,000 and less used as a reference) SES by family income Children asthma hospitalization Total children population Rate Ratio (95% CI) Low SES by percentage of persons with annual family income of $15,000 and less (21.4 – 39.7 %) 283 42281 RR=2.33 (2.01-2.69) High SES by percentage of persons with annual family income of $15,000 and less (3.1 21.4 %) 498 173216 RR=1* Total 781 215,497 Pearson correlation analysis revealed a strong significant association between income level and crude asthma hospitalization rates by zip code area of residence for both adult (correlation coefficient r=0.78, p<0.0001) and children (correlation coefficient r=0.79, p<0.0001) population groups in 1999. Th ere was strong association between asthma hospitalizations and pe rcentage of people with tota l annual family income equal to or less than $15,000 shown in 1997 and 1998. Th e percentage of people with family income equal or less than $15,000 was significant ly associated with hospitalizations for childhood (r=0.8, p<0.0001) and adult asthma (r=0.84, p<0.0001) in 1997. Accordingly,

PAGE 134

119 family income was also shown to be strongly correlated with childhood (r=0.77, p<0.0001) and adult asthma (r=0.76, p<0.0001) in 1998. The association between environmental e xposure to both coarse particulate matter and sulfur dioxide and asthma hospitalizations within separate socioeconomic categories (strata) by family income, defined as the per centage of people with average annual family income of equal to or less than $15,000, is pr esented separately for children and adult groups in Appendix F. 7 of 44 zip code areas of residence were under the low socioeconomic status category by lower family income. 6 of 7 residence areas with low socioeconomic status by family income had relatively higher environmental exposure to ambient particles and sulfur dioxide. Low so cioeconomic status areas were strongly associated with higher ambient air concentration of PM10 (correlation coefficient r= 0.42, p<0.05) and SO2 (r=0.38, p<0.05) by zip code area of residence over the total area of study. Slight socioeconomic confounding eff ect by family income was confirmed by decrease in adjusted relative risk estimates of the a ssociation between environmental exposure to ambient coarse particulate matters and asthma hospitalizations for children as compared to the total crude estimate. More noticeable socioeconomic confounding effects were shown on the asso ciation between environmental exposure to sulfur dioxide and asthma hospital admissions for both ch ildren and adult population groups within separate socioeconomic categories (strata) by poverty status. Soci oeconomic confounding effect was homogeneous in each category and there was no interaction discovered. Although there was a slight heterogeneity discovered of the association between environmental exposure to particulate matters and asthma hospitalizations for adults within different socioeconomic strata by poverty status, the hom ogeneous effect of environmental exposure within different so cioeconomic strata was assumed. Positive confounding effect indicated that the strength and magnitude of the cr ude relative risk estimates and also crude (una djusted) effects of envir onmental exposure to ambient coarse particulate matters and sulfur di oxide was overestimated (see Appendix F).

PAGE 135

120 White-collar occupation White collar occupation described the percentage of popul ation employed in professional and managerial positions in di fferent zip code areas of residence in Hillsborough County, FL. The average mean va lue for the percentage of white-collar professionals was 32.2 percent and varied fr om 14.4 to 57.7 percent over the total study area in Hillsborough County, FL, in 1999. Picture 14. Percentage of people em ployed in white-collar occ upations by zip code area of residence. Separate socioeconomic status geographical areas of the percentage of white-collar employees were grouped into two socioeconom ic status categories (strata) by using 50th percentile as an average cut-off value. Based on descriptive statistical data analysis, the 50th percentile value was 36.1 pe rcent of population employed in white-collar occupation within the given zip code area of residence (see Picture 14). The low socioeconomic

PAGE 136

121 status category (stratum) by the percentage of white-c ollar occupation employees included all zip code areas of residence with equal to or more than 14.5 and less than 36.1 percent of population employed in professional or manageri al position. Respectively, the high socioeconomic status by white-collar occupation category included geographical areas with equal to or more than 36.1 and less than or equal to 57.7 percent of total population employed in white-collar occupati ons. The numbers of adults and children asthma hospital admissions and total referen ce population within separate socioeconomic status by white-collar occ upation categories are given in Table 40 and Table 41. Table 40 Rate ratio (RR) and corresponding 95% c onfidence intervals (95% CI) of adult asthma hospital admissions in low and high socioeconomic status categories by whitecollar occupation (SES – socioeconomic stat us; -high socioeconomic status by the percentage of population employed in white-c ollar occupations used as a reference) SES by percentage of whitecollar occupation employees Adult asthma hospitalization Total adult population Rate Ratio (95% CI) Low SES by the percent of population employed in whitecollar occupations (14.5–36.1 %) 715 502,510 RR=1.92 (1.65-2.24) High socioeconomic status by the percent of population employed in white-collar occupations (36.1-57.7 %) 216 291,848 RR=1* Total 931 794,358 Table 41 Rate ratio (RR) and corresponding 95% confidence intervals (95% CI) of childhood asthma hospital admissions in low and high socioeconomic status categories by white-collar occupation (SES – socioeconom ic status; high so cioeconomic status by the percentage of population employed in white-collar occupations used as a reference) SES by percentage of whitecollar occupation employees Children asthma hospitalization Total children population Rate Ratio (95% CI) Low SES by the percent of population employed in whitecollar occupations (14.5–36.1 %) 597 137,033 RR=1.86 (158-2.19) High SES by the percent of population employed in whitecollar occupations (36.1-57.7 %) 184 78,464 RR=1* Total 781 215,497

PAGE 137

122 Separate socioeconomic category analysis revealed that the excess relative risk fraction of 86 percent of children and 92 pe rcent of adult asthma hospital admissions could be attributed to the lo w socioeconomic status describe d by the lower percentage of people employed in white-collar occupations. The Pearson correlation analysis of crude asthma hospital admission rates and percenta ge of population employed in white-collar occupations by zip code area of residence revealed a st rong statistically significant inverse association for children (corre lation coefficient r=-0.48, p<0.0001) but no significant negative associati on for adults (correlation coefficient r=-0.24). The comparative studies in 1997 and 1998 disclo sed that employment in a professional occupation was a more important factor for asthmatic adults than children. In 1997, the percentage of people employed in professiona l occupations was asso ciated with hospital admissions for childhood (r=-0.34, p<0.05) and adult asthma (r=-0.41, p<0.05). A significant inverse association was also re vealed between professional occupation and hospitalizations for childhood (r=-0.35, p<0.05) and adult asthma (r=-0.49, p<0.05) in 1998. The employment in white-collar occupatio n as an area socioeconomic deprivation status indicator was further evaluated in the stratified analysis and multivariate regression analysis (modeling) to explore possible c onfounding effect in environmental asthma studies. 28 of 44 postal zip code areas of residence were characterized by low socioeconomic status based on lower percen tage of population employed in professional or managerial occupations (w hite-collar occupati ons). There were 17 and only 13 of 28 low socioeconomic status residential area s with higher environmental exposure to ambient particles and sulfur dioxide resp ectively (see Appendix F). Higher exposure to ambient particles was significantly inversel y correlated with higher number white-collar occupation employees by zip code area of residence in the stud y area (r=-0.31, p<0.05). However, exposure to sulfur di oxide was not associated with socioeconomic status by the percentage of people employed in profe ssional and managerial occupations (r=0.15, p>0.05). The association between environmental exposure to both particulate matter and sulfur dioxide and asthma hospitalizations w ithin separate socioeconomic strata by the percentage of white-collar employees is gi ven separately for children and adult in

PAGE 138

123 Appendix F. Both confounding effect a nd slight effect modificat ion were shown within different socioeconomic strata by professional and managerial occupation as compared to crude (unadjusted) analysis results. A sli ght confounding effect co uld be noticed on the association between particulate matter a nd asthma hospitalization in the children population group. The assumption of no inter action effect on the adult sample study group could be accepted because homogeneity was retained and no significant changes were indicated within separate socioec onomic strata by the percentage of people employed in white-collar occupation in a given category (stratum) areas. The heterogeneous effect of socioeconomic stat us by white-collar occupation employees was detected on the exposure to sulfur dioxide w ith more noticeable effect modification in adults rather than in children population groups. Single parent with children The average mean for the percentage of single parents with children over the total study area was 7.76 percent and varied from zero to 15.2 percent by separate zip code area of residence. The study area was divided into two socioeconomic status categories (strata) by the percentage of single parents with children in the geographical zip code area of permanent residence during the period of study. The 50th percentile estimate was used as an arithmetic average cut-off point valu e to divide the tota l study area into two socioeconomic status categories (strata) by single parent family status. The low socioeconomic status category included zip code areas of residence with equal to or more than 7.6 and less than or equal to 15.2 percen ts of single parents with children population in the given area. The high socioeconomic stat us category area covered zip code areas of residence with corresponding valu es of equal to or more th an zero (no single parent families with children present) and less than 7.6 percent of single parent families with children. The number of hosp ital admissions and referen ce population within each socioeconomic stratum is pres ented separately for adult (T able 42) and children (Table 43) population groups.

PAGE 139

124 Table 42 Rate ratio (RR) and corresponding 95% c onfidence intervals (95% CI) of adult asthma hospital admissions in low and high socioeconomic status categories by single parent with children status (S ES – socioeconomic status; -high socioeconomic status by the percentage of single parent status used as a reference) SES by the percentage single parent living with children Adult asthma hospitalization Total adult population Rate Ratio (95% CI) Low SES by single parent living with children (7.615.2 %) 580 359,344 RR=2 (1.75-2.28) High SES by single parent living with children (0-7.6 %) 351 435,014 RR=1* Total 931 794,358 Table 43. Rate ratio (RR) and corresponding 95% confidence intervals (95% CI) of childhood asthma hospital admissions in low and high socioeconomic status categories by single parent with children status (SES – socioeconomic status; high socioeconomic status by the percentage of single parents used as a reference) SES by the percentage single parent living with children Children asthma hospitalization Total children population Rate Ratio (95% CI) Low socioeconomic status by the percentage of single parent living with children (7.6-15.2 %) 520 102,943 RR=2.18 (1.88-2.53) High socioeconomic status by the percentage of single parent living with children (0-7.6 %) 261 112,554 RR=1* Total 781 215,497 The calculated crude rate ratio of hos pital admissions for childhood and adult asthma show that there was a two-fold hi gher risk of asthma hospital admissions for asthmatic adults in low socioeconomic categor y areas described by a higher percentage of single parent families with children. In addition, there was 2.2 times higher risk (RR=2.18, 95% CI: 1.88-2.53) of asthma hospitaliz ations for children less than 15 years old in the low socioeconomic status categ ory areas represented by higher number of single parent families living with children. The Pearson simple correlation analysis confirmed a strong statistically significant association betw een the percentage of single

PAGE 140

125 parent families with children and crude rates of asthma hospital admissions both for children (correlation coefficient r=0.67, p< 0.0001) and adults (correlation coefficient r=0.69, p<0.0001). A similar significant associ ation was revealed in 1997 and 1998. In 1997, the percentage of single families with ch ildren by geographical area of residence was strongly associated with both childhood (r=0.66, p<0.001) and adult (r=0.76, p<0.001) asthma hospitalizations. Correspondi ngly, single parent family with children was also associated with hospital admi ssions for childhood (r=0.67, p<0.001) and adult asthma (r=0.72, p<0.001) in 1998. The association between environmental expos ure to ambient particulate matter with a diameter of equal or less than 10 microns (PM10) and to sulfur dioxide (SO2) with corresponding asthma hospitaliz ations within separate so cioeconomic strata by the percentage of single parent families with children is presented in Appendix D. 21 of 44 postal zip code areas of resi dence were defined as low socioeconomic status areas by single parent family with children status. Th ere were 17 and 14 low socioeconomic status residential areas which also had higher environmental expos ure to ambient particles and sulfur dioxide respectively. There was a strong significant asso ciation between low socioeconomic status and higher environm ental exposure to both ambient particles (correlation coefficient r=0.42, p<0.05) and sulf ur dioxide (r=0.4, p<0.05) by separate zip code area of residence. The positive confounding effect by the numbe r of single parents with children on the association between environmental exposur e to ambient particulate matters and to sulfur dioxide and asthma hospitalizations fo r both children and adult population groups was found in all socioeconomic status stra ta (see Appendix D). A positive confounding effect by the percentage of single parent fa milies with children in a given category confirms that the true strength and magnitude of crude relative risk as an estimate of the association between environmental exposure to both coarse particulate matters and sulfur dioxide and corresponding asth ma hospitalizations was overestimated. Detected effect had similar (homogeneous) directions in all strata and, therefore, we assumed that homogeneity was retained and no effect modification (interaction effect) on the association was present.

PAGE 141

126 Overcrowded conditions Overcrowded housing conditions were descri bed as a percentage of persons living in the household with more than one o ccupant per room by geographical area of residence. The percentage of people living in the overcrowded conditions varied from 0.2 to 28.1 percent for separate zip code areas of residence and the corresponding average mean value was 6.6 percent (SE=5.13) over the total study area. The 33rd and 66th tertiles were used to divide the study area into sepa rate socioeconomic status categories (strata) as average cut-off point values. Because of the limited number of asthma hospital admissions within socioeconomic status strata, second and th ird categories were combined together and the 33rd percentile was chosen as a cu t-off point to divide the total study area into the low and high socioeconomic status strata. Table 44. Rate ratio (RR) and corresponding 95% c onfidence intervals (95% CI) of adult asthma hospital admissions in low and high socioeconomic status categories by overcrowding housing status (SES – socioecono mic status; high socioeconomic status by the percentage of population living in overcrowding housing status used as a reference) SES by overcrowded housing conditions Adult asthma hospitalization Total adult population Rate Ratio (95% CI) Low SES by the percentage of population living in overcrowded housing conditions (9.5-28.1 %) 299 175,645 RR=1.67 (1.45-1.91) High SES by the percentage of population living in overcrowded housing conditions (0-9.5 %) 632 618,713 RR=1* Total 931 794,358 The calculated number of hospital admissi ons for asthma and total reference population within each stratum is represented separately for children (Table 44) and adults (Table 45). Crude rate ratio estim ates for adults and children asthma hospitalizations were respec tively 1.67 and 1.79 times higher in the lower socioeconomic status category area as compared to the hi gh socioeconomic depriva tion status stratum by percentage of people living in th e overcrowded household conditions.

PAGE 142

127 Table 45. Rate ratio (RR) and corresponding 95% confidence intervals (95% CI) of childhood asthma hospital admissions in low and high socioeconomic status categories by overcrowding housing (SES – socioeconomic status; RRrate ra tio; 95% CI – 95% Confidence Intervals; -high socioeconomi c status by the percentage of population living in overcrowding housi ng used as a reference) SES by overcrowded housing conditions Children asthma hospitalization Total children population Rate Ratio (95% CI) Low SES by the percentage of population living in overcrowded housing (9.5-28.1 %) 280 51,241 RR=1.79 (1.55-2.07) High SES by the percentage of population living in overcrowded housing (0-9.5 %) 501 164,256 RR=1* Total 781 215,497 Pearson correlation analysis revealed a significant correlation between the percentage of people living in overcrowd ed conditions level and crude asthma hospitalization rates by zip code area of residence for both a dults (correlation coefficient r=0.42, p<0.05) and children (corre lation coefficient r=0.39, p<0.05). There was significant association between overcrowded housing conditions and hospitalization for childhood (r=0.37, p<0.05) and adult asthma (p<0.39, p<0.05) by zip code area of residence in Hillsborough County, FL, in 1997. In addition, the overcrowding living condition was also associated with chil dhood (r=0.43, p<0.05) and adult (r=0.44, p<0.05) asthma hospital admissions by geogra phical area of residence in 1998. The association between environmental exposure to both ambient particles and sulfur dioxide (SO2) and corresponding asthma hosp italizations within separate socioeconomic strata defined by the percentage of people living in overcrowded housing conditions was evaluated separately for adu lt and children in Hillsborough County, FL, in 1999 (see Appendix F). 10 of 44 zip code areas of residence were characterized by low socioeconomic status by overcrowded housing conditions. Correspondingly, 8 and 7 residential areas out of total low socio economic status areas also had higher environmental exposure to PM10 and SO2. There was a slight association between low socioeconomic status by overcrowded living conditions and environmental exposure to ambient particles by zip code area of reside nce, however, no statistical significance was shown to prove this association (r=0. 27, p>0.05). There was also no significant

PAGE 143

128 association between socioeconomic status and ambient air pollution by sulfur dioxide revealed (r=0.08, p>0.05). Consistent confoundi ng effect was seen within different socioeconomic strata expressed by overcrow ded conditions for both sulfur dioxide and particulate matters exposure except on the crude relationship between coarse ambient particulate matters and asthma hospitalizati on in children (see A ppendix F). By assuming that effect was retained homogeneous we also accepted that there was no effect modification (interaction) detected. Urban Deprivation Index (UDI) The Urban Deprivation Index (UDI), an example of complex multidimensional area socioeconomic status based on social, econom ic and housing gradients, was calculated by using the standard method explained previ ously (see Methodology). Small-area social, economic and housing gradients were descri bed and incorporated in the complex multidimensional index. The percentage of singl e parents with children and elderly living alone was used to represent the social area gradient. The percentage of persons unemployed and unskilled indicated economic stat us of residence area. Percentage values were obtained from the US Census 2000. The nu mber of unskilled persons was used to calculate the Urban Deprivation Index for each geographic zip code area of residence, but was not used in our study data analysis as a separate socioecono mic status indicator because of high positive co rrelation with the percen tage of people with 9th grade or less high school education (correlation coe fficient r=0.73 and corresponding p-value p<0.0001) and also because of a strong significa nt inverse opposite co rrelation with the percentage of people employed in profe ssional and managerial occupations (r=-0.97, p<0.0001) both of which were retained and explored further in our study. Finally, the percentage of both overcrowded households and households with lacking amenities were used to represent the housi ng gradient in the index.

PAGE 144

129 Picture 15. UDI score by zip code area of re sidence in Hillsborough County, FL, 1999 Calculated UDI values varied from the i ndex score of 7.9 to 25.9 with a mean value of 16.9 (SE=0.64) for separate zi p code areas of residence ov er the total study area. The average arithmetic mean value as a 50th percentile cut point estimate was used to divide the total study area into two socioecono mic status categories (strata) by urban socioeconomic deprivation index score. The low socioeconomic status category (stratum) included all zip code areas of residence with the score of UDI equal or more than 16.9 and less than or equal to 25.9. Accordingly, th e high socioeconomic st atus category by corresponding deprivation index value covered areas with the index score of equal or more than 7.9 and less than 16.9. The numbers of adults and children hospital admissions with the corresponding reference population va lues are presented in Table 46 and Table 47.

PAGE 145

130 Table 46. Rate ratio (RR) and corresponding 95% c onfidence intervals (95% CI) of adult asthma hospital admissions in low and high socioeconomic status categories by Urban Deprivation Index (SES – socioeconomic stat us; -high socioeconomic status by Urban Deprivation Index used as a reference) SES by Urban Deprivation Index (UDI) Adult asthma hospitalization Total adult population Rate Ratio (95% CI) Low SES by UDI score of 15-26 636 404,756 RR=1.97 (1.72-2.26) High SES by UDI score of 8-17 295 389,602 RR=1* Total 931 794,358 Table 47. Rate ratio (RR) and corresponding 95% confidence intervals (95% CI) of childhood asthma hospital admissions in low and high socioeconomic status categories by Urban Deprivation Index (SES – socioecono mic status; high socioeconomic status by Urban Deprivation Index used as a reference) SES by Urban Deprivation Index (UDI) Children asthma hospitalization Total children population Rate Ratio (95% CI) Low SES by UDI score of 15-26 532 110,872 RR=1.94 (1.67-2.26) High SES by UDI score of 8-17 249 104,625 RR=1* Total 781 215,497 Separate socioeconomic category analysis revealed that 97 and 94 percent of the excess fraction of asthma hospital admissi ons in adult and children population groups respectively could be attributed to the low socioeconomic status described by high score of UDI. The Pearson correlation analysis of crude asthma hosp italization rates and corresponding index scores values zip code areas of residence proved a statistically significant association for both adults (correlation coefficient r=0.68, p<0.001) and children (correlation coefficient r=0.53, p<0.001) in 1999. Statistically significant association was also shown for childhood (r=0.58 and r=0.59) and adult (r=0.64 and r=0.69) asthma hospitalizations in 1997 and 1998 respectively.

PAGE 146

131 The association between environmental e xposure to ambient particulate matter and sulfur dioxide and hospital admissions for ch ildhood and adult asthma within separate socioeconomic strata by UDI is given in A ppendix F. 22 of 44 residential zip code areas of residence formed a low socioeconomic status category area by Urban Deprivation Index (UDI). Accordingly, 14 and 11 of low socioeconomic status areas were also characterized by higher environmental exposure to PM10 ad SO2. There was a strong association between low socioeconomic st atus by UDI and ambient air pollution by particulate matter (r= 0.4, p<0.05), but there wa s no significant association with sulfur dioxide (r=0.07, p>0.05) shown by separate zi p code area of resi dence in Hillsborough County, FL. Positive confounding effect by UDI was seen on the association between environmental exposure to both coarse particul ate matters and sulfur dioxide and hospital admissions for adult asthma (see Appendix D) Heterogeneous effect within separate socioeconomic strata and effect modificati on was detected on the association between particulate matters and hospita lization of asthma within separate socioeconomic status strata, and displayed only a slight confounding effect on the adjusted estimates of the association between the environmental e xposure to sulfur dioxide and asthma hospitalizations in children population. 2.3.3. Socioeconomic Deprivation Index Principal component analysis (PCA) tech niques were used to reduce the total number of significant explanatory socioeconomic status indicators and to define main composite components of residential area so cioeconomic status w ithin the selected geographical area of study. Sel ected separate socioeconomic status characteristics were highly inter-correlated and do not represent complex economic, social, demographic and behavioral characteristics but rather more specific single socioeconomic indicators or factors of residential area. B ecause of highly significant correlations and mixed effects of various socioeconomic status indicators, the complex multidimensional area socioeconomic status index was constructe d by calculating principal socioeconomic status components and relative standard ized weight values for each selected

PAGE 147

132 socioeconomic status variable. Based on previo us preliminary crude association analysis of the association between ar ea socioeconomic status indi cators and asthma hospital admission, only significant socio economic status indicators we re selected for principal component analysis and to build comple x area socioeconomic deprivation index. A total of 12 different residential area socioeconomic deprivation status indicators were shown to be associated with asthma: (1) percentage of people living below poverty; (2) percentage of people with equal to or less than 9th grade high school education; (3) percentage of unemployed persons; (4) percen tage of people employed in professional or managerial occupations; (5) percentage of people with an average household income of $15,000 or less per year, (6) age of house repres enting the percentage of houses built in 1960 or earlier; (7) percentage of people w ith no access to personal car; (8) house heating by fuel; (9) percentage of houses with lacki ng kitchen and/or plumbi ng facilities in the house; (10) percentage of people living in overcrowded conditions; (11) percentage of single parents living with at least one child or children; and (12) percentage of ethnic minority householders. The final outcome and results of principal component analysis (PCA) is provided in Appendix G. There were total 12 factors identified. Based on the Kaiser-Guttman rule we identified two main factors with the corres ponding eigenvalues great er than 1.0. Factor 1 had an eigenvalue of 7, while Factor 2 had an eigenvalue of 1.44 out of the total value of 9.82. The scree plot clearly illustrates the existing di fference between the first (Factor 1) and the second (Factor 2) factors. Other f actors had eigenvalues less than 1. Further evaluation of final communality estimates and variance explained by each factor proved that the first factor explaine d 71.4 percent of total variance. The second factor was able to explain additional 14.3 percent of total variance. Factors able to explain 85 % or more out of total variance are considered appropriat e. Therefore, both of the aforementioned factors were retained for evaluation and inte rpretation of factor lo adings in order to develop the complex socioeconomic deprivati on status index. Sta ndardized regression coefficients equal to or greater than 0.3 in absolute value were used for the evaluation of significant socioeconomic status factor lo adings. The PROMAX oblique rotation method was performed to calculate standardized regr ession coefficients on the rotated factors

PAGE 148

133 (principal components). Only va riables with standardized regr ession coefficients equal to or more than 0.3 were retained and considered for further one-factor analysis to calculate standardized score values for the complex index. Factor 1 analysis revealed that percentages of people with an education level of 9th grade or below, with professional or managerial occupations (white-collar occ upations), and overcrowded living conditions were significant components. All af orementioned variables represented social status component of the given residential area (Fac tor 1, see Appendix G). Factor 2 disclosed that percentages of people living below pove rty level, unemployed, with low family income and without vehicle were significan t factors to explain economic (financial) status component. The percentage of people living with inadequate facilities was important for both components in Factor 1 a nd in Factor 2. The percentages of people living in older houses (house built in 1960 or before) and in hous es with fuel as the main heating source were not signifi cant or important to explain social status or economic status components, and were not included in the complex i ndex. The standardized scoring coefficients represented standard weights of each selected significant socioeconomic characteristic in the complex inde x model and are given in Table 48. Table 48. Selected residential area socioeconomic status indicators with standardized relative weights used to calculate the Socioeconomic Deprivation Index (SDI) Socioeconomic status indicator Relative weight Percentage of people living below poverty level % Percentage of people with 9th grade or less high school education % Percentage of unemployed people, % Percentage of people employed in professional or managerial occupations ( white-collar occupation ), % Percentage of families with annual family income of $15,000 or less, % Percentage of people with no vehicle % Percentage of people living in overcrowded housing conditions, % Percentage of houses with lacking plumbing or kitchen facilities, % 0.31 0.18 0.02 -0.01 0.28 0.07 0.19 0.07

PAGE 149

134 Percentage of people below poverty level, low family income, overcrowded living conditions and low education were the most influential character istics in the area socioeconomic deprivation index (SDI). Th e effect of such socioeconomic status characteristics as percentage of people with no vehicle, lack of plumbing or kitchen facilities, unemployed and employed in prof essional or managerial (white-collar) occupations was of lesser magnitude. The complex multi-dimensional socioeconom ic deprivation index was calculated for each zip code area of residence based on identified significant socioeconomic status characteristics and its standardized score va lue. Higher SDI estimate represents relatively higher socioeconomic deprivation and lower so cioeconomic status for a given residential area. Calculated SDI estimates by zip code area along with corresponding crude adults and children hospital admission rates per 10,000 of total adult and children population groups are presented in Table 49. The lower index score indicated the hi gher socioeconomic status and lower socioeconomic deprivation, and the highe r index score represented the lower socioeconomic status and higher socioeconom ic deprivation status in a given area respectively. The analysis of association between the socioecono mic deprivation index score and asthma hospital ad missions rates by zip code ar ea of residence revealed significant association with hos pitalization rates for total (correlation coefficient r=0.82, p<0.001), childhood (r=0.75, p<0.001) and adult asthma (r=0.77, p<0.001) in Hillsborough County, FL, in 1999. Calculated SDI values were also strongly correlated with environmental exposure to ambient particles (r=0.42, p<0.05) and slightly but not sign ificantly associated with ambient sulfur dioxide concentrations (r=0.28, p>0.05) by zip code area of residence in Hillsborough County, FL, 1999. The average mean index sc ore for the total study area was 10.2, and varied from 1.3 to 28.5 by separate zip code area of residence in 1999 (see Pictures 16 and 17).

PAGE 150

135 Table 49. Socioeconomic Deprivation Index (SD I) values and crude rates of hospital admissions for adult and childhood asthma by geographical area of residence in Hillsborough County, FL, in 1997, 1998, and 1999 Zip code area SDI (1999) Adult asthma per 10,000 in 1997 Childhood asthma per 10,000 in 1997 Adult asthma per 10,000 in 1998 Childhood asthma per 10,000 in 1998 Adult asthma per 10,000 in 1999 Childhood asthma per 10,000 in 1999 33510 5.0 8.06 16.02 9.78 8.01 9.21 26.03 33511 4.3 8.00 16.10 4.00 17.10 7.72 20.12 33527 14.3 10.43 46.35 10.43 17.83 10.43 24.96 33534 15.4 10.87 20.25 9.06 15.19 18.11 20.25 33547 6.9 13.99 33.41 12.44 19.09 7.77 9.55 33549 4.6 5.11 13.76 3.97 12.70 5.11 16.93 33556 2.3 3.67 0.00 7.34 22.56 6.43 29.00 33565 8.3 6.03 22.54 9.80 16.90 11.31 14.08 33566 14.2 13.47 26.81 16.53 36.39 17.15 32.56 33567 10.8 7.18 23.35 9.23 9.34 9.23 14.01 33569 6.7 8.76 22.89 9.86 9.64 11.32 25.30 33570 13.7 2.92 27.21 6.81 23.32 12.64 23.32 33572 4.3 6.28 18.37 6.28 9.18 1.57 9.18 33573 5.2 2.46 0.00 1.23 0.00 3.69 0.00 33584 6.8 7.62 38.04 12.06 14.79 14.60 10.57 33592 8.7 9.05 40.21 9.05 22.34 20.69 35.75 33594 3.7 9.10 12.21 4.41 6.11 7.45 20.93 33598 23.9 9.06 12.00 9.06 12.00 12.69 35.99 33602 23.1 26.81 112.48 22.57 58.92 36.68 133.90 33603 16.8 22.19 91.01 20.96 42.33 22.19 61.38 33604 17.7 29.17 57.64 19.21 40.35 18.50 59.95 33605 28.5 31.82 107.30 21.73 95.37 22.50 119.22 33606 5.9 8.52 9.79 4.65 34.26 10.06 48.95 33607 18.3 16.57 93.60 27.62 72.33 20.99 80.83 33609 8.2 13.32 52.47 11.84 26.24 11.84 41.23 33610 18.2 25.00 60.02 19.67 40.01 21.72 81.27 33611 7.6 8.80 49.59 8.80 18.60 13.60 47.52 33612 17.3 15.36 52.23 15.97 38.92 18.98 41.99 33613 15.8 13.13 59.30 10.26 29.65 13.54 41.51 33614 13.4 11.29 50.11 14.96 45.34 17.50 54.89 33615 10.1 12.31 27.33 14.71 27.33 13.81 43.48 33616 10.8 11.66 42.60 8.48 19.36 13.78 50.35 33617 11.0 13.19 36.98 12.29 25.78 6.89 35.86 33618 5.1 7.32 30.30 5.49 20.20 5.49 15.15 33619 15.4 15.61 28.45 15.15 34.44 11.48 44.92 33621 4.9 0.00 0.00 0.00 10.31 5.82 10.31 33624 3.8 5.04 16.04 5.04 17.11 7.00 17.11 33625 4.7 6.34 29.98 6.34 23.99 4.44 13.99 33626 1.3 1.20 0.00 3.59 3.64 9.56 21.83 33629 3.4 9.03 27.28 7.44 9.92 3.72 27.28 33634 7.5 11.17 39.70 9.20 22.33 6.57 37.22 33635 6.2 7.17 7.47 4.10 22.40 9.22 37.34 33637 9.3 9.19 36.55 8.16 10.96 16.33 21.93 33647 3.8 1.50 12.64 3.01 0.00 0.00 9.48

PAGE 151

136 Picture 16. SDI score by separate zip code area of residence in Hillsborough County, FL, 1999 Picture 17. Distribution of SDI score by zip code area of residence in Hillsborough County, FL, 1999

PAGE 152

137 Analysis of data on hospital admissions for both childhood and adult asthma in 1997 and 1998 supported the conclusion that ca lculated geographical area index is strongly associated with hosp italizations for asthma in th e selected area of study (see Table 49). There was shown strong significan t association with hospitalizations for childhood (r=0.74 in 1997 and r=0.74 in 1998) and adult asthma (r=0.78 in 1997 and r=0.77 in 1998) accordingly. Preliminary crude association analysis results and possible interaction or confounding effect were expl ored further by applying multiple non-linear regression analysis techniques. 2.3.4. Multiple regression modeling and analysis The multiple log-linear regression models were used to evaluate the significance and magnitude of residential area socioeconomic deprivation levels to explain and predict hospital admissions for adult and childhood asthma. The multiple regression model building techniques were also used to support or oppose previ ous crude correlation association and frequency table analysis resu lts. The influence of environmental exposure to ambient air pollution by particulate matter, sulfur dioxide and ozone was also explored to evaluate possible interaction (effect modification) and confounding effect in the multiple regression models with selected area socioeconomic status indicators and complex area multidimensional deprivation index as independent variables and asthma hospital admission rates for childhood and adul t asthma as main outcomes of interest. Standard stepwise backward selection model building procedures were used to develop the final best-fit regression m odel. Based on the initial model and subsequent stepwise backward parameter selection analysis, only statistically significant model parameters with significance level of p< =0.5 as a reference point value were retained in the model. Overall goodness of fit for all models used in the stepwise analyses was estimated based on the scaled deviance and general assumption that all m odels were adequate to explain for overdispersion Comparative models were designed to ev aluate association between socioeconomic status and asthma hos pital admissions in 1997, 1998, and 1999. The stepwise backward selection model building analysis and fitted model validation were

PAGE 153

138 conducted separately for children under 15 years of age and adults of 15 years or older. The scatter plot and calculat ed residual deviance analys es were conducted to identify extreme or outlying observations Empirical rule of an abso lute residual deviance value of more than 2.5 was used to define extreme cases (outliers). The si gnificant influence of selected extreme cases (outliers) on regre ssion model parameters was confirmed by the DFBETAS value of equal or higher than n 2 or 0.3, where n is a total number of cases used for validation (n=44). The final inte rpretation of devel oped regression model parameter estimates was provided only after the assessment of the influence of extreme cases (outliers) on the fitted re gression model parameters. Area Socioeconomic Status Indicators and Asthma Hospitalizations Childhood asthma hospital admissions Area socioeconomic status indicators were selected based on previous empirical studies and our preliminary crud e analysis results. Selected socioeconomic characteristics included poverty, education, unemployment, wh ite-collar occupation, da ily utilization of public transportation, annual family income of $15,000 or less, age of house (build in 1960 and earlier), house heating by fuel, lack ing kitchen and/or plumbing facilities, overcrowding conditions, single parent w ith a child or children, ethnic minority householder. Aforementioned residential area socioeconomic status and environmental exposure to ambient air pollution by ambient pa rticles, sulfur dioxide and ozone were used to design the best-fit regression model. Stepwise backward selection model building procedures and analysis results are presented in Appendix H. The total study area selected for analys is covered 44 postal zip code areas of residence as a main geographical unit of data analysis in Hillsborough County, FL. Number of hospitalizations fo r childhood asthma as a primary diagnosis was used as the main outcome of interest and as an offset variable (ti) in the log-linear regression model. The influence of environmental exposure to ambient air pollution by particulate matter, sulfur dioxide, and ozone controlling for area socioeconomic deprivation indices as independent variables was evaluated by using a log-linear function as the link function in

PAGE 154

139 the multiple regression model. Three comparat ive studies were designed to evaluate the association in the multivariable regressi on models by area of residence in 1997, 1998, and 1999. The analysis of multiple regression models with only environmental exposure to selected criteria ambient air pollutants included in the stepwise backward selection best-fit model building procedures, revealed th at sulfur dioxide was the only significant factor to explain the incr ease in childhood asthma hospita l admissions by geographical area of residence. Ambient particles and ozone were not significant factors in the fitted multiple regression model after adjustment to sulfur dioxide in 1997 and 1998 (see Appendix H). Both ambient particles and sulf ur dioxide were found significant factors in the fitted multiple regression models in 1999. The interaction term analysis has shown that there was no significant interaction between sulfur dioxide and ambient particles in the best-fit model. The initial log-linear multiple regression equations and subsequent best-fit regression models’ building analysis re sults for 1997, 1998, and 1999 are provided in Appendix H. The significance of each single model parameter during the backward selection stepwise model building procedures was evaluated by its ch i-square statistics and corresponding p-value. The least influentia l variable was excluded from the next step in the stepwise model building procedur es. The final best-fit model included only significant variables based on chi-square st atistics and significance level of p> ( =0.05). The percentages of people employed in prof essional occupations, with no vehicle available, and with relativ ely higher environmental expos ure to sulfur dioxide by geographical area of residence were signi ficant factors to explain the increase of childhood asthma hospitalizations and was re tained in the final model of 1997. The scatter plot of predicted values of childhood asthma hospital admissions versus residual deviances is presented in Figure 14. Environmental exposure to sulfur dioxide was shown to be a significant factor for childhood as thma hospital admissions in 1997 and 1999 but not in 1998. Less people employed in professi onal white-collar occupations and more families with no auto vehicle available we re the only significant explanatory area socioeconomic status factors to account for the increase in childhood asthma hospitalizations in 1998. Percen tage of people living below p overty along with relatively

PAGE 155

140 higher environmental exposure to ambient ai r pollution by sulfur dioxide by geographical area of residence were significant independent variables in the final fitted regression model of childhood asthma hospital admissi ons in 1999. Ambient particles and ozone were not significant factors to explain the increase in childhood asthma hospital admissions after the adjustment to area so cioeconomic status characteristics. Scatter plot analysis has shown tw o extreme cases (case 20 and 23). Figure 14 Distribution of residual deviances agai nst predicted values for the fitted regression model with childhood asthma hospital admissions as a dependent variable in 1997 -4 -3 -2 -1 0 1 2 3 4 010203040506070 Predicted valueResidual deviance The influence of these outli ers on the regression model parameters was evaluated by comparing regression model with and without selected outliers. DFBETAS value of equal or higher than n 2 or 0.3 (sample size n=44) was used to identify significant effect on regression parameters. Calculated DFBETAS for each regression model parameter is presented in Table 50. There was no significan t effect shown by selected extreme cases

PAGE 156

141 on the fitted model regression parameters (DFBETAS <0.3), and these outliers were retained in the final model. Final model parameter estimates were tr ansformed by using natural antilogarithm transformation for further model paramete r interpretation. Transformed parameter estimates with correspondi ng 95% confidence limits are presented in Table 51. Table 50. Calculated DFBETAS values for selected extreme cases (outliers) in the childhood asthma regression model in 1997 Regression Model Parameters Regression Coefficient, (i) Difference, MSE(i) DFBETAS Final fitted regression model with extreme cases (outliers) White-collar occupation -0.017 No vehicles available 0.043 Sulfur dioxide 0.590 Final fitted regression model w ithout extreme cases (outliers) White-collar occupation -0.017 0 5.626 0 No vehicles available 0.055 0.012 5.626 0.002 Sulfur dioxide 0.584 0.006 5.626 0.001 Table 51. Transformed best-fit childhood asthma hospital admissions regression model parameter estimates (1997) Parameter Parameter estimate Transformed estimate 95% CI White-collar occupation -0.017 1.018 1.005-1.029 No vehicles available 0.043 1.044 1.021-1.067 Sulfur dioxide (SO2) 0.590 1.804 1.247-2.610 Transformed fitted model parameter estimates suggested that the decrease in the percentage of people employed in white-colla r occupations and corresponding increase in the percentage of people without access to personal vehicle by 1 percent may increase childhood asthma hospital admission rates by 1.7% and 4.3% respectiv ely. Accordingly, the increase in ambient air pollution by sulf ur dioxide by 0.1 ppb could account for the increase in childhood asthma hospital admissions by 8%.

