1

1
1
AN EXPANDING FRAMEWORK FOR RURAL PATIENTS WHO TRAVEL
FOR HEALTH CARE
by
Sharon Sweeney Fee
Copyright © Sharon Sweeney Fee 2004
A Dissertation Submitted to the Faculty of the
COLLEGE OF NURSING
In Partial Fulfillment of the Requirements
For the Degree of
DOCTOR OF PHILOSOPHY
WITH A MAJOR IN NURSING
in the Graduate College
THE UNIVERSITY OF ARIZONA
2004
UMI Number: 3145135
Copyright 2004 by
Sweeney Fee, Sharon
All rights reserved.
INFORMATION TO USERS
The quality of this reproduction is dependent upon the quality of the copy
submitted. Broken or indistinct print, colored or poor quality illustrations and
photographs, print bleed-through, substandard margins, and improper
alignment can adversely affect reproduction.
In the unlikely event that the author did not send a complete manuscript
and there are missing pages, these will be noted. Also, if unauthorized
copyright material had to be removed, a note will indicate the deletion.
UMI
UMI Microform 3145135
Copyright 2004 by ProQuest Information and Learning Company.
All rights reserved. This microform edition is protected against
unauthorized copying under Title 17, United States Code.
ProQuest Information and Learning Company
300 North Zeeb Road
P.O. Box 1346
Ann Arbor, Ml 48106-1346
2
The University of Arizona ®
Graduate College
As members of the Final Examination Committee, we certify that we have read the
dissertation prepared by
Sharon Sweeney Fee
entitled ^ Expanding Framework for Rural Patients who Travel for Health Care
and recommend that it be accepted as fulfilling the dissertation requirement for the
<5
Doctorate in Nursing
Degree of
date
'CoeiA.M^ {j\
1
i/(
1/ J l/oy
/
date
date
Final approval and acceptance of this dissertation is contingent upon the
candidate's submission of the final copies of the dissertation to the Graduate College.
I hereby certify that I have read this dissertation prepared under my direction and
recommend that it be accepted as fulfilling the dissertation requirement.
'
Dissertaiioni Director:
1
^-Qf
date
3
STATEMENT BY AUTHOR
This dissertation has been submitted in partial
fulfillment of requirements for an advanced degree at The
University of Arizona and is deposited in the University
Library to be made available to borrowers under rules of the
Library.
Brief quotations from this dissertation are allowable
without special permission, provided that accurate
acknowledgment of source is made. Requests for permission
for extended quotation from or reproduction of this
manuscript in whole or in part may be granted by the copyright
holder.
SIGNE;
4
ACKNOWLEDGEMENT
I would like to acknowledge the following for their kindness and generosity in
aiding my Journey though the data. I would not have been able to complete my
work without the assistance and advice from them all.
Ed Welsh and Daniel Cochran, Arizona Department of Health Services,
for the data bases.
Craig Wessler and Mickey Reed, U of A School of Renewable Resources,
for their GIS expertise.
Dr. Thomas C. Ricketts, University of North Carolina, for his knowledge of rural
health policy.
Daniel C. Malone, R.Ph., Ph.D., U of A College of Pharmacy, for his assistance
with cost analysis.
Kathy Roberson, Manger, University Medical Center Health Information
Management Office, for her help with DRG financial information.
Fran Roberts, RN, Ph.D., FAAN, Arizona Hospital and Health Care Association
for her friendship and reality checks.
And finally, my dissertation committee, for their patience and tolerance:
Gerri Lamb, RN. Ph.D., FAAN, Dissertation Chair
Joyce Verran, RN, Ph.D., FAAN, PhD Advisor
Judith Effken, RN, Ph.D.
Sally Reel, RN, Ph.D.
5
DEDICATION
This work is dedicated to two very special people who loved and supported me in
ways I will never be able to repay.
First, to my husband Kerry, whose great meals, wonderful walks, and ability to
always tell me what day it was, kept me fed, rested and oriented. For you, all my
love and joy, for our dreams have come true.
Second, to my mother Gwen Kinmon RN, a woman whose convictions were
strong and whose nature was gentle. May I always walk in honor of your life,
what an example you left me.
6
TABLE OF CONTENTS
LIST OF ILLUSTRATIONS
LIST OFTABLES
ABSTRACT
1; InlToduction
Statement of Problem
Disparities in Rural Health
Health Geography
Distance
Rural Economy
Outcomes
Statement of Purpose
Research Questions
Summary
2: Theoretical Framework
Defining Migration
Migration in Biological and Social Science
Migration in Health Science
Migration for Health Care
Framework
Structure
Rural and Urban System Capacity
Patient Decision
Patient Characteristics
Process
Distance Traveled
Outcome
Rural and Urban System Outcomes
Risk Adjusted Patient Outcomes
Rural Community Outcomes
Summary.
3: Methodology
Design
Data Source
Sample and Unit of Analysis
Primary Data Collection
Reliability of Primary Data
Secondary Data Analysis
9
10
...11
12
12
15
17
20
21
.....21
22
22
23
25
25
25
26
28
28
30
31
.32
33
34
34
35
..36
36
38
38
40
40
41
42
42
43
43
7
TABLE OF CONTENTS - Continued
Decision Rules
Cost Analysis
Human Subjects
Summary
45
46
46
46
4: Creating the Analytic Data File
Patient Characteristics
Patient Access Characteristics
Patient Need Characteristics
Distance Traveled
Risk Adjusted Patient Outcomes
Outcomes
Decision Rules
Summary
47
48
49
52
..54
59
..61
62
63
5: Results
Description of the Sample
Research Question One
Patient Access Characteristics
Patient Need Characteristics
Research Question Two
Other Analysis
Cost Analysis
Summary
65
65
71
....72
.74
78
79
81
90
6; Discussion of Results
Research Question One
Research Question Two
Relationship of Findings to Conceptual Model
Patient Access Characteristics
Patient Need Characteristics
Risk Adjusted Patient Outcomes
Limitations
Implications for Health Policy and Nursing
Recommendations for Further Research
Conclusion
92
93
94
95
96
98
..99
101
102
105
106
APPENDIX A; AzDHS Hospital Discharge Data Report Requirements
109
APPENDIX B: University of Arizona Human Subjects Exemption
121
APPENDIX C: Data Coding Dictionary
123
8
TABLE OF CONTENTS - Continued
APPENDIX D: AzDHS Hospital Identification File
125
APPENDIX E: Geocoding Process Description
129
REFERENCES
...133
9
LIST OF ILLUSTRATIONS
Figure 1: A Model for Studying the Concept of Migration for Health Care
30
Figure 2: Study Model
40
Figure 3: Map of Arizona with hospital clusters and patient zip codes...
Figure 4: Model with study variables
..........57
96
10
LIST OF TABLES
Table 1: Organization of AzDHS variables
Table 2; Frequency of Patient Access Characteristics
Table 3: Frequency of DRG groups............
Table 4: Frequency of Patient Need Characteristics.
41
...66
67
—69
Table 5: Frequency of Mileage groups
69
Table 6: Frequency of Age Groups and Outcome Variables
70
Table 7: Analysis of Patient Access Characteristics.
73
Table 8: Analysis of Patient Need Characteristics
76
Table 9; Outcomes Regression Analysis
.79
Table 10: Time One Frequency and Definition of Top DRGs
83
Table 11; Time Two Frequency and Definition of Top DRGs
84
Table 12: Time One Cost Analysis Data
86
Table 13: Time Two Cost Analysis Data
87
Table 14: Length of Stay Outlier Costs
89
ABSTRACT
This exploratory study utilized Donabedian's Quality model to develop a
framework to study patients who must migrate for health care. One year of the
Arizona Department of Health Services Discharge Database was used to analyze
patient characteristics that influenced discharge travel and the impact of distance
on risk' adjusted patient outcomes.
Geographic Interface software was used to identify rural patients, defined
as those with zip codes farther than thirty miles from hospitals. Zip Code analysis
was used to create distance variables between 31 and over 300 miles. The key
findings for patients who traveled greater distances included larger hospitals,
emergency admission type, private insurance, critical care services, and
Neuro/Ortho/Trauma diagnosis group. Patients which traveled shorter distances
included smaller hospitals, referral or transfer admit source, AHCCCS insurance
(or Medicaid) and Women's Health diagnosis group.
Outcomes were risk adjusted using age and distance was significant for
both number of procedures and length of stay. Patients who traveled farther
received fewer procedures and had a greater length of stay. A preliminary cost
analysis of the length of stay outliers identified approximately four million dollars
in potentially non-reimbursable charges.
CHAPTER ONE
Introduction
Rural health research has attempted to understand the needs of rural patients, rural
caregivers, rural practitioners and rural health systems. Through comparisons with urban
systems, the disparities in access, utilization, standards of care and opportunity for
services for rural health have been well documented (Ricketts, 1999). Rural systems were
defined as those areas which were 'not urban' and with each study, there remained a lack
of understanding of the health outcomes for rural patients and a lack of recommendations
for improvement of rural health systems.
In 2001, the Institute of Medicine (lOM) reported on the need to create a more
equitable and efficient health care system. That report:, combined with the current
national concerns regarding health care financing and the distribution of health services
(Licberman, Lee, Anderson, & Crippen, 2004), demonstrates the need for rural
researchers to reframe their views of rural health care in order to develop interventions
that can improve rural health. A framework that allows for an increased understanding of
rural outcomes and the factors that affect those outcomes may allow policy makers to
develop programs that will lead to a more efficient and equitable system.
Statement of the Problem
The focus of rural health research has been on the recognition of disparities with
and comparisons to urban systems, primarily in the areas of access and utilization
(Ricketts, 1999). There have been few attempts to understand rural health outcomes.
Rural research also uses government designation of metropolitan and micropolitan
counties to define rurality by exception. Identifying the variances in health services for
areas that may or may not be 'rural' has yet to lead to recommendations or solutions for
either suppliers or users of rural health care.
Rural patients frequently travel long distances to obtain carc in urban settings
(Basu & Cooper, 2000). As a result, their health experience may be very different from
their urban counterparts. Recent studies of patient travel, or migration, have investigated
the characteristics of patients who travel and the type of systems to which they travel
(Basu and Cooper, 2000; Nemet & Bailey. 2000). Little is known about the role of
distance on health outcomes or how those distances are negotiated as patients travel.
across the boundaries defined by health care geographers. These studies, again, have
focused on the access and utilization of services for an audience of insurers and heath
care system stakeholders. The studies neither measure the impact of distance on patient
outcomes nor have they led to the development of interventions that will change the
standard of care for rural patients.
One study measured outcomes for rural patients who traveled for cardiac care
(Sweeney Fee, 2002). This exploratory study analyzed risk factors related to the
readmission of rural patients who were transferred to urban centers for cardiac
interventions. Using Verran's systems framework (1997), the records of Veterans
transferred by a rural hospital over a distance of more than 200 miles for cardiology
services were examined. All patients transferred over period of 9 months were included
(N=40) and charts were reviewed for a period of 12 months post transfer to identify
patients who were readmitted to the rural hospital (n=19). Variables for the readmitted
and non- readmitted groups were organized according to the systems framework and chisquare analysis was used to identify which variables were significantly related to
readmission.
Four variables were found to be significant (p>.05) for this rural cardiac sample:
transfer method, cardiology treatment, discharge diagnosis and proximity to the rural
hospital. For this population, patients who were transferred by strangers were
significantly more likely to be readmitted (p>.01) than patients transferred by staff
familiar to them. Those patients who received only diagnostic catheterization or medical
management were also significantly more likely to be readmitted (jp=.01) than those who
received surgical interventions. Patients who lived closer to the rural hospital and those
who received a discharge diagnosis of AMI or unstable angina were also significantly
more likely to be readmitted (p=.05).
This study indicated that risk factors related to readmission of rural cardiac
patients are different than those found for urban patients. The high incidence of
readmission in this study (48%) and the difference in type of risk factors from those
classically understood for cardiac patients demonstrates the need to study the specific
characteristics of rural populations. Developing interventions based on urban
characteristics may lead to inappropriate care for rural patients. Inaccurate identification
of risk factors related to rural health outcomes can be a costly mistake for hospitals and
patients.
The results of the pilot study demonstrated significantly different risk factors and
outcomes for rural patients who traveled for care. An attempt to find other studies of rural
patient characteristics and health outcomes resulted in an information void. Effective
interventions for the disparities in rural health cannot be developed if the characteristics
of rural patients are not understood. There is no way to measure the quality of an
intervention if health outcomes are not included. Developing rural health policy without
empirical knowledge of rural populations is inefficient and ineffective. Therefore, this
exploratory study is aimed at understanding the characteristics and the health outcomes of
rural patients who must travel for care.
Disparities in Rural Health
Rural America is older and growing at a slower pace, both economically and
demographically, than its urban counterpart. In regard to health care, rural hospitals are
paid less, are slower to integrate technology, and are less effective at integrating best
practices (Ricketts, 1999). Disparities have been recognized in three main areas: rural
health delivery systems, rural access, and rural health status.
Rural health delivery systems have had significant struggles to remain viable. In
the 1980's, rural hospitals closed at greater rates each year, from 22% in 1980 to 50% in
1989 (Ricketts, 2000). Many of the closings were a result of national health care finance
reform that led to a lower Medicare payment structure for rural hospitals. The most
vulnerable hospitals were those with fewer beds, low occupancy rates, and a high
percentage of Medicaid patients. Since that time, the surviving rural hospitals have
demonstrated two characteristics which appear essential to their success: an ability to
contribute to their local economy and an adoption of other payment options such as
outpatient care and health clinics (Ricketts, 2000).
In 2003, Younnis analyzed the economic pcrfomiance of rural hospitals in
comparison with their urban counterparts. Using Medicare Cost Report data from 1991
and 1995, the study found that rural hospitals had lower profitability, generated less
revenue per bed, received few, if any, donations, and were significantly disadvantaged in
terms of performance.
Even when the health systems are adequate, access to care issues can continue to
impact rural patient's health. Rural patients have less access to employer supported
insurance, use fewer specialists, have fewer usual sources of care and are hospitalized
more (Schur & Franco 1999). Rural patients also have fewer choices in health
professionals (Rosenblatt & Hart, 1999) and fewer options for health care coverage
(Casey, 1999).
Rural areas have been shown to have disparities in many health status measure,
from prenatal care to chronic illness. Rural patients have lower rates of prenatal care and
higher rates of both teen pregnancy and neonatal deaths as well (Lishner et al, 1999). In
addition, there are higher rates of fatal injuries to rural children (Clark, Savitz, &
Randolph, 1999); chronic shortages of rural mental health providers (Harltey, Bird, &
Dempsey, 1999); and deficiencies in rural long-term care services (Coburn & Bolda,
1999). Rural areas also have higher rates of chronic diseases and mortality from trauma.
Many of the risks of occupational injury from machinery used in and exposure to toxins
from farming, mining and forestry are unique for rural patients and the systems that care
for them (Ricketts et al, 1999).
The recognized disparities, fewer providers, fewer insurers, the lack of funding,
and the lack of new technology, result in a need for a different framework of access to
care for rural patients. Rural systems are unable to meet all the health needs of rural
populations. In 2002, the Institute of Medicine (lOM) not only officially recognized the
"dearth of clinical programs with the infrastructure required to provide the full
complement of services" (p 4) but also that "fundamental reform of health care is needed
to ensure that all Americans receive care that is safe, effective, patient centered, timely,
efficient, and equitable" (2002. p 8).
Rural research that simply measures the differences in health care for specific
geographic areas has not resulted in changing the disparities in health for rural
populations. When rural systems cannot meet the needs of their communities, rural
patients will travel outside their 'geographic area' in order to obtain needed services. A
better understanding of that travel and the relationship between distance and health
outcomes may help rural systems develop strategies to move toward effective, efficient
and equitable care. A new framework is needed to understand rural health outcomes in
the systems that patients access. This can lead to the development of interventions that
could impact, rather than merely identify, some of the disparities in rural health systems,
access and health status.
Health Geography
To date, government agencies have played a major role in the definition of
rurality. Initially, rural areas were defined by what they were not, ie: 'non-urban'. Up until
2003, there were two main accepted definitions of urban areas used by government
agencies. The metropolitan (MA) and non metropolitan (NMA) designations of the
Office of Management and Budget (0MB) divided counties between those integrated
with a population center greater than 50,000 (MA) and those that were not (NMA). The
US Bureau of the Census used the same population numbers, along with measures of
population density and the relationships between communities, such as a high number of
commuters, to develop Metropolitan Statistical Area (MSA) (Ricketts, Johnson-Webb,
Taylor, 1998). These urban definitions were then used to identify rural areas by exception
and this 'not urban' designation was the basis of much of the rural research and health
policies that were funded and developed by government agencies.
Several government agencies have expanded on the current urban definitions in
an attempt to better identify rural areas. The USDA developed Urban Influence Codes to
help reflect a clearer picture of county roles and interactions for farming communities
which were not recognized using the urban designations (ERS, 2003). These nine codes
first define 3 types of 'metro' counties based on metropolitan populations of greater than
250,000, between 250,000 and 1 million, and greater than 1 million. The 6 'non metro'
counties are based on diminishing population numbers and adjacency to a metro area.
Within this scale, the population must be less than 2,500 to be classified as 'completely
rural'. This greater degree of division allows for an increased ability to understand the
differences between metro, non metro and rural communities.
More recently, in 2003, the 0MB released new standards using Core Based
Statistical Areas (CBSAs) designations in order to better define expanding urban areas.
The new standard adds a Micropolitan component, resulting in three new categories:
Metropolitan CBS A (an urbanized area of 50,000 or more), Micropolitan CBSA (an
urbanized area of 10,000 - 49,999) and Areas Outside CBSAs. The Census and US DA
have followed with re-designation of their codes. This currently is increasing the
confusion related to rural health research by providing a pretext for "two tier" rural
designations and funding, while rural remains primarily defined by exception (Slifkin,
Randolph & Ricketts, 2004). The lack of one unit of measurement to define rural can
confound both the health systems researcher and the rural populations they hope to serve.
The lack of clear agreement on what is rural has led some scientists to conclude that
much of rural policy is actually a poorly fitted urban policy (Stauber, 2001).
Researchers continue to work on improving the definition of rural. The Montana
State University Rurality Index, for example, asks two fundamental questions: how many
miles must you drive and how long does it take to get to emergency care (Weinert &
Boik. 1994). This Index was an attempt to develop a surrogate variable that would
identify degrees of rurality. The Index also recognized the differences in distance to
emergency care within rural counties. These two variables allow for a better
understanding of access from within counties without eliminating the ability to analyze
differences between the counties. While the continuing discussion and disagreements on
defining rurality are recognized, the call is to continue to work toward the development
of one "geographic-based statistical system for the entire United States....It should reflect
both the character of the urbanization. ..and fate of its rural and non-rnetropolitan places"
(Slifkin et al, 2004 p.5).
20
Distance
The measure of distance is a variable which has been used to understand
differences in access and utilization of health care services for rural patients without
having to establish geographic designations. One study of rural resident's travel over
distance for primary care attempted to understand why patients traveled out of their
county to receive care (Borders, Rohrer, Hilsenrath, & Ward, 2000). The concepts
studied, based on Andersen's Behavioral Model of Health Services Use, were the rural
health system, predisposing and enabling patient variables and perceived need. These
were analyzed with the location for physician services as the dependent variable. This
research found that perception of quality of local providers had the strongest impact on a
patient's willingness to travel for care. Availability of providers was not found to be as a
significant variable.
In 2000, Basu and Cooper studied the relationship between distance traveled and
hospital admissions using a state hospital discharge data base. Using regression analysis,
they looked specifically at the group of patients who traveled with a primary
hospitalization diagnosis that was sensitive to a lack of ambulatory care services. The
results demonstrated that patients who traveled for care were primarily white males with
insurance coverage. The older and non-white patients were more likely to stay within
their county for care. They concluded that those very old or those on Medicaid would be
most affected by changes in rural health care.
Both of these studies demonstrate the need to understand the choices rural patients
make when deciding on their health care. Rural health systems cannot make changes
based on unclear perceptions of quality and the either greater options or limitations of
insurance providers. The opportunity to use either study to create interventions which
may improve the quality or availability of services for rural health is limited. Neither of
• the above studies measured health outcomes for the patients that traveled.
Rural Economy
While the impact of disparity of health care for rural patients has been
demonstrated; it is important also to recognize the impact of vulnerable health systems on
the economy of the rural communities. As one researcher states: "Economic improvement
and growth alone are not enough to sustain communities. They are necessary, but not
sufficient. Communities that survive and prosper also invest in building the social and
human capital of their institutions and people" (Stauber, 2001 p.70). A study of
community satisfaction found that medical services and facilities were significant
contributors to satisfaction with a community as a place to live (Beesley, 1999). The
relationship between the rural health systems and the rural community is important to the
prosperity of both. Healthy individuals are dependent on healthy care delivery systems,
and the survival of the community is based in the strength of the institutions and people.
Outcomes
Health outcomes can range from the length of stay for an individual patient to the
readmission rates for a health system. Researchers are increasingly utilizing improved
methods in risk adjustment to understand variations in health outcomes (lezzoni, 2003).
Currently, studies using risk adjustment to measure heath outcomes have improved our
understanding of a number of factors that impact patient outcomes, such as RN staffing
ratios (Neeldeman et al, 2002). These methods can also be applied to the relationship
between distance and patient outcomes.
An understanding of migration for health must include both measures of distance
and risk adjusted health outcomes in order to clearly understand variation in outcomes for
patients who travel. Combining those variables with measures used to identify issues of
rural health will lead to a more global understanding of factors related to health outcomes
for rural patients. A model of migration for health will allow researchers to find
modifiable variables that can result in the development of interventions for rural patients
and the systems that care for them.
Statement of Purpose
The purpose of this research is to look at the impact of distance on hospital
outcomes for rural patients using a framework of migration. This exploratory study will
begin with an analysis of the literature which currently defines migration. The study will
also include a secondary data analysis of the Arizona Department of Health Services
(AzHDS) Hospital Discharge Data Base (HDDB), with distance as the variable of interest
and using risk adjustment to understand the relationship between distance and risk
adjusted patient health outcomes.
Research Questions
1. What are the demographic factors that influence distance traveled for rural
patients?
2. What is the relationship between distance and risk adjusted inpatient health
outcomes for rural patients?
23
Summary
Current health research recognizes disparities in rural health without
recommending solutions for rural health systems. Health care geographers continue to
work toward developing better designations for rural populations. Our understanding of
distance in relationship to rural health is used primarily to inform providers and insurers
on patterns of access and utilization of urban services. Health system researchers have
improved methods for a better understanding of factors related to inpatient health
outcomes but have not used those methods to understand the relationship between
distance and outcomes.
A new model that develops an understanding of migration for health care may
help improve the equality and efficiency of rural health systems. It may also help health
system researchers navigate the new policy environment which Ricketts describes as:
"harbingers of the more general national health care system, in which
pressures for efficiency and cost control may create systems that are expected to
function with far fewer resources...in which a heightened concern for outcomes
places the systems in a unique ethical bind" (2000, p.652).
