EPA 600/5-74-017
July 1974
Socioeconomic Environmental Studies Series
Outpatient Medical Costs Related to Air
Pollution in the Portland, Oregon Area
Office of Research and Development
U.S. Environmental Protection Agency
Washington, D.C. 20460
-------
RESEARCH REPORTING SERIES
Research reports of the Office of Research and Development, Environmental
Protection Agency, have been grouped into five series. These five broad
categories were established to facilitate further development and appli-
cation of environmental technology. Elimination of traditional grouping
was consciously planned to foster technology transfer and a maximum inter-
face in related fields. The five series are:
1. Environmental Health Effects Research
2. Environmental Protection Technology
3. Ecological Research
4. Environmental Monitoring
5. Socioeconomic Environmental Studies
This report has been assigned to the SOCIOECONOMIC ENVIRONMENTAL STUDIES
series. This series includes research on environmental management,
economic analysis, ecological impacts, comprehensive planning and fore-
casting and analysis methodologies. Included are tools for determining
varying impacts of alternative policies, analyses of environmental plan-
ning techniques at the regional, state and local levels, and approaches
to measuring environmental quality perceptions, as well as analysis of
ecological and economic impacts of environmental protection measures.
Such topics as urban form, industrial mix, growth policies, control and
organizational structure are discussed in terms of optimal environmental
performance. These interdisciplinary studies and systems analyses are
presented in forms varying from quantitative relational analyses to manage-
ment and policy-oriented reports.
EPA REVIEW NOTICE
This report has been reviewed by the Office of Research and Development,
EPA, and approved for publication. Approval does not signify that the
contents necessarily reflect the views and policies of the Environmental
Protection Agency, nor does mention of trade names or commercial products
constitute endorsement or recommendation for use.
For sale by the Superintendent ol Documents, U.S. Government Printing Office, Washington, D.C. 20402 • Price $2
-------
EPA-600/5-74-01
July 1974
OUTPATIENT MEDICAL COSTS RELATED TO AIR
POLLUTION IN THE PORTLAND, OREGON AREA
by
John A. Jaksch
Herbert H. Stoevener
Department of Agricultural Economics
Oregon State University
Corvallis, Oregon
Contract No. 68-01-0423
Project 07AAC-02
Program Element 1H1094
Project Officer
Thomas Waddell
Washington Environmental Research Center
United States Environmental Protection Agency
Washington, D.C. 20460
Prepared for
Office of Research and Development
Washington Environmental Research Center
United States Environmental Protection Agency
Washington, D.C. 20460
-------
ABSTRACT
This study has attempted to quantify in monetary terms the effects of air
pollution on the consumption of outpatient medical services. The hypotheses
were that air pollution can aggravate a state of health resulting in increased
consumption of outpatient medical services and in the number of contacts with
the medical system for certain respiratory, cardiovascular, and other diseases
aggravated by air pollution.
The study period was 1969-1970, and centered in the Portland, Oregon area.
Statistical models were formulated, explaining individual outpatient consump-
tion of medical services. Measures of suspended particulate air pollution
and meteorological conditions, as well as socioeconomic-demographic variables
thought to influence the consumption of medical services, were included in the
models as explanatory variables.
The statistical results indicated that the procedures used in the study hold
promise for quantifying the medical costs of air pollution. The results did
show air pollution to have an effect on the consumption of outpatient medical
services used to treat certain respiratory diseases.
This report was submitted in fulfillment of Project Number 07AAC-02, Contract
Number 68-01-0423, by the Department of Agricultural Economics, Oregon State
University, Corvallis, Oregon, under the sponsorship of the Environmental Pro-
tection Agency. Work was completed as of October 1.
ii
-------
CONTENTS
Page
Abstract ii
List of Figures iv
List of Tables v
Acknowledgements.. vi
Chapters
I Conclusions 1
II Recommendations 2
III Introduction 4
IV Theoretical Framework 7
V A Description of the Study Area and the
Kaiser Plan 34
VI Compiling the Medical Data 42
VII Compiling the Socioeconomic-Demographic Data. 50
VIII Compiling the Air Pollution and
Meteorological Data 60
IX Statistical Results 71
X References 96
XI Appendices
Appendix A. 102
Appendix B 123
iii
-------
FIGURES
No. Page
1 Preference Map of a Well Consumer 10
2 Preference Map of an 111 Consumer Given
Air Quality a^ 12
3 Preference Map of an 111 Consumer Given
Air Quality a» 14
4 The Study Area 35
5 Isopleth Map for Portland Area 61
iv
-------
TABLES
No. Page
1 The Disease Classification System 45
2 Outpatient California Relative Value
Schedule Dollar Equivalency Fees 48
3 Household Income Response Categories
and Transformations 52
4 Transformations Performed on the
Response to the Drinking Question 55
5 Smoking Characteristics and Transforma-
tions on the Response Categories 56
6 Analysis of Variance Table for the Nine
Suspended Particulate Stations 72
7 Regression Results of Statistical Model
10: Respiratory Diseases 78
8 Regression Results of Statistical Model
10: Circulatory-Respiratory Diseases. . . 80
9 Regression Results of Statistical Model
11: Respiratory Diseases - One Day Lag. . 93
B-l Pretest Regression Results of Statistical
Model 10: Respiratory Diseases - Unlagged 123
B-2 Pretest Regression Results of Statistical
Model 10: Circulatory-Respiratory
Diseases - Unlagged 124
B-3 Pretest Regression Results of Statistical
Model 11: Respiratory Diseases - Lagged
One Day 125
-------
ACKNOWLEDGEMENT S
The authors are indebted to numerous individuals and organizations for pro-
viding data, technical advice, and other assistance for this study. Especially
valuable were the inputs provided by Dr. Merwyn Greenlick and his associates
at the Kaiser Health Services Research Center, the staffs of Columbia-Willamette
Air Pollution Authority, Department of Environmental Quality, and Southwest
Air Pollution Control Authority. At Oregon State University, Dr. Sheldon
Wagner (Environmental Health Science Center), Dr. John A. Edwards (Department
of Agricultural Economics), Dr. Richard W. Boubel (Department of Mechanical
Engineering), and Mr. James Sasser (Computer Center) made very important con-
tributions. Of course, the authors retain full responsibility for any inade-
quacies which the study may contain.
The U.S. Environmental Protection Agency supported this work financially. Mrs.
Audree Berrey typed the final manuscript.
Last, but not least, the senior author is indebted to his wife, Carol Jaksch,
for her unlimited assistance and encouragement throughout the two years of the
research study.
vi
-------
I. CONCLUSIONS
The results of this study indicate that air pollution and meteorological con-
ditions (lagged and unlagged) have an effect on the consumption of outpatient
medical services used to treat respiratory diseases.
The regression results indicate that the procedures used in this study hold
promise for quantifying the medical costs of air pollution.
As an economic cost, it would appear that air pollution has a minimal effect
on increasing the quantity of outpatient medical services consumed per contact
with the medical system. The main reason appears to be that many such visits
are routine, and while air pollution may affect the number of visits, it does
not materially affect the cost per visit.
-------
II. RECOMMENDATIONS
Air pollution appears to have a minimal effect on increasing the quantity of
outpatient medical services consumed per contact with the medical system.
is important to know, because it substantially reduces the need to do additional
research pn the effects of air pollution on outpatient medical consumption and
suggests other areas in which research on the medical costs of air pollution
may be more productive. The procedures developed in this study hold promise
I
in quantifying medical costs attributed to air pollution. Hence, it is recom-
mended that future research attempting to quantify such costs concentrate in
three general areas.
1. Any future research on the outpatient medical costs of air
pollution should investigate the effect of air pollution on the
number of contacts with the medical system. While the consump-
tion of medical services per visit may be fairly constant, the number
of contacts with the medical system may be affected by air pollution.
While a model on this was formulated and tested within this study,
there were not enough observations to derive meaningful results.
2. An effort should be made to determine the effect of air pollution
on the consumption of inpatient medical services. There should be
more variety in the types of inpatient services consumed per contact
with the medical system and, hence, more variability.
3. A research effort should be made to examine and quantify the total
medical costs attributable to air pollution; that is, outpatient and
Under support of the United States Environmental Protection Agency (Washington
Environmental Research Center, Washington, B.C., 20460), research in this area
is in progress at the Department of Agricultural Economics at Oregon State Uni-
versity. The research is expected to be completed by December 31, 1974. Also
in-house research is being conducted by the Center in this area.
-------
inpatient medical costs (so-called direct medical costs) plus indirect
2
costs (days lost from work because of air pollution, etc.).
2
The United States Environmental Protection Agency (Washington Environmental
Research Center, Washington, D.C., 20460) is in the process of negotiating a
contract to accomplish this objective. The area of study will be the St. Louis
Standard Metropolitan Statistical Area. The study is scheduled to be complete*
by June 1976.
-------
III. INTRODUCTION
The Problem
For informed and economically efficient decisions to be made in managing the
environment, adequate information must be provided to the decision-maker. One
of the main problems of air quality management is balancing the benefits that
society derives against the costs incurred from using the atmosphere for waste
disposal. To strike such a balance, the decision-maker must have quantitative
estimates of the costs and benefits associated with achieving certain environ-
mental quality levels. The provision of such estimates involves difficult
issues of measurement and evaluation.
While dependable systematic estimates of costs resulting from the effects of
air pollution are still quite rare, progress is being made. Within the past
decade several studies have been completed, estimating property and material
costs of air pollution and the effects of pollution on property values [25, 38,
47]. An important cost yet to be measured, with any high level of confidence,
is the medical costs attributable to air pollution. Considerable evidence in
the medical literature exists to support the hypothesis that air pollution
affects human health and, particularly, aggravates certain respiratory and
cardiovascular diseases [19, 68, 69]. Adverse health effects of air pollution
result in economic losses of unknown magnitudes. These losses to society
emanate from two main sources: (1) the medical costs incurred in treating the
disease, and (2) the value of the ill person's foregone production while sick.
It is the purpose of this research to estimate some of the outpatient medical
costs attributable to air pollution.
The data bases afforded this research appeared to offer an opportunity
to overcome some data deficiencies observed in earlier medical studies in-
vestigating the effects of air pollution on health. More specifically, in the,
Portland, Oregon, Standard Metropolitan Statistical Area (S.M.S.A.) air pollution
data were available from several air pollution control agencies, and were repre-
sentative of the main air pollution problem in the study area, suspended
-------
particulate matter. Adequate meteorological data were available from several
National Weather Service stations. An excellent source of individual medical
data was provided by Kaiser Foundation Health Plan. This allowed the research
effort to concentrate on the effects of air pollution on specific quantities
of outpatient medical services consumed. This is in contrast to other air
pollution-health effects studies which used a more aggregative approach (number
of doctor office visits, number of hospital admissions, etc.) in attempts to
delineate the effects of air pollution on health. Detailed socioeconomic-
demographic data were maintained with the medical data. Hence, the fact there
appeared to be available reasonably good data on a wide range of relevant phe-
nomena affecting medical consumption (air pollution, meteorological, medical,
and socioeconomic-demographic) held special promise for productive empirical
research.
Objectives
The objectives to be accomplished by this research are as follows:
1. To present an economic model of consumer choice, from which
relevant hypotheses about the consumption of medical services can
be derived.
2. To identify certain cardiovascular, respiratory, and other
diseases which can be aggravated by air pollution.
3. To estimate the effect of air pollution, if any, on the con-
sumption of outpatient medical services used to treat the diseases.
4. If there is an effect of air pollution on the consumption of
outpatient medical services, to quantify this effect in monetary
terms.
5. To determine the effect, if any, air pollution has on the
incidence rate of the diseases identified in (2) above.
The first of the above objectives is accomplished in Chapter IV of this report.
It presents the economic framework, and the statistical models and hypotheses
-------
to be tested. The study area and the Kaiser Health Plan are described in
Chapter V. Chapter VI details the compiling of the medical data and the coni-
struction of the dependent variables in the statistical models. The collection
of the socioeconomic-demographic data and the formation of the socioeconomic
variables included in the statistical models are discussed in Chapter VII. The
collection and organization of air pollution and meteorological data, and the
assigning of air pollution and meteorological values to the observations on
medical costs are explained in Chapter VIII. Chapter IX presents the results
of the analysis.
-------
IV. THEORETICAL FRAMEWORK
The Economic Model
This section begins by specifying some of the assumptions used to formulate
the economic model upon which this study is based. A graphic presentation
of the economic model is given. From the graphic specification, two impor-
tant implications are drawn which minimize the importance of medical service
prices and income in explaining the consumption of medical services for cer-
tain cardiovascular-respiratory diseases.
In order to isolate the effect of air pollution on health, account must be
taken of other variables which could influence the consumption of medical ser-
vices. A general framework is specified, which delineates some of these char-
acteristics. From the general framework, the hypotheses and the statistical
models to test them are derived.
Some General Assumptions
Assume there is a representative consumer-patient who is suffering from, or
has a tendency to become afflicted by, certain diseases associated with air
pollution. This consumer is a free agent to enter the market for medical care.
There are no physical constraints which would prohibit him from obtaining such
care.
This study is concerned only with the short-term effects of air pollution on
pre-existing (chronic) and acute cases of certain respiratory, cardiovascular,
and other diseases. These diseases are thought by the medical profession to be
aggravated by air pollution, or have been shown by previous studies to be asso-
ciated with air pollution.
Diseases of the cardiovascular and respiratory system affect two essential life
systems of the body. Many of these diseases, if left untreated, are fatal, or
inflict such pain, suffering, and mental anguish upon the victim as to result
in a decrease of life's enjoyment. It is assumed that any consumer presently
having any of the diseases, or newly contracting one of them, would be aware
-------
of these consequences. Given the choice between the alternatives, it is assumed
that the representative consumer will actively seek medical care to maintain or
improve his state of health.
It is argued that the consumer is unable to effectively evaluate medical care,
and has to rely primarily upon the physician's advice for the treatment required
to remedy the illness. Most consumers do not have the expertise to effectively
evaluate alternative bundles of medical services which could be used to treat
their disease, or to judge the quality of physician services provided to them.
There are so many variables which affect diseases from one incident to the next
that, in many instances, it is difficult even for the physician to determine
what to do.
Each disease incident will have a different degree of acuteness or severity, re-
quiring a different intensity of treatment. It is argued that there is a cer-
tain minimum amount of medical care, given the type of disease and its degree
2
of severity, which will maintain or improve the consumer's state of health,
and that this will be the minimum treatment level prescribed by the physician.
Graphic Specification of the Economic Model
A consumer's expenditures on medical care evolve from his preference map for
medical services and other goods. The shape of his indifference map will be
different, aeteris paribus, depending on whether he is ill or not. If he is
ill, the shape of the indifference curve will be affected by the severity of
Obviously, for this to hold, it is necessary to assume that life is preferable
to death and that, if given a choice, one would choose to be free of pain and
suffering. A second supposition is that the consumer expects the consumption
of medical services to alleviate his illness.
2
A consumer's state of health will change over time, and will be dependent upon
many socioeconomic-demographic characteristics as well as the type of disease
contracted. The consumption of medical services may do no more than preserve
a patient's health state, or retard the rate of change toward further deterior-
ation (terminal cancer, for example, where treatment just postpones the inevi-
table). In other instances, the consumption of medical services may cure the
patient of the disease and improve his health state.
-------
the disease state. The severity of the disease state can be influenced by the
air quality the patient is subject to, as well as the patient's socioeconomic-
demographic status. In the following, attention is focused only on the hypo-
thesized effect air pollution might have on the disease state.
Figure 1 refers to a consumer in a good state of health. Money is shown on the
vertical axis, and services as a numerative for all commodities other than medical
services. Medical services are shown on the horizontal axis, and are* expressed
as an index. The index denotes the physical quantities of all medical services
consumed to treat an illenss. Excluded from this index are items which would
he considered frills of medical care and which are not needed to treat the
patient's illness.
The usual assumptions about the indifference curves in Figure 1 hold: utility
is constant for movements along an indifference curve, a higher indifference
curve is preferred to a lower indifference curve (I111 is preferred to I"; I"
is preferred to I'), the indifference curves do not intersect, and they are
concave from above.
The slopes of these curves play an important role in the analysis. It can be
expected that the preference for the consumption of medical services is not
strong for the well consumer. He may need medical treatment of a preventive
nature (immunization shots and yearly physical checkups, for example), or anti-
cipate the consumption of some medical services in case he does get ill. Under
these circumstances, the marginal rate of substitution of medical services for
money will be declining rapidly. A point of satiation is reached, and additional
medical services become a discommodity. Points E, F, and G on I1'', I", and I',
respectively, in Figure 1 are such points of satiation.
OM in Figure 1 is defined as the consumer's permanent income. MW represents
his budget constraint, given the price for medical services. M'S would be the
optimum quantity of medical services consumed. If the consumer were to buy
health insurance, MJ would represent the premium costs. The lower unit costs
Of medical services with insurance would be reflected in JP. This budget
-------
M'S
MS
Medical
Services
Figure 1. Preference map of a well consumer.
10
-------
constraint continues as PT (with the same slope as MW) to the right of P as
the limits of the insurance policy are exceeded.
the optimum quantity of medical services consumed.
the limits of the insurance policy are exceeded. With insurance, MS would be
K1 is an important point in the decision to purchase medical insurance. If the
consumer expects that his consumption of medical services would be less than K*,
he would be better off not to purchase insurance. Up to K1, the total monetary
outlay for the same quantity of medical services would be less without insur-
ance than with insurance (premium cost MJ plus the cost of medical services with
insurance, as reflected by the price line JP).
Admittedly this framework is an oversimplification. It fails to address the
question of the consumer's feelings about uncertainty. However, it will be appar-
ent later that this issue is not central to this study.
Assume that the same consumer has become ill with a disease aggravated by air
pollution. Reference is made to Figure 2, where MS is defined as the amount
al
of medical services needed in attempts to preserve, improve, or prevent further
2
deterioration of the consumer's state of health. Except where it is noted,
everything in Figure 2 is defined as in Figure 1. Any treatment level to the
left of MS can be attributed to additional medical care beyond that needed to
al
maintain the person's health state.
QSI"', QLI", QNI', and QVI are defined as indifference curves in Figure 2. Given
the alternatives of premature death or continued pain and suffering when the
treatment is not obtained, a consumer would be willing to give up any amount of
money in order to receive the maintenance level of treatment. The marginal rate
of substitution of medical services for money rapidly declines. Hence, as with
It is assumed, for the purposes of this discussion, that the limits of the
policy are not exceeded. Thus, JP would be the relevant portion of the con-
sumer's budget constraint.
Hereafter the words "maintaining the consumer's state of health" will have the
same meaning as "preserve, improve, or prevent further deterioration of the con-
sumer's (patient's) state of health."
11
-------
MEDICAL
SERVICES
Figure 2. Preference map of an ill consumer given
air quality a .
12
-------
Figure 1, it does not take too much additional treatment for the consumer to
become satiated of medical services. This satiation level is represented by
E, F, G, and D on QS1"', QLI', and QVI, respectively.
Medical services, MS , are prescribed by the physician. The consumer will at
al
least purchase MS of medical services in his desire to maintain his state of
al
health. The relevant portion of the consumer's budget constraint, with insur-
ance, is JP, which is tangent to the indifference curve QLI" at H. The consumer
purchases slightly more medical services at R than the recommended MS . At
al
this tangency, utility is maximized and the consumer will pay, in addition to MJ
on health insurance, JU for the purchase of R medical services.
It should be noted that the optimum quantity of medical services purchased by
the uninsured patient is not greatly different from R, although the patient
will be worse off than he would be if he were insured. To see this in Figure 2,
construct a line (MW) running through M and parallel to PT. To consume the
required medical services MS without insurance would place the consumer on
al
an indifference curve well below QLI". More specifically, utility is maximized
at tangency point A on indifference curve QVI. The patient consumes B medical
services and pays MC for it.
What happens if air quality deteriorates from a. to a_? Assuming that the con-
sumer is afflicted with a disease aggravated by air pollution, one would expect
a tendency for the patient to become ill more often, and/or the severity of
each disease incident to increase. In both instances the result implies an
increase in demand for medical services. This situation is portrayed in Figure
3. All definitions that applied to Figure 2 are also applicable to Figure 3.
The consumer's entire preference map for medical services has shifted to the
right. Now MS is the quantity of medical services required to maintain the
32
patient's state of health. The consumer's relevant budget constraint is again
JP, which is tangent to QNI' at V. The consumer will pay the insurance premium
plus JZ for the consumption of Y medical services.
13
-------
£
Medical
Services
Figure 3. Preference map of an ill consumer given air quality a .
14
-------
In comparison with Figure 2, MS is less than MS (the amount of medical ser-
al 32
vices necessary to maintain the consumer's state of health has increased); JZ
is greater than JU (total monetary outlay for medical services has increased).
From a monetary standpoint the consumer is worse off. The increase in the con-
sumption of medical services resulting from the deterioration in air quality,
and the resulting increase in monetary outlay for medical services (the differ-
ence between JZ and JU), represents an economic cost which could have been fore-
gone if air pollution had not increased.
Two important implications can be drawn from Figures 2 and 3 with respect to
the price and income elasticities of demand for medical services. In Figure 3
the dashed line connecting points VAB represents a price consumption curve.
The curve is steeply, positively sloped, and shows that as the price of medical
services changes, the amount of medical services demanded changes proportionately
less: the price elasticity of demand for medical services is highly inelastic.
Also in Figure 3 the solid line connecting points VCD represents an income con-
2
sumption curve. The curve is highly positively sloped and illustrates that
the amount of medical services demanded will not change materially with changes
in the consumer's income: income elasticity of demand for medical services is
close to zero.
The demand for medical services to treat diseases affected by air pollution is
assumed to be price inelastic. This assumption follows from the consumer's
3
preference map discussed earlier. This assumption is particularly relevant
The price-consumption curve is the locus of utility maximization points
achieved when nominal money income and the price of one good are held con-
stant, while the price of the other good is allowed to vary. The locus of
equilibrium points is obtained graphically in Figure 3 by letting JX rotate
about J until a tangency point is obtained with each indifference curve.
2
The income-consumption curve is the locus of utility maximization points
achieved when the nominal money income of the consumer is allowed to vary
and the prices of the two goods are held constant. The locus of equilib-
rium points is obtained graphically in Figure 3 by drawing other lines par-
allel to JP and tangent to each indifference curve.
"The assumption of price inelastic demand is supported by empirical evidence.
Other studiesx have shown price elasticity of demand for medical services
(physician and hospital) to be highly inelastic [13, pp. 66-67, 72-76].
15
-------
to the Kaiser data. The main cost of participating in the system is the insur-
ance premium of a fixed amount. In many instances the premium is paid by the
employer. Hence, the costs to the Kaiser patient of using the system, in addi-
tion to the insurance premium if he should pay it, are his private costs of
time and transportation and some minimal charges for physician visits and drugs.
The assumption of low income elasticity of demand for medical services does not
deny the fact that certain amenities of medical care might be a function of in-
come. Kaiser's emphasis is to provide the necessary medical services to treat
an illness. Conveniences not necessary for such treatment are not provided.
For example, amenities such as the use of a private room, when a bed in a ward
would afford the same medical services, are not obtainable at Bess Kaiser Hos-
pital in Portland.
Medical treatments for diseases associated with air pollution are not close
substitutes for non-medical consumption goods; i.e., the cross-price elasticity
between medical treatment for these diseases and other commodities is zero, or
near zero. The acceptance of this assumption rests upon the nature of the dis-
eases being examined. An assumption of zero cross-price elasticity does not
negate the fact there might be treatment substitutes within a bundle of treat-
ments used to treat a disease incident. However, once a treatment method is
decided upon, a minimum amount is required and changes in the prices of other,
non-medical goods will not affect the quantity of medical services purchased.
This implies that shifts in the demand for medical care used in treating dis-
eases affected by air pollution will be minimal as the prices of other goods
change. Again, this assumption is not inconsistent with the Kaiser data.
With a prepaid medical system where, in many instances, the insurance pre-
mium is paid by someone other than the patient, the price of medical services
Empirical evidence on relationship between income and physician services con-
sumed has been found which, to some extent, supports this assumption. Ander-
son and Benham [2] found income elasticities, using permanent income and taking
other variables into account (demographic characteristics and preventive care,
among others) to be fairly low (less than 0.30). By substituting the quantity
of physician services consumed for the monetary outlay (dollar expenditure by
the consumer) on the same services, they found income elasticity to be as low
as 0.01 [2, p. 90]. The measures of elasticity differ because of free medical
care provided at low, or no cost to poorer consumers.
16
-------
and the patient's income play a minimal role in his consumption of services.
The assumptions with respect to price, income, and cross-elasticities of demand,
backed to some extent by empirical evidence from other studies, allow the follow-
ing argument to be made. The consumption of medical services in the treatment
of certain diseases is not affected by the price of the services or the patient's
income, regardless of the sources from which care was received. Hence, the use
of a prepaid system such as Kaiser's can be used to approximate the medical costs
of certain diseases associated with air pollution, and the estimates derived
will be representative of other medical systems.
Before closing the discussion on the theoretical model, it is important to empha-
size that the preference maps in Figures 1, 2, and 3 are based upon a prescribed
set of assumptions. The maps are drawn to explain why a hypothetical consumer
would seek medical care. This does not deny the fact that a person having a
disease affected by air pollution may not seek medical care and, hence, have a
preference map quite different from that described. While such occurrences can
result in social costs (through value of production foregone, etc.), no demand
for medical services is placed upon the system. These types of incidences and
costs are not accounted for within the present model.
