EPA-600/5-78-002
January 1978
Socioeconomic Environmental Studies
IMPACT OF AIR POLLUTION ON THE
CONSUMPTION OF MEDICAL SERVICES - •
Cost of Hospitalization in the
Portland Metropolitan Area
Environmental Research Laboratory
Office of Research and Development
U.S. Environmental Protection Agency
Corvallis, Oregon 97330
-------
RESEARCH REPORTING SERIES
Research reports of the Office of Research and Development, U.S. Environmental
Protection Agency, have been grouped Into five series. These five broad
categories were established to facilitate further development and application of
environmental technology. Elimination of traditional grouping was consciously
planned to foster technology transfer and a maximum interface in related fields.
The five aeries 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 forecast-
ing, and analysis methodologies. Included are toots for determining varying
impacts of alternative policies; analyses of environmental planning techniques at
the regional, state, and local levels; and approaches to measuring environmental
quality perceptions, as well as analysts 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
management and policy-oriented reports.
This document Is available to the public through the National Technical Infofrna-
tiqn Service, Springfield, Virginia 22161.
-------
EPA-600/5-78-002
January 1978
IMPACT OF AIR POLLUTION ON THE CONSUMPTION OF MEDICAL SERVICES
Costs of Hospitalization in the Portland Metropolitan Area
by
Gobind S. Bhagia and Herbert H. Stoevener
Oregon State University
Corvallis, Oregon 97331
Contract No. 68-01-0482
Project Officer
John Jaksch
Criteria and Assessment Branch
Corvallis Environmental Research Laboratory
Corvallis, Oregon 97330
Corvallis Environmental Research Laboratory
Office of Research and Development
United States Environmental Protection Agency
Corvallis, Oregon 97330
-------
Disclaimer
This report has been reviewed by the Corvallis Environmental Research Labor-
atory, U.S. Environmental Protection Agency, 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.
ii
-------
FOREWORD
Effective regulatory and enforcement actions by the Environmental Protection
Agency would be virtually impossible without sound scientific data on pollu-
tants and their impact on environmental stabililty and human health. Respon-
sibility for building this data base has been assigned to EPA's Office of
Research and Development and its 15 major field installations, one of which
is the Corvallis Environmental Research Laboratory (CERL).
The primary mission of the Corvallis Laboratory is research on the effects
of environmental pollutants on terrestrial, freshwater, and marine ecosystems;
the behavior, effects and control of pollutants in lake systems; and the de-
velopment of predictive models on the movement of pollutants in the biosphere.
This report completes research initially funded through the Washington En-
vironmental Research Center, Washington, D.C.
A. F. Bartsch
Director, CERL
ill
-------
ABSTRACT
This study was undertaken to ascertain the possibility of measuring from
available data the impact of air pollution on the consumption of inpatient
services which patients consume per hospital stay.
The study area was the Portland, Oregon metropolitan area, and the period of
study was 1970-72. The diseases which were thought to be associated with air
pollution were first identified. A sample of the patients was taken from a
hospital population and the medical costs incurred by them were calculated.
Suspended particulates as a measure of air pollution, temperature-humidity
index as a measure of meteorological conditions of the area, and socio-
economic characteristics of patients were used in a regression model to
explain variations in medical costs.
The analysis did not indicate a quantitative relationship between air pollu-
tion in the Portland area and the consumption of inpatient medical services
per hospital stay. Age, family income, and drinking habits of a patient, and
the number of past hospital visits by the patient do significantly affect in-
patient medical costs, but the variables included in the analysis accounted
for less than 15 percent of the statistical variation in the dependent vari-
able.
This report was submitted in fulfillment of Contract No. 68010482 by Oregon
State University under the partial sponsorship of the U.S. Environmental
Protection Agency. This report covers a period from June 7, 1973 to February
7, 1975, and work was completed during December 1975.
iv
-------
CONTENTS
Foreword .......................
Abstract ....................... lv
List of Tables .................... vi
Acknowledgements .............. ..... vii
Sections
I Introduction. .... ............. 1
II Conclusions .................. 3
III Recommendations .. ...... . ....... 4
IV An Economic Model of the Consumption of
Medical Services .............. .. 5
V The Variables of the Empirical Model ...... 10
VI Results of the Analysis ............ 17
VII References. .................. 28
VIII Appendix .................... 30
-------
TABLES
No. Page
1 Estimated Regression Equations:
Respiratory Diseases 19
2 Estimated Regression Equations:
All Diseases 21
vi
-------
ACKNOWLEDGEMENTS
This study benefited from the assistance provided by several persons and
organizations. Without implicating them in any way for the remaining in-
adequacies of this study, the following individuals must be mentioned:
Dr. John A. Jaksch of the Environmental Protection Agency for his useful
comments at several stages of this research; Drs. Merwyn Greenlick and
Donald Freeborn at the Kaiser Health Services Research Center for providing
data and helpful comments at various meetings that the authors had with them;
Dr. Sheldon Wagner of the Environmental Health Research Center at Oregon State
University for his valuable assistance regarding clinical aspects of the
study; and Mr. James Sasser of the Computer Center who provided statistical
support; Drs. W. G. Brown, F. S. Conklin, and F. W. Obermiller of the Depart-
ment of Agricultural and Resource Economics and two anonymous reviewers of the
E.P.A. who provided many helpful suggestions on an earlier draft of this manu-
script. Bette Bamford typed patiently several revisions of this Report.
This work was also supported financially by the Oregon Agricultural Experi-
ment Station.
vii
-------
SECTION I
INTRODUCTION
THE PROBLEM
Public awareness of deterioration of air quality during the past two decades
has prompted a number of studies in the estimation of the damage done by air
pollution. The most common effects of air pollution are considered to be
visibility reduction, economic damage to property, physical and psychologi-
cal damage to human health, and substantive change in the ecology of the
natural environment. The Environmental Protection Agency estimated that air
pollution costs about $16 billion per year in terms of damage to vegetation,
materials, property and health [6, p. 12].
Damage to health is foremost in most peoples' minds. There is now enough
evidence in the environmental and health literature that the increasing
levels and high concentration of carbon monoxide, photochemical oxidents,
sulfur dioxide, nitrogen dioxide, and particulates in the ambient air cause
and aggravate physical and psychological impairment to such an extent that
the person exposed may have to seek medical attention. Pollution related
illnesses may cause reductions in job productivity, loss of wages due to
absenteeism, increases in health insurance, and a general decline in living
standards, and perhaps a shortened life span.
Concerns over these and other air pollution problems has been manifested in the
passage and increased enforcement of local, state, and federal legislation and
in extensive research by universities and other institutions to find ways of
abating or controlling the emission of pollutants* On the one hand, the con-
tamination of the environment, as a by-product of urbanization and Industrial-
ization, poses dangers to human health, on the other, control of air pollution
may cause shut-downs of firms, increase unemployment, and may jeopardize the
economic stability of a community. Thus, the costs and benefits from indus-
trialization must be weighed in order to make meaningful economic decisions.
In order to achieve this, it is most important to know about the severity
of damage caused to human health by the deterioration of air quality.
OBJECTIVE
It is the objective of this report to experiment with data available for an
urban area to ascertain the possibility of determining the effects of de-
teriorating air quality on the medical services consumed by patients per
hospital stay. It is also attempted to express this effect in monetary terms.
CHARACTERISTICS OF THE STUDY AREA
The study area is the Portland Standard Metropolitan Statistical area with a
population of over 900,000 [17, p. 1]. It is composed of four counties:
Multnomah, Washington, and Clackamas Counties in Oregon, and Clark County in
the state of Washington.
-------
Portland, with a population of 381,927, is the largest city in the study area
and is an important trading, transportation, and manufacturing center. It
has seven major industries: food processing and related agricultural pro-
duction, textiles, lumber and wood products Including pulp and paper, chem-
ical, aluminum, and other metal fabricating plants, shipping, and tourism
[11, P. 59],
Like most industrial cities and metropolitan areas, Portland has air pollution
problems, though not of a very severe nature. Suspended particulates are the
greatest pollution problem faced by the area. During the fall and winter
months the particulates are trapped for several days at a time by inversions
and cause visibility, health, and other problems. The suspended particulate
problem extends southward from Portland to the entire Willamette Valley.
Sulfur dioxide, photochemical smog, and nitrogen dioxide are not presently a
serious problem in the Portland area. However, downtown Portland does suffer
from carbon monoxide pollution.
Portland has a mild winter rainfall climate. Annual temperatures average
about 52.9°F and average annual rainfall is 37.18 inches, with 88 percent of
the annual total occuring between the months of October to May [20]. Marine-
tempered air moderates temperature extremes.
ORGANIZATION OF THIS REPORT
After a brief summary of the conclusions [Section II] and recommendations
[Section III] resulting from this study, the economic model employed in this
analysis is discussed in Section IV. The empirical model is presented in
Section V. The results fo the analysis are recorded in Section VI. Sections
VII and VIII contain references and an appendix, respectively.