Rural communities deserve an accurate understanding of the issues related to their
health care systems. Without including migration, those communities may never have a
clear picture of how their citizens receive care or how to create policies that will develop
an efficient and equitable system. Improved methods are needed to ensure tbat the ethical
bind between efficiency and outcomes does not constrict our rural health systems.
Migration for health care has been a missing part of the new models of health care
delivery. Who migrates, where they are migrating and why, and the impact of that
migration on communities and health care systems are important questions that need to
be asked when developing agendas for rural health policy.
CHAPTER TWO
Theoretical Framework
A major goal of this research is to develop an expanded framework of rural
migration for health care. This chapter defines migration and explores its application in
biological, social sciences and health sciences. A new conceptual framework of migration
for health care services grounded in systems theory is presented. Support for the selection
of the variables in the framework is described.
Defining Migration
The Encyclopedia Britannica (2001) defines migration as the movement from one
country, place or locality to another; to change position in an organism or substance. One
Webster definition for migration is "the ability to pass from one region to another"
(1994). When developing a model of migration for care, an understanding of existing
migration studies assist in defining the term. Studies of migration can be found in both
social and health science literature.
Migration in Biological and Social Science
In 2002, the United Nations recognized the fundamental right of migration in"
Article 13 of the Declaration of Human Rights (UN, 2003). Models and theories of
migration are often used to better understand the impact of individual's movement on
economies, communities and the individuals themselves. Outcomes range from cost
analysis to enhanced psycho-social understanding. The result leads to a growth of
knowledge and the ability to adapt policies for migrating populations based on scientific
evidence.
By using sensitivity analysis, the risks, patterns, and equilibrating forces of the
economics of labor migration can be used by economists to characterize optimal
migration patterns for a population (Chen, Chang & Leung, 2003). Migration is a major
consideration for municipal planners who view it as a demographic event equal to births
and deaths for a community (Voss, Hammer & Meier, 2001). Psychologists have
recognized migration as a major life event which can impact a person's identity and sense
of .self (Ward & Styles, 2003).
In sociology, ten classifications of cyclical migration have been identified, which
include commuters, workers, migrants and dwellers (Roseman, 1992). These
classifications are used to categorize migrating populations in order to understand the
specific patterns and social needs of each migrating group. Standard human capital
models, which evaluate the costs and benefits associated with a change in location, have
been used for cost analysis of migration patterns in an attempt to understand its impact on
the economy of communities and individuals ( Yankow, 2003).
Migration in Health Science
Migration is not a new concept for health care researchers. Models and methods
are seen in a variety of health sciences and all can lead to a clearer understanding of the
impact of migration on the health outcomes of communities and individuals. The term
migration has long been recognized by health science researchers; in fact, Hippocrates
recognized migration, or change in location, when he noted that 'it is changes that are
chiefly responsible for disease' (Mascie-Taylor & Lasker, 1998, P. 109).
The sciences of immunology and epidemiology approach the migration of
diseases from two different viewpoints. Immunologists study the migration and
transmigration of cells and their impact on disease and illness (Immunology, 2002).
Epidemiologists track pathways of disease across populations and geographic areas and
monitor the biologic aspects of migration (Mascie-Taylor & Lasker, 1988). Public health
researchers look at the relationship between migration and mortality within different
socioeconomic groups (Stephenson. Matthews & McDonald, 2003). The terms distance,
travel, bypass, and out-of-area travel are often used in health systems research. The
majority of public health research is focused on issues of access and utilization of
services when a distance variable is used (Ricketts, 1999); health outcomes are not
typically measured.
In 1979, The World Health Organization (WHO) applied the term of migration to
the issue of physician and nurse travel across boundaries to practice (Mejia, Pizurki &
Royston). Thus, not only is migration of patients recognized, but also the migration of
those who care for them. WHO defined migration as the result of the interplay of forces
at both ends of the migratory axis. For example, a depressed economy or oppressive
political system at one end of an axis and an improved economy or balanced political
system on the other end leads to a migration of physicians and/or nurses from one axis to
the other. In a recent article, which cited the original WHO study, Kline (2003) noted a
continued concern for the effects of and ethical issues related to the migration of
providers for health systems at both ends of the axis.
28
Migration for Health Care
With regard to rural health, the concept of migration has been used in few studies.
One study discussed the methodological issues in assessing rural-urban differences in
health care and recognized geographic migration as an important environmental, or
contextual, concept which impacted rural elderly and their families which needed to be
included in studies (Wallace & Wallace, 1998). Their discussion of rural elderly patients
recognized the need to explore the impact of migration patterns on family relationships.
One study was found which did research migration and defined migration as bypassing
local health care providers (Borders et al, 2000).
For this study, migration is defined as the act of passing from one region to
another to receive health sei-vices. The key to this definition of migration is that the
patient travels over a distance, to obtain either needed or desired health care. The
previously discussed studies of distance traveled for health care services focused on one
concept in the context of migration. The boundaries crosscd and the distances traveled
become variables which need to be analyzed in the study of migration. As demonstrated
in the migration studies, the act of traveling is only part of the picture. The concept of
migration also includes the antecedents to and the outcomes of that travel
Framework
The theoretical model for this study was developed using Donabedian's Quality
Model (Donabedian, 1966). Donabedian defined structure, process and outcome for
health systems: structure is the attributes of the systems resources, process is what is
actually done in deli vering care, and outcomes are the effect of the care on health status.
While many variations have occurred over time, the original model continues to be used
as an organizing framework for health systems research. In addition to measuring
structure, process and outcome, Donabedian discussed the importance of using the model
to evaluate quality of care, or the quality of the outcome. Quality was defined as either
maximally or optimally effective where maximal was all that could be done in a perfect
setting and optimal was that which can be done within the system constraints of the
organization. Cost analysis techniques were recommended to improve the analysis of
quality. Donabedian argued that simply understanding outcomes would not improve care;
it is the measure of quality that resulted in recognizing the "best in practice"(p 1743).
Finally, his model defined a number of levels at which assessment can occur, from
patient to provider to community which allows for a multi-level analysis of the quality of
both health system and patient outcomes.
Figure 1 displays the expanded framework for this study. As a result of the
information available in the data base used for this study, the specific constructs that can
be analyzed are in bold type: patient characteristics, distance traveled and risk adju.sted
patient outcomes. A variety of studies were used for the development of this framework
and a broad description of each of the variables within the framework follows. A detailed
description of the methods used in developing the constructs for this study model will be
presented in Chapter 3.
30
Figure 1: A Model for Studying the Concept of Migration for Health Care
STRUCTUS^^
PROCESS
Rural System Capacity
Urban System
Capacity
OUTCOME^
Rural System
Outcomes
Distance Traveled
Urban System
Outcomes
Patient Decision
Risk Adjusted
Patient Outcomes
Patient
Characteristics
Rural Community
Outcomes
Structure
In Donabedian's model, structure has been defined as the attributes of the care
setting (Donabedian, 1966). The structure in the study model modifies Donabedian's
original definitions based on more recent organizational models. The development of the
study model began with the variable of interest, distance traveled, w hich is the process
which occurs in order to receive care. The antecedents to the travel were then organized
as structural variables and the consequences of the travel are the outcomes in this
framework.
This study contains both organizational and patient structure in the framework. In
structural contingency theory, structure is defined as organizational specialization,
standardization and discretion (Mitchell, Shannon, Cain, Hegyvary, 1996). Verran's
(1997) system framework defines structure as both the internal and external context of
the health care system, which includes the organizational environment. The
organizational variables in this model are health care system capacity for both the rural
and urban systems.
The patient structural variables included in the framework are the patient decision
and patient characteristics which are traditionally seen as process variables in other
models. For example, in Verran's system framework, patient variables are part of the
materials technology, or nature of the client, within the process of care (1997). Structural
contingency views process as an element of coordinated patterns such as infonnation
flow (Mitchell et al. 1996). Patient decision and characteristic variables are placed within
the structure of this model in order to represent the context and discretion used prior to
traveling for care. Distance studies have recognized that patients travel because of either
medical need or personal choice. Using two patient variables allows for recognition of
many of the varied factors that may impact a patient's decision to travel for care.
Rural & Urban System Capacity
Health system capacity has been recognized in the literature as the availability of
services and satisfaction with the care delivery system. (Borders et al, 2000). Nemet and
Bailey included the number of provider offices as another structural capacity variable
(2000). Basu and Cooper have measured capacity using the teaching status and number of
beds for each health system (2000). Another study added staff mix and a measure of
technology as system capacity variables (Aiken, 2002). In this framework, components of
Health Care System Capacity can include physician mix, payer contracts, teaching status.
the number of beds and the complexity of services offered for both the rural and urban
hospitals. All capacity variables can be measured at the group, unit or institutional level.
A definition of rural and urban systems is necessary in order to understand the
differences between the systems. As previously discussed, there are a number of issues
when defining rurality and no clear unit of measurement. For this study, due to the focus
on migration for care, rural systems will be defined as those hospitals that are greater than
thirty miles from any other hospital. Urban systems will be defined as those hospitals
which are within a 30 mile radius of each other. This will allow for recognition of rural
patients who must travel over distances to access care. Patients who live in an area with
more than one hospital within 30 miles of each other may also migrate for care, but this
study is focused on rural patients specifically and will use the concentration of hospitals
as the criteria for urban and rural designation. This rural - urban definition will be used
throughout.
Patient Decision
Patient decision can incorporate both traditional access and utilization concepts.
Borders et al (2000) discuss characteristics which can affect patient decisions that include
religion, household size, social support, income and insurance. Additionally, the
structural decision variables which they defined were perceived need of service, past use
of services, frequency of use of services and acute medical need. Schur and Franco
(1999) discuss the impact of insurance coverage, usual source and site of care, travel and
waiting times on patient access to care. They also found that the availability of
specialists, nuinber of physician visits, and employment issues such as insurance and paid
sick leave affect rural patient's access to care. This framework includes patient decision
variables in order to more clearly evaluate patient indicators which preceded traveling for
care.
Patient Characteristics
To better understand migration demands that we describe the characteristics of
those patients who travel for health care. Characteristics can include demographic
information such as age, gender, marital status, racc, education and location of residence
(Conner, Rralewski & Hillson, 1994). Basu and Cooper (2000) found that travel for care
declined with age and for nonwhite patients. Other studies have used insurer, diagnosis,
comorbidities, and patient demographics (Thomas et al, 1998). For this study, patient
characteristics will be organized to describe both access variables such as insurer and
medical need variables such as diagnosis.
Finally, it should be noted that some variables, such as insurer and satisfaction,
can be included in more than one structural concept. Two important distinctions must be
made. First, the health system capacity concepts are measured at an organizational level,
while the two patient concepts are measured at the individual level. Second, based on the
research question, decisions must be made related to the organization of the variables in
the structural concepts. Patient characteristics focus on describing a population, while the
patient decision focuses on understanding the population. Both may use similar variables
but from a different conceptual focus. One example is identifying the types of insurance
payers as a patient characteristic and evaluating the types of insurance coverage with
relation to providers, covered services and co-payments as part of the patient decision
variable.
Process
Process is defined as the activities that occur in receiving care (Donabedian,
1966). Process elements of coordination and information flow have been combined with
structural elements to identify operational patterns in a system using contingency theory
(Mitchell et al, 1996). Mitchell and colleagues found that clinical processes were a
mediator of clinical outcomes. Verran's (1997) model defines process as the technology
which includes the materials and knowledge technology in the system. The materials
technology is described as the nature of the client, or patients, while the knowledge
technology is the nature of the clinician's knowledge and the clinical tools used. In this
framework, the process is the distance traveled for care. The goal of this study is to
explore the relationship between distance traveled and risk adjusted patient outcomes.
Distance Traveled
In this model, the process variable is the distance between the patient's location
and the system where care is received. The activity of traveling over that distance is the
process through which the patient obtains needed services. The choice of geographic
boundary is not required for this measure, thus eliminating the need to choose which
geographic measure to use. The limitations related to the use of geographic boundaries to
understand health care have been recognized; the lack of agreement on a fundamental
unit of measurement and the impact of social forces and power relationships on space and
place in health service delivery can lead to irrelevant decisions (Ricketts, 2002).
Distance traveled is the variable of interest in this study. Distance has been
identified as a function of the location of residence. This can include a variable that
represents whether a patient resides in the same town or county as the health services
(Borders et al. 2000). Another component of distance is a self report of miles and/or time
(Weinert & Boik, 1994). Zip code analysis also has been used to identify the distance
between a health care system and a patient location (Mooney et al, 2000). The use of
distance allows for a more patient focused view of access, utilization and outcomes than
the more abstract systems focus of health geography.
Outcome
Outcomes are the effects of the care received, which are then used to evaluate the
quality of that care (Donabedian, 1966). Structural contingency theory defines the
organizational outcomes of quality, resource use and satisfaction (Mitchell et al, 1996).
The goal in Verran's framework is described as the either the clinical or system outcome
and can be measured at the individual, group or organizational level (1997). The
American Academy of Nursing Expert Panel on Quality has developed five generic
outcomes which include; health related quality of life, symptom management, self care,
perception of being well cared for and health promotion (Mitchel, Ferketich & Jennings,
1998). Studies of inpatient outcomes have included length of stay, mortality (Zhan &
Miller, 2003), failure to rescue (Aiken et al, 2001) and the hospital complications of
urinary tract infection, pneumonia and metabolic derangement (Needleman et al, 2002).
Outcome variables, like many of the structural variables, can be measured at both the
individual and system level.
Rural & Urban System Outcomes
These variables are included in the framework for both rural and urban systems to
help evaluate the impact of migration on the previously discussed disparities in rural
hospitals. Additionally, there is an interrelationship between the economies of
communities and their health systems (Stauber, 2001). A recognition of the dependencies
of communities and systems can be useful for both recognizing, not only the functional
areas of dependence, but also the degree and sub-areas of the dependence (FigueriaMcDonough, 2001). "A community engaged in boundary exchange can therefore
increase its capacity to mobilize resources. A locality fully dependent on the outside can
lose its ability to function as a community. Necessary bridges across boundaries can exist
at informal and formal levels." (Figueria-McDonough, 2001, p.24).
Hospital system outcomes can include hospital debt ratios, satisfaction, or
readmission rates (Shaw & Miller, 2000). System outcomes such as number of certain
procedures performed and results of JCAHO evaluations are also available from both the
Joint Commission on Accreditation (JCAHO) and the American Hospital Association
(AHA) (Basu & Cooper 2000). The study framework can measure many of these
outcomes using publicly available JCAHO indicators and AHA measures to compare
systems.
Risk Adjusted Patient Outcomes
Quality of care is defined as the "degree to which health services increa.se the
likelihood of a desired health outcome" (Shaw & Miller, 2000, p 46). Patient outcomes
are the result of their care and can include patient perceptions and the clinical result of
care (Donabedian, 1966). Patient outcomes are also recognized as a surrogate measure of
health care quality (Shaw & Miller, 2000). Patient satisfaction, length of stay, variations
in outcome using risk analysis, and readmission data measures are included in this
framework and can be used to compare quality of care.
Risk adjustment is necessary to compare health systems meaningfully by
standardizing patient populations. There are a number of risk adjustment methods
available for use in health systems research (lezzoni, 2003). Risk adjustment allows for
an increased understanding of the impact of variables of interest on the outcome while
adjusting for the variations in the type of patients studied. Selection of a method for risk
adjustment in this framework depends on the sampling and methodology used for data
collection.
Donabedian (1966) also discusses optimally and maximally effective care as
economically based outcomes measures of quality. Optimally effective care is defined as
outcomes that can be obtained if unlimited resources were available. Maximally effective
care is defined as that which is the most effective and efficient, or best practice, within
the constraints of the resources that are available. In Donabedian's framework, cost is a
construct which is used to better evaluate the outcomes. Cost analysis techniques can be
applied to any variations in the outcomes in this study's framework and can add a level of
outcome analysis for the development of effective and efficient interventions. Again, the
choice of technique is based on the sample and methods used in measuring the outcomes
of interest.
Rural Community Outcomes
Community outcome is a concept which needs to be developed and evaluated. As
Stauber stated, "Economic growth alone is not enough to sustain communities. It is
necessary, but not sufficient. Communities that survive and prosper also invest in
building the social and human capital of their institutions and people" {2001, p 70).
Donabedian included community as a level at which quality could be assessed for a
system (1966). While research has demonstrated that patients who migrate usually have
the economic means to choose (Basu & Cooper, 2000), the impact that choice has on the
rural community has yet to be studied. Current econometric models to measure the
impact of migration on human capital ajid communities, used in the social sciences, may
be appropriate for analyzing this variable in health systems research.
Summary
In conclusion, a multilevel framework of migration for care was developed to
explore the impact of distance on inpatient outcomes for rural patients. Methods exist for
the evaluation of systems and patient outcomes and the identification of variables related
to traveling for care. What has yet to be understood is the relationship between distance
and patient outcomes and the impact that travel, or migration, has on patients, health
systems and communities. This system organizing framework may allow for an improved
understanding of factors that influence health outcomes for rural patients.
An understanding of the factors related to and the patterns of, travel, or migration,
may assist with the development of improved health policies to increase both the equity
and efficiency of care designed for rural patients that must travel. This exploratory study
39
will use the franievvork to organize a secondary analysis of a statewide hospital discharge
data base in an attempt to understand the relationship between distance and risk adjusted
patient outcomes.
40
CHAPTER THREE
Methodology
Design
This descriptive, exploratory study seeks to identify relationships between
distance and risk adjusted patient outcomes through a secondary analysis of the AzDHS
Hospital Discharge Data Base. Variables in the database were organized using the
described study model. Figure 2 displays the framework with the variables available in
the data for this study in bold. This study explored the relationship between patient
characteristics; distance traveled, and risk adjusted patient outcomes.
Figure 2: Study Model
OUTCOME
Rural System Capacity
Urban System
Capacity
Rural System
Outcomes
Distance Traveled
Urban System
Outcomes
Patient Decision
Risk Adjusted
Patient Outcomes
Patient
Cliaracteristics
Rural Community
Outcomes
Data Source
The AzDHS Hospital Discharge Public Data Base was the source of data for this
study. One year of data, July 2001 - June 2002, was acquired from the Bureau of Health
Statistics at AzDHS. The publicly available de-identified database is available for
purchase from the AzDHS in 6 month increments. One full year of data was selected for
this study due to the seasonal variations of the Arizona population. The data were kept in
two 6 month increments for analysis to allow for recognition of any potential seasonal
differences in the data (July-December, 2001 and January-June, 2002). These two
databases were titled Time One and Time Two, respectively.
The AzDHS data base includes 113 vajiables ranging from, hospital and physician
information to demographic information, number of diagnoses and type of treatment for
each patient record (Appendix A). The public database does not include patient name,
social security number, date of birth, and admission or discharge dates. As a result,
readmission outcomes and any longitudinal tracking of individual patients were not
possible for this study. Table 1 displays the organization of the available variables in
relation to the constructs of the study. Chapter 4 will discuss in detail the selection and
conversion of the variables into the constructs of the study framework.
Table 1: Organization of AzDHS Variables
MODEL CONSTRUCT
Patient Characteristics: Access
Patient Characteristics: Need
Patient Characteristics; Risk Adj.
Distance Traveled
Patient Outcome - Quality
Patient Outcome - Cost
AzDHS VARIABLE
payer, admit type and source
type of service, DRG, therapeutic and diagnostic
services, secondary diagnosis codes
Age
hospital and patient zip code in miles.
procedure codes, length of stay
total charges
42
Sample and Unit of Analysis
The sample includes all inpatient records from July 2001 to June 2002 in the
discharge database that had an Arizona zip code in the patient record. To accurately
measure rural patients who must migrate for care, records from patients with zip codes
that lie within thirty miles of hospitals within thirty miles of each other were excluded
from the analysis. The unit of analysis was at group level for question one and the
individual level for question two. Records were initially analyzed at the individual patient
level then aggregated to the group level based on the distribution of the distance variable.
Primary Data Collection
All hospitals licensed by the state must submit discharge records for every
inpatient to AzDHS. These records are then compiled into the publicly available
discharge database. Each licensed institution is required to code each discharge record
into the database according to specifications established by the AzDHS (Appendix A).
This coding occurs at each hospital and is then electronically transmitted to the state
according to Arizona Revised Statute 36-125.05.
In Arizona, seventy two hospitals report to the state. This number does not
include federal institutions such as the Veteran's Affairs or Indian Health Services.
Annual descriptive analysis of the database has been completed by the AzDHS Bureau of
Health Statistics. The agency is required to ensure receipt, verification and distribution of
the data.
Reliability of Primary Data
Prior to 1990, studies of coding accuracy found a number of variations in the
accuracy for hospital discharge records (lezzoni, 2003). In a study of the reliability of
discharge data, lezzoni (1997) compared risk adjusted diagnosis codes with admission
clinical findings available in hospital records and reported that risk analysis methods
were equal to reviewing admission clinical findings in their ability to predict hospital
mortality (lezzoni, 1997). In addition, Medicare's compliance movement and legal
prosecution of health systems have led to improved coding practices at hospitals. The
posting of the Medicare Case Mix Index (CM!) on a public website has also resulted in
the increased assumption of reliability of large hospital data bases (lezzoni, 2003). When
using a large database, such as the one in this study, decisions must be made during the
secondary analysis when inconsistent patterns are recognized in the data to ensure
reliability of the results of the study.
Secondary Data Analysis
The steps in secondary data analysis of such a large database must be clearly
defined to allow for guidance and consistency throughout the analysis of the information
(Wray, Ashton, Juykendall, & Hollingsworth, 1995). For this study, the following steps
were followed and will be discussed in detail in chapter four:
1) Hospital identification numbers were converted into a zip code for each institution.
2) All patient records with a zip code from within 30 miles of hospitals that are within 30
miles of each other and records from states not bordering Arizona were removed in order
to focus the study on rural patients who access Arizona ho.spitals.
3) Also removed were any records with inaccurate zip codes such as all zeros or fewer
than 5 numbers.
4) Using Geographic Information Software (GIS) a distance variable was created by
measuring the distance between the hospital and patient zip code in each record.
5) Due to the large variation of distances (ranging from 31 to over 800 miles), frequency
analysis of the distance variable was used to create groups based on mileage in order to
be able to analyze variances between groups that must travel different distances.
6) Once grouped, inaccuracies within the data were able to be identified though
frequency analysis and the data cleaning was then completed.
7) The records were then analyzed to identify which risk adjustment method could be
used from the techniques described by lezzoni (2003) and used in a number of outcomes
studies. These methods frequently use a form of "patient-level logistic regression analysis
to predict each patient's probability of having each adverse outcome" (Needleman, et al,
2002, p.1716).
8) The risk adjusted records were then analyzed with distance as the dependent variable
and each outcome as the independent variable.
9) Differences in outcomes were analyzed to identify factors within the database which
had significant relationships with the outcome variations.
10) Cost analysis was used to compare any variations in outcomes.
45
Decision rules
When analyzing a large database, many of the questions which can affect the
validity of a study do not arise until during the analysis of data. Therefore, it is important
that a framework be developed for decision making throughout the process to ensure
content validity of the results. For this study, all data decisions were documented during
the analysis and included data decisions, analysis decisions, and supporting rationale.
This process is discussed in the following chapter.
