University of Southern California Dissertations and Theses

The effects of case mix, hospital competition, and managed care penetration on hospital costs and revenues

(USC Thesis Other)

The effects of case mix, hospital competition, and managed care penetration on hospital costs and revenues

PDF

Download

Open document

Flip pages

Download a page range

Download transcript

Copy asset link

Request this asset

Transcript (if available)

Content THE EFFECTS OF CASE MIX, HOSPITAL COMPETITION,
AND MANAGED CARE PENETRATION ON
HOSPITAL COSTS AND REVENUES
Copyright 2002
by
Keon-Hyung Lee
A Dissertation Presented to the
FACULTY OF THE GRADUATE SCHOOL
UNIVERSITY OF SOUTHERN CALIFORNIA
In Partial Fulfillment of the
Requirements of the Degree
DOCTOR OF PHILOSOPHY
(PUBLIC ADMINISTRATION)
August 2002
Keon-Hyung Lee
Reproduced with permission of the copyright owner. Further reproduction prohibited without permission.
UMI Number: 3094352
Copyright 2002 by
Lee, Keon-Hyung
All rights reserved.
®
UMI
UMI Microform 3094352
Copyright 2003 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
Reproduced with permission of the copyright owner. Further reproduction prohibited without permission.
UNIVERSITY OF SOUTHERN CALIFORNIA
The Graduate School
University Park
LOS ANGELES, CALIFORNIA 900894695
This dissertation, w ritten b y
( C E o s i — H Y LBS.
Under th e direction o f Ai.-S.. D issertation
Com m ittee, and approved b y a ll its mem bers,
has been p resen ted to an d accepted b y The
Graduate School, in p a rtia l. fu lfillm e n t o f
requirem ents fo r th e degree o f
DOCTOR O F PHILOSOPHY
o f Graduate Studies
D ate
DISSER TA TIONCOi TEE
>erson
f
Reproduced with permission of the copyright owner. Further reproduction prohibited without permission.
ACKNOWLEDGEMENTS
I would like to express my sincere appreciation for the efforts of my
dissertation committee. Glenn Melnick, the chair of the committee, provided the
opportunity and data for this research. Robert Myrtle helped greatly in providing
intellectual and emotional support throughout this research. Sami Masri provided
a thoughtful feedback.
The deepest and greatest thanks are owed to my family. It is difficult to
express sufficient appreciation for the endless supports from my parents and
parents-in-law who made my study abroad all possible. But most of all to Eunhee
my wife, and Tiffany my daughter, it is impossible to express my appreciation.
Without their support, encouragement and patience, this research could not have
been accomplished.
Reproduced with permission of the copyright owner. Further reproduction prohibited without permission.
TABLE OF CONTENTS
Acknowledgements ......................... ii
List of Tables ......... iv
Abstract ....... ix
1. Introduction .................................. 1
Contents of This Document ..................................................... 7
2. Background and Significance .......................... 8
Measure of Case Mix ......................... 8
Sources of Variation in Hospital Use ...... 12
Unexplained Variation when Using DRGs: Severity of Illness ... 13
Unexplained Variation when Using DRGs: Practice Patterns ...... 15
Managed Care Penetration and Hospital Competition .......... 18
3. Data and Methods .......................................................................... 29
Conceptual Framework ........................................................ 29
Hypotheses ........................................................................... 33
Empirical Model Specification and Estimation .................... 34
Data Sources ............................................................... 37
Construction of the Variables .............. 37
4. Results ............................................................................................ 45
For All Patients ........................................................................ 50
For Medicare ............................................................................ 71
For Medi-Cal ............................................... 77
For Third-Party and All Other Payers .......................... 82
5. Discussions and Implications .......................................................... 88
Bibliography .......................................................................................... 92
Appendices ........ 98
Regression Results for Medicare Patients .................... 98
Regression Results for Medi-Cal Patients ............ 106
Regression Results for Third-Party and All Other Patients .... 114
iii
Reproduced with permission of the copyright owner. Further reproduction prohibited without permission.
LIST OF TABLES
Table 1. Variables Used in This Study ............................................ 36
Table 2. Hospital Characteristics ....... 45
Table 3. Growth Rates in Total Operating Expenses for Hospitals
in Low, Medium and High Competition Categories ...... 46
Table 4. Growth Rates in Total Net Revenues for Hospitals in Low,
Medium and High Competition Categories ..................... 46
Table 5. Growth Rates in Total Operating Expenses for Hospitals in
Low, Medium and High HMO Penetration Categories ... 47
Table 6. Growth Rates in Total Net Revenues for Hospital in Low,
Medium and High HMO Penetration Categories ........... 47
Table 7. Average Hospital CMI in Low, Medium and High Hospital
Competition Categories ....................................................... 48
Table 8. Average Hospital CMI in Low, Medium and High
HMO Penetration Categories ...... 49
Table 9. Means and Standard Errors of Covariates in
Multivariate Regression Models ..................................... 50
Table 10. The Estimated Regression Model for Hospital Operating
Expenses (using HCFA-Weight Hospital Case Mix Index
for All Patients) .............. 54
Table 11. The Estimated Regression Model for Hospital Net
Revenues (using HCFA-Weight Hospital Case Mix Index
for All Patients) ............ 55
Table 12. The Estimated Regression Model for Hospital Inpatient
Operating Expenses (using HCFA-Weight Hospital Case
Mix Index for All Patients) ................................... 56
Table 13. The Estimated Regression Model for Hospital Inpatient
Operating Revenues (using HCFA-Weight Hospital Case
Mix Index for All Patients) ............................................... 57
iv
Reproduced with permission of the copyright owner. Further reproduction prohibited without permission.
Table 14. The Estimated Regression Model for Hospital Operating
Expenses (using All-Patient Cost-Weight Hospital Case
Mix Index) ....... 58
Table 15. The Estimated Regression Model for Hospital Net
Revenues (using All-Patient Cost-Weight Hospital Case Mix
Index) ....... 59
Table 16. The Estimated Regression Model for Hospital Inpatient
Operating Expenses (using All-Patient Cost-Weight Hospital
Case Mix Index) ............................................................. 60
Table 17. The Estimated Regression Model for Hospital Inpatient
Operating Revenues (using All-Patient Cost-Weight
Hospital Case Mix Index) ............. 61
Table 18. The Estimated Regression Model for Hospital Operating
Expenses (using HCFA’s Medicare Case Mix Index) 62
Table 19. The Estimated Regression Model for Hospital Net
Revenues (using HCFA’s Medicare Case Mix Index) .... 63
Table 20. The Estimated Regression Model for Hospital Inpatient
Operating Expenses (using HCFA’s Medicare Case Mix
Index).................................... 64
Table 21 The Estimated Regression Model for Hospital Inpatient
Operating Revenues (using HCFA’s Medicare Case Mix
Index)............................................................... 65
Table 22. Coefficient Differences between Total Operating Expenses
and Total Net Revenues for Models with CMIs Using
HCF A DRG Weights Applied to All Patients (from
Tables 10 and 11)....... 66
Table 23. Coefficient Differences between Inpatient Operating Expenses
and Inpatient Net Revenues for Models with CMIs Using
HCF A DRG Weights Applied to All Patients (from
Tables 12 and 13)............ 66
Table 24. Coefficient Differences between Total Operating Expenses
v
Reproduced with permission of the copyright owner. Further reproduction prohibited without permission.
and Total Net Revenues for Models with CMIs Using
Cost-DRG Weights Applied to All Patients (from
Tables 14 and 15) ......................................................... 67
Table 25. Coefficient Differences between Inpatient Operating
Expenses and Inpatient Net Revenues for Models with
CMIs Using Cost-DRG Weights Applied to All Patients
(from Tables 16 and 17) ................................................... 67
Table 26. Coefficient Differences between Total Operating Expenses
and Total Net Revenues for Models with HCFA’s
Medicare CMIs (from Tables 18 and 19) ................... 68
Table 27. Coefficient Differences between Inpatient Operating Expenses
and Inpatient Net Revenues for Models with HCFA’s
Medicare CMIs (Tables 20 and 21) ............... 68
Table 28. Comparison on R-squares of Total Operating Expenses and
Total Net Revenues Models ................................................. 70
Table 29. Comparison on R-squares of Inpatient Operating Expenses
and Inpatient Net Revenues Models ..................... 70
Table 30. Coefficient Differences between Medicare Operating Expenses
and Medicare Net Revenues for Models with HCFA’s
Medicare CMIs (from Appendices A1 and A2) ................ 74
Table 31. Coefficient Differences between Inpatient Operating Expenses
and Inpatient Net Revenues for Models with CMIs Using
HCF A DRG Weights Applied to All Patients (from
Appendices A3 and A4) ..................................................... 74
Table 32. Coefficient Differences between Medicare Operating
Expenses and Medicare Net Revenues for Models with
Cost-Weight CMI for Medicare Patients (from
Appendices A5 and A6) .................... 75
Table 33. Coefficient Differences between Inpatient Operating
Expenses and Inpatient Net Revenues for Models with
Cost-Weight CMI for Medicare Patients (from
Appendices A7 and A8) ................................................... 75
vi
Reproduced with permission of the copyright owner. Further reproduction prohibited without permission.
