Urn!erf Slates Office of A.r Quality EPA--150 5-85-005&
Tnvi.'ennent.i' Protection Planning and Standards August 1985
Agency Research Tnangle Park IMC 2771 1
I&EPA Ambient Ozone And M
. . . , . . ENVIRONMENTAt
Human Health:
An Epidemiological DA'LAS'TEXAS
"^ MU^If*V
Jt I • U* • Ski
Analysis
Volume I
-------
AMBIENT OZONE AND HUMAN HEALTH:
AN EPIDEMIOLOGICAL ANALYSIS
Paul R. Portney and John Mullahy
Resources for the Future
1755 Massachusetts Avenue, N.W.
Washington, B.C. 20036
Volume I
Draft Final Report
September 1983
Submitted to the Economic Analysis Branch, Office of Air Quality Planning
and Standards, Environmental Protection Agency, Research Triangle Park,
North Carolina 27711. under contract number 68^02-3583.
-------
DISCLAIMER
This report has been reviewed by the Office of Air Quality Planning
and Standards, U. S. Environmental Protection Agency, and approved for
publication as received from Resources for the Future. The analysis and
conclusions presented in this report are those of the authors and should
not be interpreted as necessarily reflecting the official policies of
the U. S. Environmental Protection Agency.
-------
ACKNOWLEDGEMENTS
This report and the analysis on which it is based have benefitted from
the efforts of a large number of individuals. Robert Fuchsberg and his
colleagues in the Division of Health Interview Statistics at the National
Center for Health Statistics were extraordinarily helpful in making
available to us and acquainting us with the intricacies of the Health
Interview Survey. Professor Raymond Palmquist of North Carolina State
University gave most generously of his time to match the individuals
interviewed in the 1979 Health Interview Survey to the air pollution
monitors nearest their homes.
Within the Environmental Protection Agency, we received support and
advice from Allen Basala, William Hunt, Bart Ostro, as well as others in
the Economic Analysis Branch, the Ambient Standards Branch, the Monitoring
and Data Analysis Branch, the Health Effects Research Laboratory, and the
Environmental Criteria Assessment Office. Thomas Walton, our project
officer, not only assisted us with countless administrative and procedural
details but also made valuable substantive suggestions at each stage of our
research. V. Kerry Smith, Raymond Palmquist, Duncan Holthausen, Jan
Laarman, and W. Kip Viscusi—all acting as consultants to EPA—also
provided valuable comments on our analysis.
A number of people knowledgeable about air pollution health effects
assisted us in the early stages of our research by identifying air
pollutants that might plausibly be associated with certain types of acute
-------
and chronic illness. We are most grateful in this regard to Gilbert Omenn,
Jane Koenig and Michael Morgan of the Department of Environmental Health at
the University of Washington, Michael Lebowitz of the College of Medicine
at the University of Arizona, and Alice Whittemore of the Stanford
University School of Medicine. John Spengler of the Harvard University
School of Public Health was also very helpful in this regard. We also
benefitted from comments received during seminars at the University of
California at Berkeley, Harvard University, Johns Hopkins University, the
University of Washington and Washington State University.
Finally, we owe several of our colleagues at Resources for the Future
a debt of thanks. Michael Hazilla wrote the programs which summarized the
mass of air quality data with which we worked and matched it to the periods
for which had individuals' health histories. Richard Carson provided
expert research assistance in the early stages of the project. Among those
who commented on early versions of the chapters, Robert Frank deserves
special mention. He read and commented extensively on both the form and
substance of every chapter. William Vaughan, Clifford Russell, Winston
Harrington and Allen Kneese also provided very helpful comments on
individual chapters. Margaret Parr-Recard, Pat Flynn, D.J. Curran, Anne
Farr and others worked very hard typing and helping produce this draft
report. We are grateful to them all for their help.
It goes without saying that none of the individuals mentioned above
bears any responsibility for the contents of this report. That is the
responsibility of the authors alone.
-------
"ABLE OF CONTENTS
VOLUME I
Page
CHAPTER 1. INTRODUCTION
1.1 General Background
1.2 Economic Valuation and Physical Effects
1.3 Identifying Air Pollution Health Effects
1.4 The RFF Study
1.5 Summary and Plan of this Report
CHAPTER 2. DATA
2.1 Overview
2.2 The Health Interview Survey
2.2.1 Main Survey
2.2.2 Smoking Supplement
2.2.3 Residential Mobility Supplement
2.2.U Documentation and Cross Referencing of HIS Data
Air Pollution Data
2.3
2.5
2.3.1 SAROAD System
2.3.2 Data Integrity and Completeness
2.3.3 Multi-Year Averaged Data
2.3.4 Description of the Pollution Variables
2.3.5 Matching Individuals to Monitors
Meteorological Data
Other Data
2.5.1 Pollen
2.5.2 Indoor Air Pollution
2.5.3 Paid Sick Leave
2.5.4 The Pollutant Standard Index
2.5.5 availability of Medical Care
2.6 Problems of Sample Selection
CHAPTER 3. METHODOLOGY
3.1 Measures of Health Status:
Variables
3.2 Explanatory Variables
3.3 Estimation Techniques
Choice of the Dependent
Appendix to Chapter 3:
Definitions of and Procedures Used
to Create Measures of Morbidity
CHAPTER 4. RESULTS AND DISCUSSION
4.1 Acute Morbidity Due to All
4.1.1 Adults
4.1.2 Children
Causes
1-1
1-4
1-5
1-15
1-20
2-1
2-9
2-9
2-16
2-19
2-21
2-22
2-22
2-24
2-37
2-42
2-46
2-4?
2-51
2-52
2-54
2-56
2-57
2-59
2-60
3-2
3-6
3-52
A3-1
4-5
4-5
4-21
-------
4.2 Acute Morbidity Due to Respiratory Disease 4-28
4.2.1 Adults 4-31
4.2.2 Children 4-55
4.3 Chronic Respiratory disease 4-59
4.3.1 Adults 4-61
4.3.2 Children 4-81
4.4 Cardiovascular and Other Chronic Diseases 4-85
4.5 Analysis of Multicollinearity 4-89
CHAPTER 5. ALTERNATIVE OZONE CONCENTRATIONS AND CHANGES IN
MORBIDITY: ILLUSTRATIVE ESTIMATES
5.1 Ozone-related Changes in Morbidity 5-2
5.1.1 Acute Morbidity 5-3
5.1.2 Chronic Morbidity 5-6
5.2 Acute and Chronic Illness: Feedback Effects 5-8
5.3 Relative Risks 5-11
5.4 Monetary Benefit Estimates 5-18
VOLUME II
APPENDIX A. The Health Interview and Smoking Supplement
APPENDIX B. Health Interview Survey
APPENDIX C. -Valuing the Benefits of Improved Human Health
-------
TABLE OF CONTENTS
VOLUME I
Page
CHAPTER 1. INTRODUCTION
1.1 General Background
1.2 Economic Valuation and Physical Effects
1.5
CHAPTER
2.1
2.2
2.3
Identifying Air Pollution Health Effects
The RFF Study
Summary and Plan of this Report
2. DATA
Overview
The Health Interview Survey
2.2.1 Main Survey
2.2.2 Smoking Supplement
2.2.3 Residential Mobility Supplement
2.2.4 Documentation and Cross Referencing of HIS Data
Air Pollution Data
SAROAD System
Data Integrity and Completeness
Multi-Year Averaged Data
Description of the Pollution Variables
Matching Individuals to Monitors
logical Data
ata
Pollen
Indoor Air Pollution
Paid Sick Leave
The Pollutant Standard Index
availability of Medical Care
2.6 Problems of Sample Selection
2.4
2.5
2.3.1
2.3.2
2.3.3
2.3.4
2.3.5
Meteoi
Other
2.5.1
2.5.2
2.5.3
2.5.4
2.5.5
CHAPTER 3. METHODOLOGY
3.1 Measures of Health Status:
Variables
3.2 Explanatory Variables
3.3 Estimation Techniques
Choice of the Dependent
Appendix to Chapter 3'
Definitions of and Procedures Used
to Create Measures of Morbidity
CHAPTER 4. RESULTS AND DISCUSSION
4.1 Acute Morbidity Due to All Causes
4.1.1 Adults
4.1.2 Children
1-1
1-4
1-5
1-15
1-20
2-1
2-9
2-9
2-16
2-19
2-21
2-22
2-22
2-24
2-37
2-42
2-46
2-4?
2-51
2-52
2-54
2-56
2-57
2-59
2-60
3-2
3-6
3-52
A3-1
4-5
4-5
4-21
-------
4.2 Acute Morbidity Due to Respiratory Disease
1.2.1 Adults
4.2.2 Children
4.3 Chronic Respiratory disease
4.3.1 Adults
4.3.2 Children
4.4 Cardiovascular and Other Chronic Diseases
4.5 Analysis of Multicollinearity
CHAPTER 5. ALTERNATIVE OZONE CONCENTRATIONS AND CHANGES IN
MORBIDITY: ILLUSTRATIVE ESTIMATES
5.1 Ozone-related Changes in Morbidity
5.1.1 Acute Morbidity
5.1.2 Chronic Morbidity
5.2 Acute and Chronic Illness: Feedback Effects
5.3 Relative Risks
5.4 Monetary Benefit Estimates
4-28
4-31
4-55
4-59
4-61
4-81
4-85
4-89
5-2
5-3
5-6
5-8
5-11
5-18
VOLUME II_
APPENDIX A.
APPENDIX B.
APPENDIX C.
The Health Interview and Smoking Supplement
Health Interview Survey
-Valuing the Benefits of Improved Human Health
-------
TABLE OF CONTENTS
VOLUME I
Page
CHAPTER 1. INTRODUCTION
1.1 General Background
1.2 Economic Valuation and Physical Effects
1.3 Identifying Air Pollution Health Effects
1.4 The RFF Study
Summary and Plan of this Report
1.5
2.1
2.2
CHAPTER 2. DATA
Overview
The Health Interview Survey
2.2.1 Main Survey
2.2.2 Smoking Supplement
2.2.3 Residential Mobility Supplement
2.2.U Documentation and Cross Referencing of HIS Data
2.3 Air Pollution Data
2.3.1 SAROAD System
2.3.2 Data Integrity and Completeness
2.3.3 Multi-Year Averaged Data
2.3.4 Description of the Pollution Variables
2.3.5 Matching Individuals to Monitors
Meteorological Data
Other Data
2.5.1 Pollen
2.5.2 Indoor Air Pollution
2.5.3 Paid Sick Leave
2.5.4 The Pollutant Standard Index
2.5.5 availability of Medical Care
2.6 Problems of Sample Selection
2.U
2.5
CHAPTER 3- METHODOLOGY
3.1 Measures of Health Status:
Variables
3.2 Explanatory Variables
3.3 Estimation Techniques
Choice of the Dependent
Appendix to Chapter 3:
Definitions of and Procedures Used
to Create Measures of Morbidity
CHAPTER 4. RESULTS AND DISCUSSION
4.1 Acute Morbidity Due to All Causes
4,1.1 Adults
4.1.2 Children
1-1
1-4
1-5
1-15
1-20
2-1
2-9
2-9
2-16
2-19
2-21
2-22
2-22
2-24
2-37
2-42
2-46
2-4?