-------
SECTION II
CONCLUSIONS
In this study it was not possible to demonstrate that deterioration of
air quality in the Portland metropolitan area independently aggravates a
person's condition in such a way as to affect significantly his or her
consumption of inpatient medical services, as measured by total medical
costs per disease episode in a hospital. Age of the patient, his family
income, drinking habits, and number of hospital visits during the past
three years do appear to be related to the variation in medical costs
incurred by the patient per hospital stay.
-------
SECTION III
RECOMMENDATIONS
The results of an earlier study conducted by Jaksch [11, 12] and of the
current one must be interpreted together. The former work focused on
outpatient medical services while the interest of the work reported here
was in the relationship between air pollution and the consumption of in-
patient medical services. Both of these efforts used data which had been
collected for other purposes. It is recommended that further refinements
in the study approach be made for the analysis of outpatient data, but not
for the specification of the relationship between variations in air quality
and the consumption of hospital services.
For reasons spelled out in this Report it was not possible to derive a
significant air pollution-medical cost relationship. This is not to say
that there may not be significant effects of air pollution on inpatient
medical costs. But a new research strategy is required to uncover them.
Specifically, a new approach is required which focuses on the medical,
psychological, economic, and administrative processes by which exposure
to air pollution may be reflected in hospital admissions and the delivery
of hospital services. Such a research program would be of relatively long
duration and expensive. Still, it appears to be a necessity for isolating
the effects of air pollution on this type of medical costs.
-------
SECTION IV
AN ECONOMIC MODEL OF THE CONSUMPTION OF MEDICAL SERVICES
REVIEW OF LITERATURE
An earlier study of the relationship between deterioration of air quality
and the consumption of medical services in the Portland-Vancouver area was
done by Jaksch and Stoevener [11, 12]. That study measured the impact of
suspended particulates on the consumption of outpatient medical services
per disease episode. To isolate this effect, air pollution and raeteor-
ogical conditions of the area in which patients resided and worked and
socioeconomic-demographic characteristics of the patients were considered*
Data were drawn from a five percent, ongoing random sample of the member-
ship of Kaiser Foundation Health Plan, a prepaid group medical plan. Each
medical procedure, treatment, and clinical visit was assigned a dollar
value according to the California Relative Value System, which was used by
Kaiser Foundation Health Services to quantify medical services performed.
Jaksch used single stage least-squares regression analysis to relate the
effects of air pollution on medical costs for respiratory diseases and
circulatory-respiratory diseases. The model used no lag and one-, two-,
and three day time lags to relate air pollution and meteorological condi-
tions to outpatient medical costs. The statistical results indicated a
significant effect of air pollution on respiratory diseases. The analysis
also brought out that the sex of the patient and his drinking habits were
important factors in explaining the variation in the medical costs for
respiratory diseases. Physical fitness, drinking habits, and the occupa-
tional exposure of patients and meteorological conditions were found to be
significant in influencing medical costs for circulatory-respiratory
diseases. Air pollution was not significant as a variable influencing
circulatory-respiratory diseases.
Several other studies concerning health effects of air pollution are re-
viewed. They help mainly in identifying diseases which may be aggravated
by the deterioration of air quality.
A medical study measuring the effect of air pollution on urban disease in-
cidence was done by Sterling, et. al. [18]. They found that the groups of
diseases most affected by air pollution are: Infectious diseases, malig-
nant neoplasms of the respiratory system, allergic disorders, diseases of
the hematopoistic system, personality disorders, diseases of the heart
(darditis, valvular, arteriosclerotic heart disease, and congestive failure),
and acute upper respiratory infections, pneumonia, bronchitis, diseases of
tonsils and adenoids, and other diseases of the respiratory system. In
all these categories there was a significant difference between days of
low pollution and high pollution. Specific pollutants were used to find
the relationship between air pollution and the occurance of the diseases
but it was concluded that the difference between the effects of specific
pollutants are less important since, at best, each one of the pollution
measurements is a general index of the level of pollution rather than a
specific indicator for a particular chemical. The only other variable
in this study was a temperature-humidity index. No socioeconomic vari-
ables were included in the analysis.
-------
A New York study by Cassell [5] concluded that it is not possible to isolate
the effects of individual pollutants. It was found that low temperature and
humidity and a wide range in barometric pressures are highly associated with
the prevalence of the common cold and cough symptoms* High temperature and
the occurence of eye irritation are positively associated with each other.
The study also found association of the common cold and cough with wind and
particulate matter. Carbon monoxide appeared to be associated with respira-
tory symptoms.
Lave and Seskin [1, p. 727] cite work by Hammond in which he studied over
50,000 men to find the relationship between emphysema, age, occupational
exposure to pollution, urban exposure, and smoking. His results indicate
that the effect of air pollution is significant. Heavy smokers have a much
higher morbidity rate in cities than in rural areas, and the effect becomes
more marked as age increases.
Epidemiological studies have shown a relationship between air pollution and
respiratory disease, both chronic (long-continued) and acute (short and rela-
tively severe). A study of 38,207 deaths in Nashville, Tennessee during 1949-
1960, adjusted for differences in income and social status, indicated that
more deaths from respiratory ailments occured in the sections of the city
where air pollution was heaviest [16, p. 64]. In another study a comparison
of selected high and low air pollution days indicated that air pollution may
well be responsible for a 16.7 percent increase in respiratory symptoms (in
particular, laryngitis, bronchitis, tonsillitis, colds, and sore throat) in
Los Angeles schools situated in the highest, as compared to the lowest, con-
taminant concentrations [7, p. 241].
A study by Zeidberg, et. al. [24] associates both morbidity and mortality
rates for heart diseases with air pollution levels in Nashville. The mor-
bidity rate was about twice as high in areas where air was polluted than
in areas of clean air. The mortality rate was found to be less closely
associated: it was 10 percent to 20 percent higher in areas of polluted
air than in areas of clean air.
Cardiovascular diseases are related to air pollution in that any pollutant
placing stress on pulmonary functions may affect the heart [8]. Air pollu-
tion (especially sulphur dioxide and photochemical smog) appears to cause
burning, tearing, and other irritations of the human eye, although the dam-
age in most cases is not permanent [8]. Some pollutants irritate sensitive
membranes, those of the eyes, nose, throat, and lungs.
A list of disease categories which are thought to be associated with air
pollution was prepared from the foregoing general discussion of several
epidem£ological and clinical studies in consultation with the medical author-
ities. For specific diseases under each category, the reader is referred to
the study done by Jaksch and Stoevener [12, pp. 102-122].
See Appendix Table 1.
2
Discussion with Dr. Sheldon Wagner, M.D., O.S.U. Environmental Health
Unit.
-------
EFFECTS OF METEOROLOGICAL CONDITIONS
Independently, or in combination with air pollution, various weather condi-
tions may place stress on the human body. Intense and continuous cold tem-
peratures can contribute to exhaustion and decrease mucous transport which is
one of the main methods of cleansing airborne matter from the lungs. Severity
of chronic bronchitis or asthma can result from exposure to severe colds. Low
temperature, accompanied by low humidity, has been associated with the common
cold and cough symptoms. High temperature and eye irritations are positively
correlated. Shortness of breath in those prone to heart attacks may increase
during spells of hot weather [12, p. 20].
Atmospheric conditions are a factor affecting the impact of odors, a source
of air pollution. Wind velocity, wind direction, temperature, and humidity
all are involved.
The structures and activities associated with urban areas tend to impair wind
velocity and favor the development of inversion layers that block the diffu-
sion that normally takes place when warm air rises. This, in turn, may re-
sult in the generalized "heat dome" effect of cities [1, p. 150]. All pollu-
tion emissions are retained below the inversion level and in many instances
a minor pollution problem becomes severe.
Meteorological conditions of an area play an important role in episodes of
air pollution. The city which is suffering from a severe smog blanket on one
day may enjoy beautiful smogfree weather after two or three days even though
the operation of factories or the number of vehicles driven may not change.
A strong wind may blow the smog away, or a heavy rain may wash it out of
the air.
FORMULATION OF THE ECONOMIC MODEL
"Medical Services" for this study is defined as the entire bundle of commod-
ities and services aimed directly at preventing, curing, and abating physical
and psychological illnesses. We follow the Jaksch formulation [11, 12] that
most medical needs are highly specific and alternative goods do not provide
the same or even approximate satisfaction. In most cases, the income and
wealth of the patient or his family and his health insurance coverage, do not
restrict him from consuming medical care. Thus the cross-elasticity of price,
income elasticity, and own price elasticity for the consumption of medical
care for a sick person is very low. Medical care is a necessity of life for
an ill person.
Existence of real physiological needs and perception of real and supposed
needs are the first ingredients of any consumption function. Before a
physiological need can be translated into a desire to consume a commodity, a
willingness to meet the need by undertaking consumption of the commodity must
exist. Finally, the consumer must be economically able to satisfy his wants.
The following discussion focuses on some of the variables which are thought to
be important in determining physiological needs, perception of needs, and will-
ingness to consume medical services to treat air pollution related diseases.