Table 34. Comparison on R-squares of Medicare Operating Expenses
and Medicare Net Revenues Models ....... 76
Table 35. Comparison on R-squares of Medicare Inpatient Operating
Expenses and Medicare Inpatient Net Revenues Models ..... 77
Table 36. Coefficient Differences between Medi-Cal Operating
Expenses and Medi-Cal Net Revenues for Models with
CMIs Using HCFA’s DRG Weights Applied to
Medi-Cal Patients (Appendices B1 and B2) 79
Table 37. Coefficient Differences between Medi-Cal Inpatient Operating
Expenses and Inpatient Net Revenues for Models with CMIs
Using HCFA’s DRG Weights Applied to Medi-Cal Patients
(from Appendices B3 and B 4 )................................................. 79
Table 38. Coefficient Differences between Medi-Cal Operating
Expenses and Medi-Cal Net Revenues for Models with
CMIs Using Cost Weights Applied to Medi-Cal Patients
(from Appendices B5 and B6) ........ ................................... 80
Table 39. Coefficient Differences between Medi-Cal Inpatient
Operating Expenses and Inpatient Net Revenues for Models
with CMIs Using Cost Weights Applied to Medi-Cal Patients
(from Appendices B7 and B8) .............................. 80
Table 40. Comparison on R-squares of Medi-Cal Operating Expenses
and Medicare Net Revenues Models .................................. 81
Table 41. Comparison on R-squares of Medi-Cal Inpatient Operating
Expenses and Medicare Inpatient Net Revenues Models ... 81
Table 42. Coefficient Differences between Third-Party Payer/Others
Operating Expenses and Net Revenues for Models with CMIs
Using HCFA’s DRG Weights Applied to Third-Party/All
Other Patients (Appendices C1 and C2) ................... 85
Table 43. Coefficient Differences between Third-Party Payer/Others
Inpatient Operating Expenses and Inpatient Net Revenues for
Models with CMIs Using HCFA’s DRG Weights Applied to
Third-Party/All Other Patients (Appendices C3 and C 4).... 85
Reproduced with permission of the copyright owner. Further reproduction prohibited without permission.
Table 44. Coefficient Differences between Third-Party Payer/Others
Operating Expenses and Medi-Cal Net Revenues for Models
with CMIs Using Cost Weights Applied to Third-Party/
All Other Patients (Appendices C5 and C6) ....... 86
Table 45. Coefficient Differences between Third-Party Payer/Others
Inpatient Operating Expenses and Inpatient Net Revenues
for Models with CMIs Using Cost Weights Applied to
Third-Party/All Other Patients (Appendices C7 and C8) ...... 86
Table 46. Comparison on R-squares of Third-Party Payers and All
Others Operating Expenses and Medicare Net Revenues
Models ............... 87
Table 47. Comparison on R-squares of Third-Party Payers and All
Others Operating Inpatient Expenses and Third-Party
Payers and All Other Inpatient Net Revenues Models ..... 87
viii
Reproduced with permission of the copyright owner. Further reproduction prohibited without permission.
ABSTRACT
Due to competition and managed care, hospitals have argued that the rate
of increase in hospital costs is faster than the rate of increase in hospital revenues.
It is important to pay hospitals based on the patient’s severity of illness that the
hospital treats. However, managed care organization pay hospitals based on
negotiated prices that do not consider the severity of patient’s illness.
The purpose of this paper is to provide a better understanding of those
factors affecting hospital costs and revenues in California using the actual hospital
costs and utilization data for several years. In this study, by developing case mix
indexes (CMIs) using all hospital discharges in California by different payers
(including all payers, Medicare, Medi-Cal and all others), this study attempted to
answer policy- and methodology-related issues including 1) the relationship
between hospital revenues and costs, 2) the differential effects of CMI on hospital
costs and revenues and 3) the performance of different CMIs to predict hospital
costs and revenues.
The major findings are:
1) For the policy purposes, this study showed that the coefficients
for CMIs in total and inpatient hospital revenue models were
greater than those in hospital cost models. Over time, the
differences in coefficients for CMIs in hospital revenue and cost
models become smaller. However, by specifically looking at
ix
Reproduced with permission of the copyright owner. Further reproduction prohibited without permission.
different payers (e.g., Medicare, Medi-Cal), in certain years,
coefficients for CMIs in hospital cost models were greater than
those in hospital revenue models for Medicare, Medi-Cal and
third-party payers. It can be possible to conclude that the rate of
increase in hospital costs was greater than the rate of increase in
hospital revenues.
2) For methodological purposes, when predicting total hospital
costs and revenues for all patients, CMIs using all-patient
weights predict better compared to HCFA’s Medicare CMI.
However, this study also showed that there is no difference in
using the all-payer cost weights or the HCFA’s Medicare
weights when predicting payer-specific hospital expenses and
revenues (for Medicare, Medi-Cal or others).
Reproduced with permission of the copyright owner. Further reproduction prohibited without permission.
1. INTRODUCTION
After the implementation of the Medicare and Medicaid programs in 1965,
health policymakers have been concerned with the rapid growth in health care
costs. In an effort to decrease the rising health care costs, in 1982, the federal
government enacted diagnosis-related groups (DRGs)-based prospective payment
system (PPS). With the DRG-based payment system, the federal government
began to reimburse hospital inpatient services for Medicare beneficiaries based on
the predetermined payment schedule rather than an open-ended payment system.
In 1982, the State of California legislature implemented a selective
contracting law. According to this law, managed care organizations (MCOs) can
selectively contract with “preferred providers” (hospitals, physicians and other
health care providers) who can provide health services at a lower cost compared
to other providers. With the implementation of this legislation, Medi-Cal, the
California’s Medicaid program, as well as third-party insurance companies were
able to selectively contract with these preferred providers. In order to control both
price and utilization, MCOs may ask for discounted prices from providers and
increase utilization review requirements during the contracting process. By
signing the contract, providers can be guaranteed a larger number of patients,
however, without this they would expect a smaller number of patients due to other
Reproduced with permission of the copyright owner. Further reproduction prohibited without permission.
constraints such as higher patient’s cost-sharing including higher deductibles and
higher copayments (Melnick and Zwanziger, 1995).
With the implementation of new policies including DRG-based PPS,
selective contracting, etc., hospitals have faced a very turbulent environment that
is seriously affecting their financial performance and ability to survive. The
increased competitive pressure due to rapid changes in the health care
environment has threatened the hospital industry. Consequently, they have been
struggling to reduce their operating costs and trying to run the organization more
efficiently (Cody, Friss and Hawkinson, 1995).
During the 1990s, one of the biggest changes in the US health care system
has been the rapid increase in managed care enrollment, in both private and public
health insurance programs. In 1996, the proportion of employer-sponsored health
insurance coverage under managed care rose from 49 percent to 77 percent.1 In
addition, the proportion of Medicaid beneficiaries under managed care rose from
9.5 percent (2.7 million beneficiaries) in 1991 to 55.6 percent (17.8 million) in
1999. In 2000, about 17 percent (total 6.9 million) of Medicare beneficiaries were
covered under managed care.