The following rules, developed by Wray et al (1995) were followed throughout
the analyses when patient, medical care or outcome variations were identified. An
example is being given with each step:
1. A plausible link should exist between quality of care and frequency of
outcome.
e.g.: Variations in outcomes in which a link is not found, such as a
disproportionate number of urinary track infections for one facility may result in
not using that outcome for further analysis if no plausible link can be found.
2. Types of care which conform to acceptable practice standards but still lead to
variation in the outcome should be excluded.
e.g.: Variations which can be based on services such as higher mortality in a
hospice setting.
3. Outcomes should be prevalent.
e.g.; Outcomes which occur in 5% or less of the cases should not be used.
4. Constraints of coding such that only those disease-outcome pairs least affected
by limitations are selected for analysis.
e.g.: Some surgical admissions are planned readmissions (carpal tunnel release)
and should not be included in readmission data.
46
5. Constraints of the structure of the database should be considered throughout
analysis.
e.g.: Knowledge of the data base is required to avoid inappropriate analysis
decisions such as distinguishing which conditions are comorbidities and which are
complications in a discharge database.
(Wray et al, 1995)
To ensure consistency in the decision making process, regular communication of data
decisions with the dissertation committee occurred throughout the data analysis.
Disagreements were settled by consensus.
Cost Analysis
Cost analysis was completed for each variation in outcomes using methods currently
accepted in the literature (Muenning, 2002). A discussion of the methods used and results
of this analysis can be found in Chapter 5.
Human Subjects
The database used was de-identified and publicly available from AzDHS. Human
subject review through the University of Arizona was obtained (Appendix B).
Summary
This exploratory research was designed to describe the relationship between
distance and risk adjusted patient outcomes and to evaluate factors that are related to
patient migration for care using a systems framework. Cost analysis measures were used
to develop an. increased understanding of the impact of migration for health care on
system and patient costs. In the next chapter there is a detailed discussion of the decisions
related to the analysis of the data and the creation of variables for this study.
CHAPTER FOUR
Creating the Analytic Data File
The purpose of this study was to explore the factors related to patient travel and to
understand the relationship between distances traveled and risk adjusted patient
outcomes. This chapter describes the four main processes used to create a data file in
order to answer the research questions:
• Examination of the scope of the data in order to determine which data elements
contained in the set could be matched to concepts in the theoretical model.
• Evaluation of the reliability and validity of the data through review of literature
and discussion with experts.
• Development of a recoding data plan, which included assessment of type of
statistical analysis to be completed.
• Cleaning of data, including creation and grouping of variables for each research
question.
This chapter will discuss each of the data processes for each construct in the study model,
beginning with the patient characteristics, followed by the distance measure and finishing
with the risk adjusted outcomes. All analysis was completed using SPSS 10.0 analytical
software. Finally, the decision rules established by Wray et al (1995) and their use in this
study will be addressed at the end of this chapter. A complete data coding dictionary is
available (Appendix C).
The process of developing a database is primarily driven by the variable of
interest. In this study, the concept is distance. While the steps in creating the data file for
this study began with creating the distance variable and reducing the number of cases,
this chapter will be organized according to the study framework, rather than according to
the chronological sequence of events. All the steps used were replicated for both the
Time One and Time Two databases to ensure both data files would have the same
variables for comparative analysis.
Additionally, it is important to discuss the process of data analysis from a
theoretical standpoint. Informatics literature discusses association and coding rules and a
variety of statistical methods that can be used in data mining (Brossette et al 1998). From
a theoretical perspective, the process of analysis is more of a synergistic relationship
between the theoretical framework, statistical rules and the content of the database.
Categorization and reduction may be necessary to manage large databases in order to
clearly identify patterns in the data. In this study, strategies used to balance the
framework with the type and scope of the database included reduction of the sample
through exclusion criteria, reduction of the number of variables through matching the
framework and data set, and the development of categories after preliminary frequency
analysis of variables to enhance the interpretation of the data analysis. These empirical
strategies, guided by the decision rules from Wray et al (1995), were used throughout the
creation of the data file.
Patient Characteristics
During initial examination of the databases for this exploratory study, it was
recognized that over 80 potential patient characteristic variables existed. With over
20,000 cases, analysis of 80 variables would yield results which would be difficult to
interpret. It has been recognized that the factors related to hospital utilization are based in
either patient access or medical need (Borders et al, 2000). Consensus was reached to
separate the variables into two groups, the first group focused on access variables which
included payer, type of adnait and source of admit. Hospital size was added to the access
characteristics in order to understand, at the patient level, the type of hospital utilized by
patients who travel. The other characteristic group focused on medical need based
variables which included the diagnosis related group (DRG), type of service, the number
of therapeutic and diagnostic services, and the number of secondary diagnoses. The
process for developing the variables for each of the characteristic groups will be
discussed separately, beginning with the access variables.
Patient Access Characteristics
A number of decisions were needed in the process of creating the patient access
characteristic variables. A preliminary examination of the data allowed for organizational
decisions to be m.ade with relation to the fit of the available variables with the theoretical
model. The variables which were finally used to represent the patient access
characteristics in the model included number of beds, payer, admit type and admit source.
When examining the AzDHS Hospital Discharge Data Base for hospital type, the
results included each hospital's identification code and county of location as the only
identification for the hospital in each case. A separate file contained the identification
code and the matching name of each hospital, but no information on hospital size,
teaching status, or physical address was included (Appendix D). An evaluation of
literature for hospital comparison variables to include found hospital size, teaching status
and urban/rural designation had been recognized as reliable methods for comparing
hospital systems (Thomas et al, 1998).
While the number of beds, a proxy for hospital size, was not in the state data, it
was available through the Arizona Hospital and Health Care Association (AzHHA,
2004). Hospital and Health Care Association data has been recognized as valid in other
studies ( Aiken, 2002). Teaching status was not readily available for the hospitals in this
system. The distance measure analysis (discussed in next section) demonstrated that only
three of the seventy two hospitals were not within 30 miles of other hospitals, meeting the
definition of rural for this study. As a result, the urban/rural designation was not used due
to the small portion of hospitals with valid identification as rural.
Hospital size was created using information on the number of beds from AzHHA
and matched with the identification codes and hospitals in the AzDHS data. The number
of beds, which ranged from 25 to over 450, was then entered into the two study
databases. To assist in the interpretation of results, a frequency analysis was used to
develop an ordinal variable of bed groups. Six bed groups, ranging from less than 50 to
over 450 beds, were created by recoding the number of beds in both databases.
The payer variables in the original databases included fifteen different codes
identifying the primary payer for the majority of charges for each case. These codes
included private insurers which included HMO and PPO, state and national programs
such as AHCCCS and Champus, Medicare, self pay and charity pay. An evaluation of
methods used for other distance studies demonstrated that payer was typically sorted by
private. Medicaid and Medicare (Basu & Cooper, 2000; Borders ct al, 2000). The
Arizona Health Care Cost Containment System (AHCCCS) is the form of Medicaid for
Arizona residents.
To address one of the research questions on the relationship between distance and
risk adjusted patient outcomes, regression analysis would be required. The variables for
payer were categorical variables in the original databases and were recoded into separate
dummy variables in order to be used in regression analysis. Frequency analysis of payer
in both databases found the majority of eases were in one of the top three: Private
insurance (19-29%), AHCCCS (23-29%) and Medicare (35-36%). Payer was converted
into three different dummy variables with codes of one representing either private
insurance, AHCCCS or Medicare. For all other payer variables, the cases were coded
zero. Therefore, in the final analytic data file, there were three payer variables.
In the AzDHS database, admit type included codes identifying emergency, urgent
care, newborn, elective, observation and information not available. The emergency and
urgent care admits contained over 70 percent of the cases in both databases. Previous
studies have focused on emergency admit type and transfer admit source (Basu &
Cooper, 2000). Admit type was recoded as a dummy variable when 1 represented
emergency and urgent admissions and 0 represented all other types.
The admit source included nine codes, which included referrals, transfers,
emergency admission, legal admit, and information not available. The referral and
transfer sources contained over 50 percent of the cases. Admit source was recoded into
another dummy variable when 1 represented referral and transfers and 0 represented all
other admits, the majority of which (90%) were emergency admissions. Additionally, in
the admit source codes, a number of cases (12%) were coded as newborn in both
databases. For each newborn case, there were different definitions of the codes for a
number of the other variables. To analyze the variables in the newborn cases, they would
need to be separated from both study databases. The rationale for the exclusion of
newborn cases was due to the focus of this study on distances relationship to medical
admissions and outcomes. Consensus was reached to delete all newborn cases from both
databases to avoid misinterpretation of the results.
in summary, the patient access characteristic variables used in this study included
hospital size and three payer variables; Medicare, private insurance, and AHCCCS. The
other access variables created were emergency/urgent care admission type and
referral/transfer admission source.
Patient Need Characteristics
When examining the databases for the medical need characteristics, it was
recognized that both Diagnostic Related Groups (DRG) and ICD-9 codes were used. The
numeric DRG code was defined as the primary diagnosis code used as the reason for
admission in the databases. The ICD-9 diagnosis codes were used to describe the major
conditions related to the incident hospital admission. There was one DRG code and nine
possible ICD-9 codes for each case. An evaluation of studies found the most frequently
occurring DRGs (Thomson et al 1998) and numbers of comorbidities, or secondary
diagnosis, have frequently been used to identify the types of populations served (Zhan &
Miller, 2003).
One way of grouping the number of DRGs is to use Medicare's Major Diagnostic
Category system which contains twenty five categories. An initial frequency analysis of
the DRGs in both databases found a very wide variation which would inhibit
interpretation. Frequency analysis was then used to identify the most frequently occurring
disease groups. The analyses were reviewed and grouping of DRGs were discussed until
consensus was reached to create six balanced groupings. Due to the previously discussed
deletions of newborn cases, the cases with newborn DRGs (385-391) were also deleted
from the study. The final six groups included in the study were: Neuro/Ortho/Trauma,
Pulmonary, Cardiac, GI/GU, Women's Health, and Miscellaneous Surgery/Skin/Burns
(see Appendix C). In order to be used in the analysis, each of the numeric DRGs was then
recoded into six new dummy variables where 1 representing one of the six DRG groups
and 0 represented all the other DRGs in the study. Therefore, there were six DRG
variables in the final analytic file.
The secondary diagnoses were in both databases as eight separate fields
containing ICD-9 codes. ICD-9 codes included strings of letters and numbers and would
need to be converted to be used in statistical analysis. The primary diagnosis was not
used in this study as the DRG groups were used instead and the two have been found to
be equally valid diagnostic variables (Thomas, et al, 1998). Each secondary diagnosis
field was converted by sorting the cases in ascending order and using 1 to replace any
secondary diagnosis and zero in any blank fields. The secondarj' diagnosis fields were
then converted into a Microsoft Excel file and the number of secondary diagnoses were
totaled for each case. The totals were then transferred into the study databases to produce
a continuous variable with a range from zero to eight.
An examination of the service codes found they were all charges in dollars per
case for ten separate types of service codes which ranged from intensive care to room and
board in a ward. There were also over sixty therapeutic and diagnostic service codes such
as laboratory diagnostics and respiratory therapy which were also entered as charges in
the databases. No studies were found that utilized specific therapeutic and diagnostic
service revenue codes.
The type of service codes were recoded into a dummy variable with a code of one
representing intensive and critical care (31% for Time One and 29% for Time Two) and
all the other services were coded zero. The remaining revenue code variables consisting
of therapeutic and diagnostic services were recoded with 1 representing any charges
entered for any services rendered. The diagnostic and therapeutic services were then
totaled for each case. These totals created continuous variables for use in the analysis of
data with up to fifteen possible tlierapeutic services and eleven possible diagnostic
services. In summary, the final patient access characteristics included DRG group,
number of secondary diagnosis, type of service with 1 representing critical care, and the
number of therapeutic and diagnostic services.
Distance Traveled
An initial examination of the databases in order to create a distance variable found
fields for hospital identification codes, and the hospital and patient zip codes. As
previously discussed, there was also a separate file which identified each hospital name
with the matching identification code (Appendix D). All the hospital zip codes were
verified using AzDHS and AzHHA infonnation. Some patient records contained invalid
zip codes, such as codes with less than five digits, and some records were missing zip
codes. Also included in the data base were zip codes entered as MEXICO and zip codes
from states that did not border Arizona, which represented patients visiting the state and
not necessarily traveling for health care. To create a reliable distance variable, cases with
an inaccurate or missing zip code were deleted as were those entered as MEXICO and
those from states not bordering Arizona. The cleaned database included over 600,000
cases.
This study defined rural hospitals as those more than thirty miles from any other
hospital. The population for this study was those patients who did not live within 30
miles of hospitals that were within 30 miles of each other. Local experts were consulted
on the use of GIS for this study. The distance data analysis was completed by the School
of Renewable and Natural Resources (SRNR) at the University of Arizona (Appendix E
contains the GIS syntax). Using the information obtained by the SRNR, the next section
will discuss how the distance variable and the rural study population were establishedMeasure of distance and travel in health care literature typically use one of two
distance measures; crossing county lines (Basu & Cooper, 2000) or zip code centroid
analysis (Mooney et al, 2000). Centroid analysis measures the distance between two zip
codes using Geographic Interface Software (GIS). The centroid is the geographic center
of the zip code area. Arizona contains counties that are geographically some of the largest
in the country. Using county lines as a measure, when counties range in size from 1847 to
18,608 square miles, would lead to an unbalanced and inaccurate sample. Zip code
centroid analysis creates distance variables using actual distances which allows for a
balanced standard measure (miles) for each case.
The hospitals in the databases were mapped by their physical address to identify
those systems that were within 30 miles of each other. Next, the patient zip codes within
those 30 mile hospital clusters were identified. After receiving the identified patient zip
codes, those cases with zip codes lying within 30 miles of the hospital were deleted from
the study data base resulting in over 57,000 cases remaining. Figure 3 displays the map of
the counties in Arizona with hospital clusters and patient zip codes. The smaller and
lighter black dots are patient zip code centroids. or the center of the zip code area for each
case. The large shaded circles are 30 mile radius around each hospital with a larger black
dot in the center of each representing the location of each hospital. Urban concentrations
result in a compressed group of hospital and zip codes. All zip codes lying outside of a
shaded circle are easier to recognize and meet the rural definition for this study
It should be noted that the urban clu.ster of hospitals does not follow county lines.
This demonstrates one reason why using county based MSA's may not be appropriate for
evaluating health care geographically. Only three hospitals in the state were not within 30
miles of another; Page and Sage hospitals, each with 25 beds in the North, and Yuma
Regional Medical Center in the South with 276 beds. Also, some of the hospitals in the
study are located near border areas of the state where some patient zip code clusters
occur, leading to the decision to include the cases with patient zip codes from bordering
states Arizona in the study population; California, Nevada, Utah and New Mexico
57
Figure 3: Map of Arizona with hospital clusters and patient zip codes:
Legend
Zipcode Caitroids
•
iH
Hospital
30-MIlebuffa-
After deleting the urban zip codes identified by the mapping process (those zip
codes in the shaded circles), the remaining 'niral' cases were again input into GIS to
create a distance variable of the mileage between the hospital location and patient zip
code centroid. This variable was then transferred into the analysis databases. Frequencies
were run on the new distance variable in the databases and ranged from 31 to over 800
miles with seventy five percent of the cases traveling 200 miles or less. Again,
interpreting an analysis using this wide range of mileage could be difficult, at best.
In order to decide how to group the distance measure, a search of the health
systems literature and discussions with experts in health care geography was undertaken.
No studies, other than Medicare reports from the late 1990's, have used distances over
100 miles (T. Ricketts. personal communication, 2004). Most studies for distances less
thanlOO miles frequently use ten mile groupings, which would not yield improvements in
the analysis for this study (Basu & Cooper, 2001). Using that information, committee
consensus was reached and the distance variable was recoded to create four mileage
categories. The final four mileage groups used in the analysis were 31-100 miles, 101200 miles, 201-300 miles, and 301 and greater miles. Actual distance remained in the
databases and was used when appropriate for analysis of specific research questions (see
chapter 5).
In summary, the development of the distance variable required a number of steps
including the mapping of patient zip code centroids to identify urban cases which were
then deleted from tlie database. A distance variable was then created using the mileage
between each remaining rural case's zip code centroid and the physical address of the
treating hospital. The resulting range of distances were also recoded into four mileage
groups ranging from less than 100 miles to over 300 miles to increase the capabilities for
interpretation of each database.
Risk Adjusted Patient Outcomes
One of the more complex issues in the literature regarding large databases is how
to risk adjust the population in order to understand patient outcomes (lezzoni, 2003). The
AzDHS database contained a number of variables that could be used for risk adjustment,
including the source and type of admission, DRGs, and 89 revenue codes which include
the type of service, such as intensive care or nursery, and a large range of therapeutic and
diagnostic services for each case. Also included were nine ICD-9 diagnosis codes and 6
ICD -9 procedure codcs. Finally, age, discharge status and length of stay were available.
A literature search was conducted to decide on a method for risk adjustment.
The Charlson index (CHI) was introduced in 1987 as a way to risk adjust patient
populations for mortality using a numeric scale which assigned weight to different
disease entities. Since 1987, researchers have recognized the limitations of CHI and have
continued to adapt the CHI for better predictability of mortality. (Librero, Peiro &
Ordinana, 1998). Other researchers have worked on developing multilevel risk models
using newer computer modeling technology to evaluate the impact of comorbities on
length of stay (Thomas et al, 1998).
Recent studies of nurse staffing have developed methods to analyze risk adjusted
patient mortality, failure to rescue (Aiken et al, 2002) and rates of complications such as
urinary tract infections and upper gastrointestinal bleeding (Needleman et al, 2002). By
using patient level logistic regression they are able to predict a patient's probability for
each of the nursing focused outcomes. These studies use both discharge databases and
nurse staffing and state board data to evaluate the impact of nurse staffing on patient
outcomes.
Finally, agencies are beginning to use the Patient Safety Indicators from the
Agency for Health Care Research and Quality (AHRQ) to risk adjust outcomes used to
evaluate quality of care (www.qualityindicators.ahrq.gov). The AHRQ has software
publicly available for evaluating patient safety indicators that requires specific coding of
the discharge data. The coding requirements were quite different from that which was
available in the study databases.
In recognition of the numerous complex algorithms used in risk adjustment, one
recent study reviewed six different methods of risk adjusting for comorbidity and
mortality and found none predicted better than simply using the age of the patient
(Schneeweiss & Maclure, 2000). The authors did state that some of the limitations were
related to the variety of coding errors that were found in the large databases used in the
study.
Adding to the complexity of choosing a method for risk adjustment is the wide
variety of software, programming and coding used in creating patient level regressions.
All the studies mentioned agreed that until an acceptable standardized method is found,
there continues to be a need to create a method for risk adjustment based on the
information and questions asked for every study (Roos, Stranc, James, Li, 1997).
No single method was found to create a 'gold standard' for risk adjustment.
Because this study was of an exploratory nature, and after discussion with experts,
consensus was reached that age would be appropriate to use as the risk adjustment
variable for this study. A frequency analysis of age resulted in ages ranging from zero to
102. Any newborns (age = 0) were deleted from the data base to remain consistent. Age
was then recoded into six ordinal groups ranging from under twenty years to greater than
seventy six.
Outcomes
The outcomes found in the databases included procedures codes, length of stay,
and discharge status. Discharge status, which included the expired or mortality variable,
was not used as an outcome for this study due to the fact that over eighty percent of the
cases were discharged to self care and less than five percent of the study population had
an expired code. It was recognized that outcomes needed to be prevalent to be used in the
study (Wray et al, 1995) and discharge status did not meet the criteria. All cases with an
expired discharge code were deleted from the study population to avoid confounding the
data. The remaining discharge variables were not included in analyses.
Procedures were originally coded in the database using the lCD-9 procedure
classifications. These codes were recoded in a process similar to that used for secondary
diagnosis. This resulted in six procedure fields with a code of one representing all
procedure codes. The data were then converted into a Microsoft Excel file where the total
number of procedures were calculated for each case, with a range from zero to six. Those
totals were then converted back to the original database with a range from zero to six and
used for all outcomes analysis. Finally, frequencies were run on length of stay which
ranged from 0 to 276 days. Over ninety percent of the cases had a length of stay of less
than ten days in both databases. Length of stay was recoded into a continuous variable
ranging from zero to eleven or more days to improve interpretation of the data.
Both number of procedures and length of stay were used for the outcomes in this
study with age as the risk adjuster. Cost analysis decisions were not made until the results
of the statistical analysis demonstrated significant outcomes in either database. A
discussion of the cost analysis completed for this study is presented in Chapter Five.
Decision rules
For this study, all decisions were documented during the analysis and included
data decisions, analysis decisions and supporting rationale. Processes included
examination and evaluation of the data, utilization of literature and experts, and the
creation of the variables for each construct. Decisions included a balance between the
theoretical framework and the type and scope of data available. The following rules,
developed by Wray et al (1995) were followed throughout the analyses. Each rule is
followed by an explanation of decisions made during the development of the analytic
data file.
1. A plausible link should exist between quality of care and frequency of
outcome.
Outcomes selected for this study included the number of procedures and length of
stay which were risk adjusted using ages. Variations in outcome were analyzed
using cost analysis methods.
63
2. Types of care which conform to acceptable practice standards but still lead to
variation in the outcomc should be excluded.
Recoding the service variables and using the number of services and not the
type of services avoided the need to evaluate acceptable practice standards
because the total number of services were used for comparison.
3. Outcomes should be prevalent.
Discharge status was eliminated from the outcomes for this study due to lack of
variation and small frequency of mortality (<5%). The remaining outcomes,
number of procedures and length of stay were prevalent in both
databases.
4. Constraints of coding such that only those disease-outcome pairs least affected
by limitations are selected for analysis.
Frequency of diagnosis groups were analyzed for the most frequent DRGs and
six diagnosis groups were created. Newborn cases were deleted due to limitations
of analyses with different coding rules than the rest of the population.
5. Constraints of the structure of the database should be considered throughout
analysis.
Expert consultation, literature review and committee discussion were used
throughout the decision process for the development of the analytic data file.
Summary
This chapter discussed the creation of the analytic data file for the constructs in
study model; patient characteristics, distance traveled and risk adjusted patient outcomes.
Patient characteristics include access variables of number of beds, three payer variables,
emergency admit type and referral/transfer admit source. Patient need characteristics
include DRG group, critical care type of service, number of secondary diagnosis and the
number of therapeutic and diagnostic services. The outcomes selected were number of
procedures and length of stay with age as risk adjuster. The next chapter will include the
64
results of statistical analysis using the variables created for the analytic data file in order
to answer the research questions and a discussion of the cost analysis.
CHAPTER FIVE
Results
This exploratory study analyzed data obtained from the Arizona Department of
Health Services (AzDHS) publicly available hospital discharge database. There were two
research questions:
1) What are the demographic factors that influence distance traveled for rural
patients?
2) What is the relationship between distance and risk adjusted inpatient health
outcomes for rural patients?
This chapter will begin with a discussion of the description of the sample followed by the
results of the analysis for each question. This is followed by a description of the process
for cost analysis and the results of the cost analysis of length of stay outliers in this study.