PAGE 157

142 Employment in professional occupation and no vehicle available in family were the only factors to explain the increase in hospita l admissions for childhood asthma in 1998. Ambient air pollution by particulate matter (PM10), sulfur dioxide (SO2), and ozone (O3) were not significant factors after contro lling for area socioeconomic status. The distribution of residual deviances was used to identify extreme cases (outliers) and is presented in Figure 15. Figure 15. Distribution of residual deviances agai nst predicted values for the fitted regression model with childhood asthma hospital admissions as a dependent variable in 1998 -4 -3 -2 -1 0 1 2 3 4 01020304050 Predicted valueResidual deviance There were two extreme cases (cases 30 and 44) seen above and below absolute cut-off value of 2.5 as acceptable upper and lower residual deviance distribution limits. Final fitted regression model with and without selected outliers were used to calculate DFBETAS values and to estimate the infl uence of these extreme cases on model parameters. Calculated DFBETAS estim ates are shown below in Table 52. Selected extreme cases were not influential to effect final model regression parameters (DFBETAS <0.3), and, therefore, were retained in the final model. Final

PAGE 158

143 model parameter estimates were transf ormed by using natural antilogarithm transformation for further model parameter interpretation. Transformed regression model parameter estimates with corre sponding 95% confidence limits are presented in Table 53. Table 52. Calculated DFBETAS values for selected extreme cases (outliers) in the childhood asthma regression model in 1998 Regression Model Parameters Regression Coefficient, (i) Difference, MSE(i) DFBETAS Final fitted regression model with extreme cases (outliers) White-collar occupation -0.014 No vehicles available 0.684 Final fitted regression model w ithout extreme cases (outliers) White-collar occupation -0.007 0.007 4.327 0.001 No vehicles available 0.072 0.612 4.327 0.141 Table 53. Transformed best-fit childhood asthma hospital admissions regression model parameter estimates (1998) Parameter Parameter estimate Transformed estimate 95% CI White-collar occupation -0.014 1.014 1.002-1.026 No vehicles available 0.684 1.071 1.054-1.087 Transformed final best-fit model parameter es timates suggested that the decrease in the percentage of people employed in white-collar occupations and corresponding increase in the percentage of people without access to person al vehicle by 1 percent could explain the increase in childhood asthma hospital admission rates by 1.4% and 7.1% respectively. Finally, the stepwise backward select ion model building analysis of childhood asthma hospital admission model revealed th at the percentage of people living below poverty and also environmental exposure to am bient air pollution by sulfur dioxide were significant predictors of hosp italizations for childhood asthma by postal zip code area of residence in 1999 (see Appendix H). Residual devi ance scatter plot was used to identify

PAGE 159

144 and explore extreme cases (outliers). The di stribution of extreme cases represented by residual deviance of equal or more than absolute value of 2.5 against corresponding predicted value of childhood asthma hospitaliz ation by the fitted model is illustrated in Figure 16. Figure 16. Distribution of residual deviances against predicte d values for the fitted regression model with childhood asthma hospital admissions as a dependent variable in 1999 -4 -3 -2 -1 0 1 2 3 4 0102030405060 Predicted valuesResidual deviance Table 54. Transformed best-fit childhood asthma hospital admissions regression model parameter estimates (1999) Parameter Parameter estimate Transformed estimate 95% CI Poverty 0.0561 1.0577 1.044-1.071 Sulfur dioxide (SO2) 0.3887 1.4751 1.239-1.755 There were no extreme case values (above or below absolute residual deviance value of 2.5) identified and the final model was accepted as accurate to predict childhood asthma hospitalizations by geographical area of residence. Final fitted model regression

PAGE 160

145 parameters were transformed for further inte rpretation and are pres ented above in Table 54. Best-fit model parameter estimates suggested that the decrease in the increase in the percentage of people living be low poverty level by 1 percent and in ambient environment air pollution by sulfur dioxi de by 0.1 ppb may explain the growth of childhood asthma hospitalizations by 5.7 and 4.75 percents respectively. Adult asthma hospital admissions The influence of environmental exposure to selected criteria ambient air pollutants on adult asthma hospitalization rates by zip c ode area of residence while controlling for residential area socioeconomic status, was expl ored separately and is presented below. Three comparative log-linear regression models of hospitalizations for adult asthma were developed to estimate the asso ciation between local enviro nmental exposure to ambient air pollution and hospitalizati on for adult asthma after the adjustment to selected area socioeconomic status characteristics in 1997, 1998, and 1999. Initial multiple model analysis results and subsequent stepwise b ackward selection model building procedures are presented in Appendix H. Percentages of people living below povert y level and single parents with children or at least one child, and households with house heating by fuel by geographical area of residence were found to be significant predicto r variables for adult asthma hospitalization rates in 1997. Accordingly, th e percentage of people living below poverty, households living in houses with fuel as main source of heat, and residing in houses with lacking kitchen or plumbing facilities were significan t independent factors in the final best-fit regression model in 1998. Fina lly, asthmatic adults living below poverty, residing in less crowded housing conditions (less than 1 person in room per house), and less likely to be employed in professional or managerial occupa tions were at higher ri sk of hospitalization for asthma. Poverty was an important factor of adult asthma hospitaliz ations and also was found to be a single significant predictor of hospital admissions in single log-linear regression analyses (p<0.0001) during th e overall period of study in 1997, 1998, and 1999 (see Appendix H).

PAGE 161

146 The distribution of residual deviances ve rsus predicted values of adult asthma hospital admissions in 1997 is given in Figure 17. Figure 17. Distribution of residual deviances against predicte d values for the fitted regression model with adult as thma hospital admissions as a dependent variable in 1997 -3 -2 -1 0 1 2 3 010203040506070 Predicted valueResidual deviance Residual analysis has disclosed only one extreme case (case 44, see Appendix H). Final fitted regression models with and without selected outli er were compared to explore the influence and effect of this extreme case on model parameters. Calculated DFBETAS values for selected outlier are presented in Table 55. Calculated difference between regres sion model parameters and DFBETAS estimates disclosed that case 44 was not influe ntial and could be retained in the model without further remedy measures. Final fitted model parameter estimates were transformed by using anti-log transformation fo r further interpretati on purposes. Obtained initial and transformed m odel parameter estimates wi th corresponding 95 percent confidence intervals are represented in Table 56.

PAGE 162

147 Table 55. Calculated DFBETAS values for selected extreme cases (outliers) in the adult asthma regression model in 1997 Regression Model Parameters Regression Coefficient, (i) Difference, MSE(i) DFBETAS Final fitted regression model with extreme cases (outliers) Poverty 0.036 House heating by fuel 0.157 Single parent family with children 0.055 Final fitted regression model w ithout extreme cases (outliers) Poverty 0.036 0 5.030 0.000 House heating by fuel 0.147 0.010 5.030 0.001 Single parent family with children 0.052 0.003 5.030 0.001 Table 56. Transformed best-fit adult asthma hospital admissions regression model parameter estimates (1997) Parameter Parameter estimate Transformed estimate 95% CI Poverty 0.036 1.037 1.015-1.059 House heating by fuel 0.157 1.082 1.003-1.364 Single parent family with children 0.055 1.057 1.011-1.104 Transformed final best-fit model parameter es timates suggested that the increase in the percentage of people living below poverty and in the houses heated by fuel in the area of residence by 1% could increase adult asthma hospital admission rates by 3.7% and 8.2% respectively. In addition, if the percentage of single parent families with children or at least one child were increased by resident ial area by 1%, hospitalization rates for adult asthma could increase by 5.7% accordingly. Ambient air pollution by particulate matter (PM10), sulfur dioxide (SO2), and ozone (O3) were not significant factors after the adjustment to area socioeconomic status in the multiple regression analysis. Percentage of people living below poverty and residing in hous es lacking kitchen and/or plumbing facilities were also shown to be significant independent variables in the multiple regression model of adult asthma hospitalizations by zip code area of residence in 1998. The percentage of people residing in the houses heated by fuel was also an

PAGE 163

148 important factor to explain the increase of hospital admissions for adult asthma in the selected study area in 1998. Environmental exposure to ambient air pollution by particulate matter (PM10), sulfur dioxide (SO2), and ozone (O3) were not significant factors after the adjustment to area socio economic status in the multiple regression analysis. The distribution of residual deviances ag ainst predicted values by the model were used to explore extreme cases (outliers) and is presented in Figure 18. There was only one extreme case identified. Final fitted models with and without extreme case (case 14) were compared to calculated DFBETAS estimates and to evaluate the influence of selected outlier on best-fit model paramete rs. Calculated DFBETAS are given below in Table 57. Estimated difference between regr ession model parameters concluded that selected outlier (case 44) was not influential and could be kept in the final fitted model. Obtained transformed model parameter estimates with co rresponding 95 percent confidence intervals are repr esented below in Table 58. Figure 18. Distribution of residual deviances agai nst predicted values for the fitted regression model with adult as thma hospital admissions as a dependent variable in 1998 -4 -3 -2 -1 0 1 2 3 4 0102030405060 Predicted valueResidual deviance

PAGE 164

149 Table 57. Calculated DFBETAS values for selected extreme cases (outliers) in the adult asthma regression model in 1998 Regression Model Parameters Regression Coefficient, (i) Difference, MSE(i) DFBETAS Final fitted regression model with extreme cases (outliers) Poverty 0.074 House heating by fuel 0.224 Lacking kitchen or plumbing facilities -0.235 Final fitted regression model w ithout extreme cases (outliers) Poverty 0.072 0.002 4,792 0.000 House heating by fuel 0.216 0.008 4.792 0.001 Lacking kitchen or plumbing facilities -0.221 0.014 4.792 0.002 Table 58. Transformed best-fit adult asthma hospital admissions regression model parameter estimates (1998) Parameter Parameter estimate Transformed estimate 95% CI Poverty 0.074 1.077 1.058-1.096 House heating by fuel 0.224 1.251 1.075-1.455 Lacking kitchen or plumbing facilities -0.235 1.264 1.434-1.115 Based on transformed fitted model parameter estimates, the increase in the percentage of people living below poverty by 1% could increase adult asthma hospital admission rates by 7.7%. Even relatively small increase by 0.1% in the percentage value of houses heated by fuel and houses lacki ng kitchen and plumbing facilities could increase adult asthma hospitalization rates within selected geographical area of residence by 2.5% and 2.6% respectively. Percentage of people living below povert y level and less crowded conditions with less chance to be employed in white-colla r professional occupations were the only significant factors in final be st-fit regression model of a dult asthma hospital admission rates by geographical area of residence in 1999. Environmental exposure to ambient air pollution by particulate matter (PM10), sulfur dioxide (SO2), and ozone (O3) were not significant factors after the adjustment to area socioeconomic status in 1999. The

PAGE 165

150 distribution of model residual deviances against predicted values was plotted to explore extreme cases (see Figure 19). Three extreme cases (cases 30, 35 and 44) were identified as outliers. The significance and influence of selected outliers on model pa rameters was explored by comparing final fitted models constructe d with and without aforementioned extreme cases. Figure 19. Distribution of residual deviances against predicte d values for the fitted regression model with adult as thma hospital admissions as a dependent variable in 1999 -5 -4 -3 -2 -1 0 1 2 3 4 5 0102030405060 Predicted valuesResidual deviance Obtained DFBETAS estimates are presente d in Table 59. There was no significant influence shown by selected extreme cases (outliers) on the fitted regression model parameters (DFBETAS <0.3), and, therefore, these outliers were retained in the final model.

PAGE 166

151 Table 59. Calculated DFBETAS values for selected extreme cases (outliers) in the adult asthma regression model in 1999 Regression Model Parameters Regression Coefficient, (i) Difference, MSE(i) DFBETAS Final fitted regression model with extreme cases (outliers) Poverty 0.067 White-collar occupation -0.016 Overcrowded housing conditions -0.043 Final fitted regression model w ithout extreme cases (outliers) Poverty 0.072 0.005 6.326 0.001 White-collar occupation -0.020 0.004 6.326 0 Overcrowded housing conditions -0.0627 0.019 6.326 0.003 Final model parameter estimates were tr ansformed by using natural antilogarithm transformation for further model parameter interpretation. Transformed regression model parameter estimates with corre sponding 95% confidence limits are presented in Table 60. Table 60. Transformed best-fit adult asthma hospital admissions regression model parameter estimates (1999) Parameter Parameter estimate Transformed estimate 95% CI Poverty 0.067 1.069 1.045-1.092 White-collar occupation -0.016 1.016 1.003-1.029 Overcrowded housing conditions -0.043 1.044 1.003-1.085 Based on the estimates of transformed m odel parameter estimates, the increase in the percentage of people living below povert y by 1% could explain the increase in hospitalizations for adult asthma by 6.9%. Fina lly, the decrease in the percentage of people employed in white-collar professiona l occupations and living in overcrowded conditions could increase adult asthma hospitalization rates over the selected geographical area of residen ce by 1.6% and 4.3% respectively.

PAGE 167

152 Area Socioeconomic Deprivation I ndex and Asthma Hospitalizations Calculated area socioeconomic deprivati on index represented complex composite socioeconomic status of residential area, and was used to evaluate the association between hospitalizations for childhood and adult asthma and environmental exposure to ambient air pollution by particulate matter, sulfur dioxide, and ozone by geographical area of residence while controlling for co mplex multidimensional area socioeconomic deprivation index. Initial l og-linear regression model de sign and best-fit stepwise regression model building techniques were appl ied in similar fashion to previous area socioeconomic status characteristics anal ysis and were explained earlier (see Methodology). Selected geogra phical study area included total 44 postal zip code areas of residence in Hillsborough County, FL. Calculat ed area socioeconomic deprivation status index and environmental exposure to ambien t air pollution by selected geographical area of residence were used to develop the be st-fit model for childhood and adult asthma hospital admissions as main dependent outcome variables of interest. Separate analyses were provided for hospital admissions for childhood and adult asthma correspondingly. The children and a dult population groups by separate zip code area of residence were defi ned as offset variables (ti) in the stepwise model building procedures. The study period covered the ye ar 1999, however, comparative log-linear multiple regression equations were designed to explore the association between asthma hospitalizations and environmental exposure to selected criteria ambient air pollutants after the adjustment to re sidential area socio economic deprivation index in 1997 and 1998. Ambient air quality data by zip code area of residence in 1999 most closely corresponded to the US Census 2000 da ta on demographic and socioeconomic characteristics of residential area which refl ected to and were colle cted during the year 1999. Therefore, the association between asthma hospitalizations and local environmental exposure controlling for socio economic deprivation status by selected residential area were explored in more details and were al so used to build and verify final best-fit predictive model in Hillsborough County, FL, in 1999. Final best-fit regression model was validated to evaluate the accuracy of model prediction by using independent data

PAGE 168

153 outside of our study area. The error mean square (MSE) and mean square prediction error (MSPR) estimates were calculated to validate fitted predictive model and also to estimate the accuracy of developed best -fit regression model (see 2.2. Methodology, Chapter 2, p. 63). The initial multiple log-lin ear regression model and final results of stepwise backward selection model building analyses are presented in Appendix I. Childhood asthma hospital admissions Number of hospital admissions for chil dhood asthma as a primary diagnosis was used as the main outcome of interest. The total children population by separate zip code area of residence was defined as offset va riable using the OFFSET option in SAS. The behavior and possible interac tion of socioeconomic depriv ation index with selected environmental exposure to ambient particle s, ozone and sulfur dioxide by area of residence was evaluated by using a log-linear function as the link f unction and interaction terms included in the fitted regression model. The initial model chosen for stepwise b ackward model analysis had 39 degrees of freedom (df=39). Poisson regression model was evaluated using the deviance estimates of different models and conducting test based on pa rtial deviances.66 Overall goodness of fit test based on the model deviance and P earson chi-square statistics suggested overdispersion. Pearson chi-s quare scaled parameter ( PSCALE option in SAS) was applied to control for the overdispersion and to assure that the model is adequate to explain for overdispersion. The model scale parameter was =1.23. The final stepwise backward selection model building analyses disclosed that only area socioeconomic deprivation index and environm ental exposure to sulfur dioxi de were significant factors to explain increase in hosp ital admissions for childhood asthma by zip code area of residence. The significance of each single model parameter was evaluated by its chisquare estimate and corresponding p-value. Next step stepwise regr ession model retained all factors but the one with highest p-value or least significant variable. Environmental exposure to ambient particle and ozone were excluded as not significant factors from the second and third step regression models resp ectively (see Appendix I). The final best-fit model included only significant variables ba sed on chi-square sta tistics and significance

PAGE 169

154 level of p> ( =0.05). The residential area SDI and ambient air pollu tion by sulfur dioxide were significant explanatory variables retain ed in the final model. The analysis of interaction terms between ar ea socioeconomic status and ambient particle, ozone and sulfur dioxide exposure disclosed that none of the interaction terms was significant in the multiple regression analyses. There was no in teraction or effect modification effect between area socioeconomic deprivation stat us and local environmental characteristics shown in the model. The residual deviance anal ysis disclosed that there is no extreme or outlying cases in the data se t used to fit the final regr ession model (see Figure 20). Figure 20. Distribution of residual deviances against predicte d values for the fitted socioeconomic deprivation index and sulfur dioxide regression model with childhood asthma hospital admissions as a dependent variable -4 -3 -2 -1 0 1 2 3 4 010203040506070 Predicted valueResidual deviance By using antilogarithm function significan t model parameters were transformed for interpretation purposes. The list of transf ormed parameters with corresponding 95% confidence intervals (95% CI) retained in the final regression model is presented in Table 61.

PAGE 170

155 Table 61. Transformed best-fit childhood asthma hospital admissions regression model parameter estimates Parameter Parameter estimate Transformed estimate 95% CI SDI 0.0593 1.061 1.046 – 1.076 SO2 0.4161 1.516 1.268 – 1.813 Based on parameter estimates we could pred ict that the increase in socioeconomic deprivation index of residentia l area by 1 single unit will result in the increase of hospital admissions for childhood asthma by 6.1% in a given area of residence. Correspondingly, the increase in sulfur dioxid e concentration by 0.1 ppb coul d increase hospital admissions for childhood asthma by 5.2%. The significant predictive effect of residential area socioeconomic deprivation status as a si ngle independent predictor variable was concluded in a subsequent simple log-lin ear regression model. The socioeconomic deprivation index by itself was able to explain an increase in hospital admissions for childhood asthma. Figure 21. Scatter plot of area socioeconomic deprivation index and childhood asthma hospitalizations 0.0 5.0 10.0 15.0 20.0 25.0 30.0 0 10 20 30 40 50 60 70 Childhood asthma hospitalizations SDI

PAGE 171

156 The scatter plot of area socioeconomic deprivation status and childhood asthma hospitalizations is presented above in Fi gure 21. The distribution of model residual deviances against predicted values was plotted to identify extreme cases (outliers) and to explore the influence of these extreme cases on regression model parameters. The accuracy of fitted regression model to predict asthma hospital admissions as our outcome of interest was evaluated by error deviances retained within statistically acceptable deviance range of plus-minus 2.5 as a reference value.66 Other cases above or below selected cut-off point value were iden tified as extreme cases or outliers. The distribution of residual deviances versus predicted values of hospital admissions for childhood asthma is represented in Figure 22. Figure 22. Distribution of residual deviances against predicted values for the fitted socioeconomic deprivation index regre ssion model with childhood asthma hospital admissions as a dependent variable -4 -3 -2 -1 0 1 2 3 4 0102030405060 Predicted valueResidual deviance Four cases were identified as extreme cas es (or outliers): case 10 representing zip code 33566 (residual deviance was equal 3.11), case 18 representing zip code 33594 (2.84), case 26 representing zip code 33609 ( 2.49), and case 27 representing zip code

PAGE 172

157 33610 (2.52). DFBETAS estimates for each outlying case were calculated to evaluate the influential effect of each outlier on regression coefficients. Differences between the regression coefficients in the fitted models ( ) and in the models without an outlier ( (i)), and error mean square values obtaine d when the outlier is deleted (MSE(i)) with its 95% Confidence limit intervals (95% CL) are given for each outlying case separately in Table 62. The DFBETAS analysis suggested that by deletin g one or all extreme cases there was virtually no change in regression coeffici ents, and that selected outlying cases are not influential to require remedial measures. Fi tted regression model parameters were not influenced by selected ex treme cases (outliers). Table 62. Calculated DFBETAS values for selected extreme cases (outliers) in the childhood asthma regression model in 1999 Extreme case (outlier) Regression Coefficient, Regression Coefficient, (i) Difference, MSE(i) DFBETAS Case 10 0.0743 0.0734 0.0006 37.93 0.00010 Case 18 0.0743 0.0783 -0.0043 37.57 0.00070 Case 26 0.0743 0.0717 0.0023 33.89 0.00040 Case 27 0.0743 0.0759 -0.0019 39.79 0.00030 All cases 0.0743 0.0774 -0.0031 42.03 0.00048 Antilogarithm transformation function was used to transform a significant fitted model parameter estimate with its 95% confid ence intervals (95% CI) for interpretation purposes. The final transformed paramete r estimated was 1.077 with 95% CI: 1.0611.094. Therefore, the increase of socioeconom ic deprivation index in the area of residence by 1 index score will result in the increase of hospital admission rates for childhood asthma by 7.7% in a given area of residence. The comparative log-linear regression models supported th e conclusion that area soci oeconomic deprivation index could be used as a single independent predic tors of childhood asthma hospitalizations in 1997 and 1998 (see Appendix I). In 1997, area so cioeconomic deprivation and exposure to ambient air pollution by sulfur dioxide were the only significant factors in the final fitted model. Accordingly, in 1998, area socioeconomic deprivation status was the only

PAGE 173

158 single significant factor to explain the incr ease in hospitalizations for childhood asthma by selected area of residence. Environmenta l exposure to ambient air pollutants by particulate matter, sulfur dioxide and ozone were not significan t variables after the adjustment to area socioeconomic status a nd were excluded in the stepwise backward selection model building procedures from the final fitted model. Transformed parameter estimates suggested that the increase in area socioeconomic deprivation index as a single independent predictor by 1 unit may increas e hospitalizations fo r childhood asthma by 7.9% in 1997 and by 8.3% in 1998 respectively. This conclusion supported our previous multiple regression model analysis results in 1999. Final fitted regression model of 1999 was further used to assess the accuracy of model prediction by using independent data outside our main study area. Adult asthma hospital admissions The initial model chosen for stepwise b ackward model analysis had 39 degrees of freedom (df=39). Scaled Pearson chi-square pa rameter was applied to control for and to assure that the model is adequate to expl ain for overdispersion. Scale parameter value was =1.54. The initial model in the stepwi se backward selection model building process indicated that area socioeconomic deprivation index wa s the only significant factor able to explain incr ease in hospital admissions fo r childhood asthma by zip code area of residence. The significance of each single model parameter was evaluated by its chi-square statistics and co rresponding p-value. Environm ental exposure to sulfur dioxide, ozone and ambient particle were not significant factors and were excluded from the second, third and fourth step model bui lding procedures respectively (see Appendix F). The final best-fit model included only va riables based on chi-square statistics and significance level of p> ( =0.05). The SDI score was a signi ficant factor to explain the increase of adult asthma hospitalizations and was retained in the final model. The scatter plot of socioeconomic indices and adult as thma hospitalizations by zip code area of residence is presented in Figure 23.

PAGE 174

159 Figure 23. Scatter plot of area socioeconomic deprivation index and adult asthma hospitalizations in 1999 0.0 5.0 10.0 15.0 20.0 25.0 30.0 010203040506070 Adult asthma hospitalizationsSDI Figure 24. Distribution of residual deviances against predicted values for the fitted socioeconomic deprivation index regression m odel with adult asthma hospital admissions as a dependent variable in 1999 -6 -4 -2 0 2 4 6 010203040506070 Predicted valueResidual deviance

PAGE 175

160 The interaction terms between socioec onomic deprivation index and ambient particles, ozone and sulfur dioxide included in the model were not si gnificant factors to explain for adult asthma hosp italization admissions by area of residence. The residual analysis was conducted to identify outlying cas es and to evaluate their influence on the regression model. The distribut ion of residual deviance versus predicted values is shown above in Figure 24 Four cases or 6.8% out of original da ta set were identified as extreme cases (or outliers). The case 16 represente d zip code area 33592 and respective residual deviance value for this outlying case was e qual 2.58 (see Appendix I). The zip code area 33605 was also defined as an outlying cas e (case 22) with corresponding residual deviance value of -2.73. The case 33 and 44 represented zip codes 33617 and 33647 and had residual deviance esti mates of -2.75 and case -5.1 accordingly. DFBETAS estimates for each outlying extreme case were obtained to evaluate influential outliers. Differences between the fitted model regression coefficients ( ) and in the regres sion models without selected outlier ( (i)), error mean square values obta ined when the outlier is deleted (MSE(i)) and its 95% confidence limits ( 95% CL) are given in Table 63. Table 63. Calculated DFBETAS values for selected extreme cases in the adult asthma regression model Extreme case (outlier) Regression Coefficient, Regression Coefficient, (i) Difference, MSE(i) DFBETAS Case 16 0.0644 0.0651 -0.0007 48.51 0.00010 Case 22 0.0644 0.0744 -0.01 40.39 0.00157 Case 33 0.0644 0.0638 0.0006 43.37 0.00009 Case 44 0.0644 0.0612 0.0032 43.90 0.00048 All cases 0.0644 0.0708 -0.0064 49.80 0.00091 All DFBETAS estimates were below establishe d reference cut-off point value of n 2 or 0.3. There was only a slight but not significant difference between the fitted model parameter estimate ( ) and parameter estimates with deleted outlying cases ( (i)). The DFBETAS analysis proved that fitted regressi on model parameter estimates were not significantly influenced by selected extreme cases (outliers).

PAGE 176

161 Antilogarithm transformation function wa s used to transform fitted regression model parameter ant its 95% confidence in tervals (95% CI) for final parameter interpretation. The transformed model para meter estimate was equal to 1.067 with 95% CI: 1.051-1.082. Therefore, the in crease of area socioeconomi c deprivation index by 1 unit will result in the increase of childhood asthma hospital admission rates by 6.7 % in a given area. The analysis of adult asthma hospitalization rates in 1997 and 1998 revealed that area socioeconomic deprivation status together with enviro nmental exposure to ambient particles over the area of residence were significant explanatory factors retained in the final best-fit models (see Appendix I). Simple log-linear regr ession analysis also concluded that area socioeconomic depr ivation status was a single significant independent predictor of adult asthma hospita lization rates and accounted for the increase in hospitalizations by 7.5% in 1997 and by 7.2% in 1998 respectively. Final fitted regression model was used to evaluate the accuracy of model prediction by using independent data outside of our study area. 1.3.5. Validation of predictive regression model Childhood asthma hospital admissions The fitted regression model was selected and used to evaluate the accuracy of model prediction by using independent data outside our study area. The best-fit regression model could be represented as: 66 ln(Yi) = 0 + *Xi + ln(ti) where Yi – predicted variable by the model; Xi – independent predicto r variable; ln (ti) – offset term representing population size (ti). The final predictive regression model of to tal number of hospital admissions for childhood asthma by postal zip code area coul d be represented as follow: Predicted childhood asthma hospital ad missions = exp(-6.489 + 0.0743*SDI + ln(children population size))

PAGE 177

162 Table 64. Actual and predicted estimates of childhood asthma hospital admissions by zip code area of residence in Pinellas County, FL, 1999 Zip Code Total Population Children Population Childhood Asthma Predicted Childhood Asthma Error Estimate SDI 33701 15374 1881 7 11 -4 18.6 33702 30058 4637 11 11 0 6.1 33703 25063 4469 14 9 5 4.4 33704 16714 2823 6 6 0 5.5 33705 28083 6180 46 31 15 16.1 33706 17376 1380 0 3 -3 4.5 33707 26542 3069 9 9 0 8.8 33708 17199 1365 1 3 -2 5.5 33709 26039 3951 9 12 -3 9.0 33710 33213 5666 13 13 0 5.8 33711 19915 4385 33 20 13 14.5 33712 26222 5968 34 25 9 13.8 33713 31273 6087 26 19 7 10.0 33714 17753 3151 13 13 0 13.2 33715 7403 501 3 1 2 2.5 33716 10409 978 5 2 3 6.9 33755 26061 5120 49 20 29 12.4 33756 29081 4740 15 14 1 9.0 33759 20071 3836 18 13 5 11.0 33760 16958 2832 6 10 -4 10.8 33761 19594 2683 3 6 -3 4.1 33762 6818 702 0 1 -1 2.6 33763 18029 1682 5 4 1 4.9 33764 23673 3476 6 8 -2 5.9 33765 13403 1946 3 5 -2 7.8 33767 9765 537 4 1 3 4.3 33770 24394 3397 12 9 3 7.2 33771 29225 3620 13 9 4 6.9 33772 23232 3359 7 7 0 4.9 33773 16369 2894 2 6 -4 4.8 33774 18431 2916 15 7 8 5.9 33776 13388 2360 4 4 0 1.7 33777 17328 3236 6 7 -1 5.1 33778 13639 2349 8 6 2 6.0 33781 25287 5208 15 15 0 8.6 33782 19527 3116 7 7 0 5.8 33785 5949 470 2 1 1 3.3 34677 19628 3983 7 8 -1 3.5 34683 34025 6605 11 13 -2 3.7 34684 27429 3974 5 9 -4 4.7 34685 17559 3760 8 7 1 2.6 34689 28752 4647 14 11 3 6.6 34695 18156 3191 5 6 -1 3.5 34698 34235 4346 9 10 -1 6.1

PAGE 178

163 The final fitted regression model was sele cted for the validation of model precision with independent data. The hospital admi ssion rates for childhood asthma by zip code areas of residence in Pinellas County, FL, in 1999, were used to validate selected predictive model based on original data set error mean square (MSE) and mean square prediction error (MSPR) values. Close values suggest that the selected best-fit regression model is not seriously biased and gives an a ppropriate indication of the predictive ability. The actual and predicted number of hos pital admissions for childhood asthma by zip code area of residence by using the selected fitted regression model and new independent data set is presented in Tabl e 64. Pearson correlation analysis revealed strong significant association between act ual and predicted by the model childhood asthma hospital admission number for childhood asthma by zip code area of residence (correlation coefficient r=0.9, p<0.0000). Calculat ed estimate of mean square prediction error (MSPR) was compared with a value of error mean square (MSE) based on original data set to estimate the predictive ability of the model. The MSPR estimate (MSPR=38.2) was very close to MSE value (MSE=41.1 with standard error estimate SE=9.77 and 95% CI : 21.4 – 60.8)). Therefore, the error mean square for the selected predictive model is not seriously biased a nd gives an appropriate indication of the predictive ability for the model. Adult asthma hospital admissions The log-linear fitted regression model wa s used to validate and to evaluate the accuracy of model prediction by using independen t data outside of our main study area of Hillsborough County, FL, in 1999. The final fitt ed regression model could be written as:66 ln(Yi) = 0 + *Xi + ln(ti) where Yi – predicted variable by the model; Xi – independent predicto r variable; ln (ti) – offset term representing population size (ti). The final predictive regression model of the total number of hospital admissions for adult asthma by postal zip code area c ould be represented as follow: Predicted adult asthma hospital admissions = exp(-7.4637 + 0.0644*SDI + ln(adult population size))

PAGE 179

164 Table 65. Actual and predicted estimates of ad ult asthma hospital admissions per 10,000 population by zip code area of residence in Pinellas County, FL, 1999 Zip Code Total Population Adult Population Adult Asthma Predicted Adult Asthma Error Estimate SDI 33701 15374 13493 15 26 10.7 18.6 33702 30058 25421 19 22 2.6 6.1 33703 25063 20594 9 16 6.7 4.4 33704 16714 13891 5 11 6.3 5.5 33705 28083 21903 28 35 7.3 16.1 33706 17376 15996 8 12 4.3 4.5 33707 26542 23473 17 24 6.8 8.8 33708 17199 15834 17 13 -4.1 5.5 33709 26039 22088 18 23 4.6 9.0 33710 33213 27547 15 23 7.9 5.8 33711 19915 15530 21 23 1.7 14.5 33712 26222 20254 30 28 -1.8 13.8 33713 31273 25186 19 27 8.4 10.0 33714 17753 14602 14 20 5.5 13.2 33715 7403 6902 4 5 0.7 2.5 33716 10409 9431 8 8 0.4 6.9 33755 26061 20941 27 27 -0.3 12.4 33756 29081 24341 35 25 -10.0 9.0 33759 20071 16235 16 19 2.9 11.0 33760 16958 14126 12 16 4.2 10.8 33761 19594 16911 10 13 2.6 4.1 33762 6818 6116 2 4 2.2 2.6 33763 18029 16347 15 13 -2.1 4.9 33764 23673 20197 22 17 -5.1 5.9 33765 13403 11457 9 11 1.9 7.8 33767 9765 9228 10 7 -3.0 4.3 33770 24394 20997 40 19 -20.8 7.2 33771 29225 25605 30 23 -7.0 6.9 33772 23232 19873 20 16 -4.4 4.9 33773 16369 13475 7 11 3.6 4.8 33774 18431 15515 9 13 4.0 5.9 33776 13388 11028 4 7 3.1 1.7 33777 17328 14092 9 11 2.2 5.1 33778 13639 11290 11 10 -1.4 6.0 33781 25287 20079 13 20 7.0 8.6 33782 19527 16411 19 14 -5.3 5.8 33785 5949 5479 4 4 -0.1 3.3 34677 19628 15645 17 11 -5.7 3.5 34683 34025 27420 18 20 2.0 3.7 34684 27429 23455 23 18 -4.8 4.7 34685 17559 13799 10 9 -0.7 2.6 34689 28752 24105 15 21 6.1 6.6 34695 18156 14965 11 11 -0.2 3.5 34698 34235 29889 37 25 -11.6 6.1

PAGE 180

165 The final best-fit model was selected for final model validation with independent data. Adult asthma hospital admission rates by zip code areas of permanent residence in Pinellas County, FL, in 1999, were used to vali date selected model based on original data set error mean square (MSE) and mean squa re prediction error (MSPR) values. Fairly close values suggest that th e selected regressi on model is not seriously biased and confirms the accurate model predictive ability.66 The actual and pred icted numbers of hospital admissions for childhood asthma by zip code area of residence by using the selected fitted regression mode l and new independent data set is presented above in Table 65. Simple correlation analysis disclosed str ong significant correlation between actual and predicted hospital admission numbers for adult asthma (correlation coefficient r=0.8, p<0.0000). Calculated estimate of mean square prediction error (MSPR) was compared with a value of error mean square (MSE) based on original data set to estimate the predictive ability of the mode l. Calculated MSPR (MSPR=35.7) was fairly close to MSE estimate and within the range of its confid ence limits (MSE=48.7 with standard error estimate SE=10.4 and 95% CI: 27.8 – 69.6). Theref ore, the error mean square for the selected predictive model is not seriously bi ased and gives an appr opriate indication of the accurate predictive ability.

