Citation
Neighbored but not wanted

Material Information

Title:
Neighbored but not wanted why do rural patients bypass rural hospitals? : a Colorado case study
Creator:
Roh, Chul-Young
Publication Date:
Language:
English
Physical Description:
212 leaves : illustrations ; 28 cm

Subjects

Subjects / Keywords:
Rural hospitals -- Case studies -- Colorado ( lcsh )
Hospital utilization -- Case studies -- Colorado ( lcsh )
Rural health services -- Utilization -- Case studies -- Colorado ( lcsh )
Hospital and community -- Case studies -- Colorado ( lcsh )
Hospital and community ( fast )
Hospital utilization ( fast )
Rural health services -- Utilization ( fast )
Rural hospitals ( fast )
Colorado ( fast )
Genre:
Case studies. ( fast )
bibliography ( marcgt )
theses ( marcgt )
non-fiction ( marcgt )
Case studies ( fast )

Notes

Bibliography:
Includes bibliographical references (leaves 202-212).
General Note:
School of Public Affairs
Statement of Responsibility:
Chul-Young Roh.

Record Information

Source Institution:
|University of Colorado Denver
Holding Location:
|Auraria Library
Rights Management:
All applicable rights reserved by the source institution and holding location.
Resource Identifier:
53378984 ( OCLC )
ocm53378984
Classification:
LD1190.P86 2002d .R63 ( lcc )

Full Text
NEIGHBORED BUT NOT WANTED: WHY DO RURAL PATIENTS
BYPASS RURAL HOSPITALS? A COLORADO CASE STUDY
by
Chul-Young Roh
B. A. In-ha University, 1987
M.A. In-ha University, 1989
M.P.A. New York University, 1992
A thesis submitted to the
University of Colorado at Denver
in partial fulfillment
of the requirements for the degree of
Doctor of Philosophy
Public Administration
2002
[At.


2002 by Chul-Young Roh
All rights reserved
1


This Thesis for the Doctor of Philosophy
degree by
Chui-Young Roh
has been approved
Sheri Eisert
Date


Roh, Chul-Young (Ph. D. Public Administration)
Neighbored But Not Wanted: Why Do Rural Patients Bypass Rural
Hospitals? A Colorado Case Study
Thesis directed by Assistant Professor M. Jae Moon
ABSTRACT
American rural hospitals play an important role in their local
communities. They provide significant health care services and
employment. But, they provide a more limited range of health care
services. These factors cause patients to bypass local rural hospitals.
For the last decade, a steadily growing trend of patients bypassing
services in local rural hospitals has appeared in rural areas. Despite
distance proximity, many rural patients tend to bypass rural hospitals and
travel to hospitals outside their areas of residence. For example, about 43
percent of all rural patients in Colorado having bypassed their local rural
health care facilities during 1990s. The effect of this phenomenon is a
reduction in occupancy rate and decrease in the competitiveness of these
facilities, thereby ultimately causing rural hospitals to close and adversely
affecting the communities that they were designed to serve.
This study is designed to explore why and how Colorado rural patients
bypass their local rural hospitals and determine the factors that influence
Colorado rural patients to bypass their local rural hospitals and their
effect on rural health care policy and management.
To figure out the factors that influence patients to bypass their local rural
hospitals, binary and multinomial logistic regressions were used. This
study found the principal variables of bypassing: (1) distance between the
nearest rural hospital and the patient's residence, (2) the proximity of
rural patients to an urban county, and (3) the lack of a network with other
rural hospitals and urban hospitals, while other variables such as patient's
IV


age or hospital ownership type are also influential in the trend towards
bypassing local hospitals. The implications for rural health care
management and policy are the reduction of health care services to better
meet what the rural community wishes and can afford, the creation of
assurance areas of essential health care services and assistance to rural
patients who would have difficulty leaving the area to gain access to full
health care services when needed.
This abstract accurately represents the contents of the candidate's thesis. I
recommend its publication.
Signed
M. Jae Moon
v


DEDICATION
This dissertation is dedicated with love to my wife, Hyun Sook Kim, my
son Jong Woo Roh: without their love, patience and support, my graduate
education would not have been possible.


ACKOWLEDGEMENT
The completion of this dissertation would have been impossible the help
and encouragement I received from a number of individuals. My
dissertation chair and academic adviser Professor M. Jae Moon provided
immeasurable guidance and support throughout the process, I can never
thank him enough.
I would thank my other committee members: Professor Linda deLeon, Dr.
Sheri Eisert, Professor Christine Martell and Professor Laura Argys.
Professor Linda deLeon helped me to develop several of the key
arguments that I make in this dissertation, and she offered enthusiastic
support and helpful advice throughout the research and writing phases.
Dr. Sheri Eisert always encouraged me to do this project. She initially
provided the research idea and data. Professor Christine Martell and
Laura Argys offered invaluable insights on the research methodology. I
feel fortunate to have had such a wonderfully supportive and
intellectually engaged dissertation committee.
Other faculty of Graduate School of Public Affairs also made important
contributions to my academic experience. In his capacity as doctoral
director, professor Peter deLeon provided support and guidance. He
played a formative role in training me as a public policy scholar.
I also wish to express my gratitude to Mr. Michael Boyson, Director of
Data Services Research of Colorado Health and Hospital Association. He
provided the research data every year. He sincerely answered my
questions regarding Colorado hospitals and research data.
My thanks go to Insoo Kang, Director of Behavioral Social Science Lab and
Jong-Yun Ahn, lab technician who provided the invaluable assistance.
Especially, I want to thank my wife and son for making me feel her
constant, unlimited, and unconditional support, patience, sacrifice and
love.


CONTENTS
Figures ................................................. xii
Tables .................................................. xiii
CHAPTER
1. INTRODUCTION ............................................ 1
Background ....................................... 1
Purpose of the Study ............................. 3
Scope of the Study ............................... 4
Outline of the Study ............................. 5
2. COLORADO RURAL HOSPITALS
AND THE RURAL ENVIRONMENT .............................. 7
The Definition of Rural Hospitals ................ 8
The Definition of Rural and Urban ....... 8
The Role of Hospitals .................. 10
Colorado Rural Hospitals ........................ 13
The Issues of Colorado Rural Hospitals .......... 24
Declining Occupancy Rate of Rural
Hospitals and Increasing Bypassing
of Local Rural Hospitals ................ 25
viii


Decreasing Profit Margins and Increasing
Tax Subsidies to Rural Hospitals .......... 28
Increasing Closure of Rural Hospitals ..... 33
3. LITERATURE REVIEW ........................................ 37
Interdisciplinary Backgrounds on
Health Care Demand.................................. 38
The Geographical Approach
and Hospital Use .......................... 38
The Economic Approach and Hospital Use
..... 45
The Primary Care Choice............................. 50
The Research of Hospital Choice..................... 55
Hospital Choice by Rural Patients .................. 65
4. THEORETICAL FRAMEWORK AND
RESEARCH HYPOTHESIS ...................................... 72
Institutional Dimensions ........................... 75
Payment Sources ........................... 75
Hospital Characteristics .................. 79
Individual Dimensions .............................. 87
Patient Conditions ........................ 87
Individual Characteristics ................ 90
ix


5. METHODOLOGY AND DATA
100
Methodology ........................................ 101
Binary Choice Model ....................... 101
Multinomial Choice Model................... Ill
Data Used in the Study ............................. 114
Colorado Inpatient Discharge Data (CIDD)... 115
Patient Selection ......................... 120
Hospital Data ............................. 122
Geographic Information Data ............... 123
Variables Definitions .............................. 126
Dependent Variables ....................... 126
Independent Variables ..................... 128
Institutional Sector ............. 128
Individual Sector ................ 130
Summary ............................................ 132
6. ANALYSIS RESULTS ......................................... 134
Descriptive Statistic Results ...................... 136
Logistic Regression Analysis ....................... 151
Binary Logistic Regression Analysis ....... 151
x


Multiple Binary
Logistic Regression Analysis ............. 162
Binary Logistic Regression
Analysis (Local Rural Hospital vs.
Other Rural Hospital)........... 164
Binary Logistic Regression
Analysis (Local Rural Hospital vs.
Urban Hospital)................. 168
Multinomial Logistic
Regression Analysis....................... 173
7. CONCLUSION AND DISCUSSION ............................ 187
Introduction ..................................... 187
Overview of Principal Research Findings .......... 188
Implications for Rural Health Care
Management and Policy ............................ 191
Limitations of the Study ......................... 194
Future Research Directions ....................... 198
Conclusion ....................................... 201
REFERENCES ............................................... 202
xi


FIGURES
Figure
2-1 Colorado Hospitals .......................................... 12
2-2 Bypassing and Occupancy Rates in Colorado
(1993-2000) ............................................. 23
2-3 The Percent of Patient Service Margin (1992-1998)............ 29
2-4 Colorado Rural Hospital Subsidies (1992 -1998) .............. 30
4-1 Patient's Hospital Choice ................................... 98


TABLES
Table
2-1 Colorado Hospitals by Size, Ownership, and Location, 1999 14
2-2 Rural and Urban Health Related Characteristics in State of Colorado, 1998 18
3-1 Selected Hospital Choice Literature 57
6-1 The Diagnostic Descriptive Statistic of Colorado Rural Patients (1993 2000) 135
6-2 Hospital Use by Colorado Rural Patients According to Major Payment Sources (1993 2000) 139
6-3 Hospital Use by Colorado Rural Patients According to Referral Types (1993 2000) 142
6-4 Hospital Use According to Colorado Rural Patients' Age (1993 2000) 144
6-5 Colorado Rural Patients' Hospital Use According to Medical Conditions (1993 2000) 145
6-6 Hospital Use of Colorado Rural Patients by Hospital Ownership Types (1993 2000) 148
6-7 Length of Stays by Colorado Rural Patients (1993 2000) 149
6-8 Binary Logistic Regression Result of Hospital Choice by Colorado Rural Patients (n=l 9,618) 152
xm


6-9 Binary Logistic Regression Result of
Hospital Choice by Colorado Rural Patients
(n=16,278) (Local Rural vs. Other Rural).............. 163
6-10 Binary Logistic Regression Result of
Hospital Choice by Colorado Rural Patients
(n=15,757) (Local Rural vs. Urban).................... 169
6-11 Mutinomial Logistic Regression Result of
Hospital Choice by
Colorado Rural Patients (n=19,618) ................... 176
6-12 The Results of the Hypotheses........................... 184
xiv


CHAPTER 1
INTRODUCTION
Background
In America, rural hospitals differ from urban hospitals in significant
ways. More specifically, the challenges that rural hospitals face include
difficulties in attracting health care practitioners, dealing with an aging
population in need of health care services (Ermann, 1990), handling
inequitable reimbursement from Medicare and Medicaid resulting in higher
proportions of uncompensated care, coping with higher outlays of monies
relative to patient revenues, and an increasing rate of bypassing by rural
residents to larger health care facilities for health care and treatment (Forti et
al., 1999). At the same time, rural hospitals occupy an especially critical
position within the local communities where they furnish health care
services. Moreover, they serve to provide a source of civic self-esteem and
are vital to the economy of many rural communities as the first, second, and
third largest employer of the rural citizenry (CRHC, 1999).
1