According to the 1996 data, Americans spent 13.6 percent of Gross
Domestic Product (GDP) on health care, which increased from 5.1 percent in
1 Foster Higgins National Survey o f Employer-Sponsored Health Plans, 1996.
2 Managed Care On-Line Information Exchange, 2000.
2
Reproduced with permission of the copyright owner. Further reproduction prohibited without permission.
1960. Particularly in 1996, however, national health expenditures increased only
4.4 percent, the slowest rate of increase since 1960. Annual average growth rate in
health spending between 1993 and 1996 kept within 5 percent, with each
a -i
subsequent year showing slower growth than in any year since 1960. The
increased enrollment in managed care plans is one of the primary reasons for the
slower rate of increase in health spending.
In California, each major third party payer establishes its own hospital
payment schedules. California hospitals, thus, often face conflicting incentives
from a variety of different payment scheme. For instance, Medicare pays hospitals
based on DRG PPS, Medi-Cal (Medicaid program in California) pays hospitals
based on a fixed per diem rate negotiated in beforehand. Blue Cross pays
retrospective costs or a negotiated per diem (depending on the plan), and
commercial indemnity plans pay hospital charges. To maximize profit (or
reimbursement from different payers), each hospital should consider all different
payment schemes.
Differences in demographic characteristics, socioeconomic status, and
health status between payer groups exist because of the structure of third party
coverage in the US. Differences in the design and targeting of public and private
insurance programs result in the segmentation of patients. For instance, Medicare
3 Levit KR, Lazenby HC, Braden BR and the National Health Accounts Team (1998) “National
Health Spending Trends in 1996” Health Affairs, 17(1):35-51.
3
Reproduced with permission of the copyright owner. Further reproduction prohibited without permission.
is designed to cover the elderly and disabled. Medicaid covers low-income single
parent families, and low-income aged, blind and disabled. Blue Cross and
commercial insurance companies offers health insurance programs to the
employed population under 65. Because each payer group serves a distinct
population in terms of age, income level, prevalent medical conditions, health
status, and other factors likely to influence the course of disease and hence
treatment costs, it is reasonable to expect to observe differences in resource use by
different payer group (Mann, 1989).
Many researchers have reported that various factors such as patient
income and socioeconomic status affect hospital costs (Epstein et al., 1988). For
example, Medicare patients are older and have more comorbidities than younger
patients. Although Medicare DRG classification makes some adjustment for age
and complicating conditions, those adjustments are not sophisticated enough.
Medicaid and uninsured low-income patients may be more severely ill due to
health complications arising from difficulties associated with lower
socioeconomic status. Lower provider payments in the Medicaid program result
in limited provider participation rates (especially for physicians). Due to a smaller
physician pool in the Medicaid program, Medicaid beneficiaries may experience
limited access to physicians.
The DRG-based prospective payment system (PPS) was developed based
on the assumption that the majority of patients in each DRG will have a relatively
4
Reproduced with permission of the copyright owner. Further reproduction prohibited without permission.
homogeneous resource consumption. However, intra-DRG resource variation is
known to persist, and there is ongoing debate on the reasons for this observation.
Early research focused on the relationship between hospital characteristics and
hospital costs, and to ensure the equity of reimbursement under the PPS. Effects
of hospital efficiency, teaching facilities, patient mix and regional differences
were among factors that were also investigated (Chulis, 1991).
The issue of clinical heterogeneity within DRGs was subject to evaluation
by the development of more refined patient classification systems. However, the
ability of these systems to explain additional variation in resource use over the
DRG classification was limited and varied considerably across DRGs (Thomas
and Ashcraft, 1989, 1991; MacKenzie et al, 1991).
Many hospitals found Medicare DRG-based PPS to be less problematic
than they expected. By the late 1980s, however, the efforts to control health care
costs by various sectors (including federal and state governments, insurance
companies, and managed care organizations, etc.) have begun to impact the
hospital industry more significantly. By the early 1990s, hospitals found
themselves in a very weak negotiating position due to a decrease in demand for
inpatient services and an increase in managed care enrollment. From 1980 to
1996, the average length of stay declined from 7.6 to 6.2 days, inpatient
admissions declined from 36.4 to 31.1 million, and occupancy rates declined from
76 percent to 63 percent. Thus, hospitals found that they needed to contract with
5
Reproduced with permission of the copyright owner. Further reproduction prohibited without permission.
managed care organizations (MCOs) to ensure their revenue source. Yet, most
MCOs have implemented per diem payment rates, (i.e., a single, all-inclusive
payment rate per patient day) or per case payment rates (i.e., a single, all-inclusive
payment rate per discharge). MCOs also have used DRG-based payment rates.
Under these payment contracts, the hospitals are at risk. If costs exceed payment
rates, the hospital will inevitably lose money (Whetsell, 1999).
Before selective contracting and managed care were introduced, the
amenities-based competition was prevalent. Using the early 1980s data, several
researchers (Robinson and Luft, 1985, 1987; Luft et al., 1986) argued that
hospitals in more competitive areas have higher costs since they have competed
based on amenities. By analyzing the rate of increase in hospital cost before and
after the implementation of selective contacting, Zwanziger and Melnick (1988)
showed that hospitals in more competitive areas had lower rates of increase in
hospital cost. Additional work using the later years of California data confirmed
that hospitals with more competitors have a lower rate of increase in hospital cost
(Zwanziger, Melnick and Bamezai, 1994; Zwanziger, Melnick and Bamezai,
2000). With the introduction of selective contracting, hospitals began to compete
in not only quality and amenities but also price. Robinson (1991, 1996) argued
that managed care penetration, especially for HMOs, contributed to the lower rate
of growth in cost per admission and overall hospital expenditures.
6
Reproduced with permission of the copyright owner. Further reproduction prohibited without permission.
The purpose of this paper is to provide a better understanding of those
factors affecting hospital costs and revenues in California using the actual hospital
costs and utilization data for several years. By developing CMI using all hospital
discharges in California by different payers (including all payers, Medicare,
Medi-Cal and all others), I will investigate the impact of different CMIs on
hospital costs and revenues, controlling for hospital competition and managed
care penetration.
Contents of This Document
Chapter 2 presents the background and significance of this study by
focusing on the case mix index measurement and those factors affecting hospital
costs and revenues including hospital competition and managed care. Chapter 3
presents the study methodology, including a conceptual model of hospital costs
and revenues, data used in the study, description of variables used, and analytical
methods. Empirical results are shown in Chapter 4. Finally, empirical results and
theories are synthesized in Chapter 5.
Reproduced with permission of the copyright owner. Further reproduction prohibited without permission.
2. BACKGROUND AND SIGNIFICANCE
Measures of Case Mix
There have been a number of studies on examining differences in hospital
resource use by payment source (Greenlick, Freeborn, Columbo et al., 1972;
Rabin and Schach,1975; Studnicki,1979). Before the development of diagnosis-
related groups (DRGs), researchers employed bed size (Lave and Lave, 1971) and
the availability or number of services or facilities provided by hospitals (Jeffers
and Siebert 1974). Measurement of medical facilities and services offered by a
hospital provided no information about the utilization of such services and the
distribution of patients across these services.
Utilizing diagnostic data can alleviate these problems. Researchers have
aggregated ICDA diagnosis codes into diagnostic categories to represent different
case types. For instance, Feldstein (1968) made 28 diagnostic categories, Evans
(1971) made 41, and Lave, Lave, and Silverman (1972) made 17 categories. Each
researcher then measured the proportion of hospital admissions falling into each
of these categories.
DRGs represent another approach for aggregating diagnostic data. DRGs
use a statistical technique to classify patients into over 500 categories, which are
mutually exclusive and collectively exhaustive, based on diagnosis depending on
performance of surgical procedures, presence of complications or comorbidities,
8
Reproduced with permission of the copyright owner. Further reproduction prohibited without permission.
and age. The hospital specific case mix can be then calculated by the summation
of the relative intensity (as measured by length of stay) of each DRG for a
hospital divided by the total number of admissions for that hospital.