PAGE 181

166 Chapter 3 SUMMARY DISCUSSION Distribution of Childhood and Adult Asth ma Hospital Admissions in Hillsborough County, FL The rate of asthma has increased dramatically during the last two decades in the United States.2 Inner-city ethnic minor ity children are specific vulnerable population groups with the highest prevalence of asth ma and asthma-associated hospitalization rates.3 Despite its importance, there is minimal surveillance of asthma across the country, and no comprehensive surveillance system has been established that measures asthma trends at the state or local level.1, 14 Such information however is crucial to identify specific high-risk population groups and also to design and evaluate more effective preventive intervention measures. Population data is usually compiled and eas ily available for specific administrative regions (administrative county districts, postal zip code ar eas or census wards). Data on health outcomes or conditions may be aggreg ated to administrative health districts or postal zip code areas. Our st udy covered all postal zip code areas in Hillsborough County, FL. The principal unit of original data analysis was postal zip code area of residence. The total number of 44 postal zip code areas of reside nce was used to represent the geographical area of Hillsborough County, FL. There were no data available on asthma hospital admissions for additional five zip code areas of residence (zip code areas 33547, 33572, 33573, 33621, and 33647). These areas were excluded from any further calculations and data analysis There were no permanent resi dents of the postal zip code area 33620, who were referred and admitted to th e hospital. Zip code 33620 is the main campus area of University of South Florida, Ta mpa, FL. Main local re sidents living in the

PAGE 182

167 aforementioned area are undergraduate and grad uate students living temporarily in the dormitory facilities but with a permanent re sidence place outside of the area. Population of the given area is very mobile and unstabl e to include into the study. A zip code area 33620 was also excluded from subsequent data analysis. Our study offered an excellent opportunity to evaluate the di stribution of asthma hospitali zations and to explore area socioeconomic status and environmental expos ure to criteria ambient air pollutants as possible risk factors for hos pital admissions for childhood a nd adult asthma in a given community. The ecological study design provide s a unique opportunity to evaluate the association in the specific geographic area and draw relevant conclusions about the preventive interventions in the population at large. The excess of asthma hospital admissions attributable to environmental and/ or socioeconomic factors within specific small-area of residence could provide the mo st valuable information and guidelines to develop proper intervention policy for asth ma management and prevention in our communities today. The existing differences in asthma hosp italizations could be described by using crude, adjusted, or category-specific rate s. The decision depends primarily on the information that an investigator is trying to obtain or impart. Crude-rates represent the actual experience of the population and provide valuable information for the allocation of health resources and public health planning.63 Estimated crude asthma hospital admission rates varied from 0.9 to 50.35 (average mean value of 15.9, SE=1.67), from 1.22 to 39.81 (average mean 13.3, SE=1.28, and from 2.28 to 56.95 (average mean 17.14, SE=1.65), per 10,000 population by separate zip code area of residence in 1997, 1998 and 1999 respectively. Represented average crude ra tes of asthma hospital admissions over the selected study area during the period of 1997-1999 correspond closely to the average crude rate of asthma hospitalizations (17 per 10,000 population) estimated by the National Center for Health Statistics (NCH S) at the Center for Disease Control and Prevention (CDC), Atlanta, Georgia.7 Unfortunately, the fact that the crude values may be confounded by demographic characteristics of underlying population subgr oups makes the interpretation of study results very difficult. Category-specific ra tes are not confounded by specific factor and

PAGE 183

168 provide the most detailed information about the pattern of the disease in a given population. Adjusted (standardized) rates pr ovide a summary value that removes the effect of the differences in population struct ure to permit for valid comparisons between groups. While the interpretation of a singl e standardized rate is simple and straightforward in the indire ct adjustment method, a probl em arises when trying to compare a number of standardized morbidity ratios from different populations with each other. To compare the relative effect of the two or more levels of exposure, the direct method of standardization, in which a common standard is used for the two or more population groups, is preferred.63,64 For more easy comparison and interpretation, the direct method was used for adjustment of as thma hospital admission rates by age, gender, and race. By using the standard direct method for adjustment, we were able to control for individual’s age, gender and race as possibl e confounders. Individual information on age, gender, and race was used to calculate adjusted rates and to compare standardized rates to crude asthma hospital admission rates by posta l zip code area of residence during the period of 1999. There were no significant diffe rences shown between crude and adjusted to both gender and race hospital admission rate s. There were no statistically significant differences and very strong correlations be tween both crude and ad justed by age (r=0.99, p<0.0000) and by gender (r=0.99, p<0.0000) asthma hospital admission rates by zip code area of residence. Race has shown some deviance and disclosed much weaker but still significant relationship between crude a nd adjusted hospitalization rates (r=0.89, p<0.001). Therefore, population-based small-ar ea aggregated data analysis and an estimated crude association between environm ental asthma triggers and socioeconomic status characteristics were not biased and confounded by indi vidual characteristics such as age, gender and race. Racial differences were accounted for and represented by the percentage value of ethnic minorities as perman ent residents living in the postal zip code area of residence within the study area. Although separate childhood and adult crude asthma hospitalization rate analyses disclo sed a very strong sta tistically significant association by postal zip code area, hospitalization rates for children were in average three-fold higher by separate zip code area of residence in 1997, 1998 and 1999.

PAGE 184

169 Therefore, data analysis wa s conducted separately for ch ildren younger than 15 years of age and adults of 15 years or older population groups. Two separate studies were conducted to explore the association between childhood and adult asthma hospital admissions and both environmental asthma triggers and local small-area socioeconomic status characteristics In the first study, the association between asthma hospital admissions and environmenta l asthma triggers was estimated by using simple correlation, simple log-linear regr ession and multiple regression analysis techniques over the period of 1997-1999. In th e second study, the spa tial distribution of asthma hospital admissions by postal zip code area of residence was evaluated in relationship to specific socioeconomic status characteristics and a complex multidimensional socioeconomic deprivation inde x of postal zip code area of residence. Separate comparative studies for each consecu tive year were desi gned to compare study results in 1997, 1998 and 1999. Relatively high and low environmental exposure areas by single zip code area of reside nce were estimated and define d to adjust for exposure to selected criteria ambient air pollutants and also to evaluate po ssible confounding and/or interaction term effect between selected soci oeconomic status characteristics and ambient air pollutants. Association between environmental trigg ers of asthma and hospital admission rates for childhood and adult asthma Previous epidemiological and clinical-bas ed studies suggested that environmental exposure to such criteria ambient air pollutant s as particulate matter, sulfur dioxide and ozone, to such meteorological conditions as temperature and humidity, and to such aeroallergens as tree, ragweed, grass pollen counts could be responsible for the increase in asthma prevalence and severe exacerba tion of asthma symptoms which would end-up in hospitalizations for asthma.27, 29 Our study indicated slight monthly and seasonal variation of environmental qual ity by selected criteria ambi ent air pollutants, ambient air temperature and pollen counts by calendar quarter over time in the Tampa Bay area, FL. Decrease in ambient air temperature was show n to be the only significant factor to

PAGE 185

170 explain an increase in asthma hospital admi ssion rates and also to explain the crude association between asthma hospital admissions and other environmental exposure triggers as a third factor because of highe r inter-correlation with these environmental factors. The average seasonal values of ambi ent temperature by calendar quarter, F, were significantly negatively asso ciated with hospitaliza tion for both childhood (r=-0.7, p<0.0001) and adult asthma (r=-0.95, p<0.0001). Simple and multiple log-linear regression analysis results supported this co nclusion. After an adjustment to selected ambient air pollutants and other environmen tal allergens, temperature was the only significant factor attributed to the increase in the number of asthma hospital admissions. An adult asthmatic group was more sensit ive to seasonal environmental changes as compared to children asthmatics. There was also a significant association of ambient temperature with environmental exposure to sulfur dioxide (r=-0.95, p<0.0000), total grass (r=0.76, p<0.0001) and tree pollen coun ts (r=-0.69, p<0.0001). Crude and adjusted associations between adult asthma hospita l admissions and both tree and grass pollen seasonal counts disclosed that the statisti cally significant crude association was not significant after adjustment to temperature and other environmental asthma triggers in the multiple linear regression model. The asso ciation could be explained by the strong significant correlation with ambient temperature which is not estimated in the crude single variable correlation analysis and c ould confound final conclusions if based only on crude analysis results. Because of a hi gh correlation among various environmental triggers, one should carefully draw conclusions on the crude associa tions between asthma hospital admissions and a single environmenta l factor. There was no association shown between total weed pollen exposure and hospital admissions for childhood or adult asthma disclosed over time in the study. Th ere was also no signi ficant association between the number of hospital admissions for both childhood and adult asthma and seasonal environmental exposure to ambient pa rticles, sulfur dioxide and ozone in both simple and stepwise backward selection multip le log-linear regression analyses over the total study period by calendar quarter. The stepwise backward-selection regressi on model building techniques were used to estimate the association between asth ma hospital admissions and environmental

PAGE 186

171 triggers of asthma after adjustment to po ssible confounders in th e model over the total period of study 1997-1999. The multiple regression analysis proved that ambient temperature was the only signi ficant factor associated w ith both childhood and adult asthma hospital admissions after adjustme nt for other possible risk factors and confounders. There were no statistically sign ificant interaction e ffects shown between temperature and other environmental asthma triggers in the regression model. Adult asthmatics were shown to be more sensitive to decrease in ambient temperature. Best-fit log-linear regression model parameter estima tes suggest that a s easonal decrease in ambient temperature by 10 F could account for up to 32% and 42.6% increase in the number of hospital admissions for childre n and adult groups respectively. A sudden decrease in temperature during the cold season could explain an increase in hospitalizations for childhood and adult asthma in the late fall and winter. Because of limited availability of indivi dual-based data, we could not estimate a daily or selected time-lag association between environmental exposure to asthma triggers and severe exacerbations of asthma, and coul d estimate a possible association based only on seasonal variations. The seas onal associations could not represent and could not be used to explain possible daily effects of am bient air pollution a nd other environmental asthma triggers on asthma. However, our st udy results strongly suggest that there are explanatory factors other than environmenta l exposure to ambient air pollution and aeroallergens which are more important in ex acerbation of asthma in the selected study area. Multiple regression analys is also suggests that one shou ld be very careful in trying to explain possible associati ons and draw final conclusions based only on the estimated crude associations and results of simple correlation an alysis. Advanced two-stage hierarchical models and adva nced statistical da ta modeling techniques were proposed to develop a standard approach for the assessment of the various chronic health effects of air pollution in the national multi-city analysis design.33-36 and should be used for future studies in order to evaluate the effect of daily changes in urban ambient air quality on hospital admissions childhood and adult asthma. John Samet, together with other colleagues from John Hopkins University, proposed an innovative approach by using Bayesian semi-parametric hierarchical models to estimate health effects over time within

PAGE 187

172 a selected city and to compare temporal pa tterns across cities a nd geographical regions.3335 Potential confounders were a ddressed in the aforementione d studies, but some of the existing limitations are obscuring interpreta tion and dissemination of study findings. The main limitations could be explained by bias in case definition, exposure assessment, and confounding effect of other possible risk fact ors. Very large geographical areas were represented with a limited number of air qua lity monitoring sites using an average air quality value at the air monitoring site as an indicator of a single individual personal exposure. The composition of particulate matter is an important characteristic of ambient particles and is known to vary by spatial domain. The composition of particulate matter in separate cities was not ev aluated and discussed in these studies. Correlation with an overall hierarchical model was not always supported by significant association within a single city. Study cases were defined and selected too br oadly to prove biological plausibility because investigators look at to tal or all cardio-respiratory mortality which makes causal interpretations of study findings very questionable. As concluded by the authors, county socioeconomic characteristic s do not reflect existi ng strong variations within the county and cannot represent indi vidual socioeconomic status. The ecological association could not be defined as causal and could be also explained by chance or another third factor. Our study extends previous work in two directions. First, the socioeconomic status was evaluated at smallgeographical areas dividing urban area or different counties into postal zip code areas of residence and includ ing different social status, economical and demographic characteri stics based on previous empirical studies and preliminary crude associa tion analysis results. Second, the accuracy of ambient air pollution exposure assessment was increased by spatial interpolation and modeling of geographically coded data. Instead of taking a simple arithmetic average of selected pollutant concentration at limited number of ambient air monitoring stations within a large geographical area we were able to es timate more accurate environmental exposure by small geographical area of residence by us ing spatial interpolation. Adjacent county areas were also used to increase the number of point sample concen tration measurements, and also to increase the accura cy of average environmental exposure assessment over the selected study area. Strongly limiting our study was the unavailability of data on daily

PAGE 188

173 hospital admissions for childhood and adult asth ma, which prohibited us from exploring daily or various time-lag associations be tween environmental exposure and asthma hospitalizations. The complex multidimensional socioeconomic deprivation index was also constructed by using standard principal component analysis techniques to represent complex composite socioeconomic status diffe rences by the area of residence. Future studies should be directed to evaluate the multi-city daily and various time-lag effects of environmental exposure to ambient air pollu tion, controlling for unbiased and accurately defined both small-area and indi vidual socioeconomic status characteristics and for other possible asthma triggers, such as home environment, behavioral and life-style characteristics, in the multiple regression model analysis. Preliminary evaluation of the geographical distribution of hospital admissions for total, childhood and adult asthma by zip code area of residence disclosed the cluster of annual asthma hospital admissions within spec ific geographical areas of residence which was consistent over the overall period of study in 1997, 1998, and 1999. The next step undertaken was to evaluate the association between diverse small-area socioeconomic status indicators and hospita lizations for childhood and adul t asthma by geographical area of residence after controlling for average surrogate estimate of environmental exposure to separate selected ambient air po llutant over the defined study area. Association between area socioecono mic characteristics and hospital admissions for childhood and adult asthma The second study was conducted to estimat e the association between different small-area socioeconomic status indicators and asthma hospital admission rates after controlling for environmental exposure to ambi ent air pollution over th e selected area of residence. The cluster of annual total, childhood and adult asthma hospital admissions was consistently high within the same zip code areas of residence, suggesting the importance of unknown spatially distributed a nd correlated risk f actors within these geographical areas. The cluster of asthma hospitalizations by selected postal zip code

PAGE 189

174 areas of residence was shown for both crude number and calculated rates of hospital admissions per 10,000 persons in 1997, 1998 and 1999. An estimated association between as thma hospital admissions and area socioeconomic status indicators was adjusted to average annual environmental exposure to selected ambient air pollutants over geogr aphical area of residenc e. Advanced spatial data analysis and modeling techniques were used (1) to provide a more accurate average surrogate estimate of environmental exposur e to ambient air pollution based on a limited number geographically located point-site measurements, and (2) to describe differences in environmental exposure to ambient polluti on over the selected area of study. Spatial interpolation techniques were applied to provide more accu rate validated prediction of average annual environmental exposure to ambi ent particles, sulfur dioxide and ozone by zip code area of residence over the tota l study area. Geographic Information Systems (GIS) provide a value-added tool for integr ated management, analysis and display of environmental, demographic, socioeconomic and public health information obtained from diverse data sources. GIS was used for the interpolation of geographically distributed sample point measurement data and spatial pr ediction of unknown environmental exposure to ambient particles, ozone and sulfur dioxi de by zip code area of residence. To use environmental quality data in community health studies, it is frequently necessary to predict unknown local envir onmental conditions and unknown environmental exposure based on limited sa mple point-site meas urements. Spatial geographically restricted lim ited data interpolation and modeling is an important objective of environmental exposure asse ssment in community health studies.38 GIS provide powerful analytical and visua lization techniques to environmental, socioeconomic and behavioral data that have a geographic or spatial component. A spatial interpolation model predicts unknown values of e nvironmental exposure over a selected geographical area from a limite d number of spatially located point measurements. A typical purpose of spatial da ta interpolation is to create a spatial elevation surface of environmental exposure and to define different environmental exposure areas over a selected study area from a set of limited sample point concentration measurements. Reliable and detailed spatial environmental exposure analysis and maps

PAGE 190

175 of environmental exposure over the study area in relation to health outcome or condition of interest can aid the devel opment of more effective rese arch and public health risk management policy by defining specific disa dvantaged geographical areas and specific sensitive population subgroups or commun ities. Interpolated ambient air pollution concentrations varied by singl e zip code area of residence an d allowed us to separate the total study area into relatively high and low exposure categories (str ata). Relatively high environmental exposure areas indicated more disadvantaged urban areas with constantly higher average annual exposure compared to re ference category areas of residence with relatively low environmental e xposure to selected criteria ambient air pollutants. The study incorporated various advanced statistical and spa tial data analysis techniques to overcome the most common obs tacles encountered in environmental and socioeconomic asthma studies. Limited initia l data, accuracy of environmental exposure assessment, confounding and effect medica tion (interaction), and multicolinearity (intercorrelation) issues ar e the most common problems in the development of proper study design and final interpretation of study results. The identification of potential confounders is usually based on a prior knowledg e of the dual association of the possible confounder with the exposure and the out come, the main two poles of the study hypothesis. Stratification and multivariate re gression analysis (modeling) are the main analytical tools used to control for confounding effect and to assess effect modification.63,64,65 Study results disclosed significant geographical differences in both total number and asthma hospital admissions ra tes by postal zip code area of residence over the study area. The association was cons istent and statistically significant when comparing the number and rates of hospita l admissions for both childhood and adult asthma over the three consecutive year s of study in 1997, 1998, and in 1999. Local variation in the socioecono mic status may be large w ithin geographical areas and consistent with the hypothesis of environmenta l risk. Under these circ umstances, there is marked potential for confounding in small-area anal yses of health data near local sources of higher environmental exposure.46 Possible confounding effect and effect modification (interaction) were evaluated by (1) stratified frequency table analysis of the association between hospitalizations for as thma and environmental exposure to ambient air pollution

PAGE 191

176 within selected socioeconomic status categ ory (strata) while adjusting to specific socioeconomic variable; and (2) multiple log-linear regression analysis (modeling) of the association between environmental exposure and asthma hospital admission rates while controlling for area socioeconomic deprivation status in the multiple regression model. Stratification is the simplest method to analyze possible pres ence of confounding, and also allows a straightforward examinati on of the possible presence of both confounding and effect modification. Stratifi ed analysis as a standard ep idemiological technique has a number of advantages: it is relatively easy to carry out; it permits the evaluation of effect modification; and perhaps the most important is that it allows both the investigator and the reader to achieve a clear understanding of the associati on among the exposure, health outcome of interest, and possible conf ounding or effect modifying variables.63, 64 A fundamental problem with stratified analys is, however, is its in ability to control simultaneously for even a moderate number of potential confounders or effect modifiers. Positive confounding effect of community soci oeconomic status by the percentages of single parent families with children, people li ving below poverty level, persons with low family income, and persons living in overcrowded housing conditions revealed lower stratum specific (adjusted) relative risk estimates as compare to relevant crude (unadjusted) estimates of the association be tween environmental exposure to ambient air pollution and asthma, and suggest the critical importance of various area socioeconomic status indicators trying to e xplain higher asthma hospitalization rates within specific low socioeconomic status areas of residence. Inte raction effect and heterogeneity of separate strata were present by some socioeconomic status indicators. Ba sed on heterogeneous effects within separate strata, the slight possible effect modification (interaction) was shown by poverty status and the score of Ur ban Deprivation Index (UDI) as a complex area socioeconomic status indicator on th e association between the exposure to particulate matters and children asthma hos pital admissions. There was decrease in magnitude and significance of the associa tion between environmental exposure to ambient particles and sulfur dioxide by c ontrolling for various socioeconomic status indicators in a given area of residence. Further multiple log-linear regression analysis results proved that stratified analysis is a straightforward but also very effective method

PAGE 192

177 to evaluate possible confounding effect by soci oeconomic status as a third factor in environmental exposure to ambient air polluti on and asthma studies. Multiple regression analysis allows for the efficient estimation of measures of associat ion while controlling for a number of confoundi ng factors simultaneously, even in situations where stratification would fail because of insuffici ent numbers. Multiple regression analyses refers to a series of analytical techni ques, each based on a more or less complex mathematical model, which are used to carry out statistical adjustme nt and “controlling” for one or more possible confounding variables.63,64 Multivariable analysis involves the construction of a mathematical model to de scribe most efficiently the association between exposure and health outcome of inte rest as well as other variables that may confound the effect of environmental exposur e. The most common way in what many factors are controlled for simultaneously is through the use of a multiple regression model. Simple correlation, simple log-linear re gression and frequency table analysis techniques were used to evaluate crude association between ch ildhood and adult asthma hospital admissions and both socioeconomic ar ea status and environmental exposure to ambient particulate matter, ozone, and sulfur di oxide. Data analysis by zip code area of residence revealed significant association of environmental exposure to ambient particles with childhood (correlation coefficien t r=0.53, p<0.001) and adult (r=0.48, p<0.001) asthma hospital admission rates in 1999. Ther e was also significant association between ambient air pollution by sulfur dioxide and hospital admissions for both childhood (r=0.59, p<0.001) and adult asthma (r=0.36, p<0.05) by zip code area. However, there was no association shown between ambi ent air pollution by ozone and hospital admissions for both childhood (r=-0.17, p>0.05) and adult (r=-0.18, p>0.05) asthma by zip code area of residence. The simple log-linear regression m odel analyses also supported correlation analysis results. The comparative simple log-linear regression models were designed for 1997, 1998, and 1999. Am bient particles and sulfur dioxide were significant predictive factors of hosp ital admissions for childhood and adult asthma in 1997, 1998, and 1999. Ozone was shown to be a significant predictor of both childhood and adult asthma hospitalizations in 1997, but not in 1998 and 1999. The

PAGE 193

178 results of multiple log-linear regression model analyses of selected criteria ambient air pollutants suggest different interpretations for childhood and adult hospital admissions. Sulfur dioxide was more important to expl ain an increase in hospitalizations for childhood asthma, while ambient particles were the only significant factors to account for the increase in hospital admissions for adult as thma after controlling for other ambient air pollutants influence. Ambient particles toge ther with sulfur di oxide were significant factors to predict childhood asthma hosp ital admissions in 1999, while only sulfur dioxide was a significant fact or to explain hospitalization rates of childhood asthma by zip code area of residence in 1997 and 1998. Am bient particles were the only significant predictor of hospitalizat ions for adult asthma after cont rolling for sulfur dioxide and ozone exposure in multiple log-linear regression equations in 1997, 1998, and 1999 accordingly. The comparison of relative high and low exposure areas by conducting frequency table analysis confirmed previous results of simple correlation and simple loglinear regression analyses. Ra te ratios of asthma hospita lizations in relatively high exposure to ambient particles as compared to relatively low level exposure areas was significant for both children (R ate Ratio of 1.8 with 95% CI: 1.5-2.1) and adult (Rate Ratio of 1.5 with 95% CI: 1.33-1.77) populat ion groups respectively. Environmental exposure to sulfur dioxide may be responsible for 140 and 39 percent of excess rate ratio fractions of asthma hospital admissions base d on crude analysis in children and adult population groups in relatively high e xposure areas accordingly. There was no association shown between environmental exposure to ozone and asthma hospital admissions both over time and over space by zip c ode area of residence. Our preliminary results of crude association analyses suppor t previous studies wh ich found that even relatively low levels of air pollution (level s below current air pollution guidelines) by PM10 and SO2 may be associated with adverse hea lth effects, including daily hospital admissions for respiratory conditions.20-23, 34-36 However, many previous air pollution and asthma studies found only small statistically significant differences in crude excess risk or rate estimates attributable to environm ental exposure by ambient air pollutants which could be overestimated and confounded by the an other third factor. Further data analysis included the evaluation of environmental exposure influence after controlling for

PAGE 194

179 socioeconomic status indicators and complex multidisciplinary socioeconomic deprivation index. More profound and appropriate definition of socioeconomic area status must be developed for future environmental asth ma studies. Analysis of various area socioeconomic status indicators rev ealed strong correlation among different socioeconomic status characteristics. A lthough many of the area so cioeconomic status characteristics were shown to have strong si gnificant association with hospital admission rates for childhood and adult asth ma, the interpretation of si mple correlation analysis results is very complicated because of the high multicolinearity or inter-correlation among different socioeconomic status indi cators. Area-based socioeconomic status indices have been thoroughly us ed to analyze health dispariti es in Europe, Australia, and New Zealand. Despite numerous cross-sectiona l studies of area soci oeconomic disparities and community health, the monitoring of morb idity and mortality trends by small-area socioeconomic characteristics still remains fa r less common in the US than in Europe, Australia or New Zealand.41 A research group from the Di vision of Cancer Control and Population Sciences at National Institutes of Health (NIH), Bethesda, MD, developed and utilized a comprehensive composite area-ba sed measure of socioeconomic status to examine the extent to which area socioeconom ic deprivation is linked with county-level all-cause and cardiovascular mortality duri ng the period of study from 1969 to 1998. The socioeconomic deprivation index was cons tructed by applying principal components analysis methods and techniques. Standa rd principal component analysis (PCA) techniques were used to reduce the numb er of significant socioeconomic status characteristics and to develop a complex multidimensional area socioeconomic deprivation index which could incorporate the most significant explanatory indicators.41 The same index development and validation tech niques were used in our study in order to develop a complex multidimensional small-area socioeconomic deprivation index. The percentages of people living below poverty level, people with education level of 9th grade or less, people with an annual household in come of $15,000 or less, and people living in overcrowded housing conditions had the highest relative we ight in the area socioeconomic deprivation index. The pe rcentages of unemployed people, people

PAGE 195

180 employed in professional and managerial occ upations (white-collar occupations), people without any vehicle in the family, and also people living in a hous e lacking kitchen or plumbing facilities had relative low standa rdize weights in the index. Correspondingly, percentages of single households with child ren (single parent families), ethnic minority householders, people living in older houses, and in dwellings which are heated by fuel as the main source of heat were significantly associated with both childhood and adult asthma hospital admissions, but had no importance in representing complex multidimensional area socioeconomic depriva tion status. The higher index score value indicated lower socioeconomic status and mo re severe area socioeconomic deprivation. Area Socioeconomic Deprivation Index (SDI) was shown to be significantly associated with total asthma hospital admission rates by zip code area of residence in 1997 (r=0.8, p<0.0001), in 1998 (r=0.81, p<0.0001), and in 1999 (r=0.82, p<0.0001) respectively. The same consistency over time and strong signifi cant association was also shown between complex area socioeconomic deprivation i ndex and hospitalizati ons for both childhood and adult asthma by area of re sidence studied separately. Simple log-linear regressi on analysis results supporte d the conclusion that the complex area socioeconomic depr ivation index could be used to explain the increase in asthma hospital admissions and to predict asth ma hospitalization rates by zip code area of residence in the study area. In a single variable simple log-li near regression analysis, area socioeconomic deprivation status index wa s a significant predictor of both childhood and adult asthma hospital admission rates by area of residence in 1997, 1998, and 1999 correspondingly. The estimates adjusted to environmenta l exposure to ambient air pollution suggested a significant effect of re latively higher sulfur dioxide exposure for childhood asthma hospitalizations, and of ambi ent particles for adult asthma hospital admissions during the period of study 19971999. The area socioeconomic deprivation index was a significant factor after controlling for environm ental exposure to selected criteria ambient air pollutants by zip code area of residence in all multiple regression models for both childhood and adult as thma hospitalizations in 1997, 1998, and 1999. Area socioeconomic deprivation status was th e only significant inde pendent variable in the childhood asthma log-linear multiple regr ession model in 1998, and in the adult

PAGE 196

181 asthma model in 1999. Crude association be tween ambient air pollutants and hospital admissions for childhood and adult asthma wa s not significant after the adjustment to complex area socioeconomic deprivation inde x in the multiple regression models. There was no significant interaction be tween environmental exposure to ambient air pollution and area socioeconomic deprivation index reve aled in the regression models by including the interaction term in the final best-fit mode l. Separate census-based area socioeconomic status characteristics were also analyzed in the multiple regression models for childhood and adult asthma controlling for environmen tal exposure to ambient air pollution. All significant socioeconomic status factors based on crude simple correlation analysis were included in the multiple log-linear regression models after adjusting to environmental exposure. The percentages of people living below poverty level; people with education level of 9th grade or less; people with an annua l household income of $15,000 or less; people living in overcrowded housing condi tions; unemployed people; people employed in professional and managerial occupations (white-collar occupa tions); people without any vehicle in the family; people living the house with lacking kitchen or plumbing facilities; single households w ith children (single parent families); ethnic minority householders; people living in older houses; a nd people residing in dwellings which are heated by fuel as the main source of heat represented area socioeconomic status indicators in the multiple re gression model. Environmental exposure was represented by average annual exposure to ambient partic les, sulfur dioxide and ozone by area of residence. Stepwise backward selection procedures were used to obtain the final best-fit models for childhood and adult asthma. Final models were constructed separately for childhood and adult asthma hospital admissions in 1997, 1998, and 1999. Only environmental exposure to sulfur dioxide out of all selected criteria ambient air pollutants, included in the multiple regression model analyses, was shown to be a significant independent fact or for childhood asthma hospital admissions in 1997 and 1999, but not in 1998. Based on other final best-f it model analysis results we saw that environmental exposure was not a significant predictor of asthma hospital admissions after controlling for area so cioeconomic status. These fi ndings supported our previous study of ambient air pollution and other e nvironmental asthma triggers by calendar

PAGE 197

182 quarter, which indicated that the seasonal changes in ambi ent air quality by ambient particles, sulfur dioxide and ozone were not associated with asthma hospitalizations during the period of 1997-1999. Poverty, profes sional and managerial occupation, no vehicle available, living c onditions in overcrowded housing conditions, single parent families with children, and houses using fuel as a main source for heat were found to be significant socioeconomic status characteristi cs associated with hospital admissions for childhood and adult asthma during the period of study. Poverty, no vehicle available, single parent families with children, houses heated by fuel were positively associated, while professional and manageri al occupation and living in overcrowded conditions had a significant inverse association with hospita lizations for childh ood and adult asthma. Complex multidimensional area socioeconomic deprivation index incorporated all of these single socioeconomic characteristics except the percentage s of single parent families with children and houses heated with fuel. Residual deviance analyses allowed us to identify extreme cases and to evaluate the magnitude of influence on the log-linear regression model parameters. Outlier analysis did not show significant effect on any of the log-linear regression model parameters and were assume d to be not significant in the final interpretation of regression model para meter estimates. The final best-fit area socioeconomic deprivation index and hospita l admissions for childhood and adult asthma models were validated by using independent data outside of our study area for the year 1999. Model validation results confirmed that both models provide accurate prediction and could be used with independent data outside of the study area.