-------
Physiological Needs
Many physiological needs are age-specific, or sex-specific [23]. For example,
with an increase in age, beyond the youth period, efficiency of lung function
declines, due mainly to a decrease in the elasticity of lung tissue and an in-
creasing inflexibility of the chest wall. This lung condition would tend to
impair the clearance of pollutants from the respiratory tract, which could re-
sult in aggravated respiratory diseases. In general, young people do not
become ill easily. Thus, a positive relationship between age and consumption
of medical services tends to exist. Some diseases are common among one sex.
Comparison of Los Angeles schools analyzed for male and female sub-groups re-
vealed that respiratory diseases among men were more highly associated with
air pollutants than the identical illnesses in women.
Perception of Physiological Needs
One of the important elements in evaluating the perception of real or supposed
needs may be the level of education which a patient or his family has. Read-
ing of medical literature is also important, but it may be included in the
education variable* Education is also relevant in willingness to secure medi-
cal attention. An educated person will better appreciate the desirability of
seeking medical care. The willingness to seek medical attention also depends
upon the environment in which a person has been brought up. The number of
physical disabilities and hospitalizations during the past years are also
important factors which may be worth examining when considering willingness to
secure medical care.
Ability to Satisfy Physiological Needs
Ability to secure medical attention is expected to depend upon various eco-
nomic and other considerations. The most Important variables are family in-
come and wealth, health insurance coverage, eligibility for a welfare program,
availability of home care, and sick leave provisions. All these variables
can dictate the length of hospitalization. In a prepaid medical system income
is not expected to play as important a role as it would otherwise. However,
higher income individuals generally are able to afford preventive care and,
therefore, be healthier and may have to spend less on medical care [23]. Thus,
income acts as a proxy for other variables which affect medical cost.
Other Considerations
The influence of smoking on the prevalence of respiratory symptoms and bron-
chitis is very important [9, 10]. Cigarette smoking is an irritant, can
cause bronchospasms, can inhibit ciliary activity, and provoke secretion of
mucus. The latter interference with bronchial clearance may be critical [12,
35]. Lawther points out that there is no doubt that the overwhelmingly major
cause of lung cancer is cigarette smoking [14]. Cigarette smoking is so
important a factor in chronic respiratory diseases that it may cloud or even
conceal the effects of air pollution. Similarly, excessive drinking by an
individual is expected to play a role in causing and aggravating pulmonary
tuberculosis and other respiratory and cardio-vascular diseases, and mental
and emotional problems.
8
-------
Allergens or unusual concentrations of similar substances may be encountered
in a patient's working environment. Many asthmatic subjects develop wheezing
when exposed to any irritating dust or fumes. Some dust from cotton may con-
tain histamine releasing agents which are considered to cause asthmatic sym-
toms and metallic substances such as platinum, nickel, etc., in an aerosol
form, are considered to cause asthma-like syndromes [15, p. 61]. High con-
centrations of some organic dusts cause an allergic response in the aveolar
tissue of the lung. Thus, the occupational exposure to air pollution may be
an important factor in the demand for medical services.
Crowding at an individual's residence and at work can place physical and
emotional stress on the individual and can cause or aggravate certain res-
piratory diseases. The marital status of an individual may be considered an
explanatory variable in the model in that a married person is expected to
spend fewer days in the hospital than a single person because of availability
of care for the patient and perhaps better nutrition at home. The physical
disabilities and hospitalization during past years may be indicative of his
physical condition and explain, in part, consumption of medical services.
This study is concerned with those cases of air pollution related diseases
which result in hospital admission. Based upon the discussions in this
section, a basic severity model was developed. This model hypothesizes the
relationship between the inpatient consumption of medical services and air
pollution, and also takes into account the impact of meteorological and
socioeconomic factors.
Mijk " f(Ajk» Wjk» V
M. , « an index of inpatient medical services consumed for
J treatment of the i „ disease episode on the j day
where
1JK.
for the ku" person;
A.. - a measure of air quality on the j day for the k
person;
W . « a measure of meteorological conditions on day j for
the k person; and
S, * the socioeconomic characteristics of the k person.
3
Specification of this form of the variable was dictated largely by con-
siderations of data availability.
-------
SECTION V
THE VARIABLES OF THE EMPIRICAL MODEL
The Oregon Region Kaiser Foundation Health Plan provides prepaid comprehen-
sive medical care to approximately 165,000 members, 18 percent of Portland's
metropolitan population [17, p. 2], Medical care includes physician service
in the hospital and office* hospitalization, surgery, laboratory tests, radi-
ology, special nursing, emergency services, mental health, and routine physi-
cal examination. During 1971, the average hospital admissions per 1,000
members was 84 and the average length of stay was 5.3 days [17, p. 12], The
total number of hospital admissions for air pollution related diseases during
1970-72 was 4,478.
Since 1965, detailed medical and socioeconomic and demographic information has
been recorded on a personal history questionnaire for each patient who is ad-
mitted to Bess Kaiser Hospital. A magnetic tape containing this information
was obtained from Kaiser Research Foundation. Identification numbers of
patients (rather than their names) were used and these were scrambled so that
the privacy of the patient's record was safeguarded.
The Kaiser tapes indicated the inpatient consumption of medical services. If
a patient had a surgical operation, the surgery code of the International
Classification of Disease, Adopted (ICDA) was given. Surgery time, which is
an indication of the extent of use of the operating room and anesthesia was
not coded, nor were drugs and injections. The inpatient medical data contained
information on dates of hospitalization and discharge, admitting and discharge
diagnoses (ICDA code) surgery, laboratory tests, radiology, and consultations,
the number of physical disabilities of the patient, hospitalization during
the last three years, and the number of serious illnesses. The information
on X-rays was sketchy, only two categories were recorded: chest X-rays and
"other" X-rays. The socioeconomic and demographic data included information
on age, sex, race, place of birth, family income bracket, number of family
members, rooms in patient's residence, employment status, occupational class,
education, and smoking and drinking habits.
THE COST OF MEDICAL SERVICES
The dependent or response variable of the model is the consumption of medical
services by a patient per hospital stay. Since for every medical service a
specific dollar amount is charged, the total dollar value for all services
rendered a patient was thought to be an appropriate representation of the
consumption of medical services. In cases where the patient does not have to
pay medical expenses personally because of the prepaid Kaiser insurance sys-
tem, the dollar amount was calculated on the basis of what he would have paid
if he were not a member of the insurance plan.
10
-------
Inpatient medical costs are composed of charges for hospital room rent, sur-
gical procedures, laboratory and radiological examinations, operating and re-
covery j^room rent, anesthesia, physician's time, medication, and consulta-
tions. Since most of these charges are based upon the California Relative
Value Schedule (CRVS) units, a note on CRVS is in order. The first edition
of the CRVS was published in 1956 by the Committee on Fees of the California
Medical Association. The main objective of the Committee was the development
of a set of guidelines to govern fee schedules. It provides a basis in the
setting of fees for medical services [12, p. 462],
The doctor's component for the primary surgical procedure is $6.50 per CRVS
unit. Each secondary procedure is charged at $3.25 per unit. The Kaiser
tape, coded according to ICDA, was decoded by CRVS units. The number of
CRVS units multiplied by $6.50 for primary surgical procedures and $3.25
for secondary procedures yielded the value of the doctor's component of
surgery costs.
CRVS fees contained information on anesthesia. However, the Bess Kaiser
Hospital charged according to time a patient spent in the operating room
where anesthesia was applied rather than by CRVS units. Medical author-
ities were consulted to relate CRVS units with anethesia time. The
following rate structure resulted: the charges for anesthesia time were
$14.00 for each first and second 15 minute period and $7.00 for each sub-
sequent 15 minute period. These charges were for 1970-72. The charge for
use of the recovery room was made on a flat $24.50 basis for 1971 and 1972,
and $23.00 for 1970.
The operating room charges to a patient were $23.40 for the first 15 minute
period and $15.25 for each second, third, and fourth 15 minute period, and
thereafter $11.25 for each subsequent 15 minute period. These charges were
for 1971 and 1972. In 1970, the operating room charges were $21.00, $13.65,
and $10.50 for the respective periods.
Charges for wound dressing, orthopedic implants, and traction were considered
to be minor and were not included in the study.
The Kaiser tape contained ten separate categories of laboratory tests. These
included urine, hematology, blood chemistry, transfusions, bacteriology, ser-
ology, coagulation-bone marrows, pap smear-cervical/vaginal, pathology, and
miscellaneous tests. The 1969 CRVS units, each valued at 35 cents, were used
to derive total laboratory test charges.
Radiological procedures on the Kaiser tape were divided into chest and"other"
X-ray categories. The hospital charged $3.15 per 1969 CRVS unit. A routine
single view of the chest represented one chest X-ray. The "other" X-ray
category was broad, the average of CRVS units of routine X-rays in those
categories of diseases was used to represent the "other" X-ray category.
1 This section has benefited from conversations with Jack Thomas, Credit
Manager, Bess Kaiser Hospital, Portland, May 15, 1974.
2 A discussion with Dr. Sheldon Wagner, M.D., on November 12, 1974.
11
-------
The number of electrocardiograms with interpretation and report was on the
Kaiser tape. The CRVS units with $3.15 per unit yielded the charges for
electrocardiograms.