The viability and accessibility of rural hospitals have repercussions for
rural residents, politicians, policy analysts, and academics for several reasons.
First, American society accepts the common norm that everybody should
have equal access to hospital use. However, rural populations often have
difficulty in accessing hospitals, with some rural residents living 30 minutes,
or more, from the nearest hospital (CRHC, 1999). Second, the availability of
health care services in rural communities make the community a more
desirable place to live especially for new residents, such as the elderly
(Cordes, 1989).
A principal aspect of the difficulty that rural hospitals confront is a low
occupancy rate. These hospitals depend on a steady flow of substantial
numbers of patients or at least enough to keep occupancy at an optimum
level. Yet, a sizable portion of rural patients seek inpatient hospital services
at health care institutions outside their community area, even though the
local hospitals can provide the health care services that the patients require.
The bypassing of local hospitals by rural patients is a primary contributing
factor to the poor financial condition of rural hospitals, and subsequently to
the closure of many rural hospitals.
2


Purpose of the Study
This bypassing of rural hospitals continually increased in Colorado
rural communities during the 1990s. To understand this phenomenon, a
detailed understanding of the factors that influence hospital choice is
required. This study seeks to understand why Colorado rural patients
bypass their local rural hospitals when they could potentially access the
health care services that they need at their nearest local hospitals. This study
will expand the knowledge of hospital choice of rural patients by
investigating their decision not to use their local rural hospital, in other words,
to bypass the closest facility. The findings of this study offer critical
implications for rural health care management and policy.
The purpose of this study is to identify empirically both individual
factors and institutional variables associated with hospital choice behavior
among Colorado rural residents during the 1990s. This study will estimate a
disaggregate hospital choice model from the Colorado Inpatient Discharge
Data (CIDD). The CIDD, which contains patients' diagnostic information
such as the length of stays, the names of diseases, total charges, payment
sources, referral sources, and demographic information, such as patient's age,
zip code, and gender, will be used. Furthermore, since CIDD provides the
3


patients' diagnostic and demographic information, the use of CIDD will
enhance the ability to generalize the findings of this study to other states'
rural patient hospital choice behavior.
This study will attempt to answer the following research questions;
(1) What individual and hospital characteristics influence rural
patients to bypass the local rural hospitals to obtain health care
services?
(2) Which individual and hospital characteristics are associated with
the choice of the local rural hospital vs. other rural hospitals or
urban hospitals?
(3) To what extent does the distance between the nearest rural hospital
and patient's residence, and the distance between the nearest rural
hospital and the second next nearest hospital, influence the rural
patient to choose a hospital in order to health care services?
To answer these questions, an in-depth comprehension of variables that
influence the hospital choice of rural patient is called for.
Scope of the Study
In the course of answering why rural patients bypass their local rural
hospital to obtain health care services, even though these facilities can
4


provide the health care services that rural patients need, this study focuses
upon rural patients in Colorado. Among Colorado rural patients, this study
limits these patients to those in need of major diagnostic code (MDC) 14
(pregnancy, childbirth, and puerperium) that Colorado rural hospitals
commonly provide.
Among a handful of research concerning the locational choice of rural
populations seeking health care services, this study focuses upon how rural
female patients select a hospital for inpatient hospital services. These factors
will be discussed, in detail, throughout this study.
Outline of the Study
As stated above, to address the primary research questions, Chapter 2
describes Colorado rural hospitals and the rural environment in which they
operate. Chapter 3 surveys the literature related to the analysis of rural
residents' hospital choice behavior for health care services in order to answer
the above stated research question regarding the factors that influence rural
patients to choose hospitals for health care services. Chapter 4 develops a
theoretical framework and research hypotheses concerning the hospital
choice of rural patients. The existing literature has identified important
5


determinants that influence patients' hospital choice. In particular, this
study surveys the determinants that influence patients to choose their
hospital based on two dimensions: (1) institutional factors and (2) individual
ones. Chapter 5 describes how these methodologies are applied to the
patient's hospital behavior in order to model the intensity dimension of the
use for hospital services, and expatiates upon the nature and characteristics of
data sources used in exploring the influence of the factors that make Colorado
rural patients choose a specific hospital to seek health care services. Chapter
6 presents the descriptive statistics and logistic regression results. The
descriptive results explain the characteristics of the Colorado rural patients
who have MDC 14 (pregnancy, childbirth, and puerperium) including their
age, length of stays, total charges, DRG weight, medical conditions, their
payment sources, as well as proximity to an urban Colorado county.
Logistic regression models will determine the factors that influence the rural
patient to choose a hospital. The implications for rural health care
management and policy are outlined in Chapter 7. The upcoming chapters
detail the significance of patient bypassing on local rural hospitals and
provide a policy context for this phenomenon.
6


CHAPTER 2
COLORADO RURAL HOSPITALS AND
THE RURAL ENVIRONMENT
The purpose of this study is to determine which factors influence
Colorado rural patients to choose specific hospitals for health care services
and treatment. Even though local rural hospitals provide the health care
services that rural patients seek, these patients bypass their local rural
hospitals, and consequently, these hospital increasingly confront closure.
This study attempts to grasp the nature of rural hospitals and the
environment in which they operate in order to examine patient behavior with
regard to hospital choice and its implications for health care policy. Before
determining the effect of factors that influence hospital choice by rural
patients, this study begins by describing the rural/ urban definition and
exploring the role of hospitals in these settings.
7


The Definition of Rural Hospitals
The Definition of Rural and Urban
The most commonly used definitions of rural and urban communities
are ones that are given by Office of Management and Budget (OMB) and the
Bureau of Census. The OMB makes a dichotomous separation of the
country into either Metropolitan Statistical Areas (MSAs) or non-metropolitan
areas, with a metropolitan area defined as either a county with a city of at
least 50,000 residents, or an urbanized area being part of a county or counties
with a minimum of 100,000 inhabitants (Cohen, 1993). The Bureau of
Census defines an urban area as any area with a central city with a
population of 50,000 or more and adjacent territory of more than 2,500 people
living outside the central city limits.
Literature covering hospital markets generally assumes that the
market area coincides with an existing geographic entity (Goody, 1993). For
rural hospitals, the county is usually assumed to be the appropriate
8


geographic unit for analysis; for urban hospitals, the MSA has been also used
as an appropriate unit (Folland, 1983; Noether 1988; Hogan 1989). When
considering the theoretical underpinning of market definition, it is unlikely
that hospital markets will conveniently coincide with these geographic
boundaries (Garnick et al., 1987; Wright et al., 1990). As a geographic unit, a
county may be an overly narrow unit of measure in many situations.
Moreover, it can be argued that MSAs are too large to represent true
geographic markets and are by definition, not applicable to rural areas.
Much of the national literature on utilization and hospital choice cited utilizes
county-based measurements or geopolitical units (Bronstein, 1990,1991;
Goldsteen, 19945; Hogan 1988). This study will also utilize the definition of
a county to describe a geographical unit, and to classify urban and rural areas
of Colorado. There are 63 counties in Colorado. Eleven counties are urban
ones, while 52 counties are rural. This study will primarily focus on the
majorities, which are rural in their make-up.
9


The Role of Hospitals
The role of hospitals is to treat illness and disease as well as to
promote overall health and prevent disease in communities. The
appearance of new medical technologies allows hospitals move to the
forefront where technological advances and well-trained professionals
provide health care services. While hospitals have principally focused on
providing medical treatment in the past, recent shifts in reimbursement
methods have compelled hospitals to increase the percentage of their services
in the outpatient arena, such as rehabilitation and one day surgery.
The role of hospitals increasingly requires an emphasis on health
promotion and preventive care. Research has shown that exercise, nutrition,
stress reduction, and avoidance of high risk behaviors play a critical role in
prolonging good health, preventing illness and reducing the severity of
diseases (McKeown, 1994). Hospitals have recognized that it makes good
sense to engage in these activities as part of a more holistic approach to health
(Friedman et al., 1999). Supporting that premise, hospitals provide a wide of
10


range of health care programs, such as childbirth education, community
health promotion programs, and worksite health promotion.
Traditionally, an important role of hospitals has been to educate new
health care professionals, such as physicians, registered nurses, pharmacists,
physical and occupational therapists, and medical technicians. The
academic health centers and large teaching hospitals provide physician
education and a wide range of clinical training. Like trained medical interns,
a number of health care administration students choose to do their
internships in hospital settings.
Another role for hospitals involves providing facilities and resources
for critical clinical research. Before approval by the Food and Drug
Administration (FDA), new medical technologies, procedures, and
pharmaceutical products require lengthy clinical tests on human subjects.
This kind of activity is often carried out at academic health centers, teaching
hospitals, and specialized research hospitals.
11


Figure 2-1 Colorado Hospitals


Another increasing role of hospitals is to be responsible for
maintaining the healthy status of their communities. The shift of the health
care system from inpatient treatment to outpatient treatment requires that
hospitals become more involved in community health issues, in new and
different ways, than before. In the past, patients rarely had the luxury of
choice about when or where they were hospitalized. Today, much more
educated and quality-concerned patients demand that hospitals provide
health care services at the lowest cost and with the highest quality care
possible. Hospitals have begun to treat their patients as valued health care
customers.
Colorado Rural Hospitals
As one of the health care providers in a community, hospitals have
become the primary setting for the delivery of health care services because
they have the most sophisticated health care technology and equipment.
This is especially true in a rural community. The 1998 national data
13


Table 2-1 Colorado Hospitals by Size, Ownership, and Location, 1999
Rural Hospital Bed Size Urban Hospital Bed Size Total
>49 50-99 100+ All >49 50-99 100+ AH
Public1) 18 (50.0%) 4 (11.1%) 0 (0.0%) 22 (61.1%) 0 (0.0%) 1 (3.4%) 3 (10.3%) 4 (13.8%) 26 (40.0%)
Non- Profit2' 5 (13.9%) 5 (13.9%) 2 (5.6%) 12 (33.3%) 2 (6.9%) 2 (6.9%) 18 (62.1%) 22 (75.9%) 34 (52.3%)
For- Profit3) 1 (2.8%) 1 (2.8%) 0 (0.0%) 2 (5.6%) 0 (0.0%) 0 (0.0%) 3 (10.3%) 3 (10.3%) 5 (7.7%)
Total 24 (66.7%) 10 (27.8%) 2 (5.6%) 36 (100.0%) 2 (6.9%) 3 (10.3%) 24 (82.8%) 29 (100.0%) 65 (100.0%)
1) Public: County, City, City-Hospital District of Authority
2) Non-Profit: Non-Government, Not-For-Profit Organization (Church Operated, Other)
3) For-Profit: Investor-Owned (Individual, Partnership, and Corporation)
Source: American Hospital Association, AHA Guide to the Health Care Field, 1998-1999 edition, Chicago, IL, 1999.