By employing the 1976/77 Maryland hospital discharge data, Studnicki
(1979) compared the average length of stay (ALOS) between Medicaid and Blue
Cross patients in 83 major diagnostic categories (MDCs) and found that Medicaid
patients had a longer ALOS in 69 of the 83 MDCs, and that the gap widened in
MDCs with higher average charge per case.
Martin, Frick, and Shwartz (1984) examined differences in average costs
between payers in the state of New York by using the 1978 hospital discharge
data from 28 hospitals (total 296,000 discharges). Aggregating across DRGs, on
average, Medicaid patients had a higher cost than Blue Cross patients.
Epstein, Stem, Tometti et al. (1988) conducted a study to identify patients
of low socioeconomic status (e.g., income, education, job prestige) have longer
LOS and/or higher charges than those of higher socioeconomic status (SES) by
using 402 admissions to a single hospital for connective tissue disorders (DRGs
240 and 241) from 1981 to 1985. Authors concluded that the differences between
patients of low SES and those of high SES were as big as 25 percent for LOS and
16 percent (though statistically insignificant) for charges.
Thorpe (1987) compared relative DRG weights set by HCF A for Medicare
patients with payer-specific relative weights for four major payer groups
9
Reproduced with permission of the copyright owner. Further reproduction prohibited without permission.
(Medicare, Medicaid, Blue Cross, and all remaining categories). After regressing
HCFA’s Medicare DRG weights on the payer-specific relative weights for each
payer in New York, Thorpe concluded that there exists systematic variations in
the average cost per discharge within DRGs for different payer categories since
the regression coefficient for both Blue Cross and Medicaid weights differed
significantly from one (while that for New York’s Medicare weights equaled
one). By comparing cost per case by payer group for each of 20 large DRGs, the
author concluded that Medicaid patients typically incur higher per case costs.
Dor and Farley (1996) investigated the capacity of hospitals to vary the
intensity of their services based on patients' expected sources of payment. While
the concept of price discrimination by hospitals based on payer generosity has
been discussed extensively, the notion that hospitals can adjust payer-specific
marginal costs to reflect differences in reimbursement policies has not been
studied in depth. To examine this issue, Dor and Farley employed a multiproduct
cost function with hospital outputs defined as admissions by payment source,
controlling for the distribution and severity of illness for each payer. Marginal
costs of casemix-adjusted discharges are obtained and compared for Medicare,
Medicaid, Private Payers, and a residual category that includes uncompensated
care. Authors found that payer-specific marginal costs generally reflect payer
generosity.
10
Reproduced with permission of the copyright owner. Further reproduction prohibited without permission.
Systematic variation in patient resource use and cost not captured by
diagnosis-related groups (DRGs) can be a significant shortcoming of a DRG-
based prospective payment system. The underlying assumption of the DRG-based
case-mix classification system is that patients in the same DRG consume
relatively similar amounts of resources. If resource use varies systematically by
factors ignored by the classification system, and if patients exhibiting these
factors are unevenly distributed across hospitals, the reimbursement system may
unfairly penalize hospitals with a larger proportion of such patients. These
hospitals will be underpaid relative to the average resource requirements of their
patients. In the long run, these hospitals will suffer from financial pressures due to
their mix of patients, not because they are more inefficient (Melnick, Mann and
Serrato, 1989).
If the distribution of varying severity levels were uniformly random across
all hospitals, the effects of severity on payments to providers might be expected to
balance out over time. However, regional differences in severity may be
systematic and attributable to local conditions resulting in the inefficient
allocation of resources among hospitals or other providers. In the current DRG-
based Medicare payment system, the hidden incentive is that providers can be
rewarded by tailoring service provision to avoid the admission and treatment of
severely ill patient cases or high risk population (Forgione and D’Annunzio,
1999).
11
Reproduced with permission of the copyright owner. Further reproduction prohibited without permission.
Using a full year (1984 and 1985) of Medicare bills for each hospital
under prospective payment system, Thorpe, Cretin and Keeler (1988) found that
the relative prices of low-cost DRGs are too high and high-cost DRGs are too low
relative to their true costs. They argued that under the current prospective
payment system, hospitals with relatively high case mix index (CMI) may be
underpaid and those with lower-CMI be overpaid.
Sources of Variation in Hospital Use
The residual variation in resource use that remains after controlling for
DRG may stem from several resources: 1) patient-related factors, such as
unmeasured differences in underlying health status (severity of illness) or
differences in psychosocial needs; 2) provider-influenced factors, such as
differences in hospital efficiency or physician practice patterns; 3) market area
factors such as differences in the availability of services provided outside of
hospitals and physician-to-population ratios (Mann, 1989).
A variation in observed lengths of stay (LOS), costs and charges within a
DRG can be occurred due to several reasons. First, some of the DRGs are not
clinically specific and they include patients with quite different medical
conditions and consequently different treatment requirements. (Some of this
variation stems from the ICD9-CM coding system, which occasionally groups
quite different patients into the same five-digit code.) Second, there is often
12
Reproduced with permission of the copyright owner. Further reproduction prohibited without permission.
considerable discretion in treatment of patients with similar clinical conditions.
Third, there is a significant amount of inappropriate care provided and to the
extent that the amount of inappropriate care rendered varies across hospitals. This
will consequently lead to increased variation with a DRG. Finally, most existing
data on hospital utilization is based on data that is at least 3 years old. These data
are built on clinical information that is frequently reported incorrectly on hospital
bills. The Institute of Medicine suggests that there is a 20 percent error rate in the
DRG principal diagnosis for the HCF A data base, and thus variation exists within
a DRG because patients are falsely coded as belonging in that DRG. (Frank and
Lave, 1985)
Unexplained Variation when Using DRGs: Severity of Illness
There are many possible definitions of severity of illness. The clinician’s
(both physicians and nurses) definition of a more severely ill patient, for example,
implies that the patient is more physiologically compromised. However, managers
and payers define the severity of illness in terms of resource consumption (Smits
et al., 1984).
Variation in patient treatment needs not captured by DRGs poses a
problem for PPS, of it is unevenly distributed across patients, hospitals, or groups
of hospitals. If high cost patients within a DRG can be systematically identified at
the time of admission, that patient may be denied access to some hospitals.
13
Reproduced with permission of the copyright owner. Further reproduction prohibited without permission.
Eisenberg (1984) insisted that case mix adjustment measures (such as DRGs)
“emphasize the point that illness severity and appropriate patient treatment vary
greatly across hospitals and patient categories.”
Using the 1981 HCFA dataset and the three years (1979 through 1981) of
Maryland Medicare data, Frank and Lave (1985) compared medical, surgical and
psychiatric DRGs to determine their relative homogeneity. Frank and Lave (1985)
found that almost 77 percent of the surgical patients were in DRGs with
coefficients of variation between .25 and .79.
DRGs do take account of severity of illness since DRG developers clearly
attempted to by assigning DRGs on the basis of surgical procedures,
comorbidities, complications, and in some cases age and sex. Although DRGs do
take account of severity, the more crucial question lies in whether DRGs adjust
for severity of illness consistently enough (Eisenberg, 1984).
Some authors have examined admissions generated by the hospital
emergency department as a proxy for patient severity. Munoz et al. (1985)
conducted a study comparing the cost of admissions from hospital emergency
rooms (ERs) with a matched sample of non-ER admissions within the same DRGs
using 11 hospitals in New York City. The authors found that the cost of
emergency admission exceeded the cost of non-emergency admission within the
same DRG in more than 70 percent of the DRGs.
14
Reproduced with permission of the copyright owner. Further reproduction prohibited without permission.
Coulton et al. (1985) examined the costliness of medical intensive care
unit (MICU) versus routine care for patients across 13 DRGs. In 10 of the 13
DRGs, MICU costs were significantly greater than those for routine care, and the
costs of MICU patients exceeded estimated payment rates for all 13 DRGs.
Average loss per patient using the MICU was $1,795 whereas average profit for
routine care patients was $337 per patient. Thus, the authors concluded that the
costs of MICU care relative to PPS reimbursement could lead to hospital
decisions to limit admission of patient requiring intensive care or to reduce the
supply of intensive care beds.