Consultations, which include the services rendered by a physician whose
opinion or advice is requested by another physician in the evaluation or
treatment of a patient's illness, are charged at $8.00 per CRVS unit. Thus,
a consultation requiring limited examination and evaluation of a given system
but not requiring a complete diagnostic history and examination, and which
has a value of 3 CRVS units, will cost a patient $24.00. Because the Kaiser
tape was not clear on the extent or intensiveness of consultation, it was
assumed that each consultation cost a patient $24.00.
Hospital room rent is charged on the number of patient days spent in the
hospital. Rent is paid on the day of admission irrespective of the time of
admittance, but the day of discharge is not counted. The hospital room rent
was $55.00 per day in 1970 and $61.50 per day in 1971 and 1972. The charge
includes meals, routine drugs, and services of a nurse.
Drugs consumed in the hospital are not coded on the Kaiser tape. A pharma-
cist at Bess Kaiser Hospital was consulted to estimate these costs. The
average drug cost per patient day was estimated at $3.50, representing an
average of 10 doses per patient per day at 35 cents per dose.
AIR POLLUTION
Suspended participates was chosen as the air pollution variable. This choice
was made for three reasons: (1) the major source of air pollution in the
Portland Metropolitan Area is particulate matter; (2) suspended particulates
consisting of particles smaller than 10 microns in size are respirable, and
as such initially affect the respiratory system; and (3) the Jaksch-Stoevener
study indicated an association between particulates and health effects [11,
12]. The total weight of suspended particulates is used as a single statis-
tical variable. Although particulates include benzene, calcium, copper, iron,
lead, manganese, nickel, chloride, fluoride, nitrate and sulfate particles,
no attempt was made to separate the effect os these components.
Data for suspended particulates from air pollution control stations were
obtained from the Department of Environmental Quality for Multnomah, Washing-
ton, and Clackamas Counties, and for Clark County in Washington. The air
pollution control stations reported particulate levels every four days. A
special procedure used by Jaksch, called the spline curve, was used in this
study to convert particulates level reporting from a four day to a single
day basis to be consistent with the daily meteorological and medical data
[12, p. 63].
Assignment of Air Pollution Values to the Patient's Residential Address
Each air pollution measurement station was assigned a geographic reference
using a two-dimensional coordinate plane. The coordinate plane was used to
assign a specific air pollution value to each patient on the basis of his
resident address. Resident addresses were obtained for all 4,478 inpatients
12
-------
admitted to Bess Kaiser Hospital during 1970, 1971, and 1972 for diseases
associated with air pollution. A computer program known as Admatch [12,
p. 64], was used to match addresses with census tracts of the area and
assign air pollution values to each address. The number of addresses used
in the analysis was 3,756. The air pollution value at a certain address
was calculated by weighting the observed air pollution values of the three
closest stations by their distances from that address , _ 1
^
[12, p. 64]. Distance squared'
The Meteorological Variable
Meteorological data for the study were obtained from the National Weather
Service of the U.S. Department of Conmerce. Data were available on dry bulb
temperature, relative humidity, wet bulb temperature, wind velocity and
direction, barometric pressure, and precipitation.
Hot and cold temperature extremes were expected to be related to the consump-
tion of inpatient medical services. Thus, degree days, the absolute difference
of the actual temperatures from 65 °F, was used as a meteorological variable in
an analysis. Preliminary investigation indicated that degree days was not a
statistically significant variable to explain variations in inpatient medical
costs. The temperature-humidity index (THI) , which takes into account both
temperature and relative humidity, was used instead.
The formula for calculating the THI is:
THI - o.4 (Td + Tw) + 15
where Td » dry-bulb air temperature in °F and Tw » wet-bulb temperature in 8F.
Empirical studies have shown that in the United States, life becomes unpleasant
for 10 percent of the population when the discomfort index (THI) reaches 70;
and 50 percent of the population suffer if the THI reaches 75 [19, p. 42].
This variable was expected to have a positive relationship with the consumption
of medical services.
Only the Portland International Airport had wet-bulb readings from which to
construct the THI. However, the Portland Airport THI was thought to be repre-
sentative of the area since the temperatures in the entire Portland Metropo-
litan Area generally fluctuate in the same direction.
SOCIOECONOMIC VARIABLES
Ten socioeconomic and demographic variables and two past medical history vari-
ables of the patient are discussed in this section. Information on these
variables comes from the questionnaire completed by the patients who were
admitted to Bess Kaiser Hospital.
The age of a patient is expressed in years on the date of admission. This
variable is expected to have a positive relationship with the consumption of
medical services.
13
-------
The sex of the patient is expressed as a one if the patient is a female and a
zero if the patient is a male. Little can be said about the relationship
between the sex of the patient and the consumption of medical services in
general because the relationship varies with the type of disease.
The education of the patient was divided into nine categories on the Kaiser
inpatient tape: (1) under high school unspecified, (2) 0-6 years, (3) 7-8
years, (4) 9-12 years, (5) high school graduate, (6) some college, (7) col-
lege graduate, (8) other post-high school degrees and (9) post-graduate col-
lege. The variable was expressed as a number from 1 to 8, leaving out the
first category which is not meaningful. This weighting system, though a
subjective one, is closely associated with the period of an individual's
formal education. An educated person may read medical literature, and take
preventive care. This may lead to more or less consumption of medical ser-
vices depending upon the individual's state of health.
The number of hospital visits in the last 3 years was expressed as the actual
number of serious illnesses. Nine or more of these were coded as 9. If the
same illness was listed by the patient more than once, each episode counted
as a separate illness. This variable was hypothesized to be positively re-
lated to consumption of inpatient medical services.
The number of disabilities a patient had were also listed as 0 to 9. This
variable, like the number of serious illnesses, is expected to have a posi-
tive relationship with the consumption of inpatient medical services.
A figure on the total number of persons residing at the patient's residence
was available. This included the patient, spouse, number of children,
brothers, sisters, and all others. The number of people at the patient's
residence, as a proxy for crowding, was expressed as this number and was
hypothesized to be positively related to medical costs incurred per hospital
stay, as crowding is thought to place an undue physical and emotional stress
on a person. Another variable reflecting the same phenomenon was also avail-
able. But this one, the number of persons per room, was discarded later as
a variable as analysis showed it to be non-significant in explaining medical
costs.
The Kaiser inpatient tape contained information only on whether the patient
smoked cigarettes, cigars, or pipe, or did not smoke at all. The number of
patients who smoked cigars or pipe was negligible. A zero was assigned to
the patients who did not smoke at all, and a one was assigned to those who
smoked. Smoking was expected to be positively related to the consumption of
medical services per hospital stay.
Some information of the patient's consumption of alcoholic beverages was
available. The patient was asked whether he consumed alcoholic beverages:
"often", "occasionally," "never (or seldom)." What (how many drinks or how
often consumed) constitutes "often" or "occasionally" was not clear from the
available data. Thus, "never (or seldom)" was entered in the analysis as a
zero, and "occasionally" and "often" were grouped together and expressed as
a one. The consumption of alcohol, like smoking, is expected to have a
positive relationship with the consumption of medical services.
14
-------
Marital status was coded on the Kaiser inpatient tape In 9 categories: adult
and never married, married, remarried, divorced, widowed, separated, pedia-
tric, new born, adopted. The last three categories were not used in the
analysis. Married or remarried persons were grouped together and were ex-
pressed as a one, and all remaining categories were assigned a zero. A
married person was expected to consume fewer medical services than an un-
married one.
Each patient specified one out of seven categories in which his family's gross
annual income fell. The categories were: under $2,500, between $2,500 to
$5,000; $5,000 to $7,500; $7,500 to $10,000; $10,000 to $12,500; $12,500
to $15,000; and over $15,000. The midpoints of the first six categories were
used in the study. For the open category (over $15,000), the figure of $25,000
was used. The latter was derived by calculating the weighted average of mid-
points of income classes over $15,000 obtained from the Federal income tax
returns for Oregon for 1970. Weights used were the number of returns filed
[21, p. 33]. These income figures were used in the analysis as the "family
income" variable. Jaksch [11] discussed the function of this variable in
explaining the consumption of medical services. Given the likely preferences
for medical services and the nature of the prepaid medical insurance plan by
which patients in this study were covered, income does not play the same role
as it does in the consumption of a normal good. Instead the income variable
was included in this study as a proxy for other socioeconomic factors which
may be associated with it. The sign on this variable is difficult to predict.
A race variable was expressed as a zero for a white and as a one for a
non-white. It was included for similar reasons as the income variable.
There were seven broad categories of occupational class on the Kaiser
questionnaire: (1) executives of large concerns, proprietors, and major pro-
fessionals; (2) business managers, proprietors of medium-sized businesses
(value $35,000 to $100,000) and lesser professionals; (3) administrative per-
sonnel; (4) clerical and sales workers, technicians, and owners of small
businesses; (5) skilled manual employees (auto body repairmen, bakers, black-
smiths, bulldozer operators, firemen, mechanics, policemen; (6) semi-skilled
employees (delivery men, taxi drivers, welders); and (7) unskilled employees
(ash removers, parking lot attendants, farm helpers, janitors, woodchoppers).