Description of Sample
All records which met the criteria for mrality for this study were included: those
patient records remaining after excluding all records with zip codes within 30 miles of
hospitals which were within 30 miles of another hospital in the database. Newborn and
expired cases were excluded. The cases obtained from AzDHS were in two six month
databases recorded from July to December 2001 (N= 12,746) and January to June 2002
(N=8180). The databases were analyzed separately to aid in recognition of possible
seasonal population differences which occur in many areas of Arizona and were
identified as Time One and Time Two, respectively. More patients in both databases had
emergency admissions, transfer and referral sources, and a fonn of payment from
66
Medicare, private insurance or AHCCCS. Fewer patients in both databases had critical
care services, more than four secondary diagnoses or diagnostic services. Most of the
patients were less than 60 years olds and received fewer than three procedures and stayed
less than five days. The patient characteristic variables for both databases are displayed in
the following tables. Table 2 displays the frequencies and percent of the totals for the
access characteristic variables for Time One and Time Two.
Table 2: Frequencies of Patient Access Characteristics
Variable
Hospital Size
(# of beds)
<=50
51-150
151-250
251-350
351-450
451=>
Admit type
ER/Urgent Care
Other
Admit Source
Transfer/Referrals
ER/other
Payer
Medicare
Private Insurance
AHCCCS
Other
Time One n
Time One %
Time Two n
Time Two %
1137
4688
1938
2567
436
1980
9
37
15
20
3
16
1975
845
1103
2481
382
1394
14
30
5
17
9459
3284
74
26
6110
2070
74
26
7402
5244
58
42
4635
3545
57
43
4465
3690
2935
1656
35
29
23
13
2978
1547
2405
1250
36
19
29
15
24
in
i. V/
The first difference between the two data bases is noticed in the hospital size
variable. In Time One, 9% of the cases were admitted to hospitals with less than 50 beds,
while almost one forth of the cases in Time Two were admitted to small hospitals. Time
One had the highest frequency in the hospital size of between 51 and 150 beds while
Time Two's highest frequency was for 251 to 350 beds.
Admit type demonstrated almost three quarters of the cases in both databases
were emergency or urgent care admits. Admit source also had similar distributions in
both databases with over 50% of the cases in both databases in transfers and referrals.
Medicare was the primary payer in both databases with 35% of the cases in Time One
and 36% of the cases Time Two. There were differences in the private insurance with
Time One having 29% and Time Two with 19%. Finally, AHCCCS also was different
between the two databases with 23% in Time One and 30% in Time Two.
Table 3 displays one of the patient need characteristics, the DRG groups, for the
two databases. The table includes the actual DRG codes used for the creation of the
groups and the frequency for each group. The top six groups are identified using an
asterisk by the group name.
Table 3: Fregucncv of DRG groups
DRG Group Name
Neuro/Ortho/Trauma *
DRG codes
1-35, 209-256,
439-446
Ear Nose & Throat
36-74
Pulmonary *
75-102
Cardiac *
103-145
GI/GU*
146-190, 302-352
Hepatic/Biliary/Endocrine 191-208, 295-301,
392-393
Women's Health *
353-384
CA/RBC/Infections
394-423
MH/Detox
424-438, 447-460
Misc Surg/Skin/Bums *
257-284, 461-511
Totals:
Time one
%
n
2254 18%
178 1%
944 7%
2183 17%
1907 15%
Time two
n
%
1368
126
1000
1229
1209
17%
2%
12%
15%
15%
535
732 6%
7%
1761 13% 1001 12%
452 4%
3%
282
694 5%
402
5%
1751 14% 1 994 12%
12746 100% 1 8180 100%
The top six DRG groups included the Neuro/Ortho/Trauma group with the highest
frequency of cases in both databases, 18% in Time One and 17% in Time Two. Next was
the Cardiac group with 17% of the Time One cases and 15% of the Time Two cases. The
GI/GU group was third with 15 % of the cases in both databases. The Women's Health
group had 13% of Time One and 12% of Time Two cases while the Miscellaneous group
had 14% of Time One cases and 12% of Time Two cases. Finally, the Pulmonary group
had the smallest number of cases in the top six DRG groups, 7% in Time One and 12% in
Time Two. In both databases, the frequencies of the DRGs displayed similar patterns
with each of the DRG groups containing less than twenty percent of the cases. Four DRG
groups were not included in further analysis, all of which contained less than 10% of the
cases in both databases; the ear nose and throat group, the hepatic/biliary/endocrine
group, the cancer/rbc/infection group, and the mental health/detox group.
Table 4 displays the frequency of the other patient need characteristics used in
this study. For the three continuous variables, number of secondary diagnosis, number of
therapeutic services, and number of diagnostic services, the table displays the number of
patient cases which have totals greater or fewer than five for each variable. Both Time
One and Time Two had fewer cases which received critical care services, 31% in Time
One and 28% in Time Two. The majority of cases in both databases also had four or
fewer secondary diagnosis {51%). More cases in Time One had between five and eleven
therapeutic services (51%) while more cases in Time Two had four or fewer (53%). Both
databases had over 60% of the cases receiving four or fewer diagnostic services. Both
databases look similar in the majority of the need characteristics, with the exception of
the number of therapeutic services variable.
Table 4: Frequencv of Patient Need Characteristics
Variable
of Service
ICU/CCU
Other
Number of 2"'^ Diagnosis
0-4
5-8
Number of Therapies
0-4
5-11
Number of Diagnostics
0-4
5-11
Time One n
Time One % 1 Time Two n Time Two %
Type
3933
8813
31
69
7276
5470
57
43
6303
6443
49
51
8225
4521
65
35
1
1
1
2270
5910
28
72
4681
3499
57
43
4355
3825
53
47
5330
2850
65
35
Table 5 displays the frequencies for each of the four mileage groups. For both
databases, the majority of cases travel fewer than 200 miles, but variations can be seen
between the two databases. Time One had the greater frequency of the cases (47%) in the
101-200 mile group while Time Two had the greater frequency of cases (48%) in the 31100 mile group. Time One had the fewest cases (9%) in the 201-300 mile group while
Time Two had an equal number of cases (12%) in both the 201 -300 and 301 or greater
mileage groups.
Table 5: Frequencv of Mileage groups
Mileage Group
31-100
101-200
201-300
301=>
Time One n
4123
6040
1198
1253
Time One %
33%
47%
9%
11%
Time Two n
3964
2260
988
941
Time Two %
48%
28%
12%
12%
The last table to describe the sample is Table 6, which displays distribution of the
age groups and outcome variables for this study. The frequency patterns for age were
consistent across both databases as was the distribution of the outcome variables.
Table 6: Frequency of Age Groups and Outcome Variables
Variable
Age Groups
1-20
21-40
41-60
61-75
76=>
Outcomes
Number of Procedures
0-3
4-6
Length of stay
0-5
6-11=>
Time One n
Time One %
Time Two n
Time Two %
1441
2597
3033
3402
2273
11%
20%
24%
27%
18%
1046
1687
2045
1964
1438
14%
21%
25%
24%
18%
11143
1603
87
1i:?o
7227
953
88
12
9775
2971
77
23
6131
2049
75
25
In both databases, the majority of all the cases were less than 60 years old. In
Time One, the greatest frequency of cases (27%) were in the 61-75 year group and in
Time Two the greatest frequency of cases (25%) were in the 41-60 year group. The
smallest group in both databases was the 1-20 year group with 11% in Time One and
14% in Time Two. Finally, both Time One and Time Two had 18% of their cases in the
76 years and older group. With regard to outcome, over 85% of the cases in both
databases had three or fewer procedures. Both databases also had similar frequencies for
length of stay with three quarters of the cases staying five or fewer days.
In summary, the two databases had similar frequencies in many of the variables,
such as admit type with both having over 70 percent of cases as emergent admissions.
Admission source had over 55 percent of cases in both databases which were transfers
and referrals. Over 80 percent of the cases in both databases had some t3/pe of insurance
also. The DRG group frequencies had similar distributions in both databases and the
majority of cases also did not rcceive critical care services. The number of secondary
diagnosis and diagnostic procedures were also similar in both databases. With regard to
the outcomes, both databases had a majority of patients which were less than 61 years
old, receiving three or fewer procedures, and staying five or fewer days.
The key differences found between Time One and Time Two included hospital
size, with only nine percent of Time One and almost one forth of patients in Time Two
traveling to hospitals with less than 50 beds. In the mileage groups, Time Two had over
40 percent of cases in the less than 100 miles group and both databases had over ten
percent of cases in the over 301 miles group. More patients used private insurance than
AHCCCS in Time One and more had AHCCCS than private insurance in Time Two.
Those differences supported keeping the databases separated for the rest of the analysis.
Research Question One
This first question is concerned with what characteristics influenced distance
traveled for patients who travel for health care. The patient characteristics were arranged
into two groups, access and need. A number of descriptive and regression analyses were
used for this question. Due to the size of the sample, significance was established at p<
.01 for all statistical tests. Initial regression analysis for both databases found significance
atp<.Ql for all the variables, but it was not possible to find exactly where the significant
differences were due to the range of the distance variable, from 31 to over 800 miles in
both databases. The results also showed extensive colinearity between variables making it
difficult to interpret any differences related to distance. In order to more completely
describe patient characteristics that influenced distance traveled, a comparison of means
was used to analyze the differences between the four mileage groups. The Time One and
Time Two databases were analyzed separately throughout the study and the results of the
final analysis will be presented for both time periods.
Patient Access Characteristics
Table 7 displays the number (n) of cases in each mileage group and the mean (M)
and standard deviation (sd) for the continuous system based demographic variable of
number of beds. For the remaining dichotomous variables, the frequency (f) and percent
(%) of cases for each variable in each mileage group are displayed for both databases.
The significance (p) value from an ANOVA test for differences between the four mileage
groups for each variable follows those columns. Finally, the significant differences in
values between the groups, determined using Tukey HSD analysis, are displayed. Due to
the number of significant relationships between the groups for these demographic
variables (/'<.01), only the'non-significant (ns) differences between the groups are
displayed in the last column of the table.
73
Table 7: Analysis of Patient Access Characteristics
Mileage Group
Variable
# of cases
31-100
(1)
Time One
4123
Time Two
3954
101-200
(2)
6040
2260
201-300
(3)
1198
988
301 +
(4)
1356
941
P
NS
Hospital Size (# of beds)
Time One
Time Two
M (sd) M (sd)
283 (134) 225(120)
239 (169) 272 (130)
M (sd)
243 (147)
310(114)
M (sd)
289 (114) <.01 1,4
308 (108) <.01 3,4
Admit Type = ER/UCC
Time One
f (%)
f (%)
2881 (70) 4587(76)
f (%)
924 (77)
f (%)
1045 (77)
Time Two
Admit Source = Ref/Tran
Time One
Time Two
Medicare
Time One
Time Two
2877 (73) 1624 (72 )
779 (78)
<.01 2,3,4
3,4
800 (85) <.01 1,2
2628 (64) 3563(59)
2489 (61) 1525(68)
688 (57)
374 (38)
508 (38)
288 (31)
1225 (30) 2519(42)
1415 (36), 753 (33)
314 (26)
413 (41)
401 (30) <.01 1,4
379(40) <.01 1-2,4
3,4
1009 (35) 1723 (28)
543 (14) 382(17)
373 (31)
307 (31)
577 (43)
313 (33)
1178 (29) 1272 (21)
1375 (35) 841 (37)
366 (31)
94 (10)
111 (8) <.01 1,3
82 (9) <.01 1,2
3,4
Private Insurance
Time One
Time Two
AHCCCS
Time One
Time Two
<.01 2,3
<.01
<.01 2,3
<.01 3,4
The mean for the number of beds in Time One was greatest in mileage groups one
and four while in Time Two the mean was greatest in mileage groups three and four. The
standard deviation was over 100 for all the groups, demonstrating the large variation in
hospital size for both databases. In Time One, cases which traveled between 100 and 300
miles went to smaller hospitals. In Time Two, cases traveling less than 200 miles went to
smaller hospitals.
For both databases, the majority of cases were emergency admit types. In
Time One and Time Two, cases from greater distances, over 200 miles, had a higher
percentage of cases (77% - 85%) that were emergency admits. The majority of admit
sources were in the transfer/referral variable for all but group four in both databases and
group three mileage in the Time Two data base. The greater the distance traveled, the
fewer number of cases were referrals or transfers.
Significant differences between the mileage groups for the three payer variables
were found in both time periods. In Time One, a higher percentage of Medicare patients
were in the two shorter mileage groups while in Time Two the higher percentage of
Medicare cases were in the two greater mileage groups, in both databases, the majority of
the cases with private insurance traveled over 301 miles. The differences between the
mileage groups for the AHCCCS variable demonstrated in both time periods, the shorter
the distance, the higher frequency of AHCCCS cases.
In summary, patients who were emergently admitted or had private insurance
traveled farther in both databases. Patients who were transferred or referred traveled
shorter distances as did those patients using AHCCCS in both databases.
Patient Need Characteristics
Table 8 displays the number (n) of cases in each mileage group for the patient
need characteristic variables. The frequency (f) and perccnt (%) of cases for each of the
dichotomous variables are presented for each mileage group in both databases. The mean
(M) and standard deviation (sd) are displayed for the continuous variables in both
databases. The significance (p) value from of an ANOVA test to evaluate for differences
between all the mileage groups for each variable follows those columns. Finally, the
significant differences in values between each of the groups, using Tukcy HSD analysis,
for the patient need characteristics are displayed. Due to the number and variety of
relationships, the non-significant (ns) relationships between four mileage groups are
displayed enclosed with parentheses while the significant relationships are not enclosed
76
Table 8: Analysis of Patient Need Characteristics
31-100
(1)
4123
Time One
3954
Time Two
Mileage Group
Variable
# of cases
DRG Groups
Neuro/Ortho/Tr
Pulmonary
Cardiac
GI/GU
Women
Misc.Surg/S/B
101-200
(2)
6040
2260
201-300
(3)
1198
988
Time One
Time Two
Time One
Time Two
Time One
Time Two
F (%)
F (%)
659 (16) 117(19)
583(15) 388 (17)
317 (8) 427 ( 7)
483 (12) 286 (13)
708 (17) 1008 (17)
572(15) 269 (12)
F (%)
212(18)
215 (22)
89 (7)
99 (10)
216(18)
178 (18)
Time One
Time Two
Time One
Time Two
549 (13)
629 (16)
695 (17)
588(15)
937 (16)
273 (12)
719(12)
307 (14)
158 (14)
154(16)
163 (14)
48 (5)
Time One
Time Two
518(13)
511 (13)
877 (15)
253 (11)
152(13)
133(14)
301 +
(4)
1356
941
P
s
(NS)
F (%)
259 (19) <01 1,2
181(19) <01 3,4
108 ( 8)
.9
130 (14)
.8
246 (18)
.4
<.01
(1-2,3)
206 (22)
(3,4)
<.01
(3-1,2)
258 (19)
145 (15)
.2
64 ( 5) <-01 (2-1,3)
156 (6) <01 (1,2)
(3,4)
203 (15) .05
93 (10) .1
-
-
-
-
-
-
Type of Service = ICU/CCU
1081(26) 1934 (32) 413 (35) 500 (37) <01 (3-2,4)
Time One
1103(30) - 494 (22) 350 (36) 317 (34) <01 (3,4)
Time Two
# of Therapeutic Services
Time One
Time Two
M (sd)
M (sd)
4.6 (2.3) 5.0 (2.1)
4.7 (2.2) 4.4 (2.2)
M (sd) M (sd)
4.3 (2.1) 4.6 (2.3) <01 (1,4)
4.8 (2.4) 4.6(2.1) <01 1,4 •
2,3
# of Diagnostic Services
Time One
Time Two
3.6 (2.1) 3.8 (1.9)
3.6 (1.9) 3.6 (2.0)
3.6(1.9) 4.1 (2.1) <01 (3-1,2)
4.2 (2.0) 4.3 (2.1) <01 (1,2)
(3,4)
# of Secondary Diagnosis
Time One
3.9 (2.7) 4.4 (2.6)
4.0 (2.8) 4.1 (2.7) <01 4-1,2,3
1,3
4.5 (2.7) 4.6 (2.7) .04
Time Two
4.2 (2.6) 3.7 (2.6)
-
Significant differences were found between tiie Neuro/Ortho/Trauma groups in
both time periods. In Time One and Time Two, more cases were in the two farthest
mileage groups. The Women's health groups also had significant differences in both time
periods with more cases in the shorter mileage groups in both time periods. The Cardiac
and GI/GU DRG groups had significant differences in only one of the time periods.
Significant differences in the Cardiac cases were found in Time Two with more cases
traveling greater distances. Significant differences found in the Time One Gi/GU cases
demonstrated the greatest number of cases traveled over 301 miles. No significant
differences were found between the Pulmonary or the Miscellaneous Surg/Skin/Burns
DRG groups in either time period.
For the type of service variable significant differences were found between the
mileage groups and more cases from over 301 miles received critical care services in both
time periods. Significant differences were also found between the groups in both time
periods for the number of therapeutic services and the number of diagnostic services. The
greater the distance traveled the greater the number of diagnostic services in both
databases. The greatest mean for Time One number of therapeutic services was in the
101-200 miles groups and in the 201-300 mile group for Time Two. The differences
between the mileage groups for number of secondary diagnosis were found in Time One
with the shortest distance have the smaller mean.
In summary, question one asked: What are the demographic factors that
influence distance traveled? Cases that traveled shorted distances in both databases that
went to smaller hospitals, had more referral and transfer admit sources, and fewer
emergency admit types, had AliCCCS coverage and a Women's Health DRG. In Time
One only. Medicare cases, fewer secondary diagnosis and greater number of therapeutic
services were also significant for the shorter distance groups.
In both Time One and Time Two, patients who traveled farther went to larger
hospitals, had private insurance, and were emergency admissions, in need of critical care,
from the Neuro/Ortho/Trauma diagnosis group. In Time One, the greater the distance, the
more GI/GU cases. In Time Two, the farther distance was significant with Medicare,
Cardiac DRG and number of therapeutic services.
Research Question Two
Question number two examined the relationship between distance and risk
adjusted inpatient outcomes. Age was used to adjust risk. The outcomes were number of
procedures and length of stay. Age was entered in the first block, for each regression
followed by distance. The independent variable was the actual mileage for each case,
which was labeled as distance, entered as the second block. Since the actual mileage was
used and interpretation of the analysis was possible with distance as an independent
variable, the data were not separated into mileage groups for this question. The data were
analyzed for both the Time One and Time Two databases with each outcome as the
dependent variable.
The first outcome, number of procedures ranged from zero to six with the
majority of cases (>= 60%) receiving one or more procedures both databases. The second
outcome, length of stay (los) ranged from zero to eleven or more days, with over 50 % of
all the cases having three days or less in both databases. Table 9 displays the results of a
79
regression analysis using each outcome as the dependent variable and entering age, the
control variable, as the first block and distance (dist), the variable of interest, as the
second block. The mean (m) and standard (sd) deviation are presented for the two
outcomes in both databases. Following are the significance ip) for all the relationships.
The beta fb) and R square are displayed for the relationships which were significant at the
.01 level.
Table 9: Outcomes Regression Analysis
Outcome
Number of Procedures
Time One
Time Two
Length of Stay
Time One
Time Two
M
SD
page B age R^ age P dist B dist R- dist
1.52 1.67 .04
1.39 1.65 .96
3.88 3.06 <.01
4.02 3.11 <01
-
-
-
-
<.01
<01
-.03
-.04
.00
.00
.03
.03
<01
<01
.01
.02
.03
.02
.18
.13
When controlling for age, distance was significant at the p = .01 level for both
databases on the outcomes of number of procedures and length of stay. The negative Beta
in the procedures analysis demonstrates that the farther the patients travel, the fewer
number of procedures they receive. The Beta and R" were consistently small for the data
in this outcome and may have been an effect of the large sample size.
Other Analysis
During the regression analysis for question two, the patient characteristic files
were included in an additional analysis of this data. A second regression entered in third
and fourth blocks using each of the patient characteristic variable groups (access and
need) to measure their impact on each outcome while controlling for age and distance.
1.
80
The patient characteristic variables which were significant with the number of
procedures (p<.01) in the Time One data included; number of beds (b=.04), admit type
(b=-.01) and admit source (b=.04), pulmonary DRG (b=-.03), type of service (b=.03) and
number of therapeutic services (b=.06). In the Time Two data base, these patient
characteristic variables were significantly related to number of procedures: women's
health DRG (b=-.05), number of therapeutic services (b=.09) and number of secondary
diagnosis (b=.22).
In the Time One data base, these characteristic variables were significant (p<.01)
with length of stay: number of beds (b=.l), admit source (b=.12), Medicare (b=.04),
AHCCCS (b=,04), and all the DE-G groups with exception of pulmonary and
miscellaneous skin/surgery/burns DRGs (b= -.17 to -.06). Also significant were type of
service (b=.07), number of therapeutic services (b=.28), number of diagnostic services
(b=.21) and number of secondary diagnosis (b=.27).The characteristic variables
significantly (p=.01) related to length of stay in the time two database include all of the
variables except private insurance and number of secondary procedures with beta ranging
from .32 for number of therapeutic services to -.19 for a cardiac DRG.
In answering research question two, distance was significant for both the number
of procedures and length of stay variable. In both databases, significance was also found
between a majority of the patient characteristics and each outcome in both databases. The
greater numbers of significant relationships were with the length of stay outcome.
81
Cost Analysis
The risk adjusted length of stay outcome was significantly related to the distance
variables in this study and was used for financial analysis. The decisions in cost analysis
include what variables will be chosen and how will the costs be applied. Many studies
have used different methods of utilizing patient diagnosis in order to analyze the cost of
health care. Glick et al (2003) discussed the number of diagnosis needed for cost
estimation studies and concluded that the rate of error reduction shrank as more DRG
diagnosis codes were added. They also found that the arithmetic mean length of stay for
US DRG payments was a good predictor of unit costs in their study. They concluded that
a small number of significant DRGs and the average national pajmients were sufficient
for use in cost analysis of inpatient costs.
Another study analyzing hospital treatments found no significant differences
between various methods of analyzing cost (Larsen & Skjoldborg, 2003). A DRG cost
method was compared with traditional charge systems and an activity based costing
(ABC) estimate and all were found to be appropriate for comparing costs. These authors
concluded all three cost methods were adequate and no one method was better at
predicting costs. Finally, a study measured length of stay outliers for a group of
diagnoses, defined as the national mean length of stay per DRG plus two days. This study
analyzed hospital structural influences on patient length of stay and found outliers were
appropriate for analyzing the differences in the influences on length of stay (Cots,
Mercade, Castells & Salvador, 2003).
82
The literature search led to the conclusion that the cost analysis for this study
would use individual DRGs and length of stay outliers for each DRG used. Frequency
analysis of each of the study diagnosis groups was used to identi fy the top three DRGs in
each group for both databases. All but one DRG group had the same diagnosis for the
first two of the top three diagnoses in both databases. These DRGs accounted for a low of
23.1 % to a high of 62.6% of the DRG groups in both databases. All the other DRGs in
the databases accounted for seven percent or less of the cases in each group. Table 10
displays the frequencies and descriptions for each of the top three diagnoses in each
disease group along with the total percent of the top three DRGs for each diagnosis group
for the time one data. Table 11 displays the same information for the time two data. All
descriptions are for age greater than 17 and without complications (CC) unless otherwise
listed.