A Broader Framework of Analysis
In order to estimate the effect of air pollution on the consumption of medical
services, account must be taken of some other conditions which could influence
the consumption of medical services and disease incidence rates. These can
be broadly grouped into socioeconomic-demographic variables and meteorological
phenomena. This section expands on the graphic presentation given above, and
examines such variables.
The Kaiser membership contains N people. On a given day (j) there is a
level of air pollution which is assumed (for illustrative purposes) to be higher
than the previous day (j-1). The effects of air pollution on health, and the
This section has been formulated with the Kaiser data base specifically in
mind. However, with some modification, the concepts illustrated have applic-
ability to other medical systems.
17
-------
hypothesized increase in demand for medical services on day j, or shortly there-
after, can show up in two ways. Fach establishes a testable hypothesis about
air pollution effects on health: (1) deterioration in air quality can increase
the severity of the disease (i), requiring additional, and perhaps more intense,
medical care beyond that which would normally be incurred; and (2) increases
in air pollution can precipitate an increase in the number of contacts with
the medical system per certain disease categories.
The general framework to be developed for testing the first hypothesis is speci-
fied in Equation (1). The other explanatory variables needed to isolate the
effect of air pollution on health are also indicated.
z«k= h(V V s«k> (1)
where I... = an index of medical services consumed (inpatient and
ijfc t-v,
outpatient) for treatment of the i— disease (per dis-
ease incident or episode) resulting from exposure to
air pollution on the j— day for the k— person;
A., = a measure of air quality on day j for the k— person;
W , = a measurement of meteorological conditions on day j
J th
for the kr1— person;
S , = the socioeconomic-demographic characteristics of the
k— person, on the j— day of exposure, for the i—
disease; and (i » 1, •••, n); (j = 1, •••, m);
(k = 1, •••, £).
Index of Medical Services - An index of medical services consumed (I) would
overcome some of the inherent disadvantages incurred by using dollar expenditures
It is recognized in the formulation of this model, and all other models to
follow, that there can be lags of several days between the onset of disease
(i) and exposure to air pollution or meteorological conditions. While the
lags are not specified within the definitions of the variables, the potential
for them to exist is understood, and is explored in the statistical analysis.
18
-------
for medical services. The index would be based on physical quantities of med-
ical services consumed, would permit aggregation of different kinds of medical
services, and would reflect the intensity of medical services used to treat
the disease state. The index, if constructed properly, would more adequately
reflect the quantity of medical services consumed than would dollar expendi-
tures on medical services, because of variations in fees charged for services
performed, and the availability of free medical care to some patients.
Air Quality - The possible effects of a deterioration in air quality (A) on the
demand for medical services are many. Exposure to most pollutants in sufficient
concentrations and for long enough time periods can result in harmful physio-
logical effects on the human body. This presentation will give a brief summary
of some of these effects.
Particulate matter can affect human health by its chemical composition and size.
The particle may be intrinsically toxic due to its innate chemical and/or phys-
ical characteristics [59, p. 141]. For example, sulfur dioxide, when absorbed
on particulate matter, can impair lung tissue [41, p. 37]. Particles smaller
than two or
[41, p. 33].
2
than two or three microns can penetrate deep into the respiratory system
A number of recent laboratory and clinical studies have led to a concern that
subtle cardiovascular and central nervous system effects may be associated with
elevated levels of carbon monoxide in the ambient air [62, p. 9-1]. Photo-
chemical air pollution can cause eye irritation [1, p. 96] and aggravate res-
piratory diseases [64,.p. 9-30]. Hydrocarbon air pollutants, which enter into
and promote the formation of photochemical smog, have been associated with eye
irritation. Aldehydes, also part of photochemical smog reactions, have been
shown to irritate the eyes and upper respiratory tract [63, p. 7-27].
Sulfur dioxide, at high enough concentrations, irritates the upper respiratory
tract [41, p. 37]. A significant correlation has been shown between mortality
For a given disease there may be several disease states, each state differen-
tiated by its relative degree of severity. The intensity of medical treatment
used in treating each state is dependent upon the severity of that state.
f\
A micron is approximately 1/25,000 of an inch [61, p. 1].
19
-------
resulting from chronic bronchitis and sulfur dioxide in England and Wales
[41, p. 69].
Meteorological - Meteorological conditions (W) can act independently of, as
well as in conjunction with, air pollution to place additional stress on the
body, and possibly increase the consumption of medical services. Temperature
and humidity have an unquestioned effect on health [8, p. 200], Extremes of
temperature can place strain on the body and nervous system; prolonged cold
spells can contribute to exhaustion [32]; cold temperatures can decrease mucous
transport which, by means of the cilia, is one of the main methods of cleansing
airborne matter from the lungs [29]. Winter, weather fronts, and fog are re-
lated to the onset and increased severity of chronic bronchitis [42, p. 11],
Exposure to cold, fog, or smog may need to be avoided by asthmatics [42, p. 64].
Low temperature, low humidity, a wide range in barometric pressure, and wind
have been shown to be associated with the daily prevalence of common cold and
cough symptoms [9, p. 524]. In another study, temperature (on the day the ill-
ness started, or a few days prior) seemed to be highly associated with upper
respiratory disease incidence [49, p. 738],
Intense radiation can lead to heat stroke; intense cold can increase pulmonary
disease [18, pp. 80-81], Many physicians believe that shortness of breath, in
those prone to heart attacks, is increased during spells of hot, muggy weather,
and that fatigue worsens in foggy weather [36, p. XIV], It was found in the
Netherlands that the highest mortality from arteriosclerotic heart disease occurs
during the coldest months of the year (January and February) [50, pp. 507-508],
Meteorological conditions can interact with air pollution to produce an environ-
mental situation which can produce adverse health effects [9], These effects are
precipitated through chemical reaction of certain pollutants with each other and
with certain environmental phenomena to produce new, and sometimes unknown, sub-
stances. These chemical reactions are affected by the presence of wind, sun,
and humidity. The best known example of these chemical reactions is the pho^o-
chemical reaction between oxides of nitrogen and other organics in the presence
of sunlight. The effects on health of the resulting photochemical air pollu-
tion, or Los Angeles smog as it is more popularly known, have already been docu-
mented.
20
-------
Certain meteorological conditions can increase the build-up of air pollutants
in the environment. Many past air pollution disasters (Meuse Valley, Belgium,
1930; Donora, Pennsylvania, 1948; London, 1952 and 1962, among others) have
been characterized by prolonged, anticyclonic high pressure with a secondary
inversion [7, p. 2], An inversion acts as a lid, and retains below it all
pollution emissions to the atmosphere, hence effectively retarding their disper-
sion. Air pollutants build up and a minor air pollution problem can become a
severe episode if the meteorological conditions under which it was incurred
continue.
Chemical reactions of pollutants with each other, the interaction of pollutants
with certain meteorological phenomena, and the resulting creation of new pol-
lutants, or substantial increases in the amounts of pollution, will usually be
reflected through adequate air monitoring. However, it is possible that meteoto-
logical conditions and air pollution may act interdependently to adversely affect
a disease state. That is, the simple air pollution or meteorological condition
by itself, and in the absence of the other, may not adversely affect a disease
state. However, in combination with each other they may adversely affect the
disease condition. Such combined effects would not be reflected by single mea-
surements of different air quality or meteorological phenomena. Hence, in pur-
suing a study of this type, one should be aware of the interdependence possibil-
ity.
Socioeconomic-Demographic - Other studies have shown that certain socioeconomic-
demographic variables (S) play an important role in the incidences of some dis-
eases affected by pollution [34, 37, 68], It would seem that, to the extent
certain variables affect the health of the individual, they should also affect
the amount of medical services consumed in treating the disease. Included
within (S) are several socioeconomic-demographic variables which would appear
to be important in explaining the consumption of medical services and the indi-
vidual's state of health.
The smoking characteristics of the patient are expected to play an important
role in his consumption of medical care. Heavy smoking can aggravate certain
21
-------
respiratory diseases (emphysema and bronchitis, for example), and place addi-
tional stress on the body's cardiovascular system.
Cessation of cigarette smoking has been shown to be associated with a decrease
in the prevalence of symptoms related to smoking (cough and shortness of breath,
for example) [65, p. 146]. Numerous other studies have shown the prevalence
of cough, sputum production, breathlessness (dyspena), and chronic bronchitis,
among other respiratory conditions, to be consistently higher for ex-smokers
when compared to nonsmokers, and lower when ex-smokers are compared to cigarette
smokers [65, pp. 195-205]. Other studies (reported on in the Surgeon-General's
report on the health consequences of smoking) reveal that the risk of develop-
ing lung cancer decreases with smoking cessation [65, p. 11]. Most studies
indicate that after ten years of not smoking cigarettes, the ex-smoker's chances
of dying a premature death due to a smoking-related disease is less than that
2
for a smoker, and slightly greater than that for a nonsmoker. The residual
effects of past smoking habits can still aggravate certain respiratory diseases.
The more recent the ex-smoker stopped smoking, the greater the potential for
consuming more medical services. The studies also indicate that there are cer-
tain long-term effects of cigarette smoking, some of which are irreversible
[65, p. 145].
The occupation of the individual could affect the consumption of medical care.
It is known that exposure to dusts can aggravate asthma [27, p. 34] and chronic
bronchitis [42, p. 13]. Hence, it would seem that dustier occupations could
increase the consumption of medical services for individuals with a respiratory
ailment.
The residential history of the patient is also important. Studies have shown
that the incidence and death rates from emphysema and cancer are higher in cities
than in the relatively non-polluted rural areas [41, pp. 71-72]. It is argued
•I ^
See particularly the following [10, 27, 41, 42] for the effects of smoking
and air pollution on certain respiratory-cardiovascular diseases.
2
Obtained from the American Cancer Society, and based on its slide series, "The
Time to Stop is Now."
22
-------
that an Individual's state of health would be better if he had lived most of
his life in a rural area or other unpolluted environment. The writers feel
that, if this were the case, when a patient does become ill his disease state
would be less chronic, and require less consumption of medical services.
The incidence rates of certain diseases are more prevalent in certain age,
race, and sex groups [10, p. 13; 42, p. 39], Hence, it is felt race, age, and
sex are three variables which could influence the demand for medical services
by an ill person. If some diseases are more predominant in certain age, race,
and sex groups, then persons within these groups having the diseases might be
more acutely ill than individuals outside the groups who might also contract
the disease. Also, with increasing age, the efficiency of the lung function
declines. This is probably due to a decrease in elasticity of lung tissue, in-
creasing inflexibility of the chest wall, and, perhaps, to impairment of the
regulation of ventilation and the efficacy of cough [10, p. 16], Such aging
lung conditions would tend to impair the clearance of pollutants from the res-
piratory tract, which could result in increased aggravation of respiratory dis-
eases and an increase in the consumption of medical services. Also, increased
aging could be one of the main causes of chronic obstructive pulmonary diseases
2
[10, p. 16], Generally, one would expect an older person to have a poorer
3
state of health when compared to a younger person.
Another social habit which might affect a patient's consumption of medical ser-
vices is his drinking habits. Excessive consumption of alcohol can affect a
person's state of health (emotionally, mentally, and physically). Alcoholism
can play a role in pulmonary tuberculosis [42, p. 32], Excessive consumption
of alcohol leaves a person's body in such a physical state that, if he did be-
come ill, the consumption of medical services could be increased. Alcoholism
4
has been extensively associated with respiratory problems.
Conversation with Sheldon Wagner, M.D., Oregon State University - Environmental
Health Unit; Good Samaritan Hospital, Corvallis, Oregon; August 25, 1971.
2
Chronic obstructive pulmonary disease is a term which applies to those patients
with chronic bronchitis, asthma, and anatomic emphysema who exhibit persistent
obstruction of bronchial flow [10, p. 7].
3Sheldon Wagner, M.D., August 25, 1971.
^Sheldon Wagner, M.D., August 25, 1971.
23
-------
It is felt that the more physically active a person is, the better his state of
health. This implies that when a person does become ill, the disease case could
be milder, which would mean fewer medical services required in the treatment of
the disease incident. There is a suspected association between lack of physical
exercise and coronary heart disease [20, p. 15].
The number of_ persons per room within the patient's home would be an indica-
tion of crowding within the patient's environment. Crowded conditions could re-
sult in increased consumption of medical care. More crowded conditions lead to
increased physical and emotional strain being placed upon the individual, and
the increased possibility of initial infection and reinfection with certain res-
piratory diseases (tuberculosis, for example) [42, p. 31.]
The income of the patient is included in an attempt to describe the consumer's
ability to make expenditures, in lieu of medical expenditures or services, pre-
serving and bettering his state of health. Examples of such expenditures would
be those made for sound housing and nutritional food. One would expect the con-
sumer to consume more medical services if he is unable to make such expenditures,
and his state of health deteriorated as a result.
The marital status of the patient could explain some consumption of medical ser-
vices. Available evidence indicates that single persons spend more total days
in the hospital than married persons [13, p. 60], This could simply be a reflec-
tion of the fact that the married person may have someone available at home to
care for him when he is ill. There may also be other reasons for a better health
r\
state of the married person compared to the single person.
The general framework specified in Equation (1) would afford a test whether air
pollution increases can cause greater utilization of medical services by increas-
ing the severity of a disease state. To determine whether deterioration in air
Sheldon Wagner, M.D., August 25, 1971.
2
For the reader seeking a more detailed presentation of these reasons, see [26,
pp. 41-42]. '
24
-------
quality can precipitate an increase in the number of contacts with the medical
system for certain diseases requires another general model, which is illustrated
by Equation (2).
niCj/NCJ = t(ICJ> V V (2>
where n. . = the number of people in geographic area c making
contact with the medical system for disease i on
day j;
N . = the total number of people living in geographic
area c on day j;
A . = an average measure of air quality over geographic
area c on day j;
W . = an average measure of meteorological conditions
over geographic area c on day j;
S = an average of socioeconomic-demographic character-
istics of people living within geographic area c
on day j; and
(c = 1, •••, p); (i = 1, •••, n); (j = 1, •••, m).
Other studies investigating the effect of air pollution on health have tested
whether air pollution precipitates an increase in the incidence of disease i.
After further Investigation, incidence was discarded as a potential dependent
variable, and the number of contacts with the medical system substituted in
its place. The reasons are as follows. Incidence is defined as an expression
of the rate with which a certain event occurs, such as the number of new cases
of a specific disease occurring during a certain time period [11, p. 730], The
problem with incidence is that the disease must be a new case before it is meat-
sured; i.e., the disease must be a new episode. Many chronic and acute respira-
tory diseases can be part of a continuing episode, and be affected by air pollu-
tion such as to precipitate additional visits and/or contacts with the medical
system. Use of the incidence would not reflect those additional contacts and,
hence, would inadequately express demands placed on the medical system as they
might be affected by air pollution. The problem is solved by using the number
of contacts with the medical system, which includes not only new but continuing
contacts for certain disease categories.
25
-------
The Independent variables in Equation (2) are expressed as a mean over a geo-
graphical area (census tract, for example). The approach outlined in (2) offers
one method of seeing whether differences in air quality, meteorological, and
socioeconomic-demographic conditions can play a role in people contracting dis-
ease i, irrespective of the amount of medical services consumed per disease
incident.
The arguments for including the explanatory variables A., W , and S in
Equation (2) are similar to the arguments given for their inclusion in Equation
(1). The principal influence of A ., W , and S in Equation (2) is that they
affect a state of health, hence making the patient more susceptible to certain
cardiovascular-respiratory problems. Therefore, A , W ., and S are expected,
cj cj cj
with some exceptions discussed below, to have the same causal relationships on
disease incidences as A., , W , 9 and S , had in the consumption of medical ser-
J K JK ij K
vices in Equation (1).
Several additional variables are included in S , which were not considered in
cj
S.,,. These variables would be expected to influence the contact rate with the
IJK.
medical system (n. ./N ,), but would not necessarily influence the consumption
of medical services (I.., ) per visit. The variables are reading of medical lit-
1JK
erature, distance to the nearest clinic or hospital, and a measure of the hypo-
chondriac tendencies of the patient to contact the medical system.
The reading of medical literature could explain differential information and
preferences toward recognizing the need for, and in appreciating the desirability
of, seeking medical care. Such reading would increase the patient's medical
awareness. This would, in many instances, Imply contact with the medical system
and the consumption of medical services. Even if medical services were not pro-
vided to the patient, contact with the system, where a disease is positively
identified, would constitute a disease incident. It is expected that the increased
awareness of the patient to medical phenomena will be positively related to dis-
ease incident, particularly with respect to patients having the cardiovascular-
respiratory problems with which this study is concerned.
Distance to the nearest hospital or clinic reflects, in part, the out-of-pocket
costs and inconvenience to the patient of obtaining medical care. All Kaiser
26
-------
clinics are located within the immediate Portland area, and the hospital is
located in North Portland. Hence, it is quite possible that patients living
in outlying areas would substitute (for Kaiser services) medical services
closer to their place of residence. Longer distances could delay the patient
from seeing the Kaiser doctor, or not seeing him as often. Such situations
would have direct bearing on the contact rate with the system. Therefore, it
is expected that the farther patients have to go in obtaining medical services
from Kaiser, the lower will be the disease incidence reflected in the Kaiser
2
data.
The medical hypochondriac has a morbid fear or anxiety about his health state.
While it is highly unlikely that such a mental state will actually affect the
marginal consumption of medical services per visit, it could affect the incident
rate because of the tendency for the patient to contact the system more often.
Hence, a measure of the patient's tendency to contact the system at the first
sign of a perceived illness should be included within the analysis.
The framework specified in Equation (2) would afford a test whether air pollu-
tion can precipitate an increase in the incidence of certain diseases. The
next section will specify the null hypotheses to be tested, and delineate the
statistical models by which they will be tested.
Specification of the Statistical Models
The first null hypothesis to be tested is as follows:
Ho..: Deterioration in air quality causes no increase
in the consumption of medical services per out-
patient contact with the medical system.
Equation (3) permits the testing of this hypothesis.
One clinic is in Vancouver and one is in Beaverton (see Figure 4 in Chapter V).
The other three clinics are located within the Portland city limits.
2
This would imply that as the out-of-pocket costs of using Kaiser facilities
increase with distance, the demand for Kaiser medical services becomes more
elastic. However, demand for medical services as a whole would still remain
inelastic because medical services from other facilities closer to the patient
could be substituted for the Kaiser services.
27
-------
Yijk - ^O + 3lXlijk + 62X2ijk + 33X3ijk
+ *6X61Jk + e7X7ijk + Vsijk + 39X9i.1k +
+ Vllijk + P12X121Jk + ei3X13ijk + PlAXM±Jk + £ijk
where Y. ., = index of consumed outpatient medical services for disease
J i resulting from contact with the Kaiser system on day j
by person k, converted to dollars.
X = age of the fc— person on day j who consumed outpatient
3 medical services for disease i.
X = sex of the k— person who consumed outpatient medical
3 services on day j for disease i (0 or 1 if the patient
is male or female, respectively).
X«. ., = marital status of the fc— person who consumed outpatient
3 medical services for disease i on day j (0 or 1 if the
patient is not married or married, respectively).
X, . ., = number of people in the fc— patient's household who con-
•* sumed outpatient medical services for disease i on day j.
J.-L
X-. ., = household income of the k— patient who consumed outpatient
3 medical services for disease i on day j.
X,. , = race (Negro) of the fc— patient who consumed outpatient
13 medical services for disease i on day j (1 if Negro, 0
otherwise).
X^ ., = race (others) of the fc— person who consumed outpatient
medical services for disease i on day j (1 if of another
non-White race, 0 otherwise).
X8iik = a 8reat deal of time and energy expended by the fc— patient,
who consumed outpatient medical services for disease i on
day j, in being physically fit (1 if a great deal of time
and energy expended, 0 otherwise).
X9iik
some t±m& and energy expended by the k— patient, who con-
sumed outpatient medical services for disease i on day j,
in being physically fit (1 if some time and energy expended,
0 otherwise) .
As will be explained in Chapter VI, in order to express the effect of air pollu-
tion on health as an economic cost, the index was converted to dollar values.
The conversions were undertaken in such a manner that, for the most part, varia-
tions in medical expenses would reflect direct variations in quantities of medi-
cal services consumed.
28
-------
X-0 . = a measure of the frequency of alcohol consumption by
the k— patient who is a heavy drinker, and consumes
outpatient medical services for disease i on day j.
^111'k = an *n<*ex expressing the cigarette smoking characteristics
of the k— patient consuming outpatient medical services
for disease i on day j.
X... , = an index of occupational exposure to job-related pollu-
tants by the fc— patient consuming outpatient medical
services for disease i on day j.
^131'k = a measure °f tne ambient air pollution that the k—
patient, who consumes outpatient medical services for
disease i, is exposed to on day j (expressed as sus-
pended particulates, micrograms per cubic meter per
day).
X.,.., = a measure of the meteorological conditions that the fc—
•* patient, who consumes outpatient medical services for
disease i, is exposed to on day j (expressed as degree
days - the absolute difference between the average daily
temperature and 65° F).
e.., = a random error term, about which the usual statistical
3 JL
assumptions are made, and
(i = 1, -•-, n); (j = 1, •••, m); (k = 1, «•-, I).
3_, 3-,, *•', 3,, are the parameters to be estimated.
The second hypothesis to be tested by this study is as follows:
Ho_: Deterioration in air quality does not precipitate an
increase in the number of contacts with the medical
system per disease category.
This hypothesis, using the framework specified in Equation (2), can not be tested
directly because of an error made while specifying the statistical model. The
independent variables were specified (per each census tract) only for Kaiser mem-
bers who contacted the system. More specifically, the number of Kaiser members
at risk for each census tract was known, and could be included within the depen-
dent variables. However, the socioeconomic-demographic characteristics were
See [12, p. 17] for a detailed presentation of these assumptions.
29
-------
known only for Kaiser members from each census tract who contacted the medical
system. That is, the remaining socioeconomic-demographic characteristics of
the population at risk were not included. Hence, a test of the model specified
in the format was meaningless. The computer and programming expense to con-
vert the error was prohibitive.
However, there did appear to be an alternative method of re-specifying the model
which provided the possibility of testing the hypothesis indirectly. As pre-
viously mentioned, the socioeconomic-demographic characteristics of the popula-
tion at risk could not be determined easily; however, the number of people at
risk per census tract was known. Hence, an alternative to Equation (2) (Equation
5) was specified. Within its dependent variable are included the number of con-
tacts per disease category which resulted in the consumption of medical services
per census tract, and the number of people at risk per census tract. The in-
direct test of the second hypothesis will become clearer with specification of
the alternative model.
The dependent variable (vic.) of Model (5) is defined by (4) as follows:
1
Jl Yic^k
Yic:i - NC. • <4>
i
where 1 Yic-k = the summation of the index of consumed outpatient
k=l -1 medical services for disease i for all k persons
in census tract c on day j, converted to dollar
values.
N. = the total number of Kaiser respondents residing in
census tract c on day j; and
(i = 1, •••, n); (c = 1, -.., p); (j = 1, ..., m); (k = i, ..., £).
This leads to the specification of the statistical model (5).
As will be discussed in Chapter IX, even if the model to test the second hypo-
thesis had been specified correctly, it could not have been tested because
there were not enough observations per census tract.
30
-------
Yicj = gO + 3lXlicj + 32X2icj + 33X3icj + 34X4icj + 35X5icj
+ 36X6icj + 37X7icj + 38X8icj + 39X9icj + 310X10icj
where X. . . = the average age of patients consuming outpatient medical
services for disease i from census tract c on day j .
X« . . = the percentage of the female sex consuming outpatient
•* medical services for disease i from census tract c on
day j.
X,. . = the percentage of married patients consuming outpatient
medical services for disease i. from census tract c on
day j.
X, . . = the average number of people per household in census tract
c for individuals consuming outpatient medical services
for disease i on day j .
the average household income in census tract c for indi-
vi duals con
i on day j.
j. . ,
•* vi duals consuming outpatient medical services for disease
X, . . = the percentage of non-white patients in census tract c
consuming outpatient medical services for disease i on
day j.
X?. . = the average frequency of alcohol consumption by heavy
drinkers in census tract c who consume outpatient medical
services for disease i on day j.
XR . = the average smoking index of patients in census tract c
C-1 who consume outpatient medical services for disease i
on day j.
Xq. . = average index of occupational exposure to job-related
1C^ pollutants for patients from census tract c consuming
outpatient medical services for disease i on day j.
Xlf). . = the average measure of meteorological conditions for
1C-' patients consuming outpatient medical services for dis-
ease i from census tract c on day j (expressed as degree
days - the absolute differences between the average daily
temperature and 65° F) .
X-- . = an average measure of ambient air pollution for patients
CJ consuming outpatient medical services for disease i from
census tract c on day j (expressed in suspended particu-
late, micrograms per cubic meter per day) .
31
-------
e . - a random error term, about which the usual statis-
3 tical assumptions are made; and
(i = 1, •••, n); (j - 1, •••, m); (c - 1, •••,?),
and £0, 3,, •••, 3,, are the parameters to be estimated. Y is as previously
defined in (4) above.
The dependent variable in (5) is expressed as the per capita index of consumed
outpatient medical services over census tract c on day j, converted to dollar
values. Three separate items can influence the dependent variable in (5):
(1) the consumption of outpatient medical services per visit (Y^.^, in (3));
(2) the number of contacts with the Kaiser system which result in medical ser-
vices being consumed; and (3) the number of people at risk per census tract.
More specifically, an increase, resulting from air pollution, in the number of
contacts consuming medical services could cause the per capita index consumed
of outpatient medical services (Y. .) to increase (the number of people at risk
and the amount of medical services consumed per visit assumed constant). Simi-
larly, an increased consumption of medical services resulting or associated with
air pollution would also increase (Y. .) (the number of people at risk and the
number of contacts with the Kaiser system assumed constant). It is highly prob-
able that both of these effects operate together.