The first three categories were thought to be of the lowest air pollution
exposure and were expressed as a one in the analysis. Fourth, fifth, sixth,
and seventh were expressed as two, three, four, and five, respectively. A
five equaled a high hazard to air pollution exposure at one's occupations.
This weighting scheme is based upon inadequate data and is highly subjective.
To the extent that it does indicate the approximate degree of exposure to
pollutants, it is expected to have a positive relationship with the consump-
tion of medical services.
The statistical model with all variables as discussed above may be written as:
Yijk " f (Xlijk; X2ijk; X3ijk; '••'• X14ijk; "ijk*
where
Y.., * index of medical costs incurred by the k patient during his
* hospital stay, who was hospitalized on the j day for the
i disease,
15
-------
X,.., - a measure of suspended particulate (raicrograms per cubic
J meter per day) in the ambient air for the^k patient's
residence, who was hospitalized on the J day for the
i disease.
X2i*k * a neasure of meteorological conditions^ (expressed as
J temperature-humidity index) for the k patient's
residence, who was hospitalized on the j day for
the i disease.
X3i1k •** X14Hk " age' sex» race» "^rital status, education, tht
J J-HIJR number of disabilities, the number of serious
illnesses, crowding at the patient's residence,
smoking habits, drinking habits, family income,
and occupational exposure to air pollution, re-
spectively, of .the k person. .who was hospital-
ized on the j day for the i disease.
e. .. » a random error term to account for the variables which may
•^ have been omitted.
16
-------
SECTION VI
RESULTS OF THE ANALYSIS
EVALUATION OF THE STATISTICAL MODELS
Due to the large number of observations a 5 percent sample of observations for
all diseases that were thought to be related to air pollution was randomly
selected for a preliminary analysis. The purpose of preliminary investigation
was to pretest the statistical model in an inexpensive way for determining
the form of the model with the best fit. The multiple regression analysis
was performed on the sample using a linear combination of variables as a first
step. Scatter diagrams between the dependent variable and each of the inde-
pendent variables indicated the transformations required in the least-squares
analysis. Several transformations were tried on the variables to obtain the
best fits. The selection of the best model-was made on the basis of calculated
t -values for the individual variables and R for the whole model.
The results indicated that the coefficient of the air pollution variable was
statistically significant at the 5 percent level and the coefficient of multi-
ple determination (R ) was .18, .17, .19, and .20 for air pollution and
temperature-humidity index unlagged, lagged one day, lagged two days and
lagged three days, respectively. Temperature-humidity index was not statis-
tically significant. Degree days (the absolute difference of the temperature
from 65 °F) was also tried but it did not improve the results. The coeffi-
cients of age, race, family income, occupational exposure, the number of
hospital visits, and education were statistically significant at least at
the 20 percent level. The coefficients of sex, marital status, the number
of people per household, drinking and smoking habits were not significantly
different from zero.
The above results from the 5 percent sample showed promise and a program was
written for performing multiple regression analysis on all available observa-
tions. The transformation of the observations on medical costs, air pollution
and meteorological variables to natural logarithms gave the best fit. This
was true in all models for unlagged, and one-day lagged, two-lagged, and
three-day lagged values of the variables. Thus, the final statistical
model became
myijk
35X5ijk + 36X6ijk + 37X7ijk + 38X8ijk + 39X9ijk
B10X10ijlc 3llXllijk + 612X12ijk + 313lnX13ijk
Eijk
17
-------
where Y. , = medical costs incurred by the kfc patient who was hospitalized
J Anl* ^V*
on the j day for the i disease; X^^ to X12 are characteristics
related to the patient and, respectively, represent age, sex,
race, marital status, family income, number of people in house-
hold, alcohol consumption, smoking habit, occupational exposure
to air pollution, the number of disabilities of the patient, the
number of hospital visits by the patient in the last 3 years, and
education of the patient. X.- and X-j, are temperature-humidity
index and air pollution values, respectively, associated with
the patient's residence.
RESULTS OF THE REGRESSION ANALYSIS
There was a total of 3650 observations in all relevant disease categories com-
bined and 2808 observations for respiratory diseases only. Respiratory ill-
nesses were analyzed separately because of their manifestation in several
studies as important in considering effects of air pollution.
In both models the coefficeint of air pollution was not statistically signifi-
cant. The coefficient of multiple determination was about .08 for all disease
categories and less than .10 for respiratory diseases only. These magnitudes
of R held in all four submodels, that is, when air pollution and meteorologi-
cal values were unlagged, lagged one day, lagged two days and lagged three days.
The difference of this magnitude between a random sample and population results
is unusual. The medical costs and air pollution values were plotted for all
diseases and respiratory diseases only to see whether or not there is any asso-
ciation between the two variables, and if there is, what sort of association it
is. The maps of plottings indicated that the sample relationships may be super-
ficial and, hence, the sample may not have been a representative one. However,
the plottings also showed that some relationship other than a linear one should
be tried for obtaining the best fit.
Results of the regression analysis are presented in Table 1 for respiratory
diseases only and Table 2 for all diseases analyzed in this study. Correla-
tion matrices of variables are shown in Appendix Tables 1 and 2.
Each coefficient is accompanied by the standard error of the estimate and the
t-statistic to indicate the statistical significance of the coefficient. The
mean value of each variable is also presented along with its standard deviation.
Both tables are divided into four parts: the first part refers to the results
of regression analysis which uses unlagged values of air pollution and meteor-
ological variables, and the second, third and fourth parts use one-day, two-day
and three-day lags, respectively, on these two variables. The results for the
individual variables of the regression analysis are discussed below.
18
-------
TABLE 1. ESTIMATED REGRESSION EQUATIONS: RESPIRATORY DISEASES
Unlagged—
Variable
Sample
size
Constant
Age
Sex
Race
Marital
Status
Family
Income
Number of
people in
household
Drinking
Smoking
Occupational
exposure
Coeffi-
cient^7
2,808
5.256
.0080
(.00064)
.0499
(.0268)
.0431
(.0530)
.0584
(.0038)
.0000072
(.0000025)
.0091
(.0075)
.110
(.034)
.045
(.037)
.012
t-sta-
tistic
12.373*
1.869**
0.813
1.728**
2.852*
1.225
3.155*
1.225
1.272
Mean—
32.87
(28.17)
0.42
(0.49)
0.05
(0.233)
0.31
(0.46)
9.149
(5.190)
3.19
(2.10)
0.25
(0.43)
0.16
(.037)
1.51
Lagged 1 day—
Coeffi-
cient^
5.205
.0080
(.00064)
.0500
(.0268)
.0435
(.0530)
.0501
(.0338)
.0000072
(.0000025)
.0091
(.0075)
.110
(.034)
.045
(.037)
.012
t-sta-
tistic
12.356*
**
1.865
0.819
1.747**
2.845*
1.222
3.151*
1.223
1.274
Lagged
Coeffi-
cient^'
5.136
.0079
(.00064)
.0502
(.0268)
.0435
(.0531)
.0598
(.0338)
.0000072
(.0000025)
.0093
(.0075)
.110
(.034)
.046
(.037)
.012
2 days^
t-sta-
tistic
12.165*
**
1.872
0.819
1.767**
2.827*
1.243
3.165*
1.231
1.288
Lagged 3
Coeffi-
cient^7
5.218
.0080
(.00064)
.0502
(.0268)
.0438
(.0530)
.0602
(.0338)
.0000072
(.0000025)
.0092
(.0075)
.109
(.034)
.045
(.037)
.012
days—
t-sta-
tistic
12.346*
1.875**
0.826
1.779**
2.851*
1.234
3.147*
1.218
1.287
Continued
-------
TABLE 1. (Continued)
to
O
Unlagged—
Variable
Number of
disabilities
Number of
hospital
visits
Education
Temperature-
Humidity
Index
Air
Pollution
Medical
Costs
R2
Coeffi-
cient^
—
-.0265
(.0097)
-.0015
(.0065)
.0804
(.0844)
.2006
(.0237)
.1292
Signifies that the
Signifies that the
t-sta-
tistic
—
-2.738*
-0.231
0.952
0.869
Mear£7
0
(0
0
(1
3
(2
54
(7
67
(34
577
(586
.16
.49)
.66
.30)
.51
.18)
.97
.91)
.08
.07)
.14
.14)
Lagged
Coeffi-
cient^
•—
-.0265
(.0097)
-.0014
(.0065)
.0720
(.0834)
.0169
(.0240)
.1291
1 day^
t-sta-
tistic
—
-2.736*
-0.225
0.863
0.707
coefficient is statistically significant at 5
coefficient is statistically significant at
Lagged
Coeffi-
cient^7
.0004
(.0258)
-.0267
(.0097)
-.0015
(.0065)
.1302-
(.0819)
.0011
(.0242)
.1294
a / ,» t
2 days— Lagged 3 days—
t-sta- Coeffi- t-sta-
tistic cienfr- tistic
0.017 — —
-2.735* -.0266 -2.748*
(.0097)
-0.231 -.0015 -0.240
(.0065)
1.589 .1101 1.236
(.0817)
0.047 .00080 0.033
(.0243)
.1293
percent level (one-tailed test).