83
Tabic 10: Time One Freciuencv and Definition of Top DRGs
DRG
Neuro/O/T
209
14
243
Pulmonary
89
88
79
Cardiac
143
•127
116
GI/GU
182
174
183
Women
373
371
359
Description
Case
n
Case
%
Major Joint and limb reattachment (LE)
IC hemorrhage & stroke with infarct.
Medical back problems
333
160
104
15
7
5
DRG group
%
27
49
Simple pneumonia/pleurisy with CC
Chronic obstructive pulmonary disease
Resp. infection/inflammation with CC
221
178
62
23
18
7
33
Chest pain
Heart failure and shock
Other cardiac pacemaker implantation
256
235
225
12
11
10
26
Esophagitis/gastroenteritis & misc with
CC
GI hemorrhage with CC
Eso/gast. & misc. digestive disorders
187
10
176
137
9
7
63
Vaginal delivery
Casearean Section
Uterine and adnexa procedure,
nonmalignant
MiscS/S/B
Rehabilitation
462
Laparoscopic Cholecystectomy
494
Laparoscopic Cholecystectomy with CC
493
683
173
359
41
11
11
269
125
101
15
7
6
28
84
Table 11: Time Two Frequency and Definition of Top DRGs
DRG
Neuro/O/T
209
14
1
Pulmonary
89
88
91
Cardiac
143
127
124
GFGU
182
Description
Case
n
Case
%
Major joint and limb reattachment (LE)
IC hemorrhage & stroke with infarct.
Craniotomy with CC
170
124
82
12
9
6
DRG group
%
28
50
Simple pneumonia/pleurisy with CC
Chronic obstructive pulmonary disease
Simple pneumonia/pleurisy 0-17 yrs
252
151
97
25
15
10
33
Chest pain
Heart failure and shock
Circ. disorders cxc.AMl with cath.& C
diag.
173
149
83
14
12
7
29
Esophagitis/gastroenteritis & misc with
CC
GI hemorrhage with CC
Eso/gast. & misc. digestive disorders
174
183
Women
Vaginal delivery
373
Vaginal delivery with CC
372
Casearean section
371
MiscS/S/B
Rehabilitation
462
Laparoscopic cholecystectomy
494
Back and neck procedures except spinal
500
fus.
132
11
112
79
9
7
369
95
95
37
10
10
90
70
63
9
7
6
56
23
In order to apply costs, a variety of different definitions were used. In the national
information the costs were identified as average payments. In the databases, the costs are
entered as charges. This analysis used both the payments and charges in the cost analysis
as both were recognized in the literature to be adequate for comparison in cost analysis
(Glick et al, 2003. Larsen et al, 2003).
The Arithmetic mean length of stay (AMLOS) and national average Medicare
payment for all eighteen DRGs was found in the 2003 edition of DRG Expert (Hart &
Schmidt). Frequencies from both databases found both the mean length of stay and mean
of total charges for each of the eighteen DRGs. The length of stay and charges from the
study and the national payment and length of stay averages were used to find the payment
and chargcs per day for each DRG. Finally, the outlier length of stay was defined as the
AMLOS plus two. Table 12 displays the results of the cost analysis for the top three
diagnoses in each DRG group for Time One.
86
Table 12: Time One Cost Analysis Data
DRG
Group
and top 3
Neuro/O/T
209
14
243
Pulmonary
89
88
79
Cardiac
143
127
116
GI/GU
182
174
183
Women
373
371
359
MiscS/S/B
462
494
493
AM
LOS
National
Payment
Payment
Per Day
Study
LOS
mean
Total
charges
Charge /
Day
5.0
6.2
4.7
9200.19
5729.87
3283.95
1840.04
921.17
698.71
4.5
4.8
3.7
29795.53
17889.06
11377.77
6635.98
3726.89
3042.18
5.9
5.1
8.5
4612.93
3996.70
7168.64
781.85
783.67
843.37
4.5
4.2
6.9
12991.75
10619.85
25997.37
2893.49
2534.57
3756.84
2.1
5.3
4.5
2386.60
444.27
10216.6
1350.76
838.54
2270.36
1.7
4.1
3
7606.65
12946.80
37628.14
4527.77
3157.76
12459.7
4.4
4.8
2.9
3535.40
4405.75
2533.57
803.50
917.86
873.64
3.3
4
2.6
13132.02
13608.69
7830.81
3991.50
3385.25
3070.91
2.3
3.6
2.6
1741.58
2986.01
3694.33
757.21
829.45
1420.90
1.7
3.1
2.2
4177.42
9054.70
10619.63
2480.37
2902.15
4916.50
11.5
2.5
5.9
5001.62
4474.37
8035.89
434.92
1789.75
1362.02
7.6
2.3
4.6
25442.94
12752.39
20747.34
3347.76
5617.79
4549.86
Table 13 displays the cost information for the Time Two cases. Both databases
display similar patterns. The study mean length of stay is consistently less than the
AMLOS while the charges per day are, at a minimum, three times greater than the
payment per day.
87
Table 13: Time Two Cost Analysis Data
DRG
Group
and top 3
Neuro/Orr
• 209
14
1
Pulmonary
89
88
91
Cardiac
143
127
124
GFGU
182
174
183
Women
373
372
371
MiscS/S/B
462
494
500
AM
LOS
National
Payment
Payment
Per Day
Study
LOS
mean
Total
Charges
Charges/
Day
5.0
6.2
11.2
9200.19
5729.87
16556.5
1840.04
921.17
1478.26
4.6
4.8
6.5
29130.10
13676.81
67764.30
6546.10
2861.26
10361.5
5.9
5.1
4.0
4612.93
3996.70
3113.95
781.85
783.67
778.49
4.5
4.0
2.8
12534.69
10122.90
6213.46
2960.94
2543.44
2203.35
2.1
5.3
4.4
2386.60
4444.27
6381.52
1350.76
838.54
1450.35
1.4
4.3
3.2
7510.80
14101.89
19961.36
5364.85
3294.83
6296.96
4.4
4.8
2.9
3535.40
4405.75
2533.57
803.50
917.86
373.64
3.1
3.0
1.9
9490.08
8535.85
7698.37
3071.22
2870.42
4161.28
2.3
3.7
3.6
1741.58
2770.86
2986.01
757.21
748.88
829.45
1.7
3.0
3.4
4701.07
8535.85
10915.88
2717.38
2870.42
3191.77
11.5
2.5
2.5
5001.62
4474.37
4200.78
434.92
1789.57
1630.31
8.4
2.1
1.8
42664.87
12698.53
13162.98
5097.35
6088.51
7192.89
In Time One, the greater charges per day were for Cardiac and
Neuro/Ortho/Trauma and the lower charges were in Pulmonary and Women's Health. In
Time Two the greater charges were in Neuro/Ortho/Trauma and Miscellaneous
Surgery/Skin/Burns while the lower charges were also in Pulmonary and Women's
Health.
Finally, all the DRG groups were evaluated for the number of patient days beyond
the outlier (AMLOS + 2), which were then multiplied by the charges per day for a total
charges per DRG. Table 14 displays the results from both databases and the total charges
for length of stay outliers for all cases in the study. The total charges were then totaled
from both databases for a grand total of $3,968,024.00 in charges for outlier lengths of
stay for rural patients during one year in one state. If divided equally between the
hospitals in the study, the potentially non-reimbursable costs would be for 55 thousand
dollars per hospital.
89
Table 14: Length of Stay Outlier Costs
DRG Group
and top 3
per cohort
(one/two)
Neuro/O/T
209
14
243/1
Pulmonary
89
88
79/91
Cardiac
143
127
116/124
GFGU
182
174
183
Women
373
371/372
359/371
Misc S/S/B
462
494
493/500
Time One
Outlier days
Time One
Outlier
Charges
Time Two
Outlier days
Time Two
Outlier
Charges
Total Outlier
Charges
74
60
38
491,064 •
223,620
115,596
28
17
0
183,288
48,637
0
674,352
272,257
115,596
48
39
16
138,864
98,865
60,112
47
33
6
139,167
83,919
13,218
278,031
182,784
73,330
17
64
54
76,976
202,112
672,840
7
64
19
37,555
210,880
119,643
114,531
412,992
792,483
52
40
20
207,587
135,400
61,420
21
10
1
64,491
28,700
4,161
272,075
164,100
65,581
20
10
0
57,040
29,020
0
16
47
29
43,472
134,890
92,568
100,512
163,910
92,568
0
7
25
0
39,326
113,750
0
3
3
12 Month
0
18,267
21,579
Total
0
57,593
135,329
3,968,024
The majority of the outlier charges were in the neurological, cardiac and
pulmonary groups in both cohorts. Some of the DRGs (243 & 462) did not have any
outliers identified due to the AMLOS of eleven days or more. The study data had
grouped all the lengths of stay over eleven days and identification of stays greater than
eleven days was not possible. As the total charges for length of stay outliers may not be
reimbursable, a potentially large debt could result from allowing rural patients to stay
under a system's care for an additional day or two.
Summary
This chapter discussed the results of the statistical analysis of the data base to
answer the two research questions:
1. What are the demographic factors that influence distance traveled for rural
patients?
2. What is the relationship between distance and risk adjusted inpatient health
outcomes for rural patients?
A number of characteristic factors had significant differences between mileage groups.
The patients who traveled greater distances went to larger hospitals, had higher rates of
private insurance, had a higher number of Neuro/ortho/trauma diagnoses, received more
critical care, and diagnostic services. Patients who traveled shorted distances had more
referrals and transfers, fewer emergency admits, more AHCCCS (or Medicaid) cases and
a higher number of Women's Health diagnoses. Patients travel due to a variety of
characteristics to a variety of hospitals for a variety of services. There were also
differences in significance between the Time One and Time Two databases in the number
of Medicare cases, the GI/GU and Cardiac diagnosis groups, and the number of
secondary diagnosis and therapeutic services. The results demonstrated that no single
factor has a strong influence on distance traveled.
One of the significant relationships between distance and outcomes in both
databases in this study was between distance and length of stay. A cost analysis of the
outliers for the Jength of stay demonstrated that while the mean length of stay was less
than the national average, the charges were consistently three times greater than the
national average payment. An analysis of the outliers for length of stay resulted in
recognizing close to four million dollars in potentially non-reimhursable costs. A
discussion of the results of the analysis and recommendations for health policy and
nursing will be discussed in the next chapter.
CHAPTER SIX
Discussion of Results
This exploratory study was conducted to develop an understanding of the factors
that influence distance traveled for health care and the relationship between distance
traveled and patient outcomes. One year of data (June 2001- July 2002) from the Arizona
Department of Health Services Hospital Discharge Database was used for this study. All
records which met the criteria for rurality for this study were included i.e.: those patient
records remaining after excluding all records with zip codes within 30 miles of hospitals
which were within 30 miles of another hospital in the database. Newborn and expired
cases were excluded.
The exploratory analysis was based on Donabedian's Systems Model which
defined structure, process and outcomes (1966). A study framework of patient migration
was created and portions of the framework were used for the analysis: patient
characteristics, distance traveled, and risk adjusted patient outcomes. The discharge data
were organized in the study framework in order to answer two research questions.
1. What are the demographic factors that influence distance traveled for rural
patients?
2. What is the relationship between distance and risk adjusted inpatient health
outcomes for rural patients?
Cases were divided into four mileage groups for analysis of question one. Two patient
characteristic groups were created, access and need. Factors in both characteristic groups
influenced distance traveled. For the second question, distance was significant for both
risk adjusted outcomes of number of procedures and length of stay. A cost analysis on the
length of stay outliers was the included to answer question two.
This chapter will discuss the results of the analysis in relation to the research
questions and the study framework. The limitations of the study will then be discussed.
Finally, implication of the results for both health policy and nursing will be addressed,
followed by recommendations for further research.
Research Question One
Patient characteristics which influenced distance traveled in both databases
included hospital size, emergency admit type, referral/transfer admit source, three
different payers; Medicare, private insurance and AHCCCS, two DRG groups;
Neuro/Ortho/Trauma and Women's health, critical care service, and the number of
therapeutic and diagnostic services. Factors which were significant in the Time One data
only included the GI/GU DRG group and number of secondary diagnosis. In Time Two
the Cardiac Diagnosis was a significant factor for distance traveled. The only patient
characteristics which did not influence distance traveled were the Pulmonary and
Miscellaneous Surgery/Skin/Burns DRG groups.
A greater number of patients who traveled over 200 miles in this study went to
larger hospitals, were emergency admits with private insurance, received critical care
services and had a Neuro/Ortho/Trauma DRG with a higher number of diagnostic
services. The majority of patients who traveled under 200 miles were referred or
transferred to smaller hospitals, used AHCCCS coverage and had a Women's Health
DRG. The other significant variables of influence, Medicare and number of therapeutic
services were different between the Time One and Time Two databases.
Research Question Two
While risk adjusting for age, distance had a significant relationship with the
number of procedures, and length of stay in this study. For the number of procedures
outcome, the patients who traveled over 301 miles had the smallest number of
procedures. For the length of stay outcome, the patients who traveled under 100 miles
had shorter lengths of stay. In this study, patients who traveled the farthest distances had
fewer procedures and longer lengths of stay.
Additional analysis demonstrated a number of the patient characteristics were also
significantly related to the outcomes but no clear relationship could be identified in both
databases. For the number of procedures outcome, only the number of therapeutic
services was significant in both databases. For the length of stay, hospital size,
emergency admit. Medicare and AHCCCS, critical care services, number of therapeutic
and diagnostic services and the top four DRG groups (Neuro/Ortho/Trauma, Women's
Health, Cardiac, GI/GU) were all found to have a significant relationships. Other studies
of length of stay have also found patients with a greater medical need have greater
lengths of stay (Lebrero et al, 1999).
A cost analysis of length of stay outliers found a total of close to four million
dollars in potentially non-reimbursable charges. Costs were created using the charges for
services in the database and the national average of payment per DRG. The charges were
consistently at least three times the average payment amount for each DRG while the
95
lengths of stay in the database were consistently less than the national average. While it
was not possible to analyze the actual costs in this study, a large financial gap between
payments and charges was demonstrated.
Relationship of Findings to Conceptual Model
la order to better understand the findings in relationship to the framework, figure
4 displays the study model with the variables studied included. Patient characteristics
included both access and need variables. The distance variable was both actual mileage
and the four groups of 100, 200, 300 and over 300 miles. Age was used as the risk
adjuster for the quality outcomes of number of procedures and length of stay and a cost
analysis was used, to better understand the financial impact of length of stay outliers.
96
Figure 4: Model with study variables
STRUCTURE
PROCESS
Rural System Capacity
Urban System
Capacity
Patient Decision
Rural System
Outcomes
Distance Traveled
Mileage Groups
and
actual mileage
Patient Characteristics
- Access Number of beds
Payer
Admit type
Admit source
OUTCOME
- Need Type of Service
DRG
# therapeutic svcs.
# diagnostic svcs.
# secondary diag.
Urban System
Outcomes
Risk Adjusted
Patient Outcomes
Age as risk adjuster
-Quality# procedures
Length of Stay
-EfficiencyCost
Rural Community
Outcomes
Patient Access Characteristics
The results of the analysis demonstrated a significant relationship between
distance and the number of beds in both databases with more patients in Time One
traveling greater distances to larger hospitals and more patients in Time Two traveling
shorted distances to smaller hospitals. As a result, one conclusion can be that rural
patients travel different distances to a variety of hospitals. There is no clear pattern of
patients traveling to specific types of hospitals in this study. Studies have tried to
understand why patients choose different hospitals for care and have only been able to
identify characteristics of the patients, such as males with private insurance (Basu &
Cooper, 2000). This study has identified a number of other characteristics which
influence distance traveled.
The emergency admission type had a greater number of cases of patients who
traveled farther in both databases. The referral/transfer admission source had fewer cases
of patients who traveled farther. Research has recognized the lack of access to health care
in rural areas (Schur & Franco 1999). Due to this lack of access, patients who live in
distant rural areas may not access health care resources until they have an emergent need.
This may also explain the fewer numbers of referrals or transfers for patients who travel
greater distances.
When discussing the payer, significant differences were found for all three,
private insurance cases were from greater distances, AHCCCS cases were from shorter
distance and Medicare cases varied between the two time periods. This study supports a
previous finding (Basu & Cooper, 2000) that people with private insurance travel greater
distances. The fewer AHCCCS cases at the greater distances may be due to a lack of rural
providers who accept the state policy or due to the fact that more patients on AHCCCS
live closer to urban areas in order to access other social services. While research has
demonstrated the disparities in the number of available providers and the lack of social
services in rural areas (Casey, 1999), this exploratory study is not able to draw
conclusions about the relationship between distance traveled and AHCCCS payer.
98
Patient Need Characteristics
Two of the DRG groups, Neuro/Ortho/Trauma and Women's Health, had a
greater number of cases and were significant with distance in both databases.
The Neuro patients traveled greater distances and the women traveled shorter distances.
Cardiac and Gl/GU were significant in one of the time periods and both had greater
numbers in the greater distances. The Pulmonary and Miscellaneous Surg//Skin/Burn
groups were the smallest of the top six and were not a significant factor influencing
distance traveled.
The top three diagnoses in each DRG group were used for the cost analysis and
some interesting patterns also ajose from identifying those diagnosis. Both databases
contained two of the same DRGs in the top three for each group. None of the DRGs
contained more than 25% of the cases in any group, demonstrating the large variation of
diagnosis for this study population. The majority of the DRGs were for acute events and
one fourth of them included complications.
Many rural programs focus on Medicare populations and patients with chronic
conditions, such as the Medicare demonstration projects of community based nursing.
The results of this study demonstrate that the majority of rural patients who travel are not
covered by Medicare and are not hospitalized primarily for chronic conditions. The
majority of cases in this study had a payer other than Medicare and traveled for care for
acute events such as trauma, stroke, chest pain and childbirth.
Other studies of rural patients have discussed the greater number of risks from
occupational injury (Ricketts et al, 1999), the top Neuro/ortho/trauma diagnosis of major
joint and limb reattachment in both databases may or may not be related to occupational
injuries. Research has also shown lower rates of prenatal care and higher rates of neonatal
deaths for rural patients (Lishner et al, 1999. Clark et al, 1999). One of the most
significant DRG groups in this study was Women's Health and childbirth was the
majority of the top diagnosis in both databases. This exploratory study cannot develop
any conclusions except to recognize this significant group which travels for care.
In both databases, patients that traveled greater distances received critical care
services and a greater number of diagnostic services. Like the earlier discussion of
emergency admits and acute diagnoses related to distance, the pattern of increased acuity
with increased distance continues with these two characteristics. The number of
therapeutic services, while significant, is highest in the middle ranges of distance and
does not increase with distance.
Risk Adjusted Patient Outcomes
After adjusting for age, there was a significant relationship between distance and
both the risk adjusted patient outcomes of number of procedures and length of stay in
both databases. Patients who traveled farther had longer lengths of stay and received
fewer procedures. When comparing the study's average length of stay with the national
average (AMLOS ) for the top DRGs, the length of stay was consistently lower than the
national average. As it has been previously stated, other distance studies have not
measured outcomes so the results of this study have no comparison.
A cost analysis of outliers demonstrated close to four million dollars in
potentially non-reimbursable charges due to lengths of stay greater than the AMLOS plus
100
two. A frequency analysis of the length of stay per hospital demonstrated a fairly equal
distribution among the 72 hospitals in the study in both databases. An equal division of
the total charges for outliers between the 72 hospitals resulted in over $55,000.00 in
potentially non-reimbursable charges for each hospital.
As previously discussed, rural health systems face a number of challenges to
remain viable ( Younnis, 2003 ). Add the increase in non-reimbursable charges for rural
patients who must travel and the economic picturc becomes darker. This study
demonstrated that a majority of rural patients were referred or transferred. Frequently
costs are shared between the treating and referring systems, which adds another element
to the financial risks that needed to be understood. The number of outlier days in this
study and the high rates of early readmissions (within 30 days) from a previous study
(Sweeney Fee, 2001) demonstrate an increased risk to the viability of rural systems which
needs greater understanding.
A more recent study discussed the problems families have paying medical bills
(May & Cunningham, 2004). Even when they are insured, patients may delay or go
without needed care due to problems with medical bills. When adding the results of this
study, rural patients may be at higher risk for avoiding care which may result in an
increase number of emergency admissions for acute illness with complications. Charges
which are not reimbursed by insurance payers are billed to patients. The high amount of
charges for length of stay outliers and the potentially added costs of early readmissions
for rural patients may add pressure to an already disparate rural health care environment.
101
Limitations
There were three major limitations in this study. 1) The limitations of the database
used in the study. 2) The variety and availability of software needed for many of the
options in the analysis of the data. 3) The lack of cost infonnation available for the cost
analysis.
The databases used in this study were de-identified for public use. The data
included patient age and zip code, hospital id and zip code, physician information, and
diagnosis and treatment information, but did not include date of birth, social security
number or physical addresses for the patient cases. Without patient identification,
readmission rates cannot be analyzed, which limits the outcomes available for analysis.
Another limitation of de-identified data is the inability to check the accuracy of the
coding information. As a result, the decision made for this exploratory study was to
convert the ICD-9 coded variables into a dichotomous 0, 1 variable and use total number
of diagnosis or procedures for analysis. This limited this study's ability to understand
more complex outcomes, which may be a consequence of treatment, such as those
measured by Needleman et al (2001). The decisions made related to the development of
the variables for the study framework had to include recognition of the limitations of the
database. Finally, the grouping of the Diagnosis Related Groups was based in traditional
medical foundation and may be problematic.
The limitations of available software for data analysis and risk adjustment were
recognized from tlie start of the analysis. A variety of experts were utilized in order to
support decisions related to the creation of the analytic data file. For example, distance
102
measurement required both Geographic Interface and Zip Code mapping software, both
of which were available from another university department. Issues related to
communicating across disciplines led to increased concern regarding the reliability of the
distance variable. A number of discussions and processes were needed to ensure the
variable of interest for this study was systematically developed to ensure accuracy of the
measure. Risk adjustment of outcomes was also limited by the lack of available software
for Charlson programming and for disease specific risk adjustment methods.
Additionally, identification of costs was limited by what information was publicly
available. Both payment and charge information were used in this study, but the actual
cost of care remains open to question. As a result, the cost analysis had to focus on
payments and charges, a method that may not allow for the best analysis of the financial
impact of health outcomes for rural patients. Also, national payment and length of stay
information was used in this study and the use of regional information may have resulted
in fewer differences than between the database and the national information.
Implications for Health Policy and Nursing
With regard to question one, a number of patient characteristics influenced
distance traveled. Rural patients traveled a variety of distances to both large and small
hospitals. The majority of the patients were less than 60 years old. With regard to payer,
more patients had private insurance or AHCCCS than Medicare coverage. Many of the
DRGs in the study were not for chronic conditions, and the smallest and least significant
was the Pulmonary group. Patients who had to travel farther had higher acuity
characteristics of emergency admissions and critical care services. Rural health systems
103
research frequently focuses on the elderly with chronic conditions (Coward & Krout,
1998). The results of this study demonstrate that many other characteristics may need to
be part of rural health research and federally funded projects.