Assuming a significant association between air pollution and the per capita index
of consumed outpatient medical services in (5), one could not determine whether
the association would be caused by an increase in the number of contacts with
the medical system or an increase in the consumption of medical services per
contact. However, the null hypothesis associated with (3) permits a test on
the consumption of outpatient medical services per visit and air pollution. If
this hypothesis should not be rejected, and there is an association between air
pollution and the per capita index in (5), then it could be Implied that there
is an association between air pollution and the number of contacts made with the
Kaiser system.
The next four chapters deal with the collection, compiling, and processing of
data for the variables in Models (3) and (5). Note that some of the socioeco-
nomic-demographic variables discussed in the section titled A Broader Framework
32
-------
of Analysis were not included in the statistical models. These variables were
excluded because responses to questions eliciting the data were not specific
enough to be included in the analysis. These variables are residential history,
reading of medical literature, medical hypochondriac tendencies, and the distance
to the nearest clinic.
See [26, pp. 111-114] for a more detailed discussion on this topic,
33
-------
V. A DESCRIPTION OF THE STUDY AREA
AND THE KAISER PLAN
The Study Area
The area of study centers within the Portland Standard Metropolitan Statistical
Area (S.M.S.A.), with a population of 1,009,129 [53, pp. 39-175]. The S.M.S.A.
is made up of Clackamas, Multnomah, and Washington Counties in Oregon, and Clark
County in Washington. The study area is enclosed in the dashed lines in Figure
4. The study area boundaries were determined by the availability of air pollu-
tion data and proximity to the Kaiser medical facilities in Portland, Beaverton,
and Vancouver. The study period was 1969 to 1970.
Portland, Oregon, population 381,927 [53, pp. 39-175], is the largest city in
the S.M.S.A. and Oregon. The city is located in the northwestern part of the
State, astride the Willamette River near its confluence with the Columbia River.
Portland is an important trade, transportation, and manufacturing center for the
Pacific Northwest. Racially, the S.M.S.A. is mostly white. The minority races
are composed mostly of Orientals and Negroes [46, p. 394],
Portland is economically diversified. A number of national companies have
branches located there [46, p. 395], Food processing and related agricultural
production, textiles, lumber and wood products including pulp and paper, chemicals,
aluminum and other metal fabricating plants, shipping [46, p. 395], and tourism
are Portland's major industries. Tourist visitation is particularly heavy in
Portland during the summer, primarily because of the city's convenient locatioa
with respect to many seasonal recreational opportunities [58] and its annual
Rose Parade held in early June.
Portland (including Vancouver) is located about 60 miles from the Pacific Ocean,
and is enclosed by two mountain ranges, the low Coast Range to the west and the
Cascade Range to the east. Each range is approximately 30 miles from the city.
The Cascade Range provides a steep, high slope for the orographic lift of moisture-
laden westerly winds and consequent heavy rainfall on the west side of the range.
The Cascade Range provides a barrier, containing the interior Columbia Basin with
34
-------
r
/V
A
^
h
to
5
^
^
TV
A
Columbia River
Scappose (x ^
\ i
OREGON
X^ancouver
^\>-^ cov*8^
/ For
, , /^X v \ . X (Troutdale
Hillsboro' X \ X \
\ Beaverton I Gresham \
\ /X....
\ / A
V Xl Milwaukie
LakesOswego
\
,1
x/ Sandy
Oregon City /
Willamette
River
^
Figure 4. The study area.
35
-------
its continental air masses. While most air flow in the Portland area usually
is northwesterly in spring and summer and southeasterly in fall and winter, there
are occasional movements of dry continental air from the east to the west through
the Cascade passes and the Columbia Gorge. When this occurs, extreme high and
low temperatures are experienced in the Portland area [58],
With the exception of these temperature extremes, Portland has a mild winter rain-
fall climate. Average annual temperature is 52.9° F. Annual rainfall averages
37.18 inches, with 88 percent of the annual total occurring between the months
of October and May. On the average, there are only five days each year with
measurable snow. Incursions of marine-tempered air are a moderating influence
to any temperature extremes; i.e., extremes of heat or cold last only a few days
before being broken [58].
The region's greatest area-wide pollution problem is suspended particulates
[4, p. 4]. Periodic inversions during the fall and winter months (September
to February) can trap these particulates for several days at a time, causing
visibility and other problems. Sulfur dioxide, photochemical smog, and nitrogen
dioxide are not presently a serious problem in the Portland area [4, pp. 3-4].
In many instances cloud cover rules out a photochemical smog problem. Only down-
town Portland has a problem with pollutant gases, mainly carbon monoxide. While
there is some heavy industry in Portland, pollution problems from these sources
are usually localized.
The meteorological and air pollution problems of the Lower Willamette Valley,
containing Lake Oswego, Milwaukie, Oregon City, Hillsboro, Beaverton, Sandy,
and the southern part of Portland, are similar to the Portland-Vancouver area.
Portland's suspended particulate problem extends southward from Portland to
the Valley. The wind flow patterns in the Valley are somewhat simpler than
they are within the Immediate Portland-Vancouver area, due primarily to the
lack of a Columbia River Gorge effect. Air flow patterns within the Valley
are primarily channeled by the Coast and Cascade Ranges. Winds are predomin-
antly from the north in summer and from the south in winter. The sea breezes
are also an important moderating force on temperatures in the Valley [43, p. 159].
36
-------
Characteristics of the Data Provided
by the Kaiser Medical System
The Kaiser System
The Kaiser-Permanente Medical Care Program is a non-profit group insurance plan
which provides comprehensive medical care to its members. Kaiser is one of the
oldest (established in 1938) and largest (2.2 million members) prepaid medical
plans in the nation [66, p. 1]. The Plan operates on a regional basis in Califor-
nia, Oregon, Hawaii, Colorado, and Ohio. It has 2,100 doctors, who represent
nearly every medical specialty [66, p. 1], The physicians are organized into
a medical group, a partnership, which contracts with the Kaiser Foundation
Health Plan to provide medical services. The monetary compensation to the
physicians is usually competitive with what could be earned in private practice.
Also, to the extent each region stays within its operating budget, the members
of the medical group receive a year-end bonus [66, p. 21],
The emphasis of the Kaiser system in providing medical services has been to
eliminate unnecessary health care and to concentrate on outpatient care. Kai-
ser's inpatient admissions, and average length of hospital stay, are well below
2
the national average.
unnecessary health care, from Kaiser's viewpoint, is providing medical services
to the patient which are of little medical value to him. For example, the use
of a private room in a hospital, when a room of multiple occupancy would serve
just as well; or dispensing to the patient services on an inpatient basis
when an outpatient visit to a clinic would accomplish just as much.
2
The national average in 1967 for hospital admissions averaged 137.9 a year per
1,000 people per year. The average length of hospitalization was 6.65 days for
Kaiser members, compared to 7.8 days for the national average [66, p. 21], As
a comparison, the Oregon Region Kaiser Foundation Health Plan's average length
of stay for hospitalization of health plan members was 5.2 days in 1970; the
average hospital admissions per 1,000 members was 87. (Information obtained
from Marilyn McCabe, Special Research Assistant, Kaiser Health Services Re-
search Center, Portland, Oregon.)
37
-------
The Oregon Region Kaiser Foundation Health Plan was established in 1943. In
1971 the plan provided prepaid, comprehensive medical care to approximately
146,000 members.1 The plan is sold mainly to organized groups.
From a socioeconomic standpoint, the membership of the Oregon Region is quite
diverse. Most occupational-socioeconomic groups are represented within the mem-
bership [15, p. 938]. It has been estimated that 15 percent of Portland-Vancouver
residents are enrolled in the plan [17, pp. 298-299], and receive most of their
medical care through Kaiser's hospital and six clinics. Kaiser also provides
medical services to non-members on a fee-for-service basis.
The partnership of physicians contracting with the Kaiser Foundation Health Plan
to provide medical services to the membership is known as the Permanente Clinic.
The Permanente Clinic is paid a negotiated fee per health plan member. The fee
is paid whether the member seeks medical care or not. The income of the Perma-
nente Clinic is then distributed to the member physicians in a manner determined
by them [15, p. 938].
The socioeconomic-demographic and medical data for this research came from the
Kaiser Health Services Research Center. Kaiser Research is a separate division
within the Kaiser system. The Center is federally funded. These monies provide
a base of funds which the Center uses to attract qualified people of different
professional backgrounds. These professionals are conducting research on ways
of improving and making more efficient the delivery of health services.
There were several advantages in using the Kaiser system for this research: (1)
the membership was large enough to provide a sufficient data base; (2) the
population base at risk for medical services was continually known; (3) the
Kaiser system provided a total range of medical services, from mental health to
drugs; and (4) the Kaiser data system was readily amenable to use by modern
data processing equipment.
T:his was the membership for the Oregon Region as of December 31, 1971. The
information was obtained from Marilyn McCabe, Kaiser Health Services Research
Center.
38
-------
The Five Percent Sample
The objectives and intent of studies conducted by Kaiser Research were to explain
the determinants of medical care utilization. To accomplish these goals a 5 per-
cent, ongoing, random sample was drawn from the total membership in the Fall of
1966. A comprehensive system for computerizing the medical information for the
sample was developed. The outpatient medical data for this study were obtained
from this sample. This section discusses the drawing of the sample, and its
socioeconomic-demographic characteristics.
All of the membership records of the Kaiser Health Plan were recorded on magnetic
tape for ease of computer processing. Each family, and member within that fam-
ily, were identified by a unique number called the Health Plan Identification
Number (H.P.I.D.). This provided a reliable sampling framework for Kaiser to
draw a random 5 percent sample of family units.
The original sample was drawn from a list of people eligible for Kaiser Health
Plan services on September 1, 1966. Each month since then a 5 percent random
sample of all new families joining the Health Plan have been added to the sample
[17, p. 300], Once an individual is chosen for the sample, he will always be a
member of the sample, even if he should become a non-member of the Health Plan.
Hence, in the aggregate, the sample was actually larger than 5 percent of the
membership at any point in time. Continuous medical care utilization is recorded
whenever members of the sample use or contact the Kaiser System. The original
sample had 1,487 member family units, with 4,123 individuals. By July 1968,
2,311 families and 6,514 individuals were sample members [17, p. 300],
The Kaiser 5 percent sample was designed to be representative of the Kaiser member-
ship at any point in time [17, p. 300], During 1967, 96 percent of the sample
members resided in the Portland S.M.S.A. Sixty-four percent of the members
A subscriber family may not necessarily include all the members of the family;
that is, some members of the primary family may not be members of the Kaiser
Health Plan.
See the following citation for a more complete presentation of this sampling
procedure [22].
39
-------
resided in Multnomah County [17, p. 302]. There was no reason to expect these
distributions to change materially between 1967 and the beginning of the study
period.
The Kaiser membership's representativeness of the population in the S.M.S.A. is
an unanswered question. Using 1960 census data, some comparisons of the Kaiser
membership to the population of the S.M.S.A. have been made. Four percent of
the total membership was non-white, compared with 3 percent for the S.M.S.A.
The Kaiser Health Plan membership had slightly more members in the under-25 age
group and had slightly fewer in the over-65 age group than the S.M.S.A. [17,
p. 302].1 Dr. Merwyn Greenlick, Director of Kaiser Health Services Research
Center, felt the Kaiser membership was not significantly different from the Port-
land S.M.S.A. However, since the Kaiser membership was not randomly drawn from
the S.M.S.A., he was wary of making inferences from membership or sample results
2
to the S.M.S.A.
The Kaiser Household Interview
One of the specific aims of Kaiser Research was to examine association between
socioeconomic-situational background characteristics and the consumption of med-
ical services in different disease states. It was hypothesized that, for cer-
tain disease situations, different background characteristics were important
determinants of medical care utilization [17, p. 299], To accomplish its re-
search objective, Kaiser developed a household interview survey to collect socio-
economic-demographic and other data from members of the 5 percent sample. It
was from this survey that the socioeconomic-demographic data for use in this
study were obtained.
Some other general characteristics of the Kaiser 5 percent sample membership
which were not compared to the S.M.S.A., showed that 81 percent were Protestants
and 16 percent were Catholic. Eighty-one percent of the sample subscribers were
married. Eighty-seven percent of the subscribers were employed, with semiskilled
(24 percent) and clerical sales (23 percent) appearing to be the main occupational
categories [17, p. 302].
2
Conversation with Merwyn R. Greenlick, Ph.D.; Director, Kaiser Health Services
Research Center, Portland, Oregon, November 22, 1971.
40
-------
The questionnaires were developed after a review of the sociological and social
psychological literature on health and medical care. Information was gathered
on four types of variables which were thought by Kaiser to primarily affect the
need for medical care: (1) bio-social factors, such as age and sex; (2) situa-
tional or physical environmental conditions, such as type and place of work;
(3) sociocultural agents, such as drinking, smoking, or conventional health
procedures; and (4) psychological stress factors, socially and environmentally
produced, including financial problems and family strife. Three types of vari-
ables thought to affect demand for the consumption of medical services were in-
cluded: (1) sociocultural factors, including the valuation the individual
places on health, the determination of normal health states, and opinions on
proper sources of medical care; (2) social components, such as pressure result-
ing from interaction with family members, friends, and colleagues at work; and
(3) psychological elements, such as sick-role orientation, optimism or pessimism,
or self-acceptance [16, p. 3],
The questionnaire survey was administered to each eligible Kaiser Health Plan
subscriber, or subscriber and covered spouse, in the 5 percent sample, beginning
February 1970 and terminating August 1971. Kaiser processed and accepted a total
of 2,409 interviews.
The questionnaires were administered separately, but simultaneously, to husband
and wife in two-parent families. A different, but related, questionnaire was
administered to each spouse (questionnaires A and B). This interviewing tech-
nique was utilized because it was felt some data would be more reliable and valid
if obtained separately from the husband/father and wife/mother. Other data,
especially that of a subjective nature (such as perceptual-attitudinal data),
must be obtained from each person independently [17, pp. 300-301]. In a one-
parent or single-subscriber household, a C questionnaire was given which con-
tained all the questions of the A-B set. On the average, it took one and one-
half hours to administer the questionnaire.
njhpublished Kaiser paper titled: West Coast Community Surveys Household Inter-
view Sample Statistics, October 1, 1971.
41
-------
VI. COMPILING THE MEDICAL DATA
Inpatient Medical Data
Since 1965, detailed socioeconomic and medical information has been obtained
on each inpatient admission to Bess Kaiser Hospital. These data were stored
on magnetic tape, and comprised 65,000 records. Annual inpatient admissions
ran approximately 9 percent of the total Kaiser membership in any given year
of the study period. It was originally intended to include in the analysis
data from the inpatient records of the 5 percent sample members. However,
several difficulties arose: (1) the type of laboratory and radiology procedures
performed were not coded in sufficient detail to be compatible with the outpatient
data; (2) major coding errors committed by Kaiser personnel would have necessi-
tated going back to the raw data in order to correct them; (3) the inpatient
service records of some patients were recorded under multiple H.P.I.D. numbers;
and (4) two "International Classification of Diseases, Adapted" editions (I.C.D.A.)
were used to index admitting and discharge diagnoses and surgical procedures.
The I.C.D.A. numbers of the two editions were not comparable. As the inpatient
medical data were to be related to socioeconomic data obtained from the house-
hold interviews of the 5 percent sample, and as it appeared that fewer than 50
hospital discharges per year were made of patients in the relevant disease cate-
gories who were members of the 5 percent sample, the decision was made not to
commit the cost and effort required to overcome the above obstacles. Therefore,
only outpatient medical data were used in the study.
The I.C.D.A. is a code used for classification of morbidity (disease) and
mortality (death) information for statistical purposes. The I.C.D.A. is
periodically revised. Prior to January 1, 1970, Kaiser used the 1962 re-
vised edition of the I.C.D.A. [40] for indexing inpatient records; the eighth
revision of the I.C.D.A. [39] has been used since January 1, 1970. The 1968,
edition of the I.C.D.A. has served as the basis for coding diagnostic data
for the official morbidity and mortality statistics in the United States
[39, pp. IX-X]. See [39, pp. IX-XXXII] for a complete history on development
and use of the I.C.D.A.
42
-------
Outpatient Medical Data
All outpatient medical services consumed by the 5 percent sample were coded and
recorded by Kaiser's medical record technicians on forms specifically prepared
for computer processing. All contacts with the Kaiser Health Plan system by the
sample members, including telephone calls and letters, have been continuously
recorded on magnetic tape since January 1967. Data recorded for each contact
include, among others, date, time, place, and type of service. The major symp-
toms, and duration of these symptoms, were recorded for each presenting and
associated morbidity. The medical procedures rendered, including laboratory
and X-ray, were coded in detail for each outpatient visit.
With one exception, the presenting and associated morbidities were coded accord-
ing to the 1962 revised edition I.C.D.A. [40], as supplemented and adapted by
Kaiser. The exception was the use of symptom codes (called "T codes") developed
by Kaiser to be used in cases where a disease condition was not diagnosed before
patient care terminated. Symptoms represented the final diagnosis for 10 percent
of the patients utilizing outpatient services at Kaiser [23, p. 249], Kaiser,
finding symptoms codes in the I.C.D.A. not extensive or unique enough for many
disease states, developed its own symptoms classification system. These codes,
which facilitate ease of coding and analyzing presenting and associated morbidity
data, were developed in the following manner. Each organ system was assigned a
single block of numbers. Additional number groups were given to non-specific and
psychiatric symptoms. Since the I.C.D.A. had some symptom codes, Kaiser used the
letter "T" to distinguish its codes from the I.C.D.A.
The coding system used to identify each medical procedure rendered was an adapta-
tion by Kaiser of the 1964 California Relative Value Fee Schedule (C.R.V.S.),
In the event care was given for two or more morbidities during the same contact
with the system, the morbidity that provoked the contact was identified as the
presenting morbidity, and the others as associated morbidities. The use of
the term "associated" did not necessarily mean that the diseases were clinic-
ally linked to each other. The term "associated morbidity" means only that
the diseases were associated in time. Nor should the distinction between pre-
senting and associated morbidities be linked with the seriousness of either
disease; a presenting morbidity could be minor, whereas the associated mor-
bidity could be a major disease [28, pp. 16-17],
43
-------
[6], Each medical procedure performed on the patient was translated into a
unique four-digit number for coding purposes. In essence, then, the Kaiser
medical data system was established in such a way that most medical procedures
rendered per outpatient contact could be associated with a specific disease.
However, there were two major shortcomings which existed with the outpatient
data. As with the inpatient data, the consumption of drugs was not coded for
the five percent sample during the study period. Also, if a patient received
medical services outside the Kaiser system, such services were not included in
the Kaiser data.
The outpatient medical data were received from Kaiser on two magnetic tapes.
The tapes were ordered by H.P.I.D. number and date of service. The tapes con-
tained 145,000 separate outpatient medical records on the five percent sample
for 1967-1970. These tapes were searched by H.P.I.D. number to obtain only
medical records for which socioeconomic data were available from the household
1 2
interview. A new tape containing 103,000 records was generated.
A thorough search was made of the I.C.D.A. and T-Codes [24] used by Kaiser to
prepare a list of over 500 diseases and symptoms which were thought to be aggra-
3
vated by, or associated with, air pollution. The codes were grouped into the
eight categories shown in Table 1.
Tlach interview was identified by the member's H.P.I.D. number at the time of
the interview. All members of the subscriber's health plan were identified
by unique person numbers (part of the H.P.I.D. number). These numbers allowed
data from the household survey to be attached to the member's medical records.
2
A precaution had to be taken to discover H.P.I.D. numbers that had changed
since the household interview date. The problem occurred, for example, when
separations, divorces, or name changes took place. It was highly probable
with such changes that covered dependents stayed under the old H.P.I.D. num-
ber. Hence, in creating the new medical tapes, medical records were accepted
under both H.P.I.D. numbers.
3
Appreciation is expressed to Sheldon Wagner, M.D., Head, Oregon State University
Environmental Health Unit, Good Samaritan Hospital, Corvallis, Oregon, for his
help in developing this list.
44
-------
TABLE 1. THE DISEASE CLASSIFICATION SYSTEM
I. Respiratory System
A. Upper respiratory neoplasms
B. Upper respiratory infections
C. Lower respiratory infections
D. Lower respiratory neoplasms
II. Allergies Affecting Respiratory System
A. Asthma
B. Hay fever
C. Specific diseases
III. Other Allergies and Skin Diseases
A. Eczema
B. Other allergies
C. Other skin diseases
D. Specific diseases - allergies
IV. Diseases of the Circulatory System
A. Heart
B. Circulatory system
V. Diseases of the Digestive System
A. Upper gastrointestinal tract - neoplasms
B. Ulcer of the upper gastrointestinal tract
C. Lower gastrointestinal tract - neoplasms
VI. Diseases of the Eye
A. Inflammatory diseases of the eye
B. Non-specific eye complaints
VII. Diseases of the Genitourinary System
A. Benign neoplasms
B. Malignant neoplasms
C, Nephritis and nephrosis
VIII. Other Diseases
A. Symptoms and diagnoses referrable to nervous system
and special senses
B. Emotional, mental, and psychotic disorders
C. Other specific diseases
45
-------
Each disease is classified more specifically within the general outline presented
in Table 1. Within the system, acute and chronic diseases were grouped separately,
as were benign and malignant neoplasms. Each disease is identified by a code,
such that any single disease or group of diseases can be selected for analysis.
For example, I-A-l-a is a malignant neoplasm of the nose, contained within the
category I-A, upper respiratory neoplasms. The complete listing of all the dis-
eases and symptoms comprising Table 1 may be seen in Appendix A. The outline
code (I-A-l-a), the I.C.D.A., and the T-Code number were coded for each disease.
These were used to search the reduced outpatient medical tape discussed above
to create a third outpatient tape containing only the medical records of household
interview respondents and covered dependents afflicted by the diseases on the list.
The "Outpatient I.C.D.A., T-Code Tape," as it is called, contained 24,000 records
for the 1967-1970 period. It became the main medical tape, and was employed for
generating other tapes used in the statistical analysis.
Derivation of the Dependent Variables
The California Relative Value Schedule Units
Chapter IV described the advantages of utilizing in the analysis, as a dependent
variable, an index of the physical quantities of medical services consumed. While
no such index was found, the California Relative Value Schedule (C.R.V.S.)
appeared to be an appropriate substitute, and was already in extensive use by
Kaiser Research for coding medical procedures. The first edition of the C.R.V.S.
was formulated in 1956 by the Committee on Fees of the California Medical Associa-
tion. The main objective of the Committee was to develop a set of principles
to govern the development of fee schedules in California and, in so doing, pro-
vide some uniformity in the setting of fees [6, p. 64]. As part of this objec-
tive, the Committee also established a uniform nomenclature of medical procedures
and a standardized code to designate each procedure.
The Committee found through a survey of the California Medical Association mem-
bership that, while the dollar value of fees varied widely throughout the state,
the relationship between fees for the same procedure remained essentially the
same [6, p. 64]. This observed stability in the relationships among various
46
-------
procedures became the basis of the C.R.V.S. Various factors could be applied
to the C.R.V.S. units to convert them to dollar values, thus allowing the mone-
tary fees charged by various physicians to differ from one another. To account
for new factors and innovations in the field of medical practice, the Committee
has issued a new edition of the C.R.V.S. every three to five years.
The C.R.V.S. was selected as the best approximation available of the desired
index of medical care provided. Its use had the following advantages for this
study: (1) all services are identified by a four-digit code, and most services
are weighted by a unit value. (2) These codes and unit values are applicable to
an entire range of medical services: medicine, anesthesia, radiology, surgery,
and laboratory. (3) Kaiser Research used the C.R.V.S. procedure codes (1964
edition) to code medical services for outpatient care. (4) Kaiser, with some
modification to fit its own purposes, had coded the procedure numbers and unit
values for the 1964 edition of the C.R.V.S. on magnetic tape. The tape was
continually updated by Kaiser to reflect changes in providing medical services.
A copy of this tape was obtained from Kaiser. (5) The C.R.V.S. units are addi-
tfve.
The C.R.V.S. tape received from Kaiser was modified for anesthesia procedures
used in surgery. There were four such procedures used in outpatient treatment
for which unit values had not been assigned. Calculation of the anesthesia
units in the 1964 edition of the C.R.V.S. was done by adding the listed basic
unit for anesthesia per surgical procedure to the time, converted to units,
spent in administering the anesthesia. Kaiser had recorded on its tape only
the C.R.V.S. units applicable to each surgical procedure, and not the basic
units applicable to the anesthesia. The C.R.V.S. tape was modified by attaching
the basic anesthesia units to each surgical procedure. Hence, whenever an out-
patient surgical procedure required anesthesia, the basic anesthesia units were
added to the units for the surgical procedure. Efforts to find some mean or
median value of time spent on each surgical procedure were not successful.
Hence, the time factor was not taken into account in calculating anesthesia
units per surgical procedure.
47
-------
Model 3
The Index of Medical Services (Y..,). It was required that, in order to express
JJ*:
the effect of air pollution on health as an economic cost, the C.R.V.S. units
per medical procedure be converted to dollar values. This had to be done in a
manner that would not distort the quantity of medical services consumed, as
approximated by the C.R.V.S. units. This was accomplished by taking each
C.R.V.S. unit, per medical category, times a constant dollar equivalency (con-
version) provided by Kaiser. The dollar equivalencies per medical category are
shown in Table 2.