Butler, Bone, and Field (1985) restricted the sample for their study of
MICU treatment costs to Medicare patients. The authors found that the average
loss per patient in this group was $10,567; and for the 28 percent of these patients
who died, a $21,651 loss per discharge. The authors concluded that hospitals in
financial difficulty may find it necessary to “decrease or discontinue provision of
medical intensive care units and other types of high technology care to severely ill
patients.
Unexplained Variation when Using DRGs: Practice Patterns
Various studies reviewed by Eisenberg (1986) focused on differences in
practice pattern from the medical and financial aspects. He found that physician’s
age, gender, medical specialty, post-graduate training, years in practice and
15
Reproduced with permission of the copyright owner. Further reproduction prohibited without permission.
personal characteristics (e.g., values, tastes, and preferences for certain types of
clinical decision making) interact with practice setting and financial incentives to
influence both the technical and interpersonal dimensions of care (Eisenberg,
1986). Researchers in this area have investigated the variations across the country,
within small areas, and among individual physicians (Chassin et al., 1986;
Holahan et al., 1990; Welch et al., 1993; Wennberg and Gittelsohn, 1982;
Wennberg et al, 1989; Feinglass et al., 1991). Because variable hospital costs are
largely determined by physicians’ diagnostic and treatment order, a greater
understanding of physician practice variation is needed.
Carter et al. (1994) analyzed the types of non-Medicare payers that use
DRG-based payment system. They found that the magnitude of the DRG use
significantly varied by type of payer and by geographical area. For the state
Medicaid programs, twenty-one states out of 51 states had adopted a DRG-based
payment system. Furthermore, 67 percent (37 of 55) of the responding Blue Cross
Blue Shield Association member plans use DRGs for at least one of their
insurance products. Although DRGs are used in all regions, they are most likely
to be used in the Mid-Atlantic States and are least likely to be used in New
England and the Southern Atlantic States.
Oday and Dobson (1990) argued that since the implementation of PPS in
1983, there has been no noticeable reduction in quality of care. However, the
figures show that the hospital’s operating margins for Medicare patients have
16
Reproduced with permission of the copyright owner. Further reproduction prohibited without permission.
declined significantly, going from 14.7 percent in 1983 and declining to -6.6
percent in 1990. They insisted that the hospital’s operating deficits are a systemic
feature of the Medicare payment system, and not an isolated problem confined to
a few hospitals or regions.
If hospitals treat a random selection of patients within the DRG, then the
law of large numbers assures that, on average, their revenues will equal costs. An
important implication of heterogeneity is that it opens up the possibility for
systematic risk in the form of nonrandom selection of hospitals by patients and
patients by hospitals. If hospitals treat patients within DRGs that are
systematically more or less costly than average, then the law of large numbers
does not guarantee that costs will equal revenues. Nonrandom selection is most
problematic in DRGs that contain heterogeneous patients. (Frank and Lave, 1985)
By using 1988 discharges from 457 California hospitals, Kominski and
Rice (1994) found that, focusing on hospital costs, Medicare patients are 11.7
percent more expensive than commercially insured patients. In contrast, on
average, other payers are less expensive than commercially insured: Medicaid, -
12.3 percent; HMOs, -4.5 percent; and all other payers, -16.2 percent. Blue Cross
patients were only slightly more expensive than commercially insured patients.
17
Reproduced with permission of the copyright owner. Further reproduction prohibited without permission.
Managed Care Penetration and Hospital Competition
Managed care can be defmed as “an organized effort by health insurance
plans and providers to use financial incentives and organizational arrangements to
alter provider and patient behavior to that health care services are delivered and
utilized in a more efficient and lower-cost manner.” As an alternative way to
contain increasing health care costs by increasing efficiency, managed care
became popular (Drake, 1997).
Gaskin and Hadley (1997) reported that the health care markets have
evolved substantially over the past decade in that the health care markets of 1990s
are significantly different from those of the 1980s. During the 1980s, managed
care grew rapidly in the US, especially in California. Nationally, enrollments in
managed care organizations (MCOs) including HMOs and PPOs have increased
dramatically over the past 15 years. Estimated HMO enrollment rose from 28.6
million individuals in 1987 to nearly 60 million members by the end of 1995.
PPO enrollments rose from 12 million in 1987 to over 91 million members by the
end of 1995. By 1988, California had the largest number of HMO enrollment,
about 7.68 million HMO enrollees which were about 28.5 percent of the whole
state population. This was even twice higher than the national average enrollment
rate (12.1 percent) in 1987 and even well above the national average enrollment
rate (19.7 percent) in 1994. The number of enrollment in other types of managed
care such as preferred provider organization (PPO) and point-of-service (POS)
18
Reproduced with permission of the copyright owner. Further reproduction prohibited without permission.
has increased significantly as well (Chung and Meltzer, 2000). The growth of
managed care organizations contributed to lower cost growth, both by delivering
health care services at lower costs due to lower service intensity (Manning et al.,
1984) and by increasing competition in hospital markets (Melnick and Zwanziger,
1995).
Anderson et al. (1999) analyzed the impact of HMO penetration on
hospital utilization and expenditures using the national data from 1982 to 1996.
By dividing metropolitan statistical areas (MSAs) into four different categories in
relation to HMO penetration rates in 1996, they compared levels and rates of
change in hospital admission rates, hospital inpatient days, emergency room
visits, total expenditures per capita, and expenditures per adjusted inpatient day
from 1982 to 1996. They found that the effects of HMO penetration on hospital
expenditures and hospital utilization at the MSA-level are small (on average, less
than one percent per year).
Although competition has always existed in health care, the special
features of health care markets have made this competition different from most
other markets. In hospital markets the complicating factor has been the presence
of conventional health insurance. It has meant that consumers (i.e., patients) have
been shielded from the cost consequences of their choice. When the decision to
use care is based on the services, amenities, and quality of the provider, one
should expect that hospitals will compete for patients based on these factors. This
19
Reproduced with permission of the copyright owner. Further reproduction prohibited without permission.
has the effect of standing traditional economics on its head. Instead of more
competition leading to prices driven closer to marginal costs, more competition in
a hospital market leads to increased service, amenities, and quality rivalry - and
higher prices. (Morrisey, 2001)
Managed care introduced selective contracting into the hospital market.
Due to selective contracting, some hospitals get contracts and some do not.
Decisions as to who gets contracts now depend on services, amenities, quality,
and price as well. If so, standard economic expectations begin to apply. One
expects that more hospital competition will lead to lower prices. The nature of the
product and even the market itself may change. Because there is now a trade-off
between low prices and more services and amenities, one should expect a slower
proliferation of services and amenities. New services are less likely to be adopted
without evidence that they improve care enough to warrant their cost. (Morrisey,
2001)
Theories of the hospital market generally view the hospital as competing
for patients, providers, or both. Typically one expects hospital to behave like
other providers of complex services, distinguishing themselves on the basis of the
bundle of services, quality, and amenities they offer and the price they charge. It
has been argued, however, that the hospital market is different. The presence of
health insurance allowed individual patient and their physicians to be less
concerned about the price of care. Patients and their physicians had an incentive
20
Reproduced with permission of the copyright owner. Further reproduction prohibited without permission.
to use every service and amenity available as long as the perceived value was
greater than the out-of-pocket cost. Thus, in competing for physicians and their
patients, hospitals put much more emphasis on the services and amenities they
offered than on the prices charged. (Morrisey, 2001)
Using data from 1982, Robinson and Luft (1985) identified the number of
hospitals in a 15-mile radius of each hospital in the United States. Controlling for
hospital and market characteristics, they estimated the effect of the number of
potentially competing hospitals on the average cost per admission at each
hospital. Hospitals with no competitors had a higher average cost per admission
($2,268) than those with 11 or more potential competitors ($2,859). Rather than
more competitors forcing lower prices, more competitors would lead to increased
services and amenities and to higher prices.