For question two, both outcomes were significantly related to distance traveled.
The length of stay outcome, if added to the high rate of readmissions found in the pilot
study, demonstrates an opportunity to develop improved interventions for rural patients.
Add the high rate of discharge to self care which was in this database and the potential to
create new interventions increases. Communication with a cohort of nurses about the
results of this study led to a discussion of the occasions when rural patients remain in a
hospital for an extra day or two due to the l8.ck of support services in their rural area.
While lower than the national average, the number of outlier cases demonstrates a need
for greater understanding of why rural patients stay longer.
As for the implications from the development of the variables for the framework,
the concepts of distance traveled and patient outcomes will be discussed. The analysis of
the distance variable resulted in a map which displayed the clusters of urban health care
systems and the distribution of patients who access care. The map in this study did not
follow any county line boundaries. Existing health care geography and the definitions of
rural populations are based in the use of county boundaries (ERS, 2003). This study
demonstrated the technologies available to form a more accurate picture of health care
systems and patient access. Perhaps through increased use of the new technologies, an
improved and more accurate definition of rurality can be developed.
104
With regard to the outcomes, strategies continue to be discussed about the type of
health services needed to improve the efficiency and equity of health care for rural
patients (Lieberman, et al 2003 ). The results of this study demonstrate the potential for
development of nursing case management models for rural patients that may improve
health outcomes and financial efficacy for rural health systems. Health policy will need to
develop new methods of reimbursement for rural case management services for patients
who are not elderly and do not have chronic conditions. Nursing may have an opportunity
to create a new scope of rural nursing with the support of rural health systems as a result
of the financial impact of extended lengths of stay and high rates of readmission for rural
patients.
Finally, the importance of developing interventions and programs from
empirically based information has been recognized in the literature (Etheredge, 2003).
This study discussed the lack of understanding of outcomes for rural patients and found a
number of patient characteristics which have not been discussed in the literature which
impact distance traveled. This study analyzed health outcomes for a population from one
state for one year and found information that has not been previously recognized and
found potentially non-reimbursable charges of close to four million dollars. Interventions
and programs for rural patients which are created without empirical support may result in
neither effective nor efficient results. With the current struggles to remain financially
viable, can rural systems afford to create programs which are not focused on rural
patients needs?
105
Recommendations for Further Research
This exploratory study utilized a framework developed using Donabedian's
Systems Theory (1996). The concepts in the framework were developed through a search
of literature discussing migration. In creating the variables for the analytic data file, the
multilevel nature of the framework was acknowledged. Concept development is needed
to better describe each concept in the framework. Testing the framework and the fit of
each concept would ensure the appropriate placement of each concept within in the
framework.
Other theoretical models may be needed if the concepts and fit of the framework
are found to not to be appropriate for Donabedian's systems model. The framework was
found to be appropriate for organizing variables for an exploratory study but further
research is needed to develop the framework in order to answer other questions related to
the migration of rural patients for health care.
Additional analysis of the database could lead to a deeper understanding of the
study population. With each answer to the research questions, more questions arose. Why
do patients that travel farther have more critical needs? What is the number of procedures
and lengths of stay for each of the top DRGs? Do the urban cases in the database have
similar characteristics and outcomes? Future investigations of the database could answer
these questions.
Studies of other state or national databases are also needed to see if similar results
occur for rural patients in other areas. It is import to understand if the characteristic
findings, such as the types of payers and the variety of diagnosis groups, and outcomes
106
are repeated in other areas. Especially considering the results of this cost analysis,
multiply $ 4 million by 50 states and the need to investigate this financial impact on the
nation's health systems can be easily recognized. A more detailed cost analysis that can
display the relationship between costs and charges using regional comparisons could lead
to a more accurate portrayal of the financial impacts of traveling for health care.
Additional analysis is also needed to verify if the results of this study are unique or
universal.
As stated previously, some of the limitations were lack of patient identification
for improved analysis of outcomes and the lack of software for improved methods of
analysis. Future research using identified patient information would allow studies which
can track patients across a continuum of care and would also allow for a random
verification of the data base. Also, an addition of both outpatient and long term care
databases with patient identification would allow for an increased understanding of the
path that rural patients travel in order to obtain their health care. Increased measurement
of quality indicators and readmission rates would also be available in a database with
identified patient infonnation. Additionally, improved software and cost identification
capabilities could improve the scope of the analysis.
Conclusion
This exploratory study developed a framework based in systems theory to
understand rural patients who must travel for health care. This study used one year of
Arizona hospital discharge data to identify a variety of factors that influenced travel for
health care. A variety of patient characteristics influenced patient travel and both the
107
number of procedures and length of stay were significant outcomes. Patients who went to
larger hospitals, were emergently admitted, used private insurance, received critical care
services, had a Neuro/Ortho/Trauma DRG and more than four diagnostic procedures
traveled over greater distances. Patients who went to smaller hospitals, were referred oar
transferred, used AHCCCS, and had Women's Health DRG traveled shorter distances.
The cost of the length of stay outliers resulted in close to four million dollars of
potentially non-reimbursable charges for the 72 private health carc systems in the state.
The limitations of the data and software available were recognized and recommendations
for health policy, nursing and future research were discussed.
In conclusion, this study began with the discussion of the lack of understanding of
rural health outcomes and the results clearly demonstrate the need for health policy
institutions to begin to use different models to analyze the needs and outcomes of rural
patients and the systems that care for them. The technology exists to allow for the
development of improved frameworks for understanding rural patients. The framework
used in this study was found to be appropriate for organizing variables in order to explore
large databases.
This study demonstrated that rural patients do not fit one narrow profile. Rural
patients travel for health care to a number of different hospitals, most have some type of
insurance, many are referred or transferred and few need critical care services. The
largest of the top six DRG groups was Neuro/Ortho/Trauma and the smallest was
Pulmonary. The farther patients travel for care, the fewer procedures they receive and the
longer they stay. The deficiencies in the limited focus of current rural health research
have been recognized. There are opportunities to increase our empirical knowledge of
rural patients and to develop more efficient models of equitable care. It is time to utilize
the technologies and information available in order to create interventions which can
recognize the needs and improve outcomes for rural patients.
APPENDIX A
Arizona Department of Health Services
Hospital Discharge Data Report Requirements
ARIZONA DEPARTMENT OF HEALTH SERVICES
Page 1 OF 11
ARIZONA DEPARTMENT OF HEALTH SERVICES
HOSPITAL DISCHARGE DATA REPORTS
REQUIRED DATA ITEMS AND FORMAT SPECIFICATIONS FOR REPORTS THAT ARE DUE BEGINNING AUGUST 15, 2004
I
OBTAIN REQUIRfiD SUBMISSION FORMS ON OUR WEBSITE AT: wwwJis.stata .az.us/plan /tirr/Sndex.litm
Hospital Discharge Data File TypeS: Hospital Inpatient (IP),Hospital Outpatient Clinic (OP); Hospital Emerigency Department (ED)
NOTE; EACH FILE TYPE MUST BE SUBMITTED AS A SEPARATE FILE - SEE FILE MAMING CONVENTION AT THE END OF THIS DATA SPECIFICATION DOCUMENT
Fixed Length Record of 860 characters FOR ALL FILE TYPES
'
,
FORMAT; ASCII TEXT - ALL ALPHA CHARACTERS MUST BE IN UPPERCASE.
Data Reporting Requirements Pursuant to Arizona Revised Statues (A.R.S.) § 36-125.05 and Arizona Administrative Code (A.A.C.) TUIe 9, Chapter 11Articles 3 & 4
Medium: Compact Disk (CD) or DIskelte '
Start
End
10
1
10
iP-OP-ED A^rlzona Facllily Identification
dumber-AZ FAGJD
17,
11
27
49
28
46
1
47
47
Number of
Characters
.
Required
For File
Type .
Data Element Name
j UnifMin CODES AHO VALUES - ALL ALPHA CHARACTERS EDIT REQUIREMENTS - LEFT JUSTIFY AND LEAVE UNUSED
Billing
MUST BE IN UPPERCASE •
SPACES BLANK UNLESS OTHERWISE INDICATED
Locator
Number
n/a
Alpha-Numeric
Must be filled In for all File Types. Right justified with leading
zeros. All AZ FAC__ID numbers begin with the alpha characters
\^ED (ALL CAPS) followed by a four-dlgll number, with no spaces
or dashes. [Example:MED12343 All Arizona hospital AZ
FAG ID's are available on website:
www.hs.stat&.az.us/plan/crr/lndex.htm
fP-OP-ED Patient's Medical Record
Number
23
Alpha-Numeric
Must be filled In for all File Types. Rightjuslified with leading
zeros.
IP-OP-ED Certificate, Social Security
Number or Health insurance
Claim Number
BO
Alpha-Numeric
Must be fllledln for all File Types. Rightjuslified with leading
zeros.
n/a
Race
1 =• Amerlean Indian or Alaskia Native
Musi be entered for File Type IP. Leave Blank for File Types OP,
and ED. All Race codes match the US Census Bureau revised
2 " Asian
standards for Race and Ethnic reporting effective Januaiy 1, 2003
IP Only
PallenCs Race
. \
3 " Black or African American
4 « Hispanic or Latino
5
White
6 « Native Hawaiian or Other Pacific islander
7 « Other (Us© if patient not described above.)'
9 =' Refused
1
48
48
IP Only
Patient's Marital Status
16
Marital Status
Must be entered for File Type IP.Leave Blank for File Types OP
S« Single
and ED.
M = Married
P = Life Partner
X »= Legally Separated
ARIZONA HOSPITAL DISCHARGE DATA SPECIFICATIONS
UPDATED MAY 2004
ARIZONA DEPARTMENT OF HEALTH SERVICES
Number of
Characters
Start
End
Required
For File
Type
Data Element Name
Page 2 OF 11
Uniform CODES AND VALUES - ALL ALPHA GHARACTERJ
EDIT REQUIREMENTS - LEFT JUSTIFY AND LEAVE UNUSED
Billing
MUST BE IN UPPERCASE
SPACES BLANK UNLESS OTHERWISE INDICATED
Locator
Number
—
D s Divorced
W " VWdowed
U = Unknown
30
49
2
39
78
IP-OP-ED Patient's Street Address
IP-OP-ED Patient's City
100 IP-OP-ED Patient's State
13
13
Alpha-Numeric
Alpha-Numeric
13
Alpha-Numeric
Patient's Zip Code
13
Alpha-Numeric
Patient's Date of Birth
14
Enter month, day and year, without dashes.
MMDDYYYY
Must be filled !n for all File Types. All dlyits must be filled In (no
dashes). If any portion of birthday is unknown enter all zeros fgr
the birthday.
15
Patient's Sex
M= Male
F= Female
U = Unknown
Must be entered for all File types.
IP-OP-ED Patient's Sex
Q
IP-OP-ED Data of Admission
8
Dale of Dlscharpe
6
Hour of Admission
10
135
137
IP Only
Must be filled In for aif File Types
Must be filled tn for all File Types.
Must be filled in for ali File Types. Use US Postal Service state
abbreviations. If a foreign resident, leave this field blank and fill in
name of the country tn Patient's Zip Code field betow according to
the Edit Requirements.
Must be filled in for all File Types. Use US Postal Zip Code for the
patient's residence at the time of admission or encounter, If Zip
plus four is used Indicate as XXXXX-YYYY. Use 00000 for
unknown zip codes. If a foreign resident, fill in with the first 5
letters of the name of the country, for example Mexico = MEXIC;
Canada = CANAD; England = ENGLA; and leave the rest of the
field blank.
:•
'
Enter month, day and year of the patient's
admission to the hospital without dashes.
MMDDYYYY
Enter month; day and year of the patient's
discharge from the hospital without dashes,
MMDDYYYY
Musi be filledIn for all File Types, Ail digits musl be filled In (no
dashes).
Time Codes for Hour of Admission
Musi be entered for File Type IP. Both digits Must be filled In.
Leave Blank for File Types OP and ED.
00 = 12:00-12:69 AM {Midnight)
01=01:00-01:59
02=02:00-02:59
03 = 03:00 - 03:59
04=^54:00-04:59
05 = 05:00-05:59
06 = 06:00-06:59
07 = 07:00-07:59
Musi be filled In for all File Types. All digits must be tilled in (no
dashes).
08 = 08:00-08:59 '
ARIZONA HOSPITAL DISCHARGE DATA SPECIFICATIONS
UPDATED MAY 2004
ARIZONA DEPARTMENT OF HEALTH SERVICES
Page 3 OF 11
Number of
Characters
Start
End
Required
For FUe
. Type
Data Element Name
\
Unfform CODES AND VALUES - ALL ALPHA CHARACTERS
EDIT REQUIREMENTS - LEFT JUSTIFY AND LEAVE UNUSED
BflDng
MUST BE IN UPPERCASE
SPACES BLANK UNLESS OTHERWISE INDICATED
Locator
'•(
Number
09 = 09:00-09:59
10 = 10:00-10:59
'
11 = 11:00-11:59
^
12 = 12:00-12:59 PM (Noon)
13 = 01:00-01:59
14=^02:00-02:59
15 = 03:00-03:59
'
[16='04:00-04:59
17 = 05:00-05:59
18 = 06:00-06:59
19-07:00-07:59
20 = 08:00 - 08:59
138
IP Onfy
.
21 =D9:00-09:59
22= 10:00- 10:59
23=11:00-11:59
99 = Hour Unknown
See lime Codes above.
'
Hour of Discharge
21
Date Bill Submitted
86
The month, day, and year the bill was submitted Must be entered for File Type IP All digits must be filled !n (no
to the patient without dashes. MiVlDDYY
dashes). Leave Blank for File Types OP and ED.
(NOTE: This date field has only six characters)
Patient's Discharge Status
22
The circumstances under which the patient left Must be entered for all File Types. All codes match the Medicare
the hospital:
Patient Discharge Status Codes and follow all Medicare
01 = Discharged to home or seif care (routine definitions of terms.
discharge)
Must be entered for File Type IP. Both digits Must be filled in.
Leave Blank for File Types OP and ED
02 = Dfscharged/transfen-ed to another short
term genera! hospital for inpatient cars (See
code 43)
03 = Discharged/transferred to sitilled nursing
(SNF).
04 = Discharged/transferred to an intermediate
care facility (ICF).
05 = Discharged/transferred to another fyps of
Institution for inpatient care'
06 = Discharged/transferred to home under
care of organized home health service org.
ARIZONA HOSPITAL DISCHARGE DATA SPECIFICATIONS
UPDATED MAY 2004
ARIZONA DEPARTMENT OF HEALTH SERVICES
Page 4 OF 11
Number of
Characters
StaH
End
Required
Data Element Name
For File
Typa
Uniform CODES AND VALUES - ALL ALPHA CHARACTERS
EDIT REQUIREIWENTS - LEFT JUSTIFY AND LEAVE UNUSED
Blillng
MUST BE IN UPPERCASE
SPACES BLANK UNLESS OTHERWISE INDICATED
Locator
Number
07 = Lefl against medical advice or patient
discon'tfnued care
08 = Discharged/transferred to home under
care of a Home !V drUjj therapy provider.
09 == AdmiHed as an Inpatient to this hospital
20 »IHxplred (or did not recover - Christian
Science patient)
43 = Discharged to a federal hospital (New
code effective October 1, 2003)
50 = Discharged home with hosptce
51 = i:>!scharged or transferred to hospice medical facility
61 = iOischarged or transferred within this
institution to a hospital-based swing bed (sldiled
care) '
62 = Discharged or transferred lo anInpatient
rehabliilation facility
63 = iJIscharged or transferred to a long term
care hospital
05 = Discharged or transferred to a psychiatric
hospital or psychiatric unit of a hospital (New
code: per C^dS, effeclivedate postponed until
2005.)
DRG Code
78
"Hie condition established after study as being
chiefly responsible for the admission of a
patief»t to the hospital for care.
Must bs entered for all File Types, Right luslllled with leading "
zeros. Leave Blank If you do not calculate a DRG for File TvDa
OP or ED.
47
The total gross charges incurred by the patient
for this visit or hospital stay.
Must be filled In for all File Types. Rlqht jusUlled with leading
zeros. Note: Whole dollars only, rounded, no commas. NOTE;
For File Type IP, MUST BE WITHIN $50 OF THE SUM OF THE
CHARGES IN THE REVENUE CODES.
50a
The Primary Payor, the expected source of
payment for the majority of the charges
associated witii this visit or hospital stay.
00 = Self pay
Total Charges
Payor Code
Musi be fiiled in for all File Types. Right justified with leading
zeros.
01 - Commerclai (Indemnity)
02 ^ HMO
03 = PPO
ARIZONA HOSPITAL DISCHARGE DATA SPECIFICATIONS
UPDATED MAY 2004
ARIZONA department OF HEALTH SERVICES
Pages OF 11
Number of
Characters
Slart
End
Required
" For File
Data Elemonf Name
Type
Uniform CODhS AND VALUES - ALL ALPHA CHARACTER
EDIT REQUIREMENTS • LEFT JUSTIFY AND LEAVE UNUSED
Billing
MUST BE IN UPPERCASE
SPACES BLANK UNLESS OTHERWISE INDICATED
Locator
Nuntber
04 = AI^CCCS Health Care Group
05 = iV!edlcare
06 = AHCCCS^Medlcatd
07 CHAMPUS/MEDI=XCEL
08 = Children's Rehab Services
09 - Workers Compensation
10 = Indian Health Services
11 = Medicare Risk
12 = Charity
13 = Foreign National
14 = Other
'
15 = Tobacco Tax
Revenue Codes
All Inclusive Rate
-6
6
6
•
6
6
6
6
6
6
6
6
6
6"
6
6
6'
6
6
6
160
172
178
184
190
196
202
208
214
220
226
232
238
244
250
258
262
28B
274
171
177
183
189
195
201
207
42
10X
Room arid board - private
Room and,board - Iwo bed
11X
Room and board- 3/4 bed
Private (deluxe)
Room and board - ward
Other room and board
Nursery
13X
249
255
IP Only
IP Only
P Only
IP Only
fP Only
IP Only
IP Only
IP Only
IP Only
IP Only
tP Only
IP Only
tP Only
IP Only
fP Only
261
!P Only
267
IP Only
Medical/Surgical supplies
Oncology
273
IP Only
213
219
225
231
237
243
279
H
Leave of Absence
Intensive Care
Coronary Care
Special charges
Incremental charges
All Inclusive ancillary
Pharmacy
IV Therapy
DME (other than renal)
IP Only' Laboratory
For each Revenue Code Category, enter the
charges Incurred by the patient for this Inpatient
hospital stay.
Edit Requirements for Revenue Codes: Must be entered for
File Type IP. Right justified with leading zeros. Fill with zerosIf
item does not apply. Leave Blank for File Types OP and ED.
See Edit Requiremenls for Revenue Codes Above
12X
14X
15X
16X.
17X
18X
20X
21X
22X
23X
24X
25X
26X
27X
28X
28X
30X
ARIZONA HOSPITAL DISCHARGE DATA SPECIFICATIONS
-—
—]See
Edit Requirements for Revenue Codes Above
See Edit Requirements for Revenue Codes Above
See Edit Requirements for Revenue Codes Above
See Edit Requirements for Revenue Codes Above
See Edit Requirements for Revenue Codes Above.
See Edit Requirements for Revenue Codes Above
See Edit Requirements for Revenue Codes Above
See Edit Requirements for Revenue Codes Above
See Edit Requirements for Revenue Codes Above
See Edit Requiremenls for Revenue Codes Above
See Edit Requirements for Revenue Codes Above.
See Edit Requirements for Revenue Codes Above
See Edit Requirements for Revenue Codes Above
See Edit Requirements for Revenue Codes Above.
See Edit Requiremenls for Revenue Codes Above.
See Edit Requirements for Revenue Codes Above.
See Edit Requirements for Revenue Codes Above
See Edit Requirements for Revenue Codes Above
UPDATED MAY 2004
ARIZONA DEPARTMENT OF HEALTH SERVICES
Rags 6 OF 11
Required
For FHe
Type
280
288
Data Element Name
p Only
291
Laboratory ftaihology
iP Only" Radiology - diagnostic
iP Only Radiology - Iherapeutlc
IP Only Nuclear Medicine
IP Only CT scan •
IP Only Operating room
;-J16
321
IP Only
Anesthesia
322
328
3^4
340
346
327
333
IP Only
iP Only
iP Only
IP Only
Blood
352
357
345
351
364
376
382
388
394
CODKS AND VALUES - ALL ALPHA CHARACTERS EDIT REQUIREMENTS - LEFT JUSTIFY AND LEAVE UNUSED
MUST BE tN UPPERCASE
SPACES BLANK UNLESS OTHERWISE JNDJCATED
375
381
387
393
399
See Edit Requirements for Revenue Codes Above,
See Edit Requirements for Revenue Codes Above.
See Ed t Requirements for Revenue Codes Above.
See Ed t Requirements for Revenue Codes Above.
See Ed t Requirements for Revenue Codes Above.
See Ed t Requirements for Revenue Codes Above.
See Edit Requirements for Revenue Codes Above.
See Ed t Requirements for Revenue Codes Above.
Blood storage/processing
Other Imaging
Respiratory services
Physical therapy
See Edit Requirements for Revenue Codes Above.
See Ed Requirements for Revenue Codes Above.
See Edit Requirements for Revenue Codes Above.
See Ed Requirements for Revenue Codes Above.
JP Only
iP Only" Occupational therapy
IP Only (speech therapy
IP Only ]Emergency room
See Ed t Requirements for Revenue Codes Above.
Requirements for Revenue Codes Above.
See Ed Requirements for Revenue Codes Above,
See Ed Requirements for Revenue Codes Above.
See Ed Requirements for Revenue Codes Above.
See Ed Requirements for Revenue Codes Above.
iP Only Pulmonary function
}P Only (Audioiogy .
IP Only' I Cardiology '
IP Only ISpeclal Ambulatory Care
IP Only [Outpatient Services
IP Oniy Clinic
See Ed Requirements for Revenue Codes Above,
See Ed t Requirements for Revenue Codes Above.
)Ed Requirements for Revenue Codes Above.
400
408
412
418
424
430
436
405
411
442
447
453
459
465
471
IP Only
IP Only
483
IP Only Home therapy services
IP Only jHosplce service
64X
495
501
IP Onfy jResp?{e care (HHA Only)
P Only jCast room,
66X
70X
460
466
472
478
417
435
441
484
490
496
IP Only
Free-Standing Clinic
IP Only Osteopathic services
IP Only_jAmbulance
IP Only "jSkliied Nursing
IP Only Medical social services
iP Only {Home heallh aide (home health)
jOther visits "(home health)
Units of servfce (home heallh)
IP Only [Oxygen (home health)
IP OnlyjMRI
IP Only |Med/Surg(Ext.of27X)
IP Only Drugs req. specltlc Id
BOX
61X
62X
63X
65X
ARIZONA HOSPITAL DISCHARGE DATA SPECIFICATiONS
See Ed Requirements for Revenue Codes Above.
See Ed Requirements for Revenue Codes Above.
Requirements for Revenue Codes Above.
See Edit Requirements for Revenue Codes Above.