indicates that out of 5,015 acute care hospitals, 2,199 or 43.8 percent are rural
facilities (AHA, 2000). In 2000,36 (or 55.4 percent) of Colorado's sixty-five
acute care hospitals were located in rural areas (CHHA, 1999).
Colorado rural hospitals occupy a uniquely critical position within the
local communities where they furnish health care services due to the lack of
health care resources in those communities. Moreover, they serve to provide
a source of civic pride and are vital to the economy of many rural
communities as the first, second, or third largest employer in these local areas
(CRHC, 1999).
Rural hospitals differ from urban hospitals in significant ways.
Specifically, the challenges that confront rural hospitals include: 1) difficulties
in attracting health care practitioners; 2) dealing with an aging population in
need of health care (Ermann, 1990); 3) suffering from inequitable
reimbursement from Medicare and Medicaid programs resulting in higher
proportions of uncompensated health care claims; 4) coping with higher
outlays of monies relative to patient revenues; and 5) increasing migration of
15


rural residents to larger rural health care facilities or urban ones to obtain
health care services (Forti et al., 1999).
Colorado rural hospitals serve counties that include many of the state's
poorer families. These counties are often geographically isolated from urban
centers. Severe winter weather adds to the problem of rural isolation as
travel becomes nearly impossible. Among 36 rural facilities, 22 rural
hospitals are publicly owned by counties or hospital districts (CHHA, 1998).
The number of beds in each of the 24 Colorado rural hospitals is under 50,
while Colorado urban hospitals are much more likely to be owned by not-for-
profit organizations with over 100 beds. Although Colorado rural hospitals
are small in number and size, they are critical to the delivery of health care
services in Colorado. They serve both as a cornerstone of the economic
infrastructure of rural communities and as the focal point for the delivery of
health care services for the residents of these communities.
Colorado rural hospitals face a very different situation from Colorado
urban hospitals. The average total inpatient charge per day at Colorado
16


rural hospitals is $1,437, which is 54.5 percent of the average total charge in
urban hospitals ($2,639) (CHHA, 1999). Although this number may seem
incongruous, it seems conceivable that rural hospitals in Colorado may
provide simpler health care services than urban facilities.
Despite being smaller, the occupancy rate for Colorado rural hospitals is
around 35.9 percent compared with 56.0 percent of Colorado urban hospitals
(CHHA, 1999). This figure suggests that Colorado rural hospitals will
inevitably have a lower profit margin compared to their urban counterparts.
Coupled with a lower occupancy rate, like other state rural hospitals,
Colorado rural hospitals depend more on public revenue such as Medicare
and Medicaid beneficiaries compared with urban hospitals. In 1994,
Medicare in Colorado rural hospitals accounted for 47.5 percent of
admissions compared with 40.5 percent of admissions at urban hospitals
(CRHC, 1999). The dependence on Medicare was made more significant by
the passage of the Social Security Amendments of 1983 (Ermann, 1990).
Medicare replaced the retrospective cost-plus reimbursement method with
17


Table 2-2 Rural and Urban Health Related Characteristics in State of Colorado, 1998
Rural Urban
Average Daily Charge Per Day1) $1,437 $2,639
Occupancy Rate2* 35.9% 56.0%
Inpatient Days to FTE Personnel2) 49.6 days 40.2 days
Average Beds3)'4) 52 beds 235.6 beds
Populations per Hospital Bed3)- 4> 339.7 persons 487.5 persons
Populations per Physician3)' 4> 778.8 persons 444.2 persons
Numbers of Types of Medical Services4) 17.9 types 34.2 types
Source: 1) Colorado health and Hospital Association, 1998, REFERENCEGUIDE to Colorado Hospital and Health
are Organization Financial & Utilization Data 1997.
2) American Medical Association, 1999, Physician Characteristics and Distribution in the US 1999 edition.
3) U.S Bureau of Census, 1998, Population Estimates Program, Washington, DC.
4) American Hospital Association, 1998, AHA Guide to the health Care Field, 1997-1998 edition, Chicago, IL,
healthcare Info Source, Inc..


the diagnostic-related group (DRG)1 based prospective payment system
(PPS)2, which offers a per case payment. The PPS was intended to promote
efficient hospital operation. It encourages the use of outpatient health care
services instead of inpatient health care services, fewer ancillary services, and
reduced length of stays. Colorado rural hospitals have experienced more
financial difficulties under the PPS reimbursement method compared with
Colorado urban hospitals.
Table 2-2 summarizes major characteristics of rural and urban
hospital including average daily charge per day, occupancy rate, inpatient
days to FTE personnel, average beds, populations per hospital bed,
population per physician, and numbers of types of medical services. The
ratio of inpatient days to full time equivalent (FTE) personnel in Colorado
rural hospitals is 49.6 days per staff person, compared with 40.2 days per FTE
1 A Yale university-derived system of classification for 383 inpatient hospital services based on principal diagnosis,
secondary diagnosis, surgical procedures, age, sex and presence of complications; this system is used as a financing
mechanism to reimburse hospital and selected other health care providers for services rendered; used to describe
patient mix in hospital and to determine hospital reimbursement policy.
2 For Medicare services as established by Tile VI of the Social Security Amendments of 1983, developed and
implemented by Health Care Financing Administration (HCFA) to pay health care facilities for Medicare patients;
this system replaced the retrospective cost-based method that was begun in 1968; the primary prevention against
premature discharge of patients is the presence of sound quality assurance program.
19


in Colorado urban hospitals (AMA, 1999). This means that Colorado rural
health care personnel bear more of a burden (9.4 inpatient days) than their
urban counterparts (AMA, 1999). In other words, health care personnel in
rural Colorado hospitals have a greater work load as compared with health
care personnel in urban Colorado. Thus, Colorado rural hospitals can
become less attractive work environments for health care practitioners.
The average number of beds in Colorado rural hospitals is 52, while
urban facilities have, on average, 235.9 beds (AHA, 1998). Colorado rural
hospitals have only 18.1 percent of the total beds in the state. This indicates
that the distribution of hospital beds in Colorado closely parallels the
Colorado population distribution. However, there is great diversity among
Colorado rural counties in terms of their demographic, health care and
economic characteristics.
The population per hospital bed in Colorado rural areas is 339.7
persons compared with 487.5 people per bed in Colorado urban hospitals
20


(USBC, 1998; AHA, 1998). This illustrates that Colorado rural hospitals are
in less demand compared with their Colorado urban counterparts.
The ratio of population per physician in Colorado rural areas in 778.8 persons,
while that ratio in Colorado urban area is 444.2 persons (USBC, 1998; AHA,
1998). This figure indicates that Colorado urban physicians have a heavy
work burden compared to Colorado rural physicians. This factor adds to
the difficulty of attracting health care practitioners, especially physicians, to
stay in Colorado rural areas. Colorado rural hospitals provide an average of
17.9 kinds of health care services, while the Colorado urban hospitals offer
34.2 types of health care services to their residents (AHA, 1998). For
example, Wesibrod Memorial County Hospital (Kiowa county) provides only
one specialized health care service (swing bed) to its rural residents, while
Portercare Hospital (Denver County) offers 56 types of specialized health care
services to urban patients. Portercare hospital provides health care services
ranging from aeromedical transportation, organ transplant, neurological
services, and nuclear medicine to sports medicine (Billian's Health DATA
21


group, 1998). In general, rural hospitals in Colorado provide basic health
care services such as ambulatory surgery services, auxiliary services, birthing
rooms, blood banks, emergency services, physical therapy services, and
swing beds. On the other hand, Colorado urban hospitals address more
complicated health care issues such as AIDS, laser surgery, intensive care
units focused on specialty care areas, in addition to providing the basic health
care services that the rural hospitals provide (Billian's Health DATA Group,
1998). This study recognizes that Colorado rural hospitals are smaller (one-
fifth the size of Colorado urban hospitals), less technologically advanced, and
less specialized. They are predominantly county or local community owned,
and are characterized as sources of primary health care services, not specialty
treatments. These data illustrate the obvious obstacles facing patients,
practitioners, and administrators of rural hospitals, when competing for
patients against their urban counterparts.
22


Figure 2-2: Bypassing and Occupancy Rates in Colorado
(1993-2000)
K)
70.0%
65.0%
60.0%
55.0%
50.0%
45.0%
40.0%
35.0%
30.0%
sources: Inpatient Discharge Data: 1993-2000, Colorado Health and Hospital Association.


The Issues of Colorado Rural Hospitals
Many changes in the regulatory, patient care, and administrative
environments of rural hospitals have left these institutions in desperate
financial situations despite their efforts to adapt to these changing situations.
Environmental changes have occurred in the following areas: new
technologies and methods of practicing medicine; regulation of health care
professionals, hospitals, and other health care businesses; reimbursement
strategies; local and national economics; competition with other hospitals;
availability of health care practitioners; mechanisms of capital financing; and
the demographics of rural areas (Ermann, 1990). All these factors have
influenced patients to bypass their local rural hospitals to receive health care
services at larger rural hospitals or urban health facilities.
The increase in bypassing activity by Colorado rural patients is key to
the declining financial status of rural hospitals and is a consequence of
decreased admissions as well as shorter lengths of stays. Over the past
decades, a steadily growing trend of patients' bypassing health care services
24


provided by local hospitals has appeared in the rural Colorado areas.
Despite proximity, many Colorado rural patients often bypass local rural
health care services and travel to hospitals outside their residential areas.
For example, Colorado Inpatient Discharge Data (CIDD) from 1993 to 2000
show that about 43 percent of all rural patients in Colorado bypassed their
local rural health care facilities (CHHA, 1993-2000). The effect of this
phenomenon is that occupancy rates decline and these rural facilities become
less competitive, thereby ultimately causing rural hospitals to face financial
crisis and closure. This result adversely affects the communities that they
were designed to serve.
Declining Occupancy Rate of Rural
Hospitals and Increasing Bypassing
of Local Rural Hospitals
Generally, rural residents lack confidence in rural hospitals and
medical practitioners, according to an interview with the director of the
Colorado Rural Health Center, which understands Colorado's rural patients'
25


opinions, beliefs, and attitudes regarding rural hospitals. Colorado rural
patients want health care services in facilities equipped with modern
diagnostic equipment, which is not always available in their local hospitals.
Rural residents in Colorado not only lack awareness about the capabilities of
local hospitals but also lack the vision that rural community members can do
anything to affect change in their rural community's health care system
(Denton, 1999).
The occupancy rate of rural hospitals in Colorado has been declining
precipitously and subsequently; so has their ability to compete. In large part,
this trend is due to the increasing numbers of rural patients choosing to
bypass their local rural hospitals, opting to seek health care services from
other regional or urban facilities. There are many reasons for bypassing
local rural hospitals. A study of practitioners in rural hospitals and public
officials of state health departments identified the principal motivations for
rural patients to bypass their local rural hospitals. First, health care services
that rural patients seek are not available at local rural hospitals. Therefore,
26


rural physicians refer their patients to facilities that offer more comprehensive
care. Second, rural patients who work in urban areas seek health care
services close to their place of employment. Third, rural patients who have
no previous experience with their local health care services continue to avoid
local rural hospitals for other health care services as well. Fourth, rural
patients have faith that better health care quality is available at bigger
hospitals with better health care technology.
It is not surprising to find that many Colorado rural patients bypass
their local rural hospitals for urban or other local hospitals where the health
care quality is perceived to be superior. As Figure 2-2 indicates, the overall
Colorado hospital occupancy rate (both urban and rural hospitals) shows a
decline from 61.3 percent in 1992 to 48.4 percent in 1998.
The declining occupancy rate is more serious in Colorado rural hospitals than
in Colorado urban hospitals. The occupancy rate in Colorado rural hospitals
has decreased by 28 percent (from 46.5 percent to 34.9 percent) from 1993 to
1998, while that of urban hospitals in Colorado has declined only 20 percent
27


(63.8 percent to 50.7 percent) during the same time period (CHHA, 1997,1998,
1999). Moreover, the bypassing rate of Colorado rural patients has been
increasing from 40.9 percent in 1993 to 45.8 percent in 2000. This bypassing
rate is expected to continually increase in coming years. The increasing
bypassing of local rural hospital causes these hospitals to lose profit margins
and to increase reliance on tax subsidies. These factors lead to the high
incidence of failing hospitals in Colorado's rural communities.
Decreasing Profit Margins and Increasing
Tax Subsidies to Rural Hospitals
Increasing bypassing of local rural hospitals and lowering occupancy
rates in rural hospitals have created higher per-patient costs, lower profits,
and a decreased ability to adopt new health care technology. As a result,
occupancy rates have decreased even further, and unfortunately, these
tendencies may be difficult to be reverse (Getzen, 1997). Therefore, many
rural hospitals do not have a sufficient patient base to financially support
themselves and are losing their competitive edge.
28


Figure 2-3: The Percent of Patient Service Margin (1992-1998)
Rural Hospital
Urban Hospital
source: REFERENCE GUIDE to Colorado Hospital Financial & Utilization Data 1997,1998, 1999, Colorado Health and Hospital Association.


o
$9,000
$8,500
$8,000
$7,500
$7,000
$6,500
$6,000
Figure 2-4: Colorado Rural Hospital Subsidies
(1992-1998)
Yr92 Yr93 Yr94 Yr95 Yr96 Yr97 Yr98
Rural Hospital Subsidy/Bed
source: REFERENCE GUIDE to Colorado Hospital Financial & Utilization Data 1997, 1998, 1999,
Colorado health and Hospital Association.