PAGE 198

183 Chapter 4 FINAL CONCLUSIONS Major findings Our study explored various lo cal environmental exposure and socioeconomic status characteristics, which could explain exis ting geographical differences in asthma distribution at small-area level. The study provided valuable information about geographical differences in hospitalization ra tes for asthma by residential area, and changes in the number of asthma hospital admi ssions over time in the selected study area. The dramatic increase in asthma morbidity and geographical area-based clustering of asthma hospital admissions within specifi c sensitive population groups have been observed during the last few decades. However, there is still limited asthma surveillance conducted at the national level and there is no continuous permanent disease surveillance conducted at the local community or state le vel. The complex small-area socioeconomic deprivation index represented multidimensiona l socioeconomic status of the selected geographical area of residence. A possible interaction and confounding eff ect with environmental exposure to ambient air pollution were explored in the mu ltiple regression models. The best-fit model for childhood asthma reflected area soci oeconomic deprivation index and average environmental exposure to ambient air pollution by sulfur dioxide as the main significant explanatory factors. Single factor regression analysis and further model validation disclosed that area socioeconomic depriva tion index could be used a significant individual independent predictor of childhood asthma hospital admissi on rates. In support of this conclusion, multiple regression model analysis of both childhood and adult asthma hospital admissions indicated that area soci oeconomic deprivation index is a single significant predictor variable of adult asthma hospital admi ssion rates after adjusting to

PAGE 199

184 environmental exposure to selected criteria ai r pollutants by area of residence in 1998 and 1999. Based on the transformed log-linear regr ession model parameter estimates we can conclude that an increase in the complex multidimensional area socioeconomic deprivation index by 1 index unit may expl ain an increase by 7.9%, or increase by 3 persons per 10,000 children for childhood asth ma hospitalizations, and 7.5% or by 1 person per 10,000 adults for adult asthma hos pitalization in Hills borough County, FL, in 1997. Accordingly, the increase of residential area socioeco nomic deprivation index by 1 unit could explain the increas e in asthma hospital admissi ons by 8.3% or by 2 per 10,000 persons in children, and 7.2% or 1 per 10,000 pe rsons in adults in 1998; and 7.7% or by 3 per 10,000 persons in children, and 6.7% or by 1 per 10,000 person in adults in 1999. Although sulfur dioxide was a significant indepe ndent variable in the multiple regression model of childhood asthma in 1997 and 1999, the results of simple and multiple regression analyses of seasonal environmenta l exposure to aforementioned criteria air pollutants by calendar quarter proved that ai r pollution by ambient particles, sulfur dioxide and ozone are not signi ficant factors in trying to ex plain an increase in asthma hospitalization in the study ar ea. The regression model valida tion techniques were used to explore the predictive accuracy of the final model, with so cioeconomic deprivation index as an independent variable, by using i ndependent data for both childhood and adult asthma models outside of our study area. Mode l validation results sugge sted that the final best-fit childhood and adult asthma predic tive models could be used to identify geographic differences in the distribution of disease and to disclose specific sensitive target population groups within selected ge ographical area of re sidence. Developed predictive regression models could provide a valuable tool for local health resource managers and policy makers for more eff ective health care resource management and asthma prevention policy developmen t at local community level. Limitations The main study limitation could be explained by limited availability of individualbased data and also our inability to estim ate personal environmental exposure to main criteria ambient air pollutants. Because of aggr egated data analysis, there is a possibility

PAGE 200

185 of confounding effect by such individual ch aracteristics as age, gender and race, and future studies based on individual data anal ysis are recommended to support or oppose our study results. There was no information on daily data on asthma hospital admissions available to explore the direct acute effect of ambient air pollu tion on exacerbation of disease symptoms. We used aggregated data to evaluate changes in asthma hospital admissions over time and to explore spatial distribution of asthma hospital admissions over the geographical area of study. Due to limited data on daily asthma hospitalizations we were not able to provide analysis of daily or selected time-lag associations between ambient air pollution and hosp ital admissions for childhood a nd adult asthma. Relatively low levels of ambient air pollution concentra tions by ambient particles, sulfur dioxide and ozone below current federal and state stan dards, and short range of variations in pollutant concentrations were shown in our study. These factors obstructed our definition of relatively high and low exposure to ambien t air pollution areas of residence. However, our study results indicated no seas onal association between sele cted criteria air pollutants and hospitalizations for asthma, and strong significant effect of both diverse area socioeconomic status characteristics and co mplex area socioeconomic deprivation index on hospital admissions for childhood and adu lt asthma. We also used surrogate environmental exposure measures at small ge ographical area level and could not avoid possible bias in environmental exposure as sessment. Final result interpretation and conclusions about the influence of ambient air pollution on asthma hospitalization should be presented very carefully, and future studi es should be directed on the evaluation of possible daily or selected time-lag associat ions to support or oppose our study findings. Consistency with other studies Calculated asthma hospitalization rates per 10,000 persons were consistent with previous asthma prevalence and hospital admi ssions for asthma survey results. Hospital admission rates for asthma per 10,000 were 15.9 in 1997, 13.3 in 1998, and 17.1 per 10,000 persons in 1999 in Hillsborough County, FL, respectively. The average number of asthma hospitalizations ove r the total period of study was 15.4 hospitalizations per 10,000. CDC National Center for Health Statistic s reported the average number of 17 per

PAGE 201

186 10,000 hospitalizations for asthma in the US.7 National Hospital Discharge Survey of asthma hospitalization rates by different geographical regions reported 15.8 hospital admissions per 10,000 persons in the S outh region. There were 14.2, 18.4 and 24.5 hospitalizations per 10,000 re ported in West, Midwest a nd Northeast correspondingly.2 Hospitalization rate for childhood asthma was fo und to be three-fold higher as compared to hospital admission rates for adult asthma. Our study also supports results of previous studies where asthmatic children were found to be more susceptible to ambient air pollution exposure because of more active outdoor physical activities and existing physiological differences from adults.71 Environmental exposure to ambient air pollution by sulfur dioxide was found to be a significant explanatory factor for children but not for the adult group in the multiple regression analysis. Other selected criteria ambient air pollutants were not significant predictors of hos pitalization for asthma in the final fitted multiple regression models in 1997, 1998 and 1999 respectively. Several previous studies have focu sed on the association between low socioeconomic status and asthma.46 Our study results support many previous studies which suggested that census-ba sed socioeconomic deprivation status indices could serve as a powerful tool for describing and monitori ng social inequalities in a given community over time.41 The area socioeconomic deprivation index provides a summary description of diverse social and economic condition in an area, and can therefore be used to assess the effects of specific preventive interventions or risk management policy programs designed to target more vulnerable communities and to reduce existing health disparities. Diverse community socioeconomic status indice s can also provide useful information in the context of health res ource planning and allocation.38 Numerous plausible biological and indirect or non-causal mechanisms were previously suggested to explain a higher prevalence of asthma in subj ects living in the re sidential areas of low socioeconomic status.48,53 Specific material and social charact eristics of people at different socioeconomic levels may contribute to differe nces in environmental exposure patterns, health behaviors, and risks of developing certain disease.68 Complex socioeconomic status index is a comprehensive indicator that refers to a broad range of factors, such as levels of social standing, inco me, education, and living conditions.68 Area socioeconomic

PAGE 202

187 status may affect health through structural and material factor s as well as sociobehavioral variables. Differ ential residence in more di sadvantageous environmental conditions related to asthma across different socioeconomic levels could contribute to risk modification. Proven exposure to i ndoor dust mites, cockroaches, and mouse allergens, viral respiratory infections, and maternal smoking are closely associated with asthma in children, and these factors exhi bit substantial socioeconomic gradient by geographical area of residence. There ar e different ways in which poverty might predispose to severe asthma attacks includi ng more intense allergenic exposure in the residence place (especially to cockroach antigen); higher ra tes of cigarette smoking and environmental tobacco smoke exposure; greater exposure to indoor and outdoor environmental pollution; fewer resources to modify and improve existing home environment; and reduced availability and use of health care resources. The causal association and geographical vari ations could be at least in part explained by significant variation of indoor allergens and antigens across different socioeconomic strata. Measures of exposure to house dust mites, co ckroaches or home pet allergens were not undertaken in this study but it has been show n in the review of previously reported studies that increased exposure to these alle rgens may result in increased sensitization, and also act as an exacerbating causal risk factor for asthma.68 Previous epidemiological studies also disclosed that cockroach allergen is more common in homes of relatively low socioeconomic status expl ained by high poverty level.52 The house dust mites are known to thrive in poor ventilation and damp condi tions. Both dust mite numbers and allergen levels have been shown to in crease with higher indoor humidity and are indicators of dampness in the home.68 Low socioeconomic status is associated with low income and short education was found to associated w ith poor health of asthmatic patients.46 Socioeconomic status factors and social char acteristics specific to ethnic minority groups were found to be powerful pred ictors of disease at both the individual and ecological small-area level.3 Belonging to a minority race, lowe r household income, lower education level, inadequate social support, living in a lacking facilities urban setting, ineffective use of medications, noncompliance a nd poor insight into the disease treatment, and certain cultural health beliefs are some of the variables repeatedly f ound to be related to personal

PAGE 203

188 knowledge about asthma control and manageme nt, and also to health care resource utilization patterns in patients with asthma A significant associ ation was confirmed between lower education level and matern al smoking during pregnancy and parental smoking after the baby was born.46 Indicators of dampness a nd poor ventilation were also more common in homes of those pare nts with low education attainment.46 The aforementioned study results have shown that the education level has an important influence on predisposition of risk factor s for the development and exacerbation of asthma and indicated a deeper understanding of lifestyle and behavioral factors in low socioeconomic status population groups. The pr evious study of the health-related quality of life of children with asthma69 revealed that low income and living in poverty was significantly associated with higher prevalence, severity and more restricted resource utilization for asthmatic patients. The st udy results highlighted our study finding and concluded that socioeconomi c status (low education level, unemployment, family income, and no health insurance available) were significantly related to health-related quality life of asthmatic children and adults. Previous epidemiological studies have also implied that the causes of higher morbidity and mortality for asthma may be more associated with poverty status and living in more socioeconomic disadvantaged urban environment than ethnic minority race alone.3,46 Public health importance The results of small-area analyses should not be applied to e xplain and confirm or reject causal association betw een exposure and disease at individual level. The study was designed to explore the spatia l distribution of asthma hospital admissions and evaluate possible association with area-based risk factors specific to these more sensitive population groups rather than to ascertain caus al effect and causal risk factors related with the development of asthma. Previous st udies identified that severe exacerbation of asthma resulting in hospital admission for this medical condition are clustering within specific geographical areas and could be explained by demographic, socioeconomic, environmental and behavioral or life-style characteristics specific to that area and population groups living in the study area.38 To develop more effective preventive asthma

PAGE 204

189 prevention programs and distribute available health care resources more effectively we need to define more vulnerable populat ion groups and to understand complex demographic, socioeconomic, environmenta l and behavioral factors within these communities. Described distribution of hosp italizations for childhood and adult asthma and local environmental and socioeconomi c triggers associated with asthma hospitalizations could be utili zed by local and state public heal th professionals and health resource managers to target the growing as thma epidemic within more vulnerable local communities. The scientific novelty and a dded value of our study could be also represented by advanced spatia l data analysis and modeling techniques, and an innovative approach could be used to describe complex multidimensional small-area socioeconomic deprivation status and also to estimate possible confounding and/or interaction effect between environmental exposure and soci oeconomic status characteristics. The most important novel features in our study are: (1) complex analysis of local environmental and socioeconomic characteri stics through the applic ation of advanced spatial data interpolation and modeling techni ques to increase the accuracy of predicted surrogate environmental exposure to ambient air pollution over th e study area; and (2) advanced statistical data anal ysis techniques to estimate the association between ambient air pollution and asthma hosp italizations while controlling for socioeconomic status. We explored diverse area socioeconom ic status indicators and also used standard statistical analysis techniques to develop a complex multidimensional area socioeconomic deprivation index based on census-derived socioeconomic characteristics. Multiple regression analysis results revealed that final conclusions made solely on crude association between ambient ai r pollutants and asthma hosp italizations could be biased and overestimated without accurate definition of socioeconomic status and further adjustment to (or standardization for) soci oeconomic status characteristics. Defined significant area-based socioeconom ic status characteristics could be employed by asthma policy makers, public health practitioners, and health care resources managers or providers at community-based level in th e presented study area. Poverty, unskilled workers, single parent families with childre n, living in less populated households with no vehicle available for personal usage and in the house heated with fuel are the main

PAGE 205

190 socioeconomic characteristics, which shoul d be addressed through more effective community-based preventative programs a nd more efficient health care resource management. The crucial aspect in our st udy was the utilizatio n and validation of predictive regression models of childhood and adult asthma, which could be applied to identify specific high risk areas and sensitive population groups in or der to develop more effective asthma management and preventive pol icy strategies for the most deprived and vulnerable community groups. Developed predictive models and small-area socioeconomic status characteristics coul d be used by local h ealth care services providers, health care resource managers and health policy developers to define more vulnerable inner-city areas and to assure more effectively health care resources management and asthma prevention programs at the local community or state level by utilizing only census-based available socioeconomic info rmation. Although all of the variables used to represent socioeconomic deprivation characteristics of the aggregate population entities (postal zip code areas) ar e easily derived from census-based data, many socioeconomic variables are not quite re ducible to the level of an individual. Means, percentages, rates, and numbers de scribing the distribution, composition, and size of the population are primary ch aracteristics of aggregates. Similarly, asthma hospital admission rates are characteristics of a co mmunity or a population, but not individuals. An explanation of the variati on in such rates therefore requi res a focus on the effects of social structural and population-level determinants.38 Such a viewpoint is consistent with the notion that it is ‘in the na ture of society itself that we must seek the explanation of social life’.38 As previous studies suggested, variations in customs, life-style, behavioral, physical environment, and social status ch aracteristics may very well be the most important explanations of popula tion variation in disease rates.38 The identification of major determinants of the socioeconomic gr adients and narrowing the socioeconomic gap between affluent and disadvantaged areas has the potential to substantially reduce mortality and morbidity for childhoo d and adult asthma in the US.41 The area socioeconomic deprivation index could be suc cessfully applied to pr ovide more effective preventive interventions for asthma manageme nt and prevention within specific high risk geographical areas and population groups. Fina lly, it can be used as a summary complex

PAGE 206

191 area socioeconomic deprivation status m easure to examine the net impact of environmental conditions, cultural, behavioral and life-style, and health care factors on asthma prevalence and severity of other impor tant chronic illnesses in future studies. Future studies directed to estimate the significance of small-area socioeconomic deprivation status and to id entify geographical areas with specific sensitive population groups within these areas could be a valuab le resource of import ant information and a noteworthy tool used to deve lop more effective preventive interventions at the local community level. Future directions In October 1999, the Robert Wood Johnson Foundation funded the Rand Health Program with a primary goal to evaluate and to outline future policy directions for childhood asthma in the United States. The progr am engaged an interd isciplinary expert panel group of nationally r ecognized leaders in childhood asthma, and identified 11 policy recommendations that form a compre hensive framework for achieving one broad future policy objective: to promote the de velopment of asthma-friendly communities where children’s environment is safe from physical and social environment risk and where children receive appropr iate continuous health care.70 Six more specific policy goals were identified to reach this objective:70 1) Reduce socioeconomic disparities in childhood asthma outcomes; 2) Improve access to and quality of asthma health care services in a given community; 3) Improve asthma awareness and knowledge about disease management among asthma patients and the general public; 4) Encourage innovations in asthma prevention; 5) Ensure asthma-friendly school environment; and 6) Promote asthma-safe home environment. Our study results are consistent with and support the conclusions of the aforementioned interdisciplinary asthma expert panel group. In order to reduce socioeconomic disparities and consequently improve quality of asthma health care

PAGE 207

192 services, future asthma prevention interven tions and risk management programs should address population groups described by su ch socioeconomic status deprivation characteristics as poverty, unskilled worker s, single parent families with children, families having no vehicle available, people livi ng in less crowded households or socially excluded conditions without adequate family members or relatives support, and residing in houses heated by fuel. Development of more effective asthma prevention programs and delivery of more efficient health care se rvices or resources should be based on the understanding of complex soci oeconomic, environmental an d behavioral asthma risk within communities and community-based di sease risk management and prevention strategies. Further stud ies could be suggested to confirm or disapprove the significance of highlighted area socioeconomic status characteristics in our study. Crude analysis results should not be used to describe the associ ation or developed preliminary conclusions, which could be biased and overestimated becau se of the other third factor or confounder not included in the study. Further environmen tal asthma studies should always include accurate definition of residential area and pr eferably individual socioeconomic status indicators in the adju sted analysis. Future studies s hould also explore the association between individual and residential area socio economic characteristics and try to explain the relationship between soci oeconomic status characteristics and asthma hospital admissions at both individual and community level. Collected individual-based information could provide additional evidence in support of community-based health care resource allocation for asthma and disease management to decrease current socioeconomic inequities and much higher asthma morbidity and mortality rates among specific inner-city population groups. The ideal model of future environmental asthma study would include accurate environmental exposure assessment by providing direct personal exposure to ambient air pollu tion assessment and preferably by using biomarkers of exposure. More accurate environmental exposure assessment in accordance with direct individua l respiratory function assessment could be implemented in a short-term longitudinal small panel group study by controlling or adjusting to both individual and residential area socioeconomic status differences. Two different panel groups representing different so cioeconomic status areas of residence could be compared

PAGE 208

193 to understand physical, social, cultural, life-sty le and behavioral factors which are crucial in asthma management and which are speci fic to different community socioeconomic status stratum.

PAGE 209

194 References 1. Trends in Asthma Morbidity and Mo rtality. American Lung Association. Epidemiology and Statistics Unit, Research and Scientific Affairs. April, 2004, www.lungusa.org (15/10/04). 2. Mannino MD, Homa DM, Petrowsk i CA, Ashizawa A, Nixon LL, Jahnson CA, Ball LB, Jack E, Kang DS. Surveillance for Asthma – Unites States, 1960-1995. CDC, MMWR 1998;47:1-28. 3. Sandel M and O’Connor G. Inner-city asthma. Immunol Allergy Clin N Am. 2002;22:737-752. 4. A Close Look at Asthma. Asthma and Allergy Foundation of America/National Pharmaceutical Council., www.aafa.org (15/10/04). 5. Centers for Disease Control and Pr evention. Forecasted state-specific estimates of self-reported asthma prev alence-United States, 1998; MMWR Morbid Mortal Wkly Rep. 1998;47(47):1022-1025. 6. Asthma and Allergy Statistics: Data Fact Sheet. National Heart, Lung and Blood Institutes of Health, US Depart ment of Health and Human Services: Washington DC, 1999. 7. Facts on Asthma. CDC National Ce nter for Health Statistics www.cdc.gov/nchs/fastats/asthma.htm 2004 (15/10/04).

PAGE 210

1958. CDC Mortality and Mortality Weekly Report. Self-reported Asthma Prevalence among AdultsUnited St ates, 2000. JAMA. 2001;286(13):1003-12. 9. Centers for Disease Control and Prevention, Asthma mortality and hospitalization among children and young adults, 1980-1993. MMWR Morbid Mortal Wkly Rep. 1996;45(17):350-353. 10. CDC. National Center for Health Statistics. New Asthma Estimates: Tracking Prevalence, Health Care, and Mortality. http://www.cdc.gov/nchs/products/pubs/ pubd/hestats/asthma/asthma.htm (01/10/04). 11. Homa DM, Mannino DM, Redd SC Regional differences in hospitalizations for asthma in the United States, 1988-1996. J Asthma. 2002;39(5):449-455. 12. Eggleston P.A, Buckley T.J, Breysse P.N. et al. The Environment and Asthma in US Inner Cities. Envir on Health Perspect 1999;107(3):439-450. 13. Castro M, Schechtman KB, Halstead J, Bloomberg G. Risk factors for asthma morbidity and mortality in a large metropolitan city. J Asthma. 2001;38(8):625-35. 14. Boss LP, Kreutzer RA, Luttinger D, et al. The public health surveillance of asthma. J Asthma 2001;38(1):83. 15. Mutius E. Worldwide asthma epidem ics. Immunol Allergy Clin N Am. 2002;22:701-711.

PAGE 211

19616. Malveaux F.J, Fletcher-Vincent S.A, Environmental Risk Factors for Childhood Asthma in Urban Centers. Envi ron Health Perspe ct, 1995;103(6):59-62. 17. Bearer CF. How Are Children Differe nt from Adults? Environ Health Perspect 1995;103(6):7-12. 18. Moeller DW. Environmental Health. Ha rvard University Press, Cambridge MA 1997, 475pp. 19. Plopper CG, Fanucchi MV. Do urban environment pollutants exacerbate childhood lung diseases. Environ He alth Perspect 2000;108(6):A252-A253. 20. Neukirch F, Segala C, Le Moullec Y. Short-term effects of low-level winter pollution on respirator y health of asthmatic adults. Arch Environ Health 1998;53(5):320-8. 21. Petroeschevsky A, Simpson RW, Tha lib L. Association between outdoor air pollution and hospital admissions in Br isbane, Australia. Arch Environ Health 2001;56(1):37-52. 22. Jedrychowski W, Flak E, Mroz E. Th e adverse effect of low levels of ambient air pollutants on lung function grow th in preadolescent children. Environ Health Perspect 1999;107(8):669-74. 23. Desqueyroux H, Pujet JC, Prosper M. et al. Short-term effects of low-level air pollution on respiratory health of a dults suffering from moderate to severe asthma. Environ Res 2002;89:29-37.

PAGE 212

19724. Wallace LA, Mitchell G, O’Connor GT. Et al. Particle Concentrations in Inner-City Homes of Children with Asth ma: The Effect of Smoking, Cooking, and Outdoor Pollution. Environ H ealth Perspect 2003;111:1265-1272. 25. Janssen NAH, Hoek G, Harssema H. Personal Exposure to Fine Particles in Children Correlates Closely with Am bient Fine Particles. Archives of Environmental Health 1999;54(2):95-101. 26. Bush KR. Environmental Asthma. Lung Biology in Health and Disease, Vol.153. Marcel Dekker, 2001, 370 pp. 27. Fin PW. Asthma and the Environment. Immunology and Allergy Clinic of North America. WB Saunder C o., November 2002, Vol.22(4), 788pp. 28. Delfino R.J. Epidemiologic Evidence for Asthma and Exposure to Air Toxics: Linkages between Occupationa l, Indoor, and Community Air Pollution Research. Environ Health Perspect 2 002;110(4):573-589; Desqueyroux H, Pujet JC, Prosper M. et al. Short-Term Effect of Low-Level Air Pollution on Respiratory Health of Adults Suffering from Moderate to Severe Asthma. Environ Research 2002;89:29-37. 29. Forsberg B. Urban air quality and indicators of respiratory problems. Umea University Medical Dissertatio n. New Series No522. Umea, Sweden 1997, 77pp. 30. Brunekreef B, Dockery DW, Krzyzanow ski M. Epidemiologic studies on shot-term effects of low levels of major ambient air pollution components. Environ Health Perspect 1995;103 (Suppl 2):3-13.

PAGE 213

19831. Steenland KK, Savitz DA. Outdoor Air III: Ozone. Topics in Environmental Epidemiology. Oxford Univ Press. 1997, 363pp. 32. D’Amato G, Liccardi G, D’Amato M, Cazzola M. The Role of Outdoor Air Pollution and Clim Resp Med 2001. 33. Samet JM, Dominici F, McDermott A, Zegn SL. New problems for an old design: time-series analysis of ai r pollution and health. Epidemiology 2003;14(1):11-12. 34. Dominici F, Zanobetti A, Zeger SL, Swartz J, Samet JM. Hierarchical bivariate time-series models: a combined anal ysis of the effect of particulate matter on morbidity and mortality. Bios tat 2004 (submitted for publication, www.biostat.jhsp.edu/~fdominici/research.html 10/15/2004). 35. Samet JM, Dominici F, Curriero F et al. Particulate air pollution and mortality: findings from 20 US citie s. New Eng J Med 2001;343(24):1742-1757. 36. Peng RD, Dominici F, Pastor-Barriusio R et al. Seasonal Analyses of air pollution in 100 US cities. Am J Epidem 2004 (submitted for publication, www.biostat.jhsp.edu/~fdominici/research.html 10/15/2004). 37. Last JM. Public Health and Hu man Ecology. Appleton& Lange, 2nd edition, 1998:464pp. 38. Elliott P, Cuzick J, English D, Stern R. Geographic and Environmental Epidemiology: methods for small-area st udies. Oxford Univ Press, 2000, 382pp. 39. Cesaroni G, Farchi S, Davoli M. et al Individual and area -based indicators of socioeconomic status and childho od asthma. Eur Respir J 2003;22(4):619-24.

PAGE 214

19940. Basagana X, Sunyer J, Kogevinas M. et al. Socioeconomic Status and Asthma Prevalence in Young Adu lts. Am J Epidemiol 2004;160:178-188. 41. Singh GK, Siahpush M. increasing inequalities in all-cause and cardiovascular mortality among US adults aged 25-64years by area socioeconomic status, 1969-1998. 42. Miller JE. The effects of race/eth nicity and income on early childhood asthma prevalence and health care us e. Am J Public Health 2000;90:428-30. 43. Lin S, Fitzgerald E, Hwang SA, Munsie JP, Stark A. Asthma Hospitalization Rates and Socioeconomic St atus in New York State (1987-1993). J Asthma 1999;36(3):239-51. 44. Persky VW, Slezak J, Contreras A. Relationships of race and socioeconomic status with prevalence, severity, and symptoms of asthma in Chicago school children. Ann Alle rgy Asthma Immunol 1998;81(3):266-71. 45. Kitch BT, Chew G, Burge HA et al. Socioeconomic predictors of high allergen levels in homes in the Greate r Boston area. Environ Health Perspect 2000;108:301-7; 46. Lannero E, Kull I, Wickman M, Persha gen G, et al. Environmental risk factors for allergy and socioeconomic stat us in a birth cohort (BAMSE). Pediatr Allergy Immunol 2002;13:182-187. 47. Maziak W, von Mutius E, Keil U et al. Predictors of Health Care Utilization of Children with Asthma in the Community. Pediatr Allergy Immunol 2004;15:177-71.

PAGE 215

20048. Brown CM, Anderson HA, Etzel RA. Asthma: the states’ challenge. Public Health Reports 1997;112:198-205. 49. Lewis SA, Weiss ST, Platts-Mills TAE. Association of Specific Allergen Sensitization with Socioeconom ic Factors and Allergic Di sease in a Population of Boston Women. J Allergy C lin Immunol 2001;107:615-22. 50. Sarinho E, Schor D, Veloso MA, Rizzo JA. There Are More Asthmatics in Homes with High Cockroach Infestati on. Braz J Med Biol Res 2004;37(4):503-10. 51. Custovic A, Woodcock A. On Allergens and Asthma (Again): Does Exposure to Allergens in Ho mes Exacerbate Asthma? (Ed itorial) Clin Exp Allergy 2001;31:670-3. 52. Crain EF, Walter M, O’Connor GT et al. Home and Allergic Characteristics of Children with asthma in Seven US Urban Communities and Design of an Environmental Interventi on: The Inner-City As thma Study. Environ Health Perspect 2002;110:939-45. 53. Leaderer BP, Belanger K, Triche E et al. Dust mite, cockroach, cat, and dog allergen concentrations in homes of as thmatic children in the Northeastern US: impact of socioeconomic factors and popul ation density. Environ Health Perspect 2002;110:419-425. 54. Togias A, Horowitz E, Joyner D et al. Evaluating the Factors that relate to asthma severity in adolescents. Int Arch Allergy Immunol 1997;113(1-3):87-95. 55. US Public Health Service. The International Classification of Diseases, 9th Revision, Clinical Modification (ICD-9-CM), 4th Ed., DHHS Publication No.911260 (PHS); Department of Health a nd Human Resources: Washington, DC, 1991.

PAGE 216

20156. US Census Bureau. US Census 2000, www.census.gov (10/15/04). 57. Melnick AL. Introduction to Geographi c Information Systems in Public Health. Aspen, 2002, 300 pp. 58. Florida Department of Environmenta l Protection. Air Monitoring Report, Tallahassee, FL, 1999. 59. Jarman B. Identification of underprivile ged areas. British Medical Journal, 286, 1705-09, 1983 60. SAS Institute, Inc. SAS/STAT software, version 8.1. Cary, NC, 2001. 61. Epi-Info version 3.2, Center for Dise ase Control and Prevention (CDC), Atlanta, GA, 2001. 62. GIS version 3.2, Environmental Sciences Research Institute (ESRI), Redland, CA, 1999. 63. Hennekens CH, Buring JE. Epidemiology in Medicine. Little, Brown and Co., Boston, MA. 1987, 383pp. 64. Szklo M, Nieto FJ. Epidemiology: beyond the basics. Aspen Publishers, Inc., Gaithersburg, MD. 2000, 495pp. 65. Friis RH, Sellers TA. Epidemiology for public health practice. Aspen Publishers, Inc., Gaithersburg, MD. 1996, 430pp. 66. Neter J, Kutner MH, Nachtsheim CJ, Wasserman W. Applied Linear Regression Models. Irwin, 1996, 3rd ed., 720 pp.

PAGE 217

20267. Fanta CH. Fatal asthma and the envi ronment. Immunol Allergy Clin N Am 2002;22:911-924. 68. Crain EF. Walter M, ‘Connor GT et al Home and allergic characteristics of children with asthma in seven US urban communities and design of an environmental intervention: the inner-city asthma study. Environ Health Perspect 2002;110;:939-45. 69. Ericson et al. Influence of Socioeconomics on the HQL. J Asthma 2002;39(2), 107-117. 70. Lara M, Nicholas W, Morton S et al. Improving childhood asthma outcome in the US: a blueprint for po licy action. Rand Health. MR-1330-RWJ, 114 pp, www.rand.org/publicationsMR/MR1330

PAGE 218

203 Appendices

PAGE 219

204 Appendix A. IRB Exemption Certificate (Instituti onal Review Board, USF Division of Research Compliance, Research Protocol No.102536, May 28, 2004)

PAGE 220

205 Appendix B. Direct Adjustment Calculations Table B-1. Total number of asthma hospital admissions within different age strata by zip code area of residence in Hillsborough County, FL, in 1999 Zip Code 0-4 5-14 1524 2534 3544 4554 5564 6574 7584 85 0-14 15 T otal 33510 2 11 2 4 2 1 5 0 1 1 13 16 29 33511 10 10 6 5 7 2 3 3 1 0 20 27 4 7 33527 5 2 0 1 3 2 0 0 3 0 7 9 1 6 33534 1 3 2 0 2 4 2 0 0 0 4 10 14 33547 1 1 0 0 4 1 0 0 0 0 2 5 7 33549 10 6 1 2 1 3 5 3 3 0 16 18 34 33556 3 6 1 0 1 1 2 1 1 0 9 7 1 6 33565 4 1 3 6 1 0 2 1 1 1 5 15 20 33566 11 6 8 6 0 3 5 3 2 1 17 28 45 33567 7 2 3 2 3 2 2 3 3 0 9 18 2 7 33569 13 8 5 1 6 7 5 6 0 1 21 31 52 33570 5 1 2 3 4 1 2 0 0 1 6 13 19 33572 1 0 0 0 0 0 1 0 0 0 1 1 2 33573 0 0 0 0 0 1 0 1 4 0 0 6 6 33584 4 1 3 3 6 2 4 3 2 0 5 23 28 33592 7 1 0 1 4 6 2 3 0 0 8 16 24 33594 12 12 4 4 7 2 7 1 1 1 24 27 51 33598 9 0 2 0 0 3 1 1 0 0 9 7 1 6 33602 23 2 2 3 4 2 2 7 5 1 25 26 51 33603 10 19 9 3 6 4 7 3 3 1 29 36 65 33604 30 22 3 11 10 15 4 6 3 0 52 52 104 33605 27 23 3 3 8 4 5 4 2 0 50 29 79 33606 5 5 0 0 5 3 2 1 2 0 10 13 23 33607 23 15 1 4 12 6 4 4 5 2 38 38 7 6 33609 9 2 0 3 4 2 1 3 3 0 11 16 2 7 33610 33 32 6 4 14 16 5 5 3 0 65 53 118 33611 12 11 13 2 2 6 3 5 3 0 23 34 5 7 33612 21 20 4 9 14 16 9 6 3 2 41 63 104 33613 9 12 9 10 2 6 2 1 3 0 21 33 54 33614 25 21 7 12 9 13 10 6 4 1 46 62 108 33615 22 13 4 11 10 6 9 4 2 0 35 46 81 33616 7 6 2 2 0 2 6 1 0 0 13 13 2 6 33617 16 16 5 1 8 3 3 3 0 0 32 23 55 33618 2 4 0 0 2 1 3 0 3 0 6 9 15 33619 21 9 5 4 2 6 4 1 3 0 30 25 55 33621 0 1 0 0 0 0 1 0 0 0 1 1 2 33624 8 8 1 6 3 5 4 2 3 1 16 25 41 33625 4 3 1 1 1 1 1 1 1 0 7 7 14 33626 3 3 0 1 4 0 2 1 0 0 6 8 14 33629 8 3 1 1 0 2 2 0 1 0 11 7 18 33634 8 7 3 0 2 3 1 1 0 0 15 10 25 33635 7 3 0 1 1 3 2 2 0 0 10 9 19 33637 5 1 2 5 6 1 0 2 0 0 6 16 22 33647 4 2 0 0 0 0 0 0 0 0 6 0 6 Total 447 334 123 135 180 167 140 98 74 14 781 931 1712

PAGE 221

206 Appendix B. Direct Adjustment Calc ulations (continued) Table B-2 Total population distribution within sepa rate age strata by zip code area of residence in Hillsborough County, FL, in 1999 Zip Code 0 4 5 14 15 24 25 34 35 44 45 54 55-64 65-74 75-84 85 Total 335101529 3465 2820 3228 3949 3258 1975 1250 659 241 22374 335113153 6786 6083 7407 7936 6176 3546 2215 1173 452 44927 33527898 1907 1762 1619 1768 1412 1069 558 344 94 11431 33534712 1263 1109 1145 1149 914 583 377 210 34 7496 33547619 1476 1044 1194 1456 1262 799 412 214 51 8527 335492957 6493 5041 6437 8581 7255 3955 2406 1265 282 44672 33556905 2198 1339 1534 2927 2587 1333 745 354 73 13995 335651121 2429 2009 1926 2644 2060 1757 1723 933 212 16814 335661729 3492 3214 2968 3044 2643 1793 1373 941 355 21552 335672093 4331 3627 3708 4032 3199 2240 1592 843 255 25920 335692665 5636 4301 5250 6415 4808 3049 2152 1158 255 35689 33570887 1686 1569 1496 1553 1427 1403 1585 1031 220 12857 33572309 780 656 630 1180 1381 1135 799 490 101 7461 3357318 38 41 81 124 427 2038 5427 6147 1980 16321 335841394 3338 2529 2802 3668 2991 1891 1111 611 155 20490 33592702 1536 1218 1320 1616 1349 1003 678 416 132 9970 335943210 8255 5377 5460 9344 7752 4020 2554 1438 311 47721 33598874 1627 1467 1210 1056 775 504 310 157 39 8019 33602623 1244 1146 1377 1458 1177 743 627 399 161 8955 336031507 3218 2769 3300 3388 2533 1495 1245 1078 414 20947 336042722 5952 4734 5650 6073 4822 2809 2069 1527 427 36785 336051249 2945 2345 2043 2440 2099 1492 1334 825 309 17081 33606643 1400 2473 3098 2586 2066 1075 826 564 229 14960 336071499 3202 2817 3178 3087 2540 2127 2226 1584 541 22801 33609836 1832 1592 2501 2850 2384 1456 1279 1099 351 16180 336102477 5521 4461 3933 4752 4116 3030 2350 1330 427 32397 336111627 3213 3006 5189 5467 4081 2668 2391 1669 526 29837 336123417 6347 6625 6792 6466 5148 2918 2413 2074 761 42961 336131931 3128 6549 5254 3908 3195 1980 1349 1288 842 29424 336142946 5435 6588 8351 6986 5140 3642 2672 1586 457 43803 336152560 5489 5347 7157 7060 5441 3871 2586 1408 430 41349 33616814 1768 1718 2172 2101 1498 947 640 297 59 12014 336172934 5989 7752 7036 6386 5264 3163 2120 1313 324 42281 336181206 2754 2564 2755 3517 3329 2020 1281 738 194 20358 336191907 4771 4435 4135 4647 3684 2369 1588 733 190 28459 33621356 614 629 586 418 67 13 4 1 1 2689 336242814 6539 5537 7038 8206 7285 3885 2140 1270 351 45065 336251553 3450 2350 3204 4215 3060 1474 822 489 164 20781 336261217 1532 788 2347 2444 1435 745 378 186 44 11116 336291494 2538 1752 3510 4286 3517 1900 1701 1610 550 22858 336341203 2827 2589 3204 3251 2566 1869 1148 491 107 19255 33635933 1745 1173 2190 2428 1651 1051 783 401 84 12439 33637946 1790 2073 2476 2111 1428 833 516 274 87 12534 336472145 4184 3523 4676 5399 3399 1665 822 408 69 26290 Total 69334 146163132541152567168372 13460185333 64577 43026 13341 1009855

PAGE 222

207 Appendix B. Direct Adjustment Calculations (cont.) Table B-3. Total number of asthma hospital admi ssions and standard weight (Standard 1,000,000 population) for separate age category (s trata) group used to calculate ageadjusted rates by direct adjustment technique 0 4 5 14 15 24 25 34 35 44 45 54 Total asthma hospitalizations 447 334 123 135 180 167 Total study population 2145 4184 3523 4676 5399 3399 US standard population 19175798 41077577 39183891 39891724 45148527 37677952 Standard 1,000,000 population 69135 145565 138646 135573 162613 134834 Table B-3. Total number of asthma hospital admissi ons and standard weight for separate age category group used to calcu late age-adjusted rates by di rect adjustment technique (cont.) 55 to 64 65 to 74 75 to 84 85 Total Total asthma hospitalizations 140 98 74 14 1712 Total study population 1665 822 408 69 26290 US standard population 24274684 18390986 12361180 4239587 281421906 Standard 1,000,000 population 87247 66037 44842 15508 1000000 Table B-4. Total number of asthma hospital admissi ons within different strata by race (White, Black, Hispanic and Other) and by gende r (Male and Female) by zip code area of residence in Hillsborough County, FL, in 1999 Zip Code White Black Hispanic Other Male Female Total Total Population 33510 21 2 4 2 11 18 29 22374 33511 36 6 1 4 19 28 47 44927 33527 7 0 9 0 10 6 16 11431 33534 13 1 0 0 5 9 14 7496 33547 6 0 0 1 2 5 7 8527 33549 27 2 3 2 15 19 34 44672 33556 13 3 0 0 10 6 16 13995 33565 19 0 1 0 11 9 20 16814 33566 24 13 8 0 21 24 45 21552 33567 20 4 3 0 10 17 27 25920 33569 42 7 2 1 23 29 52 35689 33570 10 0 9 0 9 10 19 12857 33572 1 1 0 0 0 2 2 7461 33573 6 0 0 0 2 4 6 16321 33584 23 2 2 1 8 20 28 20490 33592 16 6 1 1 6 18 24 9970 33594 43 3 3 2 26 25 51 47721 33598 5 2 8 1 9 7 16 8019 33602 7 27 16 1 27 24 51 8955 33603 20 17 27 1 25 40 65 20947 33604 29 45 21 9 36 68 104 36785