10 percent level (one-tailed test) .
— The standard error of the coefficient is in parenthesis
— Means of untransformed variables. The standard deviation Is in parentheses.
-------
TABLE 2. ESTIMATED REGRESSION EQUATIONS: ALL DISEASES
NJ
Unlagged—
Variable
Sample
size
Constant
Age
Sex
Race
Marital
status
Family
income
Number of
people in
household
Drinking
Smoking
Occupational
exposure
Coeffi-
cient^
3,650
5.440
.0074
(.00058)
.0297
(.0247)
.0678
(.0501)
.0325
(.0289)
.0000062
(.0000023)
.00249
(.00730)
.120
(.030)
.039
(.033)
.0024
(.0087)
t-sta-
tistic
12.734*
1.201
1.352
1.124
2.605*
.341
3.924*
1.184
.280
Mean2-'
40.43
(29.51)
0.432
(0.495)
0.055
(0.229)
0.349
(0.476)
8,648
(5,171)
2.873
(2.068)
0.257
(0.437)
0.180
(0.384)
1.483
(1.410)
Lagged 1
Coeffi-
cient
5.419
.0074
(.00058)
.0297
(.0247)
.0682
(.0501)
.0327
(.0289)
.0000062
(.0000023)
.00251
(.00730)
.120
(.030)
.039
(.033)
.0024
(.0087)
day^
t-sta-
tistic
12.737*
1.201
1.359
1.131
2.596*
.344
3.927*
1.179
.283
Lagged
Coeffi-
cient^7
5.277
.0074
(.00058)
.0300
(.0247)
.0684
(.0501)
.0331
(.0289)
.0000062
(.0000023)
.00263
(.00730)
.120
(.030)
.039
(.033)
.0026
(.0087)
2 days3-7
t-sta-
tistic
12.742*
1.212
1.364
1.147
2.592*
.361
3.941*
1.178
.300
Lagged 3
Coeffi-
cient^7
.0074
(.00058)
.0298
(.2047)
.0682
.0336
(.0289)
.00000619
(.00000239)
.00258
(.00730)
.120
(.030)
.039
(.033)
.0025
(.0087)
days^7
t-sta-
tistic
12.738
1.207
1.362
1.163
2.588*
.354
3.925*
1.176
.293
Continued
-------
TABLE 2. (Continued)
to
NS
Variable
Number of
disabilities
Number of
hospital
visits
Education
Temperature-
humidity
index
Air
pollution
Medical
costs
R2
Unlagged—
Coeffi- t-sta-
cient— tistic
.020 .927
(.021)
-.019 -2.154
(.0088)
.0047 .778
(.0060)
.060 .778
(.077)
.0049 .224
(.0220)
.1061
c/
0.199
(0.538)
0.749
(1.341)
3.240
(2.211)
54.97
(7.91)
67.08
(34.07)
605.84
(587.88)
Lagged 1 day—
Coeffi- t-sta-
cient— tistic
.020 .924
(.021)
-.019* -2.156
(.0088)
.0047 .783
(.0060)
.064 .848
(.075)
.00589 .266
(.0221)
.1062
Lagged
Coeffi-
cient^
.020
(.021)
*
-.019
(.0088)
.0047
(.0060)
.111
(.074)
-.0054
(.0221)
.1065
2 days^
t-sta-
tistic
.932
-2.170
.779
1.489
-.245
Lagged
Coeffi-
cient^
.020
(.021)
*
-.019
(.0088)
.0046
(.0060)
.106
(.741)
-.0063
(.0221)
.1065
3 days^
t-sta-
tistic
.930
-2.172*
.766
1.438
-.286
b/
Signifies that the coefficient is statistically significant at 5 percent level (one-tailed test).
Refers to the respective lags in meteorological and air pollution variables.
— The standard error of the coefficient is in parentheses.
c/
— Mean of untransformed variables. The standard deviation is in parentheses.
-------
Age of the Patient
Age of the patient was the most important factor in explaining the variation
in the consumption of inpatient medical services per hospital stay. The
coefficient was positive as expected and it was statistically significant at
1 percent level for respiratory diseases only (model 1) and also for all
diseases (model 2) analyzed in this study. The positive coefficient implies
that as the patient grows older he consumes more medical services per hospi-
tal stay. The average age of the patient was about 33 years for the patients
hospitalized for respiratory diseases and about 40 for the patients hospital-
ized for all diseases analyzed.
Sex of the Patient
The coefficient of sex of the patients was positive and was statistically
significant at the 5 and 10 percent levels, respectively, for model 1 and 2,
implying that women consume more medical services when hospitalized than men.
About 43 percent of the patients who were treated for all diseases analyzed
here were women and the percentage of women in the respiratory disease category
was 42.
Jaksch and Stoevener [12] found that 68 percent of women consume out-patient
services for respiratory illnesses, but women consume fewer out-patient medi-
cal services than men per episode of illness. The results of this study
indicate that the percentage of women admitted to the hospital is lower than
that of men. Yet when they are admitted to the hospital they consume more
services there than men. This may have some implications about differences
in the rate of affliction with respiratory diseases between men and women,
as well as about sex differences in the costs of work loss on the job and
Che availability of medical care in the home.
Race of the Patient
Race of the patient was statistically significant at the 10 percent level
for all diseases combined but was not significantly different from zero for
respiratory diseases only. The positive coefficient on the race variable
implies that blacks and other non-whites consume more in-patient medical
services than whites. Of all patients who consumed inpatient medical services
for the diseases analyzed in this study, only about 5 percent were nonwhite.
Marital Status of the Patient
Marital status of the patient was statistically significant at the 10 per-
cent level for respiratory illnesses but not significant for all diseases
taken together. The positive coefficient indicates that married persons
consume more in-patient medical services per hospital visit. About 35
percent and 31 percent of the patients admitted for all diseases combined
and respiratory diseases only, respectively, were married persons. Jaksch
and Stoevener [12] found that over 84 and 81 percent of the patients con-
suming out-patient medical services for respiratory and circulatory respira-
tory diseases, respectively, were married. Married persons may consume fewer
inpatient medical services because of the availability of home care for the
patient. Unless a married person is very sick, he may not go to the hospital
but he may consume more medical services in the hospital once he is admitted.
23
-------
Family Income
The coefficient of family income has a positive sign and is statistically
significant at the 1 percent level in both disease categories. As the
household income of a patient rises he or she consumes more medical ser-
vices per hospital stay. The average family income was over $9,000 for
patients with respiratory illnesses and about $8,600 for patients in the
all diseases category.
Number of Persona in the Patient's Household
The coefficient was positive but not significantly different from zero in
both models indicating that there is no relationship between the number of
people in the patient's household and the consumption of inpatient medical
services. This variable appears to have been a poor proxy for the effect
of crowding in living conditions which it was supposed to measure.
Drinking Habits of the Patient
The coefficient of the drinking variable is positive as expected and statis-
tically significant at the one percent level. Though data on the amount of
drinking the patients do were not available for this study, this result is
consistent with the popular notion of association between drinking and bad
health. About 25 percent of the patients analyzed in this study consumed
alcoholic beverages occasionally or often.
Smoking Habits of the Patient
The coefficient of this variable was positive but not statistically signi-
ficant for any of the models. Similar results were obtained by Jaksch and
Stoevener [12, p. 87] for out-patients consuming medical services for res-
piratory and circulatory-respiratory diseases.
This is a surprising result because it is known that in general smoking de-
teriorates the condition of health, especially for patients who have respir-
atory illnesses.
The results for smoking in this analysis may be attributed to poor data
available for this variable. The inpatient questionnaire only contained
data on whether a person smoked or not. As in the case of drinking, the
amount of smoking and the number of years smoked by the patient were not
known. Only 16 to 18 percent of the patients said that they smoked. As
the percentage of smokers in the general population is much higher [22]
the possibility of a measurement error suggests itself.
Occupational Exposure to Pollution
The coefficient of occupational exposure to job related pollution was posi-
tive for both models. It was significantly different from zero at the 20
percent level in only the case of respiratory diseases with two- and three-
day lags of the air pollution and meteorological variables. It was not
significantly different from zero in the remaining cases. The positive
coefficient implies that workers exposed to dust, dirt, etc., spend more on
inpatient medical services than white collar workers in offices.
24
-------
The exact occupation of the patient was not known and the categories of
occupations were very broad. This may also have contributed to the non-
significance of the coefficients.
Number of Disabilities of the Patient
The coefficient of the number of disabilities a patient had was not statis-
tically significant from zero in any of the models. The nature of disabilities
may be more important than the number of disabilities in explaining the ex-
penditure on medical services. The nature of the disabilities was not known.
Number of Hospital Visits
The number of times in the preceding three years a patient was admitted to a
hospital is negatively related to the medical expenses under consideration
here. This variable had a coefficient which was statistically significant
at the 5 percent level in all models. The negative coefficient of this
variable indicates that the more frequently the patient had been to the
hospital, the less will be his expenditure on medical services during the
stay under consideration. The expected sign of the coefficient was positive.