Table 2. OUTPATIENT CALIFORNIA RELATIVE VALUE
SCHEDULE DOLLAR EQUIVALENCY FEES
Medical Dollar
category equivalency
Surgery 6.50
Pediatrics 7.00
Laboratory and radiology 7.00
All other 8.00
The dollar equivalencies came from two sources. The credit office, Bess Kaiser
Hospital, provided the dollar conversions for surgery, pediatrics, and all
other. The laboratory and radiology conversion factors were obtained from Kai-
ser Research Center. The credit office and research center used the C.R.V.S.
units and dollar equivalencies to bill non-members and certain public grants
for medical services rendered. These dollar equivalencies were established
January 1, 1969. The equivalencies, taken times the C.R.V.S. units, reflected
the going fee for service rates for professional outpatient medical services in
Portland. The resulting dollar values may not be representative of Kaiser's*
Conversation with Jack Thomas, Credit Manager, Bess Kaiser Hospital, Portland,
Oregon, July 13, 1972.
48
-------
costs of providing the services. Kaiser did not have a detailed enough cost
accounting system to provide such information.
The C.R.V.S. units were converted to dollar values (Y.., ) as follows: by code
IJK
number, each medical procedure on the Kaiser outpatient tape was matched against
the same procedure on the C.R.V.S. tape. The units on the C.R.V.S. tape were
then multiplied by the appropriate dollar equivalencies. For pediatrics and sur-
gical procedures, the appropriate dollar equivalencies to assign were determined
by the numeric codes assigned to the doctors. The laboratory and radiology pro-
cedures were grouped on the outpatient tape within a given numeric field; pro-
cedures in these fields were matched and converted to dollar values. All other
procedures and respective units were multiplied by the $8 equivalency.
Certain coded medical procedures in the 1964 C.R.V.S. edition did not have unit
values assigned to them. This seemed to occur primarily when the services
rendered were unusual and difficult to quantify. Kaiser sometimes modified the
C.R.V.S. and assigned units to the procedures. However, no attempt was made by
the authors to proxy values to procedures which Kaiser had failed to modify. The
problem was not widespread enough to be considered important. However, to the
extent these services were used by outpatients, Y . would represent an under-
ijfc
statement of the dollar value of medical services consumed.
Y.«, then represents the summation of dollar values (converted C.R.V.S. units)
ijk
for all outpatient medical services rendered to Kaiser questionnaire respondents
in treating a presenting and/or associated morbidity in the same office visit,
per individual, per day.
Model 5
Per Capita Index of Consumed Outpatient Medical Services, (Yic-)- The dependent
variable of (5) represents the summation of dollar values (converted C.R.V.S.
units) expressed on a per capita basis, for all outpatient medical services
rendered (to Kaiser questionnaire respondents) in treating a presenting and/or
associated morbidity in the same office visit, per day, per census tract.
49
-------
VII. COMPILING THE SOCIOECONOMIC-DEMOGRAPHIC DATA
This section will discuss the derivation of the socioeconomic-demographic vari-
ables specified for the statistical models. Each variable in Equation (3) is
also represented in Equation (5). The variables are the same, except for speci-
fication as individual or aggregate variables in the two equations. Except where
noted, each variable has the same hypothesized relationship with the dependent
variable.
The socioeconomic-demographic variables are matched in the statistical analysis
by H.P.I.D. members. For purposes of this study, it was decided to use only
the medical records of actual respondents to the household interviews for whom
complete and specific socioeconomic-demographic data existed.
Most questions on the Kaiser household interview were coded with a numeric code,
with each number representing a verbal response to the questions. The data were
transformed into continuous data wherever possible. The transformations and other
modifications performed on each variable will be discussed below.
Age of the Patient: Xli>k - Model 3, Xj_ic. - Model 5
It is expected that older patients will consume more medical services. Hence,
higher values of this variable are expected to be positively related to the con-
sumption of medical services. The age of the patient is expressed in years as of
the date of medical service.
Sex of the Patient: X2i>k - Model 3, 2±C' ~ Model 5
This variable is expressed as a zero if the patient is male, or a one if the
patient is female in (3) . In (5) the variable is expressed as the percentage of
female patients seeking medical care. The variable has been included because
the incidence of certain diseases is more prevalent in one sex than in the other.
For example, the incidence of emphysema has been reported to be ten times greater
in men than in women [10, p. 13]. A priori, little can be said about the effect
of sex on the consumption of outpatient medical services generally, since the
relationship varies with the type of disease.
50
-------
Marital Status: X,.., - Model 3, X.. . - Model 5
JlJ K. JlCJ
In (3), this variable is expressed as a zero if the patient is not married, and
as a one if he is. In (5) the variable is expressed as the percent of married
patients seeking medical care per census tract. For reasons detailed in Chapter
IV, this variable is expected to be negatively related to the consumption of
medical services.
Number of People in the Patient's Household;
X,... - Model 3, X.. . - Model 5
4ijk *_ 4icj
It was desired to have some kind of density measure, such as the number of per-
sons per room, as an indication of crowding within the patient's home. Increased
crowding could result in more physical and emotional strain being placed upon all
household members, and could also effect an increased potential for initial infec-
tion and re-infection of household members with certain respiratory diseases.
This could result in the increased consumption of medical services.
A density measure could not be obtained or specified from the Kaiser data. The
number of people residing in the patient's household was the best measure that
could be obtained. The variable includes not only members of the immediate family,
but anyone else living in the household at the time of the interview. Higher
values of this variable are expected to be positively related to the dependent
variable.
Household Income: X5 k - Model 3, X-. . - Model 5
Household income is Included to measure the ability of the patient's household to
make expenditures in lieu of consuming medical services to preserve and improve
its state of health. Since all Kaiser members have the same opportunity to con-
sume Kaiser medical care, it is expected that this variable will be negatively
related to the consumption of medical services. An estimate of the 1969 house-
hold income, before taxes and from all sources, was used as a measure of this
variable.
It should be pointed out that, in most instances, the income question was asked
of only one respondent per household. For example, in the A-B questionnaire
51
-------
only the A questionnaire was given to the male respondent. If B required medical
service, A's answer would be assigned to her as a measure of household income.
If two C questionnaires were administered to respondents living in the same house-
hold,1 and the responses to the household income question were different between
the two respondents, then the two responses were averaged to obtain an estimate
of household income. Out of the 2,439 questionnaires, this happened approximately
five times.
The household income question asked the respondent to indicate into which one of
nine income classes his 1969 before-tax income fell. The midpoint of the relevant
class was assigned to the respondent as his income. An exception to this rule had
to be made for the highest open-ended class (greater than $20,000). Census data
for the Portland S.M.S.A. relating to "incomes of families and unrelated individ-
uals" [53, pp. 39-191] were used to estimate a median income of $32,000 for house-
holds indicating an income over $20,000. Table 3 shows the income response cate-
gories from the questionnaire, and the corresponding transformations per response
category.
TABLE 3. HOUSEHOLD INCOME RESPONSE CATEGORIES AND TRANSFORMATIONS
Response category Transformation
Under $2,500 per year $ 1,250
$ 2,500 - $ 3,499 3,000
3,500 - 4,999 4,250
5,000 - 6,499 5,750
6,500 - 7,499 7,000
7,500 - 9,999 8,750
10,000 - 14,999 12,500
15,000 - 19,999 17,500
Over $20,000 32,000
This happened only when husband and wife both worked, were separate subscribers
to the Kaiser Health Plan, and were both members of the 5 percent sample.
52
-------
Racei X6ijk» X7i1k - Model 3» X6icj "
In Model (3) the race variable is expressed as a dummy variable in the following
manner:
Race
White ,
X6ijk
1
0
0
X7ijk
o
1
0
On each variable the statistical test is whether the consumption of medical ser-
vices is different for Negroes or other non-whites (mostly Oriental), as compared
to the consumption of medical services by whites.
In Model (5) the Negro and other non-white variables are combined, and express
the percentage of non-white patients consuming medical services. The variable
permits a test on whether the consumption of medical services by non-whites is
materially different from whites.
As with the sex variable, the relationship of the race variables to the consump-
tion of medical services will depend upon the diseases aggregated in the depen-
dent variable. If the relationship of race to disease incidence is not known,
then, a priori, the sign of the variable's coefficient cannot be determined.
Physical Fitness: Xg.., and Xg. ., - Model 3
A physically fit patient would be expected to consume fewer medical services.
There was one question in the household interview which elicited information on
physical fitness. It was stated to the respondent as follows:
"How much time and energy do you spend being physically
fit - a great deal, some, or very little?"
Dummy variables were used to describe the qualitative responses as follows:
53
-------
Physical fitness
Great deal of time and energy... 1 0
Some time and energy ............ 0 1
Very little time and energy..... 0 0
The responses to the physical fitness question were subjective. Hence, there
was difficulty in standardizing the answers as between responses. The question
was the best assessment on physical fitness available from the interview. How-
ever, the use of qualitative (dummy) variables does permit explaining some differ-
ences between individual responses. The variables were expected to be negatively
related to the dependent variable.
Physical fitness was not included as a variable in (5) because only qualitative
variables could be used to represent each response. Aggregation of qualitative
variables over census tracts, where variations in group behavior are being ex-
plained, presents a problem of adequately interpreting the results.
Consumption of Alcoholic Beverages ;
X10i.1k " Model 3» *7icj " M°del 5
This variable expresses the number of times in the 12 months prior to the inter-
view date the patient has had the equivalent of about six drinks , a bottle of
table wine, or eight cans of beer. Higher values of this variable are expected
to be associated with a greater consumption of medical services on the part, of
the patient.
Table 4 shows the transformations on each response category to the question. The
transformations were constructed by taking the midpoint of each time period indi-
cated in the response.
Cigarette Smoking: X< - Model 3, X. - Model 5
Each patient's smoking characteristics were embodied in a single index. The
index took a value of zero if the patient did not smoke, and went to a high of
1960 for current smokers smoking the longest period of time and the largest
54
-------
Table 4. TRANSFORMATIONS PERFORMED ON THE RESPONSES
TO THE DRINKING QUESTION
Response categories Transformations
Every day or nearly every day 365.0
Three or four times a week 182.5
Once or twice a week 78.2
Two or three times a month 30.0
About once a month 12.0
Six to eleven times a year 8.5
One to five times a year 3.0
Never in the last twelve months. 0.0
number of cigarettes per day. The index also incorporated the smoking character-
istics of former smokers, including the amount they smoked, how long they smoked,
and how long it had been since they stopped smoking.
Two sets of questions from the household interview were used to obtain informa-
tion on the smoking habits of current and former cigarette smokers. The first
question asked the respondent if he currently smoked cigarettes, and if he did,
how many cigarettes he smoked each day; and how many years he had smoked. The
responses to this question are shown under (I) in Table 5. The second question
asked former smokers how many cigarettes they used to smoke, how many years they
had smoked, and how long it had been since they had stopped smoking. The respon-
ses to this question are under (II) in Table 5.
In Table 5, the transformations were constructed by taking the midpoint of each
response interval. The following criteria were followed in quantifying the open-
ended responses. A frequency distribution of the responses showed that fewer than
nine respondents out of 2,439 indicated they currently smoked or had ever smoked
more than 61 cigarettes a day. It was decided that 70 cigarettes would be a rea-
sonable number to use for this group.
A frequency distribution on age was generated for the questionnaire respondents.
Their median age was calculated to be 46. The legal age for smoking in Oregon is
55
-------
Table 5. SMOKING CHARACTERISTICS AND TRANSFORMATIONS ON
THE RESPONSE CATEGORIES
Response categories Transformations
I. Current smokers:
A. Number cigarettes per day:
Under 10 5
10 to 20 15
21 to 40 30
41 to 60 50
61 and over 70
B. Number of years smoked:
Under 1 year 0.5
1 to 3 years 2.5
4 to 5 years 5.0
6 to 10 years 8.5
11 to 20 years 15.0
Over 20 years 28.0
II. Former smokers:
A. Number cigarettes per day:
Under 10 5
10 to 20 15
21 to 40 30
41 to 60 50
61 and over 70
B. Number of years smoked:
Under 1 year 0.5
1 to 3 years 2.5
4 to 5 years 5.0
6 to 10 years 8.5
11 to 20 years 15.0
Over 20 years 28.0
C. Number of years since cessation
of smoking:
Less than 3 months 0.13
4 to 6 months 0.42
7 months to 1 year 0.79
More than 1 to 3 years 2.50
4 to 5 years 5.00
6 to 10 years 8.50
Over 10 years 13.50
56
-------
18. While some people start smoking earlier and others later, 18 is probably
a good estimate of the median age when people start smoking. Therefore, 28
years (the difference between 46 and 18) appeared to be a reasonable number of
years for those who indicated they smoked for 20 years or more.
The last response category in Table 5 (II-C) asked the former smoker how long
it had been since he had stopped smoking. The American Cancer Society and other
researchers have established that after 10 years of not smoking, the death rates
of former smokers and lifelong abstainers are virtually the same [3, p. 131; 65,
p. 145], Given this evidence and the structuring of the last response category
in (II-C), it was felt that 13.5 years would be a reasonable number of years
for those who indicated they had quit smoking over 10 years ago.
There is a direct association between the number of cigarettes smoked per day,
the number of years smoked, and mortality and morbidity incidences of lung cancer
and coronary diseases [64; 65, p. 11]. It appears that as the number of years
since cessation of smoking increases, the ex-smoker's state of health becomes com-
parable to that of a person who has never smoked. There is also an Immediate
response of the body to cessation of smoking. Within six months after discontinu-
ance, the bronchial system improves to a steady state; i.e., pulmonary function
or vital capacity stabilizes. There is an immediate response of the cilia to
2
cessation of smoking. Each cigarette paralyzes the cilia for 20 minutes. It is
the opinion of some medical authorities that it takes from 3 to 6 weeks after
discontinuing smoking to make a reasonable difference in the ability of the body
3
to ward off infection.
It was desired to construct an index which would have at least some potential
for reflecting the conditions described in the preceding paragraph. Extensive
reading of the medical literature and conversation with medical authorities failed
Conversation with James F. Morris, M.D., Veterans Administration Hospital, Port-
land, Oregon, December 3, 1971.
2Cilia are hairlike cells that line the airways and, by their movement, propel
the dirt and germ-filled mucous out of the respiratory tract [41, p. 91],
Conversation with Arthur Koski, Head and Professor, Department of Health Educa-
tion, Oregon State University, Corvallis, October 19, 1971.
57
-------
to give any clues on the existence or on means of constructing such an index.
Finally an index, described by Equation (6) and developed by the authors, was
constructed for the Kaiser data:
V
t*i
where S, = smoking index of the k — person seeking outpatient
medical care.
A, = the number of cigarettes per day the k — patient
seeking outpatient medical care currently smokes
per day.
B = the number of cigarettes per day the fc — patient
seeking outpatient medical care used to smoke per
day.
C. = the number of years the k — patient seeking out-
patient medical care has smoked.
D = the number of years the fc — patient seeking out-
patient medical care used _tp_ smoke before quitting.
E, = the number of years since the k — patient seeking
outpatient medical services has quit smoking; and
(k = !,••-, £).
If a person currently smoked 70 cigarettes per day, and had done so for 28 years,
his index would be 1960.0. If a patient quit smoking 13.5 years ago and, at the
time he quit smoking, smoked 70 cigarettes per day for a period of 28 years, then
his index would be 135.2. Patients who have never smoked in their lives assumed
a value of zero in the index.
Higher values of the index represent more intense cigarette smoking. Because of
the relationship between intense cigarette smoking and cardiovascular-respiratory
problems, this index is expected to be positively related to the dependent variable
in both models .
Occupational Air Pollution Exposure;
X12ijk " Ifodel 3' X9icj - Model 5
Job-related pollution can aggravate certain respiratory problems [27, p. 34; 42,
p. 13], Job-related pollution would not, in most instances, be measured by the
58
-------
regional, ambient air pollution stations. To take account of this fact, an
occupation exposure index was constructed for each employed patient seeking out-
patient medical care. A value of zero was assigned to patients who did not work.
It is expected that higher values of this variable will be positively related to
the dependent variable in both models.
The occupational exposure index was constructed in the following manner: the
household interview contained a question asking each respondent to identify his
specific job. The question was coded, using the numerical classification of
occupations published in the 1960 Census of Population, Alphabetical Index of
Occupations and Industries [52]. A listing of the occupations was submitted to
Mr. Darryl D. Douglas, who rated each classification on a scale of one to four.
The ratings were assigned to each class, based on the probability of that occupa-
tion being characterized by a particulate problem which could aggravate certain
cardiovascular-respiratory problems. The meaning of the scales was as follows:
four equaled a high hazard to particulate exposure; three equaled an intermediate
hazard; two equaled a moderate hazard; and one equaled a low hazard. Each weight,
with its respective occupation code, was multiplied by the average number of hours,
including regular overtime, the patient worked per day. The resulting product
was the occupational exposure index.
. Douglas is the Director, Occupational Health Section, Health Division,
Department of Human Resources, State of Oregon, Portland.
59
-------
VIII. COMPILING THE AIR POLLUTION AND METEOROLOGICAL DATA
The Air Pollution Data
The air pollution data were collected from the various state and regional air
pollution control agencies having jurisdiction in the Portland S.M.S.A. All
of the agencies had similar methods of sample scheduling and collection, analy-
sis, and data reporting. Hence, the air pollution data were comparable among
the originating agencies.
Each air pollution station was identified by a unique number. The stations were
located on a grid coordinate system, expressed in state plane coordinates.
The coordinates were used by one of the regional agencies to plot air pollution
isopleth maps of the form shown in Figure 5. The suspended particulate stations
operated continuously for 24 hours every fourth day. All stations in the Port-
land S.M.S.A. were operated simultaneously. The air pollution data were coded
on 80-column cards for ease of data processing. Stations which had error in
their data, or which were poorly sited such that they did not provide repre-
sentative measures of ambient air pollution, were not used.
.Selecting the Proper Measure of Air Pollution
Portland's main air pollution problem is particulate matter. The effects of
particulate matter and other air pollutants were presented very briefly in Chap-
ter IV. More specifically, particulate matter can physiologically affect human
health in several ways. These effects would initially occur in the respiratory
system, but the cardiovascular system could also be affected by additional
In order to insure the location of original land survey measurements, the United
States Coast and Geologic Survey has worked out for each state a system of state
plane coordinates [48, pp. 29-30], The coordinates are expressed in feet. To
limit the size of the grids and to keep scale variations to a minimum, each state
usually has two or more overlapping zones. Each zone is covered by a single
coordinate system [30, pp. 345-346]. All Oregon air pollution and meteorological
stations are located in the Oregon North Zone; all Washington air pollution and
meteorological stations are located in the Washington South Zone.
60
-------
WILLAMETTE R.
HM-LSBORO
40 WASHINGTON
I CO.
FIGURE 5; ISOPLETH MAP FOR PORTLAND AREA
-------
stress being placed upon it by an impaired respiratory system. Particulate mat-
ter may exert a toxic effect on the human body via one or more of the following
three mechanisms:
1. The effect of particles on human health can be due to the
particle's inherent chemical and/or physical characteristics
[59, p. 141]. The most important toxic aerosol is sulfur tri-
oxide. Others of increasing importance are lead, beryllium,
and asbestos [59, p. 129].
2. The particles may interfere with one or more of the clearance
mechanisms in the respiratory tract [59, p. 141], The particle
size will determine the location of the toxic effect in the res-
piratory system. Small particles of less than two or three mi-
crons can penetrate deep into the respiratory system where no
protective mucous blanket exists [41, pp. 33-34].
3. The particle can act as a carrier of an adsorbed toxic substance
159, p. 141]. The toxic effect of sulfur dioxide seems to be
greater when the gas combines with an aerosol than either is alone
[1, p. 62; 60, p. 111]. Sulfur oxides can produce immediate air-
way constriction [41, p. 66].
Some air contaminants (gaseous or particulate) could weaken the body's defense
mechanism, making it more susceptible to germs, bacteria, or viruses which can
precipitate an active disease [41, p. 65]. A review of numerous epidemiological
studies indicated an association between air pollution, as measured by particu-
late matter accompanied by sulfur dioxide, and health effects of varying severity
[60, p. 146].
There are two main types of atmospheric particulate matter in urban areas: sus-
pended particulate matter and dustfall (particle fallout) [59, pp. 11-17]. In
this study, suspended particulate data were selected over particle fallout as the
measure of air quality in the study area. There are several reasons for this.
Dustfall consists of particulates 10 microns or larger [59, p. 16], which settle
out of the air fairly rapidly [44, p. 7; 59, p. 28]. These particle sizes are
too large to be respirable by the human body. Suspended particulate matter con-
sists of particle sizes smaller than 10 microns [59, p. 16], Suspended particu-
late matter is respirable. Hence, because of the settled particles' larger sizes,
particle fallout measurements are not capable of assessing the health risks of air
62
-------
pollution. The large sampling period usually associated with collecting particle
fallout (30 days) would prohibit measuring, or approximating, day-to-day changes
in air quality.
High-volume samplers are used by the air pollution control agencies to collect
suspended particulate matter. The operation of high-volume samplers consists of
drawing air through a filter of low air resistence. The filters are felts of
glass or some synthetic organic fiber [59, p. 22], The filters can trap particle
sizes as small as 0.3 microns [61]. The samples are collected over a 24-hour
period. The filters are then analyzed, and the results reported as total weight
in micrograms per cubic meter. The total weight of suspended particulate mat-
ter was used as the air pollution parameter in this study.
Preparing the Air Pollution Data for Use
in the Statistical Models
The suspended particulate stations were operated every fourth day during the
study period. The medical and meteorological data were daily observations. It
was desired to observe changes in the consumption of medical services with changes
in air quality. This required the formulation of a procedure to estimate air
quality on days for which data were not available. Two alternatives were con-
sidered to estimate the missing data: linear interpolation and spline fit. The
spline fit was selected over linear interpolation for the reasons detailed below.
The spline fit is executed by connecting each pair of adjacent points (observa-
tions) with a third degree polynomial, matching up the sections such that the
first and second derivatives are continuous at each point [45, pp. 404-405]; i.e.,
the spline fit is a piecewise cubic function which is twice continuously differ-
entiable [67, p. 27].
The spline curve is a higher order fit of data points than is linear interpola-
tion. The curve is of piecewise construction, which means that only part of the
data for a given station is fitted at a time. More specifically, the spline fit,
Some agencies perform analyses on the total weight of particulates to determine
their chemical composition. These analyses are also reported in micrograms per
cubic meter.
63
-------
as employed in this study, used four observation points to fit each segment. For
example, observation points A, B, C, and D would be used to fit one segment of
the curve. Points B, C, D, and E would be used to fit the next segment. While
estimated points between D and E would be most influenced by the actual observed
values at D and E, points B and C would also have some influence. Using linear
interpolation would have resulted in only two observation points being used to
determine each linear segment.
Assigning Air Pollution Values to Residential
and Job Addresses
A method was needed to assign air quality values to each household interview
respondent for outpatient medical services consumed in the treatment of the dis-
ease categories listed in Table 1. The procedure had to be capable of accounting
for air pollution levels at the patient's residence and his place of work.
The method for assigning air quality values to each patient made use of an Admatch
computer program developed by the U.S. Census Bureau and local agencies. The
program, in one of its forms, geocoded street addresses to state plane coordi-
nates. The meteorological and air pollution stations were also geographically
located by state plane coordinates. These coordinates were used to assign air
quality and meteorological values to each residence and job address.
Two techniques were considered for estimating air quality levels for each outpati-
ent. The first was a least squares analysis employing various mathematical forms.
This technique was not successful. In the second method, air pollution values of
a certain address were calculated by weighting the observed air pollution values
of the three closest stations by their distances from the address in question.
The distance from each address to each station was calculated using state plane
coordinates. The procedure is shown in Equation (7).
(S31>
See [54] for a detailed description of the Admatch program.
64
-------
where P = the estimated air pollution value for address i on day j.
D. . = the distance (in feet) the closest station is from address i
on day j.
S.. = the observed air pollution values of the closest air pollu-
^ tion station to i on day j.
D- . = the distance (in feet) of the second closest station from
^ address i on day j.
S^. = the observed value of air pollution on day j for the
~* second closest air pollution station to i.
D_ = the distance (in feet) of the third closest station from
•* address i on day j.
S,. = the observed value of air pollution on day j for the
^ third closest air pollution station to i; and
(i - 1, •••, d); (j = 1, ••-, m).
The use of distance to weight each air pollution observation is simple and in-
expensive to employ. The biggest disadvantage to its use is that only distance
is taken into account. Admittedly, topographical and meteorological conditions
can play important roles in fluencing the distribution of pollutants. Another
disadvantage of using the distance system is that the air pollution values calcu-
lated for any address are constrained to a range of values, the limits of which
are determined by the highest and lowest air pollution observations of the three
closest stations used in Equation (7). However, given the structuring of the
study's air pollution network, this disadvantage is not critical. The study's
air pollution station network was designed to measure ambient air pollution. The
air pollution station sampling sites were, for the most part, representative of
the surrounding terrain, and were not subjected to contamination by a point source
air pollution emission.
Several stations located in the immediate Portland Metropolitan area did not
operate continuously during the two-year study period. The computer program used
to calculate the weighted air pollution values took account of this non-operation.
If a non-operating station was one of the three closest stations to a patient's
work or residence address, then the non-operating station was passed over and the
next closest station to the address was used in calculating the weighted air
pollution value.