Salkever and Seidman (1978) reviewed hospital competition literature and
generalized this finding by insisting that “the major conclusion which emerges
from the foregoing description of the hospital services market is that competition
among hospitals is based primarily upon the availability and sophistication of
services and facilities rather than price”
In 1983 California allowed insurers to selectively contract with providers,
and it implemented a selective contracting program for the state Medicaid
program. These contracts were based, in part, on the prices the hospitals were
willing to accept. Melnick and Zwanziger (1988) analyzed the rate of increase in
21
Reproduced with permission of the copyright owner. Further reproduction prohibited without permission.
hospital costs in the periods before and after the selective contracting legislation
took effect. In the pre-legislation period, hospital costs behaved as the service
competition model suggested. Hospitals with more competitors had more rapid
rates of cost increase than did hospitals with fewer competitors. This suggests that
the competition had been in the form of services, amenities, and quality
enhancements. In the post-legislation period, however, hospitals in the more
competitive markets had lower rates of increase. Indeed, their real costs actually
dropped slightly. In contrast, the hospitals in less competitive markets had more
rapid rates of cost increase. Additional work using the California data found that
by the end of the 1980s the levels of average costs were lower in the more
competitive areas (Zwanziger, Melnick, and Bamezai, 1994).
Using 1982 data from 5732 hospitals in the US, Robinson and Luft (1987)
found that costs were substantially higher in hospitals operating in more
competitive local environments than in hospitals in less competitive
environments. Average costs per admission were 26% higher in hospitals in the
most competitive markets (more than ten hospitals within a 24-km radius) than in
hospitals with no competitors within a 24-km radius. Average costs per patient-
day were 15% higher in hospitals in the most competitive markets than in
hospitals with no competitors.
Robinson (1991) studied the impact of HMO-induced price competition
on the rate of inflation in average cost per admission based on 298 private, non-
22
Reproduced with permission of the copyright owner. Further reproduction prohibited without permission.
HMO hospitals between 1982 and 1988. He concluded that HMO coverage
increased from an average of 8.3% of all admissions in local hospital markets in
1983 to 17.0% of all admissions in 1988. The average rate of growth in costs per
admission between 1982 and 1988 was 9.4% lower in markets with relatively
high HMO penetration compared with markets with relatively low HMO
penetration. Cost savings for these 298 hospitals are estimated at $1.04 billion for
1988. Using private nonprofit and for-profit hospitals with 25 or more beds in
California, Robinson (1996) also compared the impact of HMOs on hospital
utilization and expenditures between 1983 and 1993. He concluded that hospital
expenditures grew 44 percent less rapidly in markets with high HMO penetration
than in markets with low HMO penetration.
Price competition between HMOs and conventional health insurers can
significantly reduce hospital cost inflation if legislative barriers to selective
contracting are removed. The HMOs have pioneered a style of medical practice
that uses less intensive hospital care per enrollee than conventional insurance
plans. Market pressures in a deregulated legal environment can spread HMO’s
more conservative hospital utilization patterns to FFS plans. Thus, HMOs can
exert indirect effects on hospital behavior by stimulating more price-competitive
behavior on the part of other health insurance plans (Robinson, 1991).
Hadley, Zuckerman and Feder (1989) investigated the role of fiscal
pressure in determining hospitals’ response to PPS incentives using hospital-level
23
Reproduced with permission of the copyright owner. Further reproduction prohibited without permission.
data. Their key variable was the ‘fiscal pressure’ index, an estimate of each
hospital’s potential profit margin on Medicare inpatients if it did not alter costs or
volume from the previous years. Hadley et al. found that among hospitals entering
PPS in FY 1985, the LOS reduction was 9.4% for highly pressured hospitals,
compared to 6.8% at those hospitals facing low pressure.
Hadley, Zuckerman and Iezzoni (1996) analyzed the effects of financial
pressure and market competition on hospital performance using 1,435 acute care
hospitals between 1987 and 1989. Over the observation period, the least
profitable hospitals constrained their growth in total expenses to half that for the
most profitable hospitals (13.3% versus 27.6%) by limiting the growth of their
staffs and their total assets. These changes were associated with a reduction in
inefficiency of 1.8% (11.2%) compared with a very slight increase in inefficiency
for the highest profit group. Similarly, hospitals in highly competitive markets
controlled expenses relative to those in the least competitive areas. However, they
also experienced slower revenue growth than did less competitive hospitals so
that, in relative terms, their profit rates fell. The authors found no evidence to
suggest that financial pressures created by either low profits or market
competition resulted in hospitals engaging in cost-shifting. The authors conclude
that health care reforms or market forces that put financial pressures on hospitals
can result in cost-containment and improved efficiency without significant cost-
shifting.
24
Reproduced with permission of the copyright owner. Further reproduction prohibited without permission.
Hadley and Swartz (1989) analyzed hospital expenses in 43 large
SMS As between 1980 and 1984. The authors found that that hospital rate
regulation was the single most important factor leading to the slowdown in the
rate of increase in hospital costs between 1980 and 1984. In 1984, hospital costs
covered by Medicare's PPS were 12.5% lower than they would have been in the
absence of rate regulation, and in the four states covered by all-payer rate
regulation, hospital costs were between 11% and 15% lower. In contrast, changes
in the proportion of people either covered by employer-group health insurance or
enrolled in HMOs, reduced hospital costs by less than 1%. Measures of
competition suggest that hospital costs are higher when there is more competition.
The authors found that almost all of the effect of regulation on costs came from
gains in the efficiency of producing hospital care and/or from reductions in the
quality of care. It appears that controlling hospital payment rates gave hospitals a
strong incentive to provide care at a lower cost.
Gaskin and Hadley (1997) showed that the higher HMO penetration
reduces hospital cost inflation. During the period 1985-1993, hospitals in areas
with high HMO penetration had a slower rate of growth in expenses than
hospitals in areas with low HMO penetration (8.3% vs. 11.2%). Decline in
Medicare PPS margins also lowered hospital cost inflation.
By using 1990-1994 HCFA’s Medicare data, Baker (1999) analyzed the
impact of HMO market share on Medicare’s traditional fee-for-service (FFS) plan
25
Reproduced with permission of the copyright owner. Further reproduction prohibited without permission.
expenditures. The author found that increases in system-wide HMO market share
were associated with declines in both Medicare Part A and Part B FFS
expenditures per Medicare beneficiary. When HMO market share increases from
10% to 20%, Medicare Part A FFS expenditures decreased by 2.0% and Part B
FFS expenditures decreased by 1.5%. Thus, HMO penetration has reduced
Medicare’s traditional FFS expenditures through spillover effects.
Williams, Hadley and Pettengill (1992) argued that the number of closed
hospitals increased substantially after the implementation of DRG-based PPS and
that hospital profitability is associated with the Medicare case-mix index and the
share of Medicare patients. The findings also suggest that the case mix index may
be rewarding some small hospitals in excess of the costs attributable to case-mix.
For both urban and rural hospitals, a low share of Medicare patients reduced
hospital profitability. The low share of Medicare patients also indirectly affected
hospital closure, through its effects on profit.
Dor and Farley (1996) investigated the capacity of hospitals to vary the
intensity of their services based on patient’ expected sources of payment and
concluded that payer-specific marginal costs generally reflected payer generosity.
Reinhardt (1996) suggested that hospitals and insurers price bed-days well
above marginal cost, particularly for days after the patient’s first. This induces us
to substitute high-cost home health or rehabilitation hospital services for low cost
hospital beds, wasting resources. In an ideal competitive market, price
26
Reproduced with permission of the copyright owner. Further reproduction prohibited without permission.
competition would force the bed-day prices down to marginal cost, however this
has not happened yet. The Medicare DRG payment system makes it worse by
allowing $0 for extra hospital days. Meantime, Medicare pays by the day or
service for home health or stays in a rehabilitation facility.
The number of Americans who do not have health insurance continues to
increase, reaching 43 million in 1997. This is up from 40 million in 1994 and 34
million in 1988. As price competition forces providers to minimize cost,
Weissman (1996) argued that those hospitals who give less uncompensated care
have a cost advantage over those who give more uncompensated care.
Hospitals in California became more increasingly subject to tighter budget
constraints with the implementation of Medicare PPS and increased competition
due to Medicaid (Medi-Cal) selective contracting and the growth of managed care
(Chung and Meltzer, 2000). Hospital revenues have declined but patients’ acuity
has been increased. Due to various utilization management techniques that have
been employed by managed care organizations, the acuity of patients who are
staying in the hospitals gets worse (Chung and Meltzer, 2000). Carpenter et al.