Rural hospitals have also increased the government's financial burden
because many of these hospitals are either public or non-profit and rely
heavily on public funding to provide their health care services. This rising
trend of bypassing local rural hospitals in Colorado seems to threaten the
financial viability of local rural hospitals and may eventually put more tax
burden on the government in the form of subsidies to bail out those local
rural hospitals. The percent of patient service margin of the total operating
revenue3 is one of the main indicators regarding the financial viability of
hospitals. The patient service margin reflects total patient charges collected
plus other operating revenue from hospital services for non-health care
services, such as revenue from cafeteria sales, less the expenses to the hospital
for health care services provided (CHHA, 1997).
As figure 2-3 indicates, the average percent of the patient service
margin based on the total operating revenue in Colorado was 3.74 percent for
rural hospitals, and 5.11 percent for urban hospitals. In the 1990s, this trend
3 (Patient Service Margin Amount) *- (Total Operating Revenue). Profit or loss in healthcare has
traditionally been called "margin".
31


was aggravated in many Colorado hospitals, especially for rural ones.
Hospitals with lower operating margins do not fare well as they have little
time for strategic planning and face other operational struggles such as older
equipment, minimal investments in new health care services, a weakened
ability to attract capital, and limited financial means to act on rural
community needs or market challenges. Therefore, many Colorado rural
hospitals that are public or non-profit owned which constitute 95 percent of
all rural hospitals are in need of financial subsidies from government to
maintain an adequate quality of health care services. For example, in 1998,
Colorado rural hospitals received $8,832 of governmental subsidies per bed.
As depicted in Figure 2-4, this trend had incrementally increased since 1995.
Although immediate closure of rural hospitals often can be averted by
increasing tax subsidies for public rural facilities, this alternative is not
financially viable over the long run. Rural communities have higher
proportions of the poor and unemployed, making increased taxation and
donations a less feasible alternative than in more populated and affluent
32


areas. Poor financial conditions make it difficult for a facility to gain access
to capital for improvement, maintain licensure and accreditation standards,
and attract qualified health care practitioners. Budget deficits of hospitals in
the rural areas result in impediments to maintaining higher health care
quality.
Increasing Closure of Rural Hospitals
The decreasing profit margins and increasing tax subsides to
Colorado rural hospitals have caused the cost per patient to increase and have
hindered the efforts of such rural facilities to introduce new technologies and
health care services (Merlis, 1989; Mayer, 1987). In particular, smaller and
less competitive rural hospitals appear to be less sustainable and more
susceptible to closure. National statistics indicate that within a seven-year
time period (1993-1999), 50 rural community hospitals, out of 2,249, closed
their doors. In the state of Colorado, such decline is more significant. In
comparison to the national decline of 2.2 percent, Colorado faced a more
33


drastic reduction of 12.2 percent as five (out of 41) rural hospitals stopped
providing health care services, while two urban hospitals were established
during the same period. These figures indicate that the closure of rural
hospitals in Colorado is six times the national rural hospital closure rate from
1993 through 1999. It logically follows that Colorado rural hospitals have an
increased risk of closure.
The closure of hospitals in rural areas has three effects: (1) weakening
a rural community's economy, (2) reducing access to local health care services,
and (3) accelerating health care costs. Rural hospital closure has a negative
impact on the economy of the rural communities involved as a result of loss
of employment for those dependent upon the hospital for their livelihoods,
unemployment among those who provided goods and services to those who
made their living through the rural hospital, outward migration of those left
unemployed, loss of tax base and decline in the town's attractiveness to
business and new residents.
34


The closure of rural hospitals impedes access to both doctors and
hospital health care and reduces the overall quality of health care services
that remain. Physician care is influenced since doctors are less likely to stay
in a community once the rural hospital is closed and recruiting new
physicians becomes impossible because physicians in rural areas have to
work for hospitals in order to make income near the national average. The
hospital closure rate also threatens ready access to emergency care
particularly needed by the residents in rural occupations with high related
injury rates, such as farming and mining.
The closure of rural hospitals not only has a substantial impact on the
economic viability of the rural community but also potentially increases the
health care costs by diverting less emergent patients into bigger rural
hospitals or urban hospitals. When Medicare case-mix, teaching activity,
and relative wages are held constant, the health care costs of big rural
hospitals or urban hospitals exceed those of rural hospitals.
Considering the significance of bypassing as well as the resulting
loss in competitiveness that confronts Colorado rural hospitals, this study
will examine both institutional and individual dimensions that might
contribute to the increased rate of bypassing Colorado's rural hospitals. The
35


following chapter will review literature that is significant to the study of this
topic.
36


CHAPTER 3
LITERATURE REVIEW
This chapter surveys the literature related to the analysis of rural
residents' hospital choice behavior in order to identify the factors that
influence rural patients to choose hospitals for health care services. This
study reviews the previous literature related to hospital utilization and
geographic access, and the hospital choices of rural patients.
First, this chapter explores analyses that deal with distance and the
demand for health care services. It reviews analyses of the effect of distance
and geographic access on the choice of health care services. Second, this
chapter reviews literature that deals with patients' behavior in selecting
primary care services. Third, it reviews the research that investigates
various institutional and individual factors determining patients' hospital
choice in general. Lastly, this chapter surveys selected literature concerning
rural patient's hospital choice.
37


Interdisciplinary Backgrounds and
Health Care Demand
Many health care studies have borrowed their research methods from
other academic disciplines. The principal academic areas that influence
health care researchers are geography (medical geography) and economics
(health economics). This section chronologically reviews those two areas of
study.
The Geographical Approach and Hospital Use
One of the primary public policy issues related to rural health care
services is whether rural patients have inferior health care services compared
to their urban counterparts. Many health care studies have focused on the
inquiries that are associated with distance and time and how these factors
influence rural patients' access to hospitals (Shannon et al., 1969). The
gravity model is a popular geographical method that describes and predicts
patient flows from origin point to a destination, for example, from the
patient's residence to the hospital. This model uses the principles of central
place theory. The concept of central place theory in the case of health care
states that health care facilities are spatially ordered in hierarchical fashion,
including larger health care centers that offer various health care services on a
38


regional basis, local hospitals that provide basic medical services to smaller
local areas and, at the lowest level of the hierarchy, an ubiquitous distribution
of doctors and health care professionals (Shannon et al., 1969).
While some factors may work in opposition to the spatial organization
of health care services, such as geographical constraints (rivers or mountains)
and variation in health care needs within the population, the principle of the
gravity model provides a strong basis for the study of patient flows in health
care market studies. Further, the gravity model is shown to incorporate the
concepts of the efficiency of service provision and equity of access to services
from location theory (Leonardi, 1981). Other critical bases for the
development of the gravity model are entropy maximization (Wilson, 1967)
and cost efficiency (Smith, 1976).
Jehlik et al. (1952) investigated how the distance between the nearest
hospital and a Missouri farm family influences choice of hospitals. They
found that distance was not associated with the length of a patient's stay.
They conjectured that there would be no relation between length of stay and
hospital access, because there are not many alternatives when rural patients
choose a hospital for health care services.
39


Altman (1954) examined the travel patterns of populations in 27
western Pennsylvania counties. He found that distance and isolation did not
reduce the number of doctor visits to general practitioners. However, in the
case of medical specialists, patients who lived in a county, that required more
travel time in order to visit specialists visited specialists less often.
While studies conducted in the 1950s were interested in the health
implications of distances to hospital facilities, these studies did not focus on
how patients chose a specific hospital or the implications of that choice on
rural health care policy. Since the 1960s, the role of time and distance as an
explanation of the consistent spatial pattern of health care services has
become a more popular research topic among health care researchers and
geographers.
Geography provides a pattern of analysis to explore how patients
choose a hospital for health care services. Health care research maintains the
idea that hospitals were ordered spatially according to a hierarchy of size and
complexity from the central place theory suggested by Christaller (1966).
These studies provide an important implication for the decision of where to
locate hospitals. The bigger and most complex treatment hospitals should
40


be located in a central area, such as a medical complex in the urban core, with
smaller community hospitals dispersed further from the central area.
Related to the hierarchical point of view of spatial organization of
hospitals is the areal unit problem. Shannon et al. (1969) argued the
difficulties of using aggregate and areal analysis to investigate the effect of
distance on the behavior of hospital choice among patients. The researchers
observed the aggregate measures from a zone or other areal unit that assumes
the factors within the unit are the same. They examined the role of distance,
race, and other demographic variables on hospital choice behavior in
Cleveland, Ohio. To examine the effect of distance on hospital use, they
relied upon the gravity model. The implication of this gravity, or spatial
interaction model, is that the migration between two centers would be
inversely related to the distance separating them. They found that
significant differences in hospital use within the Cleveland metropolitan area
resulted from ethnic, religious, and income differences among residents. For
example, although an area inventory of hospital beds per capita showed that
the residents of Cleveland had easy access to hospital services compared with
rural residents, inner-city African-Americans in Cleveland could not easily
access the same level of hospital services that their Caucasian neighbors did.
41