2-51
2-52
2-54
2-56
2-57
2-59
2-60
3-2
3-6
3-52
A3-1
4-5
4-5
4-21
-------
4.2 Acute Morbidity Due to Respiratory Disease
4.2.1 Adults
4.2.2 Children
4.3 Chronic Respiratory disease
1.3.1 Adults
U.S.2 Children
4.4 Cardiovascular and Other Chronic Diseases
4.5 Analysis of Multicollinearity
CHAPTER 5. ALTERNATIVE OZONE CONCENTRATIONS AND CHANGES IN
MORBIDITY: ILLUSTRATIVE ESTIMATES
5.1 Ozone-related Changes in Morbidity
5.1.1 Acute Morbidity
5.1.2 Chronic Morbidity
5.2 Acute and Chronic Illness: Feedback Effects
5.3 Relative Risks
5.4 Monetary Benefit Estimates
4-28
4-31
4-55
4-59
4-61
4-81
4-85
4-89
5-2
5-3
5-6
5-8
5-11
5-18
VOLUME II
APPENDIX A.
APPENDIX B.
APPENDIX C.
The Health Interview and Smoking Supplement
Health Interview Survey
-Valuing the Benefits of Improved Human Health
-------
TABLE OF CONTENTS
Page
VOLUME I
CHAPTER 1. INTRODUCTION
1.1 General Background 1-1
1.2 Economic Valuation and Physical Effects 1-4
1.3 Identifying Air Pollution Health Effects 1-5
1." The RFF Study 1-15
1.5 Summary and Plan of this Report 1-20
CHAPTER 2. DATA
2.1 Overview 2-1
2.2 The Health Interview Survey 2-9
2.2.1 Main Survey 2-9
2.2.2 Smoking Supplement 2-16
2.2.3 Residential Mobility Supplement 2-19
2.2.4 Documentation and Cross Referencing of HIS Data 2-21
2.3 Air Pollution Data 2-22
2.3.1 SAROAD System 2-22
2.3.2 Data Integrity and Completeness 2-24
2.3-3 Multi-Year Averaged Data 2-37
2.3.4 Description of the Pollution Variables 2-42
2.3-5 Matching Individuals to Monitors 2-46
2.4 Meteorological Data 2-47
2.5 Other Data " 2-51
2.5.1 Pollen 2-52
2.5.2 Indoor Air Pollution 2-54
2.5.3 Paid Sick Leave 2-56
2.5.4 The Pollutant Standard Index 2-57
2.5.5 availability of Medical Care 2-59
2.6 Problems of Sample Selection 2-60
CHAPTER 3. METHODOLOGY
3.1 Measures of Health Status: Choice of the Dependent
Variables 3-2
3.2 Explanatory Variables 3-6
3.3 Estimation Techniques 3-52
Appendix to Chapter 3: Definitions of and Procedures Used
to Create Measures of Morbidity A3-1
CHAPTER 4. RESULTS AND DISCUSSION
4.1 Acute Morbidity Due to All Causes 4-5
4.1.1 Adults 4-5
4.1.2 Children 4-21
-------
4.2 Acute Morbidity Due to Respiratory Disease
4.2.1 Adults
4.2.2 Children
4.3 Chronic Respiratory disease
4.3.1 Adults
U.S.2 Children
4.4 Cardiovascular and Other Chronic Diseases
4.5 Analysis of Multicollinearity
CHAPTER 5. ALTERNATIVE OZONE CONCENTRATIONS AND CHANGES IN
MORBIDITY: ILLUSTRATIVE ESTIMATES
5.1 Ozone-related Changes in Morbidity
5.1.1 Acute Morbidity
5.1.2 Chronic Morbidity
5.2 Acute and Chronic Illness: Feedback Effects
5.3 Relative Risks
5.4 Monetary Benefit Estimates
4-28
4-31
4-55
4-59
4-61
4-81
4-85
4-89
5-2
5-3
5-6
5-8
5-11
5-18
VOLUME II_
APPENDIX A. The Health Interview and Smoking Supplement
APPENDIX B. Health Interview Survey
APPENDIX C. -Valuing the Benefits of Improved Human Health
-------
OF CONTENTS
VOLUME I
Page
CHAPTER 1. INTRODUCTION
1.1 General Background
1.2 Economic Valuation and Physical Effects
1.3 Identifying Air Pollution Health Effects
1.14 The RFF Study
1.5 Summary and Plan of this Report
CHAPTER 2. DATA
2.1 Overview
2.2 The Health Interview Survey
2.2.1 Main Survey
2.2.2 Smoking Supplement
2.2.3 Residential Mobility Supplement
2.2.4 Documentation and Cross Referencing of HIS Data
2.3 Air Pollution Data
SAROAD System
Data Integrity and Completeness
Multi-Year Averaged Data
Description.of the Pollution Variables
Matching Individuals to Monitors
Dlogical Data
)ata
Pollen
Indoor Air Pollution
Paid Sick Leave
The Pollutant Standard Index
availability of Medical Care
2.6 Problems of Sample Selection
CHAPTER 3. METHODOLOGY
3.1 Measures of Health Status:
Variables
3.2 Explanatory Variables
3.3 Estimation Techniques
2.4
2.5
2.3-1
2.3.2
2.3.3
2.3.5
Meteoi
Other
2.5.1
2.5.2
2.5.3
2.5.4
2.5.5
Choice of the Dependent
Appendix to Chapter 3:
Definitions of and Procedures Used
to Create Measures of Morbidity
CHAPTER 4. RESULTS AND DISCUSSION
4.1 Acute Morbidity Due to All Causes
4.1.1 Adults
4.1.2 Children
1-1
1-4
1-5
1-15
1-20
2-1
2-9
2-9
2-16
2-19
2-21
2-22
2-22
2-24
2-3?
2-42
2-46
2-47
2-51
2-52
2-54
2-56
2-57
2-59
2-60
3-2
3-6
3-52
A3-1
4-5
4-5
4-21
-------
4.2 Acute Morbidity Due to Respiratory Disease 4-28
I*. 2.1 Adults 4-31
4.2.2 Children 4-55
4.3 Chronic Respiratory disease 4-59
4.3.1 Adults 4-61
4.3.2 Children 4-81
4.4 Cardiovascular and Other Chronic Diseases 4-85
4.5 Analysis of Multicollinearity 4-89
CHAPTER 5. ALTERNATIVE OZONE CONCENTRATIONS AND CHANGES IN
MORBIDITY: ILLUSTRATIVE ESTIMATES
5.1 Ozone-related Changes in Morbidity 5-2
5.1.1 Acute Morbidity 5-3
5.1.2 Chronic Morbidity 5-6
5.2 Acute and Chronic Illness: Feedback Effects 5-8
5.3 Relative Risks 5-11
5.4 Monetary Benefit Estimates 5-18
VOLUME II_
APPENDIX A. The Health Interview and Smoking Supplement
APPENDIX B. Health Interview Survey
APPENDIX C. -Valuing the Benefits of Improved Human Health
-------
Chapter 1
INTRODUCTION
\. 1 General Background
This report presents the methodology and results of a study undertaken by
Resources for the Future (RFF) for the Economic Analysis Branch of the
Environmental Protection Agency's Office of Air Quality Planning and
Standards (OAQPS), The study was designed to identify the benefits in the
form of improved human health associated with possible alternative air
quality standards for ozone.
One act of Congress and one executive order provide the impetus for
this study. In 1970, Congress amended the Clean Air Act of 1963 in such a
way as to fundamentally alter the approach by which air quality was pursued
in the United States. Among other things, the 1970 Amendments to the Clean
Air Act directed the Administrator of the Environmental Protection Agency
(EPA) to establish two kinds of National Ambient Air Quality Standards
(NAAQS) for a handful of common (or "criteria") air pollutants. Primary
standards were those designed to protect human health with an "adequate
margin of safety." Secondary standards were also to be issued where
appropriate; these were to be set at levels which afforded protection to
-------
1-2
vegetation, aquatic ecosystems, visibility, and other "welfare"
considerations (where "welfare" refers to non-health concerns, rather than
to general well-being as in economists' parlance).
Accordingly, in 1971 EPA promulgated NAAQSs for total suspended
particulates, sulfur dioxide, carbon monoxide, nitrogen dioxide, and
photochemical oxidants. A standard was also issued for hydrocarbons
although it was intended only to assist states and localities in meeting
the NAAQS for photochemical oxidants. In 1978 a NAAQS was also issued for
airborne concentrations of lead. To this date, these same six pollutants
are the only ones governed by NAAQSs.
Although the 1970 amendments to the Clean Air Act directed the
administrator of EPA to revise the ambient standards whenever new data
accumulated to so warrant, in 1977 Congress provided more specific
guidance. The 1977 amendments directed EPA to review each NAAQS at least
every five years and to revise the existing standard where it seemed called
for, or retain the existing standard if no compelling evidence suggested
that it should be changed. In 1978, in the first of these mandatory
revisions, EPA proposed to change the basis of the photochemical oxidant
standard to ozone (the most common photochemical oxidant), and to relax the
standard from .08 parts-per-million (ppm), not to be exceeded during the
highest hour each day more than once per year, to .10 ppm not to be
exceeded (on an expected value basis) more than three times over a three
year period. In 1979, EPA issued a final revised NAAQS for ozone. It was
set at .12 ppm, to be measured in the way suggested in the proposed
standard.
-------
1-3
Because of the five-year mandatory review schedule established in the
1977 amendments to the Clean Air Act, EPA is now reviewing the scientific
evidence relating ozone to human health, agricultural output, visibility
impairment, and other effects. One purpose of the present study is to
provide some additional information to the EPA, and specifically to the
OAQPS, as part of that review.
However, the study reported here was also undertaken on account of
Executive Order 12291, issued by President Reagan on February 17, 1981.
That order, a successor to Executive Order 12044 issued by President
Carter, directed all federal regulatory agencies in the executive branch
to prepare both preliminary and final Regulatory Impact Analyses (RIAs) to
accompany all proposed and final "major" federal rules and regulations.
These RIAs are, among other things, to describe the potential benefits
expected to result from proposed or final rules, identify the potential
costs, calculate the net benefits (the former minus the latter), and
describe alternative approaches that might accomplish the same things as
the rule in question, but at a lower cost—along with an explanation of the
legal reasons why such alternatives could not be selected if, in fact, they
were not.
Because the promulgation of NAAQSs have been designated as major
rulemakings, Regulatory Impact Analyses have generally accompanied them.
Thus, another purpose of the present study is to provide information on the
human health benefits associated with possible alternative NAAQSs for ozone
so as to facilitate the compilation of the overall benefits—health,
agricultural, materials damage, aesthetic, and other—associated with
-------
1-U
different ozone standards. Such information, compiled from a wide variety
of sources, could then be combined with data on the costs associated with
these same alternatives to provide information on the overall effects
expected to result from varying possible ozone standards.