65
-------
Derivation of the Ambient Air Pollution Variable;
X13iik " Model 3» Xllicl - Model 5
Many of the Kaiser respondents work, and the place of work (in all but a few
instances) was different from the residence address. One would expect, there-
fore, that on the same day air pollution values would be different as between
the addresses, generally reflecting a suburban-urban gradient of ambient air
quality. It was desired to reflect such differences in the air pollution values
assigned to each patient. Hence, the air pollution variables in each model were
weighted to reflect the patient's potential exposure to air pollution at both
residence and job addresses .
The weighted air pollution value assigned to each patient was calculated by Equa-
tion (8) :
(J. .. + T. .. ) P' + [24 - (J.., + T.M )] P*
ijky ijk xjk ijk/J ijk
ijk - 24
where P.., = the weighted air pollution value assigned to the k —
^ patient who consumes outpatient medical services for
disease i on day j.
J..k = the number of hours worked per day (including regular
overtime) by the k — patient consuming outpatient
medical services for the i — disease on day j.
Ti'k = the avera6e transit time to and from work by the k —
patient consuming outpatient medical services for the
i — disease on day j .
f*Vi
Piik = estimated air pollution at the job address for the k—
patient consuming outpatient medical services for the
i — disease on day j .
*
Piik = es£imated air pollution at the residence address for the
k— — patient consuming outpatient medical services for the
i — disease on day j ; and
(i = 1, •••, n); (j - 1, ••-, m); (k = 1, •••, 0).
66
-------
Whether the patient worked or not, and the transit time to and from work, was
determined from the household interview. For purposes of calculating the weighted
air pollution, it was always assumed that the patient was either at work, in tran-
sit, or at his place of residence. The transit time was added to the number of
hours on the job, because it was felt the pollution in transit to and from the
job site would be higher than air pollution at the place of residence.
A question in the household interview asked the respondent what days of the
week he normally worked. Answers to the question allowed account to be taken in
Equation C8) of normal work week patterns. If the patient indicated he worked
Monday through Friday, then Saturday and Sunday were considered as days off;
i.e., if he became ill on Saturday or Sunday, and consumed out patient medical
services, the value of air quality assigned to him would be his residence air
pollution. Unless it was indicated otherwise, Sunday was always considered a
day off. An exception to this was when the patient indicated he worked through
the weekend, but had a day off in the middle of the week. In this case, Wednes-
day was assumed to be that day. It was assumed that patients who had Mondays,
Tuesdays, Thursdays, and Fridays off would be normally distributed about Wednes-
day. Patients who indicated they worked less than five days a week were assigned
their residence air pollution values, because there was no way of telling, from
the Kaiser data, which days they worked.
The Meteorological Data
Collecting and Analyzing the Meteorological Data
An attempt was made, via the selection of the meteorological data and the meteoro-
logical station sites, to account for the meteorological influences of the Columbia
River Gorge. Areas affected by the Gorge are generally characterized by lower
annual rainfall and cooler annual mean temperatures than other geographical areas
located in the Willamette Valley. The differences result from the interaction
of the dry, cold east Gorge winds with the moist marine air from the west.
Given the wide divergence of meteorological conditions between the Willamette
Valley and the Columbia Gorge, and the initial decision to reflect in the analysis
the effect of such variation on air pollution and health, it was important that
67
-------
any meteorological data selected for use in this study be representative of a
large geographical area and be of good quality. Data from the National Weather
Service (N.W.S.) of the U.S. Department of Commerce were used in this study.
The data were compiled from published sources [55, 56, 57].
There were many stations and data which could be coded within the study area.
Several simple statistical tests and visual inspections of the data were conducted
to determine what stations should be used and the frequency with which the data
should be coded. The analyses dealt with variations of meteorological conditions
in the study area over space and time. The period of record for these tests was
1967 through 1970.2
When one year was compared to another during the 1967 to 1970 period, mean annual
temperature and total annual precipitation of all meteorological stations in the
study area, with one exception, fluctuated together. That is, if Hillsboro was
greater in mean annual temperature for 1967 compared to 1968, then so were the
other stations. The exception was annual precipitation at Oregon City between
1969 and 1970 (see Figure 4 for a map of the study area).
A statistical test was conducted to determine whether annual weather patterns (as
measured by temperature) were significantly different for the 1967-1970 period.
For all meteorological stations taken as a group, the first test compared the
highest and lowest annual mean temperatures recorded during the four-year period.
The two stations exhibiting the largest difference between each other were Port-
land City (annual mean temperature 1967) versus Vancouver (annual mean tempera-
ture 1970). The test showed the means not to be statistically different at the
5 percent level.
In investigating and retrieving data available through the National Weather Ser-
vice, the researchers are indebted to Stanley Holbrook, Oregon State Climatolo-
gist, National Weather Service, Portland, Oregon.
2
The stations used in these analyses were located in or near the towns of Warren,
Portland, Hillsboro, Oregon City, Troutdale, and McMinnville in Oregon; and
Vancouver, Washington.
68
-------
For all stations taken as a group, a similar test compared the highest and low-
est monthly mean temperatures recorded (for the same months, but over the four-
year observation period) in order to determine whether there were significant
monthly fluctuations in temperature between the various stations. The statistical
test showed the means to be statistically different from each other at the 1 per-
cent level. The two stations exhibiting the largest difference were again Port-
land City and Vancouver for February 1970.
From these analyses it appeared that, while there may be monthly fluctuations
between stations, the annual meteorological conditions within the study area
were not materially different with respect to temperature. Where changes did
occur from one year to the next with temperature and precipitation, the changes,
with the exception noted, were all in the same direction.
Since this study was using daily medical records, and the air pollution values
could be estimated daily, it was decided to use daily meteorological conditions.
However, a judgment was made that, since all the stations seemed to fluctuate in
the same direction, coding of daily data for all the stations would not materially
add to the statistical analyses. Hence, the decision was made to code the daily
meteorological observations of two stations, and assume that the stations would
represent the meteorological changes occurring over the study area. The stations
coded were Portland International Airport, located on the Columbia River, and
Portland City Weather Bureau, located in downtown Portland. The Portland Air-
port station was assumed to be representative of the meteorological conditions
affected by the Columbia Gorge. The Portland City station was selected because
it was somewhat removed from the influence of the Gorge. Hence, data from the
downtown station were considered representative of the southern portion of the
study area. Patients consuming outpatient medical services were assigned the
meteorological values of the station closest to their residence address (as de-
termined using state plane coordinates).
Derivation of the Meteorological Variable;
- Model 3» *LOiej - Model 5
The effect of temperature extremes on cardiovascular-respiratory problems was
detailed in Chapter IV. It was desired to specify a measure which would incor-
porate both temperature extremes into one variable. Degree days, the absolute
69
-------
difference between the average daily temperature and 65° F, appeared to fulfill
this requirement. As the average ambient temperature becomes colder, the value
of the variable will increase. Similarly, as the average ambient temperature
becomes warmer (above 65° F), the value of the variable will also increase.
Higher values of this variable were expected to be positively related to the
dependent variables of both statistical models.
70
-------
IX. STATISTICAL RESULTS
Preliminary Analyses on the Air Pollution Data
An initial question which had to be answered was whether air quality levels vary
over time and space in the Portland area. Several preliminary analyses were
conducted on the air pollution data to determine whether such variation existed.
The analyses included computing the monthly means, variances, and standard devia-
tions for nine suspended particulate stations which had been in operation through-
out the 1969-1970 study period. The results indicated quite an amount of within-
month-variation for each station.
An analysis of variance was performed on the monthly means of nine suspended
particulate stations operating during the entire 1969-1970 period. The analysis
of variance may be seen in Table 6. It should be noted that the following holds
for any analysis of variance containing interaction terms: if any of the first
order interactions are significant (year by month, year by station, and month by
station in Table 6), then the main effects (year, month, and station in Table 6)
do not have any meaning.
Looking at the first interaction term (Y x M), the following hypothesis is tested:
Ho-: The differences between months are the same from one year to the next. The
null hypothesis is rejected. The F test is significant at the 1 percent level.
The differences between identical months are not the same from one year to the
next. One could tentatively conclude from the analysis that the depth of the
pollution surface (amount of pollution) in Portland changes from one year to the
next within the same months.
The interaction M x S tests the following hypothesis: Ho2: The differences be-
tween stations are the same from one month to the next within the same year. The
F test was not significant; hence the null hypothesis is not rejected. One can
tentatively conclude that the differences between stations do not change signifi-
cantly from one month to the next. This would tend to indicate that while the
depth of the pollution surface might change by month from one year to another,
the shape of the surface will not change significantly from station to station
over time.
71
-------
Table 6. ANALYSIS OF VARIANCE TABLE FOR THE
NINE SUSPENDED PAKEICULATE STATIONS
Source
Year by month
Month by statJ
Year by static
Year by month
(Y x M x
TOTAL...,
(Y x M)....
Lon (M x S).
in (Y x S)..
by station
S)
Sums of
squares
3967.2244
17909.9682
102040.3072
16389.1217
5398.1639
3881.0485
40618.4404
238784.2743
Degrees <
freedor
1
11
8
11
88
8
88
215
>f Mean
a square
3967.2244
1628.1789
12755.0384
1489.9202
613.3882
485.1311
461.5732
F
mmf
fmi^m
__
3.2279
1. 3289
1.0510
__
The interaction Y x s tests the following hypothesis: Ho,: The differences be-
tween stations are the same from one year to the next. The F statistic is not
significant, and Ho_ is not rejected. This would tend to corroborate the M x s
interaction: while there may be changes in the depth of the pollution surface,
the relative shape of the surface does not change from one year to the next.
The tests of the above hypotheses allowed the following conclusions to be advanced
about air pollution patterns in Portland. If annual or monthly isopleths of
pollution loadings in Portland were compared, one would find that while the amount
of pollution (depth of the pollution surface) in Portland may vary over time, the
relative loadings (shape of the pollution surface) would remain unchanged; i.e.,
areas with high pollution would remain high in comparison to all other areas,
and areas with low pollution would generally remain low in comparison to all other
areas.
Preliminary Examination of the Statistical Models
Only outpatient medical records of respondents to the household interview were
examined in this study. To obtain these records, a sort was made on the "Out-
patient I.C.D.A., T-Code Tape," discussed in Chapter VI, to create a second
72
-------
magnetic tape containing only respondent outpatient medical records for 1969 to
1970. This tape will be referred to as "The C.R.V.S. Severity Tape." It con-
tained approximately 6,000 outpatient medical records.
The medical records were matched to the socioeconomic-demographic data on the
"Household Interview Tape." As the medical and socioeconomic-demographic records
were being matched, the records of respondents not living in the study area were
deleted.
Air pollution and meteorological values were estimated for each patient on the
day of contact with the medical system, and for each of the three days prior to
that contact. The lagging of air pollution and meteorological values was an
attempt to determine how soon after meteorological and air pollution conditions
change do morbidity patterns change.
There are two principal types of delays which can effect a postponement between
exposure and the seeking of medical care: (1) a short time may elapse between
exposure to environmental conditions and disease onset or aggravation; (2) once
a patient is ill, it may take him awhile to recognize the fact and seek medical
care. The quest for determining a time sequence in causality between exposure
to air pollution and meteorological phenomena and the onset of an illness pre-
supposes some reasonable length of time between cause and effect. Because
several medical studies, using daily data, have used one-, two-, and three-day
lags [14, pp. 1062-1063; 21, p. 594], this study did likewise.
Two statistical models were postulated for this study. The transformations
which might be required on the models' variables, in order to obtain the best
fit of the data, were not known beforehand. In most instances there were numer-
ous transformations which could have been performed on each variable. Using all
data to test each transformation would have been expensive and time-consuming.
Hence, to pretest the statistical models, two 10-percent systematic samples were
drawn from The C.R.V.S. Severity Tapes. One sample was used to derive the final
specification of Model 3; the other was used on Model 5.
Many transformations were tried with each of the variables in the models. One
method was to examine scatter diagrams between the dependent and each of the
73
-------
independent variables. The form of the scatter would often provide clues to
transformations needed on the variables. These transformations were then tried
in regression. Where several transformations were tried on the same variable,
the criterion used to decide between the specifications was to choose the trans-
formation giving the highest level of significance; i.e., the largest t statis-
tic on the coefficient. Where this did not discriminate, the transformation
2
which yielded the highest coefficient of multiple determination (R ) was chosen.
The following variables in both models were transformed to natural logarithms:
the dependent variables, the air pollution variables (including all lags), and
the meteorological variables (including all lags). The meteorological variable
was respecified from degree days to a Temperature Humidity Index (T.H.I.).
Degree days was dropped from the models because the coefficients were not statis-
tically significant and had negative instead of the expected positive signs.
The discussion on the meteorological variable in Chapter IV demonstrated there
are many meteorological conditions which can affect a state of health. A priori,
it is difficult to identify the meteorological conditions which will affect a
state of health within a given region; i.e., different meteorological variables
must be tried. Degree days was included within the model because it was a
variable which took account of extremes of temperature at both spectrums. Prob-
ably the prime reason the variable was not significant was the lack of varia-
tion exhibited in the variable. For example, on the 10 percent sample, degree
days (unlagged) had a mean of 13.65 and a standard deviation of 8.89. This
would tend to indicate that temperature extremes in Portland were not severe
enough to affect states of health.
Temperature and humidity have been associated with both cardiovascular and res-
piratory problems. The T.H.I, allows both conditions to be expressed as one
measure. Originally, the main purpose for developing the T.H.I, was to have a
simple method of determining the effect of summer conditions in the United
States on human comfort [51, p. 41]. There are several ways to calculate the
T.H.I. The following formula [51, p. 41] was used in this study: »
T.H.I. - 0.4(Td + TW) + 15 (9)
74
-------
where Td = dry-bulb (air) temperature in °F.
T = wet-bulb temperature in °F.
Vw
Higher values of the T.H.I, mean increased discomfort to the patient. Empirical
studies in the United States have shown that 10 percent of the general popula-
tion becomes uncomfortable when T.H.I, reaches a value of 70. Fifty percent of
the population becomes uncomfortable when the T.H.I, reaches 75 [51, p. 42],
For patients susceptible to circulatory and respiratory problems, higher T.H.I.
values could aggravate existing health conditions and result in increased con-
sumption of medical services. Hence, it is expected the coefficient of this
variable will be positively related to the dependent variable.
The respecified statistical models resulting from the pretest are given below.
Model 3 has been respecified as Model 10, and Model 5 has been respecified as
Model 11.
ln Y * In3 + 6X + 62X2ijk + 33X3ijk + 34X4ijk
65X5ijk + 66X6ijk + 37X7ijk + 38X8ijk + 39X9ijk
310X10ijk + 3llXllijk + 312X12ijk + 313lnX13ijk
+ 314lnX14ijk + Eijk
where X,,. ., = a measure °f the meteorological conditions that the
k — patient, who consumes medical services for disease
i, is exposed to on day j , expressed as a Temperature
Humidity Index; and
(1-1, ••', n); (J - 1, •••, m); (k = 1, •••,£),
Atmospheric humidity can be measured by use of a mercury thermometer with a
moistened wick surrounding the mercury reservoir. With adequate ventlation,
the wick's temperature is lowered by evaporation to a constant value, which
is called the wet-bulb temperature. This, together with air temperature and
suitable tables, can be used to determine dew point or relative humidity
[36, p. 7],
The only meteorological station in Portland which measured wet-bulb temperature
was at the Portland International Airport. Hence, both wet-bulb and dry-bulb
temperatures used in calculating the T.H.I, were measured at the Portland Air-
port and inferred as being representative of the entire study area.
75
-------
and
Xlijk» "•' X13ijk; 30' •"• 313; and £ijk are M Previously defined
in Model 3.
Model 5 Is respecified as follows:
where ^-in- • = t*ie avera8e measure of meteorological conditions for
lOicj patients consuming outpatient medical services for
disease i from census tract c on day j, expressed as
a Temperature Humidity Index; and
(i - 1, •••, n) (c = 1, •••, p); (j = 1, •••, m),
£icj «" as pre"
viously defined in Model 5.
There were several results using the pretest which influenced the testing and
respecification of the statistical models. Statistical tests were conducted to
determine whether an interaction existed between the air pollution and meteoro-
logical variables. The interaction variable was included in the model separately
from, and with, variables measuring meteorological and air pollution phenomena.
The tests were not statistically significant.
Two air pollution variables were examined during the pretest. One variable was
composed only of the patient's estimated exposure to air pollution at his place
of residence. The second variable tested was weighted, as discussed in Chapter
VIII, to reflect exposure to air pollution at place of residence and work. The
weighted air pollution variable proved to be statistically more significant.
It was felt that if a relationship between air pollution and the consumption of
medical services were to exist, then the association would most likely be re-
flected through medical services used to treat respiratory and circulatory ill-
nesses (Categories I and IV in Table 1, Chapter VI). Air pollution was found
76
-------
to be statistically significant at the 5 percent level for respiratory diseases
analyzed alone, and for circulatory-respiratory diseases analyzed together. Air
pollution was not statistically significant when circulatory diseases were analy-
zed alone. For this reason, only respiratory and respiratory-circulatory dis-
eases together were analyzed in the respecified models using all the data.
In the pretest regression on Model 11, air pollution was not statistically signifi-
cant in affecting the consumption of outpatient medical services used to treat
2
respiratory illnesses. The highest R occurred when the air pollution and
meteorological variables were lagged one day. Hence, given a severe budget
constraint on the funds available for computer use, this model was not extended
to analyze other disease categories.
It was expected there might be a problem with autocorrelation in the analysis.
However, a plot of the residuals indicated that such a problem did not exist for
either of the disease categories tested.
The results of the pretest regressions may be seen in Appendix Tables B-l, B-2,
and B-3. Attention is now turned to discussing the analysis associated with the
respecified statistical models using all of the data.
The Regression Results; Statistical Model 10
The respecified statistical models were tested for each of the selected disease
categories. To be more specific, Model 10 was tested with respiratory diseases
alone and with circulatory-respiratory diseases combined. The regression results
for each disease category can be seen in Tables 7 and 8. Submodels 1 through 4
indicate the lags for the air pollution and meteorological variables. Submodel
1 is not lagged; Submodel 2 is lagged by one day; Submodel 3 is lagged by two
days; and Submodel 4 is lagged by three days.
Before proceeding directly to the regression results, it is best to summarize
again the characteristics assumed inherent in the dependent variable (Vi1k) of
Model 10. All the outpatient medical services used to treat a single disease
77
-------
Table 7. REGRESSION RESULTS OF STATISTICAL MODEL 10s RESPIRATORY DISEASES
Variables
Sample size
Constant
Age
Sex
Marital status
No. of people
in household
Household income
Race-Negro
Race - other
non-white
Physically fit
Somewhat phys-
ically fit
Drinking
Smoking index
Occupational ex-
posure index
Air pollution
Temp, humidity
index
Outpatient medi-
cal services
R2
Submodel 1 -
Coefficients-''
1700
4.0995E-01
9.2984E-04
(1.7641E-03)
-9.2609E-02
(4.4077E-02)
4.9017E-02
(7.0571E-02)
-2.2342E-04
(1.6566E-02)
1.3839E-06
(3.2721E-06)
-1.2499E-01
(1.0598E-01)
1.1757E-01
(1.8843E-01)
-3.6543E-02
(7.1782E-02)
-6.6018E-02
(5.5690E-02)
1.4501E-03
(6.4242E-04)
1.1919E-04
(9.8823E-05)
1.3806E-06
(4.5747E-03)
8.1786E-02
(3.6601E-02)
3.9050E-01
(1.5010E-01)
2.1991E-02
• unlagged
t statistics
0.5271
2.1011
0.6946
0.0135
0.4229
1.1793
0.6240
0.5091
1.1855
2.2573
1.2061
0.3018
2.2346
2.6017
Mean^7
46.0506
(16.0231)
0.6818
(0.5537)
0.8424
(0.3645)
3.3906
(1.7783)
12085.4412
(7525.6879)
0.0524
(0.2228)
0.0159
(0.1251)
0.1753
(0.3803)
0.5624
(0.4962)
12.4064
(37.9688)
152.0516
(247.8353)
3.4162
(5.3893)
61.2532
(34.8633)
53.0533
(8.0905)
14 .'75 39
(12.0912)
Submodel 2 - lagged 1 day
a/
Coefficients— t statistics
1700
1.7224E-01
9.4703E-04 0.5373
(1.7625E-03)
-9.1826E-02 2.0851
(4.4040E-02)
5.0257E-02 0.7127
(7.0520E-02)
-5.5410E-04 0.0335
(1.6546E-02)
1.3071E-06 0.3999
(3.2683E-06)
-1.2354E-01 1.1669
(1.0587E-01)
1.1621E-01 0.6177
(1.8814E-01)
-3.6802E-02 0.5132
(7.1711E-02)
-6.7155E-02 1.2070
(5.5637E-02)
1.4583E-03 2.2721
(6.4184E-04)
1.4583E-03 2.2721
(9.8761E-05)
1.5975E-03 0.3497
(4.5681E-03)
8.4906E-02 2.3605
(3.5969E-02)
4.4765E-01 3.0250
(1.4798E-01)
2.3885E-02
Meai£7
£/
£/
£/
£/
£/
£/
£/
£/
£/
£/
£/
£/
61.7633
(35.2088)
52.9020
(8.1066)
(continued)
78
-------
Table 7. (CONTINUED)
Submodel 3 - lagged 3 days
Variables
Sample size
Constant
Age
Sex
Marital status
No. of people
in household
Household income
Race - Negro
Race - other
non-white
Physically fit
Somewhat phys-
ically fit
Drinking index
Smoking index
Occupational ex-
posure index
Air pollution
index
Temp, humidity
index
Outpatient med-
ical services
R2
Coefficients^
1700
3.2088E-01
9.8659E-04
(1.7643E-03)
-9.3276E-02
(4.4085E-02)
4.6437E-02
(7.0569E-02)
-4.1014E-04
(1.6564E-02)
1.2893E-06
(3.2721E-06)
-1.2677E-01
(1.0600E-01)
1.2076E-01
(1.8824E-01)
-3.2048E-02
(7.1742E-02)
-6.5781E-02
(5.5702E-02)
1.4623E-03
(6.4238E-04)
1.2433E-04
(9.8795E-05)
1.7257E-03
(4.5709E-03)
6.4037E-02
(3.6881E-02)
4.3036E-01
(1.4653E-01)
2.1868E-02
t statistics Mean—
0.5592 £/
2.1158 £/
0.6580 £/
0.0248 £/
0.3940 £/
1.1959 £/
0.6415 £/
0.4467 £/
1.1809 £/
2.2763 £/
1.2585 £/
0.3776 £/
1.7363 62.1665
(35.1278)
2.9370 52.9290
(8.1986)
Submodel 4 - lagged 4 days
Coefficients^
1700
2.8884E-01
1.0105E-03
(1.7641E-03)
-9.6809E-02
(4.4086E-02)
4.3497E-02
(7.0582E-02)
2.3213E-04
(1.6565E-02)
1.1461E-Q6
(3.2724E-06)
-1.2602E-01
(1.0601E-01)
1.2032E-01
(1.8821E-01)
-3.1869E-02
(7.1726E-02)
-6.3444E-02
(5.5724E-02)
1.4348E-03
(6.4241E-04)
1.2607E-04
(9.8778E-05)
1.9866E-03
(4.5709E-03)
4.6301E-02
(3.6696E-02)
4.5616E-01
(1.4446E-01)
2.1725E-02
t statistics Mean—
0.5728 £/
2.1959 £/
0.6163 £/
0.0140 £/
0.3502 £/
1.1887 £/
0.6393 £/
0.4443 £/
1.1385 £/
2.2334 £/
1.2763 £/
0.4346 £/
1.2617 62.1979
(35.2225)
3.1577 53.0916
(8.3442)
—' The standard error of the regression coefficient is in parentheses. E-01, E 01, E 00, etc., indi-
cate that the decimal place of the number is to be shifted to the left by one place; shifted to
the right by one place; or not shifted at all, respectively.
—/ Means are untransformed; their standard deviations are in parentheses.
—* The mean for this variable is the same as its mean in Submodel 1. Only the name of the lagged
variables changed per model.