While Shannon et al. (1971) focused upon aggregate patient patterns from one
area of a city to another, the research unit of analysis was at the individual-
level analysis.
Roughman and his colleagues (1979) found positive correlations
between hospital characteristics, such as hospital size, specialization, health
care quality, strength of demand and shorter distances traveled for health
care services. Mayer (1983) adopted the gravity concept to calculate the
hierarchical structure of a hospital's health care services based on the central
place theory. However, he found that the location of a hospital was not
consistent with central place patterns, resulting in differences in spatial access
to health care services across population groups.
McLafferty (1988) analyzed geographic patterns of hospital utilization
before and after the closure of Sydenham Hospital in New York City. She
wished to determine whether the patterns of patients' hospital use after a
hospital closure could be predicted using standard spatial interaction
modeling procedures. Using three variables distance, hospital size, and
ownership type the model accurately described the pattern of patients'
hospital use in each year following the hospital's closure. The geographic
42


distribution was analyzed along with the implications for spatial modeling
efforts.
The gravity model has been applied in other countries where the
health care system is more centrally planned and controlled than in the
United States. A critical assumption of this study was that the need for
health care services is limited by the supply of health care services, implying
that all available health care service resources were utilized. In this
approach, total hospital demand could not exceed capacity or patient access
to hospitals and must be set proportionally to the population of the origin
zones. Wilson and his colleague (1990) applied the government resource-
allocation formula to particular specifications of the destination-constrained
model and found that differences in sensitivity to distance by service category
exist.
The destination-constrained approach is not proper in the United
States' health care market, where there is freedom with regard to hospital
choice. The role of doctors in influencing patients' hospital choice has been
demonstrated (Burns et al. 1992), although patient convenience, measured by
distance or travel time, is shown to be the most reliable predictor of hospital
patronage 0avalgi et al., 1991).
43


Lowe and his colleague (1996) formulated a model to predict the
hospital patient flow system. They found that universal health coverage
dramatically improved access to health care services for the lower income
patient group. The study also reduced that health care reform had
differential effects on different types of hospitals. Medical school hospitals
are expected to lose patients after implementation of health care reform, with
important implications for graduate medical education. Hospitals at risk for
closure were forecast to receive increased patient flows under this reform.
Hospital closures were shown to have negative effects on access to services in
poor neighborhoods but relatively little effect on health care access in the
system as a whole.
Congdon (2001) examined the impact of patient referral flows on
reconfigurations of emergency services in some hospitals. Such system
changes involved construction of new hospital sites, or an expanded number
of beds at the same hospitals. In the context of patient hospitalization, this
study facilitates the prediction of patient flow following hospital
reconfiguration.
In summary, previous studies using the geographical approach
(gravity model) to patient flow confirmed the importance of distance as a
44


distinct measure in model formulation, which suggests that nonspatial factors,
such as hospital characteristics, also be included in the model. These
researchers also found differential sensitivity to distance according to types of
hospital health care services and showed that origin-constrained versions of
the model would be appropriate for forecasting applications in United States'
health care markets, although the problems of unit measurement in
geographic analysis were also recognized as occurring with respect to such
measures as beds per capita in a county.
The Economic Approach and Hospital Use
If the geographical approach (gravity model) is the economic model of
spatial supply, the neoclassical economic model of spatial demand represents
the time-cost, or travel-cost, model and its variants.
The time-cost model for health care services originates from the
household production model developed by Becker (1952). This method of
research sought to introduce the cost of time systematically into decisions
about non-work activities. Lancaster (1966) developed abstract notions of
consumption that lent themselves to economic explanations of innovations in
products and services. Quandt et al. (1966) developed an abstract product
45


approach for the analysis of new transportation modes and changes in the
configuration of retail establishments. Some studies have applied these
insights concerning the role of time in non-work activities, in combination
with the consumers' demand for product characteristics in health care
services.
Owing to the influence that economists employed to try to explain the
increasing importance of nonmonetary factors as determinants of health care
service demand in the late 1960s and early 1970s, Acton (1975) developed his
time-price analysis of the demand for health care services according to the .
household production model of consumer behavior. He explored how
nonmonetary factors influence New York City residents in their health care
decisions. His model emphasizes the belief that distance and other variables
associated with time-costs serve to guide consumer choice when out-of-
pocket prices decline. He found that as the monetary cost per unit of health
care service drops to zero, the demand for health care services becomes more
responsive to increases in the time-cost of the health care services. He also
recognized that while an increase in non-earned income leads to an increase
in health care demand, an increase in the patient's wage rate may or may not
lead to a demand increase.
46


A number of empirical studies involving the time-price model were
researched with regard to health economics. Phelps et al. (1974) investigated
the change in the location of student health care services in conjunction with
demand for health care services. They found that the elasticity of the
demand for health care service visits with respect to time showed a change of
-.28 to 0.51. They also assign the time-cost of visits to a Palo Alto health
clinic and found that the time-costs of physician visits were high. They
concluded that time acts as a strong rationing device in the absence of price
fluctuations.
Salkever (1975) examined the quantitative impacts of financial and
physical access barriers on the decision to enter the health care system for
treatment. He used the probit regression model to analyze the demand for
entry into the health care system. In the probit model, he specifies time as
the time traveled to the individual's usual source of ambulatory care and he
measures the patients' wage as the estimated wage for employed adults
based upon their occupation, age, race, education level, and gender. He
found little support for the notion that physical and financial access barriers
strongly discourage entry into the health care system.
47


Using the binary probit model, Christianson (1976) estimated the
variable, willingness-to-pay (WTP), for outpatient health care services at the
Marshfield clinic in rural Wisconsin. He found that individuals with high
travel costs to get to the clinic were less likely to choose the clinic as their
primary source of health care. He reported that as travel costs increase from
$1 to $11, the probability that the clinic will be selected as the principal source
of health care decreases from .673 to .002.
Holtmann et al. (1976) analyzed the effect of the demand for dental
services on Pennsylvania and New York State populations. They employed
the household production model to estimate the demand for dental visits
over previous years. They used waiting time, travel time, income, price,
education level, and family size as variables. They calculated the price
elasticity of dental care to be estimated in the range of -.032 to -.187 and
found that the elasticity of waiting time is larger than the price elasticity.
Travel time is not statistically significant in their model. One problem with
the measurement must be noted: the study overlooked that many dental
patients travel to the dentist from their work place, while the researchers only
measured the distance from the patient's home.
48


Colle et al. (1978) investigated the demand for pediatric health care
services through a nationwide household survey data set. They explored
the factors that determined the use of pediatric care services as measured by
a) contacting a physician in the past year; b) having preventative physical
examinations in the past year; c) the number of physician visits by children;
and d) the average quality of the visits. They found that the time-cost
coefficient is positive in relation to the demand curve for visits. Their
measurement of time-cost is statistically insignificant in their regression
models.
Miners (1981) examined how the household structure and interaction
affects of household members' influence on the choice of primary care
services in rural North Carolina. He found that the elasticity for visits, with
respect to time traveled, illustrate that few of the coefficients of travel time are
statistically significant in this regression model.
Coffey (1983) examined how time cost affects female patients' choices
for health care services in the Dallas area. She estimated a three-equation
system that modeled the choice of health care provider, entry demand, and
visit demand. She found that failure to control for the type of health care
provider and the opportunity cost of time will lead to understated estimates
49


of the time price elasticity. The result indicated a time price elasticity of
entry into the health care system of -.09. This figure means that a ten
percent increase in the opportunity cost of time should show a 9 percent
decrease in the probability of seeking health care services, if all other
independent variables remain constant.
Samuels and his colleagues (1991) examined the effect of hospital
closures on travel time to hospitals and found that, contrary to popular belief,
even in rural communities hospital closures have had minimal effect on travel
times. It seems that closures are most likely to occur in markets with
declining populations, declining demand for inpatient days, and an
increasing supply of open beds. The implication is that hospitals close,
primarily, in reaction to overall decreases in the demand for health care
services in a given health care market, which results in a surplus of beds.
The Primary Care Choice
Much of the literature concerning rural health care services focuses on
the issue of rural patients' access to primary care. Braden et al. (1994)
analyzed the 1987 National Medical Expenditure Survey (NMES) that
compares health status, access to health care, and use of health care services
50


for rural and urban patients. They found that rural residents showed worse
health status than urban residents in almost all cases. Twenty-eight percent
of rural adults perceived their health status as fair/poor, while 72 percent
perceived their status as excellent/good. On the other hand, 79.4 percent of
urban residents responded that their health status was excellent/good.
Similar differences held for the prevalence of chronic conditions, mobility and
physical activity limitations, medical treatment for life-threatening or chronic
conditions, mobility and self-care limitations.
Braden et al. (1994) also reported that 76.2 percent of rural residents
sought ambulatory medical services in 1987, while 77.0 percent of
metropolitan area residents needed those services. However, rural
populations demonstrated a tendency to obtain their health care services
from a general practitioner with more frequency than did urban residents.
Their analysis indicated that rural residents were more likely to use a routine
source of health care than urban residents. The striking difference in access
to health care between rural and urban residents, in their analysis, appears in
the case of travel times to seek primary care services. They reported that
11.2 percent of rural residents traveled over 30 minutes to visit a primary
physician, while 6.8 percent of metropolitan residents made such a lengthy
51


trip. In general, the results from the 1987 NMES indicated that rural
residents have poorer health status relative to urban residents, and their
access to health care services is poorer overall than that of their urban
counterparts.
The researchers focusing on the geographic access to physician
services employed the standard location theory presented by Christaller
(1966) and Greenhut et al. (1975). The economic theory of location choices of
physician services provides a behavioral explanation of the distribution of
physicians across rural and urban areas.
Newhouse (1990) established three implications that can be compared
with data on the geographic distribution of physicians across urban and rural
communities. First, there is a critical population size required to acquire a
given number of physicians. Second, in general, the towns with the highest
rates of growth in the number of physicians will be those towns whose size is
just below the critical threshold. Third, medical specialists will tend to
locate in the larger towns and cities, because they can maximize the demand
for their unique health care services. Generalists will tend to locate in
smaller towns, because in the larger towns and cities specialists will compete
against them for some of the health care services they offer. Moreover, the
52


health care market areas of specialists will cover a larger area than the health
care markets of generalists.
The empirical research supports the view that Newhouse's proposition
in regard to physicians' location choice. Schwartz et al. (1980) found that the
location pattern of specialists became more diffuse geographically, as the
supply of board-certified physicians increased from 1960 to 1977.
Furthermore, the theory's prediction regarding rates of change was
confirmed by the data. As the supply of specialists increases, smaller
communities experienced a greater increase in specialists per capita than did
larger communities.
With respect to access to physician services in rural areas, Newhouse
et al. (1982a) found that there are no towns with more than 2,500 residents
that do not have easy geographic access to a physician. William et al. (1983)
reported the driving distances of rural patients from sixteen Northeast, North
Central High Plains, and Southeast states to various types of physicians.
They found that 80 percent of the rural residents were within a ten-mile
radius of a physician, while 98 percent of the rural residents resided within a
25-mile radius. However, they recognized that 74 percent of the sample
communities with 20,000-30,000 residents from 23 states, had access to
53


specialists, including an internist, a surgeon, a pediatrician, and an
obstetrician/ gynecologist. Smaller communities tended to have much lower
levels of access for this basic set of health care services. For example, only 4
percent of communities with a population of 5,000-10,000 had all of these
health care services, while 22 percent of communities with a population of
10,000-20,000 had all of these health care services. Although the supply of
physicians results from the competitive force of the health care market,
persons who live in smaller towns have different access to physicians than
those who live in larger towns. This is especially true with regard to health
care specialists. Furthermore, William et al. (1983) pointed out that the
health care market alone would not solve the problems of low income and
under-insured rural residents and the special needs of residents in very
sparsely settled areas.
Newhouse et al. (1982b) argued that the economic analysis of primary
care access has implications for public policy decisions that aim to decrease
the differences in the number of physicians per capita across rural and urban
communities. First, in the absence of externalities, a public policy that has to
place a physician in a town that she/he would not have been located in
otherwise exacts a loss of efficiency. Because the physician is less busy in
I
54