1.2 Economic Valuation and Physical Effects
The original purpose of this study was to estimate, in dollar terms where
possible, the value of any reduced human morbidity that might accompany
reductions in ambient ozone concentrations (or the additional morbidity
that might arise from a relaxation of the standard). We were to
concentrate on both acute as well as chronic morbidity. That is, we were
to concern ourselves with the temporary or day-to-day changes in individual
health status that might result from human exposure to ambient ozone, and
how these might be valued. In addition, we were to attempt to place dollar
values on any long-term, or chronic, illness that might result, perhaps not
so much from occasional exposures to peak ozone concentrations, but rather
from prolonged exposures to ozone as well as other air pollutants.
Although the overall focus of the work reported below is still the
same, the emphasis has shifted somewhat during the course of our study.
Specifically, because of the availability of an extraordinary data set, it
became clear during the initial phase of the RFF study that we might not
only say something original about the economic .valuation of ozone health
effects, but might also be able to provide information about the
relationship between ambient ozone concentrations and the health effects
themselves. The question of economic valuation has received considerable
-------
1-5
thought during the course of our study. Appendix C to this report presents
a model of household behavior from which are derived measures of the
valuation of health improvements. However, more work than was originally
anticipated was devoted to identifying the acute and chronic health effects
that might be related to ambient exposures to ozone and other air
pollutants. Thus, this report concentrates primarily on the identification
and quantification of health effects resulting from exposure to ozone.
1.3 Identifying Air Pollution Health Effects
There are three primary means of ascertaining the relationship between
ozone or other air pollutants and both acute and chronic morbidity. In
this report we shall refer to these three approaches as toxicological,
clinical, and epidemiological. Although we make use of but one of these
approaches in our study, it is useful to discuss all three albeit briefly.
Under the toxicological approach, laboratory animals are exposed under
carefully controlled conditions to varying levels of specific air
pollutants. Dose-response relationships are estimated either by exposing
otherwise identical groups of animals to different air pollution regimes
and then observing differences in biological end-points (which may be
functional, biochemical, structural, or behavioral), or by exposing the
same animal(s) to alternative ozone or other air pollution levels at
different times and observing changes in the same end-points under
different regimes. The information gleaned from such animal tests is used
to provide insight into the effects of these same air pollutants on human
health.
-------
1-6
As with the other approaches to health effects estimation, there are
both advantages and disadvantages to the toxicological method. On the
positive side, animals can be subjected to carefully controlled air
pollution levels, at least some of which experimenters would be unwilling
to administer to human subjects. In addition, since the lifetimes of some
laboratory animals, as for example mice, may be only about two years, it is
possible to observe the effects of a simulated "lifetime" exposure to air
pollution in a manageable period of time. Finally, laboratory animals can
be sacrificed during the course of their lives to observe directly the
timing of physiological changes resulting from the exposure to air
pollution that may occur anywhere within the organism.
There are, however, several serious drawbacks to toxicological
studies as a means of determining human susceptibility. The most obvious
of these involves the uncertainty of extrapolating the results seen in
animals to human beings.- Researchers are unsure whether humans will
respond even qualitatively to air pollution (or other environmental
stimuli) in the same way as do experimental animals, much less whether the
quantitative dose-response relationships estimated in animal studies are
representative of likely human responses. In addition, the cost of
toxicological studies sometimes necessitates that animals be given "doses"
of air pollution much greater than levels to which humans might ever be
exposed to make sure an effect is observed if it exists. Thus, there often
arises the additional problem of extrapolating from effects observed at
high doses to those observed at much lower doses, even before the
inter-species extrapolations must be made.
-------
1-7
A second approach to human health effects estimation is the clinical.
At the risk of over-simplification, this involves controlled
experimentation in exposure chambers or other laboratory settings using
human subjects. Typically, volunteer subjects are exposed acutely—usually
for periods lasting up to a few hours—to a variety of different levels of
air pollution, perhaps under different exercise or other conditions as
well. Their symptomatic, physiological, or in some instances biochemical
responses are monitored to provide evidence about the effects of air
pollution on human health.
There are several desirable features of clinical studies. First, and
perhaps foremost, the dose of any pollutant(s) can be carefully controlled
and measured. Thus, we can be quite sure that a certain individual was
exposed to, say, .20 ppm ozone for exactly two hours and in the absence of
•
other pollutants, because the air inflow to a chamber can be controlled
quite closely. Also, because individuals have the opportunity to report
symptoms as they are experienced, or because physiological changes can be
measured in "real time" (i.e., contemporaneously), the likelihood of
establishing a causal link is quite good. In other words, establishing
such a link does not depend on an individual's recollection of a symptom
sometime in the recent or distant past. Finally, if the pollutant(s) in
question are thought to be particularly harmful to "sensitive
individuals"—those who because of age, underlying disease, or some other
distinguishing characteristic, are thought to be particularly adversely
affected by the pollutant—these individuals can be selected for study.'
In other words, the experimenter need not rely on good fortune to ensure
-------
1-8
that there are enough children, asthmatics, or other sensitive individuals
in the sample group.
In spite of these attractive aspects, there are several drawbacks to
the clinical approach. First, and least serious, concerns the air
pollution dose that subjects receive. Although it is generally carefully
measured, it is also artificial and typically is not representative of the
entire mixture of air pollutants and other substances to which individuals
are normally exposed in the course of their everyday lives. For example,
it may not contain any of the allergens or infectious material to which we
are all exposed from time to time. Thus if there are important synergistic
effects between several pollutants, they may be overlooked in clinical
studies.
Another problem with clinical studies concerns the numbers of
participants. Perhaps on account of cost, clinical studies are
generally—although not always—characterized by small numbers of subjects.
It is not uncommon to find studies with fewer than te.i subjects, while it
is unusual to see studies in which more than fifty individuals participate.
This creates two problems. First, it limits one's ability to generalize
the results of a particular study to larger, potentially more diverse
populations. Such generalizations will be especially difficult if the
individuals studied are more sensitive in some respect than the overall
population.
Small samples create another problem. For any given level of
statistical significance, the smaller the number of experimental subjects
the larger must be any observed effect. Since at least some air pollution
-------
1-9
health effects can be quite subtle, this means that potentially important
relationships can sometimes be missed. Larger sample sizes make this less
likely.
These problems pale in comparison to the most serious drawback of the
clinical approach—its general inability to shed light on the possible
long-term or chronic health effects associated with ozone or other air
pollutants. For while subjects can be placed in artificial settings where
exposures are controlled for hours or perhaps even days or weeks, legal,
ethical and practical considerations prevent use of exposure regimes likely
to cause persistent or irreversible damage, or that would take years to
complete. Yet one important concern about air pollution is that prolonged
exposures, even to levels well below those at which acute effects are
thought to begin, may eventually result in chronic respiratory or other
kinds of disease. Clinical studies cannot resolve this concern. In fact,
air quality standards designed primarily to guard against acute health
impairments associated with air pollution might still permit the onset and
exacerbation of chronic illnesses that ought also to be taken into account
in setting health-based air quality standards. It is here that reliance
must be placed on epidemiological research.
An epidemiological study is one that looks at different groups of
individuals at one point in time or the same individuals at different times
(or both) and attempts to correlate differences in their health status to
other differences between them (environmental, personal, and other). Of
course, health status is affected by much more than exposure to air
pollution (or even all environmental factors) and these other factors can
-------
1-10
vary dramatically across different individuals or population groups at any
one time; they can also differ for specific individuals over time—income
may grow, personal habits (particularly smoking) may change, weather
varies, marital status, occupation, education and a host of other factors
may change and all may significantly influence health. A goal of
epidemiological research is to hold constant as many of these other
identifiable influences as possible. This may be done either through
statistical means in the case of a heterogeneous population, or by
deliberately selecting sample populations that are alike in as many ways as
possible save for the characteristic(s) whose potential influences on
health is to be isolated.
Epidemiological studies can be distinguished in a number of ways. One
distinction revolves around the units of observation, whether individuals
or aggregate populations. In the former, data are available on specific
individuals—their health status, socioeconomic characteristics,
environmental exposures, and so on. In aggregate epidemiology, on the
other hand, a researcher tries to link variations in morbidity or mortality
rates for different cities, counties, states or even countries to air
pollution after attempting to control for the variation in other
potentially important explanatory variables. Individual data are always
to be preferred where it is available because it enables one to avoid the
so-called "ecological fallacy" which can arise when averaged or pooled data
obscure important individual characteristics.
Another distinction that can be drawn among epidemiological studies
concerns the extent to which the study is controlled or uncontrolled. In a
-------
1-11
controlled (or ex ante) study, the individuals or groups to be studied are
identified ahead of time. They are then monitored over the course of the
study period for changes in health status, personal characteristics, and so
on. In addition, ozone concentrations, say, are carefully monitored
during the study period so that they can be linked to health status (or
other dimensions of interest). In other words, such controlled or ex ante
studies are not too unlike clinical studies save for the fact that the air
pollution exposures take place in real rather than artificial settings, and
for the possibly extended duration of epidemiological studies.
In uncontrolled (or ex post) epidemiological studies, the researcher
may examine the health status of individuals or groups for whom health
status, other information, and air pollution exposures are known or can be
constructed, even though some or all of this information may have been
collected for reasons completely unrelated to air pollution or other kinds
of epidemiology.' In either controlled or uncontrolled studies, the
quality of the results depend critically upon the breadth and reliability
of the data on both the variables of interest (exposure to air pollution
and health status) as well as the potentially confounding influences.
There are a number of desirable features to the epidemiological
approach. Foremost among these is that, in principle, epidemiological
studies have the capability of shedding light on any possible chronic
health effects associated with air pollution or other factors. If one
could identify individuals who have lived in one location their whole
lives, if one had reliable measures of their health status, if one knew the
levels of ozone and other air pollutants to which they were exposed over
-------
1-12
this period, and if one could control for the many other influences on
health, then one could draw some (cautious) inferences about the link
between air pollution and health. Obviously, these are very big "ifs" and
the likelihood of their being satisfied is discussed below. Nevertheless,
we emphasize again that, unlike clinical studies, epidemiological studies
can in principle make an important contribution to the study of air
pollution and chronic illness.
Another advantage to epidemiological studies is that, in general, they
include larger numbers of individuals than typically participate in
clincical studies. This is not inherent in the nature of such studies, but
>
it is generally the case. Such large numbers make it more likely that any
statistically significant relationships will be identified—be they between
air pollution and health status or between other dependent and independent
variables.
There is a final, less important advantage to epidemiological studies.
By definition, such studies are based on ambient or "real world" exposures
to air pollution and other environmental factors, rather than to the
artificial exposures administered in clinical studies. Thus, if the
exposure data are measured correctly, they accurately (indeed
definitionally) reflect the conditions under which people actually receive
doses of pollution. Such studies avoid the problems that may arise if
laboratory settings induce by themselves certain behavioral or
physiological changes.