Perhaps the total length fo stay in the hospital during previous visits would
be a more appropriate variable.
Education of the Patient
Education was not a significant variable in any of the models. This is not
surprising given the difficulty of predicting the effect of this variable and
its correlation with some other variables included in the analysis [Appendix
Tables 2 and 3].
Temperature-Humidity Index
The coefficient of this variable was statistically significant at the 20
percent level for two- and three-day lags in both models. The sign of the
coefficient was positive as expected in all cases.
Air Pollution
None of the coefficients in the eight submodels evaluated was statistically
significant. This implies that there is no relationship between suspended
particulates and the consumption of inpatient medical services in the Port-
land Metropolitan Area, given the specifications of these models.
One of the primary shortcomings explaining the nonsignificance of the re-
gression coefficient of air pollution may be a time lag of greater than
three days between the onset of illness and the patient's admission to a
hospital. The aggravation of disease as an effect of deterioration in air
quality may be slow to be noticed and the patient may not feel the necessity
of seeking medical help immediately. A hospital bed or surgeon's time (in
case of surgery) may also not be available on short notice. For all of these
reasons the association of the decline in air quality which may have caused
The hospitalization with the date of admission to the hospital, is very
difficult to make.
25
-------
GENERAL COMMENTS
Only about 10 to 13 percent of the variation in total medical costs for in-
patient services was explained by the deterioration in air quality, meteor-
ological conditions, and socioeconomic and general characteristics of the
patient.
As discussed by Jaksch and Stoevener [12] the low coefficient of multiple
determination (R ) may be attributed to the fact that the models used in
this study are a simple representation of a much more complex medical
phenomenon. In both models (and all submodels) age of the patient, family
income, drinking habits of the patient, and the number of hospital visits by
the patient in the last 3 years were the significant variables in explaining
the variation in the consumption of inpatient medical services. The in-
fluence of unspecified factors on inpatient medical costs is greater than
the explanation offered by these few variables.
There are several limitations to this study which indicate that its results
should be used with caution. First, there were a number of inadequacies in the
data used. The specific laboratory tests and radiological procedures were not
known for a patient, instead only broad categories of these tests and proce-
dures were known. Thus, the estimates were made on the basis of categories
of these tests and procedures. The number of drugs consumed per day were not
known for each patient separately. The pharmacy staff at the hospital reported
that, in general, the expenditure on drugs is the same per day for each patient
and if this amount varies it does not vary a great deal. Only approximation
could be used in calculating the time involved in surgery. There was also no
indication whether the consultations a patient had with a physician were of
limited or extensive nature*
Second, occupational addresses for the hospital admitted patients were not
recorded on the questionnaire forms, hence, the air pollution and meteor-
ological values at the job were not taken into account. Thus, air pollution
and meteorological values at only residential addresses were considered in
this study.
Third, data on the socioeconomic characteristics of patients came from the
questionnaire administered to patients at the hospital. The questionnaire
was not designed for the purpose of this study. There may also have been
important measurement errors with several variables.
Finally, and most importantly, the time lag between the onset of the illness
and the patient's admission to the hospital may vary from illness to illness
and according to the availability of medical services. The lags are likely
to be much greater in some instances than could be considered in an analysis
using the data studied here. It becomes apparent that much more specific
data are required for the analysis of inpatient medical costs' relationship
to air pollution. Such data would have to be collected at the time of the
patient's admission to the hospital. The information would have to focus
on the history of the patient's well-being during the time between the onset
of the current disease episode and his admission to the hospital. Only if
such information is available will it be possible to relate in a meaningful
26
-------
way changes in air pollution levels to medical services provided by hospi-
tals. It is unlikely that such data will become available unless a consid-
erable commitment is made to generate them as patients' hospital medical
records develop. This would involve a long-run cooperative research project
between the medical profession and social scientists.
27
-------
SECTION VII
REFERENCES
1. Brubaker, Sterling. To Live on Earth: Man and His Environment in
Perspective. New York: The Johns Hopkins Press, 1972.
2. Burch, George E. and Nicholas P. Depasquale. Hot Climates, Man and
His Heart. Springfield, Illinois: Charles C. Thomas, 1962.
3. California Medical Association. Committee on Fees of the Commission
on Medical Services. 1964 California Relative Value Studies. 4th ed.
San Francisco: Six Ninety Sutter, 1964. 79 p.
4. California Medical Association. Committee on Fees of the Commission
on Medical Services. 1969 California Relative Value Studies. 5th ed.
San Francisco: Six Ninety Sutter, 1969. 135 p.
5. Cassell, Eric J. Air Pollution, Weather, and Illness in a New York
Population, Archives of Environmental Health. AMA: Chicago, Illinois,
18:523-533, April, 1969.
6. Conservation Foundation, The. A Citizen's Guide to Clean Air.
Washington, D.C., 1972.
7. Durham, William H. Air Pollution and Student Health. Archives of
National Health. AMA: Chicago, Illinois, 28:241-254, May, 1974.
8. Faith, W. L. and Arthur A. Atkisson, Jr. Air Pollution. 2nd ed.
New York: Wiley-Interscience, 1972.
9. Holland, W. W. Air Pollution and Respiratory Disease. Connecticut:
Technomic, 1972*
10. Holland, W. W. and D. D. Reid. The Urban Factor in Chronic Bronchitis.
The Lancet, London, 1:7383, 1965.
11. Jaksch, John A. Outpatient Medical Costs Related to Air Pollution in
the Portland - Vancouver Area. Unpublished Ph.D. thesis, Oregon State
University, Corvallis, 1972.
12. Jaksch, John A. and H. H. Stoevener. Outpatient Medical Costs Related
to Air Pollution in the Portland, Oregon Area. Washington, D.C., U.S.
Environmental Protection Agency, Socioeconomic Environmental Studies
Series, July, 1974.
13. Lave, Lester B. and Eugene P. Seskin. Air Pollution and Human Health.
Science, 169: 723-733, 1970.
14. Lawthen, P. J. Air Pollution and Tobacco Smoke. Clinical Aspects of
Inhaled Particles* London: William Heinemann Medical Books Limited,
1972.
28
-------
15. Muir, D.C.F. Airborne Allergens. Clinical Aspects of Inhaled Particles.
London: William Heinemann Medical Books Limited, 1972.
16. National Tuberculosis and Respiratory Disease Association. Air Pollution
Primer. New York, 1969.
17. Saward, E., Janet Blank, and Henry Lamb. Some Information Descriptive of
a Successfully Operating HMO. Rockville, Maryland. Health Service and
Mental Health Administration, DHEW Publication (HSM) 73-13011, 1973.
18. Sterling, Theodor D., J. Weikam, and S. V. Pollack, "Measuring the Effect
of Air Pollution on Urban Mortality," Archives of Environmental Health,
AMA, Chicago, Illinois. 18:485-494, April 1969.
19. Tromp, S. W. Medical Biometeorology: Weather, Climate, and the Living
Organism. Amsterdam: Elsevier, 1963*
20. U.S. Department of Commerce, National Oceanic and Atmospheric Administra-
tion, Environmental Data Service, Local Clinatological Data: Annual
Summary with Comparative Data, Portland, Oregon. Washington, B.C., 1970.
21. U. S. Public Health Service. National Center for Health Statistics.
Eighth Revision of International Classification of Diseases. Vol. 2.
Washington, D.C. Government Printing Office, PHS Publication 1693, 1969.
22. . . Monthly Vital Statistics. Vols. 18 and 21,
Washington, D.C. 1969 and 1972.
23. Wirick, Grover and Robin Barlow. "The Economic and Social Determinants
of the Demand for Health Services," Proceedings of the Conference on
the Economics of Health and Medical Care, 1962. Ann Arbor, the Univer-
sity of Michigan, 1964.
24. Zeidberg, L.D., R.J.M. Horton, and E. Landau, "The Nashville Air Pollu-
tion Study: VI* Cardiovascular Disease Mortality in Relation to Air
Pollution," Archives of Environmental Health, AMA, Chicago, Illinois,
15: 225-236, 1967.