the forcedly placed location than he/ she would be in a location that he/ she
chooses, the physician would lose efficiency. Another public policy
implication is that the placement of physicians, in general, in isolated
communities will be not effective because the new physician will displace
established physicians. The third public policy implication is the difficulty
of trying to assist the isolated populations in getting health care services that
concentrate on the characteristics of the particular populations. Current
health care options make it difficult for residents to obtain health care services,
taking into account the socio-economic, ethical, and cultural factors that
combine to engender differences in health care across individuals.
The Research on Hospital Choice
McFadden's discrete choice model has examined the way in which
transportation decisions influence the way health economists have modeled
the choice of health care facilities by health care consumers (patients).
McFadden employed the conditional logistic model to explore the probability
that urban residents would choose a particular mode of transportation in the
San Francisco Bay area (McFadden, 1974). Furthermore, he theorized that
the correct economic framework to analyze travel decisions was the
55


household production model developed by Becker. Health care researchers
and health economists who examined hospital choice heavily depended upon
McFadden's explanation that demonstrates links between the utility
maximization theory and the conditional logistic model. Table 3-1 below
lists a number of these studies, their data sources, and a brief description of
the study.
The logistic regression models of hospital choice originated in the
early 1980s, and research into hospital choice has continued to the present.
The early models tended to be aggregate choice models that were evaluated
using patient flow data. Examples of such studies are Folland et al (1983),
McGuirk et al (1984), Cohen et al. (1985), Lee et al. (1985), Erickson et al.
(1985), Garnick et al. (1989), Luft et al. (1990), and Luft et al. (1991).
Although the aggregate choice model has its applications, it has been
criticized because the stratification of data did not consider the possibility of
individual characteristics, other than illness measures, that may have
influenced the hospital choice. Nevertheless, the aggregate choice model
was popular among health care researchers and health economists because
the sources of aggregate patient data needed to run these models existed.
56


Table 3-1 Selected Hospital Choice Literature
Authors Data Variables Method
Garnick et al. (1989) California Cardiac Catheterization Patients Dep. Variable: the probability of patient choice among hospitals. Ind. Variables: distance, charge, mortality, hospital type, staff/bed. Conditional Logistic Regression Model of hospital choice
Luft et al. (1990) California patients from Three Clusters: (San Diego, San Francisco, and Modesto) Dep. Variable: the likelihood of going to a particular hospital. Ind. Variables: quality, charges, ownership, distance. Conditional Logistic Regression Model of hospital choice for 7 surgical procedures and 5 medical diagnoses estimated effects of hospital quality measures and charges
Phibbs et al (1993) 1985 Data All Deliveries in San Francisco Bay Area (n=61,436) Dep. Variable: the number of patients from a given zip code who chose a given hospital Ind. Variables: quality, charges, ownership, Medi-Cal contracting, socioeconomic status, distance. Conditional Logistic Regression Model of hospital choice for delivery by high-risk and low- risk women
Chernew et al. (1998) 1991 Hospital Discharge Data on undergoing CRAB who resided in 3 California markets (LA, S. & N. California) (n=20,567) Dep. Variable: the probability that patients would receive care at any given hospital. Ind. Variables: quality, distance, bed size, ownership, medical school affiliation, Conditional Logistic Regression Model in the choice for coronary artery bypass graft surgery
Bronstein et al (1991) 1983,1988 Rural Alabama Obstetrics Patients Dep. Variable: the probability of bypassing the nearest hospital Ind. Variables: distance, bed size, ownership, race, age, income, Two-level Binary Logistic Regression Model of bypassing rural hospitals for obstetrics care


Table 3-1 Selected Hospital Choice Literature
Authors Data Variables Method
Gamick et al. (1989) California Cardiac Catheterization Patients Dep. Variable: the probability of patient choice among hospitals. Ind. Variables: distance, charge, mortality, hospital type, staff/bed. Conditional Logistic Regression Model of hospital choice
Luft et al. (1990) California patients from Three Clusters: (San Diego, San Francisco, and Modesto) Dep. Variable: the likelihood of going to a particular hospital. Ind. Variables: quality, charges, ownership, distance. Conditional Logistic Regression Model of hospital choice for 7 surgical procedures and 5 medical diagnoses estimated effects of hospital quality measures and charges
Phibbs et al (1993) 1985 Data All Deliveries in San Francisco Bay Area (n=61,436) Dep. Variable: the number of patients from a given zip code who chose a given hospital Ind. Variables: quality, charges, ownership, Medi-Cal contracting, socioeconomic status, distance. Conditional Logistic Regression Model of hospital choice for delivery by high-risk and low- risk women
Chernew et al. (1998) 1991 Hospital Discharge Data on undergoing CRAB who resided in 3 California markets (LA, S. & N. California) (n=20,567) Dep. Variable: the probability that patients would receive care at any given hospital. Ind. Variables: quality, distance, bed size, ownership, medical school affiliation, Conditional Logistic Regression Model in the choice for coronary artery bypass graft surgery
Bronstein et al (1991) 1983,1988 Rural Alabama Obstetrics Patients Dep. Variable: the probability of bypassing the nearest hospital Ind. Variables: distance, bed size, ownership, race, age, income, Two-level Binary Logistic Regression Model of bypassing rural hospitals for obstetrics care


The first application of the aggregate choice model to hospital choice
was Folland's work (1983). He investigated patient travel behavior in South
Dakota to understand the patients' inter-city movements.
He found that distance between the patient and the hospital is the
critical factor in rationing health care market shares. Also, he discovered
that the number of hospital beds and the presence of psychiatric and
intensive care units (ICU) were positively associated with a rise in a hospital's
market share.
McGurik et al. (1984) examined the patients' hospital choice in
Allegheny County, which includes the city of Pittsburgh, Pennsylvania.
They analyzed the data on hospitalizations in 4 categories: a) medical-
surgical; b) obstetrics-gynecological; c) pediatrics; and d) psychiatric services.
The researchers found that distance and time elasticity have strong effects on
the probability that a particular hospital will be chosen. Because these
elasticities resulted from an aggregate choice model, they were interpreted as
the expected percentage in patients choosing a hospital owing to a 1 percent
increase in distance or time from their residence. The selection elasticity for
a 200-bed acute-care hospital that provides inpatient diagnostic and short-
term treatment of patients, ranged from -1.223, at 2 miles from the patient's
59


residence, to -6.048, at 15 miles from the patient's residential community.
Psychiatric diagnoses were less elastic and ranged from -0.548, at 2 miles for
the same hospital, to -2.068, at 15 miles. For a large general hospital with
500 beds, the elasticity was slightly smaller than for the acute hospital with
200 beds at the same distance. For instance, the elasticity with a medical-
surgical diagnosis ranged from -0.959, at 2 miles, to -4.460, at 15 miles, for
large acute hospitals. The researchers included travel times instead of
distances in a regression model specification, and the elasticity of these
measures were similar to the distance elasticity.
Cohen et al. (1985) explored the hospital choice behavior of Rhode
Island populations and the degree that significant factors such as distance,
illness, and health care service provided by the hospital had on the choice.
They employed the multinomial logistic regression model, including all 14
Rhode Island acute hospitals. They found that travel time deters patients
from seeking health care services at more distant hospitals. The effect of an
increase in travel time by one minute was to reduce the probability of
hospitalization from 1.93 percent to 1.71 percent.
They also tried to measure the effect of physicians in health care
location choice. To figure out this effect, they specify the number of
60


physicians within 30 minutes of the patient's residence and with admission
privileges at the specific hospital under consideration. They found that the
physician variable is significant in choosing hospitals. However, as Drave et
al. (2993) commented, concerning Burns et al. (1992)'s treatment of the
physician's location, the location of a physician may be endogenous.
Therefore, the distance between a physician's office and a patient's residence
may be a joint choice, if the physician works out of, or close to the hospital in
question.
Garnick et al. (1989) exhibited the value of the maximum estimation
(MLE) technique over the least squares estimation techniques. The
technique used here involves discarding the zero flow observations from the
model. Using California health data regarding the choice of hospitals for
cardiac catheterization, they estimated a qualitative choice model (a logistic
model) for specific types of patients and for the entire relevant market area.
With the existence of many zero values for the dependent variable
observation, they figured that the linear estimation technique became
sensitive to the proportion of zeros, and biased estimates resulted.
Therefore, they suggest that the preferred estimation method is to apply the
61


maximum likelihood estimation in the hospital choice model to the logistic
regression model.
Luft et al. (1990) and Luft et al. (1991) investigated whether hospital
quality and charges play roles in choosing a hospital. As Granick et al.
(1989) did, they estimated the hospital discharges on seven surgical
procedures and five medical diagnoses from three hospital clusters in
California, by employing the logistic regression model. Health care quality
was measured through the inclusion of a severity-adjusted mortality rate and
a severity-adjusted complication rate. The hospital clusters were defined
within the San Francisco area, which covers an area that takes three hours to
drive from north to south. They found that health care quality is a
significant factor in choosing a hospital. With respect to distance between a
hospital and a patient's residence, and elasticity of the hospital choice, they
reported that a 10 percent reduction in distance increases the likelihood of
admissions from 12.62 percent to 13.90 percent.
Recent hospital choice research has depended upon the conditional
logistic specification model and has utilized individual patient level data.
Adams et al. (1991), Burns et al. (1992), Phibbs et al. (1993) and Chernew et al.
62


(1996) are examples of researchers who employed the conditional logistic
regression model to explore hospital choice behavior.
Using patient discharge data released from the Arizona Department of
Health Services (DPS) in 1989, Burns et al. (1992) examined the effects of
physician, patient, and hospital characteristics on hospital choice, by using
the conditional logistic regression model. They focused on whether the
inclusion of physician variables improved the fit of these choice models and
influenced the estimates of patient and hospital measures. They found that
physician characteristics were a strong determinant of hospital choice that
accounted for much of the explained variation. Also, health care quality of
hospitals and patient charges exerted critical influence over hospital choice.
Phibbs et al. (1993) examined 61,436 female patients in the San
Francisco Bay area to explore the hospital choice for maternal deliveries in
1985. They wanted to determine whether or not the choice of hospital was
sensitive to the patient's insurance source and the nature of risk involved in
the patient's medical condition (defined by the ICD-9 code). They found
that the risks involved with pregnancy conditions and payment source are
statistically significant to the probability that patients will choose a particular
hospital. Also, they found that the distance between the hospital and the
63


patient's residence appears highly associated with hospital choice among all
the logistic regression models. However, they did not explore the effects of
probability elasticity with regard to the insurance types, risk status, and
distance, thereby making it difficult to measure the size of the estimated
effects.
Using discharge abstracts complied by the California Office of
Statewide Health Planning and Development, Chernew et al. (1996)
employed a conditional choice regression model to estimate the choice of
hospital for open-heart surgery. They investigated whether a health
maintenance organization (HMO) member's choice of hospital is influential
by their insurance coverage. They found that distance is a statistically
significant variable with regard to the probability of choosing a particular
hospital. In other words, the distance between a hospital and a patient's
residence acted as a stronger constraint to non-HMO members in their choice
of hospital, than to HMO members. They reported that non-HMO members
were approximately 60 percent less likely to travel 15 miles if an identical
hospital were available 5 miles away, while HMO patients were 50 percent
less likely to travel to the further hospital under the same conditions.
64