Like the toxicological and clinical approaches, however, the
epidemiological approach is not without problems. It, too, has drawbacks
-------
1-13
that can seriously limit the confidence one can place in the results of
epidemiological studies. The most serious of these concerns the
measurement of individual exposures to any particular air pollutant. In
contrast to clinical studies, where concentrations can be carefully
controlled, epidemiological studies typically rely for exposure data on
readings from ambient air pollution monitors in the areas where individuals
live, work and recreate. In some cases, in fact, no monitoring data may be
available from the area in which one or more individuals live. And even
where data are available for a particular region, they may be suspect for a
number of reasons. Since individuals may in the course of their daily
activities travel all about a particular metropolitan area, one must decide
whether to characterize air pollution dose by the monitor nearest their
homes or offices, or take some kind of weighted average based on relative
amounts of time spent in various locations.
Often no data whatsoever are available about such activity patterns,
so that even in the presence of comprehensive ambient monitoring data, it
would be very difficult to create a weighted exposure index. Complicating
the estimate of exposure still further is the matter of indoor versus
Q
outdoor exposure to air pollution. It is now known, for example, that
individuals may be exposed to much higher peak concentrations of nitrogen
dioxide, particulate matter, and carbon monoxide in indoor than in outdoor
environments even though these are three pollutants for which ambient
standards exist. In addition, individuals may be exposed to radon,
formaldehyde, carbon dioxide and other air pollutants in indoor
environments, and these indoor concentrations may make it difficult to
-------
1-14
isolate the effects on health of ambient air quality. (Because this study
deals specifically with the health effects associated with ozone, it is
noteworthy that ozone is one of the air pollutants that is almost
non-existent in indoor environments. With the exception of certain
occupational settings (for example, welding), almost all exposures result
from contact with ambient air.)
The seriousness of the problem of estimating exposure varies between
epidemiological studies. In some studies the nearest air pollution
monitors are so far removed from the subjects, or their data so unreliable,
that little confidence can be placed in the results. In other studies,
however, monitoring is done nearby and very carefully. In these cases,
exposures tcay be well represented. Nevertheless, ambient monitoring data
can be no more than a proxy for actual total exposure to ozone and other
•
air pollutants.
Another problem, about which less will be said, concerns the controls
in epidemiological studies for variables other than air pollution.
Clearly, the greater the extent of controls for such variables, the greater
the confidence that can be placed in any air pollution health effect that
is identified. The extent of these controls varies markedly between
studies—in some, such obvious and potentially confounding influences as
personal smoking habits, pre-existing respiratory disease, or exposures to
other air pollutants are missing. In other studies, earnest efforts are
made to account for these and other much more subtle influences on health.
As in the case of exposure data in epidemiological studies, the severity of
-------
1-15
the "control problem" depends on the specific measures taken in each study
to address the problem.
1.U The RFF Study
The study reported here is an epidemiological one. It combines four
different types of data to shed light on acute and chronic human morbidity
that may be associated with ozone, other criteria (and some non-criteria)
air pollutants, and other possible contributory factors.
As described in detail below, the data on acute and chronic morbidity
as well as socioeconomic characteristics are specific to individuals rather
than to population groups. Thus, ours is an individual (or micro-)
epidemiological study. The individual health and socioeconomic data for
this study come from the 1979 Health Interview Survey (HIS) of more than
110,000 individuals, a survey that is conducted annually by the National
Center for Health Statistics (NCHS) in the Department of Health and Human
Services (HHS). The survey is reproduced in its entirety in Appendix A to
this report.
The health data are of two types. First, information about acute
illness was elicited from all individuals surveyed, information that was
specific to the two-week period ending the Sunday night before the week of
the interview. Individuals were asked about physical conditions that
confined them to bed for most of a day, prevented normal activity (work or
school) without requiring confinement to bed, or forced them to restrict
their normal activity in some way even though it was not serious enough to
force them to stay in bed or miss work or school. Individuals were also
-------
1-16
asked about doctor visits during the two-week period as well as short-stay
hospital episodes, although these are not analyzed in our report.
In addition to information about acute health status during a specific
two week period, invididuals participating in the 1979 HIS were also asked
whether they had any chronic health impairments that limited their
activities. Adult respondents could also identify chronic illnesses or
impairments through their responses to a separate section of the 1979 HIS.
Both acute and chronic impairments were identified by cause and recorded
with the International Classification of Disease (ICD) codes.
One of the advantageous features of the HIS is the form of the
questions about acute health status. Because such concepts as a. "work loss
day" or a "bed disability day" are easily understandable, individuals may
be able to value the prevention of such occurrences without much
difficulty. This is important in a study like this one—one aim of which
is to translate physical improvements in health into dollar terms—because
these individual valuations are the basis of benefit calculations in
welfare economics.
This would be difficult to accomplish if each of the respondents in
the HIS were given a test of lung function as part of the examination. In
such a case, we would in our statistical analysis be identifying a link
between ozone concentrations and lung function. If we calculated a change
in the latter for a given change in ozone, the resulting effect—although
important to a clinical appraisal—would not lend itself to valuation as
easily as a practical endpoint like one extra or fewer day spent sick in
-------
1-17
bed. Thus, from the standpoint of applied welfare economics the HIS
elicits health information in a useful form.
The socioeconomic data collected for each individual in the HIS
include age, race, sex, income, education, occupation, industry of
employment, marital status, history of military service, as well as
other individual and household-specific characteristics discussed below.
In addition to the health and socioeconomic information that is
elicited each year as part of the HIS, several supplements to the 1979
survey made it particularly useful for epidemiological purposes.
Specifically, the 1979 HIS contained a supplement that went to one-third of
all the adults interviewed (26,271 of a total of 79,743 adults) which
provided detailed data on lifetime smoking history, including the tar and
nicotine content of the brands most commonly smoked. (The smoking
supplement is included in Appendix A.) This is of great importance if one
is interested in examining respiratory and cardiovascular disease.
Also, the 1979 HIS included a supplement (again to one-third of all
adults surveyed) designed to provide detailed information on residential
histories. This information is important in examining the possible links
between air pollution and chronic respiratory disease, since it makes it
possible to identify individuals who have lived in their present location
for a long period of time. Supplemental questions were also asked about
eye care and home health care in the 1979 HIS.
To this rich set of individual health, socioeconomic, smoking and
residential history data, RFF added three other types of data. The most
important of these were data on air quality in the United States in 1979»
-------
1-18
consisting of all the hourly or 24-hour readings from the air pollution
monitors that reported to EPA's SAROAD file (JStorage a.nd Retrieval of
Aerometric Data) in 1979, as well as summary measures of this data. We
also merged with the individual HIS data, other annual average pollution
data dating back in some cases as far as 197** (again from EPA's SAROAD
network). Thus, we used air pollution data that were contemporaneous with
the period for which we had information on individuals' acute morbidity (as
opposed to using annual average air pollution to proxy exposure during a
specific part of the year). In addition, we had annual average air
pollution data from the urban areas in which the individuals in our sample
resided (for both ^979 as well as previous years in many cases) to assist
in exploring the possible role of air pollution in chronic disease
etiology.
In addition to this contemporaneous matching of individuals to air
quality data, we also tried for a finer spatial resolution of air quality
data. Using a procedure discussed in Chapter 2, we matched those
individuals in the 1979 Health Interview Survey who lived in Standard
Metropolitan Statistical Areas (SMSAs) to the nearest ten monitors for each
of eight pollutants—ozone, total suspended particulate matter, nitrogen
dioxide, total oxides of nitrogen, sulfur dioxide, carbon monoxide, lead
and sulfates. In addition, we knew the distance between the centroid of
the census tract in which an individual resided and the location of these
monitors. Thus, although we were free to use SMSA-wide averages to
represent air pollution exposures at any point in time, or construct our
own distance-weighted average exposures, we had the capability to match
-------
1-19
each individual in our sample to the air pollution monitor(s) nearest his
or her home and use that as a measure of exposure. The latter was the
approach used in our study.
We also added to our data base information on weather conditions
specific to the two-week period for which we had acute health data for a
particular individual, as well as averaged over the year 1979. These data
came from the National Oceanographic and Atmospheric Administration (NOAA)
and were measured at one or more weather stations in each SMSA, generally
at airports. Included in the weather data were observations on temperature
(both central tendencies and extremes), precipitation, humidity and other
measures.
Finally, we compiled a third data set, one much more heterogeneous
than the air pollution or meteorlogical data. This included a number of
potential influences on acute and chronic health status which were
unrelated to socioeconomic status, air pollution concentrations, or weather
conditions. Included in this category were: ragweed concentrations for a
subset of the metropolitan areas from which the individuals in our sample
were drawn; the probability that an individual came from a household where
piped or bottled natural gas was the predominant cooking fuel (because of
its potential for indoor air pollution); a measure of the availability of
doctors in the area in which an individual lived; an estimate of the annual
amount of paid sick leave to which an individual was entitled (because sick
leave may affect work loss); as well as a number of other unrelated but
potentially important factors.
-------
1-20
A final observation about the RFF study concerns our reliance on
medical and epidemiological experts. Every study should try to build upon
and learn from the advances made in previous studies. The present study is
no exception. We have tried to draw on the best features of previous
epidemiological, clinical and toxicological studies. These were useful in
helping identify possible explanatory variables and in constructing
hypotheses about the kinds of acute and chronic health impairments expected
to result from short-term peak exposures or longer-term exposures to ozone
and other air pollutants. Of particular assistance were a number of
doctors and/or epidemiologists who identified specific diseases and
symptoms as being likely to result from exposures to certain air
pollutants. Although they were not acting as paid consultants to our
study, their assistance was most important in the construction of the
hypotheses we tested below.
1.5 Summary and Plan of this Report
The study described in this report was an ambitious one. It used an
unusually large and detailed individual data set; combined these individual
data with air pollution and meteorological data that were as finely
resolved in time and space as the SAROAD network, the NOAA system and the
HIS permitted; included explanatory variables which, to our knowledge, were
not used before; experimented with a number of functional forms, model
specifications, and definitions or measurements of individual health
status; and reported the results of a great many sensitivity tests designed
-------
1-21
to shed light on the confidence one can place in the results reported
below.
Nevertheless, it is important to keep this study in its proper
persepctive. Because it is an epidemiological study, it suffers from some
of the problems identified earlier. For instance, although individuals
were matched to the air pollution monitors nearest their homes, we still
lacked information on their daily activity patterns. Thus, we could not
characterize precisely their exposure to ambient pollutant concentrations.
Even if we had, potential exposure to indoor pollutants would still have
complicated the interpretation of our results. Until personal monitors are
a practical solution, however, some imperfection in characterizing exposure
and dosage must be expected. Thus, our results must be interpreted
carefully in view of the imperfect exposure measures we used.
Second, in spite of the detaiLed socioeionomic data contained in the
HIS, we lacked data on certain individual characteristics or habits which
may play an important role in determining acute or chronic health status.
For instance, individuals surveyed by NCHS were asked nothing about their
exercise or sleeping habits, dietary practices, or alcohol consumption,
even though all of these may influence health in one way or another.