79
-------
Table 8. REGRESSION RESULTS OF STATISTICAL MODEL 10: CIRCULATORY-RESPIRATORY DISEASES
Variables
Sample size
Constant
Age
Sex
Marital status
No. of people
in household
Household income
Race - Negro
Race - other
non-white
Physically fit
Somewhat phys-
ically fit
Drinking index
Smoking index
Occupational ex-
posure index
Air pollution
index
Temp, humidity
index
Outpatient med-
ical services
R2
Submodel 1 •
Coefficients—
3363
1.3190E 00
-1.6382E-03
(1.2708E-03)
2.4635E-04
(2.9157E-02)
2.9929E-03
(4.4089E-02)
1.7845E-03
(1.3087E-02)
-1.1933E-06
(2.2427E-06)
-9.6164E-02
(8.9548E-02)
8.0489E-02
(1.2691E--01)
-6.2167E-02
(4.5984E-02)
-6.0504E-02
(3.6337E-02)
7.9656E-04
(4.0020E-04)
6.0772E-05
(6.6437E-05)
6.2953E-03
(3.4506E-03)
2.1134E-02
(2.5430E-02)
2.4138E-01
(1.0057E-01)
8.0063E-03
- unlagged
T /
t statistics Hear.—
1.2891 54.3717
(16.2918)
0.0084 0.6375
(0.5613)
0.0679 0.8112
(0,3914)
0.1364 2.8944
(1.6443)
0.5321 10765.5367
(7/81.8663)
1.0739 0.0324
(0.1771)
0.6342 0.0167
(0.1280)
1.3519 0.1861
(0.3893)
1.6651 0.5031
(0.5001)
1.9904 10.8700
(40.2684)
0.9147 153.9242
(247.1794)
1.8244 2.6489
(4.8684)
0.8311 62.3909
(34.3420)
2.4001 54.3396
(8.1619)
13.9063
(12.6867)
Submodel 2 - lagged 1 day^
a/
Coefficients— t statistics
3363
1.2441E 00
-1.6180E-03 1.2742
(1.2698E 00)
1.3022E-04 0.0045
(2.9151E-02)
2.8894E-03 0.0655
(4.4082E-02)
1.8327E-03 0.1401
(1.3082E-02)
-JU1682S-06 0.5207
(2.2433E-06)
-9.6459E-02 1.0774
(8.9532E-02)
3.1491E-02 0.6425
(1.2684E-01)
-6.2172E-02 1.3523
(4.5976E-02)
-6.0654E-02 1.6696
(3.6328E-02)
7.9294E-04 1.9818
(4,(1011E-02)
6.0766E-05 0.9148
(6.6422E-05)
6.3640E-03 1.8453
(3.4489E-03)
1.9713E-02 0.7881
(2.5015E-02)
2.6135E-01 2.6355
(9.9166E-02)
8.3310E-03
Mean^-
£/
£/
£/
£/
£/
£/
£/
£/
sJ
sJ
sJ
£/
62.6497
(34.6665)
54.2546
(8.2262)
«i
(continued)
80
-------
Table 8. (CONTINUED)
Submodel 3 - lagged 3 days
Variables
Sample size
Constant
Age
Sex
Marital status
No. of people
In household
Household income
Race - Negro
Race - other
non-white
Physically fit
Somewhat phys-
ically fit
Drinking index
Smoking index
Occupational ex-
posure index
Air pollution
index
Temp, humidity
index
Outpatient med-
ical services
R2
Coefficients^'
3363
1.4245E 00
-1.5595E-03
(1.2698E-03)
-9.2519E-04
(2.9156E-02)
2.6729E-03
(4.4103E-02)
1.9317E-03
(1.3086E-02)
-1.2578E-06
(2.2455E-06)
-9.6901E-02
(8.9581E-02)
8.3529E-02
(1.2687E-01)
-6.1298E-02
(4.5995E-02)
-6.1317E-02
(3.6338E-02)
7.9628E-04
(4.0027E-04)
6.2927E-05
(6.6441E-05)
6.3350E-03
(3.4490E-03)
-1.3033E-03
(2.5491E-02)
2.3656E-01
(9.8931E-02)
7.6485E-03
t statistics
1.2281
0.0317
0.0606
0.1476
0.5601
1.0817
0,6584
1.3327
1.6874
1.9894
0.9471
1.8368
0.0511
2.3912
Mean^
£/
£/
£/
£/
£/
£/
£/
£/
£/
£/
£/
£/
62.8473
(34.6274)
54.2279
(8.2513)
Submodel 4 - lagged 4 days
Coefficients^'
3363
1.3690E 00
-1.5932E-03
(1.2701E-03)
-1.7524E-03
(2.9143E-02)
1.4193E-03
(4.4101E-02)
2.0741E-03
(1.3083E-02)
-1.3074E-06
(2.2458E-06)
-9.6205E-02
(8.9583E-02)
8.1589E-02
(1.2684E-01)
-6.0687E-02
(4.5995E-02)
-6.0791E-02
(3.6335E-02)
7.9438E-04
(4.0019E-04)
6.1176E-05
(6.6455E-05)
6.3415E-03
(3.4484E-03)
-8.4495E-03
(2.5556E-02)
2.5834E-01
(9.9073E-02)
7.9576E-03
t statistics Meai£'
1.2544 £/
0.0601 c/
0.0322 £/
0.1585 c/
0.5821 £/
1.0739 £/
0.6432 £/
1.3194 cj
1.6731 £/
1.9850 £/
0.9206 £/
1.8390 £/
0.3306 62.7691
(34.5243)
2.6076 54.3663
(8.2780)
— The standard error of the regression coefficient is in parentheses. E-01, E 01, E 00, etc., Indi-
cate that the decimal place of the number is to be shifted to the left by one place; shifted to
the right by one place; or not shifted at all, respectively.
—' Means are untransformed; their standard deviations are in parentheses.
— The mean for this variable is the same as its mean in Submodel 1. Only the name of the lagged
variables changed per model.
81
-------
incident per contact with the medical system were expressed in California Rela-
tive Value units. The units were transformed to dollar values. Hence, the dol-
lar values would reflect the quantity of medical services consumed per contact.
This section will discuss the regression results of Model 10 by independent vari-
able. The regression results for the respiratory disease model will be analyzed
first (table 7), then the results of the circulatory-respiratory diseases com-
bined will be discussed (table 8).
Age of the Patient (
Respiratory Diseases - The estimated coefficients were not significantly differ-
ent from zero and, hence, do not appear to influence the consumption of out-
patient medical services per contact with the Kaiser system. The average age
of the patient consuming outpatient medical services for respiratory diseases
was 46 years.
Circulatory-Respiratory Diseases - The estimated coefficients for age were signifi-
cantly different from zero at the 20 percent level for Submodel 1 in Table 8.
The coefficients in the other models were not statistically significant.
The signs of the coefficients in Model 10 were negative. The hypothesized sign
for this variable was positive, i.e., the older the patient becomes, the more
medical services he consumes. The negative coefficient indicates that he con-
sumes fewer medical services.
Two explanations can be given for the negative sign on the coefficient. One,
since the variable is barely significant at 20 percent, the results should be
discounted; i.e., there is a one in five chance of committing a Type I error on
the statistical test. However, the average age for patients consuming outpatient
medical services for circulatory-respiratory illnesses is 54 years; for respira-
tory diseases it was 46. This indicates that an older group of patients are
Because of the nature of the study and the type of data being used, it was de-
cided that any coefficient not significantly different from zero at the 20
percent level would be considered not statistically significant.
82
-------
obtaining treatment for circulatory-respiratory problems. Hence, the second
reason the age coefficient may be negative is as follows: Most circulatory-
respiratory diseases in older people may be severe enough to require hospitaliza-
tion. Therefore, fewer medical services are consumed on an outpatient basis.
Younger patients, with a probable better state of health, would most likely not
be hospitalized. Treatment of the diseases for the younger patients would be at
a clinic and, proportionately, they would consume more outpatient medical ser-
vices per visit.
Sex of the Patient (X_ .,)
Respiratory Diseases - The estimated coefficients (for all lags) of this variable
were significantly different from zero at the 5 percent level in Model 10. The
signs of the coefficients were negative, which indicated that women consumed fewer
medical services than men. This is not unexpected when one considers that the
incidence of many respiratory diseases are lower for women than men.
Circulatory-Respiratory Diseases - The regression coefficients were not statistic-
ally significant. This would tend to indicate that, at least with circulatory-
respiratory diseases combined, sex does not have a discriminating role in the con-
sumption of outpatient medical services. In all the disease categories, over 60
percent of the patients consuming outpatient medical services were women.
Marital Status (X- ., )
JJ.J K
Respiratory Diseases - The coefficients of the marital status variable were not
statistically significant. This implies that marital status has no effect on the
consumption of medical services. Over 84 percent of the patients receiving out-
patient medical services were married.
Circulatory-Respiratory Diseases - As with respiratory diseases, the marital status
coefficients were not statistically different from zero. Over 81 percent of the
This is not to argue that older patients may not visit a clinic more than younger
patients. It is argued that, proportionately, they may consume fewer medical
services per visit.
83
-------
patients receiving treatment for circulatory-respiratory diseases were married.
Number of People in the Patient's Household
Respiratory Diseases - The estimated coefficients of this variable were not sta-
tistically significant. Although the average number of people within the patient's
household was 3.39, this appeared not to influence the consumption of outpatient
medical services per visit to the clinic. However, while crowding appears to have
no effect on the severity of the disease state, it may still affect the number
of visits to the clinic. The latter point was not tested by this model.
Circulatory- Respiratory Pis eases - Since crowding had no effect on the consumption
of medical services used in treating respiratory diseases, it was not surprising
to find it statistically insignificant with circulatory-respiratory diseases.
The average number of people in the patient's household who were suffering from
circulatory- respiratory diseases was 2.89. Once more this indicates that an older
set of people were being analyzed in the circulatory- respiratory disease category.
Household Income
Respiratory Diseases - Household income was included in this study in an attempt
to describe the patient's ability to make expenditures in lieu of consuming medi-
cal services which would preserve or improve his state of health. Examples of
such expenditures would be those made for sound housing and nutritional food.
The coefficients of the household income variables were not statistically signi-
ficant in any of the submodels of 10.
It would appear, at least for outpatient medical services, that household income
does not play a significant role in this respect. One reason for this, perhaps,
is the availability of income supplements within the Portland area to low- income
families. These would include welfare, availability of surplus foods, and/or
food stamps. The income supplements, and the opportunity for all Kaiser members
to consume the same kinds of medical services, would tend to minimize any differ-
ences in the consumption of medical services as between families having differ-
ent income levels. The mean household income of patients consuming medical ser-
vices for respiratory illnesses was $12,085.
84
-------
Circulatory-Respiratory Diseases - The coefficients were not significantly dif-
ferent from zero. The reasons for the nonsignificance are the same as those given
above for the respiratory diseases.
The household income for patients consuming outpatient medical services for cir-
culatory-respiratory illnesses was $10,766. Again, this reflects the older age
group being examined in this disease category. Based on age, family size, and
income, it would appear that more patients in the circulatory-respiratory cate-
gory are retired than is the case for those afflicted with respiratory diseases
only.
Race (X6i1k' X7ijk>
Respiratory Diseases - The coefficients of the two race variables were not sta-
tistically significant. This would imply that race is not important in the con-
sumption of outpatient medical services as between Negroes (X,..,), other non-
ul J iC
whites (Xyj-fi,)* ^d whites. Of the contacts with the Kaiser system which re-
sulted in outpatient medical services being consumed, approximately 5.2 percent
of the contacts were by Negroes, 1.6 percent by other non-whites, and 93.2 per-
cent by whites.
Circulatory-Respiratory Diseases - Neither of the race coefficients were statis-
tically significant. Of the contacts with the Kaiser system resulting in the
consumption of medical services, approximately 3.2 percent of the contacts were
by Negroes, 1.7 percent by other non-whites, and the other 95.1 percent by whites.
Physical Fitness (X8ijk» X9jHk>
Respiratory Diseases - Physical fitness was expressed by two qualitative vari-
ables. Xfl.., represented "a great deal of time and energy expended by the patient
OIJ K
in being physically fit;" X , represented "some time and energy expended by
yx j K
the patient in being physically fit." Statistically these were tested against
the third category, expressed as "very little time and energy expended being
physically fit." It was expected that the more physically fit patient would con-
sume fewer medical services.
85
-------
With respiratory diseases, the coefficients were not significant at the 20 per-
cent level. It would appear that physical fitness (as measured here) does not
influence the consumption of outpatient medical services for respiratory dis-
eases .
Of the contacts with the Kaiser system which resulted in the consumption of
medical services, 17.53 percent indicated they spent a great deal of time and
energy being physically fit; 56.24 percent indicated they spent some time and
energy being physically fit; 26.23 percent spent very little time and energy
being physically fit.
Circulatory-Respiratory Diseases - The coefficient of Xgi fc was significantly
different from zero at the 20 percent level in all submodels. XQ. ., was statis-
tically significant at the 10 percent level in all submodels. The signs on both
of the coefficients were negative, as expected. The coefficient of Xg. .- was
larger than Xg , (-0.0622 versus -0.0605, respectively). This would imply that
more physically fit patients consume fewer medical services. There is a sus-
pected association between lack of physical exercise and coronary heart disease
[20, p. 15], and lack of exercise and circulatory problems in general. Hence,
with circulatory-respiratory diseases it was not a surprise that the physical
fitness variables were statistically significant, but not statistically signifi-
cant when the respiratory diseases were analyzed alone.
Of the patients contacting the Kaiser system and consuming outpatient medical
services for circulatory-respiratory diseases, 18.6 percent, 50.3 percent, and
31.1 percent indicated they spent a great deal, some, and very little time and
energy, respectively, being physically fit.
Drinking
Respiratory Diseases - The estimated coefficients for Xin , were significantly
JLUZLj 1C
different from zero at the 5 percent level in all submodels. The variable mea-
sured how many times a year a respondent drank six drinks, eight cans of beer,
or a bottle of table wine. The estimated coefficients of the variable had the
expected positive signs. This would tend to support the hypothesis that heavy
drinking can affect a state of health and influence the consumption of outpatient
medical services.
86
-------
The average number of times a year patients consumed alcohol excessively (as
defined in the paragraph above) was 12.4. The mean reflects many zero values
assigned to respondents who indicated they had not consumed alcoholic beverages
excessively during the 12 months prior to the household interview date. The
standard deviation of the mean was 37.9.
Circulatory-Respiratory Diseases - For all the submodels, the coefficients were
statistically significant at the 5 percent level. All the coefficients had the
expected positive signs. The average number of times per year patients indicated
they drank excessively was 10.9. The standard deviation for the mean was 40.3.
Cigarette Smoking Index (X,,..,)
Respiratory Diseases - While the coefficients had the right positive signs, none
were significantly different from zero at the 20 percent level. The coefficient
in Submodel 4 came closest to being significantly different from zero at the 20
percent level. The index took a value of zero if the patient indicated he had
never regularly smoked cigarettes in his life. Of the 2,439 respondents to the
household interview, 991, or 40.6 percent, indicated they had not smoked regu-
larly in the past. The mean smoking index for patients consuming medical ser-
vices was 152.1, with a standard deviation of 247.2.
If any disease category would have been most affected by cigarette smoking, res-
piratory diseases should have been. There were a number of potential problems
with the cigarette smoking index variable which could have resulted in its not
being significant. Perhaps not enough weight in the index was attached to the
cigarette smoking habits of former smokers. The variable could be respecified
to include only the number of cigarettes smoked per day, the number of years
the patient has smoked, or whether the patient has ever smoked or not.
It is also quite possible that patients suffering from diseases most affected
by cigarette smoking (chronic bronchitis and emphysema) would receive propor-
tionately more of their medical services in the hospital. This would particu-
larly be true in more severe disease cases. Also, once a patient has a dis-
ease affected by smoking, he may visit the doctor more often but the amount of
medical services consumed per visit would remain constant.
87
-------
Circulatory- Respiratory Diseases - The coefficients were not statistically signi-
ficant in any of the models. If the coefficients were not statistically signifi-
cant with respiratory diseases, the fact they were not significant with circula-
tory-respiratory diseases combined was not surprising. Again, the lack of sta-
tistical significance may be due to the same reasons indicated for the respira-
tory disease category above. The mean smoking index for patients consuming
medical services was 153.9, with a standard deviation of 247.2.
Occupational Exposure Index
Respiratory Diseases - This variable was expressed as an index. A value of zero
was assigned to patients who did not work. Values greater than zero were assigned
to each working patient, based upon the number of hours worked per day times a
weighting factor representing job-related exposure to particulates.
The coefficients were not significantly different from zero at the 20 percent
level. This is not surprising given the fact that 23 percent of the subscribers
in the 5 percent sample were employed in clerical-sales [17, p. 301], resulting
in a low occupational exposure index. Also, approximately 39 percent of the
respondents to the questionnaire did not work at all. The mean occupational ex-
posure index of those seeking medical services was 3.4; the standard deviation
was 5.4.
Circulatory-Respiratory Diseases - The coefficients were statistically significant
at the 10 percent level. The signs of the coefficients were positive, as expected.
Why the coefficients would be statistically significant for circulatory-respira-
tory diseases, and not for respiratory diseases alone, is not known. The mean
occupational exposure index for those consuming medical services was 2.6; the
standard deviation was 4.9.
Air Pollution
Respiratory Diseases - The coefficients of the air pollution variables were signi
ficantly different from zero for each of the submodels in Table 7, as follows:
Submodel 1 - significantly different from zero at the 5 percent level;
Submodel 2 - significantly different from zero at the 2 percent level;
88
-------
Submodel 3 - statistically significant at the 10 percent level;
Submodel 4 - not statistically significant at the 20 percent level.
All coefficients had positive signs, as expected. The null hypothesis that
deterioration in air quality causes no increase in the consumption of medical
services per outpatient contact with the medical system is rejected.
The highest level of statistical significance for the air pollution coefficient
occurred in Submodel 2, where the variable was lagged by one day. This model
2
also had the highest R of the four models. This would tend to confirm the
belief expressed earlier that there is a delay between exposure to air pollution
and contact with the medical system.
Using the coefficients from each of the submodels in which they were statistically
significant, a 20 microgram increase in air pollution, from 60 to 80 micrograms
per cubic meter, would increase outpatient medical costs per contact as follows:
Submodel 1 - an increase of 3.3 cents;
Submodel 2 - an increase of 3.5 cents;
Submodel 3 - an increase of 2.4 cents.
The average exposure to air pollution (lagged one day) of patients seeking out-
patient medical services was 61.8 micrograms per cubic meter. The standard de-
viation was 35.2. The means and standard deviations for each of the variables
may be seen in Table 7.
Circulatory-Respiratory Diseases - The air pollution coefficients were not sta-
tistically significant. The null hypothesis that deterioration in air quality
causes no increase in the consumption of medical services per outpatient con-
tact with the medical system is not rejected.
One reason, perhaps, why air pollution is not statistically significant is that
many circulatory problems may be severe enough to require hospitalization. Any
outpatient medical services consumed for treatment of circulatory problems would
be routine, thereby limiting the amount of medical services consumed. The aver-
age unlagged exposure to air pollution for patients obtaining medical care was
89
-------
62.4 micrograms. The standard deviation about the mean was 34.3. The means and
standard deviations for each of the other lags may be seen in Table 8.
Meteorological Conditions (xiA.t.i,)
Respiratory Diseases - The estimated coefficients for the T.H.I, variable were
statistically significant at the 1 percent level for all the models in Table 7,
with the most significance occurring when the variable was lagged three days.
All of the coefficients had the expected positive signs. This would tend to
support the hypothesis that the combined factors of temperature and humidity
influence the consumption of medical services.
The mean T.H.I, value (unlagged) of those consuming outpatient medical services
for respiratory diseases was 53.0, with the standard deviation being 8.2. The
mean does not compare to the empirical evidence (previously cited) establishing
a T.H.I, of 70 as causing 10 percent of the population to become uncomfortable.
However, since this study is dealing with respiratory diseases, one would ex-
pect the patients to be considerably more susceptible to temperature and humidity
than would members of the general population. This would imply that the T.H.I.
values for these patients could be lower than those cited in the empirical stud-
ies, still cause discomfort to the patient afflicted with respiratory diseases
and, hence, precipitate an increase in the consumption of outpatient medical ser-
vices used to treat the diseases.
Circulatory-Respiratory Diseases - The coefficients of the T.H.I, variable were
significantly different from zero in each of the models in Table 8, as follows:
Submodel 1 - significantly different from zero at the 2 percent level;
Submodel 2 - significantly different from zero at the 1 percent level;
Submodel 3 - statistically significant at the 2 percent level;
Submodel 4 - statistically significant at the 1 percent level.
The signs of the coefficients were positive, as expected.
The mean T.H.I, value (lagged two days) for patients suffering from circulatory-
respiratory diseases was approximately 54.2, with a standard deviation of 8.3.
90
-------
The same argument applies to circulatory-respiratory diseases as was used with
the respiratory diseases above. One would expect that patients suffering from
circulatory-respiratory illnesses would be more susceptible to temperature and
humidity and, hence, consume more medical services.
Summary
Respiratory Diseases - The following coefficients of the variables were signifi-
cantly different from zero in Model 10 at 20 percent or higher: Sex (X? ) ;
*• J-J K
Drinking habits (x10i'tJ; Air pollution (X13±.k)» except in Submodel 4 of (10);
and Meteorological conditions - T.H.I. (X., ., ). The overall F tests for regres-
J.'HjK
sion on all four models were statistically significant at the 1 percent level.
Circulatory-Respiratory Diseases - In Model 10 the following coefficients were
statistically significant at 20 percent or higher: Age (Xi.!! significant only
in Submodel 1 and with a negative sign; Physical fitness (Xg ., and Xq . .. ) ;
Drinking habits (X._ ,); Occupational exposure (Xio--^* Meteorological condi-
tions - T.H.I. (X,,. .,). The overall F tests for regression on all four models
were statistically significant at the 5 percent level.
The Regression Results; Statistical Model 11
Model 11 was specified in an attempt to determine whether air pollution has an
effect on the number of contacts with the medical system where the cost per con-
tact may be more or less constant. If the hypothesis that air pollution has no
effect on the consumption of medical services per contact with the medical system
were accepted, then it was felt the effects of air pollution on the number of
contacts with the system could be tested indirectly with (11). The main fault
with (11) was that only socioeconomic-demographic characteristics of patients
contacting the Kaiser system were accounted for. That is, the socioeconomic-
demographic characteristics of those who were at risk (at risk to contract a
disease and use the Kaiser facilities) in each census tract were not taken into
account within the independent variables unless they contacted the Kaiser system.
As a result of the pretest regressions, only respiratory diseases were included
in the analysis. Given a severe budget constraint on the funds available for
91
-------
computer use, Model 11 was tested using only a one-day lag. The regression re-
sults of these models may be seen in Table 9, which also contains the means and
standard deviations of the variables used in the model.
The following variables were statistically different from zero at the 20 percent
level or higher in statistical Model 11: Sex (X2ic.); Household income (X5ic.j),
which had a positive relationship with the dependent variable instead of the
negative relationship hypothesized; Drinking habits (X7ic-s)» and Smoking (X8ic-j)'
The overall F test for regression was statistically significant at the 5 percent
level.
The results of Model 11 should be interpreted with caution. There were too few
multiple observations per census tract per day resulting in the consumption of
outpatient medical services. Table 7 shows that 1700 patients with respiratory
illnesses contacted the Kaiser system and consumed outpatient medical services.
Table 9 shows that even after aggregation over the census tract for the two-year
study period, there were still 1,569 observations that entered regression. This
means that during the study period only 131 observations came from the same cen-
sus tract on the same day; the other 1,569 observations were single observations
per census tract. In most instances, then, the independent variables were indi-
vidual patient characteristics and not aggregate data as originally specified.
Hence, a lack of a sufficient number of observations per census tract prohibited
testing whether air pollution had an effect on the number of contacts with the
medical system.
Statistical Results; A Discussion
The statistical results of the air pollution coefficients in Model 10 (respira-
tory diseases) tend to lead to a rejection of the null hypothesis that deteriora-
tions in air quality cause no increase in the consumption of medical services
(per contact with the medical system) to treat respiratory diseases. Also, the
results would tend to confirm that there is a delay between exposure to air
pollution and meteorological conditions and contact with the medical system.
Using the best prediction model - Submodel 2 - an increase in air pollution by
20 micrograms (from 60 to 80 micrograms per cubic meter) would result in an esti-
mated 3.5 cent increase in outpatient medical costs (for respiratory diseases)
per contact with the medical system.
92
-------
Table 9. REGRESSION RESULTS OF STATISTICAL MODEL 11:
RESPIRATORY DISEASES - ONE DAY LAG
Variables
Sample size
Constant
Age
Sex
Marital status
No. of people in household
Household income
Race - non-white
Drinking
Smoking index
Occupational exposure index
Air pollution
Temperature humidity index
Per capita costs
R2
a/
Coefficients—
1569
-2.4077E 00
1.0900E-03
(2.2164E-03)
9.6563E-04
(5.9370E-04)
-4.7287E-04
(8.9162E-04)
3.5658E-03
(2.0714E-02)
6.9822E-06
(4.1046E-06)
3.1119E-04
(1.1727E-03)
1.0742E-03
(7.8481E-04)
1.7571E-04
(1.2237E-04)
4.0458E-03
(5.7227E-03)
-4.7287E-04
(8.9162E-04)
3.1119E-04
(1.1727E-03)
1.1830E-02
t Statistics
0.4918
1.6265
0.5303
0.1722
1.7011
0.2654
1.3687
1.4359
0.7070
0.5303
0.2654
Mean*-7
46.0153
(16.0014)
65.6474
(51.8897)
84.4806
(36.1985)
3.3996
(1.7804)
12112.6119
(7540.2050)
7.0746
(25.6481)
12.8749
(39.1389)
154.0181
(250.5819)
3.4413
(5.4109)
61.7121
(34.9227)
53.0149
(8.0738)
1.6083
(2.4151)
a/
— The standard error of the regression coefficient is in parentheses. E-01,
E 01, E 00, etc., indicate that the decimal place of the number is to be
shifted to the left by one place; shifted to the right by one place; or
not shifted at all, respectively.
— The standard deviation about the mean is in parentheses; means are untrans-
formed.
93
-------
The regression results indicate that the procedures used in the study hold prom-
2
ise for quantifying the medical costs of air pollution. The R 's for each regres-
sion are low; however, this is not surprising. While the X fc's may be the perti-
nent explanatory variables, their influence on Y.., may be weak compared to the
influence of the random disturbances. This particularly seems to be the case for
relationships describing household behavior that have been estimated from cross-
sections data [31, p. 234], Other health studies research using multiple regres-
2
sion have also been characterized by low R 's [33, 34, 35]. In a study of this
kind there is more concern about the reliability of the estimated structural
parameters than about the size of the coefficient of multiple determination
[5, p. 248].
The theoretical framework for this study, and the statistical models derived from
it, were a simplified description of a complex phenomenon. The consumption of
medical services is dependent not only on happenings in the present time period,
but also on occurrences which may have happened in the patient's past and, hence,
are difficult to determine and quantify.