(1999) argued that hospitals with the higher values of severity variables suffered
more from the financial losses since payers did not compensate adequately for
severity.
Bamezai et al. (1999) showed that HMOs and PPOs have significantly
restrained cost growth of hospitals located in competitive hospital markets using a
27
Reproduced with permission of the copyright owner. Further reproduction prohibited without permission.
national database between 1989 and 1994. Wholey, Feldman and Christianson
(1995) insisted that more competition reduced HMO premiums for group-model
HMOs and more market penetration reduces premiums for IP As.
28
Reproduced with permission of the copyright owner. Further reproduction prohibited without permission.
3. DATA AND METHODS
Conceptual Framework
Hospital case mix can be an important predictor of hospital costs. Using
hospital level data, the DRG case mix index for Medicare patients (in
combination with the wage index, teaching activity, and urban and rural status)
explained 72 to 75 percent of the variation in cost per case (Pettengill and
Vertrees, 1982; Cotterill et al., 1985; Horn et al., 1984). Pettengill and Vertrees
(1982) argued that whether cost-based DRG weights or charge-based DRG
weights are used, the models explain 72 percent of average hospital cost per case
for Medicare patients. Cotterill et al. (1985) confirmed the Pettengill and
Vertrees’ study and argued that charge-based DRG weights can be an effective
basis for DRG weight determination.
In addition to hospital case mix, competition among hospitals and
managed care penetration significantly affect hospital costs as well as hospital
revenues. Previous studies on hospital costs focused on either the relationship
between hospital costs and competition (Zwanziger and Melnick, 1988;
Zwanziger et al., 1993; Zwanziger et al., 2000; Robinson and Luft, 1987; Hadley
and Swartz, 1989; Hadley et al, 1996) or the relationship between hospital costs
and managed care penetration (Robinson 1991; Robinson, 1996; Gaskin and
29
Reproduced with permission of the copyright owner. Further reproduction prohibited without permission.
Hadley, 1997) but there were few research studies that focus on two important
issues at the same time.
If hospital competition and HMO penetration are positively correlated and
if hospital competition tends to lower hospital-operating costs, HMO penetration
would be correlated with decreasing costs in the absence of adjusting for
competition (Gaskin and Hadley, 1997). Thus, it is important to control for both
hospital competition and HMO penetration.
Due to competition and managed care, hospitals have argued that the rate
of increase in hospital costs is faster than the rate of increase in hospital revenues.
By employing various utilization management programs (e.g., preadmission
screening and triage), managed care has been able to treat low intensity patients at
outpatient settings. Thus, hospitalized patients are more likely to be those who
truly need inpatient services.
By using a random sample of 85,232 patients in 1986 from hospitals in
New Jersey, Averill et al. (1992) argued that severity of illness was found to be a
significant determinant of hospital cost in 76 DRGs that accounted for 41.4
percent of the total direct hospital patient care costs. Due to hospital competition
and managed care, however, hospitals in California have been paid from managed
care organizations (MCOs) based on the number of days that patients stayed
without considering the severity of illness. If hospitals admit a severely ill patient,
they would expect to lose money due to a lower rate of payment.
30
Reproduced with permission of the copyright owner. Further reproduction prohibited without permission.
Keeler et al. (1990) developed disease-specific measures of sickness at the
stage of admission by using Medicare patient medical records in one of five
medical conditions (i.e., congestive heart failure, acute myocardial infarction,
cerebrovascular accident, pneumonia, and hip fracture). They revealed that the
expected 30-day mortality was 1.0% higher in the 1985-1986 period than in the
1981-1982 period (16.4% vs 15.4%). In addition, the expected 180-day mortality
was 1.6% higher in the 1985-1986 period than in the 1981-1982 period (30.1% vs
28.5%). After implementing the DRG-based PPS, thus, Keeler et al. argued that
sickness at admission increased significantly for the above mentioned five
conditions.
The PPS used by Medicare and some other payers in the US has been
criticized for not adjusting for differences in severity of illness within DRGs.
Using a patient-level data, Carpenter et al. (1999) argued that two measures of
severity (disease stage and number of unrelated diseases) were significant
predictors of cost per case, and often had better predictive power than DRGs. In
most instances, they argued that payers did not compensate adequately for
severity so that higher values for the severity variables resulted in financial losses
for the hospital.
It is important to pay hospitals based on the patient’s severity of illness
that the hospital treats. Yet, it might be difficult to adjust payment in California
based on the severity of illness since there is no case mix index using all
31
Reproduced with permission of the copyright owner. Further reproduction prohibited without permission.
Californian patients. One of the possible alternatives is to use the CMI based on
Medicare patients as a proxy but it is well known that the elderly use significantly
more health services than the non-elderly. Another possibility is to use the CMI
based on all patients in States of New York and New Jersey since they once used
an all-payer DRG payment system. However, when calculating hospital case mix
index in California, if one applies the Medicare DRG weights or the NY/NJ all
payer DRG weights to California, it may mislead the true case mix of the
Californian population. Instead of using Medicare DRG weights or NY/NJ all
payer DRG weights, hospital discharge data from all acute-care hospitals in
California will be used to calculate DRG weights for selected years in 1986-1998.
By using DRG weights based on Californian hospital discharge data, hospital case
mix index for different payers such as Medicare, Medi-Cal and others4 can be
calculated.
Previous studies (Zwanziger and Melnick, 1988; Zwanziger et al., 2000)
on hospital costs did incorporate hospital CMI as a control variable to explain the
variation in hospital costs. Yet, those studies used either a hospital CMI that was
based on a small sample patient discharges or HCFA’s Medicare CMI that is
based on only Medicare patients. Hospitals have a different mix of patients. Thus,
4 According to Carter, for the DRG development, whenever the population base changed, it is
necessary to recalibrate using the new database.
32
Reproduced with permission of the copyright owner. Further reproduction prohibited without permission.
hospital CMI based on either a small subset of discharges or only Medicare
patients may mislead the true case mix of patients that each hospital treats.
The purpose of this paper is to provide a better understanding of those
factors affecting hospital costs and revenues in California using the actual hospital
costs and utilization data for several years. In this study, by developing CMI
using all hospital discharges in California by different payers (including all
payers, Medicare, Medi-Cal and all others), I will investigate 1) the relationship
between hospital revenues and costs, 2) the differential effects of CMI on hospital
costs and revenues and 3) the performance of different CMIs to predict hospital
costs and revenues.
Hypotheses
There are three hypotheses in this study.
Hypothesis #1: Competition is dominating the pricing practice in the
hospital industry. Hospital expenses are driven by CMI but hospital revenues are
not driven by CMI.
Hypothesis #2: Over the time, the differential effects of CMI on hospital
expenses and revenues will grow.
33
Reproduced with permission of the copyright owner. Further reproduction prohibited without permission.
Hypothesis #3: Weights based on all patients in California perform better
to predict hospital expenses and revenues than those based on Medicare
population in the nation.
Empirical Model Specification and Estimation
Earlier empirical studies of hospital cost functions have tended to adopt
one of two approaches in specifying the model to be estimated (Breyer, 1987).
One is based on ad hoc, with variables based on knowledge of the hospital
industry; the other is based on the flexible functional forms used in the analysis of
neoclassical production theory (McFadden, 1978). Several researchers
(Zwanziger and Melnick, 1988; Zwanziger, Melnick, and Bamezai, 2000)
combined both of these approaches. The combined approach includes the
logarithm of the multiple output and input price levels. After performing a
Hausman specification test for the “random effects” specification versus the
“fixed effects” specification, the Hausman specification test rejected the random
effects specification. In this study, thus, a hospital fixed-effects estimator of
hospital costs and revenues with a model specification based on the translog
structure will be used because of the flexibility and theoretical reasoning of the
translog structure.5
5 The primary problem in modeling hospital expenses and revenues is that the distribution of
hospital expenses and revenues are highly skewed. Upon examination, the distribution of the
natural logarithm of hospital expenses and revenues are approximately normally distributed.
34
Reproduced with permission of the copyright owner. Further reproduction prohibited without permission.