Similarly, while we had information on individuals' occupations and the
industries in which they are employed, it would have been useful to have
data on exposures to hazardous substances in the workplace. In some of
these instances, we constructed crude proxies for these factors. But these
proxies were second-best alternatives to more detailed information.
Throughout the study we point to respects in which the estimates could be
-------
1-22
improved by additional information. Nevertheless, we must emphasize at the
outset that we lacked perfect control measures.
Because of these (almost inevitable) limitations, as well as others
identified below, the results of this study must be carefully interpreted.
It is important when reading the following chapters to pay as much
attention to the caveats about our estimates as to the estimates
themselves. Similarly, one must pay close attention to the sensitivity of
a particular result to the form of the estimating equation, the
specification of the model, the measurement of the dependent and
independent variables, and the sample over which the results were
estimated.
No single epideiniological or other kind of study can ever come up with
the "right" answer. Thus, the results of this study must be viewed
alongside those from other epidemiological, clinical, and toxicological
studies. One can have more confidence in results emerging from a variety
of different kinds of studies than in results specific to a particular
study or class of studies.
The plan of this report is as follows. Chapter 2 discusses the data
used in this report, and is divided into five sections. The first of these
deals with the Health Interview survey, and goes into some detail about the
main survey (in particular the definitions of acute and chronic morbidity),
the smoking supplement, and the supplement on residential mobility. A
second section deals with the air pollution data used in the analysis.
Included is a discussion of the EPA air pollution monitoring data, the
completeness and quality assurance methods to used to screen these data,
-------
1-23
the procedure we used to create a measure of pollution exposure over time,
our approach to naming the pollution variables in the regressions below,
and the procedure used to match individuals in the HIS to the air pollution
monitoring sites. Two sections then follow on the meteorological and other
data employed in our analysis and the purposes for which they are used. A
final section takes up the question of sample selection bias in the method
used to create our final sample of adults and children.
Chapter 3 of the report presents the methodology used in our analysis
of acute and chronic morbidity. Included there is: a discussion of the
dependent variables used in the analysis and the way in which they were
created from the HIS data tapes; a presentation of the independent
variables used to help explain acute and chronic morbidity (and a
discussion about possible collinearity between them); and a review of the
estimating techniques employed in the analysis.
In Chapter U are presented the results of our study. Separate
sections are devoted to the analysis of acute morbidity due to all causes,
acute morbidity due specifically to respiratory disease, chronic
respiratory disease, and cardiovascular and other diseases. We also
discuss in Chapter 4 the diagnostic tests we have conducted to identify any
collinearity that may exist between independent variables.
Finally, in Chapter 5 we present some estimates of the changes in both
acute and chronic health status that might result from changes in ambient
ozone concentrations in the urban areas of the U.S. These predicted
changes might form the basis for estimates—in dollar terms, perhaps—of
-------
1-24
the benefits (or health costs) that might result from alternative National
Ambient Air Quality Standards for ozone.
-------
1-25
Footnotes
Technically, the executive order applies to the regulatory agencies
governed by 4U United States Code 3502(1), excluding those agencies
specified in 4U United States Code 3502(10).
See Schneidennan, Mantel, and Brown (1975).
•3
On sensitive populations, see Friedman (1981).
For a general introduction to epidemiology, see Lilienfeld and
Lilienfeld (1980).
For example, see Lave and Seskin (1979).
"See Whittemore and Korn (1980), for an example.
7
'See Seskin (1979), for instance.
"For an introduction to and discussion of indoor air pollution and-its
possible effects on health, see National Academy of Sciences (1981) and the
General Accounting Office (1980).
-------
1-26
References
Friedman, Robert, Sensitive Populations and Environmental Standards
(Washington, D.C.: The Conservation Foundation), 1981.
General Accounting Office, "Indoor Air Pollution: An Emerging Health
Problem," report no. CED-80-111 (1980).
Lave, Lester and Eugene Seskin, Air Pollution and Human Health (Baltimore,
Md.: The Johns Hopkins University Press), 1977.
Lilienfeld, Abraham and David Lilienfeld, Foundations of Epidemiology (New
York: Oxford University Press), 1980.
National Academy of Sciences, Indoor Pollutants (Washington, D.C.:
National Academy Press) 1981.
Schneiderman, Marvin, Nathan Mantel, and Charles Brown, "From Mouse to
Man—or How to Get from the Laboratory to Park Avenue and 59th
Street," Annals of the New York Academy of Sciences, vol. 246 (1975)
pp. 237-248.
Seskin, Eugene, "An Analysis of Some Short-Term Health Effects of Air
Pollution in the Washington, D.C. Metropolitan Area," Journal of Urban
Economics, vol. 6 (1979) pp. 275-91.
Whittemore, Alice and Edward Korn, "Asthma and Air Pollution in the Los
Angeles Area," American Journal of Public Health, vol. 70 (1980), pp.
687-96.
-------
Chapter 2
DATA
2.1 Overview
The purpose of the data used in this study is to assist in determining
the relationships, if any, between individuals' acute and chronic
morbidity and their exposure to air pollutants, while controlling for
other factors. As Chapter 1 pointed out, this is not an easy task
given imperfections in the data, problems in measuring exposure, and an
array of other potentially confounding factors. Nonetheless, we.
believe the data used in this study improve upon those used in prior
epidemiologic investigations in several important respects.
The basic approach in this study is the following. A survey of
individuals' health and socioeconomic status serves as the foundation
of the data base. With the help of the National Center for Health
Statistics we matched each of those interviewed in the 1979 Health
Interview Survey (HIS) to the nearest ten monitors for each of eight
different pollutants by assuming that each household lived at the
centroid of the census tract in which its house or apartment was
located. This was accomplished via a computer algorithm that uses
input data on the geographic coordinates of census tracts from the
Urban Atlas Files of the Bureau of Census, as well as data on the
-------
2-2
geographic coordinates of air pollution monitors from the EPA's
Directory of Air Quality Monitoring Sites. Details of the procedure
are described later in this chapter.
Because of the need to preserve confidentiality, we could not be
given either the address or census tracts of individuals in the HIS.
To get around this problem we provided the NCHS with a tape which
matched each census tract in every SMSA in the United States with the
nearest ten monitors for each of eight different pollutants. Using the
census tract identifiers on their "master" HIS data tape, NCHS matched
individuals to monitors and included this monitoring information on the
HIS tape they sent us (a tape which did not include the census tract
identifiers'). Matching weather stations and other area-specific data
was accomplished at the same time.
Finally, pollution data gathered from these stations were
summarized and combined with the individual health and socioeconomic
data. The pollution data were averaged over three periods: the
two-week period in 1979 for which data were available on each
individual's acute morbidity; the entire year 1979; and, in some cases,
the period 197^-79 (the uses of these data are explained below). A
similar procedure was used for matching summarized meteorological data
to the individuals interviewed by NCHS.
Two distinct groups of individuals were analyzed in this
study—men and women aged seventeen and above (designated "adults"),
and children and adolescents under seventeen (designated "children").
Both groups were drawn from the 1979 Health Interview Survey. This was
-------
2-3
a survey of 110,530 individuals of all ages and from all geographical
regions of the United States. In forming the "adults" data set, the
original sample of 110,530 was first cut by requiring that all adults
had been given the supplemental smoking survey, since estimation of the
effects of air pollution on respiratory and cardiovascular health was
the primary focus of the study. The smoking supplement was
administered to approximately one-third of the adults in the main
survey, or 26,271 individuals. This number was reduced further because
air pollution data were lacking for a portion of these adults
(described below). This reduced the number of adults to 14,416.
Finally, five individuals with internally inconsistent data were
deleted, bringing the number of adults to be studied to 14,4^1. Table
2-1 indicates the distribution across U.S. metropolitan areas of those
in the final RFF sample of adults.
The data set for "children" was created the same way, although no
observations were lost for lack of smoking data since the smoking
supplement was only administered to those aged seventeen or above. The
number of children used in the analysis below is 15,711. Later in this
chapter we discuss the possible bias introduced by our sample selection
procedures for adults and children.
-------
2-4
Table 2-1.
SMSA
Number
0080
0160
0200
0240
0320
0360
0400
0460
0480
0520
0600
0640
0680
C720
0760.