As an economic cost it would appear that air pollution has a minimal effect on
increasing the quantity of outpatient medical services consumed in treating res-
piratory diseases. The reasons are that many such visits are routine and, while
air pollution may affect the number of visits, it does not dramatically affect
the cost per visit. This is important to know, because it substantially reduces
the need to do additional research on the effects of air pollution on outpatient
medical consumption, and suggests other areas in which research on the medical
costs of air pollution may be more productive. These recommendations for future
research have been previously addressed in Chapter II.
One cannot finish this section without a brief comment concerning the adequacy
of using the C.R.V.S. in the models as a proxy for the quantity of medical ser-
vices rendered, particularly as it refers to outpatient medical services. There
are a variety of different medical services that can be provided within a given
office call. To adequately reflect the multitude of potential outpatient medical
services, the C.R.V.S. (Kaiser adapted), while detailed, would have to be infi-
nitely more so than they are now. As a first step toward quantifying the output
of physician and associated medical personnel in rendering medical services, a
94
-------
detailed study of each office and medical procedure would have to be undertaken.
This would require a detailed manpower study. Some kind of weighting scheme
could then be developed for each medical procedure, reflecting the difficulty
of the tasks performed. While the C.R.V.S. is a good approximation of the quan-
tity of medical services consumed, it does have limitations which should be recog-
nized in its use.
95
-------
X. REFERENCES
1. American Association for the Advancement of Science, Air Conservation,
Air Conservation Commission. Baltimore: Horn-Shafer, 1965 (Publica-
tion No. 85).
2. Andersen, Ronald, and Lee Benham, "Factors Affecting the Relationship
Between Family Income and Medical Care Consumption." In: Empirical
Studies in Health Economics, edited by Herbert E. Klarman. Baltimore:
Johns Hopkins Press, 1970.
3. Auerbach, Oscar, and Steven M. Spencer, "New Hope for Heavy Smokers."
Readers Digest 96:129-131, Feb. 1970.
4. Berg, Neil J., and John I. Kowalczyk, "An Air Quality Index Designed
to Serve the Needs of a Regional Air Pollution Control Authority."
Paper presented at the annual meeting of the Pacific Northwest Inter-
national Section of the Air Pollution Control Association, Portland,
Oregon, November 23-25, 1969.
5. Brown, William G., and Farid Nawas, "Impact of Aggregation on the Esti-
mation of Outdoor Recreation Demand Functions." American Journal of
Agricultural Economics 55(2):246-249, 1973.
6. California Medical Association, Committee on Fees of the Commission on
Medical Services, 1964 California Relative Value Studies, 4th ed., San
Francisco: Six Ninety Sutter Publications, 1964.
7. Cassell, Eric J., "Epidemiology of Air Pollution in Urban Population."
Paper presented at Air Pollution and Respiratory Disease; A Progress
Report, St. Vincent's Hospital and Medical Center of New York, New York
City, Oct. 28, 1966.
8. , "The Health Effects of Air Pollution and Their Implications
for Control." Law and Contemporary Problems 33:197-216, 1968.
9. , e£ al., "Air Pollution, Weather, and Illness in a New York
Population." Archives of Environmental Health 18:523-530, 1969.
10. Committee of the Oregon Thoracic Society, Chronic Obstructive Pulmonary
Disease; A Manual for Physicians. Rev. ed., Portland, Oregon Thoracic
Society, 1966.
11. Dorland's Illustrated Medical Dictionary. 24th ed., Philadelphia: W. B.
Saunders Company, 1965.
12. Draper, N. R., and H. Smith, Applied Regression Analysis. New York:
Wiley, 1966.
13. Feldstein, Paul J., "The Demand for Medical Care." In: Report of the Com-
mission on the Cost of Medical Care. Chicago: American Medical Associa-
tion, Vol. I, 1964.
96
-------
14. Greenburg, Leonard, e£ al., "Intermittent Air Pollution Episodes in New
York City. 1962." Public Health Reports 78:1061-1064, 1963.
15. Greenlick, Merwyn R., and Ernest W. Saward, M.D., "Impact of a Reduced
Charge Drug Benefit in a Prepaid Group Practice Plan." Public Health
Reports 81:938-940, 1966.
16. Greenlick, Merwyn R., Principal investigator, Supplemental Progress Re-
port; Medical Care Utilization Project, June 1966 to August 1968. Kaiser
Foundation Research Institute. Research Grant Number 5 R01 CH 00225-03 HSR.
17. , e£ al., "Determinants of Medical Care Utilization." Health
Services Research Winter:296-315, 1968.
18. Griffiths, John F., Applied Climatology; An Introduction. London: Oxford
University Press, 1966.
19. Hagstrom, R. M., H. A. Sprague, and E. Landau, "The Nashville Air Pollu-
tion Study: VII. Mortality from Cancer in Relation to Air Pollution."
Archives of Environmental Health 15:237-248, 1967.
20. Hayman, Charles, M. S., and G. L. Dunnahoo, M.D., "Cardiovascular Disease:
A Public Health Problem." In: Cardiovascular Disease, New Perspectives;
Continuing Education in Public Health. San Francisco: Western Regional
Office, American Public Health Association, March 1962 (Continuing Educa-
tion Monograph No. 2).
21. Hodgson, Thomas A. Jr., "Short-Term Effects of Air Pollution and Mortality
in New York City." Environmental Science and Technology 4:589-597, 1970.
22. Hopkins, C. E., «£ ail., "Intrafamily Correlation and Its Significance in
the Interpretation of Sample Survey." American Journal of Public Health
53:1112-1120, 1973.
23. Hurtado, Arnold W., and Merwyn R. Greenlick, "Approaches and Techniques:
A Disease Classification System for Analysis of Medical Care Utilization,
With a Note on Symptom Classification." Health Services Research Fall:
235-250, 1971.
24. , The Clinical-Behavior Disease Classification Sys-
tem.Kaiser Foundation Hospitals, Portland, Oregon, n.d.
25. Jaksch, John A., and H. H. Stoevener, Effects of Air Pollution on Resi-
dential Property Values in Toledo, Oregon. Corvallis, Agricultural Ex-
periment Station, Oregon State University, September 1970.
26. Jaksch, John August, "Outpatient Medical Costs Related to Air Pollution
in the Portland-Vancouver Area." Unpublished Ph.D. thesis. Oregon State
University, Corvallis, 1972.
27. Joint Committee of the Allergy Foundation of America and the American
Thoracic Society, Asthma; A Practical Guide for Physicians. U.S.A.,
National Tuberculosis and Respiratory Disease Association in coopera-
tion with the Allergy Foundation of America, 1968.
97
-------
28. Kaiser, "Recording and Coding Instructions for Medical Care Utiliza-
tion Study." Rev. ed., unpublished. Kaiser Foundation Hospitals Medical
Care Research Unit, August 1968.
29. Kilburn, Kaye H., "Cilia and Mucous Transport as Determinants of the
Response of Lung to Air Pollutants." Archives of Environmental Health
14:77-91, 1967.
30. Kissam, Philip, Surveying for Civil Engineers. New York: McGraw-Hill,
1956.
31. Kmenta, Jan, Elements of Econometrics. New York: Macmillan, 1971.
32. Kutschenreuter, Paul Herbert, "A Study of the Effect of Weather on Mor-
tality in New York City." Master's thesis. New Brunswick, Rutgers,
1960. Cited in: Hodgson, Thomas A. Jr., "Short-Term Effects of Air
Pollution on Mortality in New York City." Environmental Science and
Technology 4:589-597, 1970.
33. Lave, Lester B. and Eugene Seskin, "Acute Relationships Among Daily
Mortality, Air Pollution, and Climate." A paper. Graduate School of
Industrial Administration, Carnegie-Mellon University, Pittsburgh,
September 1972.
34. , "Air Pollution and Human Health." Science
169:723-733, 1970.
35. Lave, Lester B., Samuel Leinhardt, and Martin B. Kaye, "Low-Level En-
vironmental Radiation and U.S. Mortality." A paper. Graduate School
of Industrial Administration, Carnegie-Mellon University, Pittsburgh,
July 1971.
36. Licht, Sidney (ed.), Medical Climatology, Vol. 8. New Haven: Elizabeth
Licht, 1964.
37. McCarroll, James R., et al., "Health and the Urban Environment. V. Air
Pollution and Illness in a Normal Urban Population." Archives of En-
vironmental Health 15:178-184, 1967.
38. Michelson, Irving, and Boris Tourin, "Comparative Method for Studying
Costs of Air Pollution," Public Health Reports 81:505-512, 1966.
39. National Center for Health Statistics, Eighth Revision International
Classification of Diseases, Adapted for Use in the United States. Wash-
ington, D.C., Public Health Service Publication No. 1693. 2 Vols. 1968.
40. National Center for Health Statistics, International Classifications
of Diseases, Adapted for Indexing Hospital Records by Diseases and
Operations, rev. ed., Washington, D.C., Department of Health, Education,
and Welfare, Public Health Service Publication No. 719, 2 Vols. 1962.
41. National Tuberculosis and Respiratory Disease Association, Air Pollution
Primer. New York, 1969.
98
-------
, Introduction to Respiratory Diseases. 2nd ed. , U.S.A.,
~~ - - -
1964.
43. Olsson, Lars E. , et al. , The Study of the Natural Ventilation of the
Columbia-Willamette Valleys ; II. Oregon State University, Corvallis,
Departments of Atmospheric Sciences and Oceanography and Air Resources
Center. Technical report no. 71-2, April 1971.
44. Oregon-Washington Air Quality Committee, Criteria for Particle Fallout
Objectives and Standards. Portland: Oregon State Sanitary Authority,
April 1969.
45. Pennington, Ralph H. , Introductory Computer Methods and Numerical Analy-
sis. New York: Macmillan, 1965.
46. Portland. In: The Encyclopedia Americana, International ed., Vol. 22.
New York, 1970.
47. Ridker, Ronald G. , Economic Costs of Air Pollution; Studies in Measure-
ment. New York: Frederick A. Praeger, 1967.
48. Robinson, Arthur H. , and Randall D. Sale, Elements of Cartography, 3rd
ed., New York: John Wiley and Sons, 1969.
49. Thompson, Donovan J., et^ al. , "Health and the Urban Environment: VIII.
Air Pollution, Weather, and the Common Cold." American Journal of Pub-?
lie Health 60:731-737, 1970.
50. Tromp, S. W. , "The Geographical Distribution of Arteriosclerotic Heart
Diseases in the Netherlands (Period 1936-1945)." Leiden, Bioclimatolog-
ical Research Center, Nov. 20, 1958. Cited in: Tromp, S. W. , Medical
Biometeorology; Weather, Climate, and the Living Organism. Amsterdam:
Elsevier, 1963.
51. _ , Medical Biometeorology; Weather, Climate, and the Living
Organism. Amsterdam: Elsevier, 1963.
52. U.S. Bureau of the Census, 1960, Census of Population, Alphabetical
Index of Occupations and Industries, rev. ed. , Washington, D.C., U.S.
Government Printing Office, 1960.
53. U.S. Bureau of the Census, 1970 Census of Population; General Social and
Economic Characteristics. Final Report PC(1)-C39 Oregon. Washington, D.C.,
U.S. Government Printing Office, 1972.
54a t Census Use Study; Admatch Users Manual. Washington, D.C.
1970
55. U.S. Department of Commerce, Environmental Data Service, Climatological
Data; Oregon, Vol. 73, No. 1 through Vol. 76, No. 12 (all months inclu-
sive from January 1967 through December 1970).
99
-------
56. , Environmental Data Service, Climatological Data; Washington,
Vol. 71, No. 1, through Vol. 74, No. 12 (all months inclusive from January
1967 through December 1970).
57. , Environmental Data Service, Local CHmatological Data; Port-
land^ Oregon International Airport. Washington, D.C., U.S. Government
Printing Office, 1967-1970.
58. , National Oceanic and Atmospheric Administration, Environmental
Data Service, Local Climatological Data; Annual Summary with Comparative
Data, Portland, Oregon. Washington, D.C., U.S. Government Printing Office,
1970.
59. U.S. Public Health Service, Consumer Protection and Environmental Health
Service, Air Quality Criteria for Particulate Matter. Washington, D.C.,
National Air Pollution Control Administration, January 1969 (National Air
Pollution Control Administration Publication No. AP-49).
60. . . , Consumer Protection and Environmental Health Service, Air
Quality Criteria for Sulfur Oxides. Washington, D.C. National Air Pollu-
tion Control Administration, January 1969 (National Air Pollution Control
Administration Publication No. AP-50).
61. , Division of Air Pollution, Air Pollution Measurements of the
National Air Sampling Network; Analysis of Suspended Particulates, 1957-
1961. Washington, D.C., Robert A. Taft Sanitary Engineering Center, 1962
(Public Health Service Publication No. 978).
62. , Environmental Health Service, Air Quality Criteria for Carbon
Monoxide. Washington, D.C., National Air Pollution Control Administration,
March 1970 (National Air Pollution Control Administration Publication
No. AP-62).
63. , Environmental Health Service, Air Quality Criteria for Hydro-
carbons . Washington, D.C., National Air Pollution Control Administration,
March 1970 (National Air Pollution Control Administration Publication
No. AP-64).
64. , Environmental Health Service, Air Quality Criteria for Photo-
chemical Oxidants. Washington, D.C., National Air Pollution Control Admin-
istration, March 1970) (National Air Pollution Control Administration
Publication No. AP-63).
65. U.S. Public Health Service, Health Services and Mental Health Administra-
tion, The Health Consequences of Smoking; A Report of the Surgeon General;
1971. Washington, D.C., U.S. Government Printing Office (Department of
Health, Education, and Welfare Publication No. HSM 71-7513).
66. Wall Street Journal, "The Health Factory: Mammoth Prepaid Plan Trims
Medical Costs, Wins Praise for Care." San Francisco edition, p. 1, col. 6,
April 26, 1971.
100
-------
67. Wendroff, Burton, Theoretical Numerical Analysis. New York: Academic
Press, 1966.
68. Winkelstein, Warren Jr., e£ al., "The Relationship of Air Pollution and
Economic Status to Total Mortality and Selected Respiratory System Mortality
in Men." Archives of Environmental Health 14:162-171, 1967.
69. Zeidberg, L. D., R. J. M. Horton, and E. Landau, "The Nashville Air Pollu-
tion Study: VI. Cardiovascular Disease Mortality in Relation to Air Pollu-
tion." Archives of Environmental Health 15:225-236, 1967.
101
-------
XI. APPENDICES
Page
A. Disease Data Retrieval System 102
B. Pretest Regression Results 123
-------
APPENDIX A
DISEASE DATA RETRIEVAL SYSTEM
Disease Classified I.C.D.A. T Codes
I. Respiratory System
A. Upper respiratory neoplasms
1. Malignant neoplasms, upper respiratory
a. Malignant neoplasm of nose (internal
and nasal cavities) 160.0
b. Malignant neoplasm of maxillary sinus. 160.2
c. Malignant neoplasm of other specified
accessory sinuses 160.8
d. Malignant neoplasm of site unspeci-
fied 160.9
e. Malignant neoplasms of larynx 161.0
2. Benign neoplasms, upper respiratory
a. Benign neoplasm, nose, nasal cavi-
ties, middle ear, and accessory
sinuses 212.0
b. Benign neoplasm of larynx 212.1
B. Upper respiratory infections
1. Influenza - acute
a. Influenza with pneumonia 480.0
b. Influenza, unqualified 481.0
c. Influenza, respiratory 481.1
d. Influenza, respiratory and digestive
system 481.2
2. Other and unspecified upper respiratory
infection - chronic
a. Hypertrophy of tonsils and adenoids... 510.0
b. Chronic pharyngitis 512.0
c. Chronic nasopharyngitis 512.1
d. Chronic maxillary sinusitis 513.0
e. Chronic frontal sinusitis 513.1
f. Chronic sinusitis, other specified.... 513.8
g. Chronic sinusitis, unspecified and
pansinusitis 513.9
102
-------
Disease Classified I.C.D.A. T Codes
h. Deflected nasal septum.... 514.0
i. Nasal polyp 515.0
j. Chronic laryngitis 516.0
k. Turbinate hypertrophy 517.6
3. Other and unspecified upper respiratory
infections - acute
a. Acute nasopharyngitis (common cold)... 470.0
b. Acute maxillary sinusitis 471.0
c. Acute frontal sinusitis 471.1
d. Acute sinusitis, not specified 471.4
e. Acute sinusitis, viral 471.5
f. Acute sinusitis, bacterial 471.6
g. Acute sinusitis, not specified,
antibiotic given 471.7
h. Acute sinusitis, other specified
sites 471.8
i. Acute sinusitis, unspecified and
pansinusitis 471.9
j. Acute pharyngitis, not specified 472.0 311
k. Acute pharyngitis, viral 472.1
1. Acute pharyngitis, bacterial 472.2
m. Acute pharyngitis, not specified,
antibiotic given 472.3
n. Acute pharyngitis, other 472.9
o. Acute tonsillitis, not specified 473.0
p. Acute tonsillitis, viral 473.1
q. Acute tonsillitis, bacterial 473.2
r. Acute tonsillitis, not specified,
antibiotic given 473.3
s. Acute laryngitis and tracheitis, not
specified, no antibiotic given 474.0
t. Acute laryngitis and tracheitis,
viral 474.1
u. Acute laryngitis and tracheitis,
bacterial 474.2
v. Acute laryngitis and tracheitis, not
specified, antibiotic given 474.3
103
-------
Disease Classified I.C.D.A. T Codes
w. Acute upper respiratory infection
of multiple or unspecified sites 475.0
x. Epistaxis 783.0
y. Change in voice 783.5 335,337
z. Stridor 783.6 336
4. Nonspecific respiratory complaints
a. Hemoptysis 783.1
b. Dyspnea 783.2
c. Cough NOS 783.3
d. Excess of sputum 783.4
e. Chest pain 783.7 401
f. Sighing respiration, wheezing 407
g. Choking 408
h. Sneezing 409
i. Purulent sputum 414
j. Other respiratory symptoms 419
k. Other diseases of the respiratory
system 517.9
11. Nasal discharge 284
22. Nasal obstruction 285
33. Nasal congestion 290
44. Pain of sinuses 291
55. Post-nasal discharge 294
66. Chronic post-nasal drip 298
77. Other symptoms of nose 299
88. Recurrent sore throat 310
C. Lower respiratory infection
1. Tuberculosis
a. Pulmonary tuberculosis, active,
minimal 002.0
b. Pulmonary tuberculosis, active,
moderately advanced 002.1
c. Pulmonary tuberculosis, active,
far advanced 002.2
d. Pulmonary tuberculosis, active,
stage unspecified 002.3
104
-------
Disease Classified I.C.D.A. T Codes
e. Pulmonary tuberculosis, miliary NOS... 002.4
f. Status following surgical collapse
of lung 002.5
g. Chromogenic acid-fast bacilli 002.6
h. Activity unspecified 002.9
i. Pleurisy specified as tuberculosis.... 003.0
j. Pleurisy with effusion, without
mention of cause 003.1
k. Primary tuberculosis complex with
symptoms 004.0
1. Other respiratory tuberculosis 007.0
m. Suspected tuberculosis 793.2
2. Acute bronchitis
a. Bronchitis, not specified, no anti-
biotic given. 500.0
b. Bronchitis, viral 500.1
c. Bronchitis, bacterial 500.2
d. Bronchitis, not specified, anti-
biotic given 500.3
e. Bronchitis, unqualified 501.0
3. Chronic bronchitis and emphysema
a. Bronchitis with emphysema 502.0
b. Emphysema without mention of
bronchitis 527.1
4. Chronic bronchitis without mention of
emphysema
a. Other (without emphysema) 502.9
b. Bronchiectasis (with or without
bronchitis) 526.0
5. Pneumonia
a. Friedlander's B. (lobular) 490.0
b. Pneumococcus (lobular) 490.1
c. Staphylococcus (lobular) 490.2
d. Streptococcus (lobular) 490.3
105
-------
Disease Classified I.C.D.A. T Codes
e. Other specified organism or cause
(lobular) 490.8
f. Unspecified organism or cause
(lobular) 490.9
g. Bronchopneumonia, Friedlander's 491.0
h. Bronchopneumonia, pneumococcus 491.1
i. Bronchopneumonia, staphylococcus 491.2
j. Bronchopneumonia, streptococcus 491.3
k. Bronchopneumonia, other specified
organism or cause..... 491.8
1. Bronchopneumonia, unspecified
organism or cause. 491.9
m. Primary atypical pneumonia 492.0
n. Friedlander's B 493.0
o. Pneumococcus 493.1
p. Other specified organism or cause.... 493.8
q. Unspecified organism or cause 493.9
6. Respiratory diseases of the newborn
a. Pneumonia of newborn... 763.0
b. Hyaline membrane (disease) (lung).... 773.0
c. Other respiratory distress 773.1
7. Other and unspecified diseases of the
respiratory system - chronic
a. Nonspecific respiratory disease,
secondary to smoking.., 503.0
b. Empyema 518.0
c. Abscess of lung 521.0
d. Pain in chest 783.7
8. Other and unspecified diseases of the
respiratory system - acute
a. Unspecified respiratory infection,
bacterial 476.2
b. Unspecified respiratory infection,
organism unspecified, antibiotic
given 476.3
c. Pleurisy, without mention of effu-
sion or tuberculosis 519.0
106
-------
Disease Classified I.C.D.A. T Codes
d. Pleurisy with effusion, with bac-
terial cause other than tuberculo-
sis 519.1
e. Pleurisy with effusion of unknown
etiology 519.2
f. Pleurisy, other specified forms of
effusion, except tuberculosis 519.9
g. Spontaneous pneumothorax 520.0
h. Pulmonary congestion and hypostasis.. 522.0
i. Pulmonary collapse (1 year and over). 527.0
j. Acute pulmonary edema without men-
tion of heart 527.2
9. Specific diseases to be analyzed
a. Spontaneous pneumo thorax 520.0
b. Acute pulmonary edema without men-
tion of heart 527.2
c. Hemoptysis 783.1
d. Dyspnea 783.2
e. Cough NOS 783.3
f. Excess sputum 783.4
D. Lower respiratory
1. Malignant neoplasm, lower respiratory
a. Malignant neoplasm of trachea
(primary or NOS) 162.0
b. Malignant neoplasm of bronchus and
lung 162.1
c. Malignant neoplasm, pleura, speci-
fied as primary 162.2
d. Malignant neoplasm of lung, unspeci-
fied as to whether primary or secon-
dary 163.0
e. Malignant neoplasm of mediastinum... 164.0
2. Benign neoplasm of the lower respira-
tory system
a. Benign neoplasm of trachea 212.2
b. Benign neoplasm of bronchus and
lung 212.3
107
-------
Disease Classified I.C.D.A. T Codes
c. Benign neoplasm of pleura 212.4
d. Benign neoplasm, mediastinum 212.5
e. Benign neoplasm, unspecified site.... 212.9
II. Allergies affecting respiratory system
A. Asthma
1. Asthma due to pollen 241.0
2. Asthma due to dander or dandruff 241.1
3. Asthma due to feathers 241.2
4. Asthma due to dust 241.3
5. Asthma due to food 241.4
6. Asthma due to cosmetics 241.5
7. Asthma due to drugs 241.6
8. Asthma with multiple allergens 241.8
9. Asthma due to other and unspecified
causes 241.9
B. Hay Fever
1. Hay fever due to pollen 240.0
2. Hay fever due to dander or dandruff 240.1
3. Hay fever due to feathers 240.2
4. Hay fever due to dust. 240.3
5. Hay fever due to cosmetics 240.5
6. Hay fever due to drugs 240.6
7. Hay fever with multiple allergens........ 240.8
8. Hay fever due to other and unspecified
causes 240.9
C. Specific diseases
1. Hay fever due to dust 240.3
2. Hay fever due to other and unspecified
causes 240.9
3. Asthma due to dust 241.3
4. Asthma due to other and unspecified
causes 241.9
108
-------
Disease Classified I.C.D.A. T Codes
III. Other Allergies and Skin Diseases
A. Eczema
1. Allergic eczema or dermatitis, pollen 244.0
2. Allergic eczema or dermatitis, dander
or dandruff 244.1
3. Allergic eczema or dermatitis due to
internal agent, feathers 244.2
4. Allergic eczema or dermatitis due to
internal agent, dust 244.3
5. Allergic eczema or dermatitis due to
internal agent, food 244.4
6. Allergic eczema or dermatitis due to
internal agent, cosmetics 244.5
7. Allergic eczema or dermatitis due to
internal agent, drugs 244.6
8. Allergic eczema or dermatitis due to
other and unspecified internal agents 244.9
B. Other allergies
1. Angioneurotic edema due to pollen 242.0
2. Angioneurotic edema due to dander or
dandruff 242.1
3. Angioneurotic edema due to dust 242.3
4. Angioneurotic edema due to food 242.4
5. Angioneurotic edema due to cosmetics 242.5