See Ed t Requirements for Revenue Codes Above.
See Ed t Requirements for Revenue Codes Above.
See Ed Requirements for Revenue Codes Above.
See Ed Requirements for Revenue Codes Above.
Requli^ements for Revenue Codes Above.
t Requirements for Revenue Codes Above.
See Ed t Requirements for Revenue Codes Above.
See Ed t Requirements for Revenue podes Above.
See Ed t Requirements for Revenue Codes Above.
See Ed t Requirements for Revenue Codes Above.
See Ed Requirements for Revenue Codes Above.
See Ed t Requirements for Revenue Codes Above.
UPDATED MAY 2004
ARIZONA DEPARTMENT OF HEALTH SERVICES
'
Page 7 OF 11
Number pf
Characters
!
6
' 6
5
6
8
s
Start
End
Required
For File
Type
507
513
519
525
5^ 531
532
538^ 543
544 549
550 555
IP Only
IP Only
IP Only
IP Only
IP Only
.502
508
514 .
520
8
6
6
5
6
556
552
561
567
IP Only
IP Oniy
IP Only
IP Only
IP Only
5
6 -
574
6
580
Data Element Name
Recovery room
Labor/Delivery
EKG/ECG
EEG
.
"
Gastro-lntesllna' services
Trealment/observadon room
Preventative care services
Lithotripsy
Inpatient renal dialysis
Organ acquisition
Haematolysis - outpatient or
home
Peritoneal dialysis - outpatient
or home
^Jnlform CODES AND VALUES; - ALL ALPHA CHARACTERS
EDIT REQUIREMENTS - LEFT JUSTIFY AND LEAVE UNUSED
BHHna
MUST B£ (N UPPERCASg
SPACES BLANK UNLBSS OTHERWISE INDICATED
Locator
Number
71X
72X
73X
74X
75X
See Edit Requirements for Revenue Codes Above
See Edit Requirements for Revenue Codes Above
See Edit Requirements for Revenue Codes Above
See Edit Requirements for Revenue Codes Above
76X
77X
79X
SOX
81X
82X
See Edit Requirements for Revenue Codes Above
,See Edit Requirements for Revenue Codes Above
See Edit Requirements for Revenue Codes Above
See Edit Requirements for Revenue Codes Above
See Edit Requirements for Revenue Codes Above.
See Edit Requirements for Revenue Codes Above.
a3X
See Edit Requirements for Revenue Codes Above.
84X
See Edit Requirernpnte for Revenue Codes Above.
peritoneal dialysis (CAPD) outpatient or home
85X
See Edit Requirements for Revenue Codes Above.
dialysis (GCPD) - outpatient or
home
6
586
591
597
603
609
615
IP Only
IP Only
IP Only
IP Only
IP Only
822
627
,IP Only
6
628
633
IP Only
Other (herspeut/c services
Professional fees (96X)
Professional fees {97X)
Professional fees (9aX)
6
634
6
22
646
639 ^IPOnly
645
IP Only
667
IP Only
Patient convenience Items
Ail other charges
Attending Physidan Name
8
6
598
504
610
6
6
6
Mlsceiianeous dialysis
Psychiatric treatment
Psychiatric services
Other diagnostic services
88X
90X
91X
92X
94X
96X
97X
98X
99X
See Edit Requirements for Revenue Codes Above
See Edit Requirements for Revenue Codes Above
See Edit Requirements for Revenue Codes Above
See Edit Requirements for Revenue Codes Above
See Edit Requlremenis for Revenue Codes Above
See Edit Requlremenls (or Revenue Codes Above
See Edit Requirements for Revenue Codes Above
82
See Edit Requirements for Revenue Codes Above
See Edit Requirements for Revenue Codes Above
Attending physician's name. Last name, one
Left Justified. No commas or other punctuation. Must be entered
space, first name, one space, and middle initial. for File Type IP. Leave Blank for File Types OP and ED,
Hyphenated names are acceptable.
Attending Physician Stgte
License Number
02
AUendlng pliyslclan's Arizona License Number
Slate Licensing Board
n/a
state Licensing Board Codes:
Must be entered for Pile Type IP, All digits ivlust be filled In, Riglit
justified with leading zeros. Fill with zeros if unl^nown. Can be
Alpha-numerlo, Leave Blanit for File Types OP and ED
Must be fined in for File Type iP, Leave Blank for File Types OP
and ED.
ARIZONA HOSPITAL DISCHARGE DATA SPECIFICATIONS
UPDATED MAY 2004
ARIZONA DEPARTMENT OF HEALTH SERVICES
Page 8 OF 11
["sTrt"J——^
Characters
' uata Elament Name
For File
Type
Uniform CODES AND VALINES • ALL ALPHA CHARACTERi
EDIT f^EQUIREfflENTS - LEFT JUSTIFY AND tEAVE UNUSED
Billing
MUST BE IN UPPERCASE
SPACES BLANK UNLESS OTHERWISE INDICATED
Locator
Number
1 - Me.i^ical Examiners
2 = Dental Examiners
'
3 = Podiatry Examiners
'
4 Osteopathic Examiners
5 = Nursing
IP Only
704
/U4
iP Only
B ^ other
"
Operating or Other Physician
Name
83
Primary procedure physician's OR other
IMust be tilted In for File Type iP If a Procedure was, dnns No
practlfionsr's name. Last name, one space, first commas or other punctuation. Left lustiiled. Leave Blanl<If no
name, one space, and middle Initial.
Procedure was done. Leave Blank for File Types PP and ED.
Hyphenated names are acceptable.
Operating or Other Physician
Stale License Number
83
Physician OR other practilioner's /^Izona
License Number who performed the primary
procedure.
Stats Licensing Board
n/a
State Lioenslna Board Codes:
1 = Medical Examiners
2 = Dental Examiners
3 = Podiatry Exarhlners
4 = Osteopathic Examiners
5 =: Nursing
9 •= Olher
19
Indicates tha priority {type) of admission:
1 = Emergency
2. = Urgent •
Type of Admission
Must be entered for File Type IP Ea Procedutesasjcang Aii
digits must be filled in. Right Justified with leading zeros. Can be .
Alpha-numerlo. Fill with zeros If unitnown. Leave Blanl< If no
Procedure was done. Leave Blank for Fiie Types OP and ED
Must be entered for File Type IP If a Procedure was dnne Leave
Blank If no Procedure was done. Leave Blank for File Types OP
and ED.
Must be filled In for File Type IP. If 4 ((Newborn), see SOURCE
OF ADMISSIOI^ below. Leave Blank fbr File Types OP and ED
3 - Elective
4 = Newborn
5-
705
/Ob
Observation
9 = Information not available
!P Only
Source of Admission
20
indicates the source of admission - adults and
pediatrics:
1 = Physician referral
2 = Clinic referral
Must be entered for File Type IP. Leave Blank for File Types OP
and ED.
.
,
3 - HMO/AHCCCS health plan referral
4 = Transfer fr^im a hospital
5 = Transfer from SNF\
a = Transfer from another health care faciilly
(other than acute cars or SNF)
7 = Emergency room
ARIZONA HOSPITAL DiSCHARGE DATA SPECIFICATIONS
UPDATED MAY 2004
ARIZONA DEPARTMENT OF HEALTH SERVICES
Page 9 OF 11
Number of
Characters
SlaH
End
Required
For File
Type
Data Element Name
Uniform CODES AND VALUES - ALL ALRHA CHARACTER!
EDIT REQUIREMENTS - LEFT JUSTIFY AND LEAVE UNUSED
Billing
MUST BE IN UPPERCASE
SPACES BLANK UNLESS OTHERWISE INDICATED
Locs'lor
Numiier
8 - Court/Law Enrorce/nent
9 = Information nof availablu
NOTU; IF TYPE OF ADMISSION = Newborn "
(4), use;
1 = Normal Delivery
2 « Premature Delivery
3 = yick baby
>
4 s Extramural birth
IP-OP-ED Principal Diagnosis Code
Second Diagnosis Code
6
718
6
724
6
730
6
6
^ 6
736
742
748
e.
754
6
760
723 IP-OP-^ED Third Diagnosis Code
729 IP-OP-ED Fourth Diagnosis Code
735 IP-OP-ED Fifth Diagnosis Code
741 IP-OP-ED
747 IP-OP-ED Seventh Diagnosis Code
753 IP-OP-ED Eighth Dtagnbsis Code
759 • IP-OP-ED Ninth Diagnosis Code
765 IP-OP-ED External Cause of Injury
67
63
69
70 '
71
73
74
75
77
9 « Information not available
Enter the iCD-9-CM code describing the
condition chiefly responsible for causing this
encounter.
iVlust be filled in for all File Types,Include letter V code if
applicable. DO NOT place E-Codes In this field, EXCLUDE
DECIMAL POINTS. (Note: iF Principle Diagnosis Code Is 800 0
through 999.9, see EXTERNAL CAUSE OF INJURY below )
Enter the ICD-9-GM code describing additional
Leave blank if not applicable. Othenvise, must be filled in for all
conditions.
File Types. Include letter V code If applicable. DO NOT place ECodes in this field. EXCLUDE DECIMAL POINTS.
Same as the Second dlaanosis code.
Same as (he instructions for the second diagnosis code
Same as the Second dlapnosls code.
Same as the Instructions for the second diagnosis code.
Same as the Second diagnosis code.
Same as the instructions for llie second diagnosis code.
Same as the instructions for the second diagnosis code.
Same as the Second diagnosis code.
Same as the instructions for the second diagnosis code.
Same as the Second diagnosis code.
Same as the instructions for the second diagnosis code
Same as the Second diagnosis code.
Same as the Instructions for the second diagnosis code.
Enter the ICD-9-CM E-Code describing the
IF the PRINCIPLE DIAGNOSIS CODE above equals 800,0
external cause oflnjury.
through 999.9, THEN an E-Code should be enteredin this
EXTERNAL CAUSE OF INJURY Field. The External Cause of
Injury E-Code ranges are E800.0 through EB48.9. and E850.0
through E999.9. Include the letter E. EXCLUDE DECIMAL
POINTS. When there are muitfpie E-Codes in the record, the ECode associated with the PRINCIPLE DIAGNOSIS CODE should
be entered here. NOTE: An E-Code is to be reported only for the
first hospitalization or visit during which the injury, poisoning
and/or adverse effect was diagnosed or treated. SEE ALSO: The
ADDITIONAL EXTERNAL CAUSE OF INJURY 1.2 and 3 Fields
below.
ARIZONA HOSPITAL DISCHARGE DATA SPECIFICATIONS
UPDATED MAY 2004
ARIZONA DEPARTMENT OF HEALTH SERVICES
Page 10 OF 11
Number of
Characters
Start
End
Required
For File
Type
Data Element Nama
Uniform CODES AND VALUES - ALL ALPHA CHARACTERJ
EDIT REQUIREfyiENTS - LEFT JUSTIFY AND LEAVE UNUSED
BHIIng
MUST BE IN UPPERCASE
SPACES BLANK UNLESS OTHERWISE INDICATED
Locator
Number
6
771
8
n/a
Enter theiCD-Q-CM E-Code describing the
Place where the Injury or Poisoning occurred.
80
Enter the month, day and year of the patient's
Principal Procedure wilhout dashes
MMDDYYYY
80
Enter Ihe ICD-9-CM code that identifies the
If Procedure was done for this patient, must be entered for all Flis
principal procedure performed. NOTE: For File
Types. Leave blank If not applicable. EXCLUDE DECIMAL
Types OP and ED only, IF the ICD-9-CM code POINTS.
is NOT available, THEN enter the CPT4 Code
or HCPCS Code.
Principal Procedure Date
5
(Note: See
Codes and
Values)
5
4
790^1 794 IP-OP^ED Third Procedure Code
795 799 IP-OP'ED Fourth Procedure Code
804 IP-OP-ED Fifth Procedure Code
805 809 IP-OP-ED Sixth Procedure Code
010 813
iP Only Newborn Birth Weight
1
314
5
If Procedure was done for this patient, must be entered for'all File
Types. All digits must be filled In (no dashes). If no Procedum
81A
Enter the fcb-9~CM code that identifies the
Leave biani<: if not applicable. Otherwise, must be entered for all
principal procedure performed. NOTE: For File File Types. EXCLUDE DECIMAL POINTS.
Types OP and ED only. IF the ICD-9-CM code
Is NOT available. THEN enter the CPT4 Code
or HCPCS Code.
>
SIB
81C
81D
81E
n/a
Same as Second procedure code.
Same as Second procedure code.
Same as Second procedure code.
Same as Second procedure code.
Birth weight In grgms.
n/a
1 - yes
2 = no
(Note: See
Codes and
Values)
5
6
For Plaos of Injury Code Only, The Place of InJury.Code Range Is
EB40.O through E849.g. Must be entered (or all Fife Types
Include the letter E. EXCLUDE DECIMAL POINTS. Leave blank If
no EXTERNAL CAUSE OF INJUt^Y
Same as instructions for the Second procedure code
Same aa instructions for the Second procedure code.
Same as instructions forth® Second procedure code;
Must be entered for all newborns. See TYPE OF ADMISSION
and SOURCE OF ADMISSION fields above. Right justify and
leave unused Spaces blank. Leave Blank for File Types OP and
ED.
Must be entered for File Type IP. Leave Blank for File Types OP
and ED.
9 = hot recorded
815
844
IP Only
Pallenl name
*Optlonal Additional External Cause Of
for File Injury 1
Types IPOP-ED
12
n/a
ARIZONA HOSPITAL DISCHARGE DATA SPECIFICATIONS
Last name, one space, first name, one space,
and middle Initial. Hyphenated names are
acceptable.
Alpha-numeric
Must be entered for File Type IP. No commas or other
punctuation. Leave Blank for File Types OP and ED.
IF there is an E-code In the paRent record that WAS NOT placed
In the External Cause of Injury field above, IT MAY be entered
here. The E-Code ranges for this field are E800.0 through
E848.9. and E05O.O through E999.9. Include the letter E.
EXCLUDE DECIMAL POINTS. Fill wlUi blanks If not used.
UPDATED MAY 2004
ARIZONA DEPARTMENT OF HEALTH SERVICES
Page 11 OF 11
Number of
Characters
Start
End
Required
For F»e
Type
5
850
854
'Optional Additional External Cause Of
for File Injury 2
Types IPOP-ED
n/a
Alpha-f)umerlc
IF there is an E-code In the patient record that WAS NOT placed
in the External Cause ol Injury,or the Additional Cause of Injury 1
Fields above, IT MAY be entered here. The E-Code ranges for
this field are E800.0 through E848.9, and E850.0 through E999.9.
Include the letter E. EXCLUDE DECIMAL POINTS. Fill with
bianks if not used.
B
855
859
"Optional AddiUona! External Cause Of
for File Injury 3
TypesIPOP^ED
n/a
Alpha-numeric
1
860
B60 IP-OP-ED Type of Record
n/a
1 = Hospital Inpatient
2 « Hospital Outpatient - ALL Non-Emergency
Department Outpatient visits
3 = Hospital Emergency Department
IF there Is an E-code Inthe patient record that WAS NOT placed
in the External Cause of Injury, Additional Cause of Injury 1, or
Additional Cause of injury 2 Fields above, IT MAY be entered
here. The E-Code ranges for this field are E800.0 through
E848.9, and E8S0.0 through E999.9. Include the letter E.
EXCLUDE DECIMAL POINTS.Fiil with blanks If not used.
Must be entered for all File Types.
Data Element Namo
Uniform ] CODES AND VALUES - ALL ALPHA CHARACTERS EDIT REQUIREMENTS - LEFT JUSTIFY AND LEAVE UKUSEO
BIIMng
f;iUST BE IN UPPERCASE
SPACES BLANK UNLESS OTHERWISE INDICATED
Locaior
Nuiriber
ARIZONA HOSPITAL DISCHARGE DATA SUBMISSION FILE NAMING CONVENTION: [Facility ID]JFll0 Type] [Reporting Period]
EXAMPLE: MED1234JP_2004-01
This example tells us that the submission file Is from Facility I\/iED1234 and that it is an Inpatient File for the First Half (Jan-June) of 2004
The list of Facility ID's for each Arizona Hospital is available on our website listed below.
Hospital Discharge Data File Types: Hosplta) Inpatient (iP), Hospital Outpatient Clinic (OP); Hospital Emergency Department (ED)
NOTE*. EACH FILE TYPE WUST BE SUBMiTVED AS A SEPARATEFILE - SEE FiLE NAWimG CONVENTION ABOVE
Fixed Lengtli Record of 860 characters FOR ALL FILE TYPES
FORMAT: ASCII TEXT - ALL ALPHA CHARACTERS MUST BE IN UPPERCASE,
Dala Reporting Requiremenls Pursuant lo Arizona Revised Statues (A.R.S.) § 36-125.05 and Arizona Administrative Code (A.A.C.) Title 9, Chapter 11. Articles 3 & 4
Medium; Compact Disk (CD) or Diskette
Check website for lipdates at: www.hs.state.az.us/plan/crr/index.htm
ARIZONA HOSPITAL DISCHARGE DATA SPEClFtCATiONS
UPDATED MAY 2004
APPENDIX B
University of Arizona Human Subjects Exemption
122
THE UNivERsnvoF
HiTTnart Sxibjects ?Totec!ion Program
h.cqj://'Www-irb-arizotia.ed'U
ARIZONA.
TUCSON ARIZONA
1350 N. Vine Avenue
P.O. Box 245137
T-acson, AZ 85724-5137
(520) 626-6721
22 Maich 2004
Sharon Sweeney Fee, MSN
Advisor: Gerri Lamb, PhD
College of Nursing
PG BOX 210203
BJE;
MIGRATION FOR HEALTH
Dear Ms. Sweeney Fee:
We have received documents concerning your above referenced project. Regulations published by
the U.S. Department of Health and Human Services [45 CFR Part 46.101 (b) (4)] exempt this type
of research, firom review by the Institutional Review Board.
Exempt status is granted with the imderstanding that no furfher changes or additions -will be made
to the procedures followed or to the consenting instrument used, without the review and ^proval
of the Human Subjects Committee and your College or Departmental Review Committee. Any
research related physical or psychological harm to any subject must also be reported to each
committee,
"
Thank you for informing us of your work. If you have any questions conceming the above, please
Gont2ct tiis ofScs.
Sincerely,
Rebecca Dahl, R.N., Ph.D.
Director
Human Subjects Protection Program
cc: DeparimenfColIege Review Committee
APPENDIX C
Data Coding Dictionary
124
Data Coding Dictionary
AzDHS Data
Element
DRG Code;
Study Data Element
Conversion Notes
6 top DRG groups:
DRG codes:
N=
1-35, 209-256,
Neuro/ortiio/trauma
DRG codes
Pulmonary
439-446
P
=
75-102
Cardiac
1-511
C= 103-145
GI/GU
1 per case
G= 146-190,302-352
Women's Health
W= 353-384
Misc. Surg/SMn/Bum
S = 257-284, 461-511 all others were baseline
I = Commercial, HMO
Recoded into 3 dummy
Payor Code:
variables:
PFO
15 different
A
=
AHCCCS
Health
Medicare
elements from
AHCCCS
care grp, Medicaid
private to
Insurance
charity pay
service = 1 for icu/ccu.
Revenue Codes:
Conversion:
icu/'ccu =1 for dummy variable
10 service codes
0 for aU other
diagnostic/therapeutic codes converted
26 diagnostic and
0-11 diagnostic services
into 1 and added for continuous variable 0-15 therapeutic svcs
therapeutic codes
Recoded into one dummy variable:
Admit Type:
er/ucc = 1, all others = 0
Newborn cases dropped
9 item scale
Recoded into one dummy variable:
court/law enforcement
Admit Source:
referral/transfer = 1, er = 0
& not available dropped
9 item scale
Conversion:
Range of 0-8
Diagnosis Codes:
ICD-9 for Principle 8 secondary codes converted into 1, and Primary diagnosis
dropped, using DRGs
and 8 secondary
added for continuous variable
Conversion:
Age:
range from 0 to 106 recoded into 5 groups from 1-20 to 76=> Age = 0 dropped
Conversion:
Procedures:
codes converted into 1 and added for
Range of 0-6
6 possible ICD-9
continuous variable
codes
Length of stay:
range from 0 to 276
Range of 0-11=>
Days over 11 combined to 11=>
93% 11 or less
Over prospective
payment:
Range of 0-11=>
range -31 to 71
Days over 11 combined to 11=>
less than 0 dropped
days, 98% 0-11
see Appendix 1
not used in this study
All other data
elements
APPENDIX D
Arizona Department of Health Services
Hospital Identification
ARIZONA DEPARTMENT OF HEALTH SERVICES
"ARIZONA !
^FACIUTY ID
MED1442
MED2312
MED1864 •
MED0209
MED2157
MED0239
MED0216
MED0219
MED0227
MED0238
MED2615
MED0194
MED0343
MED0253
MED0254
MED0258
MED2124
'MED0205
MED0195
MED 1397
MBD0217
MED0248
MED0201
MED2568
MED0244
MED0225
MED0454
MED0342
MED0240
ARIZONA LICENSED HOSPITALS REQUIRED TO REPORT
DISCHARGE; DATA ON FEBRUARY ISTH & AUGUST 15THEACH YEAR
i/
AKIZONA HEART HOSPITAL
ARIZONA SPINE AND JOINT
ARIZONA SURGICAL HOSPITAL
ARROWHEAD COMMUNITY HOSPITAL
BANNER BAYWOOD HEART HOSPITAL [was Lutheran Heart]
BANNER BAYWOOD MEDICAL CENTER [was Valley Lutheran]
BANNER DESERT MEDICAL CENTER [was Desert Samaritan]
BANNER GOOD SAMARITAN MEDICAL CENTER [was Good Samaritan]
BANNER MESA MEDICAL CENTER [was Mesa Lutheran]
BANNER THUNDERBIRD MEDICAL CENTER [was Thuhderbird Samaritan]
BARIATRIC CARE CENTERS OF ARIZONA IppenedJuly 2003 ]
BENSON HOSPITAL
CARONDELET HOLY CROSS HOSPITAL
CARONDELET ST JOSEPH'S HOSPITAL
CARONDELET ST MARY'S HOSPITAL
CASA GRANDE REGIONAL MEDICAL CENTER
CHANDLER REGIONAL HOSPITAL
COBRE VALLEY COMMUNITY HOSPITAL
COPPER QUEEN COMMUNITY HOSPITAL
CORNERSTONE HOSPITAL OF SE ARIZONA [was Summit Hospital]
DEL E. WEBB MEMORIAL HOSPITAL
EL DORADO HOSPITAL
FLAGSTAFF MEDICAL CENTER
GREENBAUM SURGICAL SPECIALTY HOSPITAL [OpenedMy 2003 ]
HAVASU REGIONAL MEDICAL CENTER
HEALTHSOUTH MERIDIAN POINT REHAB HOSPITAL - SCOTTSDALE
HEALTHSOUTH REHAB. HOSPITAL OF SOUTHERN AZ [1921 W. Hospital Dr]
HEALTHSOUTH REHAB INSTITUTE OF TUCSON [2650 North Wyatt Dr]
HEALTHSOUTH VALLEY OF THE SUN REHAB
www.hs.statei9z.us/plan/crr/lndex,h(m
.