0800
0840
0920
0960
1000
1120
1160
1170
1200
1240
1280
1320
1440
1520
1600
1640
1680
1720
1740
1760
1800
1840
1880
1920
1960
2000
2080
Residence of Individuals in RFF's
Location of SMSA
Akron, Ohio
Albany-Schenectady-Troy, NY
Albuquerque, NM
Allentown-Bethlehem-Easton. PA
Amarillo, TX
Anaheim-Santa Ana-Garden Grove ,
Anderson, IN
Appleton-Oshkosh, WI
Ashville, NC
Atlanta, GA
Augusta, GA-SC
Austin, TX
Bakersfield, CA
Baltimore, MD
Baton Rouge , LA
Bay City, MI
Beaumont-Port Arthur-Orange , TX
Biloxi-Gulf port , MS
Binghamton, NY-PA
Birmingham, AL
Boston, MA
Bridgeport, CT
Bristol , CT
Brockton, MA
Brownsville-Harlingen-San Benito
Buffalo, NY
Canton, OH
Charleston, SC
Charlotte, NC
Chicago, IL
Cincinnati, OH-IN-KY
Cleveland, OK
Colorado Springs , CO
Columbia, MO
Columbia, SC
Columbus , GA-AL
Columbus , OH
Corpus Christi , TX
Dallas-Ft. Worth, TX
Adults Data Set
Number of
Individuals
65
82
31
69
38
CA 170
65
30
64
146
26
37
40
199
34
34
30
50
38
93
276
37
c
25
, TX 36
145
42
21
47
703
157
218
49
45
32
19
92
42
193
Davenport-Rock Island-Moline, IA-IL 40
Dayton, OH
Denver, CO
89
124
Percent
of Total
0.45
0.57
0.22
0.48
0.26
1.18
0.45
0.21
0.44
1.01
0.18
0.26
0.28
1.38
0.24
0.24
0.21
0.35
0.26
0.64
1.91
0.26
0.04
0.17
0.25
1.00
0.29
0.15
0.33
4.87
1.09
1.51
0.34
0.31
0.22
0.13
0.64
0.29
1.34
0.28
0.62
0.86
-------
2-5
SMSA
Number
2120
2160
2240
2320
2360
2400
2440
2480
2520
2600
2640
2680
2760
2840
2880
2920
2960
3000
3040
3080
3120
3160
3240
3230
3320
3360
3400
3440
3480
3520
3560
3600
3640
3680
3720
3760
3840
3880
4000
4040
4120
4160
4280
4320
Location of SMSA Number of
Individuals
Des Moines , IA
Detroit, MI
Duluth-Superior, MN-WI
El Paso, TX
Erie, PA
Eugene, OR
Evans ville, IN-KY
Fall River, MA-RI
Fargo-Moorhead , MN-ND
Fitchburg-Leominster , MA
Flint, MI
Ft. Lauderdale-Hollywood, FL
Ft. Wayne, IN
Fresno, CA
Gads den, AZ
Galveston-Texas City, TX
Gary-Hammond-East Chicago, IN
Grand Rapids , MI
Great Falls, MT
Green Bay, WI
Greensboro-Winston Salem, NC
Greensville, SC
Harrisburg, PA
Hartford, CT
Honolulu, HI
Houston, TX
Huntington-Ashland, OH-WVA-KY
Huntsville, AL
Indianapolis, IN
Jackson, MI
Jackson, MS
Jacksonville, FL
Jersey City, NJ
Johnstown, PA
Kalamazoo, MI
Kansas City, KS-MO
Knoxville, TN
Lafayette, LA
Lancaster, PA
Lansing-East Lansing, MI
Las Vegas, NV
Lawrence-Haverhill , MA-HN
Lexington, KY
Lima, OH
26
407
16
49
31
29
22
22
48
14
57
73
20
51
42
40
51
53
34
52
66
45
34
70
69
230
32
33
103
37
24
58
61
24
39
132
39
45
44
46
35
30
41
35
Percent
of Total
0.18
2.82
0.11
0.34
0.22
0.20
0.15
0.15
0.33
0.10
0.40
0.51
0.14
0.35
0.29
0.28
0.35
0.37
0.24
0.36
0.46
0.31
0.24
0.49
0.48
1.59
0.22
0.23
0.71
0.26
0.17
0.40
0.42
0.17
0.27
0.91
0.27
0.31
0.31
0.32
0.24
0.21
0.28
0.24
-------
2-6
SMSA
Number
4400
4440
4480
4520
4560
4640
4720
4760
4800
4880
4920
4960
5000
5040
5080
5120
5160
5360
5400
5^40
5480
5560
5601-08
5640
5680
5720
5760
5880
5920
5960
6000
6040
6080
6120
6160
6200
6280
6320
6400
6440
6480
6600
6640
6680
Location of SMSA
Little Rock, N. Little Rock, AR
Lorain-Elyria, OH
Los Angeles-Long Beach, CA
Louisville, IN-KY
Lowell, MA
Lynchburg, VA
Madison, WI
Manchester, NH
Mansfield, OH
McAllen-Pharr-Edinburg, TX
Memphis, TN-AR
Meridian, CT
Miami, FL
Midland, TX
Milwaukee, WI
Minneapolis-St. Paul, MN
Mobile, AL
Nashville, TN
New Bedford, MA
New Britain, CT
New Haven-West Haven, CT
New Orleans , LA
New York, NY
Newark, NJ
Newport News-Hampton, VA
Norfolk-Portsmouth, VA-NC
Norwalk, CT
Oklahoma City, OK
Omaha, NE-IA
Orlando, FL
Oxnard-Ventura, CA
Patterson-Clifton-Passaic, NJ
Pensacola, FL
Peoria, IL
Philadelphia, PA-NJ
Phoenix, AZ
Pittsburgh, PA
Pittsfield, MA
Portland, ME
Portland, OR-WA
Providence-Pawtucket-Warwick, RI
Racine, WI
Raleigh -Durham, NC
Reading, PA
Number of
Individuals
38
27
754
88
1Q
38
36
13
46
37
76
4
142
41
151
180
38
68
17
11
39
116
1,184
210
30
64
9
72
43
41
38
178
31
33
483
88
279
11
47
70
98
39
16
32
Percent
of Total
0.26
0.19
5.22
0.61
0.31
0.26
0.25
0.09
0.32
0.26
0.53
0.03
0.98
0.28
1.05
1.25
0.26
0.47
0.12
0.08
0.27
0.80
8.20
1.45
0.21
0.44
0.06
0.50
0.30
0.28
0.26
1.23
0.22
0.23
3.35
0.61
1.93
0.08
0.33
0.49
0.68
0.27
0.11
0.22
-------
2-7
SMSA
Number
6760
6800
6840
6880
6920
7040
7120
7160
7240
7280
7320
7360
7400
7480
7500
752C
7600
7680
7800
7840
8000
S040
8080
8120
8160
8200
8280
8320
8360
8400
8440
8480
8520
8560
8680
8720
8840
8880
8920
8960
9000
9040
9120
9160
Location of SMSA Number of
Individuals
Richmond, VA
Roanoke , VA
Rochester, NY
Rockford, IL
Sacramento, CA
St. Louis, MO-IL
Salinas-Monterrey, CA
Salt Lake City, UT
San Antonio, TX
San Bernardino-Riverside-Ontario, CA
San Diego , CA
San Francisco-Oakland, CA
San Jose, CA
Santa Barbara, CA
Santa Rosa, CA
Savannah, GA
Seattle-Everett, WA
Shr eve port , LA
South Bend, IN
Spokane , WA
Springfield-Chicopee-Holyoke, MA-CT
Stamford, CT
Steubensville-Weirton, OH-WVA
Stockton, CA
Syracuse, NY
T a coma, WA
Tampa-St. Petersburg, FL
Terre Haute, IN
Texarkana, TX-AR
Toledo, OH-MI
Washington, DC-VA-MD
Trenton, NJ
Tucson, AZ
Tulsa, OK
Utica-Rome, NY
Vallejo-Napa, CA
Washington, DC-MD-VA
Water bury, CT
Waterloo , IA
West Palm Beach, FL
Wheeling, OH-WVA
Witchita, KS
Wilkes Barre-Hazelton, PA
Wilmington, DE-NJ-MD
27
26
104
30
82
217
27
52
96
134
146
328
61
43
48
44
170
29
28
29
57
32
14
35
70
37
90
50
38
67
36
28
39
51
31
21
258
26
37
37
13
35
38
51
Percent
of Total
0.19
0.18
0.72
0.21
0.57
1.50
0.19
0.36
0.67
0.93
1.0
2.27
0.42
0.30
- 0.33
0.31
1.18
0.20
0.19
0.20
0.40
0.22
0.10
0.24
0.49
0.26
0.62
0.35
0.26
0.46
0.25
0.19
0.27
0.35
0.22
0.15
1.79
0.18
0.26
0.26
0.09
0.24
0.26
0.35
-------
2-8
SMSA
Number
9200
9240
9280
9320
Location of SMSA
Wilmington, NC
Worcester, MA
York, PA
Youngstown-Warren , OH
Number of
Individuals
41
29
39
58
Percent
of Total
0.28
0.20
0.17
0.40
-------
2-9
2.2 The Health Interview Survey
2.2.1 Main Survey
The 1979 Health Interview Survey was administered by the National
Center for Health Statistics (NCHS) of the Public Health Service of the
U.S. Department of Health and Human Services. The following general
description may be useful:
The National Health Interview Survey utilizes a
questionnaire that obtains information on personal
and demographic characteristics, illnesses,
injuries, impairments, chronic conditions, and
other health topics. The population covered by the
sample for the National Health Interview Survey is
the civilian, noninstitutionalized population of
the United States living at the time of the
interview. The sample does not include members of
the Armed Forces or U.S. nationals living in
foreign countries.
The sampling plan of the survey follows a
multistage probability design which permits a
continuous sampling of the civilian
noninstitutionalized population of the United
States. The sample is designed in such a way that
the sample of households interviewed each week is
representative of the target population and that
weekly samples are additive over time. This
feature of the design permits both continuous
measurement of characteristics of samples and more
detailed analysis of less common characteristics
and smaller categories of health-related items.
The continuous collection has administrative and
operational advantages as well as technical assets
since it permits fieldwork to be handled with an
experienced, stable staff.
The overall sample was designed so that tabulations
can be provided for each 'of the four major
geographic regions and for selected places of
residence in the United States.
The first stage of the sample design consists of
drawing a sample of 376 primary sampling units
(PSU's) from approximately 1,900 geographically
-------
2-10
defined PSU's. A PSU consists of a county, a small
group of contiguous counties, or a standard
metropolitan statistical area. The PSU's
collectively cover the fifty states and the
District of Columbia.
With no loss in general understanding, the
remaining stages can be combined and treated in
this discussion as the ultimate stage. Within
PSU's, then, ultimate stage units called segments
are defined in such a manner that each segment
contains an expected six households. Three general
types of segments are used.
- Area segments which are defined
geographically
- List segments, using 1970 census
registers as the frame
- Permit segments, using updated lists
of building permits issued in sample
PSU's since 1970
Census address listings were used for all areas of
the country where addresses -were well defined and
could be used to locate housing units. In general .
the list frame included the larger urban areas of
the United States from which about two-thirds of
the NHIS sample was selected.
The usual NHIS sample consists of approximately
12,000 segments containing about 50,000 assigned
households, of which 9»000 Were vacant, demolished,
or occupied by persons not in the scope of the
survey. The U2,000 eligible occupied households
yield a probability sample of about 111,000
persons.
In 1979, there was a total noninterview rate of approximately 3.9
percent, consisting of individuals refusing to respond or who could not
be located at home after repeated calls.
-------
2-11
This is a cursory description of the HIS; the interested reader
is referred to more detailed information available elsewhere.
The HIS provides a rich description of the socioecononiic
characteristics of those surveyed. Each year information is sought
about individuals' health status and household and family
characteristics, educational attainment, income, occupation, number of
hours worked, and many other subjects of potential interest to
researchers. Furthermore, the 1979 HIS contained detailed supplemental
surveys on smoking behavior and residential history, each going to
one-third of the adults surveyed (the selection of subsamples produced
considerable overlap between those who received the smoking survey and
those who received the residential migration survey.) Appendix A to
this report presents the 1979 Health Interview Survey in its entirety,
as well as the smoking and residential history supplements.
It is for the provision of detailed information about
individuals' health status, however, that the HIS was selected as the
foundation of our data base. It is designed to provide a detailed
picture of interviewees' health status at the time of the interview,
rather than broader and more general information. Because the data are
self reported, response biases are possible with respect to health and
other types of information that could confound estimation results.
Further, because the HIS was not designed solely for the purposes of
epidemiologic research, but rather to provide a detailed picture of
health in the U.S., it does not elicit certain information one would
collect in a survey designed explicitly for epidemiological purposes.
-------
2-12
Because of the size of the HIS sample, however, and the details
available on individuals' health and other characteristics, the HIS is
a very rich data base.
As discussed above, it is difficult to estimate the determinants
of acute and chronic morbidity. In fact it is hard to define sickness
or health, much less summarize it by a quantitative or qualitative
index that can be utilized in statistical estimation.5 The problem is
compounded when using the HIS because health measures are self-reported
rather than physician-diagnosed. We now turn to a discussion of the
health status measurements available in the HIS and utilized in the
present study.
It is not uncommon in health surveys to ask individuals if they
consider their health to be excellent, good, fair, or poor. The "EGFP"
question, as it has come to be known, is included in the HIS. The
response to such a question conveys too little objective information
about health, however, and two individuals in equivalent states of
health might answer differently for totally unrelated reasons. Indeed,
the same individual might answer the question differently—without any
change in objective health status—at two different times during the
same day.°
The HIS, however, goes much further in eliciting health
information. Survey questions are aimed at ascertaining not only
whether the respondent did or did not have any of a detailed set of
chronic or acute conditions or impairments defined below, but also the
severity of such conditions. Severity is quantified in a variety of
-------
2-13
ways, including time since onset of condition, duration of activity
limitation, and degree of physical disability due to limitations. The
following describes the measures of health status compiled in that
survey.