6. Angioneurotic edema due to drugs 242.6
7. Angioneurotic edema with multiple
allergens 242.8
8. Angioneurotic edema, other and unspeci-
fied 242.9
9. Urticaria due to pollen 243.0
10. Urticaria due to dander or dandruff 243.1
11. Urticaria due to feathers 243.2
12. Urticaria due to dust 243.3
13. Urticaria due to food 243.4
14. Urticaria due to cosmetics 243.5
15. Urticaria due to drugs 243.6
16. Urticaria with multiple allergens 243.8
109
-------
Disease Classified I.C.D.A. T Codes
17. Urticaria, other and unspecified 243.9
18. Other allergic disorders due to pollen.... 245.0
19. Other allergic disorders due to dander
or dandruff 245.1
20. Other allergic disorders due to feathers.. 245.2
21. Other allergic disorders due to dust 245.3
22. Other allergic disorders due to food 245.4
23. Other allergic disorders due to cosmetics. 245.5
24. Other allergic disorders due to drugs 245.6
25. Other allergic disorders with multiple
allergens , 245.8
26. Other allergic disorders due to other
and unspecified causes 245.9
C. Other skin diseases
1. Eczema 701.0
2. Other dermatitis due to plants 703.0
3. Other dermatitis due to oils and greases.. 703.1
4. Other dermatitis due to solvents „. 703.2
5. Other dermatitis due to drugs in contact
with skin 703.3
6. Other dermatitis due to other chemicals
in contact with skin 703.4
7. Other dermatitis due to radiation 703.5
8. Other dermatitis due to cosmetics 703.6
9. Other dermatitis due to dyes 703.7
10. Other dermatitis due to other specified
agents in contact with skin 703.8
11. Other dermatitis due to unspecified agent
in contact with skin 703.9
D. Specific diseases - allergies
1. Angioneurotic edema due to dust 242.3
2. Urticaria due to dust 243.3
3. Allergic eczema or dermatitis due to
internal agent - dust 244.3
4. Other allergic disorders due to dust 245.3
5. Other allergic disorders due to other
and unspecified causes 245.9
110
-------
Disease Classified I.C.D.A. T Codes
6. Other dermatitis due to other specified
agents in contact with skin 703.8
IV. Diseases of the Circulatory System
A. Heart
1. Hypertensive cardiovascular disease
a. Hypertensive heart disease with
arteriolar nephrosclerosis 442.0
b. Other hypertensive heart disease 443.0
2. Arteriosclerotic heart disease, including
coronary heart disease
a. Arteriosclerotic heart disease so de-
scribed, with or without angina
pectoris 420.0
b. Myocardial infarction 420.1
c. Healed coronary occlusion 420.2
d. Coronary insufficiency 420.3
e. Angina pectoris without mention of
coronary disease 420.4
f. Aneurysm of coronary artery and heart. 420.5
g. Arteriosclerotic heart disease with
atrial fibrillation 420.6
h. Arteriosclerotic heart disease with
angina pectoris 420.9
i. Arteriosclerotic heart disease with
myocardial infarction and angina
pectoris 421.5
3. Hypertensive and arteriesclerotic heart
disease
a. Hypertensive heart disease and arterio-
sclerotic heart disease 436.0
b. Hypertensive and arteriosclerotic
heart disease with atrial fibrilla-
tion 436.6
c. Hypertensive and arteriosclerotic
heart disease with angina pectoris.... 436.9
4. Chronic rheumatic heart disease
a. Other endocarditis, specified as
rheumatic (chronic), inactive 414.0
111
-------
Disease Classified I.C.D.A. I Codes
b. Other myocarditis, specified as
rheumatic (chronic) , inactive 415.0
c. Rheumatic heart disease with multi-
ple valve involvement 417.0
5. Other heart diseases
a. Congestive heart failure 434.1
b. Left ventricular failure 434.2
c. Other disease of heart 434.3
d. Cardiac enlargement or hypertrophy... 434.5
e. Cardiac or ventricular dilatation.... 434.6
f. Cor pulmonale (right ventricular
failure) 434.7
g. Past congestive heart failure (com-
pensated at time of service) 434.8
h. Other unspecified diseases of heart.. 434.9
i. Precardial pain 782.0
j. Palpitation 782.1
k. Tachycardia 782.2
6. Specific heart diseases to be examined
a. Myo car dial infarction 420.1
b. Coronary insufficiency 420.3
c. Angina pectoris without mention of
coronary disease 420.4
d. Arteriesclerotic heart disease with
angina pectoris 420.9
e. Arteriosclerotic heart disease with
myocardial infarction and angina
pectoris 421.5
f. Cor pulmonale 434.7
B. Circulatory system
1. Diseases of arteries, veins, and other
diseases of the circulatory system
a. Pulmonary embolism and infarction.... 465.0
b. Pallor and cyanosis (not of newborn). 782.3
c. Edema and dropsy (not of newborn).... 782.6
112
-------
Disease Classified I.C.D.A. T Codes
2. Hypertensive disease
a. Hypertension with arteriolar
nephrosclerosis 446.0
b. Other hypertensive diseases 447.0
c. Hypertensive heart disease with
atrial fibrillation 447.6
3. Acute brain syndromes associated with
circulatory disturbance
a. Acute brain syndrome associated with
circulatory disturbance 303.0
b. Acute brain syndrome associated with
metabolic disturbance 305.0
c. Cerebral arteriosclerosis. 334.0
d. Cerebral encephalopathy due to
arteriosclerosis or hypertension 334.1
4. Chronic brain syndromes associated with
a circulatory disturbance
a. Chronic brain syndrome cerebral
arteriosclerosis 313.0
b. Chronic brain syndrome, senile brain
disease.... 315.0
c. Chronic brain syndrome, presenile
brain disease 315.1
5. Arteriosclerosis
a. Arteriosclerosis not further
specified 450.0
b. Other 450.9
6. Non-specific circulatory complaints
a. Precardial pain 782.0 351
b. Dyspnea, orthopnea 783.2 368
V. Diseases of the Digestive System
A. Upper gastrointestinal tract - neoplasms
1. Benign neoplasms, upper gastrointestinal
tract
a. Benign neoplasms of the esophagus.... 211.0
b. Benign neoplasms of the stomach 211.1
113
-------
Disease Classified I.C.D.A. I Codes
c. Benign neoplasm of small intestine,
including duodenum 211.2
d. Benign neoplasms of liver and bilary
passages 211.5
e. Benign neoplasm of pancreas 211.6
f. Benign neoplasm of peritoneum 211.7
2. Malignant neoplasms
a. Malignant neoplasm of esophagus 150.0
b. Malignant neoplasm of stomach......... 151.0
c. Malignant neoplasm of duodenum 152.0
d. Malignant neoplasm of jejunum 152.1
e. Malignant neoplasm of ileum 152.2
f. Malignant neoplasm, part unspecified.. 152.9
g. Malignant neoplasm, liver stated to
be primary site.. 155.0
h. Malignant neoplasm gall bladder,
extrahepatic gall ducts including
ampulla of vater 155.1
i. Malignant neoplasm of liver, secon-
dary or unspecified 156.0
j. Malignant neoplasm of pancreas 157.0
k. Malignant neoplasm of peritoneum 158.0
B. Ulcer of the upper gastrointestinal tract
1. Ulcer of stomach, without perforation
and without hemorrhage 540.0
2. Ulcer of stomach, without perforation
but with hemorrhage 540.1
3. Ulcer of stomach, with perforation but
without hemorrhage 540.2
4. Ulcer of stomach, with perforation and
with hemorrhage 540.3
5. Peptic ulcer without mention of stomach
or duodenum 540.8
6. Ulcer of duodenum, without perforation
and without hemorrhage 541.0
7. Ulcer of duodenum, without perforation
but with hemorrhage 541.1
114
-------
Disease Classified I.C.D.A. T Codes
8. Ulcer of duodenum, with perforation
but without hemorrhage 541.2
9. Ulcer of duodenum, with perforation and
with hemorrhage 541.3
10. Gastrojejunal ulcer, without perforation
and without hemorrhage 542.0
11. Gastrojejunal ulcer, without perforation
but with hemorrhage 542.1
12. Gastrojejunal ulcer, with perforation
but without hemorrhage. 542.2
13. Gastrojejunal ulcer, with perforation and
with hemorrhage 542.3
C. Lower gastrointestinal tract - neoplasms
1. Benign neoplasms, lower gastrointestinal
tract
a. Benign neoplasm of large intestine,
excluding rectum 211.3
b. Benign neoplasms of rectum 211.4
2. Malignant neoplasms, lower gastrointes-
tinal tract
a. Malignant neoplasm of cecum, appendix,
and ascending colon 153.0
b. Malignant neoplasm of transverse
colon, including hepatic and splenic
flexures 153.1
c. Malignant neoplasm of descending
colon 153.2
d. Malignant neoplasm, Sigmoid 153.3
e. Malignant neoplasm, large intestine,
part unspecified 153.8
f. Malignant neoplasm, intestinal tract,
part unspecified 153.9
g. Malignant neoplasm of rectum 154.0
VI. Diseases of the Eye
A. Inflammatory diseases of the eye
1. Infective conjunctivitis 370.0
2. Conjunctivitis of unknown etiology, with
antibiotic given 370.1
115
-------
Disease Classified I.C.D.A. T Codes
3. Conjunctivitis of unknown etiology, with
antibiotic not given 370.2
4. Allergic conjunctivitis 370.4
5. Infectious blepharoconjunctivitis 370.5
6. Unspecified blepharoconjunctivitis, with
antibiotic given 370.6
7. Allergic blepharoconjunctivitis 370.8
B. Non-specific eye complaints
1. Diseases of the eyeball, ocular muscles,
and orbit 388.2
a. Pain around eye 221
b. Pain of eye.... 231
2. Discharge of eye «, 224
3. Excessive tearing, watery 234
4. Blepharitis 371.0
a. Inflammation of eyelid 255
5. Other pruritic conditions 708.9
a. Itching of eye 237
6. Feeling of foreign body in eye 258
VII. Diseases of the Genitourinary System
A. Benign neoplasms
1. Benign neoplasm of ovary, other and
unspecified 216.9
2. Benign neoplasm of kidney and ureter 219.0
3. Benign neoplasm of bladder 219.1
4. Benign neoplasm, other and unspecified
urinary organs 219.9
B. Malignant neoplasms
1. Malignant neoplasm of ovary 175.0
2. Malignant neoplasm of kidney and ureter... 180.0
3. Malignant neoplasm of bladder 181.0
4. Malignant neoplasm, other urinary organs.. 181.9
116
-------
Disease Classified I.C.D.A. T Codes
C. Nephritis and nephrosis
1. Acute nephritis 590.0
2. Nephritis with edema, including
nephrosis 591.0
3. Functional edema, idiopathic 591.1
4. Chronic nephritis 592.0
5. Nephritis not specified as acute or
chronic 593.0
6. Other renal sclerosis 594.0
VIII. Other Diseases
A. Symptoms and diagnoses referrable to nervous
system and special senses
1. Neurologic symptom - tic, tremor, other
abnormal voluntary movements, distur-
bances 780.4
a. Tic 165
b. Tremor 166
c. Other abnormal involuntary move-
ments 167
d. Twitching of muscles 709
e. Muscle spasm 710
f. Nocturnal leg cramp 719
g. Leg cramps 720
2. Neurologic symptom - other abnormal
voluntary movements, disturbances of
coordination 780.5
a. Other abnormal voluntary movements... 168
b. Disturbance of coordination 170
c. Other disturbances in writing 198
3. Tetany 788.5 169
4. Neurologic symptom - vertigo 780.6
5. Neurologic symptom - disturbance of
sleep.....«..«..••«•••«•••••••••••••••••• 780.7
a. Hypersomnia 025
b. Insomnia.. 026
c. Somnambulism 027
117
-------
Disease Classified I.C.D.A. T Codes
d. Nightmares
e. Sleepiness
6. Neurologic symptom - disturbance of
e. Sleepiness ............. . .............. °68
memory ............. .... ................... 780.0 175
a. Amnesia ............................... 176
7. Neurologic symptom - meningismus ....... ... 780.9
8. Disturbance of vision, except defective
sight ..................................... 781.0
a. Spots in field of vision .............. 248
9. Occulomotor disturbances .................. 781.1
a. Diplopia .......... . ............. . ..... 246
b. Drooping of eyelid .................... 247
10. Photophobia ....... ......... ............... 781.2 241
11. Disturbance of hearing, except deafness... 781.3
a. Tinnitus .............................. 268
b. Other extraneous noises in ear; and
sensitivity to noises ................. 269
12. Disturbance of cranial nerves, except
optic, occulomotor, and auditory ....... ... 781.4
13. Other disturbances of sensation ........... 781.7 190
a. Lightheadedness ....................... 020
b. Loss of sense of smell ...... . ......... 185
c. Loss of taste ..... ....... ............. 186
d. Anesthesia ............................ 187
e. Hyperesthesia ......................... 188
f. Paresthesia ........................... 189
g. Frigidity ............................. 626
h. Burning skin... ....................... 661
B. Emotional, mental, and psychotic disorders
1. Non-specific psychosomatic symptoms
a. Nervousness ........................... 790.0 100
b. Debility and undue fatigue ....... «.... 790.1
11. Cachexia ......................... 028
22. Fatigue .......................... 030
33 . Weakness ......................... 035
118
-------
Disease Classified I.C.D.A. T Codes
c. Lethargy 029
d. Depression of functional activity 790.2
e. Illness which prevents sleep,
except insomnia 790.4
f. Headache.... 791.0 151
g. Observation, mental 793.0
h. Malingering 795.3 115
i. Multiple chronic symptoms 059
j. Multiple complaints 060
k. Vague feeling of not feeling well 066
1. Euphoria 105
m. Loss of affect. 109
n. Slow learning in school 120
o. Other psychiatric symptoms. 149
2. Non-specific psychosomatic diagnoses
and/or symptoms of nervous system
a. Simple schizophrenic reaction,
negativism 320.0 117
b. Paranoid state, delusions 321.1 107
c. Psychoneurotic, anxiety reaction 324.0
d. Dissociative reaction 324.1
e. Conversion reaction 324.2
f. Phobic reaction 324.3 077
g. Obsessive compulsive reaction 324.4
h. Depressive reaction 324.5 076
i. Tranquilizers or sedatives pre-
scribed ; reason not stated 324.7
j. Hypochondriasis 324.8
k. Psychoneurotic disorders, other and
unspecified 324.9
1. Inadequate personality 325.0
m. Schizoid personality 325.1
n. Cyclothymic personality 325.2
o. Paranoid personality. 325.3
p. Other personality pattern disturbance,
abnormal sexual behavior 325.4 125
119
-------
Disease Classified I.C.D.A. T Codes
q. Emotionally unstable personality 325.5
r. Passive-aggressive personality 325.6 089
s. Compulsive personality 325.7 079
t. Other and unspecified personality
disturbance 325.9
11. Feelings of excess demands
on self 078
22. Mental Immaturity 098
33. Bizarre or disorganized
behavior 108
44. Failure to adjust to school 119
u. Antisocial reaction 326.0 087
v. Dissocial reaction 326.1 095
w. Alcohol addiction 326.3
x. Special symptom reactions 327.0
y. Hyperactive behavior 327.1
z. Transcient anxiety provoking situa-
tion; sedative or tranquilizer
prescribed 327.8
aa. Gross stress reaction.. 328.0
bb. Adult situation reaction 328.1
cc. Other and unspecified. 328.9
11. Hostile behavior 085
22. Asocial behavior 086
33. Failure to conform with be-
havior standards of family 088
44. Lack of mature behavior 118
3. Adjustment reactions
a. Adjustment to infancy 328.2
Hyperactive behavior 080
Tantrums 090
Cruelty 096
Destructiveness 097
b. Adjustment reaction of childhood 328.3
Hyperactive behavior 080
Tantrums 090
120
-------
Disease Classified I.C.D.A. T Codes
Cruelty 096
Destructiveness 09 7
Breathholding 116
c. Adjustment reaction of adolescence... 328.4
Hyperactive behavior 080
Tantrums 090
Cruelty 096
Destructiveness 097
d. Adjustment to late life 328.5
Hyperactive behavior 080
Marital problems 099
e. Hyperactive behavior 080
f. Tantrums 090
g. Cruelty 096
h. Destructiveness 097
i. Marital problems 099
j. Breathholding 116
4. Non-specific psychosomatic diagnoses
of other systems
a. Skin reaction 323.0
b. Musculoskeletal reaction 323.1
c. Respiratory reaction 323.2
d. Cardiovascular reaction 323.3
e. Hemic and lymphatic reaction 323.4
f. Gastrointestinal reaction 323.5
g. Endocrine reaction 323.7
h. Nervous system reaction 323.8
i. Reaction of organs of special sense.. 323.9
5. Psychotic disorders
a. Involutional psychotic reaction...... 318.0
b. Manic depressive, reaction manic
type 319.0
c. Manic depressive, depressive type.... 319.1
d. Manic depressive, other 319.2
e. Psychotic depressive reaction 319.3
121
-------
Disease Classified I.C.D.A. T Codes
f. Schizophrenic reaction, hebephrenic
type 320.1
g. Schizophrenic reaction, catatonic
type 320.2
h. Schizophrenic reaction, paranoid type. 320.3
i. Schizophrenic reaction, acute
undifferentiated type 320.4
j. Schizophrenic reaction, chronic
undifferentiated type 320.5
k. Schizophrenic reaction, schizo-
af fective 320.6
1. Schizophrenic reaction, childhood 320.7
m. Schizophrenic reaction, residual...... 320,8
n. Schizophrenic reaction, other and
unspecified 320.9
o. Paranoia 321.0
p. Other psychotic reaction... 322.0
C. Other specific diseases
1. Sarcoid of Boeck 138.0
2. Porphyria (except due to drugs) 289.4
3. Disaccharidase deficiency 289.7
4. Hyperventilation syndrome 324.6
5. Cirrhosis of liver without mention of
alcoholism 581.0
6. Cirrhosis of liver with alcoholism 581.1
122
-------
APPENDIX B
Table B-l. PRETEST REGRESSION RESULTS OF STATISTICAL MODEL 10:
RESPIRATORY DISEASES - UNLAGGED
Variables
Sample size
Constant
Age
Sex
Marital status
No. of people in household
Household income
Race - Negro
Race - other non-white
Physically fit
Somewhat physically fit
Drinking
Smoking index
Occupational exposure index
Air pollution
Temp, humidity index
Outpatient medical services
R2
a/
Coefficients—
153
-1.8736E 00
(1.9241E 00)
-6.7890E-05
(5.9527E-03)
1.1035E-01
(1.4408E-01)
2.8038E-01
(2.2013E-01)
2.4386E-02
(5.0846E-02)
-7.7191E-06
(1.0137E-05)
-2.4771E-01
(3.4051E-01)
4.6733E-01
(6.6923E-01)
-1.9884E-01
(2.4558E-01)
-4.8518E-01
(1.6731E-01)
4.8631E-04
(2.1145E-03)
4.5499E-04
(4.0694E-04)
7.7237E-03
(1.3866E-02)
2.7269E-01
(1.2159E-01)
7.0107E-01
(4.5198E-01)
0.1156
t Statistics
0.9738
0.0114
0.7659
1.2737
0.4796
0.7615
0.7275
0.6983
0.8097
0.2900
0.2300
1.1181
0.5570
2.2427
1.5511
Mean^-7
45.5294
(16.0538)
0.7386
(0.5823)
0.8039
(0.3983)
3.5033
(1.9639)
12178.1046
(7869.8121)
0.0523
(0.2233)
0.0131
(0. 1140)
0.1307
(0.3382)
0.5556
(0.4985)
13.5157
(38.2885)
118.9319
(198.2665)
4.3030
(5.9661)
60.2792
(32.0161)
52.7171
(8.5303)
13.8386
(11.2557)
—• The standard error of the regression coefficient is in parentheses. E-01,
E 01, E 00, etc., indicate that the decimal place of the number is to be
shifted to the left by one place; shifted to the right by one place; or
not shifted at all, respectively.
— The standard deviation about the mean is in parentheses; means are untrans-
formed.
123
-------
Table B-2. PRETEST REGRESSION RESULTS OF STATISTICAL MODEL 10:
CIRCULATORY-RESPIRATORY DISEASES - UNLAGGED
Variab les
Sample size
Constant
Age
Sex
Marital status
No. of people in household
Household income
Race - Negro
Race - other non-white
Physically fit
Somewhat physically fit
Drinking
Smoking index
Occupational exposure index
Air pollution
Temp, humidity index
Outpatient medical services
R2
a/
Coefficients—
333
-5.9326E-01
(1.2102E 00)
9.6186E-04
(3.9557E-03)
2.3956E-01
(8.3891E-02)
2.2654E-01
(1.2799E-01)
3.5947E-02
(3.8202E-02)
-9.1927E-06
(6.5999E-06)
-2.1188E-01
(2.6783E-01)
-5.1321E-03
(3.3190E-01)
-3.1567E-02
(1.3458E-01)
3.4824E-02
(1.0458E-01)
-6.9087E-04
(1.0267E-03)
3.5109E-04
(2.1380E-04)
1.4589E-02
(9.6653E-03)
2.1064E-01
(7.4181E-02)
3.9523E-01
(2.9039E-01)
0.0864
t Statistics
0.4902
0.2432
2.8557
1.7713
0.9410
1.3929
0.7911
0.0155
0.2346
0.3330
0.6729
1.6421
1.5095
2.8396
1.3610
Mean^7
54.4924
(16.0471)
0.6666
(0.5859)
0.8018
(0.3992)
2.9399
(1.7552)
10704.9549
(7900.4679)
0.0300
(0.1709)
0.0210
(0.1437)
0.1772
(0.3824)
0.4955
(0.5007)
12.3330
(46.1372)
136.7889
(224.4347)
3.1037
(5.2676)
59.8635
(31.8933)
54.1314
(8.2007)
13.6919
(11.4310)
b/
The standard error of the regression coefficient is in parentheses. E-01,
E 01, E 00, etc., indicate that the decimal place of the number is to be
shifted to the left by one place; shifted to the right by one place; or
not shifted at all, respectively.
The standard deviation about the mean is in parentheses; means are untrans-
formed.
124
-------
Table B-3. PRETEST REGRESSION RESULTS OF STATISTICAL MODEL 11:
RESPIRATORY DISEASES - LAGGED ONE DAY
Variables
Sample size
Constant
Age
Sex
Marital status
No. of people in household
Household income
Race - non-white
Drinking
Smoking index
Occupational exposure index
Air pollution
Temp, humidity index
Per capita costs
R2
Coefficients^'
166
-2.2162E 00
(2.6239E 00)
-1.1754E-02
(7.2070E-03)
-2.2007E-04
(1.9796E-03)
7.2529E-04
(2.8123E-03)
-1.0366E-01
(7.4044E-02)
-2.8430E-06
(1.4136E-05)
9.3073E-04
(4.0313E-03)
3.1881E-04
(2.9840E-03)
-2.1284E-04
(3.9639E-04)
2.0080E-02
(1.7792E-02)
2.0219E-01
(1.4319E-01)
5.1959E-01
(6.3601E-01)
0.0603
t Statistics
0.8466
1.6309
0.1112
0.2579
1.4000
0.2011
0.2309
0.1068
0.5370
1.1286
1.4121
0.8170
Meai£'
44.9036
(16.1028)
68.0723
(52.8453)
80.1205
(40.0301)
3.3373
(1.6933)
11974.3976
(7154.9744)
6.6265
(24.9497)
11.9235
(35.8459)
156.8643
(260.8312)
3.8560
(5.9076)
52.8506
(7.8279)
62.1369
(3415265)
1.7724
(3.3421)
—• The standard error of the regression coefficient is in parentheses. E-01,
E 01 E 00, etc., indicate that the decimal place of the number is to be
shifted to the left by one place; shifted to the right by one place; or
not shifted at all, respectively.
—^ The standard deviation about the mean is in parentheses; means are untrans-
formed.
125
-------
BIBLIOGRAPHIC DATA
SHEET
1. Report No.
EPA 600/5-74-017
S.N^ecipient's Accession No.
. Title and Subtitle
Outpatient Medical Costs Related To Air Pollution in the
Portland, Oregon Area
"3. Report Date
July 1974
6.
Author(s)
John A. Jaksch and Herbert H. Stoevener
8. Performing Organization Rept.
No.
Performing Organization Name and Address
Department of Agricultural Economics
Oregon State University
Con/all is, Oregon
10. Project/Task/Work Unit No.
1H1094-07AAC-02
11. Contract/Grant No.
Contract No. 68-01-0423
2. Sponsoring Organization Name and Address
Washington Environmental Research Center
Office of Research and Development
Environmental Protection Agency
Washington. D.C. 20460
13. Type of Report & Period
Covered
14.
15. Supplementary Notes
16. Abstracts
[his study has attempted to quantify in monetary terms the effects of air pollution on
the consumption of outpatient medical services. The hypotheses were that air pollution
can aggravate a state of health resulting in increased consumption of outpatient medical
services and in the number of contacts with the medical system for certain respiratory,
cardiovascular, and other diseases aggravated by air pollution. The study period was
1969-1970, and centered in the Portland, Oregon area. Statistical models were formulatec
explaining individual outpatient consumption of medical services. Measures of suspended
particulate air pollution and meteorological conditions, as well as socioeconomic-
demographic variables thought to influence the consumption of medical services, were
included in the models as explanatory variables. The statistical results indicated that
the procedures used in the study hold promise for quantifying the medical costs of air
pollution. The results did show air pollution to have an effect on the consumption of
outpatient medical services used to treat certain respiratory diseases.
17. Key Words and Document Analysis. 17a. Descriptors
Air Pollution
Economic Analysis
Benefit/Cost Analysis, Air Pollution
Economic Effects, Air Pollution
Economic Effects, Health
17b. Identifiers/Open-Ended Terms
Air Pollution Economics
Economic Impact
Air Pollution Effects (Health)
17c. COSATI Field/Group
18. Availability Statement
Unlimited
19. Security Class "(This
Report)
UNCLASSIFIED
20. Security Class (This
Page
UNCLASSIFIED
21- No. of Pages
132
22. Price
FORM NTIS-35 (KEY. 3-72)
THIS FORM MAY BE REPRODUCED
USCOMM-DC 14952-P72
------- |