PAGE 1 OF 3
CITY
'hoenix
viesa
Phoenix
Glendale
Mesa
VIesa
Mesa
Phoenix
Mesa
Glendale
Soottsdale
Benson
Nogales
Tucson
Tucson
Casa Grande
Chandler
Globe
Bisbee
Tucson
Sun City West
Tucson
Flagstaff
Scottsdale
Havasu City
Scottsdale
Tucson
Tucson
Glendale
"COUNTY
Vlaricopa
"ilaticopa
Vlarioopa
Maricopa
Vlaricopa
Maricopa
Maricopa
Maricopa
Maricopa
Maricopa
Maricopa
Cochise
Santa Cruz
Pima
Pima
Pinal
Maricopa
Gila
Cochise
Pima
Maricopa
Pima
Coconino
Maricopa
Mohave
Maricopa
Pima
Pima
Maricopa
ZIP
85016
85206
85015
85308
85206
85206
85202
85006
85201
85306
85255
85602
85621
85711
85745
85222
85224
85501
85603
85710
85375
85712
8600)
85251
86403
85260
85704
85712
85304
UPDATED APRIL 2004
ARIZONA DEPARTMENT OF HEALTH SERVICES
' ARIZONA
FACILITY ro
PAGE 2 OF 3
, ARIZONA LICENSED HOSPITALS REQTJIBBD TO REPORT
DISCHARGE DATA Off 1?EBKX)AR¥ 15TH & AUGUST 15TH EACH YEAH
CITY
"COUNTY
ZIP
,,
•
MED0230
IVIED0222
MED0340
MED0565
MED2199
MEDQ245
MED0250
MED0207
MED0483
MED0223
MED0224
MED1574
MED0226
MED0206
MED0246
MED0196
MED0251
MED0203
IVffiD2149
MED0204
MED0228
MED2170
MED2660
MED2590
.MED0192
MED0235
MED0236
MED1724
MED2131
lOHN C LINCOLN HOSPITAL - DEER VALLEY
lOHN C. LINCOLN HOSPITAL - NORTH MOUNTAIN
KINDRED HOSPITAL - PHOENIX
KINDRED HOSPITAL - TUCSON
KINDRED HOSPITAL - SCOTTSDALE
KINGMAN REGIONAL MEDICAL CENTER
KINO COMMUNITY HOSPITAL
LA PAZ REGIONAL HOSPITAL
LOS NINOS HOSPITAL
MARICOPA MEDICAL CENTER '
MARYV LE HOSPITAL MEDICAL CENTER
MAYO CLINIC HOSPITAI.
MESA GENERAL HOSPITAL
MOUNT GRAHAM COMMUNITY HOSPITAL
NAYAPACHE REGIONAL MEDICAL CENTER
NdRTHERN COCHISE COMMUNITY HOSPITAL
NORTHWEST MEDICAL CENTER
PAGE HOSPITAL
PARADISE VALLEY HOSPITAL
PAYSON REGIONAL MEDICAL CENTER
PHOENIX BAPTIST HOSPITAL
PHOENIX CHILDRENS HOSPITAL
PHOENIX MEMORIAL HOSPITAL
PROMISE SPECIALTY HOSPITAL [was Phoenix Specialty] [OpenedMy 2003 ]
A
SAGE MEMORIAL HOSPITAL
SCOTTSDALE HEALTHCARE-OSBORN , •
SCOTTSDALE HEALTHCARE - SHEA
SELECT SPECIALTY HOSPITAL - MESA
SELECT SPECIALTY HOPSITAL - ARIZONA [Phoenix Downtown -1012 E, Willetta]
www.hs.state,az.us/planterr/index.htin
'hoenix
'hoenix
'hoenix
vlaricopa
^laric'opa
>(Iaric6pa
Tucson
Scottsdale
Kingman
Tucson
Parker
Phoenix
Phoenix
Phoenix
Phoenix
Mesa
Safford
Show Low
Wilioox
Tucson
Page
Phoenix
Payson
Phoenix
Phoenix
Phoenix
Phoenix
'ima
Maricopa
Mohave
Pima
LaPaz •
Maricopa
Maricopa
Maricopa
Maricopa :
Maricopa
Graham
Navajo
Cochise
Pima
Coconino
Maricopa
Gila
Maricopa
Maricopa
Maricopa
Maricopa
Apache
Ganado
Scottsdale
Scottsdale
Mesa
Phoenix
Maricopa
Maricopa
Maricopa
Maricopa
85027
85020
85012
85711
85260
86401
85713
85344
85016
85008
85031
85054
85201
85546
85901
85643
85741
. 86040
85032
85541
85015
85016
85007
85006
86505
85251
85260
85201
85006
UPDATED APRIL 2004
ARIZONA DEPARTMENT OF HEALTH SERVICES
* ARIZONA
FACILITY ID
ARIZONA LICENSED HOSPITALS REQUIRED TO REPORT
DISCHARGE DATA ON FEBRUARY I5TH & AUGUST 15TH EACH YEAR
MBD0532 • SELECT SPECIALTY HOSPITAL - PHOENIX [350 W. Thomas Rd]
MED2U7
SELECT SPECIALTY HOPSITAL - SCOTTSDALE
MED0198
SIERRA VISTA REGIONAL HEALTH CENTER
MHD0199 ^ SOUTHEAST ARIZONA MEDICAL (XNTOR
MED2125
ST JOSEPH'S HOSPITAL AND MEDICAL CENTER
MED0234
ST LUKE'S MEDICAL CENTER
MBD0237
TEMPE ST LUKE'S HOSPITAL
MEmi29
TUCSON HEART HOSPITAL
MED0256
TUCSON MEDICAL CENTER
MED0257
UNIVERSITY MEDICAL CENTER
MBD0260
VERDE VALLEY MEDICAL CENTER
MED0241
WALTER 0. BOSWELL MEMORIAI, HOSPITAL
MED2640
WEST VALLEY HOSPITAL MEDICAL CENTER [Opened Jum 2003 ]
MED0243
WESTERN ARIZONA REGIONAL MEDICAL CENTER
MED0193
WHITE MOUNTAIN REGIONAL MliDICAL CENTER
MED2277
WICKENBURG REGIONAL HOSPITAL
MED0247
WINSLOW MEMORIAL HOSPITAL
MED0261
MED0262
MBD2543
YAVAPAI REGIONAL MEDICAL CENTER
YUMA REGIONAL MEDICAL CENTER
YUMA REHABILITATION HOSPITAL [Opened April 2003 ]
PAGE 3 OF 3
CITY
'hoenix
Scottsdale
Sierra Vista
•ouglas
Phoenix
Phpenix
Tempe
Tucson
Tucson
Tucson
Cottonwood
Sun City
Goodyear
Bullhead City
Springerville
"Wiokenburg
Winslow
Prescott
Yuma
Yuma.
"•COUNTY
ZIP
Vlaricopa
Maricopa
Cochise
Cochise
Maricopa
Maricopa
Maricopa
Pima
Pima
Pima
Yavapai
Maricopa
Maricopa
Mohave
Apache
Maricopa
Navajo
Y avapai
Yuma
Yuma
85013
85251
85635
85607
85013
85006
85281
85704
85733
85724
86326
'85351
85338
^442
85938
85390
' 86047
86301
85364
85364
* The AZ FAC_ID is the hospital's peiwaneiit Arizona facility ID number for all discharge
data suhmissions. AZ FAC ID numbers begin with the alpha characters MED (ALL CAPS)
followed by a four-digit number, with no spaces or dashes. E?;ample: MED1234. Check our
Website at: www.hs.state.az.us/plan/orr/index.htm for Data Specification Updates and contact
infonnation.
** Arizona Counties'. Apache, Cochise, Coconino, Oila, Graham, Greenlee, La Paz,
Maricopa, Mohave, Navajo, Pima, Pinal, Santa Cruz, Yavapai, Yuma
www.hs.state.az.us/pian/crriindex.htm
UPDATED APRIL 2004
APPENDIX E
Geocoding Process Description
130
GEOCODING PROCESS DESCRIPTION
Zip code centroids
Origin:
Geographic Data Technology^ Inc. (GDT)
Title: U.S. ZIP Code edition 2003
ESRI Data & Maps CD 2003
Redlanas, Cal.ifornia, USA
This was a nationwide dataset and had to be reselected. A polygon
representing a circle around the centroid of the state with a radius of
700 miles. This polygon was used to clip the nationwide extent to a
more local extent (AZ). The zip code shapefiles do not arrive with x
and y coordinates represented in tabular form so they were processed
with third-party add-on scripts that extracted them. Continent-wide
geographies are represented in degree-minutes units, which are useless
at more local extents because the units are too large. The data had to
be re-projected into a coordinate system coincident with the geocoding
data set (see below).
Geocoding Source files
Origin: U.S. Department of Commerce Bureau of the Census Geography
Publication Date: 2001
Title: (For Example) Pinal County Roads
Edition: Redistricting Census 2000
Geospatial Data Presentation Form: vector digital data
Series Name: TIGER/Line Files
Publication Place: Washington, DC
Horizontal Datum. Name: D North American 1983 HARN
Ellipsoid Name: Geodeti.c Reference System 80
131
These files (for Arizona) are distributed on a county-by-county basis
by the Arizona Land Information System (ALRIS) and by Southern Arizona
Data Service Program http;//sdfsnet.srnr.arizona.edu/. It was decided
to geocode the hospital locations on the county basis to prevent
matches from emerging due to streets in different counties having the
same name.
At least fifteen geocoding passes were required to assemble
the preliminary results of the address-matching. The address matching
process was very successful on the whole.
There were cases where
addresses were not matched and these points had to be entered and
attributed by hand.
For example: Sage hospital in Ganado does not have
an address so an arbitrary point on hwy 264 near town center was
chosen. Sam.e was done for Cobre Valley Community Hospital, 5880 S.
Hospital Dr. not digitized. Placed on N. Hospital Dr. in arbitrary
location. Some missing zip codes were looked up. Fortunately, most
Arizona websites have web pages referring to them and the web
application MapQuest was also useful, as that organization has access
to proprietary value-added addressing sources. Once the address
matching procedure was perfected and examined for quality control,
points representing hospitals were buffered to 30 miles to generate
polygons representing "hospital clusters" or "urban hospitals". The
larger clusters were removed from further analysis, requiring further
reselection/clipping.
132
Hospital Addresses:
2 .dbf files supplied by Sharon Sweeney Fee
The files contained the zip codes of the serving hospitals and of the
patients they served.
Some of the fields in these tables were
eliminated. The remaining fields contained hospital names, addresses
and zip codes, and matching patient- zip codes. These tables were
converted to INFO format so that x-y coordinates could be applied to
them, and brought back to their native dbase format.
A spreadsheet
algorithm was developed to match the list of hospital and patient zip
codes with the list of all zip codes with their coordinates.
Then
Pythagorean geometry was applied to determine the distances between
sets of corresponding zip codes.
Mickie Reed
School of Natural and Renewable Resources
University of Arizona
133
REFERENCES
Aiken, L., Clarke, S., Sloane, D., Sochalski, J., & Silber, H. (2002). Hospital
staffing and patient mortality, nurse burnout, and job dissatisfaction. JAMA (16), 19871993.
Baker, J. (2002). Medicare payment system for hospital inpatients: diagnosticrelated groups. Journal of Health Care Finance 28(3). 1-13.
Basu, J., & Cooper, J. (2000). Out-of-area travel from rural and urban counties; a
study of ambulatory care sensitive hospitalizations for New York state residents. Agency
for Health Care Quality and Research, AHRQ 00-R05. 129-139.
Beesley, K. (1999). Living in the rural-urban fringe: toward and understanding of
life and scale. In: N. Walford, J. Everitt, & D. Napton (Eds.), Reshaping the
Countryside: Perceptions and Processes of Rural Change (pp. 91-104). New York, NY:
CABI.
Blendon, R., Schoen, C., DesRocher, C.. Osborn., Zapert, K., Raleigh, E., (2004).
Confronting competing demands to improve quality: a five-country hospital survey.
Health Affairs (3). 119-135.
Borders, T., Rhorer, J., Hilsenrath, P., & Ward, M. (2000). Why rural residents
migrate for family physician care. The Journal of Rural Health (16). 337-347.
Brossette, S. Sprague, A., Hardin, J., Waites, K., Jones, W., Moser, S. (1998).
Associtaiton rules and data mining in hosptial infection control and public health
surveillance. Journal of the American Medical Informatics Association (5) 4. 373-381.
Buczko, W. (2001). Rural medicare beneficiaries' use of rural and urban
hospitals. The Journal of Rural Health (17). 53-58.
Casey, M. (1999). Rural Managed Care. In: T. C. Ricketts III (Ed.), Rural Health
in the United States (pp. 113-118). New York, NY: Oxford University.
Chen K., Chiang S.,& Leung S. (2003). Migration, family, and risk
diversification. Journal of Labor Economics (21). 353-380.
Clark, S. Savitz, L., & Randolph, R. (1999). Rural Children's Health. In: T. C.
Ricketts III (Ed.), Rural Health in the United States (pp. 150-158). New York, NY:
Oxford University.
134
Cobum, A., & Bolda, E. (1999). The Rural Elderly and Long Term Care. In: T. C.
Ricketts III (Ed.), Rural Health in the United States (pp. 179-189). New York, NY:
Oxford University.
Conner, R., Kralewski, J. & Hillson, S. (1994). Measuring geographic access to
health care in rural areas. Medical Care Review (51). 337-377.
Cots, F., Mercade, L. Castells, X., Salvador, X. (2004). Relationship between
hospital structural level and length of stay outliers implications for hospital payment
systems. Health policy (68). 159-168.
Coward, R. T. & Krout, J. A. Eds. (1998). Aging in Rural Settings. New York,
NY: Springer
Donabedian, A. (1966). The quality of care how can it be assessed? Journal of the
American Medical Association (260). 1743-1748.
Economic Research Service (2003). Measuring Rurality: rural-urban continuum
codes. Retrieved on February 12, 2002 from
w w w.ers-usda. gov/briefing/rurality/RuralUrbCon/
Etheredge, L. (2003). The need for evidence-based health policy to address health
care variations. Health Affairs Web Exculsive. Retrieved on September 17, 2003 from
www.healthaffairs.org
Figueria-McDonough, J. (2001). Community Analysis and Praxis: Toward a
Grounded Civil Society. Philadelphia, PA: Brunner-Routledge.
Glick, H., Orzol, S., Tooley, J. Polsky, D., Mauskopf, J. (2002). Design and
analysis of unit cost estimation studies: how many hospital diagnosis? How many
countries? Health Economics (12). 517-527.
Hart, A.C. & Schmidt. K. (Eds.). (2003) DRG Expert: A Comprehensive
Reference to the DRG Classification system (19"^ ed.). Reston, VA: Igenix.
Hartley, D., Bird, D., Dempsey, P. (1999). Rural Mental Health and Substance
Abuse. In: T. C. Ricketts III (Ed.), Rural Health in the United States (pp. 159-178). New
York, NY: Oxford University.
Hussey, P,. Anderson, G., Osbom, R., Peek, C., McLaughlin, V., Millar, J.,
Epstein, A. (2004). How does the quality of care compare in five countries? Health
Affairs (3). 89-99.
135
lezzoni, L. (1997). Assessing quality using administrative data. Annals of Internal
Medicine(127). 666-673.
lezzoni, L. Ed. (2003). Risk Adjustment for Mearsuring Health Outcomes (3rd
Ed.) Health Administration Press: Chicago, 11.
Institute of Medicine (2001). Crossing the Quality Chasm: a new health care
system for the
century. Washington, DC: National Academy of Sciences.
Institute of Medicine (2002). Guidance for the National Healthcare Disparities
Report. Washington, DC: National Academy of Sciences.
Kline, D., (2003). Push and pull factors in international nurse migration. Journal
of Nursing Scholarship (35). 107-111.
Larsen, J. & Skjoldborg, U. (2004). Comparing systems for costing hospital
treatments the case of stable angina pectoris. Health Policy (67). 293-307
Librero, J., Peiro, S., & Ordinana, R. (1999). Chronic comorbidity and outcomes
of hospital care: length of stay, mortality, and readmission at 30 and 365 days. Journal of
Clinical Epidemiology (3). 171-179.
Lieberman, S., Lee, J., Anderson, T., & Crippen, D. (2004). Reducing growth of
medicare spending: geographic vs patient-based strategies. Health Affairs Web Exclusive
(3). Retrieved on January 17, 2004 from http://www.HealthAffairs.org.
Lishner, D., Larson, E., Rosenblatt, R., & Clark, S. (1999). Rural Maternal and
Perinatal Health . In: T.C. Ricketts III (Ed.), Rural Health in the United States, (pp. 134149). New York, NY: Oxford University.
C. G. N. Mascie-Taylor & G.W. Lasker (Eds.), (1988). Biological Aspects of
Human Migration. Great Britain: Cambridge; 1988.
May, J.H. & Cunningham, P.J (2004). Tough trade-offs: medical bills, family
finances and access to care. Center for Studying Health System Change Issue Brief (85).
Retrieved on June 20, 2004 from www.hschange.org
Mejia A., Pizurki H, & Royston E. (1979). Physician and Nurse Migration:
analysis and policy implications. Geneva: World Health Organization; 1979
Migration and Immunity (2002). Immunology, (suppl)l. 48-54.
Mitchell, P., Ferketich, S., Jennings, B. (1998). Quality health outcomes model.
Image: Journal of Nursing Scholarship. 30(1). 43-46.
136
Mitchell, P., Shannon, S., Cain, K., & Hegyvary, S. (1996). Critical care
outcomes; linking structures, processes and organizational and clinical outcomes.
American Journal of Critical Care (5). 353-363
Mooney, C., Zwanziger, J., Phibbs, C., & Schmitt, S. (2000). Is travel distance a
barrier to veterans' use of VA hospitals for medical surgical care? Social Science and
Medicine (50). 1743-1755.
Muenning. P. (2002) Designing and Conducting Cost-effectiveness Analyses in
Medicine and Health Care. San Francisco, CA: Jossey-Bass.
Needleman, J., Buerhaus, P., Mattke, S., Stewart, M., & Selevinsky, K. (2002).
Nurse-staffing levels and the quality of care in hospitals. New England Journal of
Medicine (346). 1715-1722.
Nemet, G., & Bailey. A. (2000). Distance and health care utilization among the
rural elderly. Social Science and Medicine (50). 1197-1208.
Reinhardt, U., Hussey, P. & Anderson, G. (2004). U.S. health care spending in an
international context. Health Affairs (3). 10-23.
Ricketts, T., Johnson-Webb, & Taylor P. (1998). Definition of Rural: A
Handbook for Health Policy Makers and Researchers. A Technical issue paper prepared
for the Federal Office of Rural Health Policy.
Ricketts, T. (1999). Rural Health in the United States. Oxford University Press,
NY, NY.
Ricketts, T. (2000). The changing nature of rural health care. Annual Review of
Public Health (21). 639-657.
Ricketts, T. (2002) Geography and disparities in health care. In Institute of
Medicine (Ed.), Guidance for the National Health Care Disparities Report (pp. 149-173).
Retrieved on September 7, 2003 from http://www.naD.edu .
Roos, L. Stranc, L., James,R., & Li, J. (1997). Complications, comorbidities, and
mortality: improving classification and prediction. Health Services Research (2), 229236.
Roseman, C. (2003). Cyclical and polygonal migration in a western context. In
P. C. Jobes, W. F. Stinner, & J. M. Wardwell (Eds.), Community, Society and Migration:
Noneconomic Migration in America (pp. 33-46). Lantham, MD: University; 1992.
137
Rosenblatt, R., & Hart, L. (1999). Physicians and Rural America. In: T.C.
Ricketts III (Ed.), Rural Health in the United States (pp. 38-51). New York, NY: Oxford
University.
Schneeweiss, S., & Maclure, M. (2000). Use of comorbidity scores for control of
confounding in studies using administrative databases. International Journal of
Epidemiology (29), 891-898.
Schultz, A. (1997). Identification of needs of and utilization of resources by rural
and urban elders after hospital discharge to the home. Public Health Nursing (14), 28-36.
Schur, C., & Franco, S. (1999). Access to health care. In: T. C. Ricketts III (Ed.),
Rural Health in the United States (pp. 25-37). New York, NY; Oxford University.
Shaw, L., Miller, D. (2000). Defining quality health care with outcomes
assessment while achieving economic value. Topics in Health Information Management,
44-54.
Slifkin, R., Randolph, R., & Ricketts, T. (2004). The changing metropolitan
designation process and rural America The Journal of Rural Health (20,) 1-6.
Stauber, K. (2001). Why invest in rural america - and how? A critical public
policy questions for the 21'^' century. Economic Review, 57 - 87. Retrieved September 7,
2003 from http://vvww.kc.frb.org.
Stephenson R., Matthews Z., & McDonald J.W. (2003). The impact of rural-urban
migration on under two mortality in India. Journal of Biosocial Science (35), 15-31.
Sweeney Fee, S. (2002). Factors related to the hospital readmission of rural
patients after transfer to urban centers for acute cardiac interventions. Masters Thesis,
available at the University of Arizona.
Thomas, J., Bates, E., Hoffer, T., Perkins, A.. Foltz-Murphy, N., Webb, C. (1998).
Interpreting risk-adjusted length of stay patterns for VA hospitals. Medical Care (12),
1660-1675.
United Nations levels and trends of international migration and related policies.
(2003). Population and Development Revue (29), 335-341.
Verran, J.A. (1997). The value of theory-driven rather than problem-driven
research. Seminars for Nurse Managers (5). 169-172
Voss P., Hammer R., & Meier A. (2001). Migration analysis; a case study for
local public policy. Population, Residency and Policy Review (20), 587-603.
138
Wallace, R. & Wallace, R. (1998) Rural-urban contrasts in elder health status. In
R. Coward, & J. Krout (Eds.), Aging in Rural Setting (pp 67-84). New York, NY;
Springer.
Ward C, & Styles, 1. (2003) Lost and found: reinvention of the self following
migration. Journal of Applied Psychoanalytic Studies (5), 349-367.
Wienert, C, & Boik, R. (1994). MSU Rurality Index: development and evaluation.
Research in Nursing and Health (18), 453-464.
Wray, N., Ashton,C., Kuykendall, D., & Hollingsworth, J. (1995). Using
administrative databases to evaluate the quality of medical care: a conceptual framework.
Social Science Medicine (40), 1707-1713.
Yankow J. (2003) Migration, job change, and wage growth: a new perspective on
the pecuniary return to geographic mobility. Journal of Regional Science (43), 483-516.
Younnis, M. (2002). A comparison study of urban and small rural hospitals
financial and economic performance. Online Journal of Rural Nursing and Health Care
(1). Retrieved September 7, 2003, from http://www.mo.org/iournal/issue.s/Vol-3/issue1ounnis.htm
Zhan, C., & Miller, M., (2003). Excess length of stay, charges, and mortality
attributable to medical injuries during hospitalization. JAMA, 14, 1868-1874.
Was this manual useful for you? yes no
Thank you for your participation!

* Your assessment is very important for improving the work of artificial intelligence, which forms the content of this project

Download PDF

advertisement