Terms Relating _t£ Conditions
Condition—A morbidity condition, or simply a condition, is
any entry on the questionnaire that describes a departure from a
state of physical or mental well-being. It results from a
positive response to one of a series of "medical-disability
impact" or "illness-recall" questions. In the coding and
tabulating process, conditions are selected or classified
according to a number of different criteria (such as whether they
were medically attended, whether they resulted in disability, or
whether they were acute or chronic) or according to the type of
disease, injury, impairment, or symptom reported.
Conditions except impairments are classified by type
according to the ninth revision of the International
Classification of Diseases, with certain modifications adopted to
make the code more suitable for a household interview survey.
Chronic Condition—A condition is considered chronic if (1)
the condition is described by the respondent as having been first
noticed more than three months before the week of the interview,
or (2) it is one of the following conditions always classified as
chronic regardless of the onset.
Tuberculosis
Neoplasms (benign and malignant)
Diseases of the thyroid gland
Diabetes
Gout
Psychoses and certain other diseases of
the central nervous system
Multiple sclerosis and certain other
diseases of the central nervous system
Certain diseases of the eye
Certain diseases of the circulatory
system (includes rheumatic fever,
hypertension, stroke, and all heart
conditions)
Emphysema, asthma, hay fever, and
bronchiectasis
-------
2-m
Ulcers and certain other diseases of the
esophagus, stomach, and duodenum
Hernia of abdominal cavity (includes
rupture)
Gastroenteritis and colitis (with
exceptions)
Calculus of kidney, ureter, and other
parts of the urinary system
Diseases of the prostate
Chronic cystic diseases of the breast
Eczema and certain other dermatitis
Arthritis and rheumatism
Cyst of the bone (except jaw)
All congenital anomalies
Impairments—Impairments are chronic or permanent defects,
usually static in nature, that result from disease, injury, or
congenital malformation. They represent decrease or loss of
ability to perform various functions, particularly those of the
musculoskeletal system and the sense organs.
Onset of_ condition—A condition is considered to have had
its onset when it was first noticed. This could be the time the
person first felt sick or became injured, or it could be the time
when the person or family was first told by a physician that the
person had a condition of which he or she had been previously
unaware.
Terms_ Relating to Disability
Disability—Disability is the general term used to describe
any temporary or long-term reduction of a person's activity as a
result of an acute or chronic condition.
Disability day—Short-term disability days are classified
according to whether they are days of restricted activity, bed
days, hospital days, work-loss days, or school-loss days. All
hospital days are, by definition, days of bed disability; all
days of bed disability are, by definition, days of restricted
activity. The converse form of these statements is, of course,
not true. Days lost from work and days lost from school are
special terms that apply to the working and school-age
populations only but these too are days of restricted activity.
Hence, "days of restricted activity" is the most inclusive term
used to describe disability days.
Restricted-activity day—A day of restricted activity is one
on which a person cuts down on his or her usual activities for
the whole of that day because of an illness or an injury. The
-------
2-15
term "usual activities" for any day means the things that a
person would ordinarily do on that day. For children under
school age, usual activities depend on whatever the usual pattern
is for the child's day, which will in turn be affected by the age
of the child, weather conditions, and so forth. For retired or
elderly persons, usual activities might consist of almost no
activity, but cutting down on even a small amount for as much as
a day would constitute restricted activity. On Sundays or
holidays, usual activities are the things the person usually does
on such days—going to church, playing golf, visiting friends or
relatives, or staying home and listening to the radio, reading,
looking at television, and so forth. Persons who have
permanently reduced their usual activities because of a chronic
condition might not report any restricted-activity days during a
two-week period. Therefore, absence of restricted-activity days
does not imply normal health.
Restricted activity does not imply complete inactivity, but
it does imply only the minimum of usual activities. A special
nap for an hour after lunch does not constitute cutting down on
usual activities, nor does the elimination of a heavy chore such
as cleaning ashes out of the furnace or hanging the wash. If a
farmer or housewife carries on only the minimum of the day's
chores, however, this is a day of restricted activity.
A day spent in bed or a day home from work or school because
of illness or injury is, of course, a restricted-activity day.
Bed-disability day—A day of (bed)-disability is one on
which a person stays in bed for all or most of the day because of
a specific illness or injury. All or most of the day is defined
as more than half of the daylight hours. All hospital days for
inpatients are considered to be days of bed disability even if
the patient was not actually in bed at the hospital.
Work-loss day—A day lost from work is a day on which a
person did not work at his job or business for at least half of
his normal workday because of a specific illness or injury. The
number of days lost from work is determined only for persons
seventeen years of age and over who reported that at any time
during the two-week period covered by the interview they either
worked at or had a job or business.
School-loss day—A day lost from school is a normal school
day on which a child did not attend school because of a specific
illness or injury. The number of days lost from school is
determined only for children six through sixteen years of age.
-------
2-16
2.2.2 Smoking Supplement
In its administration of the annual Health Interview Survey, the
NCHS includes what are termed rotating supplemental surveys which are
administered at less than an annual frequency. One such survey of
relevance to the present study is the smoking supplemental survey
included as a part of the 1979 HIS. Due to the nature of our
study—where analysis of the effects of pollutants on individuals'
health must necessarily be concentrated largely on respiratory
conditions—the availability of detailed information on individuals'
smoking histories is a necessity if the effects of air pollution on
health are to be properly assessed.
Detailed data are available from the smoking supplement.
Individuals' responses to questions about the following were recorded
in the 1979.survey:
1. Smoking status (Never, Former, Occasional, Current)
2. Current smoking status (Occasional Smokers in 1)
3. Number of cigarettes now smoked a day
4. Age started smoking regularly
5. Number of cigarettes smoked a day at peak period
6. Last smoked regularly
7. Interval since last smoked regularly (Former smokers in 1)
8. Number of brands smoked
9. Type of filter
10. Kind of-cigarette smoked (plain, menthol)
11. Package type
-------
2-17
12. Cigarette size
13. Ever seriously attempted to quit
14. Total number of serious attempts to quit
15. Number of serious attempts in past twelve months
16. Starting time of last attempt
17. Length of time stayed off cigarettes
18. Tar level of brand smoked most
19. Nicotine level of brand smoked most
In 1979, 26,271 adults (aged seventeen and above) were surveyed
about their smoking habits, this representing approximately one-third
of all adults interviewed in the main survey. A facsimile of the 1979
smoking survey is included with the facsimile of the main survey in
Appendix A to this report.
*
One of the drawbacks of the smoking supplement is that typically
only one adult in each household (or two adults if there were four or
more adults in residence) was surveyed about his or her smoking
behavior. This makes it difficult to estimate the effects of
sidestream smoking on the health of children and other resident adults.
In assembling the data used in the present study, with the exception of
the data on children, we have deleted those individuals for whom no
smoking data were collected. In the estimation phase of the study,
this presents some problems. For example, if a respondent reports him-
or herself to be a nonsmoker, there exists the possibility that the
individual is exposed to sidestream smoke in the house, apart from any
-------
2-18
sidestream smoke experienced while out of the house. Trie Surgeon
General of the United States has reported that:
Children of smoking parents had more bronchitis and
pneumonia during the first year of life; and acute
respiratory disease accounted for a higher number
of restricted activity days (1.1 days) and bed
disability days (0.8 day) in children whose
families smoked than in those whose families did
not. A reduction in exercise tolerance with
exposure to sidestream cigarette smoke has been
demonstrated in patients with angina pectoris, and
a decrease in small airway function of the lung
equivalent to that observed in light smokers (one
to ten cigarettes a day) has been reported in
adults who never smoked themselves nor lived with
smokers, but who were exposed to cigarette smoking
in the workplace.0
The results of the smoking data should be interpreted carefully
since there is evidence of an increased underreporting of cigarette
consumption over time. One researcher has concluded:
Data from four major surveys, spanning the years
since the Surgeon General's Report, suggest a
significant reduction in rates of cigarette
smoking. These data, however, conflict with
production and sales data which record only a
slight reduction. Explanations for this
discrepancy range from problems in the surveys'
methodology to increased underreporting of
cigarette consumption because of both a growing
awareness of the threat to health and the social
undesirability of smoking.'
Smoking has become more stigmatized, which may account for
whatever response bias exists. Future research should control for the
possible errors-in-variables problem with the quantitative or
qualitative smoking measures used in analysis.
-------
2-19
Table 2-2 summarizes the smoking habits of all individuals
receiving the 1979 HIS smoking supplement as well as the
characteristics of the adults used in the present study. The latter
are members of the former for whom air pollution data were available.
It is clear that the smoking habits of our subsample are virtually
identical to those of the universe of individuals answering the smoking
supplement.
2.2.3 Residential Mobility Supplement
As part of the 1979 Health Interview Survey, a subset of adults
was surveyed about their residential histories. They were asked about
how long they had been at their present address, how many times they
had moved in the past three years, how far they had moved, and when
they had moved. Data were summarized for 25,519 individuals, most of
whom were also given the smoking supplemental survey.
The reason for our interest in information about individuals'
residential histories centers on the analysis of the possible
relationship between air pollution and chronic morbidity. Previous
investigations of this relationship have proceeded by examining
associations between chronic disease and current air pollution
concentrations in the area where respondents currently live. Current
air pollution levels are taken as representative of lifetime exposures.
-------
2-20
Table 2.2. Comparison of Smoking Habits of the General
Population (as Sampled in the HIS) and Those in the
HFF Subsample
Main Survey RFF Subsample
(n = 26,271) (n = 14,441)
Smoking Status
% Never i»3.1 41.9
» Occasional 2.2 2.1
% Former 18.0 18.3
% Current 30.5 31.6
% Unknown 6.2 6.2
Number of Cigarettes
Per Pay, Present
Smokers, Median 20 20
Age Started Smoking
Regularly, Median 17 17
Tar Level of_ Brand
Smoked Most (mgs),
Median 16.9 16.9
Nicotine Level of_
Brand Smoked Most
(mgs), Median 1.05 1.05
-------
2-21
But on average individuals move every five years and thus live in a
nunber of places in the course of their lifetimes. Thus, a chronic
illness that may have been initiated and/or promoted by poor air
quality in one area might be falsely associated with air quality in the
area where the individual is living at the time of the study. Data on
residential histories can be helpful in avoiding this potential source
of spurious correlation.
In some of the analysis below, we used information from the
residential migration survey to identify individuals who had lived in
the same area over the course of the exposure periods we were
interested in, typically five to six years. Using summarized air
pollution data from these periods, we tested hypotheses about the
health status of individuals living in geographical proximity to these
levels—rather than to pollution in some other geographical area.