PAGE 223

208 Appendix B. Direct Adjustment Calculations (cont.) Table B-4. Total number of asthma hospital admissi ons within different strata by race (White, Black, Hispanic and Other) and by ge nder (Male and Female) by zip code area of residence in Hillsborough County, FL, in 1999 (cont.) Zip Code White Black Hispanic Other Male Female Total Total Population 33605 9 50 17 3 28 51 79 17081 33606 9 9 4 1 10 13 23 14960 33607 9 34 32 1 35 41 76 22801 33609 15 3 9 0 15 12 27 16180 33610 31 73 8 6 56 62 118 32397 33611 39 6 7 5 27 30 57 29837 33612 43 38 18 5 37 67 104 42961 33613 28 8 14 4 18 36 54 29424 33614 41 5 58 4 41 67 108 43803 33615 42 17 21 1 34 47 81 41349 33616 10 6 8 2 13 13 26 12014 33617 18 26 8 3 18 37 55 42281 33618 9 1 5 0 6 9 15 20358 33619 31 20 3 1 25 30 55 28459 33621 1 0 0 1 1 1 2 2689 33624 28 1 7 5 20 21 41 45065 33625 8 1 4 1 8 6 14 20781 33626 10 0 3 1 5 9 14 11116 33629 14 0 2 2 13 5 18 22858 33634 10 3 11 1 15 10 25 19255 33635 16 1 0 2 6 13 19 12439 33637 13 4 5 0 11 11 22 12534 33647 6 0 0 0 2 4 6 26290 TOTAL 828 447 362 75 729 983 1712 1009855 Table B-5. Total population distribution within di fferent strata by race (White, Black, Hispanic and Other) and by gender (Male and Fe male) by zip code area of residence in Hillsborough County, FL, in 1999 Zip Code White Black Hispanic Other Male Female Total 33510 18526 2013 452 1383 10811 11563 22374 33511 32761 4263 5752 2151 21817 23110 44927 33527 7692 98 3462 179 5939 5492 11431 33534 5692 104 1509 191 3877 3619 7496 33547 7713 110 556 148 4252 4275 8527 33549 36671 1825 4502 1674 22160 22512 44672 33556 12098 431 982 484 7070 6925 13995 33565 14879 259 1434 242 8329 8485 16814 33566 13305 4437 3532 278 10312 11240 21552 33567 17806 1782 5743 589 13027 12893 25920 33569 27963 2773 3819 1134 17827 17862 35689 33570 8419 129 4119 190 6516 6341 12857 33572 6634 63 566 198 3718 3743 7461 33573 15986 22 195 118 6960 9361 16321 33584 16650 1554 1784 502 10197 10293 20490 33592 7931 1163 627 249 5012 4958 9970 33594 38691 2853 4543 1634 23438 24283 47721 33598 2651 346 4936 86 4254 3765 8019 33602 3106 4145 1578 126 4454 4501 8955 33603 8736 5883 5973 355 10143 10804 20947 33604 18815 9121 7714 1135 17955 18830 36785

PAGE 224

209 Appendix B. Direct Adjustment Calculations (cont.) Table B-5. Total population distribution within di fferent strata by race (White, Black, Hispanic and Other) and by gender (Male and Fe male) by zip code area of residence in Hillsborough County, FL, in 1999 (cont.) Zip Code White Black Hispanic Other Male Female Total 33605 1905 10549 4523 104 8418 8663 17081 33606 11618 1706 1163 473 7483 7477 14960 33607 4224 8774 9433 370 10802 11999 22801 33609 10956 1138 3465 621 7927 8253 16180 33610 10519 18362 2943 573 15217 17180 32397 33611 23184 1846 3133 1674 14546 15291 29837 33612 21211 12446 7669 1635 20733 22228 42961 33613 17108 5617 5033 1666 14542 14882 29424 33614 17379 3909 20676 1839 21709 22094 43803 33615 24247 3449 11840 1813 20218 21131 41349 33616 7136 2137 1608 1133 6042 5972 12014 33617 22716 11576 5887 2102 20242 22039 42281 33618 15394 916 3174 874 9834 10524 20358 33619 11371 11416 5152 520 14814 13645 28459 33621 1519 659 323 188 1428 1261 2689 33624 31712 2894 8008 2451 21501 23564 45065 33625 14064 1518 4182 1017 10185 10596 20781 33626 8650 607 1322 537 5453 5663 11116 33629 20091 269 1931 567 10922 11936 22858 33634 9638 1554 7209 854 9297 9958 19255 33635 8968 726 2017 728 6116 6323 12439 33637 8191 2122 1594 627 6046 6488 12534 33647 19977 1541 2432 2340 12995 13295 26290 Total 644503 149105 178495 37752 494538 515317 1009855 Table B-6. Total number of asthma hospital admi ssions and standard weight (Standard 1,000,000 population) within separate race and ge nder category (stratum) group used to calculate raceand gender-adjusted ra tes by direct adjustment technique White Black Hispanic Other Male Female Total Asthma hospitalizations 828 447 362 75 7299831712 Total study population 644503 149105 178495 37752 4945385153171009855 US population 194552774 36419434 35305818 15143880 138053563143368343281421906 Standard 1,000,000 population 691321 129412 125455 53812 4905575094431000000

PAGE 225

210 Appendix C. Percentage distribution of socioeconom ic status characteristics by area of residence. Table C-1. Distribution of socioeco nomic status indicators by zip code areas of residence, Hillsborough County, FL. Zip Code Poverty Education Unemployment Uskilled Divorced Professional Occupation Family Income 33510 4 3.9 2.4 66.8 11.2 34.3 8.7 33511 4 2.8 2.8 60.6 10.7 38.8 6.8 33527 12.7 17.1 5.8 87.3 10.5 19.5 15.5 33534 16.5 12.1 5 90.5 14.7 16.8 19.7 33547 7 5.4 2.6 78.4 9.2 29.6 11.5 33549 3.9 2.2 2.9 55.9 11 42 9 33556 1.9 2 0.8 53.4 6.8 42.7 5.2 33565 6.6 9.1 2.2 82.6 8.4 25.6 12.6 33566 12.9 13.5 3.4 84.6 13.1 21 19.5 33567 9.6 13.6 3.1 79.5 9.9 22.2 12.3 33569 6.6 5 2.1 71.5 11.7 31.8 10.9 33570 9.5 14.9 3 88.2 10.6 18 20.9 33572 2.4 2.4 2.9 71.2 11.7 38.6 10.2 33573 2.2 3 0.6 66.5 5.6 35 12.8 33584 5.4 5.3 3.3 78.2 12.6 25.2 11.7 33592 9.6 6.8 3.3 84.9 15.3 21.5 10.6 33594 3 2.9 2 57.6 7.3 42 6.8 33598 23.1 31.3 5.2 90.4 8.6 14.5 17.3 33602 27.9 8.8 8 66.2 18.1 40.3 33.9 33603 17.7 10.5 6 78.2 14.1 23.9 23.8 33604 19.6 8.4 4 79.5 16.4 24.4 25.3 33605 28.3 16.9 6.1 89 14.5 15.5 39.7 33606 3.8 2.1 11.1 35.2 13.1 55 13 33607 18.9 17.5 4.1 81.7 13.3 28.8 18.7 33609 6.1 5.5 2.5 54.5 14.1 43.4 15 33610 18.8 11 5.1 86.3 15.1 16.9 26.7 33611 5.9 2.9 2.3 63.1 16.2 34.4 15.9 33612 17.3 7.1 5.1 77.2 13.5 24.4 26.4 33613 14.5 6.6 5.1 65 14.2 31.6 26.1 33614 13.4 8.4 3.3 75.8 14.5 25.3 17.4 33615 13 5.8 3.1 67.8 13.9 28.7 12 33616 13.6 4.4 5 77.1 17.5 24.5 13.8 33617 10.2 4.9 3.7 61.3 12.6 34.9 17.7 33618 4.4 2 2.2 49.8 11.3 45.2 9.3 33619 15.9 11.1 3.8 88.3 14.9 15.9 20.5 33621 4.1 0.8 2.8 70.2 3 27.9 3.1 33624 3.4 2.7 2.1 53 12.8 42.8 6.2 33625 4.8 3.7 3.9 60.8 11.4 38 6.2 33626 1.7 0.6 1.5 39.7 6.6 55.1 3.4 33629 2.3 1.1 2 37.8 11.3 57.7 8.9 33634 6.5 6.1 2.8 72.3 13 29.9 9.9 33635 6.3 2.7 2.1 59.3 12.2 41.9 11.4

PAGE 226

211 Appendix C. Percentage distribution of socioec onomic status characteristics by area of residence. (cont.) Table C-1. Distribution of socioeco nomic status indicators by zip code areas of residence, Hillsborough County, FL (cont.) Zip Code Poverty Education Unemployment Uskilled Divorced Professional Occupation Family Income 33637 10.8 2.8 1.7 63.2 15.1 36.7 15 33647 3.9 0.7 2.1 33.1 8.1 55.8 8.1 Table C-2. Distribution of socioeco nomic status indicators by zip code areas of residence, Hillsborough County, FL. Zip Code House Age 40 Moved last year No vehicle Heating with gas Heating with fuel Heating with wood Lacking facilities 33510 5.6 22.5 4.2 6.3 1.1 0 0.1 33511 2.5 31 3 7.6 0.7 0.1 0.8 33527 17.9 13.5 5.6 11.5 1 0 1.8 33534 14.4 29.7 10.4 12.5 0.7 0.5 0.4 33547 8.3 21.9 2.7 8.9 1.1 0.9 0.5 33549 4.6 25 3.6 5.5 0.2 0.1 0.4 33556 6.1 19.1 0.8 4.8 0.2 0.2 0.5 33565 11.2 15.3 4.7 8.6 0.6 0 1.5 33566 29.8 23.4 8.9 8.2 0.5 0.1 0.7 33567 15 23.6 3.7 8.3 0.7 0 0.7 33569 4.7 29.9 3 8 0.2 0.3 0.6 33570 10.9 20.5 4.1 10.9 0.3 0.4 1 33572 3.6 13.1 1.9 3.5 0.3 0 0.9 33573 1.7 18.1 9.2 1.2 0.1 0 0.8 33584 9.6 16.8 5.2 3.3 0.6 0.7 0.7 33592 6.5 22.6 4.8 6.5 0.7 0.3 0.5 33594 2.8 19.4 2.6 3.4 0.3 0.2 0.5 33598 4.7 17.2 7.3 18.9 0 0 3.8 33602 42.2 27.7 26 20.5 1.1 0 2 33603 63.5 23.7 12.2 11.2 1.8 0.2 1 33604 42.1 24.8 14.8 6.1 0.9 0.6 1.2 33605 60.4 21.1 27.8 19.3 2.8 0 4.7 33606 57 33.4 9.6 9.4 0.4 0.2 1.2 33607 40.4 23.8 21.6 9.8 1.7 0 1.1 33609 43.4 21.3 9.9 5.4 1.3 0 0.7 33610 31.9 21.2 14.5 6 2.1 0.2 1.8 33611 34.3 23.8 9 4.2 0.4 0.1 0.5 33612 22.8 30.8 16.6 6.3 0.8 0.1 1.1 33613 6 41.5 15.3 4 0 0.2 2.6 33614 15.4 35.4 8.3 5.3 0.3 0 1 33615 3.9 29.4 8.2 1.9 1.1 0 1 33616 25.9 30.9 7.8 4.3 0.1 0.2 0.8 33617 12.4 33.8 9.2 4 0.8 0.1 0.4 33618 5 23 5 1.2 0.8 0 0 33619 23.5 20 10 3.5 1.5 0.4 1

PAGE 227

212 Appendix C. Percentage distribution of soci oeconomic status characteristics by area of residence. (cont.) Table C-2. Distribution of socioeconomic stat us indicators by zip code areas of residence, Hillsborough County, FL. (cont.) Zip Code House Age 40 Moved last year No vehicle Heating with gas Heating with fuel Heating with wood Lacking facilities 33621 67.2 51.8 5.3 51.6 0 0 0 33624 1.2 24.4 2.6 1.6 0.1 0 0.9 33625 2.3 20 1.9 1.1 0.4 0 0.6 33626 0.2 35.5 0.5 42.7 0 0 0.4 33629 55.3 18 4.7 6.2 1.5 0 0.5 33634 5.6 25.8 5.1 0.9 0.5 0 0.8 33635 1.3 22.4 4.4 9.1 0.2 0.4 0 3.4 39.5 4.9 2 0.1 0 0.3 33637 33647 0.7 42.1 2.7 14.9 0 0 0.2 Table C-3. Distribution of socioeco nomic status indicators by zip code areas of residence, Hillsborough County, FL. Zip Code Overcrowded conditions Children under 5 yr Elderly 65 yr. Black Single w children Ethnic minority 33510 3.2 6.8 9.6 9.6 7.6 14.8 33511 3.8 7 8.5 10.3 7.5 15.4 33527 12.7 7.9 8.7 1.2 5.7 11 33534 9.1 9.5 8.3 1.6 10 9.3 33547 2.6 7.3 7.9 1.5 5 4.3 33549 2.5 6.6 8.8 4.5 5.2 8.9 33556 1.1 6.5 8.4 3.4 3 5.9 33565 4.3 6.7 17.1 1.7 4.8 5.1 33566 8.8 8 12.4 21 9.6 26.5 33567 9.1 8.1 10.4 7.2 7.1 16.1 33569 3.7 7.5 10 8.3 7.4 11.1 33570 10.1 6.9 22.1 1.1 4.9 9.6 33572 2.1 4.1 18.6 1.4 3.5 3.8 33573 0.2 0.1 83 0.2 0 0.8 33584 3 6.8 9.2 7.6 8 11.3 33592 6.2 7 12.3 12.3 8.1 14.6 33594 2.6 6.7 9 6.4 4.6 9.7 33598 28.1 10.9 6.3 4.4 8 32.7 33602 7.4 7 13.3 47.5 12.4 45.8 33603 9.4 7.2 13.1 29.4 11 32.3 33604 10.1 7.4 10.9 26.2 12.6 29.1 33605 15.9 7.3 14.4 63.2 14.6 69.2 33606 1.4 4.3 10.8 11.8 4.4 13.4 33607 12.9 6.6 19.1 39.8 11.4 44.6 33609 3.7 5.2 16.9 7.9 5.3 12.4 33610 8.7 7.6 12.7 57.7 15.2 58.6 33611 2.1 5.5 15.4 7.1 5.8 11.2 33612 10.4 8 12.2 30.6 13.5 35

PAGE 228

213 Appendix C. Percentage distribution of socioeconomic status characteristics by area of residence. (cont.) Table C-3. Distribution of socioeconomic status indicators by zip code areas of residence, Hillsborough County, FL. (cont.) Zip Code Overcrowded conditions Children under 5 yr Elderly 65 yr. Black Single w children Ethnic minority 33613 8.5 6.6 11.8 20.2 9 28.3 33614 11.9 6.7 10.8 9.9 8.7 22.5 33615 5.8 6.2 10.7 9.1 7 18.4 33616 7 6.8 8.3 19.4 10.4 26.6 33617 8 6.9 8.9 28.7 10.8 33.9 33618 4.2 5.9 10.9 5 6.1 11.2 33619 9.7 6.7 8.8 41.3 13.2 42.5 33621 12.8 13.2 0.2 26.9 10 35.4 33624 3.5 6.2 8.3 7.2 7.1 13.8 33625 4.9 7.5 7.1 8.1 7.1 15.1 33626 1 10.9 5.5 5.5 3.3 11.9 33629 0.9 6.5 16.9 1.4 4.1 4 33634 7.4 6.2 9.1 9.1 8.1 19.1 33635 3 7.5 10.2 6.4 5.9 12.1 33637 5.6 7.5 7 17.7 9.5 22.3 33647 2.5 8.2 4.9 6.5 4.8 15.6

PAGE 229

214 Appendix D Results of Stratified Frequency Table Analyses Poverty Environmental exposure to particulate maters (PM 10 ) Table D-1. Association between environmental e xposure to particulate matters with a diameter of 10 microns or less (PM10) and children asthma hospital admissions stratified by poverty status (SES – socioeconomic status ; RRrate ratio; 95% CI – 95% Confidence Intervals) Poverty status category (strata) Environmental exposure to PM10, g/m3 Children hospitalizations for asthma Total population Number of zip code areas of residence Rate Ratio (95% CI) Low SES by poverty (15-28.3 % persons living below poverty) 22 and < 25 25 and 28 177 261 74,444 87,976 2 8 RR=1.25 (1.03-1.52) High SES by poverty (1.7 – 15 % persons living below poverty) 22 and < 25 25 and 28 13 330 4,476 48,601 19 15 RR=2.34 (1.34-4.07) Total 22 and < 25 25 and 28 190 591 78,920 136,577 21 23 RR=1.8 (1.53-2.12) Table D-2 Association between environmental e xposure to particulate matters with a diameter of 10 microns or less (PM10) and adult hospital admissi ons stratified by poverty status (SES – socioeconomic status; RRrate ratio; 95% CI – 95% Confidence Intervals) Poverty status category (strata) Environmental exposure to PM10, g/m3 Adults hospital admissions for asthma Total population Number of zip code areas of residence Rate Ratio (95% CI) Low SES by poverty (15-28.3 % persons living below poverty) 22 and < 25 25 and 28 255 337 297,065 324,469 2 8 RR=1.21 (1.00-1.42) High SES by poverty (1.7 – 15 % persons living below poverty) 22 and <25 25 and 28 17 322 11,039 161,785 19 15 RR=1.29 (0.79-2.1) Total 22 and < 25 25 and 28 272 659 308,104 486,254 21 23 RR=1.54 (1.33-1.77)

PAGE 230

215 Appendix D Results of Stratified Frequency Table Analyses (cont.) Environmental exposure to sulfur dioxide (SO2) Table D-3 Association between environmenta l exposure to sulfur dioxide (SO2) and children asthma hospital admi ssions stratified by poverty st atus (SES – socioeconomic status; RRrate ratio; 95% CI – 95% Confidence Intervals) Poverty status category (strata) Environmental Exposure to SO2, ppm Children hospital admissions for asthma Total population Number of zip code areas of residence Rate Ratio (95% CI) Low SES by poverty (15-28.3 % persons living below poverty) 3.5 and < 5.2 5.2 and 7.0 257 181 116,221 46,199 2 8 RR=1.8 (1.47-2.14) High SES by poverty (1.7 – 15 % persons living below poverty) 3.5 and < 5.2 5.2 and 7.0 13 330 4,476 48,601 9 25 RR=2.3 (1.34-4.07) Total 3.5 and < 5.2 5.2 and 7.0 270 511 120,427 94,289 11 33 RR=2.4 (2.08-2.79) Table D-4. Association between environmenta l exposure to sulfur dioxide (SO2) and adult asthma hospital admissions stratified by poverty status (SES – socioeconomic status; RRrate ratio; 95% CI – 95% Confidence Intervals) Poverty status category (strata) Environmental Exposure to SO2, ppm Adults hospital admissions for asthma Total population Number of zip code areas of residence Rate Ratio (95% CI) Low SES by poverty (15-28.3 % persons living below poverty) 3.5 and < 5.2 5.2 and 7.0 381 211 428,948 192,586 2 8 RR=1.23 (1.00-1.46) High SES by poverty (1.7 – 15 % persons living below poverty) 3.5 and < 5.2 5.2 and 7.0 17 322 11,039 161,785 9 25 RR=1.29 (0.79-2.1) Total 3.5 and < 5.2 5.2 and 7.0 398 533 439,589 353,838 11 33 RR=1.66 (1.46-1.89)

PAGE 231

216 Appendix D Results of Stratified Frequency Table Analyses (cont.) Family Income Environmental exposure to particulate maters (PM 10 ) Table D-5. Association between environmental e xposure to particulate matters with a diameter of 10 microns or less (PM10) and children asthma hospital admissions stratified by family income (SES – socioeconomic status; RRrate ra tio; 95% CI – 95% Confidence Intervals) SES categories by family income Environmental exposure to PM10, g/m3 Children hospital admissions for asthma Total population Number of zip code areas of residence Rate Ratio (95% CI) Low SES by annual family income of $15,000 and less (21.4 – 39.7 %) 22 and < 25 25 and 28 169 329 73,861 99,355 1 6 RR=1.5 (1.2-1.75) High SES by annual family income of $15,000 and less (3.1 21.4 %) 22 and < 25 25 and 28 21 262 5059 37,222 20 17 RR=1.7 (1.09-2.64) Total 22 and < 25 25 and 28 190 591 78,920 136,577 21 23 RR=1.8 (1.53-2.12) Table D-6. Association between environmental expos ure to particulate matters with a diameter of 10 microns or less (PM10) and adult asthma hospita l admissions stratified by family income (SES – socioeconomic status ; RRrate ratio; 95% CI – 95% Confidence Intervals) SES categories by family income Environmental exposure to PM10, g/m3 Adults hospital admissions for asthma Total population Number of zip code areas of residence Rate Ratio (95% CI) Low SES by annual family income of $15,000 and less (21.4 – 39.7 %) 22 and < 25 25 and 28 239 400 283,739 364,350 1 6 RR=1.3 (1.11-1.53) High SES by annual family income of $15,000 and less (3.1 21.4 %) 22 and < 25 25 and 28 33 259 24,365 121,904 20 17 RR=1.57 (1.09-2.25) Total 22 and < 25 25 and 28 272 659 308,104 486,254 21 23 RR=1.54 (1.33-1.77)

PAGE 232

217 Appendix D Results of Stratified Frequency Table Analyses (cont.) Environmental exposure to sulfur dioxide (SO 2 ) Table D-7 Association between environmenta l exposure to sulfur dioxide (SO2) and children asthma hospital admissi ons stratified by family in come (SES – socioeconomic status; RRrate ratio; 95% CI – 95% Confidence Intervals) SES categories by family income Environmental Exposure to SO2, ppm Children hospital admissions for asthma Total population Number of zip code areas of residence Rate Ratio (95% CI) Low SES by annual family income of $15,000 and less (21.4 – 39.7 %) 3.5 and < 5.2 5.2 and 7.0 249 249 115638 57578 1 6 RR=2.0 (1.69-2.39) High SES by annual family income of $15,000 and less (3.1 21.4 %) 3.5 and < 5.2 5.2 and 7.0 21 262 5059 37,222 25 12 RR=1.7 (1.09-2.64) Total 3.5 and < 5.2 5.2 and 7.0 270 511 120,427 94,289 26 18 RR=2.4 (2.08-2.79) Table D-8. Association between environmenta l exposure to sulfur dioxide (SO2) and adult asthma hospital admissions stratified by family income (SES – socioeconomic status; RRrate ratio; 95% CI – 95% Confidence Intervals) SES categories by family income Environmental Exposure to SO2, ppm Adults hospital admissions for asthma Total population Number of zip code areas of residence Rate Ratio (95% CI) Low SES by annual family income of $15,000 and less (21.4 – 39.7 %) 3.5 and < 5.2 5.2 and 7.0 365 274 415622 232467 1 6 RR=1.3 (1.15-1.57) High SES by annual family income of $15,000 and less (3.1 21.4 %) 3.5 and < 5.2 5.2 and 7.0 33 259 24,365 121,904 25 12 RR=1.5 (1.1-2.2) Total 3.5 and < 5.2 5.2 and 7.0 398 533 439,589 353,838 26 18 RR=1.7 (1.46-1.89)

PAGE 233

218 Appendix D Results of Stratified Frequency Table Analyses (cont.) White-collar occupation Environmental exposure to particulate maters (PM 10 ) Table D-9. Association between environmental expos ure to particulate matters with a diameter of 10 microns or less (PM10) and children asthma hospital admissions stratified by the percentage of white collar employees (SES – socioeconomic status; RRrate ratio; 95% CI – 95% Confidence Intervals) SES categories by percentage of white-collar occupation employees Environmenta l exposure to PM10, g/m3 Children hospital admissions for asthma Total population Number of zip code areas of residence Rate Ratio (95% CI) Low SES by employment in white-collar occupations (14.5–36.1 %) 22 and < 25 25 and 28 108 489 37503 99530 11 17 RR=1.7 (1.39-2.1) High SES by employment in white-collar occupations (36.1-57.7 %) 22 and < 25 25 and 28 82 102 41417 37047 10 6 RR=1.4 (1.04-1.86) Total 22 and < 25 25 and 28 190 591 78,920 136,577 21 23 RR=1.8 (1.53-2.12) Table D-10. Association between environmental expos ure to particulate matters with a diameter of 10 microns or less (PM10) and children asthma hospital admissions stratified by the percentage of white collar employees (SES – socioeconomic status; RRrate ratio; 95% CI – 95% Confidence Intervals) SES categories by percentage of white-collar occupation employees Environmental exposure to PM10, g/m3 Adults hospital admissions for asthma Total population Number of zip code areas of residence Rate Ratio (95% CI) Low SES by employment in white-collar occupations (14.5–36.1%) 22 and < 25 25 and 28 181 534 150776 351734 11 17 RR=1.3 (1.07-1.5) High SES by employment in white-collar occupations (36.1-57.7 %) 22 and < 25 25 and 28 98 125 157328 134520 10 6 RR=1.6 (1.22-2.12) Total 22 and < 25 25 and 28 272 659 308,104 486,254 21 23 RR=1.5 (1.33-1.77)

PAGE 234

219 Appendix D Results of Stratified Frequency Table Analyses (cont.) Environmental exposure to sulfur dioxide (SO 2 ) Table D-11. Association between environmenta l exposure to sulfur dioxide (SO2) and children asthma hospital admi ssions stratified by the pe rcentage of white collar employees (SES – socioeconomic status; RRrate ratio; 95% CI – 95% Confidence Intervals) SES categories by percentage of white-collar occupation employees Environmental Exposure to SO2, ppm Children hospital admissions for asthma Total population Number of zip code areas of residence Rate Ratio (95% CI) Low SES by employment in white-collar occupations (14.5–36.1 %) 3.5 and <5.2 5.2 and 7.0 149 448 55579 81454 15 13 RR=2.1 (1.71-2.47) High SES by employment in white-collar occupations (36.1-57.7 %) 3.5 and <5.2 5.2 and 7.0 121 63 65118 13346 11 5 RR=2.5 (1.86-3.49) Total 3.5 and <5.2 5.2 and 7.0 270 511 120,427 94,289 26 18 RR=2.4 (2.08-2.79) Table D-12. Association between environmenta l exposure to sulfur dioxide (SO2) and adult asthma hospital admissions stratified by the percentage of white collar employees (SES – socioeconomic status; RRrate ra tio; 95% CI – 95% C onfidence Intervals) SES categories by percentage of white-collar occupation employees Environmental Exposure to SO2, ppm Adults hospital admissions for asthma Total population Number of zip code areas of residence Rate Ratio (95% CI) Low SES by employment in white-collar occupations (14.5–36.1 %) 3.5 and <5.2 5.2 and 7.0 260 455 210280 292230 15 13 RR=1.3 (1.08-1.47) High SES by employment in white-collar occupations (36.1-57.7 %) 3.5 and <5.2 5.2 and 7.0 138 78 229707 62141 11 5 RR=2.0 (1.57-2.76 Total 3.5 and <5.2 5.2 and 7.0 398 533 439,589 353,838 26 18 RR=1.7 (1.46-1.89)

PAGE 235

220 Appendix D Results of Stratified Frequency Table Analyses (cont.) Single parent families with children Environmental exposure to particulate maters (PM 10 ) Table D-13. Association between environmental expos ure to particulate matters with a diameter of 10 microns or less (PM10) and children asthma hospital admissions stratified by single parent with children status (SES – socioeconomic st atus; RRrate ratio; 95% CI – 95% Confidence Intervals) SES categories (strata) by the percentage single parent living with children Environmental exposure to PM10, g/m3 Children hospital admissions for asthma Total population Number of zip code areas of residence Rate Ratio (95% CI) Low SES by single parent living with children (7.615.2 %) 22 and < 25 25 and 28 139 122 64164 48390 4 17 RR=1.2 (0.91-1.48) High SES by single parent living with children (0-7.6 %) 22 and < 25 25 and 28 51 469 14756 88187 17 6 RR=1.5 (1.15-2.05) Total 22 and < 25 25 and 28 190 591 78,920 136,577 21 23 RR=1.8 (1.53-2.12) Table D-14. Association between environmental expos ure to particulate matters with a diameter of 10 microns or less (PM10) and adult asthma hospita l admissions stratified by single parent with children status (SES – soci oeconomic status; RRrate ratio; 95% CI – 95% Confidence Intervals) SES categories (strata) by the percentage single parent living with children Environmenta l exposure to PM10, g/m3 Adults hospital admissions for asthma Total population Number of zip code areas of residence Rate Ratio (95% CI) Low SES by single parent living with children (7.615.2 %) 22 and < 25 25 and 28 194 157 256369 178645 4 17 RR=1.2 (0.94-1.43) High SES by single parent living with children (0-7.6 %) 22 and < 25 25 and 28 78 502 51735 307609 17 6 RR=1.1 (0.85-1.38) Total 22 and < 25 25 and 28 272 659 308,104 486,254 21 23 RR=1.5 (1.33-1.77)

PAGE 236

221 Appendix D Results of Stratified Frequency Table Analyses (cont.) Environmental exposure to sulfur dioxide (SO 2 ) Table D-15. Association between environmenta l exposure to sulfur dioxide (SO2) and children asthma hospital admissions stratified by the percentage of single parent living with children (SES – socioeconomic status; RRrate ratio; 95% CI – 95% Confidence Intervals) SES categories (strata) by the percentage single parent living with children Environmental Exposure to SO2, ppm Children h ospitalizations for asthma Total population Number of zip code areas of residence Rate Ratio (95% CI) Low SES by single parent living with children (7.615.2 %) 3.5 and < 5.2 5.2 and 7.0 206 55 98971 13583 7 14 RR=1.95 (1.41-2.62) High SES by single parent living with children (0-7.6 %) 3.5 and < 5.2 5.2 and 7.0 64 456 21726 81217 19 4 RR=1.9 (1.47-2.47) Total 3.5 and < 5.2 5.2 and 7.0 270 511 120,427 94,289 26 18 RR=2.4 (2.08-2.79) Table D-16. Association between environmenta l exposure to sulfur dioxide (SO2) and adult asthma hospital admissions stratified by th e percentage of singl e parent living with children (SES – socioeconomic status; RRrate ratio; 95% CI – 95% Confidence Intervals) SES categories (strata) by the percentage single parent living with children Environmental Exposure to SO2, ppm A dults h ospitalizations for asthma Total population Number of zip code areas of residence Rate Ratio (95% CI) Low SES by single parent living with children (7.615.2 %) 3.5 and < 5.2 5.2 and 7.0 281 70 364762 70252 7 14 RR=1.3 (1.00-1.68) High SES by single parent living with children (0-7.6 %) 3.5 and <5.2 5.2 and 7.0 117 463 75225 284119 19 4 RR=1.1 (0.86-1.28) Total 3.5 and < 5.2 5.2 and 7.0 398 533 439,589 353,838 26 18 RR=1.7 (1.46-1.89)

PAGE 237

222 Appendix D Results of Stratified Frequency Table Analyses (cont.) Overcrowded living conditions Environmental exposure to particulate maters (PM 10 ) Table D-17. Association between environmental expos ure to particulate matters with a diameter of 10 microns or less (PM10) and children asthma hospital admissions stratified by overcrowded housing conditions (SES – socio economic status; RRrate ratio; 95% CI – 95% Confidence Intervals) SES categories (strata) by overcrowding Environmental exposure to PM10, g/m3 Children hospital admissions for asthma Total population Number of zip code areas of residence Rate Ratio (95% CI) Low SES by living in overcrowded conditions (9.5-28.1 %) 22 and < 25 25 and 28 84 326 38022 90410 2 8 RR=1.6 (1.28-2.07) High SES by living in overcrowded conditions (0-9.5 %) 22 and < 25 25 and 28 15 265 5074 46167 19 15 RR=1.9 (1.16-3.26) Total 22 and < 25 25 and 28 190 591 78,920 136,577 21 23 RR=1.8 (1.53-2.12) Table D-18. Association between environmental expos ure to particulate matters with a diameter of 10 microns or less (PM10) and adult asthma hospita l admissions stratified by overcrowded housing conditions (SES – socio economic status; RRrate ratio; 95% CI – 95% Confidence Intervals) SES categories (strata) by overcrowding Environmental exposure to PM10, g/m3 Adults hospital admissions for asthma Total population Number of zip code areas of residence Rate Ratio (95% CI) Low SES by living in overcrowded conditions (9.5-28.1 %) 22 and < 25 25 and 28 141 380 154160 326411 2 8 RR=1.3 (1.00-1.54) High SES by living in overcrowded conditions (0-9.5 %) 22 and < 25 25 and 28 20 279 15802 159843 19 15 RR=1.4 (0.88-2.17) Total 22 and < 25 25 and 28 272 659 308,104 486,254 21 23 RR=1.5 (1.33-1.77)

PAGE 238

223 Appendix D Results of Stratified Frequency Table Analyses (cont.) Environmental exposure to sulfur dioxide (SO 2 ) Table D-19 Association between environmenta l exposure to sulfur dioxide (SO2) and children asthma hospital admissions stratifie d by overcrowded housing conditions (SES – socioeconomic status; RRrate ratio; 95% CI – 95% Confidence Intervals) SES categories (strata) by overcrowding Environmental Exposure to SO2, ppm Children hospital admissions for asthma Total population Number of zip code areas of residence Rate Ratio (95% CI) Low SES by living in overcrowded conditions (9.5-28.1 %) 3.5 and < 5.2 5.2 and 7.0 248 253 112818 51438 3 7 RR=2.2 (1.88-2.66) High SES living in overcrowded conditions (0-9.5 %) 3.5 and < 5.2 5.2 and 7.0 22 258 7879 43362 23 11 RR=2.1 (1.38-3.29) Total 3.5 and < 5.2 5.2 and 7.0 270 511 120,427 94,289 26 18 RR=2.4 (2.08-2.79) Table D-20 Association between environmenta l exposure to sulfur dioxide (SO2) and adult asthma hospital admissions stratified by overcrowded housi ng conditions (SES – socioeconomic status; RRrate ratio; 95% CI – 95% Confidence Intervals) SES categories (strata) by overcrowding Environmental Exposure to SO2, ppm Adults hospital admissions for asthma Total population Number of zip code areas of residence Rate Ratio (95% CI) Low SES by living in overcrowded conditions (9.5-28.1 %) 3.5 and < 5.2 5.2 and 7.0 369 263 415559 203154 3 7 RR=1.5 (1.24-1.71) High SES by living in overcrowded conditions (0-9.5 %) 3.5 and < 5.2 5.2 and 7.0 29 270 24428 151217 23 11 RR=1.5 (1.03-2.21) Total 3.5 and < 5.2 5.2 and 7.0 398 533 439,589 353,838 26 18 RR=1.7 (1.46-1.89)