One of the features of the hospital choice literature that is reviewed
here is the emphasis on the value of distance or travel-time rather than the
time-price notion suggested by Acton and other researchers in the 1970s.
Hospital Choice by Rural Patients
This study focuses on hospital choice by rural patients. It examines
how institutional factors such as hospital ownership type, the number of beds,
the number of health care services, and insurance types as well as individual
factors such as the patient's age, severity of illness, and distance between the
hospital and the patient's residence influence a patient in choosing a hospital
for health care services. Few empirical research studies address this
question.
Hogan (1988) explored the hospital choice behavior of rural New York
state residents during 1983 in order to estimate whether rural patients treated
in urban hospitals and in rural referral centers had more complex illnesses
than patients treated in other rural hospitals. He used hospital discharge
data available from the State Department of Health. He found that rural
patients hospitalized in urban hospitals or rural referral centers generally had
more serious illnesses. Moreover, he recognized that 71 percent of rural
65


patients were hospitalized within their home county. Of the 29 percent that
crossed the county border for hospitalization, approximately two-thirds of
the patients traveled to an urban hospital. Older patients, especially those
over 75 years of age, were less likely to cross a county line for hospitalization.
Hart et al. (1989) surveyed rural residents from Alaska, Washington,
Idaho, and Montana, in order to explore the relationship between socio-
economic characteristics and rural hospital utilization. They found that
rural patients who had been hospitalized at other, large rural hospitals or
urban hospitals were more likely to be wealthier than rural patients who had
been hospitalized at local rural hospitals. They reported that patients with
higher household incomes are more likely than their less affluent
counterparts to leave their local community for hospital care.
Bronstein et al. (1990; 1991) examined the behavior of rural Alabama
female patients in choosing a hospital for obstetric care services in 1983 and
1988. From research conducted in 1990, they found that in 1988,50 percent
of pregnant rural female patients bypassed their local rural hospital. These
patients were motivated by the tendency of rural hospitals to drop obstetric
services, and the trend of rural physicians to discontinue obstetric services.
The dependent variable in their research was the distance that a patient
66


traveled to obtain health care services. The independent variables included
the distances to neighboring hospitals, health care services at the neighboring
hospitals, public ownership of the hospitals, per capita income of the county,
the percent of women who receive AFDC payments, age, and ethnic
background. They found that rural hospitals do not exist in isolated markets,
and that the closure of some rural hospitals may not influence a large number
of female patients and their choice for obstetrical care. They also recognized
that the closure of an obstetrics unit at the local rural hospital would lead to
travel distance increases of 0.7 of a mile and 0.4 of a mile for blacks and
whites, respectively. Income also played an important role in choosing a
hospital. A one percent increase in the per capita income of the county
produced a 4 percent increase in distance traveled to get to a hospital.
In the 1991 research of Bronstein et al., the binary logistic regression
model was employed to explore the bypassing effect of rural Alabama
women seeking obstetric care in 1983 and 1988. They found that in 1983,
40.3 percent of pregnant rural women bypassed their nearest rural hospital,
and in 1988 49.9 percent of pregnant rural female patients bypassed their
nearest rural hospital. The independent variables in this study include the
distance to the nearest hospital, population of the town where the hospital
67


was located, services offered at the hospital, hospital ownership, birth volume
in the hospital in the preceding year, per capita income in the county where
the women live, patient age, race of patient, and percent of women receiving
AFDC. They recognized that women who resided farther from their nearest
hospital and closer to urban areas were more likely to bypass and go to the
urban hospital. They reported that income was also statistically significant
in bypassing the nearest rural hospital. Rural women, in counties with a
$1,000 higher average income, were 25 percent more likely to go to an urban
hospital. A one percent increase in AFDC coverage in the county of
residence is associated with an 8 percent increase in the probability of
bypassing the nearest rural hospital.
Adams et al. (1991) examined the hospital choice behavior of 12,000
western Minnesota rural patients in three overlapping areas. They
employed the conditional logistic regression model to explore the patterns of
rural patients' hospital choice. To determine how the severity of illness
influenced hospital choice, they utilized an illness severity scale and
incorporated this measurement into their conditional logistic regression
model of the patient's hospital choice. They combined individual hospitals
into groups, in order to form a set of hospital alternatives. The dependent
68


variable in their study was the type of hospital, and the independent
variables were the severity index of illness, distance, age of patient, and
hospital characteristics. Using MEDPAR (Medicare) data from three states
(Minnesota, South and North Dakota), they recognized that the severity of
illness highly influenced Medicare patients' choice of hospital. Medicare
beneficiaries with surgical diagnoses were more likely to bypass the nearest
rural hospital. Also, distance, hospital characteristics, and patient's age were
statistically significant in their conditional logistic regression model. The
probability of using a rural hospital increased with age, holding case severity
and other variables constant. They found an effect of distance on hospital
choice. If two hospitals had the same characteristics, in terms of size, the
number of health care services, and teaching attributes, the odds of the rural
Medicare beneficiary going to a hospital ten miles farther away were
approximately 50 percent lower. However, this study did not include
income as an independent variable because it was not reported in the data set.
In order to examine the effect of case-mix and demographic variables
on the hospital use of rural Medicare beneficiaries, Buczko (1992) analyzed a
1987 MEDPAR data set that contained discharge records for all short-stay
hospitalizations by beneficiaries paid under Medicare Part A. He found that
69


rural Medicare patients were more likely to have received surgical and other
complex treatment in an urban hospital, because when complex surgical
procedures were required, they were often only performed in large urban
affiliated teaching hospitals. Patients with chronic conditions and the
elderly were more likely to choose inpatient care services in rural hospitals.
This study showed the strong effect of the severity of illness on hospital
choice patterns.
In using the records of 2,171 rural patients in Illinois who received
inpatient care services for mental illness or substance abuse, Goldsteen et al.
(1994) investigated variables that influenced the decision to bypass a local
rural hospital to obtain health care services. Like Bronstein et al. (1990), they
incorporated community variables such as migration, per capita income,
unemployment rate, and urban proximity, as well as individual and hospital
characteristics, into their model. They used the binary logistic regression
model to assess the factors that caused patients to seek health care services in
hospitals that are farther away from the individual's residence than a local
rural hospital. They found that the patient's age, insurance type, per capita
income of the community area, the service orientation of the local hospital,
70


and the proximity of the patient's residence to an urban area were statistically
significant in the decision of patients to bypass their local rural hospitals.
In summary, the literature review of hospital choice by patients, or
patient outmigration, has revealed several principal findings. These studies
reach agreement on several aspects of the patterns of patients' hospital choice,
such as bypassing being more likely for patients who suffer from complex
illnesses and less likely for the elderly. Additionally, hospital characteristics
play an important role in the reasons why rural patients bypass their local
rural hospitals. As Bronstein et al. (1990) demonstrates, if local rural
hospitals provide a more complete set of health care services, the rural
patients are less likely to bypass local rural hospitals.
In conclusion, the review of previous studies regarding hospital
choice by patients reveals the opportunity to extend the analysis of rural
patients' demand for health care services in several ways.
This study develops a binary choice model and a multinomial choice
model that is appropriate for the analysis of a Colorado rural patient
discharge data set.
71


CHAPTER 4
THEORETICAL FRAMEWORK AND
RESEARCH HYPOTHESIS
This chapter develops a theoretical framework and research
hypotheses regarding the hospital choice of rural patients. The existing
literature has identified important determinants that influence a patient's
choice of hospital. In particular this study surveys the determinants that
influence patients to choose their hospital based on two dimensions: (1)
institutional factors and (2) individual factors.
When patients seek health care services, they tend to have a
preference as to where they are treated; however, several factors influence a
patient to choose one hospital over another. If a patient's insurance not
cover medical expenses at a certain hospital, the patient will consider an
alternative hospital for health care services. For example, the hospital
chosen by a Medicaid beneficiary will be limited to those hospitals that
accept Medicaid patients. The patient without any insurance will look for
the facilities that accept indigent cases.
72


The attributes of hospitals themselves also influence a patient's
behavior in choosing a hospital. For example, the ranges of health care
services provided by a hospital and the number of beds in a facility are
important factors for patients selecting a hospital. Additionally, the
medical condition of patients also influences their selection of a hospital.
Surgical patients are more likely to go to the bigger, urban facilities than
patients with simple medical diagnoses.
The theoretical framework describes how institutional dimensions,
such as payment sources and hospital characteristics, and individual
attributes, such as patient medical conditions and individual characteristics,
serve to influence rural patients to choose a hospital in a manner consistent
with the theory of consumer choice.
Some scholars have examined how and why patients select certain
hospitals to access better health care services. Many assume that patients
often bypass their local health care facilities and travel to other hospitals
become the quality of health care is perceived to be superior. The existing
literature has identified some important determinants of the tendency of
patients to bypass local hospitals as follows: (1) individual characteristics,
such as gender, age, and distance between the patient's residence and their
73


local hospital (Garnick et al., 1989; Luft et al., 1991; Adams et al., 1991,
Zwanziger et al., 1993; Phibbs et al., 1993; Meadow et al., 1998; Hawes et al.,
1984; Goldsteen et al., 1994; Hogan, 1988), (2) payment sources, such as type
of insurance (Burns et al., 1992; White et al., 1998; Chernew et al., 1998), and
(3) hospital characteristics, such as ownership type, the number of services
offered, the number of beds, and affiliation with other health institutes
(Chernew et al., 1998; Luft et al., 1991; Adams et al., 1991; Bronstein et
al.,1990,1991; Garnick et al., 1989; Burns et al., 1992; Goldsteen et al., 1994;
Hogan 1988; White et al., 1998; Phibbs et al., 1993; Buczko, 1994)
Exploring a theoretical framework with which rural patients' hospital
choice is estimated, this study investigates both institutional and individual
dimensions. The institutional dimension is composed of payment sources
as well as local rural hospital characteristics such as ownership, the number
of beds, and existing networks with other hospitals. The individual
dimension is composed of individual attributes such as age, the distance
between the nearest hospital and patient's residence, the distance between
the nearest local rural hospital and the next nearest hospital, proximity to an
urban county, and patient conditions such as severity of illness. This study
combines a comprehensive assessment of institutional and individual
74


dimensions in an attempt to determine factors that affect rural patients'
choice of a hospital for health care services.
Institutional Dimensions
Payment Sources
Various payment sources such as Medicare, Medicaid, managed care,
commercial insurance, or self-payment predetermine the pool from which
patients can select a hospital for medical treatment. Payment source is
critical in hospital choice because an insurer's contract provisions determine
the scope of medical coverage and potentially restrict a patient's hospital
selection. This study defines the payment sources as the primary source that
is expected to be responsible for the largest percentage of the patient's
current bill for medical treatment.
The payment sources can be classified into the following categories:
commercial insurance, managed care programs or discount care programs
such as Health Maintenance Organizations (HMOs) or Preferred Providers
Organizations (PPOs), Medicare, Medicaid, and self-pay. Commercial
payment sources are composed of Blue Cross/ Blue Shield (BC/BS) plans,
75