2.2.U Documentation and Cross Referencing of HIS Data
Appendix B to this report contains a listing of the variables
used in our analysis of acute and chronic morbidity that originate in
the HIS. This list cross-references the variable of interest, its
location on the HIS Public Use Data Tape (available from the National
Center for Health Statistics), and the corresponding question number on
the HIS itself.
-------
2-22
2.3 Air Pollution Data
2.3.1 SAROAD System
The pollution data used in the present study came from two
sources. The first was the Environmental Protection Agency's Storage
and Retrieval of Aerometric Data (SAROAD) system from which we obtained
every hourly or 24-hour reading from every air pollution monitor in the
U.S. for the year 1979. The second source was the National Aerometric
Data Branch (NADB) time series data, also maintained by EPA. The
following is a description of the SAROAD system as of 1979:
In accordance with requirements of the Clean Air
Act and U.S. Environmental Protection Agency (EPA)
regulations for State Implementation Plans (SIPs),
ambient air quality data resulting from air
monitoring operations of state, local, and Federal
networks must be reported each calendar quarter to
the EPA. The EPA Storage and Retrieval of
Aerometric Data (SAROAD) format is the established
medium for transmittal of air data to EPA Regional
Offices within 45 days after the end of each
reporting period. EPA Regional Offices must,
within an additional 30 days, forward data they
have received to the EPA Aeormetric and Emissions
Report System (AEROS), of which the SAROAD system
is an operational party. AEROS is managed by the
National Air Data Branch (NADB), Monitoring and
Data Analysis Division, Office of Air Quality
Planning and Standards.1^
-------
2-23
In 1979, the SAROAD files contained data on approximately 12,000
monitoring sites, 4,000 of which were then operational. The codes
identifying each site are uniquely defined by five criteria:
(i) State Code
(ii) Area Code (city, or if not in city, county)
(111) Site Code (001-999)
(iv) Agency Code
A-E ... EPA Headquarters Groups
F ... State Agency
G ... County Agency
H ... City Agency
I ... District Agency
J ... Private
K ... Institution
L ... Military
M ... International Agency
N ... Other Federal Nonmilitary Agency
P ... EPA Regional Group
Q ... World Meteorological Organization
R ... World Health Organization
Z ... Other
(v) Project Classification Code
01 ... Population-oriented
02 ... Source-oriented
03 ... Background
04 ... Complaint investigation
05 ... Special studies
06 ... Episode monitoring
08 ... Global surveillance
09 ... Duplicate sampling
10 ... Continuous Air Monitoring Program
Thus, for example, a site identifier of 310660004F01 indicates that the
monitoring site is located in New Jersey (31), specifically in
Burlington County (0660), its number is 004, it is operated by a state
agency (F), and the monitor is population-oriented (01). An identifier
of 482890001F03 indicates a location in Virginia (48), specifically
Shenandoah National Park in Page County (2890), at site 001, operated
by a state agency (F), for background surveillance (03).
-------
2-2U
Other considerations are the type of sampler or monitor used in
the data collection procedure and the frequency with which the data are
collected (or, if continuous, are summarized). Insofar as collection
methods are concerned, the number of different types of monitors varies
with the pollutant in question. In 1979 suspended particulate monitors
were all of the same type (HI-VOL Gravimetric), for example, whereas
there were twelve types of sulfur dioxide monitors, eleven types of
nitrogen dioxide monitors, and three kinds of carbon monoxide monitors.
Two different monitoring methods were used for ozone: instrumental
chemiluminescence and instrumental ultraviolet, with approximately
twice as many monitoring sites using the former as the latter. Both
methods produce data at one-hour frequencies, whereas for some of the
other pollutants, the data frequency varies with the monitoring process
being utilized.
2.3.2 Data Integrity and Completeness
Three basic statistics were used in our study to summarize
individual pollution exposure over the two-week period for which we had
acute health data and all were derived directly from the SAROAD raw
data. These were the average of the fourteen highest daily one-hour
readings for a two-week reference period; the highest single hourly
reading over the reference period; and the average hourly concentration
over the entire two-week period. Two measures of the completeness of
the data at each monitor were also included in the air pollution data
we matched to individuals. The first was the number of
-------
2-25
days during a two-week reference period for which there was at least
one hourly or 24-hour reading (bounded between 0 and 14). The second
was the total number of hourly observations available from a monitor
during a reference week (336 maximum).
Tables 2-3 and 2-4 compare for two randomly chosen monitors the
annual data derived from our method of summarizing pollutant
concentrations by two-week reference periods. Our summary data compare
exactly with the monitoring data summarized for each site in the NADB
"Quick Look" Report (included following the monitoring data for each of
the two sites). The annual summary data appear in the last row of
Tables 2-3 and 2-4.
In addition to our concern about data completeness, we also
analyzed the quality of the data in an effort to identify miscoded or
otherwise inaccurate data. Because ozone is the focus of this study,
the ozone data were subjected to the most detailed analysis.
Nevertheless, all the air pollution data were screened using one or
more outlier tests. Air quality data are usually screened by federal,
state, or local reporting agencies prior to submittal of the data to
the SAROAD system. Thus, it is not surprising that the ozone data
exhibited reasonable consistency. Given more time the screening tests
applied to the ozone data could be applied to any of the other
-------
2-26
•e 3
c c
^ ^
VI <
C i
. -
~ —
a
«
£
~
i
^
;•
1
-
*,
J1
r
-^ *»
A j»
3 *
C.
IT ^
£ s
-
^
— •
-
c
^
&
Cj
d,
^
T t r- -o *x <%' r; «r rj — —
~ i .•*! r-j ri "^ "i «^! « « ^*i
i
u*
o ' c% M »r CN r: ?3 ry >. r^
o i CM r4 rj »*i rr ro f*5 r^ f*J
i
w i
— I ^ •< r; -c n n o
Si — ^ — XO^O"^S =
^ICCCOOCOCC
£ :
1
1
»««cs«r««ein* —
«• r-i r; — — M —
i
IK-!rj9.-c
-• n rg -. — rj
:15iili';£
~ i
t •
•« rg •: ^ .-. -c o. c >
•? ^ rs — • o s it" r» *• o* ^ *^ ii
^COfjCSC^^ CK*«^^9hO^ Ow
CM «• M rj f4 P-< fM n n — r>
*^"*^U")G~'**0I»O^G^ "J
rt'*^rjr4r^^'*i'*^^r*5r*i1^^r
o ^ o o v o cs o o o o o o
oooeeeooooees
o ^~ .- s c c r>» -c ;•> :: >.: c -c
C %fv rv ^ j*; *N a c r-* r; o^ ^ v.
ooeooooo — — o .— —
<»
rvo^-.csrj"*c— «c ^^"'C:
o*"O^. SO*-- rv*c< v^ ic x *r
oocoecoceceoc
oin-.esiftrjo-«c-!Ofvmo
*• r* *• r* t*w «• rg TX ^
— t
cc^^_^c.cc<:<: = =
r = cc«««zsi=-:-:
1 1 ; 1 ; ; ! , : 1 ! 1 !
u-5!"<3«-crg&-<":e'v«r — s
— - M — rj n «. r.1 r.
«<:c<£r^i.4.i«ce — r.
-------
2-27
c
fj
")
•)
.1
-1
.t-
111
u in
t. o
r o
o -
o
o
— o
•v 03
(N CM
•o
•c
Cl
O o-
— a
c
[••;
; OP
IT i.'
1 -
•s. -
> •»
m "5
OS
r*> *r * v
o o o o
r-i
a
c
o
c o c e o
in L~ -o < T
s — rj r-j o
— — — — CN
O O
s in
*o «n
0
e
o
c
o
0
co«ooc*oc ~
o o u- ® © * f> in e c
O.O-c'^r^-COO'r^'
ooo'Oeo — — o o
a
~v
O
SB
^
O
o — — —
o rv
— —-co-
o- -c >c •<.
o c o o
in
~a
o
ri
••^
in
< ro
— ri
5 E
« «
c
c
«• — CD
in
(N
M
C
LU
I
PJ
i. i.
^. Ui
tn CR
u u u u c
» — o in CM
n — CM
V.
r.
-------
2-28
c -. tn
u S 1 -O -C -0 -O
O 3 "? l *^ ^ "^ M
~~ ~ Si 1
1
1
j
I
*. !
8 Ut t CO C4 ^ &»
£• U"5 1 OJ O" f* fl
3 O 1 « *« — -.
00 1
«• * 1
1
1
i
1
t
t
1
1
1
31 ~ i
«M C 1
•. •*• 1
5 w 1
C *5 1 "-O f*) M -C
S *• 1 ^ *V FJ C»»
-> 1 — *« ~ O
z c i c c o c
ft S !
"" S i
W :
!
!
i
i #
5 i
u ;
C !
z ' : * o c e
C i If! V. O O
— i — 1 ^ V » -5
in T O O O O
— a
wee .
•«• '
• ,
J
I
— «t T i
3 m A i
s/ (y i
a _ £ i
=. tt Z i
e i
1
!
I
t
1
1
7 1
— e i
o r i « > n >
e. e • > r^ ^ o*
c " i o c b c
& s
z: >
i
t
i
i
IT fvi O- -O
! (M —
;
I
1 S> U U U
a, i e uJ uj uj
-> ' Z ,=, C. C.
i t
Ail t t l
l
r-i > -c PI
— «•• r-j
^ ^ ^ u
t = e e a.
t. s x s: s.
u & :
«j b
^» 3
1. ' -o rv 3 o.
C : ^- «• «r v
"O -c t T
fi -I ! c '
i i
i !
1 !
1 1
1 1
f-i t*l 1 » 1
a n < r*. i
" (N ' CC 1
1 -0 1
1 1
1 1
t I
1 1
! 1
. 1 1
i 1
I
' f
1 !
1 1
1 f
1 1
1 1
ffl — l O l
•• •« i rv i
— — 1 CM I
o o • o i
1 !
i 1
] 1
I ,
1 1
1
1
1 1
1 1
1 1
1 1
; l
I 1
0010!
^ ^ ! ^ I
**5 ••* t O i
O O i (N i
; i
, '
! i
1 !
! 1
J !
I i
I ;
, f
l i
1 ;
; i
1
; i
T :
i i
; 1
1 1
i t
i '•
f. CM i e i
O «r i rt •
IN M • '.'. '.
0 C : 0 i
i i
'. '
i l
l i
l t
l t
1 i
not i
ts n \ i
' !
1 !
U U 1 '
UJ -. 1 i
OH, 1
i u i
l lie;:
1 — !
O Px 1 -i 1
•• •• i ml
l «• i
! W I
U U I fC
UJ LL' i ->
s. e. i in ;
i re :
i ~ i
C i
b~ mi c i
in a
< u»
in
in in
— ui
O X
M <
5 ""
S "*
S Q
in *•
(M in
m in
-t <
— > =
3 O
Z X
UJ
>•
O
a
UJ
u
*r
3E
Z N
u
u
MJ
e
5
w
- a
S
u
a
IM
in
e
5
i
in
e
&
|
o
u
in
O
u
O^^O^OF1^
»»>O"»«>O>9'»
h»r^b*r^r^r«r^*F«r^
_ = |