PAGE 239

224 Appendix D Results of Stratified Frequency Table Analyses (cont.) Urban Deprivation Index (UDI ) Environmental exposure to particulate maters (PM 10 ) Table D-21. Association between environmental expos ure to particulate matters with a diameter of 10 microns or less (PM10) and children asthma hospital admissions stratified by UDI index (SES – socioeconomic status; RRrate ratio; 95% CI – 95% Confidence Intervals) Socioeconomic status categories (strata) by UDI Environmental exposure to PM10, g/m3 Children hospital admissions for asthma Total population Number of zip code areas of residence Rate Ratio (95% CI) Low socioeconomic status by UDI of 1726 22 and < 25 25 and 28 82 161 41417 56879 8 14 RR=1.4 (1.1-1.86) High socioeconomic status by UDI of 8-17 22 and < 25 25 and 28 108 424 37503 73369 13 9 RR=2.0 (1.63-2.48) Total 22 and < 25 25 and 28 190 591 78,920 136,577 21 23 RR=1.8 (1.53-2.12) Table D-22. Association between environmental expos ure to particulate matters with a diameter of 10 microns or less (PM10) and children asthma hospital admissions stratified by UDI index (SES – socioeconomic status; RRrate ratio; 95% CI – 95% Confidence Intervals) Socioeconomic status categories (strata) by UDI Environmental exposure to PM10, g/m3 Adults hospital admissions for asthma Total population Number of zip code areas of residence Rate Ratio (95% CI) Low socioeconomic by UDI of 17-26 22 and < 25 25 and 28 91 204 157328 212313 8 14 RR=1.6 (1.3-2.13) High socioeconomic by UDI score of 8-17 22 and < 25 25 and 28 181 455 150776 253980 13 9 RR=1.5 (1.25-1.78) Total 22 and < 25 25 and 28 272 659 308,104 486,254 21 23 RR=1.5 (1.33-1.77)

PAGE 240

225 Appendix D Results of Stratified Frequency Table Analyses (cont.) Environmental exposure to sulfur dioxide (SO 2 ) Table D-23. Association between environmenta l exposure to sulfur dioxide (SO2) and children asthma hospital admi ssions stratified by UDI index (SES – socioeconomic status; RRrate ratio; 95% CI – 95% Confidence Intervals) Socioeconomic status categories (strata) by UDI Environmental Exposure to SO2, ppm Children hospital admissions for asthma Total population Number of zip code areas of residence Rate Ratio (95% CI) Low socioeconomic by UDI of 17-26 3.5 and < 5.2 5.2 and 7.0 136 107 67090 32106 11 11 RR=1.7 (1.31-2.18) High socioeconomic by UDI score of 8-17 3.5 and < 5.2 5.2 and 7.0 128 404 47278 63594 15 7 RR=2.4 (2.08-2.79 Total 3.5 and < 5.2 5.2 and 7.0 270 511 120,427 94,289 26 18 RR=2.4 (2.08-2.79) Table D-24 Association between environmenta l exposure to sulfur dioxide (SO2) and adult asthma hospital admissions stratified by UDI index (SES – socioeconomic status; RRrate ratio; 95% CI – 95% Confidence Intervals) Socioeconomic status categories (strata) by UDI Environmental Exposure to SO2, ppm Adults hospital admissions for asthma Total population Number of zip code areas of residence Rate Ratio (95% CI) Low socioeconomic by UDI of 17-26 3.5 and < 5.2 5.2 and 7.0 169 126 237134 132507 11 11 RR=1.3 (1.06-1.69) High socioeconomic by UDI score of 8-17 3.5 and < 5.2 5.2 and 7.0 229 407 182892 221864 15 7 RR=1.4 (1.25-1.72) Total 3.5 and < 5.2 5.2 and 7.0 398 533 439,589 353,838 26 18 RR=1.7 (1.46-1.89)

PAGE 241

226 Appendix E. Principal Component Analysis (PCA) Statistical Principal Component Analysis Outcome 00 : 23 Tuesday, July 20 2004 The FACTOR Procedure Means and Standard Deviations from 44 Observations Variable Mean Std Dev Poverty 9.863636 7.037312 Education_9 7.009091 6.045249 Unemployment 3.520455 1.918524 Occupationprof 32.227273 11.573755 Inchouse_15000 14.759091 7.905027 Houseyr_1960 18.800000 19.724356 Novehicle 7.809091 6.215989 Househeating_fuel 0.681818 0.627017 Lackingfacilities 0.954545 0.906938 Occupants_over1 6.634091 5.133064 Singleh_children 7.756818 3.364410 Ethnicminorhh 20.345455 15.022161 Factorial Analysis: All Factors The FACTOR Procedure Initial Factor Method: Principal Factors Prior Communality Estimates: SMC Education_ Inchouse_ Houseyr_ Poverty 9 Unemployment Occupationprof 15000 1960 Novehicle 0.9418095 0.9091614 0.57277793 0.74331869 0.90053438 0.64458572 0.89357079 Househeating_ Occupants_ Singleh_ fuel Lackingfacilities over1 children Ethnicminorhh 0.68351854 0.77796501 0.91752546 0.92464636 0.90680777 Eigenvalues of the Reduced Correlation Matrix: Total = 9.81622159 Average = 0.81801847 Eigenvalue Difference Proportion Cumulative 1 7.00939119 5.57098156 0.7141 0.7141 2 1.43840963 0.80534525 0.1465 0.8606 3 0.63306438 0.26009731 0.0645 0.9251 4 0.37296707 0.00816266 0.0380 0.9631 5 0.36480440 0.16341926 0.0372 1.0002 6 0.20138515 0.14776257 0.0205 1.0208 7 0.05362258 0.02876487 0.0055 1.0262 8 0.02485770 0.06118532 0.0025 1.0288 9 .03632762 0.01409821 0.0037 1.0251 10 .05042583 0.01171828 0.0051 1.0199 11 .06214411 0.07123884 0.0063 1.0136 12 .13338295 0.0136 1.0000 5 factors will be retained by the PROPORTION criterion. Factorial Analysis: All Factors The FACTOR Procedure Initial Factor Method: Principal Factors

PAGE 242

227 Appendix E. Principal Component Analysis (PCA) (cont.) Scree Plot of Eigenvalues ‚ 7 ˆ 1 ‚ ‚ ‚ ‚ 6 ˆ ‚ ‚ ‚ ‚ 5 ˆ ‚ ‚ ‚ ‚ E 4 ˆ i ‚ g ‚ e ‚ n ‚ v 3 ˆ a ‚ l ‚ u ‚ e ‚ s 2 ˆ ‚ ‚ ‚ 2 ‚ 1 ˆ ‚ ‚ 3 ‚ 4 5 ‚ 6 0 ˆ 7 8 9 0 1 ‚ 2 ‚ ‚ ‚ 1 ˆ ‚ Šƒƒˆƒƒƒƒƒƒˆƒƒƒƒƒˆƒƒƒƒƒƒˆƒƒƒƒƒƒˆƒƒƒƒƒƒˆƒƒƒƒƒƒˆƒƒƒƒƒƒˆƒƒƒƒƒƒˆƒƒƒƒƒƒˆƒƒƒƒƒƒˆƒƒƒƒƒƒˆƒƒƒƒƒƒˆƒƒ 0 1 2 3 4 5 6 7 8 9 10 11 12 Number Factorial Analysis: All Factors 85 00 : 23 Tuesday, July 20 2004 The FACTOR Procedure Initial Factor Method: Principal Factors Factor Pattern Factor1 Factor2 Factor3 Factor4 Factor5 Poverty 0.95604 0.06621 0.01235 0.15941 0.00415 Education_9 0.73141 0.57556 0.15277 0.15096 0.12220 Unemployment 0.62415 0.19115 0.27527 0.06449 0.22187 Occupationprof 0.64353 0.45874 0.26610 0.04511 0.18968 Inchouse_15000 0.88870 0.14902 0.10512 0.24773 0.14719 Houseyr_1960 0.52884 0.47715 0.10317 0.38155 0.06389 Novehicle 0.85725 0.35359 0.12557 0.15122 0.03057

PAGE 243

228 Appendix E. Principal Component Analysis (PCA) (cont.) Househeating_fuel 0.61610 0.34419 0.05225 0.21273 0.37786 Lackingfacilities 0.74323 0.19508 0.38053 0.10437 0.00532 Occupants_over1 0.76940 0.51148 0.02832 0.17098 0.22991 Singleh_children 0.82061 0.15962 0.47709 0.05277 0.12293 Ethnicminorhh 0.86893 0.19307 0.22265 0.02420 0.16491 Variance Explained by Each Factor Factor1 Factor2 Factor3 Factor4 Factor5 7.0093912 1.4384096 0.6330644 0.3729671 0.3648044 Final Communality Estimates: Total = 9.818637 Education_ Inchouse_ Houseyr_ Poverty 9 Unemployment Occupationprof 15000 1960 Novehicle 0.9439808 0.92727862 0.55526567 0.73339843 0.90608104 0.66764444 0.89947938 Househeating_ Occupants_ Singleh_ fuel Lackingfacilities over1 children Ethnicminorhh 0.68880176 0.74616427 0.93648280 0.94439265 0.86966679 Factorial Analysis: All Factors The FACTOR Procedure Prerotation Method: Varimax Orthogonal Transformation Matrix 1 2 3 4 5 1 0.53703 0.52930 0.49518 0.34411 0.26043 2 0.79384 0.15683 0.21923 0.45533 0.29972 3 0.01816 0.56813 0.75663 0.29172 0.13894 4 0.28134 0.60947 0.23362 0.66820 0.21984 5 0.04402 0.03189 0.28225 0.37772 0.88017 Rotated Factor Pattern Factor1 Factor2 Factor3 Factor4 Factor5 Poverty 0.52072 0.58566 0.50665 0.19029 0.19216 Education_9 0.90031 0.29556 0.05065 0.08890 0.13749 Unemployment 0.19682 0.47035 0.19026 0.50902 0.00049 Occupationprof 0.72597 0.09605 0.35536 0.10656 0.24395 Inchouse_15000 0.29765 0.70916 0.40953 0.18321 0.33659 Houseyr_1960 0.01163 0.17878 0.21731 0.70842 0.29405 Novehicle 0.14076 0.67368 0.43370 0.38003 0.30545 Househeating_fuel 0.13317 0.23279 0.26372 0.35291 0.65022 Lackingfacilities 0.53130 0.64243 0.06322 0.21021 0.05459 Occupants_over1 0.85669 0.19940 0.31523 0.22470 0.11376 Singleh_children 0.28506 0.21657 0.84935 0.22706 0.20804 Ethnicminorhh 0.29526 0.37320 0.69327 0.36809 0.16463 Variance Explained by Each Factor Factor1 Factor2 Factor3 Factor4 Factor5 2.9583918 2.3423882 2.1996925 1.4006714 0.9174928 Final Communality Estimates: Total = 9.818637

PAGE 244

229 Appendix E. Principal Component Analysis (PCA) (cont.) Education_ Inchouse_ Houseyr_ Poverty 9 Unemployment Occupationprof 15000 1960 Novehicle 0.9439808 0.92727862 0.55526567 0.73339843 0.90608104 0.66764444 0.89947938 Househeating_ Occupants_ Singleh_ fuel Lackingfacilities over1 children Ethnicminorhh 0.68880176 0.74616427 0.93648280 0.94439265 0.86966679 Factorial Analysis: All Factors The FACTOR Procedure Prerotation Method: Varimax Scoring Coefficients Estimated by Regression Squared Multiple Correlations of the Variables with Each Factor Factor1 Factor2 Factor3 Factor4 Factor5 0.93604051 0.83994683 0.88751084 0.69969055 0.68736877 Standardized Scoring Coefficients Factor1 Factor2 Factor3 Factor4 Factor5 Poverty 0.0900004 0.74127728 0.19069678 0.5388132 0.2747786 Education_9 0.55541344 0.1823784 0.3581759 0.1566692 0.80261666 Unemployment 0.0247237 0.06631889 0.1180558 0.37687599 0.2623286 Occupationprof 0.129206 0.09742638 0.0250212 0.2284694 0.2346521 Inchouse_15000 0.0147006 0.45994343 0.0813797 0.2599096 0.17046002 Houseyr_1960 0.0228924 0.1270672 0.0827875 0.40599101 0.22886721 Novehicle 0.2186091 0.19940346 0.0526004 0.35525167 0.11053001 Househeating_fuel 0.04968591 0.1963758 0.1359695 0.24064348 0.40083712 Lackingfacilities 0.0299189 0.30458739 0.1100149 0.0803787 0.0589363 Occupants_over1 0.51178599 0.3343239 0.04559336 0.6459242 1.0150688 Singleh_children 0.01835532 0.6188101 0.94380078 0.1482017 0.25840976 Ethnicminorhh 0.1247892 0.13613231 0.17822734 0.1234784 0.0752074 Factorial Analysis: All Factors The FACTOR Procedure Rotation Method: Promax (power = 3 ) Target Matrix for Procrustean Transformation Factor1 Factor2 Factor3 Factor4 Factor5 Poverty 0.18837 0.52966 0.21240 0.01153 0.01609 Education_9 1.00000 0.06992 0.00022 0.00121 0.00605 Unemployment 0.02255 0.60818 0.02493 0.48909 0.00000 Occupationprof 0.74539 0.00341 0.10702 0.00296 0.04807 Inchouse_15000 0.03741 1.00000 0.11928 0.01094 0.09195 Houseyr_1960 0.00000 0.02533 0.02818 1.00000 0.09692 Novehicle 0.00400 0.86674 0.14323 0.09872 0.06947 Househeating_fuel 0.00505 0.05337 0.04806 0.11797 1.00000 Lackingfacilities 0.28472 0.99480 0.00059 0.02211 0.00053 Occupants_over1 0.84892 0.02116 0.05177 0.01921 0.00338 Singleh_children 0.03088 0.02677 1.00000 0.01957 0.02040 Ethnicminorhh 0.03884 0.15500 0.61537 0.09435 0.01144

PAGE 245

230 Appendix E. Principal Component Analysis (PCA) (cont.) Procrustean Transformation Matrix 1 2 3 4 5 1 1.15487157 0.2771411 0.1386402 0.02884896 0.05014828 2 0.3720236 1.76239364 0.1923852 0.3351254 0.2485143 3 0.2812513 0.2591224 1.12552624 0.2345505 0.2570311 4 0.091164 0.3026886 0.1101431 1.2636828 0.12853892 5 0.08668962 0.1842499 0.2458185 0.02258624 1.20471094 Normalized Oblique Transformation Matrix 1 2 3 4 5 1 0.33540944 0.40264935 0.32927514 0.14244965 0.11903458 2 0.961974 0.19714757 0.23923978 0.39862331 0.26854944 3 0.0366815 0.90325145 1.1341057 0.30915943 0.0714459 4 0.69206311 1.0669414 0.3684512 0.97756941 0.544094 5 0.1590138 0.0554133 0.59534021 0.35079378 1.018255 Factorial Analysis: All Factors The FACTOR Procedure Rotation Method: Promax (power = 3 ) Inter-Factor Correlations Factor1 Factor2 Factor3 Factor4 Factor5 Factor1 1.00000 0.51644 0.45615 0.12274 0.12942 Factor2 0.51644 1.00000 0.59589 0.44133 0.40197 Factor3 0.45615 0.59589 1.00000 0.35469 0.47038 Factor4 0.12274 0.44133 0.35469 1.00000 0.08452 Factor5 0.12942 0.40197 0.47038 0.08452 1.00000 Rotated Factor Pattern (Standardized Regression Coefficients) Factor1 Factor2 Factor3 Factor4 Factor5 Poverty 0.27292 0.53060 0.37416 0.04840 0.00595 Education_9 0.92850 0.16473 0.19849 0.02670 0.12815 Unemployment 0.04491 0.45653 0.04740 0.39108 0.08487 Occupationprof 0.70876 0.10930 0.27439 0.19590 0.19011 Inchouse_15000 0.01054 0.75464 0.21271 0.07531 0.15338 Houseyr_1960 0.02394 0.01044 0.06873 0.69284 0.32626 Novehicle 0.14780 0.69135 0.26197 0.14333 0.13688 Househeating_fuel 0.08094 0.06270 0.04113 0.28422 0.67000 Lackingfacilities 0.37783 0.71558 0.19189 0.04559 0.05331 Occupants_over1 0.83083 0.01178 0.23697 0.14475 0.18483 Singleh_children 0.04813 0.01956 0.94209 0.02457 0.02075 Ethnicminorhh 0.05458 0.20351 0.69191 0.16610 0.00990 Reference Axis Correlations Factor1 Factor2 Factor3 Factor4 Factor5 Factor1 1.00000 0.41149 0.29513 0.20508 0.21409 Factor2 0.41149 1.00000 0.23666 0.37870 0.27819 Factor3 0.29513 0.23666 1.00000 0.22186 0.37412 Factor4 0.20508 0.37870 0.22186 1.00000 0.19923 Factor5 0.21409 0.27819 0.37412 0.19923 1.00000

PAGE 246

231 Appendix E. Principal Component Analysis (PCA) (cont.) Factorial Analysis: All Factors The FACTOR Procedure Rotation Method: Promax (power = 3 ) Reference Structure (Semipartial Correlations) Factor1 Factor2 Factor3 Factor4 Factor5 Poverty 0.21968 0.36117 0.26849 0.04160 0.00499 Education_9 0.74736 0.11213 0.14243 0.02295 0.10738 Unemployment 0.03615 0.31076 0.03401 0.33616 0.07111 Occupationprof 0.57049 0.07440 0.19690 0.16838 0.15930 Inchouse_15000 0.00848 0.51367 0.15263 0.06473 0.12852 Houseyr_1960 0.01927 0.00711 0.04932 0.59553 0.27338 Novehicle 0.11897 0.47059 0.18798 0.12320 0.11469 Househeating_fuel 0.06515 0.04268 0.02952 0.24430 0.56141 Lackingfacilities 0.30412 0.48708 0.13769 0.03919 0.04467 Occupants_over1 0.66874 0.00802 0.17004 0.12442 0.15487 Singleh_children 0.03874 0.01332 0.67603 0.02112 0.01739 Ethnicminorhh 0.04393 0.13853 0.49650 0.14277 0.00830 Variance Explained by Each Factor Eliminating Other Factors Factor1 Factor2 Factor3 Factor4 Factor5 1.4955581 0.9889773 0.9456430 0.6151745 0.4879279 Factor Structure (Correlations) Factor1 Factor2 Factor3 Factor4 Factor5 Poverty 0.71245 0.87553 0.80046 0.35248 0.42646 Education_9 0.94289 0.58925 0.39295 0.15379 0.22342 Unemployment 0.33932 0.64645 0.43872 0.60772 0.15980 Occupationprof 0.77804 0.41020 0.55250 0.04375 0.35042 Inchouse_15000 0.50789 0.91525 0.71263 0.34744 0.55177 Houseyr_1960 0.12928 0.45506 0.45080 0.73724 0.40985 Novehicle 0.36404 0.88940 0.72174 0.53479 0.53098 Househeating_fuel 0.25368 0.52376 0.53138 0.39304 0.74905 Lackingfacilities 0.65855 0.79505 0.39796 0.33521 0.19682 Occupants_over1 0.92668 0.54809 0.57333 0.30996 0.04166 Singleh_children 0.47346 0.58586 0.97086 0.35775 0.46433 Ethnicminorhh 0.49440 0.71332 0.89233 0.50719 0.41847 Variance Explained by Each Factor Ignoring Other Factors Factor1 Factor2 Factor3 Factor4 Factor5 4.3496607 5.5946183 5.0331196 2.2175111 2.1267576 Factorial Analysis: All Factors The FACTOR Procedure Rotation Method: Promax (power = 3 ) Final Communality Estimates: Total = 9.818637 Education_ Inchouse_ Houseyr_ Poverty 9 Unemployment Occupationprof 15000 1960 Novehicle 0.9439808 0.92727862 0.55526567 0.73339843 0.90608104 0.66764444 0.89947938 Househeating_ Occupants_ Singleh_ fuel Lackingfacilities over1 children Ethnicminorhh 0.68880176 0.74616427 0.93648280 0.94439265 0.86966679

PAGE 247

232 Appendix E. Principal Component Analysis (PCA) (cont.) Factorial Analysis: All Factors The FACTOR Procedure Rotation Method: Promax (power = 3 ) Scoring Coefficients Estimated by Regression Squared Multiple Correlations of the Variables with Each Factor Factor1 Factor2 Factor3 Factor4 Factor5 0.95675346 0.94571859 0.95551410 0.74420049 0.76240783 Standardized Scoring Coefficients Factor1 Factor2 Factor3 Factor4 Factor5 Poverty 0.04535 0.46318 0.14429 0.35685 0.14204 Education_9 0.45682 0.03264 0.08781 0.25423 0.65744 Unemployment 0.01601 0.06954 0.06793 0.37596 0.24430 Occupationprof 0.10377 0.05406 0.03150 0.24545 0.20378 Inchouse_15000 0.04897 0.33412 0.00792 0.18625 0.19472 Houseyr_1960 0.02990 0.01386 0.02570 0.35287 0.20572 Novehicle 0.15867 0.20684 0.04199 0.36885 0.14028 Househeating_fuel 0.01364 0.04811 0.01965 0.16656 0.33728 Lackingfacilities 0.06221 0.21073 0.04953 0.03293 0.04129 Occupants_over1 0.44848 0.14311 0.02938 0.62327 0.94836 Singleh_children 0.05322 0.27885 0.72284 0.15544 0.37408 Ethnicminorhh 0.06080 0.14238 0.16912 0.16790 0.00288 Factorial Analysis: Only One Factors Means and Standard Deviations from 44 Observations Variable Mean Std Dev Poverty 9.863636 7.037312 Education_9 7.009091 6.045249 Unemployment 3.520455 1.918524 Occupationprof 32.227273 11.573755 Inchouse_15000 14.759091 7.905027 Houseyr_1960 18.800000 19.724356 Novehicle 7.809091 6.215989 Househeating_fuel 0.681818 0.627017 Lackingfacilities 0.954545 0.906938 Occupants_over1 6.634091 5.133064 Singleh_children 7.756818 3.364410 Ethnicminorhh 20.345455 15.022161 Factorial Analysis: Only One Factors The FACTOR Procedure Initial Factor Method: Principal Factors Prior Communality Estimates: SMC Education_ Inchouse_ Houseyr_ Poverty 9 Unemployment Occupationprof 15000 1960 Novehicle 0.9418095 0.90916140 0.57277793 0.74331869 0.90053438 0.64458572 0.89357079 Househeating_ Occupants_ Singleh_ fuel Lackingfacilities over1 children Ethnicminorhh 0.68351854 0.77796501 0.91752546 0.92464636 0.90680777

PAGE 248

233 Appendix E. Principal Component Analysis (PCA) (cont.) Eigenvalues of the Reduced Correlation Matrix: Total = 9.81622159 Average = 0.81801847 Eigenvalue Difference Proportion Cumulative 1 7.00939119 5.57098156 0.7141 0.7141 2 1.43840963 0.80534525 0.1465 0.8606 3 0.63306438 0.26009731 0.0645 0.9251 4 0.37296707 0.00816266 0.0380 0.9631 5 0.36480440 0.16341926 0.0372 1.0002 6 0.20138515 0.14776257 0.0205 1.0208 7 0.05362258 0.02876487 0.0055 1.0262 8 0.02485770 0.06118532 0.0025 1.0288 9 .03632762 0.01409821 0.0037 1.0251 10 .05042583 0.01171828 0.0051 1.0199 11 .06214411 0.07123884 0.0063 1.0136 12 .13338295 0.0136 1.0000 1 factor will be retained by the NFACTOR criterion. Initial Factor Method: Principal Factors Scree Plot of Eigenvalues ‚ 7 ˆ 1 ‚ ‚ ‚ ‚ 6 ˆ ‚ ‚ ‚ ‚ 5 ˆ ‚ ‚ ‚ ‚ E 4 ˆ i ‚ g ‚ e ‚ n ‚ v 3 ˆ a ‚ l ‚ u ‚ e ‚ s 2 ˆ ‚ ‚ ‚ 2 ‚ 1 ˆ ‚ ‚ 3 ‚ 4 5 ‚ 6 0 ˆ 7 8 9 0 1 ‚ 2 ‚ ‚ 1 ˆ ‚ Šƒƒˆƒƒƒƒƒˆƒƒƒƒƒˆƒƒƒƒƒˆƒƒƒƒƒˆƒƒƒƒƒˆƒƒƒƒƒˆƒƒƒƒƒˆƒƒƒƒƒˆƒƒƒƒƒˆƒƒƒƒƒˆƒƒƒƒƒˆƒƒƒƒƒˆƒƒ 0 1 2 3 4 5 6 7 8 9 10 11 12

PAGE 249

234 Appendix E. Principal Component Analysis (PCA) (cont.) Number Factorial Analysis: Only One Factors The FACTOR Procedure Initial Factor Method: Principal Factors Factor Pattern Factor1 Poverty 0.95604 Education_9 0.73141 Unemployment 0.62415 Occupationprof 0.64353 Inchouse_15000 0.88870 Houseyr_1960 0.52884 Novehicle 0.85725 Househeating_fuel 0.61610 Lackingfacilities 0.74323 Occupants_over1 0.76940 Singleh_children 0.82061 Ethnicminorhh 0.86893 Variance Explained by Each Factor Factor1 7.0093912 Final Communality Estimates: Total = 7.009391 Education_ Inchouse_ Houseyr_ Poverty 9 Unemployment Occupationprof 15000 1960 Novehicle 0.9140166 0.5349540 0.38956819 0.41413509 0.78978776 0.27966755 0.734883 Househeating_ Occupants_ Singleh_ fuel Lackingfacilities over1 children Ethnicminorhh 0.37957771 0.55238347 0.59197810 0.67340173 0.75503785 Factorial Analysis: Only One Factors The FACTOR Procedure Prerotation Method: Varimax Factorial Analysis: Only One Factors The FACTOR Procedure Prerotation Method: Varimax Scoring Coefficients Estimated by Regression Squared Multiple Correlations of the Variables with Each Factor Factor1 0.98355880

PAGE 250

235 Appendix E. Principal Component Analysis (PCA) (cont.) Standardized Scoring Coefficients Factor1 Poverty 0.18148 Education_9 0.17949 Unemployment 0.02474 Occupationprof 0.01270 Inchouse_15000 0.15021 Houseyr_1960 0.07877 Novehicle 0.11313 Househeating_fuel 0.04261 Lackingfacilities 0.07980 Occupants_over1 0.07838 Singleh_children 0.16597 Ethnicminorhh 0.11620 Factorial Analysis: Only One Factors The FACTOR Procedure Prerotation Method: Varimax Scoring Coefficients Estimated by Regression Squared Multiple Correlations of the Variables with Each Factor Factor1 0.98355880 Standardized Scoring Coefficients Factor1 Poverty 0.18148 Education_9 0.17949 Unemployment 0.02474 Occupationprof 0.01270 Inchouse_15000 0.15021 Houseyr_1960 0.07877 Novehicle 0.11313 Househeating_fuel 0.04261 Lackingfacilities 0.07980 Occupants_over1 0.07838 Singleh_children 0.16597 Ethnicminorhh 0.11620 Final Model. Factorial Analysis: Only One Factors The FACTOR Procedure Means and Standard Deviations from 44 Observations Variable Mean Std Dev Poverty 9.863636 7.037312 Education_9 7.009091 6.045249 Unemployment 3.520455 1.918524 Occupationprof 32.227273 11.573755 Inchouse_15000 14.759091 7.905027 Novehicle 7.809091 6.215989 Lackingfacilities 0.954545 0.906938 Occupants_over1 6.634091 5.133064

PAGE 251

236 Appendix E. Principal Component Analysis (PCA) (cont.) Final Model. Factorial Analysis: Only One Factors The FACTOR Procedure Initial Factor Method: Principal Factors Prior Communality Estimates: SMC Education_ Poverty 9 Unemployment Occupationprof 0.91682363 0.81604912 0.42460573 0.65941184 Inchouse_ Occupants_ 15000 Novehicle Lackingfacilities over1 0.89737430 0.85117698 0.66203687 0.83345444 Eigenvalues of the Reduced Correlation Matrix: Total = 6.06093292 Average = 0.75761661 Eigenvalue Difference Proportion Cumulative 1 5.07848078 4.11205732 0.8379 0.8379 2 0.96642345 0.73567943 0.1595 0.9974 3 0.23074402 0.18496420 0.0381 1.0354 4 0.04577982 0.07725335 0.0076 1.0430 5 .03147353 0.02608598 0.0052 1.0378 6 .05755951 0.01131250 0.0095 1.0283 7 .06887202 0.03371807 0.0114 1.0169 8 .10259009 0.0169 1.0000 1 factor will be retained by the NFACTOR criterion.

PAGE 252

237 Appendix E. Principal Component Analysis (PCA) (cont.) Final Model. Factorial Analysis: Only One Factors The FACTOR Procedure Initial Factor Method: Principal Factors Scree Plot of Eigenvalues 6 ˆ ‚ ‚ ‚ ‚ ‚ 5 ˆ 1 ‚ ‚ ‚ ‚ ‚ 4 ˆ ‚ ‚ ‚ E ‚ i ‚ g 3 ˆ e ‚ n ‚ v ‚ a ‚ l ‚ u 2 ˆ e ‚ s ‚ ‚ ‚ ‚ 1 ˆ 2 ‚ ‚ ‚ ‚ ‚ 3 0 ˆ 4 5 6 7 ‚ 8 ‚ ‚ ‚ ‚ 1 ˆ Šƒƒƒƒƒƒƒƒˆƒƒƒƒƒƒƒƒˆƒƒƒƒƒƒƒƒˆƒƒƒƒƒƒƒƒˆƒƒƒƒƒƒƒƒˆƒƒƒƒƒƒƒƒˆƒƒƒƒƒƒƒƒˆƒƒƒƒƒƒƒƒˆƒƒƒƒƒƒƒƒˆƒƒƒƒƒƒƒ 0 1 2 3 4 5 6 7 8 Number Final Model. Factorial Analysis: Only One Factors The FACTOR Procedure Initial Factor Method: Principal Factors

PAGE 253

238 Appendix E. Principal Component Analysis (PCA) (cont.) Factor Pattern Factor1 Poverty 0.95604 Education_9 0.81773 Unemployment 0.59216 Occupationprof 0.67651 Inchouse_15000 0.87736 Novehicle 0.79626 Lackingfacilities 0.78550 Occupants_over1 0.81650 Variance Explained by Each Factor Factor1 5.0784808 Final Communality Estimates: Total = 5.078481 Education_ Poverty 9 Unemployment Occupationprof 0.91401946 0.66868009 0.35064893 0.45766550 Inchouse_ Occupants_ 15000 Novehicle Lackingfacilities over1 0.76975271 0.63402929 0.61701199 0.66667280 Final Model. Factorial Analysis: Only One Factors The FACTOR Procedure Prerotation Method: Varimax Final Model. Factorial Analysis: Only One Factors The FACTOR Procedure Prerotation Method: Varimax Scoring Coefficients Estimated by Regression Squared Multiple Correlations of the Variables with Each Factor Factor1 0.97303066 Standardized Scoring Coefficients Factor1 Poverty 0.31882 Education_9 0.18410 Unemployment 0.01822 Occupationprof 0.00349 Inchouse_15000 0.27464 Novehicle 0.06785 Lackingfacilities 0.06860 Occupants_over1 0.19064 Final Model. Factorial Analysis: Only One Factors

PAGE 254

239 Appendix E. Principal Component Analysis (PCA) (cont.) The FACTOR Procedure Prerotation Method: Varimax Scoring Coefficients Estimated by Regression Squared Multiple Correlations of the Variables with Each Factor Factor1 0.97303066 Standardized Scoring Coefficients Factor1 Poverty 0.31882 Education_9 0.18410 Unemployment 0.01822 Occupationprof 0.00349 Inchouse_15000 0.27464 Novehicle 0.06785 Lackingfacilities 0.06860 Occupants_over1 0.19064

PAGE 255

240 Appendix F. Poisson Regression Analysis Outc ome for Adult Hospital Admissions Adult Asthma and SDI Regression Model 57 19:47 Saturday, October 9, 2004 The GENMOD Procedure Model Information Data Set WORK.ASTHMA Distribution Poisson Link Function Log Dependent Variable Asthma_adult Offset Variable ln1 Observations Used 44 Criteria For Assessing Goodness Of Fit Criterion DF Value Value/DF Deviance 39 110.1364 2.8240 Scaled Deviance 39 46.2700 1.1864 Pearson Chi-Square 39 92.8316 2.3803 Scaled Pearson X2 39 39.0000 1.0000 Log Likelihood 885.6341 Analysis Of Parameter Estimates Standard Wald 95% Confidence ChiParameter DF Estimate Error Limits Square Pr > ChiSq Intercept 1 -8.6305 4.1189 -16.7033 -0.5576 4.39 0.0361 SDI 1 0.0583 0.0088 0.0410 0.0756 43.77 <.0001 PM10_99 1 0.0583 0.0493 -0.0384 0.1550 1.40 0.2375 O3_99 1 -0.0029 0.0447 -0.0905 0.0848 0.00 0.9487 SO2_99 1 0.0029 0.1313 -0.2544 0.2601 0.00 0.9824 Scale 0 1.5428 0.0000 1.5428 1.5428 NOTE: The scale parameter was estimated by the square root of Pearson's Chi-Square/DOF. Adult Asthma and SDI Regression Model The GENMOD Procedure Model Information Data Set WORK.ASTHMA Distribution Poisson Link Function Log Dependent Variable Asthma_adult Offset Variable ln1 Observations Used 44 Criteria For Assessing Goodness Of Fit Criterion DF Value Value/DF Deviance 40 110.1376 2.7534 Scaled Deviance 40 47.4608 1.1865 Pearson Chi-Square 40 92.8241 2.3206 Scaled Pearson X2 40 40.0000 1.0000 Log Likelihood 908.4158 Analysis Of Parameter Estimates Standard Wald 95% Confidence ChiParameter DF Estimate Error Limits Square Pr > ChiSq Intercept 1 -8.6785 3.4479 -15.4361 -1.9208 6.34 0.0118 SDI 1 0.0584 0.0085 0.0417 0.0750 47.21 <.0001 PM10_99 1 0.0589 0.0392 -0.0179 0.1357 2.26 0.1326 O3_99 1 -0.0024 0.0374 -0.0756 0.0709 0.00 0.9498 Scale 0 1.5234 0.0000 1.5234 1.5234 NOTE: The scale parameter was estimated by the square root of Pearson's Chi-Square/DOF.

PAGE 256

241 Appendix F. Poisson Regression Analysis Outc ome for Adult Hospital Admissions (cont.) Adult Asthma and SDI Regression Model The GENMOD Procedure Model Information Data Set WORK.ASTHMA Distribution Poisson Link Function Log Dependent Variable Asthma_adult Offset Variable ln1 Observations Used 44 Criteria For Assessing Goodness Of Fit Criterion DF Value Value/DF Deviance 41 110.1468 2.6865 Scaled Deviance 41 48.6381 1.1863 Pearson Chi-Square 41 92.8495 2.2646 Scaled Pearson X2 41 41.0000 1.0000 Log Likelihood 930.8698 Analysis Of Parameter Estimates Standard Wald 95% Confidence ChiParameter DF Estimate Error Limits Square Pr > ChiSq Intercept 1 -8.8871 0.9424 -10.7342 -7.0400 88.92 <.0001 SDI 1 0.0584 0.0084 0.0419 0.0748 48.35 <.0001 PM10_99 1 0.0590 0.0387 -0.0169 0.1349 2.32 0.1275 Scale 0 1.5049 0.0000 1.5049 1.5049 NOTE: The scale parameter was estimated by the square root of Pearson's Chi-Square/DOF. Adult Asthma and SDI Regression Model The GENMOD Procedure Model Information Data Set WORK.ASTHMA Distribution Poisson Link Function Log Dependent Variable Asthma_adult Offset Variable ln1 Observations Used 44 Criteria For Assessing Goodness Of Fit Criterion DF Value Value/DF Deviance 42 115.4582 2.7490 Scaled Deviance 42 48.9626 1.1658 Pearson Chi-Square 42 99.0398 2.3581 Scaled Pearson X2 42 42.0000 1.0000 Log Likelihood 892.8460 Analysis Of Parameter Estimates Standard Wald 95% Confidence ChiParameter DF Estimate Error Limits Square Pr > ChiSq Intercept 1 -7.4637 0.1066 -7.6726 -7.2547 4901.98 <.0001 SDI 1 0.0644 0.0076 0.0495 0.0792 72.43 <.0001 Scale 0 1.5356 0.0000 1.5356 1.5356 NOTE: The scale parameter was estimated by the square root of Pearson's Chi-Square/DOF.