29
-------
SECTION VIII
APPENDIX
APPENDIX TABLE 1. Classification of Diseases
I. Diseases of the respiratory system
Upper respiratory infections
Lower respiratory
II. Allergies affecting the respiratory system
Asthma
Hay Fever
Specific diseases
III. Diseases of the circulatory system
Heart
Circulatory system
IV. Other allergies and skin diseases
Eczema
Other allergies
Other skin diseases
Specific diseases - allergies
V. Diseases of the eye
Inflamatory disease of the eye
Non-specific eye diseases
30
-------
APPENDIX TABLE 2. Correlation Matrix of Variables - Respiratory Diseases
Marital Family Persons in
Variables Age Sex Race Status Income Household Drinking
Age 1.000 -.045 -.095 .473 -.246 -.567 -.309
Sex 1.000 -.008 -.071 -.029 .024 -.097
Race 1.000 -.077 -.071 .054 -.048
Marital Status 1.000 .022 -.101 -.481
Family Income 1.000 .236 .055
Persons in
Household 1.000 -.111
Drinking 1.000
Smoking
Occupational
Exposure
Number of
Disabilities
U> Number of
*~* Hospital visits
Education
THI unlagged
THI lagged
1 day
THI lagged
2 days
THI lagged
3 days
Air Pollution
Unlagged
Air pollution
Lagged 1 day
Air pollution
Lagged 2 days
Air pollution
Lagged 3 days
Number
Occupational Number of Hospital
Smoking Exposure Disabilities Visits
-.269 -.032 .202 .182
-.106 -.380 -.019 .013
.017 .005 -.061 .055
.344 .057 .096 .149
-.027 -.002 -.107 -.127
-.109 .033 -.061 -.028
.450 .119 .066 .089
1.000 .137 .039 .093
1.000 -.025 -.037
1.000 .153
1.000
Medical Costs
Continued
-------
APPENDIX TABLE 2. (continued)
THI
Variables Education unlagged
Age -.347 .002
Sex .032 .002
Race -.040 -.015
Marital Status .016 -.004
Family Income .309 .017
Persons in
Household .418 .010
Drinking .066 -.037
Smoking -.008 .004
Occupational
Exposure 0 -.018
Number of
Disabilities .034 .003
CO Number of
10 Hospital visits -.002 .023
Education 1.000 .018
THI unlagged 1.000
THI lagged
1 day
THI lagged
2 days
THI lagged
3 days
Air pollution
Unlagged
Air pollution
Lagged 1 day
Air pollution
Lagged 2 days
Air pollution
Lagged 3 days
V*-. J4 _ *• 1 S^JH A^ A
THI
lagged
1 day
.009
-.001
-.019
-.004
.024
-.022
-.034
.003
-.023
.006
.022
.011
.925
1.000
THI
lagged
2 days
.005
.011
-.010
-.017
.019
-.002
-.031
-.004
-.018
0
.016
.011
.845
.927
1.000
Till
lagged
3 days
-.002
.011
-.015
-.023
.018
.010
-.023
.007
.016
.009
'.019
.025
.802
.849
.931
1.000
Air
pollution
unlagged
-.030
.023
.007
.007
.020
.035
.004
.004
.011
-.003
-.007
.028
.198
.165
.125
.114
1.000
Air pollu-
tion lagged
1 day
-.022
.027
.001
-.002
.018
.043
0
.010
.015
..003
-.007
.027
.215
.189
.151
.128
.924
1.000
Air pollu-
tion lagged
2 days
-.014
.017
-.003
-.011
.014
.042
-.003
.014
.021
.007
-.101
.027
.220
.210
.180
.156
.729
.910
1.000
Air pollu-
tion lagged
3 days
-.009
.002
-.008
-.014
.007
.035
-.004
.020
.027
.012
-.010
.027
.212
.214
.200
.185
.504
.716
.916
1.000
Medical
Costs
.332
-.002
-.027
.224
-.011
-.156
.193
.148
.016
.057
.016
-.082
.017
.017
.027
.021
.013
.012
.004
.004
1.000
-------
APPENDIX. TABLE 3. Correlation Matrix of Variables - All Diseases
Marital Family Persons in
Variables Age Sex Race Status Income household Drinking
Age 1.000 -.012 -.108 .391 -.301 -.600 .218
Sex 1.000 -.006 -.118 -.044 -.006 -.116
Race 1.000 -.065 -.062 .087 -.022
Marital Status 1.000 .021 -.042 .404
Family income 1.000 .250 .064
Persons in
household 1.000 -.063
Drinking 1.000
Smoking
Occupational
exposure
Dumber of
disabilities
U» Number of
hospital visits
Education
THI unlagged
THI lagged
1 day
THI lagged
2 days
THI lagged
3 days
Air pollution
Unlagged
Air pollution
Lagged 1 day
Air pollution
Lagged 2 days
Air pollution
Lagged 3 days
Number of
Occupational Number of hospital
Smoking exposure disabilities visits
.184 -.067 .201 .192
-.125 -.376 -.031 -.019
.029 .013 -.058 .077
.295 .070 .100 .168
-.013 .023 -.096 -.147
-.063 .061 -.053 -.025
.436 .109 .064 .096
1.000 .152 .026 .097
1.000 0 -.024
1.000 .185
1.000
Medical Costs
Continued
-------
APPENDIX TABLE 3. (continued)
XHI
THI lagged
Variables Education unlagged 1 day
Age -.389 -.002 -.003
Sex .017 .004 0
Race .017 -.017 -.022
Marital Status .032 .001 -.004
Family Income .306 .008 .014
Persons in
Household .441 .010 .029
Drinking .095 -.034 -.032
Smoking .029 .010 .011
Occupational
Exposure .026 -.018 -.021
Number of
Disabilities .051 -.006 -.005
.p- Number of
Hospital visits .002 .003 .004
Education 1.000 .014 .007
THI uolagged 1.000 .026
Till lagged
1 day 1.000
THI lagged
2 days
THI lagged
3 days
Air pollution
Unlagged
Air pollution
Lagged 1 day
Air pollution
Lagged 2 days
Air pollution
Lagged 3 days
Medical Costs
Till
lagged
2 days
-.007
.012
-.015
-.016
.012
0
-.032
.006
-.017
-.009
.001
.009
.851
.931
1.000
•nil
lagged
3 days
-.015
.014
-.014
-.023
.019
.010
-.025
.010
-.012
-.008
.001
.024
.806
.851
.931
1.000
Air
pollution
unlagged
-.018
.029
.008
.023
.009
.024
.016
.023
.018
-.013
-.015
.022
.190
.155
.115
.109
1.000
Air pollu-
tion lagged
1 day
-.014
.038
.001
.011
.012
.027
.006
.019
.018
-.006
-.018
.022
.204
.177
.140
.120
.024
1.000
Air pollu-
tion lagged
2 days
-.008
.034
-.005
0
.011
.025
-.008
.015
.019
-.003
-.019
.023
.208
.198
.170
.147
.731
.912
1.000
Air pollu-
tion lagged
3 days
-.005
.023
-.009
-.003
.011
.020
-.016
.016
.023
0
-.019
.023
.202
.205
.192
.178
.509
.723
.918
1.000
Medical
Costs
.303
0
-.020
.170
-.034
-.163
.156
.111
-.007
.071
.033
-.078
.010
.010
.018
.016
.004
.005
0
.001
1.000
-------
TECHNICAL REPORT DATA
(Please read Instructions on the reverse before completing)
1. REPORT NO.
EPA-600/5-78-002
2.
3. RECIPIENT'S ACCESSION-NO.
4. TITLE AND SUBTITLE
IMPACT OF AIR POLLUTION ON THE CONSUMP-
TION OF MEDICAL SERVICES — Cost of Hospitalization in
the Portland Metropolitan Area
5. REPORT DATE
January 1978
6. PERFORMING ORGANIZATION CODE
7. AUTHOR(S)
Gobind S. Bhagia
Herbert H. Stoevener
8. PERFORMING ORGANIZATION REPORT NO.
9. PERFORMING ORGANIZATION NAME AND ADDRESS
Department of Agriculture and Resource Economics
Oregon State University
Corvallis, OR 97331
10. PROGRAM ELEMENT NO.
1HA094
11. CONTRACT/GRANT NO.
68-01-0482
12. SPONSORING AGENCY NAME AND ADDRESS
Environmental Research Laboratory—Corvallis
Office of Research and Development
U.S. Environmental Protection Agency
Corvallis, Oregon 97330
13. TYPE OF REPORT AND PERIOD COVERED
Final
14. SPONSORING AGENCY CODE
EPA/600/02
15. SUPPLEMENTARY NOTES
16. ABSTRACT
This study was undertaken to ascertain the possibility of measuring from available
data the impact of air pollution on the consumption of inpatient services which
patients consume per hospital stay.
The study area was the Portland, Oregon metropolitan area, and the period of study
was 1970-72. The diseases which were thought to be associated with air pollution were
first identified. A sample of the patients was taken from a hospital population and
the medical costs incurred by them were calculated. Suspended particulates as a meas-
ure of air pollution, temperature-humidity index as a measure of meteorological condi-
tions of the area, and socio-economic characteristics of patients were used in a regre
sion model to explain variations in medical costs.
The analysis did not indicate a quantitative relationship between air pollution in the
Portland area and the consumption of inpatient medical services per hospital stay.
Age, family income, and drinking habits of a patient do significantly affect inpatient
medical costs, but the variables included in the analysis accounted for less than 15
percent of the statistical variation in the dependent variable.
7.
KEY WORDS AND DOCUMENT ANALYSIS
DESCRIPTORS
b.lDENTIFIERS/OPEN ENDED TERMS
c. COSATI Field/Group
Air Pollution Effects (Health)
Economic Analysis
Economic Effects, Air Pollution
Economic Effects, Health
Environmental Economics
Air Pollution Economics
Health Economics
05C
Behavioral and
Social Science/
Economics
18. DISTRIBUTION STATEMENT
Unlimited
19. SECURITY CLASS (ThisReport)
21. NO. OF PAGES
20. SECURITY CLASS (Thispage/
Unclassified
22. PRICE
EPA Form 2220-1 (9-73)
35
------- |