any private, indemnity, self-insured, self-funded, commercial insurance
carrier, as well as any liability, or no fault auto or home casualty insurance.
The commercial payment sources do not include HMOs, PPOs or managed
care, except any liability, no fault auto or home casualty insurance for the
purposes of this study.
The important distinction between public insurance (Medicare and
Medicaid) and commercial insurance, including HMO and PPO, is that the
former pays hospitals at a lower rate for services than the latter (Dranove et
al., 1993). The average Medicaid payment per diem in California was
approximately 50 percent to 60 percent of the average private insurance
payment. This can explain why hospitals tend to avoid patients with
Medicare or Medicaid insurance. Studies have shown that around 40
percent of obstetricians refused Medicaid beneficiaries during the 1980s
(Institute of Medicine, 1989). For Medicaid and Medicare beneficiaries, the
choice of hospitals may be limited to those facilities that will accept
Medicaid and Medicare patients, even though their out-of-pocket price of
health care is low. Medicaid beneficiaries are usually employed, but may
be less likely to have their own transportation. Dranove and his colleagues
(1993) found that Medicaid patients tend to go to public hospitals because
76


they are among the facilities that accept this form of insurance. In rural
areas, most of the hospitals are owned by public authorities or local
governments. So, it would be conceivable that the Medicaid beneficiaries
in rural communities are less likely to bypass their local rural hospitals than
those people not receiving Medicaid.
Because commercial insurance pays higher rates, compared with
public insurance (Medicare and Medicaid), commercial insurance is more
flexible to the insured in their choice of hospitals than public insurance.
This is especially true because the managed care providers, such as an HMO
and a PPO, own the medical facilities where their clients are treated, or
contract with facilities in urban or major rural cities. It is conceivable that
the patients in managed care programs are in a position to more frequently
bypass their local rural hospitals than patients in any other health insurance
program.
Self-payment is defined as payment by the patient, guarantor,
relatives, or friends directly to the hospital, for health care service charges.
A self-payer has no related restrictions in choosing hospitals. Although the
use of a hospital by a self-payer is less constrained than a patient using any
other payment source, it is conceivable that a self-payer is more likely to go
77


to their nearest local rural hospital if the nearest local rural hospital provides
the health care services that the self-payer is looking for. There are no
reasons that a self-payer would bypass their local rural hospital, assuming
that it offers the health care services that s/he wants.
It appears that the selection of hospitals by rural patients is highly
sensitive to variations in payment sources and related co-payment
provisions. In contrast to the limitations that exist with the inpatient
benefits offered by public insurance, such as Medicare and Medicaid, the
choice of hospitals for rural patients who retain managed care, such as HMO
and PPO, and commercial insurance, including Blue Cross and Blue Shield,
is relatively generous. Based on this information, the following hypothesis
is proposed.
Hypothesis 1: Commercial and HMO-PPO patients are more likely to
bypass their local rural hospitals than are rural
Medicaid and Medicare beneficiaries, and rural self-
payers, even though their local rural hospitals provide
the health care services that rural patients look for.
78


Hospital Characteristics
In addition to payment sources, hospital characteristics should also be
considered to be a set of institutional factors. Hospitals can be
characterized by the hospital size, such as the number of beds, the number
of health care services, and organizational factors, such the ownership type,
network with other hospitals, and medical school affiliation. These factors
should be considered to help differentiate hospital choice characteristics.
Some studies on hospital choice include the following hospital
attributes: the number of hospital beds (Chernew et al., 1998; Buczko,
1994;Goldsteen et al., 1994; Adams et al., 1991), ownership type (Chernew et
al., 1998; Bronstein et al., 1990; 1991; Garnick et al., 1989; Phibbs et al., 1993;
Luft et al., 1990; Burns et al., 1992), medical school affiliation (Chernew et al.,
1998; Luft et al., 1990; Burns et al., 1992; Adams et al., 1991), staff per bed
(Garnick et al., 1989), the number of out-of-state admissions (Chernew et al.,
1998; Luft et al., 1990; Burns et al., 1992), and the number of provided
services (Goldsteen et al., 1994; Bronstein et al., 1990; 1991; Buckzo, 1994).
This study also assesses the influence of several of the above hospital
characteristics on the propensity to seek health care services at certain
hospitals. Hospital characteristics discussed in this study include
79


ownership type, the number of beds, the number of services provided, and
the network with other hospitals.
One of the characteristics that make a hospital's attributes unique is
the ownership status of the facility. The ownership status of hospitals is
classified as for-profit1, non-profit2, and public owned3. Ownership status
may influence a patient's choice of hospital. The effect of ownership on
choosing hospitals varied according to the patient group, the diagnoses, and
the residence of patients. Because public owned hospitals primarily serve
indigent populations and Medicaid or Medicare beneficiaries (Sloan et al.,
1988), they often are less attractive to privately insured patients for this
reason (Phibbs et al., 1993).
Luft and his colleagues (1990) examined the influence hospital
ownership has on the choice of hospital for patients undergoing seven
surgical procedures (total hip replacement, cholecystectomy, hysterectomy,
transurethral prostatectomy, colon surgery, mastectomy, and coronary
artery bypass graft surgery) and five medical diagnoses (acute myocardial
infraction, stroke, pneumonia, chronic obstructive pulmonary disease, and
fracture of the femur) in hospitals in three geographic areas (San Francisco,
1 Investor-Owned (Individual, Partnership, and Corporation).
2 Non-Government, Not-for-Profit organization (Church Operated, Other).
3 County, City, City-Hospital District of Authority.
80


San Diego, and Modesto-Merced) in California in 1983. They found that
public and for-profit hospitals were routinely less likely to be chosen than
not-for profit hospitals. This analysis shows that patients are more likely to
bypass the nearest local hospitals if the nearest local hospitals are public or
for-profit hospitals.
Bronstein and her colleague (1990; 1991) examined the tendency of
rural pregnant women to bypass the nearest rural hospital in Alabama from
1983 to 1988. They found that when the nearest rural hospital was publicly
owned, rural patients were less likely to bypass these facilities. When the
next-nearest hospital or a hospital in the closest urban area was a public
owned hospital, rural patients were more likely to bypass their nearest rural
hospitals. This illustrates the preference rural patients have for publicly
owned hospitals.
Chernew and his colleagues (1998) examined the effect of the
relationship between hospital characteristics and patient choice patterns,
with particular attention to whether these HMO enrollees are more or less
likely than other patients to receive care at high-quality hospitals and
whether HMO enrollees travel farther to receive care. They found that
81


holding other observed factors constant, HMO patients were less likely to
receive health care services at public owned and for-profit hospitals.
This study excludes the private hospital as outlier from the data set
since there is only one private hospital among 36 Colorado rural hospitals.
This study examines how two types of hospital (public and non-profit)
influence to rural patients' hospital choice. Like Bronstein and her
colleague's studies (1990; 1991), this study examines the rural female
patient's hospital choice. Therefore, this study expects when the nearest
rural hospital was publicly owned, rural patients were less likely to bypass
these facilities.
The number of hospital beds relative to a patient's choice of hospital
will also be examined in this study. Bed quantity has been found to affect a
patient's choice of hospital positively and significantly (Adams et al., 1991).
A hospital's size is measured by the number of acute care beds, since this
study focuses on inpatient services. Patients take into account that the
higher the number of beds, the better the health care services provided by a
hospital. Bed quantity is a commonly used measure of the scope of services
available at a hospital (Bronstein et al, 1991). Therefore, it can also be a
82


proxy measure of a potential patient's perception of the quality of health
care in a particular hospital.
Buckzo (1994) examined hospital choice by aged rural Delaware
Medicare beneficiaries living in a zip code area that contained a local
hospital. He found that, while beneficiaries bypassing local rural hospitals
for other rural hospitals appear to be bypassing hospitals that are smaller
than the average local hospital, in reality, the local hospitals chosen by
beneficiaries treated in urban areas were typical of the average rural hospital
size. Rural patients treated in other local hospitals or urban hospitals were
treated in facilities that were significantly larger than the local rural
hospitals in their zip code areas. Rural patients bypassing local rural
hospitals for urban hospitals were treated in large hospitals with the
capacity for delivering many technologically advanced services.
Adams and her colleagues (1991) examined the effect of a patient's
severity of illness on hospital choice. They found that all hospital attributes
included in their model had significant and positive effects on a patient's
hospital choice. For example, the effect of bed size on hospital choice is
fairly stable across specifications and indicates that an increase in 10 beds
increases the odds of choosing a hospital by 1.7 percent, holding other
83


factors constant. Goldsteen and his colleagues (1994) examined factors that
influence the tendency of patients to seek health care services at other rural
hospitals or urban hospitals rather than local rural hospitals in using the
records of 2,171 rural Illinois patients who received inpatient treatment for
mental illness or substance abuse. They found that as a characteristic of a
local rural hospital, number of hospital beds is primarily significant in
choosing hospitals for health care services. In other words, a larger number
of hospital beds reduces the propensity by patients of choosing other rural
hospitals or urban hospitals.
One of the measurement scales of hospital characteristics should
include the number of health care services that each hospital can provide.
It is measured by the number of special units or facilities available at each
hospital. This measure anticipates that an increased number of health care
services provided by local rural hospitals will reduce the possibility that
rural patients may seek health care services elsewhere. This study adds, as
an independent variable, the number of health care services that each
hospital provides, in assessing the influence of characteristics of the hospital
in patient care choice. It is expected that, as with bed size, a larger number
of provided health care services, and a greater concentration of facilities in
84


local hospitals diminish the possibility that patients will bypass local
hospitals.
To support these statements, above, Goldsteen and his colleagues
(1994) found that when there is a larger number of health care services in
local rural hospital, patients are less likely to bypass their local rural hospital.
Phibbs and his colleagues (1993) found that patients had stronger
preferences for hospitals that provide more health care services. Patients
were willing to travel farther to reach a hospital that provided more health
care services.
Hospitals belong to networks with other hospitals. Some rural
hospitals have made network connections with other rural hospitals or
urban hospitals. The hospital that maintains a network with other
hospitals easily refers patients into urban hospitals or other rural hospitals
and can be medically supported by resources from other hospitals. These
hospitals that form networks with other facilities can provide better health
care services to patients than hospitals without networks, especially in rural
areas. Although no previous study adds the network as an independent
variable, this study expects that rural patients will be less likely to bypass
85


their local rural hospital if that hospital is involved in a network with other
rural hospitals or urban hospitals.
This study recognizes that rural hospitals are smaller, less
technologically advanced, and less specialized. They are predominantly
county or local government owned, and characterized as sources of primary
health care services. Often, they are located in more isolated communities,
lacking amenities and convenience, limiting the ability of these facilities to
recruit and attract qualified health professionals (Hogan, 1988). Overall,
this study proposes the following hypotheses to test the relationship
between hospital characteristics and patients' hospital choice.
Hypothesis 2: If the nearest local rural hospital is a public hospital, the
rural patient is less likely to bypass her nearest local
rural hospital to obtain health care services.
Hypothesis 3: The greater number of health care services that the
nearest local rural hospitals provide, the less likely rural
patients are to bypass their local rural hospitals.
Hypothesis 4: The greater quantity of beds that the nearest local rural
hospital has to offer, the less likely rural patients are to
bypass their local rural hospitals.
86