PAGE 257

242 Appendix F. Poisson Regression Analysis Outc ome for Adult Hospital Admissions (cont.) Observation Statistics Asthma_ Observation adult Pred Xbeta Std HessWgt Resraw Reschi Resdev StResdev StReschi Reslik 1 16 13.752662 2.6212324 0.0754123 5.8321178 2.2473376 0.6060032 0.5905303 0.3910986 0.4013461 0.3914428 2 27 26.466093 3.2758644 0.0794333 11.223526 0.5339074 0.1037818 0.1034357 0.0698779 0.0701116 0.0698944 3 9 12.419346 2.5192554 0.052284 5.2666956 -3.419346 -0.970272 -1.020851 -0.669624 -0.636447 -0.669158 4 10 8.5320629 2.1438312 0.055123 3.6182082 1.4679371 0.5025513 0.4890851 0.3202615 0.3290794 0.3203598 5 5 5.7516383 1.7494847 0.0654182 2.4391082 -0.751638 -0.31341 -0.320637 -0.2099 -0.205169 -0.209851 6 18 27.162534 3.3018386 0.0776913 11.518868 -9.162534 -1.758047 -1.874103 -1.265206 -1.186857 -1.259917 7 7 7.2439365 1.9801648 0.0916375 3.07195 -0.243936 -0.090634 -0.09115 -0.060138 -0.059798 -0.060129 8 15 12.979341 2.5633589 0.0592171 5.5041739 2.020659 0.5608758 0.5471927 0.3598257 0.3688236 0.3600015 9 28 23.36183 3.1511035 0.0520841 9.9070957 4.6381703 0.9596067 0.9302242 0.6140775 0.633474 0.6146068 10 18 22.40801 3.1094185 0.0518087 9.5026074 -4.40801 -0.931196 -0.96454 -0.636283 -0.614287 -0.635732 11 31 24.177728 3.1854319 0.0663946 10.253095 6.8222723 1.3874627 1.3288666 0.8856141 0.924665 0.8874162 12 13 14.245583 2.6564469 0.0512408 6.0411514 -1.245583 -0.330014 -0.335007 -0.219911 -0.216633 -0.219859 13 1 4.8199938 1.5727726 0.0794333 2.044024 -3.819994 -1.739961 -2.12001 -1.389557 -1.140454 -1.386629 14 6 13.037114 2.5678002 0.0742928 5.5286736 -7.037114 -1.948964 -2.182141 -1.443218 -1.289 -1.438757 15 23 14.000755 2.6391112 0.0659039 5.9373267 8.9992454 2.4050847 2.1988921 1.4507643 1.5868041 1.4544323 16 16 7.7633777 2.0494175 0.0576803 3.2922304 8.2366223 2.9561318 2.5822924 1.6908955 1.9356871 1.6937685 17 27 26.38636 3.2728472 0.0829936 11.189714 0.6136399 0.1194604 0.1190018 0.0806659 0.0809767 0.0806899 18 7 14.736989 2.6903606 0.1003075 6.2495431 -7.736989 -2.015429 -2.247592 -1.511958 -1.355782 -1.502616 19 26 17.979984 2.8892591 0.0951226 7.624806 8.0200163 1.8913879 1.7718671 1.1958428 1.276508 1.2015819 20 36 27.432999 3.3117466 0.060223 11.633564 8.5670006 1.6356567 1.5600012 1.0380198 1.0883606 1.040193 21 52 50.373469 3.9194646 0.0642134 21.361973 1.6265307 0.2291719 0.2279549 0.1554501 0.15628 0.1555234 22 29 46.276172 3.8346272 0.1315622 19.624425 -17.27617 -2.53962 -2.728942 -2.186926 -2.035207 -2.1366 23 13 10.83067 2.382382 0.07049 4.592983 2.1693296 0.6591707 0.6388217 0.4208356 0.4342409 0.4211463 24 38 33.711281 3.5178325 0.0671258 14.296007 4.2887191 0.7386522 0.7237665 0.487278 0.4972998 0.4879298 25 16 13.137193 2.5754474 0.0596191 5.5711145 2.862807 0.789843 0.763485 0.5021846 0.5195217 0.5025337 26 53 45.151639 3.8100266 0.0666278 19.147542 7.8483608 1.1679983 1.1363992 0.7736426 0.7951547 0.7754943 27 34 23.382973 3.1520081 0.0621697 9.9160618 10.617027 2.1955993 2.0547289 1.3644587 1.4580048 1.3681619 28 63 57.975388 4.0600186 0.0623798 24.585733 5.024612 0.6599039 0.6507004 0.4455917 0.4518941 0.4461985 29 33 38.635226 3.6541645 0.0564247 16.384114 -5.635226 -0.906608 -0.930106 -0.622136 -0.606419 -0.621326 30 62 48.12887 3.8738822 0.0508632 20.410102 13.87113 1.9994428 1.9133766 1.2802651 1.3378531 1.2833705 31 46 36.58775 3.5997135 0.053314 15.515837 9.41225 1.5560574 1.4956162 0.9961716 1.0364291 0.9979812 32 13 10.840806 2.3833173 0.0518087 4.5972811 2.1591943 0.6557842 0.6356449 0.4165149 0.4297114 0.4166803 33 23 38.837221 3.6593791 0.0514706 16.469774 -15.83722 -2.541294 -2.752406 -1.83282 -1.692241 -1.826912 34 9 13.059396 2.5695078 0.0748508 5.5381228 -4.059396 -1.123311 -1.190669 -0.78769 -0.743129 -0.786345

PAGE 258

243 Appendix F. Poisson Regression Analysis Outc ome for Adult Hospital Admissions (cont.) 35 25 33.66 3.5163102 0.055123 14.274261 -8.66 -1.49266 -1.564699 -1.041788 -0.993824 -1.039753 36 1 1.3515051 0.3012189 0.0759772 0.5731354 -0.351505 -0.302359 -0.317132 -0.206861 -0.197225 -0.20683 37 25 26.158266 3.2641652 0.0823936 11.092986 -1.158266 -0.226466 -0.228169 -0.154518 -0.153365 -0.154431 38 7 12.246258 2.5052204 0.0771168 5.193294 -5.246258 -1.49916 -1.631616 -1.07932 -0.991699 -1.07672 39 8 5.2177698 1.6520701 0.0980454 2.2127096 2.7822302 1.2180093 1.1284871 0.7428225 0.8017501 0.7441245 40 7 13.439159 2.5981728 0.0848088 5.6991698 -6.439159 -1.75648 -1.935624 -1.287151 -1.168024 -1.282486 41 10 14.15057 2.6497549 0.0626167 6.0008594 -4.15057 -1.103369 -1.165223 -0.76789 -0.727128 -0.766956 42 9 8.3439856 2.121541 0.0689205 3.5384499 0.6560144 0.2271049 0.2242225 0.1472583 0.1491513 0.1472903 43 16 10.225097 2.3248452 0.0556046 4.3361762 5.7749034 1.805972 1.6667296 1.0927375 1.1840273 1.0940118 44 0 14.621 2.6824589 0.0823936 6.2003554 -14.621 -3.823742 -5.407587 -3.598003 -2.544172 -3.559939 Plot of Asthma_adult*SDI. Symbol used is '*'. Asthma_adult ‚ ‚ 80 ˆ ‚ ‚ ‚ 60 ˆ ‚ ‚ ‚ 40 ˆ ‚ * ‚ * ‚ * * 20 ˆ * ‚ ** * ‚ *** ** ‚ *** * 0 ˆ ** ‚ Šˆƒƒƒƒƒƒƒˆƒƒƒƒƒƒƒˆƒƒƒƒƒƒƒˆ 0 10 20 30 Poisson Regression Analysis. Extreme Ca ses (Outliers) Analysis for Adult Asthma Ouliers analyses. Adult Asthma. Outlier 1 13 21:21 Friday, October 8, 2004 The GENMOD Procedure Model Information Data Set WORK.ASTHMA Distribution Poisson Link Function Log Dependent Variable Asthma_adult Offset Variable ln1 Observations Used 43 Criteria For Assessing Goodness Of Fit Criterion DF Value Value/DF Deviance 41 108.6927 2.6510 Scaled Deviance 41 48.8987 1.1927 Pearson Chi-Square 41 91.1354 2.2228 Scaled Pearson X2 41 41.0000 1.0000 Log Likelihood 935.9454 Algorithm converged.

PAGE 259

244 Appendix F. Poisson Regression Analysis Outc ome for Adult Hospital Admissions (cont.) Analysis Of Parameter Estimates Standard Wald 95% Confidence ChiParameter DF Estimate Error Limits Square Pr > ChiSq Intercept 1 -7.4821 0.1045 -7.6869 -7.2773 5126.65 <.0001 SDI 1 0.0651 0.0074 0.0507 0.0796 77.94 <.0001 Scale 0 1.4909 0.0000 1.4909 1.4909 NOTE: The scale parameter was estimated by the square root of Pearson's Chi-Square/DOF. Observation Statistics Observation Resraw Reschi Resdev StResdev StReschi Reslik 1 2.4470915 0.6647127 0.6460794 0.440776 0.4534882 0.4412069 2 0.9321654 0.1825747 0.1815024 0.1263266 0.1270729 0.12638 3 -3.325601 -0.947253 -0.995506 -0.672596 -0.639995 -0.672136 4 1.5252714 0.5239432 0.5093 0.3435007 0.3533769 0.343611 5 -0.676272 -0.283851 -0.289787 -0.195402 -0.1914 -0.19536 6 -8.759885 -1.693387 -1.80128 -1.252846 -1.177803 -1.247737 7 -0.124115 -0.046501 -0.046637 -0.031694 -0.031601 -0.031691 8 2.1771231 0.6079813 0.5918909 0.4009338 0.4118331 0.4011494 9 4.8162709 1.0002746 0.9683344 0.6584396 0.680158 0.6590352 10 -4.179905 -0.887537 -0.917863 -0.623736 -0.603128 -0.623213 11 7.1427034 1.4623523 1.3972162 0.9593143 1.0040361 0.9614017 12 -1.13162 -0.301026 -0.305184 -0.206349 -0.203537 -0.206304 13 -3.747463 -1.719913 -2.092775 -1.412893 -1.161164 -1.409912 14 -6.849702 -1.910844 -2.135573 -1.454952 -1.301845 -1.450482 15 9.1837515 2.4707287 2.2530748 1.531288 1.6792151 1.5353232 16 1.0225237 0.2006203 0.1993253 0.1391993 0.1401037 0.1392696 17 -7.732642 -2.014594 -2.246577 -1.556714 -1.395967 -1.547074 18 8.0362245 1.8960651 1.7759486 1.2346155 1.3181188 1.240568 19 8.7223947 1.6700621 1.591169 1.0905075 1.1445768 1.0928436 20 1.8776654 0.2652177 0.2635871 0.1851356 0.1862809 0.1852368 21 -17.42422 -2.557297 -2.749175 -2.273137 -2.114483 -2.220159 22 2.3193534 0.7096891 0.6860822 0.4655694 0.4815889 0.4659446 23 4.4415156 0.766708 0.7506652 0.5205319 0.5316564 0.5212553 24 3.0221584 0.8389112 0.8091558 0.5482456 0.5684064 0.5486567 25 8.05642 1.2017338 1.1682734 0.8191715 0.8426333 0.8211911 26 10.911169 2.270755 2.1200108 1.4503402 1.5534673 1.4544733 27 5.3311459 0.7020207 0.6916009 0.4877971 0.4951463 0.4885049 28 -5.387248 -0.869508 -0.891136 -0.613964 -0.599063 -0.613195 29 14.267018 2.0650184 1.9731763 1.3600955 1.4234015 1.3635329 30 9.8039234 1.6295555 1.563232 1.0727171 1.1182294 1.0747892 31 2.2695496 0.6928364 0.6703434 0.452452 0.4676338 0.4526446 32 -15.44771 -2.491316 -2.694511 -1.848524 -1.709125 -1.842597 33 -3.870687 -1.078915 -1.141172 -0.777679 -0.735252 -0.776387 34 -8.43381 -1.458582 -1.527425 -1.047508 -1.000295 -1.0455 35 -0.331774 -0.287493 -0.300872 -0.202142 -0.193153 -0.202112 36 -0.754871 -0.148745 -0.149481 -0.10429 -0.103777 -0.104252 37 -5.065638 -1.458341 -1.583969 -1.079342 -0.993738 -1.076781 38 2.872428 1.2685082 1.1713636 0.7941835 0.8600474 0.7956449 39 -6.227895 -1.712363 -1.883067 -1.289897 -1.172965 -1.285287 40 -3.971506 -1.062512 -1.119962 -0.760292 -0.721291 -0.759387 41 0.7697205 0.2683031 0.2642761 0.1787816 0.1815059 0.1788281 42 5.8905 1.8526242 1.7060541 1.152147 1.25113 1.153547 43 -14.39552 -3.794143 -5.365729 -3.677684 -2.600515 -3.638535 Ouliers analyses. Adult Asthma. Outlier 2 Distribution Poisson Link Function Log Dependent Variable Asthma_adult Offset Variable ln1 Observations Used 43 Criteria For Assessing Goodness Of Fit

PAGE 260

245 Appendix F. Poisson Regression Analysis Outc ome for Adult Hospital Admissions (cont.) Criterion DF Value Value/DF Deviance 41 104.7554 2.5550 Scaled Deviance 41 47.7589 1.1649 Pearson Chi-Square 41 89.9303 2.1934 Scaled Pearson X2 41 41.0000 1.0000 Log Likelihood 931.0165 Analysis Of Parameter Estimates Standard Wald 95% Confidence ChiParameter DF Estimate Error Limits Square Pr > ChiSq Intercept 1 -7.5617 0.1143 -7.7857 -7.3377 4377.71 <.0001 SDI 1 0.0744 0.0087 0.0574 0.0913 73.77 <.0001 Scale 0 1.4810 0.0000 1.4810 1.4810 NOTE: The scale parameter was estimated by the square root of Pearson's Chi-Square/DOF. Observation Statistics Observation Resraw Reschi Resdev StResdev StReschi Reslik 1 2.8928565 0.799048 0.7720685 0.5309289 0.5494819 0.5316064 2 1.9519549 0.3900164 0.3851086 0.2707061 0.274156 0.2709744 3 -3.988895 -1.106793 -1.172232 -0.798291 -0.753728 -0.797557 4 0.9780504 0.32562 0.3199875 0.2175605 0.2213901 0.2176137 5 -0.586719 -0.248229 -0.252775 -0.171622 -0.168535 -0.171588 6 -7.784337 -1.533005 -1.621862 -1.139033 -1.076628 -1.134431 7 0.2798225 0.1079425 0.1072061 0.0734465 0.073951 0.073461 8 2.2152438 0.6195485 0.6028368 0.4111458 0.4225434 0.4113753 9 3.5911981 0.7268857 0.7100729 0.4871667 0.4987016 0.4875335 10 -4.630308 -0.97334 -1.009731 -0.690835 -0.665937 -0.690198 11 7.5624091 1.5620815 1.4876946 1.029366 1.0808358 1.0318804 12 -1.809839 -0.470289 -0.480391 -0.327331 -0.320448 -0.327208 13 -3.56174 -1.667619 -2.021898 -1.374921 -1.134007 -1.371823 14 -6.450033 -1.828002 -2.034822 -1.397145 -1.255139 -1.392696 15 9.4142748 2.5541445 2.3214826 1.5892356 1.7485108 1.5937753 16 8.322389 3.0035506 2.6176691 1.7774711 2.0394956 1.7806233 17 2.1766652 0.4368794 0.4307172 0.3039802 0.3083292 0.3043509 18 -9.964309 -2.419241 -2.745141 -1.956134 -1.723904 -1.933694 19 5.467324 1.2065686 1.1582135 0.8292431 0.8638639 0.8331434 20 6.583285 1.2137956 1.1722858 0.8149593 0.8438164 0.8166212 21 -2.503927 -0.339162 -0.34181 -0.246428 -0.244519 -0.246194 22 2.584463 0.80081 0.7707093 0.5268579 0.5474348 0.5273695 23 1.3051972 0.2154636 0.2142048 0.1517737 0.1526656 0.1518559 24 3.0726802 0.8546004 0.823715 0.5619449 0.5830152 0.5623828 25 3.9014203 0.556786 0.5496461 0.3958046 0.400946 0.3964292 26 11.128099 2.3268578 2.1685242 1.494302 1.6034077 1.4988041 27 0.5210076 0.0659139 0.0658226 0.0476845 0.0477506 0.0476932 28 -8.017141 -1.251805 -1.296317 -0.906016 -0.874906 -0.903974 29 12.114524 1.7152179 1.6520013 1.149941 1.1939454 1.152588 30 9.3067974 1.5364112 1.4774957 1.020739 1.0614411 1.0225965 31 2.0516483 0.6200526 0.6020589 0.4091014 0.4213283 0.4092577 32 -16.30096 -2.600229 -2.820851 -1.9487 -1.79629 -1.942146 33 -3.458858 -0.979927 -1.031497 -0.708463 -0.673044 -0.707304 34 -10.59266 -1.775517 -1.876737 -1.303046 -1.232768 -1.299329 35 -0.286782 -0.252813 -0.263202 -0.178036 -0.171009 -0.178011 36 0.3666493 0.0738736 0.0736915 0.0519507 0.0520791 0.0519613 37 -4.636517 -1.35919 -1.468912 -1.008901 -0.933541 -1.006463 38 3.2076109 1.4652295 1.3354317 0.9125966 1.0012967 0.9148024 39 -5.605238 -1.578769 -1.725005 -1.191968 -1.090919 -1.18759 40 -3.827466 -1.029295 -1.08328 -0.740576 -0.70367 -0.739693 41 0.9517481 0.3354836 0.3291774 0.2242788 0.2285754 0.2243564 42 5.8270672 1.8269512 1.6844368 1.1451888 1.2420793 1.146566 43 -13.76866 -3.710614 -5.247601 -3.628037 -2.56541 -3.585886 Ouliers analyses. Adult Asthma. Outlier 3 17 21:21 Friday, October 8, 2004 The GENMOD Procedure

PAGE 261

246 Appendix F. Poisson Regression Analysis Outc ome for Adult Hospital Admissions (cont.) Model Information Data Set WORK.ASTHMA Distribution Poisson Link Function Log Dependent Variable Asthma_adult Offset Variable ln1 Observations Used 43 Criteria For Assessing Goodness Of Fit Criterion DF Value Value/DF Deviance 41 107.5897 2.6241 Scaled Deviance 41 48.7014 1.1878 Pearson Chi-Square 41 90.5760 2.2092 Scaled Pearson X2 41 41.0000 1.0000 Log Likelihood 932.5803 Analysis Of Parameter Estimates Standard Wald 95% Confidence ChiParameter DF Estimate Error Limits Square Pr > ChiSq Intercept 1 -7.4390 0.1031 -7.6412 -7.2369 5201.20 <.0001 SDI 1 0.0638 0.0073 0.0496 0.0780 77.03 <.0001 Scale 0 1.4863 0.0000 1.4863 1.4863 NOTE: The scale parameter was estimated by the square root of Pearson's Chi-Square/DOF. Observation Statistics Observation Resraw Reschi Resdev StResdev StReschi Reslik 1 1.9441362 0.5185587 0.5072451 0.3472642 0.3550096 0.347532 2 -0.060253 -0.011583 -0.011587 -0.008097 -0.008094 -0.008096 3 -3.626823 -1.020656 -1.076507 -0.729739 -0.691879 -0.729189 4 1.330775 0.4519751 0.441094 0.2984642 0.3058268 0.2985487 5 -0.872154 -0.359911 -0.369421 -0.249901 -0.243467 -0.249832 6 -9.767636 -1.853617 -1.982244 -1.384152 -1.294335 -1.377913 7 -0.414912 -0.152371 -0.153826 -0.104892 -0.1039 -0.104866 8 1.7591458 0.4834414 0.473289 0.3216726 0.3285728 0.3218126 9 4.2465508 0.8713106 0.847111 0.5780489 0.5945622 0.5785156 10 -4.827323 -1.010366 -1.04954 -0.715731 -0.689016 -0.715032 11 6.3128868 1.2705532 1.2214695 0.8417343 0.8755586 0.8433474 12 -1.488465 -0.391046 -0.398045 -0.27004 -0.265291 -0.269962 13 -3.928202 -1.769496 -2.160198 -1.463127 -1.1985 -1.459927 14 -7.323038 -2.006271 -2.252388 -1.539836 -1.371579 -1.53482 15 8.7050775 2.3024051 2.1135223 1.4413524 1.5701643 1.4449418 16 8.081987 2.8721693 2.5192014 1.7046616 1.9435035 1.707573 17 0.0121485 0.0023385 0.0023383 0.0016396 0.0016397 0.0016396 18 -7.90237 -2.047056 -2.286088 -1.588977 -1.422834 -1.57902 19 7.8100465 1.8312086 1.7192197 1.1989558 1.2770551 1.2045284 20 8.1479613 1.5439046 1.476549 1.0156034 1.0619321 1.0176487 21 0.8829999 0.1235031 0.1231501 0.086848 0.0870969 0.0868704 22 -17.67438 -2.58705 -2.783268 -2.302423 -2.140104 -2.248786 23 1.9361597 0.5820874 0.56624 0.3855386 0.3963287 0.3857965 24 3.8026896 0.6502718 0.6387465 0.4445617 0.4525832 0.4450918 25 2.5973583 0.7094743 0.6882325 0.4678828 0.4823237 0.4681846 26 7.1948105 1.0630704 1.0369177 0.7299242 0.7483341 0.7315351 27 10.136489 2.0750108 1.9492818 1.3383628 1.4246874 1.3419025 28 4.1556153 0.5417299 0.5355343 0.3793213 0.3837096 0.3797524 29 -6.247484 -0.997239 -1.025623 -0.709339 -0.689708 -0.708299 30 13.042275 1.8639864 1.7892525 1.2383086 1.2900306 1.2412066 31 8.7128952 1.4268659 1.3760972 0.9479163 0.9828881 0.9495548 32 1.9563342 0.5886897 0.5724789 0.3876704 0.398648 0.3878139 33 -4.346561 -1.189765 -1.265111 -0.865121 -0.813597 -0.86352 34 -9.201121 -1.573333 -1.653213 -1.13802 -1.083033 -1.135617 35 -0.381379 -0.324489 -0.341458 -0.230125 -0.218689 -0.230087 36 -1.753051 -0.338928 -0.342735 -0.24008 -0.237413 -0.239874 37 -5.518364 -1.559684 -1.702562 -1.164158 -1.066463 -1.161175 38 2.6560678 1.14897 1.0693648 0.7274446 0.7815967 0.7286687 39 -6.747835 -1.819899 -2.011485 -1.382779 -1.251074 -1.377482 40 -4.44219 -1.168909 -1.238146 -0.843381 -0.796219 -0.842262

PAGE 262

247 Appendix F. Poisson Regression Analysis Outc ome for Adult Hospital Admissions (cont.) 41 0.4778199 0.1636775 0.1621826 0.1100771 0.1110916 0.1100948 42 5.5747587 1.7265653 1.5993638 1.0836492 1.1698346 1.0849008 43 -14.95345 -3.866969 -5.46872 -3.761705 -2.659927 -3.720755 Ouliers analyses. Adult Asthma. Outlier 4 19 21:21 Friday, October 8, 2004 The GENMOD Procedure Model Information Data Set WORK.ASTHMA Distribution Poisson Link Function Log Dependent Variable Asthma_adult Offset Variable ln1 Observations Used 43 Criteria For Assessing Goodness Of Fit Criterion DF Value Value/DF Deviance 41 85.5800 2.0873 Scaled Deviance 41 42.6665 1.0406 Pearson Chi-Square 41 82.2373 2.0058 Scaled Pearson X2 41 41.0000 1.0000 Log Likelihood 1057.1157 Analysis Of Parameter Estimates Standard Wald 95% Confidence ChiParameter DF Estimate Error Limits Square Pr > ChiSq Intercept 1 -7.4084 0.0991 -7.6026 -7.2141 5585.11 <.0001 SDI 1 0.0612 0.0070 0.0474 0.0750 75.39 <.0001 Scale 0 1.4163 0.0000 1.4163 1.4163 NOTE: The scale parameter was estimated by the square root of Pearson's Chi-Square/DOF. Observation Statistics Observation Resraw Reschi Resdev StResdev StReschi Reslik 1 1.6931747 0.4476413 0.4392206 0.3156871 0.3217394 0.3159004 2 -0.593486 -0.112981 -0.11339 -0.083236 -0.082936 -0.083213 3 -3.545779 -1.001066 -1.054838 -0.750291 -0.712044 -0.749748 4 1.4109719 0.4814446 0.469091 0.3330733 0.3418449 0.333172 5 -0.947599 -0.388556 -0.399626 -0.283719 -0.27586 -0.283635 6 -10.29277 -1.935061 -2.074873 -1.521892 -1.419341 -1.514609 7 -0.600373 -0.217773 -0.220739 -0.158046 -0.155923 -0.157988 8 1.6376679 0.4480069 0.4392936 0.3133263 0.319541 0.3134517 9 4.3928823 0.9041242 0.8780576 0.6285905 0.6472512 0.6291053 10 -4.887768 -1.021666 -1.061709 -0.75963 -0.73098 -0.758896 11 5.9827282 1.1961325 1.1526621 0.8337905 0.8652353 0.835301 12 -1.417907 -0.373419 -0.379805 -0.270355 -0.26581 -0.270283 13 -4.025314 -1.795635 -2.195826 -1.561108 -1.276595 -1.557582 14 -7.553878 -2.051815 -2.308419 -1.656742 -1.472579 -1.651151 15 8.5176602 2.2382106 2.0597658 1.4743455 1.6020733 1.4779276 16 8.0176343 2.8377911 2.493202 1.7704633 2.0151617 1.7734222 17 -0.562544 -0.107151 -0.107519 -0.079218 -0.078947 -0.079196 18 -7.442375 -1.958359 -2.178339 -1.588629 -1.428201 -1.579063 19 8.3348814 1.9830842 1.8516137 1.3548543 1.4510532 1.3616931 20 8.505693 1.6221399 1.5477377 1.1168605 1.1705497 1.1191979 21 1.6572724 0.2335746 0.2323103 0.1718337 0.1727688 0.1719169 22 -15.69687 -2.347872 -2.510477 -2.172193 -2.031499 -2.126227 23 1.7648926 0.5265379 0.5135835 0.3670508 0.3763092 0.367275 24 4.3730819 0.7541263 0.7386078 0.5393021 0.5506331 0.5400442 25 2.4708854 0.6717658 0.6527329 0.4656877 0.4792666 0.4659703 26 7.9471249 1.183992 1.1515173 0.8502727 0.8742519 0.8523516 27 9.8737632 2.0101937 1.8922461 1.3635225 1.4485137 1.3670092 28 4.986742 0.6547165 0.6456575 0.4796037 0.4863328 0.4802568 29 -5.844069 -0.937676 -0.962798 -0.698462 -0.680238 -0.697514 30 13.24275 1.896525 1.8191425 1.3202841 1.3764462 1.32335 31 8.5461803 1.3964455 1.3478307 0.9739757 1.009106 0.9755927

PAGE 263

248 Appendix F. Poisson Regression Analysis Outc ome for Adult Hospital Admissions (cont.) 32 1.9270911 0.5791238 0.5634381 0.4003697 0.4115157 0.4005123 33 -16.64368 -2.643397 -2.871105 -2.074002 -1.909513 -2.066948 34 -4.581333 -1.243142 -1.325208 -0.951368 -0.892453 -0.949503 35 -8.884735 -1.526309 -1.601571 -1.156467 -1.102122 -1.15414 36 -0.406408 -0.342694 -0.361577 -0.255751 -0.242394 -0.255705 37 -2.315653 -0.443065 -0.449557 -0.330884 -0.326106 -0.330506 38 -5.751798 -1.610712 -1.762648 -1.265347 -1.156277 -1.261945 39 2.5081711 1.0702831 1.0012698 0.7151605 0.7644535 0.7163191 40 -7.051524 -1.881141 -2.08508 -1.505313 -1.35808 -1.499216 41 -4.604981 -1.204973 -1.278441 -0.913939 -0.861418 -0.91269 42 0.3526327 0.119917 0.1191155 0.0848555 0.0854264 0.0848656 43 5.5063818 1.699823 1.576554 1.1209505 1.2085963 1.1222077 Poisson Regression Analysis Outcome for Childhood Hospital Admissions Children Asthma. Regression Model 69 19:47 Saturday, October 9, 2004 The GENMOD Procedure Model Information Data Set WORK.ASTHMA Distribution Poisson Link Function Log Dependent Variable Asthma_14 Offset Variable ln2 Observations Used 44 Criteria For Assessing Goodness Of Fit Criterion DF Value Value/DF Deviance 39 59.0396 1.5138 Scaled Deviance 39 38.5878 0.9894 Pearson Chi-Square 39 59.6703 1.5300 Scaled Pearson X2 39 39.0000 1.0000 Log Likelihood 1113.1594 Analysis Of Parameter Estimates Standard Wald 95% Confidence ChiParameter DF Estimate Error Limits Square Pr > ChiSq Intercept 1 -15.7508 3.9648 -23.5217 -7.9799 15.78 <.0001 SDI 1 0.0589 0.0075 0.0441 0.0737 61.13 <.0001 PM10_99 1 0.0612 0.0439 -0.0249 0.1473 1.94 0.1637 O3_99 1 0.0736 0.0431 -0.0108 0.1580 2.92 0.0875 SO2_99 1 0.2711 0.1223 0.0314 0.5108 4.91 0.0267 Scale 0 1.2369 0.0000 1.2369 1.2369 NOTE: The scale parameter was estimated by the square root of Pearson's Chi-Square/DOF. Children Asthma. Regression Model 70 The GENMOD Procedure Model Information Data Set WORK.ASTHMA Distribution Poisson Link Function Log Dependent Variable Asthma_14 Offset Variable ln2 Observations Used 44 Criteria For Assessing Goodness Of Fit

PAGE 264

249 Appendix F. Poisson Regression Analysis Outc ome for Adult Hospital Admissions (cont.) Criterion DF Value Value/DF Deviance 40 62.0381 1.5510 Scaled Deviance 40 39.7461 0.9937 Pearson Chi-Square 40 62.4343 1.5609 Scaled Pearson X2 40 40.0000 1.0000 Log Likelihood 1090.1967 Analysis Of Parameter Estimates Standard Wald 95% Confidence ChiParameter DF Estimate Error Limits Square Pr > ChiSq Intercept 1 -13.0342 3.4686 -19.8326 -6.2358 14.12 0.0002 SDI 1 0.0613 0.0073 0.0470 0.0756 70.24 <.0001 O3_99 1 0.0544 0.0410 -0.0259 0.1347 1.76 0.1840 SO2_99 1 0.3663 0.0995 0.1712 0.5614 13.54 0.0002 Scale 0 1.2493 0.0000 1.2493 1.2493 NOTE: The scale parameter was estimated by the square root of Pearson's Chi-Square/DOF. Children Asthma. Regression Model 71 19:47 Saturday, October 9, 2004 The GENMOD Procedure Model Information Data Set WORK.ASTHMA Distribution Poisson Link Function Log Dependent Variable Asthma_14 Offset Variable ln2 Observations Used 44 Criteria For Assessing Goodness Of Fit Criterion DF Value Value/DF Deviance 41 64.8020 1.5805 Scaled Deviance 41 40.5218 0.9883 Pearson Chi-Square 41 65.5667 1.5992 Scaled Pearson X2 41 41.0000 1.0000 Log Likelihood 1063.2029 Analysis Of Parameter Estimates Standard Wald 95% Confidence ChiParameter DF Estimate Error Limits Square Pr > ChiSq Intercept 1 -8.4790 0.4527 -9.3664 -7.5917 350.74 <.0001 SDI 1 0.0593 0.0072 0.0452 0.0733 68.61 <.0001 SO2_99 1 0.4161 0.0912 0.2374 0.5949 20.82 <.0001 Scale 0 1.2646 0.0000 1.2646 1.2646 NOTE: The scale parameter was estimated by the square root of Pearson's Chi-Square/DOF. Children Asthma. Regression Model with Residuals 72 19:47 Saturday, October 9, 2004 The GENMOD Procedure Model Information Data Set WORK.ASTHMA Distribution Poisson Link Function Log Dependent Variable Asthma_14 Offset Variable ln2 Observations Used 44 Criteria For Assessing Goodness Of Fit

PAGE 265

250 Appendix F. Poisson Regression Analysis Outc ome for Adult Hospital Admissions (cont.) Criterion DF Value Value/DF Deviance 41 64.8020 1.5805 Scaled Deviance 41 40.5218 0.9883 Pearson Chi-Square 41 65.5667 1.5992 Scaled Pearson X2 41 41.0000 1.0000 Log Likelihood 1063.2029 Analysis Of Parameter Estimates Standard Wald 95% Confidence ChiParameter DF Estimate Error Limits Square Pr > ChiSq Intercept 1 -8.4790 0.4527 -9.3664 -7.5917 350.74 <.0001 SDI 1 0.0593 0.0072 0.0452 0.0733 68.61 <.0001 SO2_99 1 0.4161 0.0912 0.2374 0.5949 20.82 <.0001 Scale 0 1.2646 0.0000 1.2646 1.2646 NOTE: The scale parameter was estimated by the square root of Pearson's Chi-Square/DOF. Observation Statistics Observation Asthma_14 Pred Xbeta Std HessWgt Resraw Reschi Resdev StResdev StReschi Reslik 1 13 12.098996 2.4931225 0.0717861 7.5657145 0.9010038 0.2590311 0.255912 0.2064318 0.2089478 0.2065304 2 20 22.624961 3.1190537 0.0747635 14.147784 -2.624961 -0.55186 -0.563083 -0.463993 -0.454745 -0.463269 3 7 8.4889528 2.1387657 0.0978199 5.308291 -1.488953 -0.511039 -0.527189 -0.427894 -0.414786 -0.427238 4 4 7.1681755 1.9696512 0.0798468 4.4823858 -3.168175 -1.183327 -1.29209 -1.036666 -0.949404 -1.034275 5 2 5.5175064 1.707926 0.0614311 3.4501935 -3.517506 -1.497488 -1.725079 -1.373109 -1.191954 -1.370905 6 16 19.167513 2.9532168 0.0742098 11.985782 -3.167513 -0.723495 -0.744934 -0.609532 -0.59199 -0.60839 7 9 3.8877996 1.3578433 0.1201157 2.4311092 5.1122004 2.5927219 2.2100818 1.7791476 2.0871784 1.7908488 8 5 7.0141586 1.9479308 0.0998986 4.3860764 -2.014159 -0.760512 -0.802115 -0.648644 -0.615 -0.647208 9 17 10.578086 2.3587845 0.1787384 6.6146623 6.4219144 1.9745175 1.8129405 1.6142983 1.7581715 1.6457502 10 9 14.002941 2.6392674 0.1126371 8.7562844 -5.002941 -1.336952 -1.431472 -1.200618 -1.121341 -1.192071 11 21 18.521268 2.9189197 0.0692782 11.581674 2.4787317 0.5759624 0.5637829 0.4587552 0.4686657 0.4593117 12 6 8.3040758 2.1167465 0.0773712 5.1926842 -2.304076 -0.79956 -0.84161 -0.676112 -0.642331 -0.675087 13 1 1.8068332 0.5915757 0.0928447 1.1298444 -0.806833 -0.60024 -0.656136 -0.521398 -0.47698 -0.520984 14 0 0.0916915 -2.389326 0.1012944 0.0573363 -0.091691 -0.302806 -0.428232 -0.338733 -0.23952 -0.338683 15 5 11.834477 2.471017 0.0622045 7.4003059 -6.834477 -1.986693 -2.247924 -1.803602 -1.594005 -1.79794 16 8 5.9840225 1.789093 0.0589629 3.7419142 2.0159775 0.8241174 0.7833401 0.623511 0.6559683 0.623944 17 24 22.323508 3.1056403 0.0776261 13.959281 1.6764918 0.3548299 0.3505221 0.2896313 0.2931908 0.2899324 18 9 14.652624 2.6846195 0.1265477 9.1625429 -5.652624 -1.4767 -1.59127 -1.362233 -1.264154 -1.348289 19 25 21.075932 3.0481317 0.0974474 13.17915 3.9240677 0.8547578 0.8301017 0.701802 0.7226473 0.7044446 20 29 30.446916 3.4159847 0.0634531 19.03899 -1.446916 -0.262224 -0.264343 -0.217538 -0.215794 -0.217405 21 52 52.03665 3.9519483 0.0537543 32.53943 -0.03665 -0.005081 -0.005081 -0.004221 -0.004221 -0.004221 22 50 57.312511 4.0485189 0.1079549 35.838518 -7.312511 -0.965921 -0.987638 -1.023443 -1.00094 -1.014105 23 10 11.042084 2.4017138 0.1883012 6.9048087 -1.042084 -0.313601 -0.318738 -0.290042 -0.285367 -0.288905 24 38 31.105321 3.4373789 0.0587801 19.450702 6.8946791 1.2362227 1.1943174 0.9778601 1.0121705 0.9802035

PAGE 266

251 Appendix F. Poisson Regression Analysis Outc ome for Adult Hospital Admissions (cont.) 25 11 10.63164 2.3638345 0.0915952 6.6481509 0.3683599 0.1129724 0.1123292 0.0914125 0.0919359 0.0914418 26 65 47.213023 3.8546698 0.0549229 29.523131 17.786977 2.5886374 2.447309 2.0276528 2.1447467 2.0383537 27 23 15.306349 2.7282677 0.0663696 9.5713283 7.6936511 1.9665124 1.8289662 1.4777798 1.5889152 1.4826336 28 41 52.198783 3.9550592 0.0528157 32.640814 -11.19878 -1.550032 -1.611094 -1.336292 -1.285646 -1.331761 29 21 23.054764 3.1378724 0.052293 14.416548 -2.054764 -0.427939 -0.434544 -0.350605 -0.345275 -0.350396 30 46 38.643999 3.6543915 0.0495979 24.16477 7.3560009 1.183316 1.1484723 0.9364369 0.9648477 0.9381498 31 35 24.683589 3.2061386 0.0522785 15.435081 10.316411 2.0764645 1.9523934 1.5775268 1.6777757 1.5818841 32 13 10.247584 2.327042 0.0564662 6.4079938 2.7524162 0.8598117 0.8250819 0.6592194 0.6869676 0.659798 33 32 33.527998 3.5123809 0.0498155 20.965645 -1.527998 -0.263888 -0.265931 -0.215984 -0.214324 -0.215898 34 6 9.3349444 2.2337648 0.07009 5.8373044 -3.334944 -1.091523 -1.168685 -0.937704 -0.875792 -0.935986 35 30 32.569148 3.4833655 0.0470868 20.366059 -2.569148 -0.45018 -0.456302 -0.369263 -0.364309 -0.369041 36 1 2.62484 0.9650199 0.081686 1.6413585 -1.62484 -1.002904 -1.148756 -0.913418 -0.797445 -0.912227 37 16 19.257553 2.9579034 0.0762238 12.042086 -3.257553 -0.74232 -0.764878 -0.627181 -0.608684 -0.625905 38 7 9.7918693 2.2815524 0.076498 6.1230275 -2.791869 -0.8922 -0.94061 -0.757502 -0.718516 -0.756139 39 6 3.3980949 1.223215 0.1152717 2.1248883 2.6019051 1.4114766 1.2722908 1.020601 1.1322525 1.0239204 40 11 11.988379 2.4839377 0.1162062 7.4965433 -0.988379 -0.285459 -0.289522 -0.241495 -0.238106 -0.241154 41 15 11.755083 2.4642857 0.0608174 7.3506593 3.2449173 0.946435 0.9072485 0.7273814 0.7587989 0.7282535 42 10 5.1842137 1.6456182 0.0837555 3.2417797 4.8157863 2.1150749 1.872903 1.4981688 1.6918865 1.5028516 43 6 8.9161836 2.187868 0.053283 5.5754459 -2.916184 -0.976619 -1.03878 -0.828016 -0.778467 -0.827255 44 6 13.584958 2.6089631 0.076634 8.4949123 -7.584958 -2.057899 -2.315917 -1.878824 -1.669502 -1.868937 Children Asthma. Regression Model with Residuals. Only SDI The GENMOD Procedure Model Information Data Set WORK.ASTHMA Distribution Poisson Link Function Log Dependent Variable Asthma_14 Offset Variable ln2 Observations Used 44 Criteria For Assessing Goodness Of Fit Criterion DF Value Value/DF Deviance 42 99.3550 2.3656 Scaled Deviance 42 43.6816 1.0400 Pearson Chi-Square 42 95.5302 2.2745 Scaled Pearson X2 42 42.0000 1.0000 Log Likelihood 739.9266 Analysis Of Parameter Estimates Standard Wald 95% Confidence ChiParameter DF Estimate Error Limits Square Pr > ChiSq Intercept 1 -6.4894 0.1171 -6.7188 -6.2600 3073.09 <.0001 SDI 1 0.0743 0.0078 0.0590 0.0895 91.37 <.0001 Scale 0 1.5082 0.0000 1.5082 1.5082

PAGE 267

252 Appendix F. Poisson Regression Analysis Outc ome for Adult Hospital Admissions (cont.) Observation Statistics Observation Asthma_14 Pred Xbeta Std HessWgt Resraw Reschi Resdev StResdev StReschi Reslik 1 13 11.000832 2.3979709 0.084505 4.836532 1.9991685 0.6027492 0.5857507 0.3952752 0.4067461 0.3956769 2 20 20.784476 3.0342063 0.0887587 9.1379256 -0.784476 -0.172072 -0.173172 -0.119194 -0.118437 -0.11914 3 7 12.32847 2.5119112 0.054449 5.42023 -5.32847 -1.517567 -1.653156 -1.10506 -1.014425 -1.103662 4 4 9.4194857 2.2427805 0.0562268 4.1412909 -5.419486 -1.765812 -1.996768 -1.332735 -1.178584 -1.330832 5 2 5.3143626 1.6704131 0.0737489 2.3364675 -3.314363 -1.43772 -1.64914 -1.100497 -0.959414 -1.098818 6 16 20.207174 3.0060377 0.0869195 8.8841143 -4.207174 -0.935918 -0.971587 -0.666996 -0.642509 -0.66538 7 9 5.5932235 1.7215558 0.1015376 2.459069 3.4067765 1.4404977 1.3223031 0.8880994 0.9674825 0.8901994 8 5 9.9920984 2.3017946 0.0668103 4.3930409 -4.992098 -1.579264 -1.749466 -1.171548 -1.057571 -1.16942 9 17 22.777402 3.1257689 0.0543512 10.014119 -5.777402 -1.210544 -1.268036 -0.853506 -0.814808 -0.852386 10 9 21.771024 3.0805799 0.0575391 9.5716632 -12.77102 -2.737072 -3.105101 -2.092294 -1.844307 -2.084888 11 21 20.746554 3.0323802 0.0748168 9.1212535 0.2534458 0.0556432 0.0555305 0.0377977 0.0378744 0.0378016 12 6 10.815866 2.3810141 0.0540273 4.7552115 -4.815866 -1.464346 -1.600211 -1.068481 -0.977762 -1.067274 13 1 2.277321 0.8229998 0.0887587 1.0012276 -1.277321 -0.846424 -0.953227 -0.634557 -0.563458 -0.634027 14 0 0.1252035 -2.077815 0.083315 0.0550459 -0.125204 -0.353841 -0.500407 -0.331864 -0.234664 -0.331833 15 5 11.914782 2.4777798 0.0742807 5.2383516 -6.914782 -2.003251 -2.268512 -1.526385 -1.347902 -1.52152 16 8 6.489207 1.8701403 0.0650265 2.8529895 1.510793 0.5930744 0.5720431 0.381609 0.3956389 0.3817813 17 24 22.930612 3.1324728 0.0925036 10.081478 1.0693881 0.2233199 0.221617 0.1537261 0.1549073 0.1538284 18 9 22.426769 3.1102553 0.0980248 9.8599626 -13.42677 -2.835229 -3.227845 -2.249474 -1.975862 -2.224995 19 25 15.775784 2.7584761 0.0928984 6.9358469 9.2242161 2.3223838 2.1381193 1.4621411 1.5881493 1.4699875 20 29 25.004783 3.2190671 0.0601931 10.99339 3.9952168 0.7989669 0.7789934 0.5271258 0.5406414 0.5276708 21 52 49.076422 3.8933787 0.0636006 21.576522 2.923578 0.4173287 0.4132848 0.2868364 0.2896431 0.2870825 22 50 52.925895 3.9688927 0.1293698 23.268949 -2.925895 -0.402184 -0.405978 -0.344503 -0.341284 -0.343253 23 10 4.8114673 1.5710021 0.0792493 2.1153687 5.1885327 2.3654065 2.0626669 1.3768531 1.5789352 1.3797319 24 38 27.809842 3.32539 0.066186 12.226638 10.190158 1.9323316 1.8293696 1.2468355 1.3170109 1.2506939 25 11 7.4539826 2.0087485 0.0672714 3.2771546 3.5460174 1.2988131 1.2121035 0.8097268 0.8676518 0.8106161 26 65 46.963869 3.8493786 0.0657392 20.647735 18.036131 2.6318519 2.4857333 1.7270502 1.8285712 1.7363504 27 23 12.932817 2.559768 0.0701547 5.685932 10.067183 2.7993772 2.51973 1.6946173 1.8826911 1.7001634 28 41 53.626327 3.9820401 0.0620119 23.576894 -12.62633 -1.724201 -1.79952 -1.251264 -1.198892 -1.246606 29 21 24.855813 3.2130916 0.0571772 10.927895 -3.855813 -0.773396 -0.794808 -0.536681 -0.522223 -0.536171 30 46 34.454024 3.5396258 0.0539665 15.147763 11.545976 1.9670286 1.8701559 1.268322 1.3340202 1.271292 31 35 25.896118 3.2540931 0.0596446 11.385267 9.1038822 1.7889957 1.6970814 1.1487753 1.2109932 1.1513606 32 13 8.7504335 2.1691032 0.0575391 3.8471411 4.2495665 1.4365801 1.3389795 0.8935354 0.9586668 0.8943948 33 32 30.692751 3.4240265 0.0570186 13.494114 1.3072487 0.235961 0.2343151 0.1588899 0.160006 0.158939 34 6 8.7881601 2.1734054 0.0839085 3.8637277 -2.78816 -0.940521 -0.998283 -0.671115 -0.632284 -0.670088 35 30 31.849785 3.4610306 0.0562268 14.002805 -1.849785

PAGE 268

253 Appendix F. Poisson Regression Analysis Outc ome for Adult Hospital Admissions (cont.) -0.327769 -0.331021 -0.224513 -0.222308 -0.224416 36 1 2.1209133 0.7518468 0.0851044 0.9324627 -1.120913 -0.769681 -0.859147 -0.571601 -0.512078 -0.57122 37 16 18.845962 2.9362987 0.0918736 8.2856554 -2.845962 -0.655571 -0.673206 -0.462856 -0.450732 -0.462018 38 7 10.777799 2.3774884 0.0863117 4.7384754 -3.777799 -1.150731 -1.230245 -0.830519 -0.776841 -0.828684 39 6 4.6004164 1.5261468 0.1081986 2.0225799 1.3995836 0.6525297 0.6230446 0.4180969 0.437883 0.4185762 40 11 7.8865073 2.0651534 0.0944067 3.4673147 3.1134927 1.1086782 1.0456274 0.7042836 0.7467515 0.7056342 41 15 10.688752 2.369192 0.0706537 4.6993257 4.311248 1.3186803 1.2422765 0.8335415 0.8848069 0.8347802 42 10 6.4490707 1.863936 0.0775584 2.8353435 3.5509293 1.3982778 1.292719 0.8645575 0.9351542 0.8658098 43 6 8.2947334 2.1156208 0.0625452 3.6467919 -2.294733 -0.796766 -0.838529 -0.560006 -0.532115 -0.559618 44 6 12.75271 2.5457438 0.0918736 5.6067479 -6.75271 -1.890936 -2.111305 -1.434276 -1.284573 -1.427546 Scatter Plot 19:47 Saturday, October 9, 2004 78 Plot of Asthma_14*SDI. Symbol used is '*'. Asthma_14 ‚ 75 ˆ ‚ ‚ ‚ ‚ ‚ 50 ˆ * ‚ ‚ ‚ ‚ ** ‚ ** 25 ˆ * ‚ * ‚ ** * ‚ ** * ‚ ** * * ‚ *** ** 0 ˆ ** Šˆƒƒƒƒƒƒƒˆƒƒƒƒƒƒƒˆƒƒƒƒƒƒƒˆ 0 10 20 30 SDI Poisson Regression Analysis. Extreme Ca ses (Outliers) Analys is for Childhood Asthma Ouliers analyses. Childhood Asthma 3 21:21 Friday, October 8, 2004 The GENMOD Procedure Model Information Data Set WORK.ASTHMA Distribution Poisson Link Function Log Dependent Variable Asthma_14 Offset Variable ln1 Observations Used 43 Criteria For Assessing Goodness Of Fit Criterion DF Value Value/DF Deviance 41 89.4698 2.1822 Scaled Deviance 41 42.5281 1.0373 Pearson Chi-Square 41 86.2551 2.1038

PAGE 269

254 Appendix F. Poisson Regression Analysis Outc ome for Adult Hospital Admissions (cont.) Scaled Pearson X2 41 41.0000 1.0000 Log Likelihood 797.2076 Analysis Of Parameter Estimates Standard Wald 95% Confidence ChiParameter DF Estimate Error Limits Square Pr > ChiSq Intercept 1 -6.4608 0.1124 -6.6812 -6.2404 3301.30 <.0001 SDI 1 0.0734 0.0074 0.0588 0.0880 97.49 <.0001 Scale 0 1.4504 0.0000 1.4504 1.4504 NOTE: The scale parameter was estimated by the square root of Pearson's Chi-Square/DOF. Observation Statistics Observation Resraw Reschi Resdev StResdev StReschi Reslik 1 1.7303893 0.5154534 0.5030411 0.3531395 0.3618531 0.3534523 2 -1.305497 -0.282833 -0.285798 -0.204755 -0.202631 -0.204599 3 -5.526108 -1.56139 -1.704566 -1.185017 -1.08548 -1.183443 4 -5.561173 -1.798501 -2.037404 -1.414178 -1.248354 -1.412086 5 -3.435055 -1.473437 -1.694467 -1.175951 -1.022557 -1.174076 6 -4.708222 -1.034631 -1.078099 -0.770312 -0.739254 -0.768211 7 3.256404 1.3587712 1.2536728 0.875808 0.9492291 0.8777964 8 -5.206364 -1.629667 -1.810247 -1.260858 -1.135082 -1.258443 9 -6.144597 -1.277229 -1.341105 -0.938986 -0.894263 -0.937659 10 -0.221478 -0.048078 -0.048162 -0.034112 -0.034053 -0.034109 11 -4.995096 -1.506414 -1.649822 -1.145664 -1.046079 -1.144304 12 -1.334408 -0.873374 -0.98656 -0.68295 -0.604597 -0.682351 13 -0.12824 -0.358106 -0.506438 -0.34923 -0.246943 -0.349196 14 -7.186452 -2.058619 -2.337528 -1.636093 -1.440878 -1.630628 15 1.373989 0.5337737 0.5167584 0.3585106 0.3703153 0.3586596 16 0.4820765 0.0994069 0.0990701 0.071545 0.0717881 0.0715665 17 -13.59342 -2.859813 -3.258684 -2.361271 -2.072246 -2.335434 18 9.0957282 2.2807652 2.1030806 1.4954482 1.6217953 1.5033171 19 3.6505425 0.7250585 0.708624 0.4987714 0.5103389 0.4992455 20 2.2867264 0.3243232 0.3218833 0.2324512 0.2342132 0.2326077 21 -3.102433 -0.425741 -0.429991 -0.37859 -0.374848 -0.377146 22 5.0749025 2.2867584 2.0037504 1.3910036 1.5874678 1.3938724 23 9.84424 1.8552335 1.7603749 1.2479956 1.3152443 1.2517355 24 3.3855037 1.2268818 1.149567 0.7986836 0.8523995 0.7995309 25 17.447751 2.5301946 2.3952165 1.7314727 1.8290467 1.7405168 26 9.781668 2.6904479 2.4321973 1.7015429 1.8822126 1.7070127 27 -13.34147 -1.80983 -1.892686 -1.3695 -1.309548 -1.364088 28 -4.220756 -0.840449 -0.865695 -0.608045 -0.590313 -0.607407 29 10.965733 1.8526405 1.7667744 1.2467104 1.3073012 1.2495237 30 8.5907056 1.6716696 1.5914875 1.1208944 1.1773671 1.1233093 31 4.0816911 1.3667808 1.2784898 0.8872942 0.9485698 0.8881264 32 0.7239534 0.1294508 0.1289561 0.0909901 0.0913391 0.0910059 33 -3.002081 -1.000578 -1.065742 -0.745256 -0.699688 -0.744021 34 -2.328866 -0.40959 -0.414661 -0.292587 -0.289008 -0.292425 35 -1.172925 -0.795697 -0.890899 -0.616367 -0.550502 -0.615935 36 -3.326944 -0.75677 -0.780206 -0.558327 -0.541555 -0.557139 37 -4.044062 -1.216896 -1.305514 -0.916852 -0.854617 -0.91467 38 1.2717175 0.5848425 0.56119 0.3916983 0.4082072 0.3921078 39 2.9093501 1.0228319 0.9692247 0.6790851 0.7166449 0.6803084 40 4.0743069 1.2326201 1.1659232 0.8137198 0.8602689 0.8148747 41 3.400378 1.3236332 1.2291047 0.8549267 0.9206778 0.8561229 42 -2.465102 -0.847264 -0.894367 -0.6212 -0.588483 -0.620731 43 -7.078181 -1.957258 -2.192289 -1.549576 -1.383448 -1.541918 Ouliers analyses. Childhood Asthma. Outlier 2 5 21:21 Friday, October 8, 2004 The GENMOD Procedure Model Information Data Set WORK.ASTHMA Distribution Poisson Link Function Log Dependent Variable Asthma_14 Offset Variable ln1

PAGE 270

255 Appendix F. Poisson Regression Analysis Outc ome for Adult Hospital Admissions (cont.) Observations Used 43 Criteria For Assessing Goodness Of Fit Criterion DF Value Value/DF Deviance 41 88.1088 2.1490 Scaled Deviance 41 42.0507 1.0256 Pearson Chi-Square 41 85.9072 2.0953 Scaled Pearson X2 41 41.0000 1.0000 Log Likelihood 800.7606 Analysis Of Parameter Estimates Standard Wald 95% Confidence ChiParameter DF Estimate Error Limits Square Pr > ChiSq Intercept 1 -6.5251 0.1135 -6.7476 -6.3026 3304.10 <.0001 SDI 1 0.0783 0.0076 0.0634 0.0933 105.83 <.0001 Scale 0 1.4475 0.0000 1.4475 1.4475 NOTE: The scale parameter was estimated by the square root of Pearson's Chi-Square/DOF. Observation Statistics Observation Resraw Reschi Resdev StResdev StReschi Reslik 1 2.1681157 0.6587649 0.6384411 0.4488503 0.4631388 0.4493495 2 -0.407323 -0.090167 -0.090469 -0.064866 -0.064649 -0.06485 3 -5.605194 -1.578759 -1.724994 -1.20178 -1.0999 -1.200148 4 -5.673927 -1.824242 -2.069477 -1.439624 -1.269027 -1.437412 5 -3.273178 -1.425389 -1.633527 -1.135744 -0.991031 -1.134021 6 -3.864624 -0.867096 -0.897771 -0.642036 -0.620099 -0.640594 7 3.5525851 1.5221213 1.3900832 0.9726156 1.0650003 0.9750431 8 -4.971053 -1.574265 -1.743451 -1.216465 -1.098418 -1.214254 9 -6.279231 -1.301434 -1.367707 -0.959732 -0.913228 -0.958336 10 -12.94631 -2.763537 -3.138133 -2.203558 -1.940521 -2.195614 11 0.4308971 0.0950092 0.0946804 0.0671446 0.0673778 0.0671565 12 -5.031792 -1.514955 -1.659916 -1.155055 -1.054183 -1.153669 13 -1.235997 -0.826574 -0.928773 -0.644166 -0.573284 -0.64364 14 -0.123381 -0.351256 -0.496751 -0.34324 -0.242708 -0.343208 15 -6.817658 -1.983214 -2.243601 -1.572858 -1.390315 -1.567882 16 1.5139591 0.5944623 0.5733319 0.3985006 0.4131875 0.3986817 17 1.5401423 0.3249805 0.3213686 0.232189 0.2347986 0.2324139 18 8.284726 2.0263824 1.8863175 1.3488043 1.4489572 1.3557005 19 3.1737231 0.6245082 0.6123311 0.4323041 0.4409011 0.4326728 20 1.1261029 0.1578813 0.1573041 0.1141592 0.114578 0.1141985 21 -7.318023 -0.966603 -0.98835 -0.915095 -0.89496 -0.906225 22 5.2451213 2.4053918 2.0923756 1.4551745 1.6728664 1.4582734 23 9.1014375 1.6930581 1.6141537 1.148851 1.2050101 1.1521748 24 3.5647297 1.3073089 1.219454 0.8487809 0.9099308 0.8497224 25 16.21732 2.3219147 2.2083731 1.6048756 1.6873889 1.6130067 26 10.130967 2.8240864 2.5394689 1.7794729 1.9789119 1.7853647 27 -14.50044 -1.946404 -2.041992 -1.484631 -1.415134 -1.478024 28 -4.568623 -0.903507 -0.932641 -0.656817 -0.636299 -0.656054 29 10.900825 1.8399708 1.7552827 1.2411763 1.3010601 1.2439629 30 8.9693032 1.7579876 1.6692525 1.1774843 1.2400778 1.1801061 31 4.1791134 1.40711 1.313497 0.9133191 0.9784114 0.9141874 32 1.0350518 0.186006 0.1849839 0.1307315 0.1314538 0.1307637 33 -2.656701 -0.902956 -0.956301 -0.669791 -0.632428 -0.668805 34 -2.71012 -0.473857 -0.480637 -0.340056 -0.335259 -0.339834 35 -1.087495 -0.752687 -0.838487 -0.581218 -0.521744 -0.580838 36 -2.466543 -0.573979 -0.587525 -0.420773 -0.411072 -0.420106 37 -3.599388 -1.105575 -1.17914 -0.829301 -0.777562 -0.827537 38 1.5376248 0.7278926 0.6911511 0.4831609 0.5088457 0.4837775 39 3.2847809 1.1825847 1.1107865 0.7794281 0.8298082 0.7810225 40 4.3682722 1.3397006 1.2608251 0.881439 0.9365806 0.8827735 41 3.6190279 1.4326777 1.3218309 0.9210496 0.9982875 0.9224192 42 -2.310862 -0.801588 -0.843847 -0.58719 -0.557784 -0.586778 43 -6.495964 -1.837633 -2.046503 -1.448276 -1.300462 -1.441663

PAGE 271

256 Appendix F. Poisson Regression Analysis Outc ome for Adult Hospital Admissions (cont.) Ouliers analyses. Childhood Asthma. Outlier 3 9 21:21 Friday, October 8, 2004 The GENMOD Procedure Model Information Data Set WORK.ASTHMA Distribution Poisson Link Function Log Dependent Variable Asthma_14 Offset Variable ln1 Observations Used 43 Criteria For Assessing Goodness Of Fit Criterion DF Value Value/DF Deviance 41 92.4871 2.2558 Scaled Deviance 41 42.2137 1.0296 Pearson Chi-Square 41 89.8280 2.1909 Scaled Pearson X2 41 41.0000 1.0000 Log Likelihood 675.5512 Analysis Of Parameter Estimates Standard Wald 95% Confidence ChiParameter DF Estimate Error Limits Square Pr > ChiSq Intercept 1 -6.4803 0.1159 -6.7075 -6.2531 3124.83 <.0001 SDI 1 0.0717 0.0079 0.0562 0.0871 82.72 <.0001 Scale 0 1.4802 0.0000 1.4802 1.4802 NOTE: The scale parameter was estimated by the square root of Pearson's Chi-Square/DOF. Observation Statistics Observation Resraw Reschi Resdev StResdev StReschi Reslik 1 2.0427495 0.6171125 0.5992898 0.4121171 0.4243733 0.4125501 2 -0.74007 -0.162505 -0.163486 -0.114701 -0.114012 -0.114651 3 -4.984566 -1.439847 -1.562434 -1.064863 -0.981315 -1.063468 4 -5.130425 -1.697883 -1.91267 -1.301549 -1.155389 -1.29957 5 -3.267072 -1.423554 -1.631206 -1.109146 -0.967952 -1.107455 6 -4.148188 -0.924146 -0.958935 -0.670981 -0.646638 -0.669359 7 3.3894569 1.4309606 1.3143337 0.8996377 0.9794667 0.9017861 8 -4.866989 -1.549416 -1.713596 -1.169315 -1.057283 -1.167203 9 -5.147817 -1.09385 -1.140949 -0.783433 -0.751093 -0.782421 10 -12.35833 -2.674091 -3.026673 -2.078946 -1.836766 -2.071503 11 0.4273045 0.0942089 0.0938856 0.0651229 0.0653472 0.0651345 12 -4.530668 -1.396157 -1.52018 -1.034755 -0.950335 -1.033549 13 -1.272456 -0.844102 -0.950363 -0.644634 -0.572557 -0.644091 14 -0.124642 -0.353047 -0.499284 -0.337379 -0.238563 -0.337346 15 -6.811845 -1.98201 -2.242106 -1.537267 -1.358936 -1.532375 16 1.5987431 0.6318967 0.6080039 0.4132899 0.429531 0.4134917 17 1.0824462 0.2261113 0.2243654 0.1586665 0.1599011 0.1587747 18 -12.26063 -2.65904 -3.00797 -2.144699 -1.89591 -2.120616 19 10.013194 2.5865334 2.3577304 1.6471748 1.8070228 1.6580083 20 4.8511331 0.9871765 0.9566187 0.6609828 0.682097 0.6619255 21 4.7148972 0.685662 0.674716 0.4795662 0.4873462 0.4803226 22 0.4261048 0.0605187 0.0604324 0.0531422 0.0532181 0.0531733 23 5.2188607 2.3867669 2.078558 1.4137471 1.6233777 1.4167551 24 11.247247 2.1745128 2.0439209 1.423661 1.5146227 1.4292092 25 3.6374221 1.3405369 1.2481302 0.8496048 0.9125062 0.8505795 26 10.205712 2.85322 2.5626869 1.7562493 1.9553561 1.7621658 27 -10.72301 -1.490991 -1.547548 -1.102207 -1.061925 -1.098236 28 -3.067857 -0.62534 -0.639385 -0.440696 -0.431016 -0.44032 29 12.428146 2.1449575 2.0296347 1.4046879 1.4845015 1.4085484 30 9.5482194 1.8926195 1.7896721 1.2348765 1.3059103 1.2378926 31 4.4154404 1.5070061 1.3995383 0.9517718 1.0248565 0.9527649 32 1.9048085 0.3472184 0.343649 0.2376004 0.2400683 0.2377125 33 -2.751056 -0.929971 -0.986475 -0.675788 -0.63708 -0.674756 34 -0.872393 -0.15701 -0.157758 -0.10926 -0.108742 -0.109235 35 -1.113064 -0.765709 -0.854312 -0.579145 -0.51908 -0.578757

PAGE 272

257 Appendix F. Poisson Regression Analysis Outc ome for Adult Hospital Admissions (cont.) 36 -2.830305 -0.652236 -0.669692 -0.469355 -0.45712 -0.468499 37 -3.743527 -1.142109 -1.220471 -0.839635 -0.785725 -0.837774 38 1.3732538 0.6384296 0.6102129 0.4173359 0.4366338 0.4178133 39 3.1117972 1.1079553 1.0449873 0.717309 0.760532 0.718702 40 4.4229737 1.3599801 1.2786825 0.8742528 0.929837 0.8756058 41 3.5966055 1.4213063 1.3122218 0.8942362 0.9685737 0.895564 42 -2.169483 -0.75903 -0.797002 -0.542386 -0.516545 -0.542021 43 -6.742115 -1.888754 -2.108647 -1.45998 -1.307731 -1.453052 Ouliers analyses. Childhood Asthma. Outlier 4 11 21:21 Friday, October 8, 2004 The GENMOD Procedure Model Information Data Set WORK.ASTHMA Distribution Poisson Link Function Log Dependent Variable Asthma_14 Offset Variable ln1 Observations Used 43 Criteria For Assessing Goodness Of Fit Criterion DF Value Value/DF Deviance 41 92.7786 2.2629 Scaled Deviance 41 42.8550 1.0452 Pearson Chi-Square 41 88.7626 2.1649 Scaled Pearson X2 41 41.0000 1.0000 Log Likelihood 756.2126 Analysis Of Parameter Estimates Standard Wald 95% Confidence ChiParameter DF Estimate Error Limits Square Pr > ChiSq Intercept 1 -6.5243 0.1167 -6.7530 -6.2955 3124.24 <.0001 SDI 1 0.0759 0.0077 0.0609 0.0909 98.08 <.0001 Scale 0 1.4714 0.0000 1.4714 1.4714 NOTE: The scale parameter was estimated by the square root of Pearson's Chi-Square/DOF. Observation Statistics Observation Resraw Reschi Resdev StResdev StReschi Reslik 1 2.2904417 0.6998959 0.6769396 0.4684213 0.4843064 0.4689917 2 -0.21145 -0.047034 -0.047116 -0.033268 -0.033209 -0.033263 3 -5.18243 -1.484795 -1.61483 -1.106531 -1.017427 -1.10514 4 -5.324343 -1.74364 -1.969267 -1.347289 -1.192925 -1.345371 5 -3.189444 -1.400086 -1.601553 -1.095651 -0.957824 -1.093963 6 -3.659523 -0.82535 -0.853174 -0.600816 -0.581221 -0.599492 7 3.5784021 1.5368288 1.4022153 0.9655031 1.058192 0.968 8 -4.779162 -1.528273 -1.688243 -1.159155 -1.049319 -1.15704 9 -5.503977 -1.160238 -1.213127 -0.837118 -0.800621 -0.836048 10 -12.39269 -2.679371 -3.033238 -2.095936 -1.851418 -2.088419 11 0.7476092 0.1661256 0.1651189 0.1152843 0.1159871 0.1153213 12 -4.677463 -1.43145 -1.561543 -1.068841 -0.979795 -1.067637 13 -1.214536 -0.816148 -0.915962 -0.625043 -0.556931 -0.624524 14 -0.121928 -0.349181 -0.493817 -0.335682 -0.237363 -0.335649 15 -6.632849 -1.94472 -2.195848 -1.515041 -1.341773 -1.510183 16 1.6450049 0.6525437 0.627054 0.4288419 0.4462743 0.4290629 17 1.7230328 0.3650611 0.3605005 0.256547 0.2597924 0.256834 18 -13.50501 -2.846789 -3.242341 -2.316565 -2.033954 -2.291179 19 9.1894828 2.3110962 2.1286281 1.4921578 1.6200671 1.5001411 20 4.192133 0.8416671 0.8194904 0.5683993 0.5837811 0.5690204 21 3.2397197 0.4639534 0.4589529 0.3264745 0.3300316 0.3267861 22 -3.503866 -0.479021 -0.484398 -0.423293 -0.418595 -0.421443 23 5.3091607 2.4513233 2.1262996 1.4550545 1.6774724 1.4583117 24 10.342699 1.9666586 1.859981 1.2993139 1.3738349 1.303405 25 3.7060354 1.3722326 1.2753768 0.873488 0.9398232 0.8745388

PAGE 273

258 Appendix F. Poisson Regression Analysis Outc ome for Adult Hospital Admissions (cont.) 26 18.301225 2.6781039 2.5267694 1.7992761 1.9070392 1.809134 27 -12.2467 -1.678312 -1.749735 -1.247027 -1.196124 -1.2425 28 -3.620545 -0.729668 -0.748745 -0.518257 -0.505052 -0.517789 29 12.003225 2.0586337 1.9524643 1.3577777 1.4316097 1.3611763 30 9.5824584 1.9006851 1.7968525 1.2474984 1.3195861 1.2505879 31 4.4016307 1.501086 1.3944659 0.953999 1.0269412 0.9549905 32 1.8309474 0.3333458 0.3300566 0.2295541 0.2318418 0.2296577 33 -2.556845 -0.874073 -0.924135 -0.637003 -0.602495 -0.636068 34 -1.528082 -0.272143 -0.274387 -0.190781 -0.189221 -0.190712 35 -1.064426 -0.740825 -0.824104 -0.562035 -0.50524 -0.561663 36 -2.311689 -0.540213 -0.552223 -0.389451 -0.380982 -0.388854 37 -3.487383 -1.076877 -1.146773 -0.79384 -0.745456 -0.792146 38 1.5478925 0.7335977 0.6962741 0.4789893 0.5046654 0.4796208 39 3.3419863 1.2076653 1.1327688 0.7822788 0.8340015 0.7839585 40 4.5524458 1.4084371 1.321203 0.9089724 0.9689884 0.9104672 41 3.7095872 1.4790608 1.3608797 0.9331035 1.0141358 0.9345818 42 -2.131001 -0.747329 -0.78416 -0.536904 -0.511686 -0.536542 43 -6.391177 -1.815618 -2.019815 -1.407104 -1.264851 -1.400573

PAGE 274

About the Author Tomas Tamulis received a Bachelor’s of Science Degree in Environmental Engineering from Kaunas University of T echnology in Kaunas, Lithuania, in 1997, and a M.Sc. in Environmental Sciences from Vytautas Magnus University in Kaunas, Lithuania, in 1999. While in the Doctor of Philosophy program at the University of South Florida, Mr. Tomas Tamulis distinguished himself as an outstanding student and graduate teaching assistant. He has taught both regular and we b-based graduate courses in Environmental and Occupational Health, and Analytical Me thods in Industrial Hygiene at the USF College of Public Health during the period 2002-2004. Since 2001, Mr. Tomas Tamulis has been a consultant in biostatistics and study design in the Division of Asthma, Aller gy and Immunology at the USF College of Medicine. Mr. Tamulis has also coauthored one publication and made several conference presentations during his doctoral studies.