The regression model used in this study has the following form:
C lt —f(F , P, Z, M , H, B , T) + hj + e u6
Where:
C = hospital costs (or revenues, both total and inpatient);
F = a vector of hospital flow variables (e.g., inpatient discharges,
outpatient
visits, case mix);
P = the input price index (e.g., Medicare area wage index);
Z = a vector of hospital control variables (e.g., ownership, payer mix);
M = measures for competition in inpatient hospital markets;
H = measures for HMO penetration in county;
B = measures that capture cost-cutting pressure from the Medicare PPS
and Medicaid selective contracting programs;
T = a vector of time dummies;
hi = a hospital-specific constant; and
ejt = the error that is i.i.d. (0, s2 ).
6 This model is basically based on Zwanziger, Melnick, and Bamezai (2000). I added payer-
specific case mix index and HMO penetration rates to analyze the relationship between case mix
index and hospital costs (and revenues).
35
Reproduced with permission of the copyright owner. Further reproduction prohibited without permission.
Table 1: Variables Used in This Study
Variable Specific measure
Dependent variables 1) Operating expense (total and inpatient)
2) Net revenue (total and inpatient)
3) Medicare operating expense (total and inpatient)
4) Medicare net revenue (total and inpatient)
5) Medi-Cal operating expense (total and inpatient)
6) Medi-Cal net revenue (total and inpatient)
7) Others operating expense (total and inpatient)
8) Others net revenue (total and inpatient)
Flow (output)
variables
inpatient discharge
outpatient visits
Case mix
- All-Payer CMI
- Medicare CMI
- Medicaid CMI
- Others CMI
Input price variables Medicare Wage Index
Control variables
1) Ownership
2) Payer mix
1) Dummy variables for ownership status
2) Proportion of discharges by major payer categories
Market
competitiveness
variable
Hirschman-Herfmdahl Index (HHI)
Managed Care
Penetration variable
HMO penetration rate by county in California
(available from InterStudy)
Cost cutting pressure
variable
Medicare pressure index
Time Year dummy variables
36
Reproduced with permission of the copyright owner. Further reproduction prohibited without permission.
Data Sources
There are two major data sources. One is the annual and quarterly
hospital-level data for years 1986, 1989, 1992,1995 and 1998 from California’s
Office of Statewide Health Planning and Development (OSHPD). The other is the
discharge data (patient census data) for the same years that are also available from
OSHPD. Medicare DRG weights were available from Federal Registrar. The
California state government has required hospitals to report detailed cost, revenue,
and utilization information such as operating expenses, net revenues, number of
admissions, number of discharges, etc.
Construction of the Variables
Dependent variables: Total expenses, total revenues, inpatient expenses
and inpatient revenues for all patients, Medicare, Medi-Cal and Others (including
third-party payers)7 were used as dependent variables. These dependent variables
7 For All Patients in 1986 and 1989:
• Inpatient net revenue=[(gri_tot)/(gri_tot+gro_tot)]*net_tot, where gri_tot=gross inpatient
revenues total, gro_tot=gross outpatient revenues total, and net_tot=net revenues total.
For Medicare in 1986 and 1989:
• Medicare operating expense=(net_mcar/net_tot)*tot_exp, where net_mcar=Medicare net
revenues, tot_exp=total expense. For Medicare operating expense, although it is better to use
[(grCmcar+gro_mcar)/(gri_tot+gro_tot)]*tot_exp, there is no information on gri_mcar,
gromcar, where gri_mcar=Medicare gross inpatient revenues, and gro_mcar=Medicare gross
outpatient revenues.
• Medicare inpatient operating expense=[(gri_tot)/(gri_tot+gro_tot)]*(Medicare operating
expense). It is better to use [(gri_mcar)/(gri mcar+gro mcar)]* (Medicare operating expense)
but there is no information on grimcar, gromcar.
37
Reproduced with permission of the copyright owner. Further reproduction prohibited without permission.
were not deflated because in order not to restrict hospital responses to increasing
input prices to be exactly proportional to inflation. Instead, an input price index
was included as a covariate to control for inflation. Since these variables were
highly skewed, their logged form was used in the model. The primary problem in
modeling hospital expenses and revenues is that the distribution of hospital
expenses and revenues are highly skewed. Upon examination, the distribution of
• Medicare inpatient net revenue=[(gri tot)/(gri_tot+gro_tot)]*net_mcar. It is better to use
[(gri_mcar)/(gri_mcar+gro_mcar)]*(net_mcar) but there is no information on gri mcar,
grom car.
For Medi-Cal in 1986 and 1989:
• Medi-Cal operating expense=(net_mcal/net_tot)*tot_exp, where netjmcal=Medi-Cal net
revenue. For Medi-Cal operating expense, although it is better to use
(gri_mcal+gro_mcal)/(gri_tot+gro_tot)]*tot_exp, there is no information on grim cal, grojncal,
where gri_mcal=Medi-Cal gross inpatient revenues, and gro_mcal=Medi-Cal gross outpatient
revenues.
• For Medi-Cal inpatient operating expense=[(gri_tot)/(gri_tot+gro_tot)]*(Medi-Cal operating
expense). It is better to use [(gri_mcal)/(gri_mcal+gro_mcal)] * (Medi-Cal operating expense)
but there is no information on gri mcal, grom cal.
• For Medi-Cal inpatient net revenue=[(gri_tot)/(gri_tot+gro_tot)]*net_mcal. It is better to use
[(gri_mcal)/(gri_mcal+gro_mcal)]*(net_mcal) but there is no information on gri mcal,
grom cal.
For Third-party payers and Others in 1986 and 1989:
• Others operating expense=(net91oth/net_tot)*tot_exp, where net91oth=net revenues for third-
party payers and others. For inpatient operating expense for Others, it is better to use
[(gri_cnty+gri_thrd+gri_oth+gro_cnty+groJhrd+gro_oth)/(gri_tot+gro_tot)]*(tot_exp), but
there is no information on these.
• Others inpatient operating expense=[(gri_tot)/(gri_tot+gro_tot)]*(others operating expense). For
inpatient operating expense for Others, it is better to use
[(gri_cnty+gri_thrd+gri_oth)/(gri_cnty+griJ:hrd+gri_oth+gro_cnty+gro_thrd+gro_oth)]*(others
operating expense), but there is no information on these.
• Others inpatient net revenue =[(gri_tot)/(gri_tot+gro_tot)]*(others total revenues). For inpatient
net revenues for Others, it is better to use
[(gri_cnty+griJhrd+gri_oth)/(gri_cnty+gri_thrd+gri_oth+gro_cnty+gro_thrd+gro_oth)]*(others
total revenues), but there is no information on these.
For years 1992, 1995 and 1998,1 used the above mentioned better ways o f calculation due to the
availability of data in OSHPD quarterly financial and utilization data.
38
Reproduced with permission of the copyright owner. Further reproduction prohibited without permission.
the natural logarithm of hospital expenses and revenues are approximately
normally distributed.
Input prices variables: The Medicare area wage index (from HCFA) was used to
control for the relative cost of labor in each hospital’s geographic area.
Case mix index: To control for differences in hospital’ patient mix, payer-
specific case mix indices were included. Previous studies (Zwanziger et al., 2000)
applied CMI that is derived from using either HCFA’s DRG weights or NY/NJ
all-payer DRG weights as a proxy to CMI of Californian hospitals. Instead of
using HCFA’s case mix index or
NY/NJ all-payer CMI, the payer-specific CMIs (all-payer, Medicare, Medi-Cal
and all other payers) were calculated by using the all discharges from all acute
care hospitals in Calfomia between 1986 and 1998.
HCFA defines a CMI as a measure of the average relative costliness of
Medicare cases treated by a hospital compared to the national average cost of all
Medicare hospital cases. The CMI is based on the distribution of cases across
DRGs and the relative costliness of cases in each DRG. By adopting this
definition, a payer-specific CMI using all patients in California was developed
Q
(excluding outliers ).
8 Outliers are removed using three standard deviations o f hospital charges.
39
Reproduced with permission of the copyright owner. Further reproduction prohibited without permission.
The step-by-step approach to calculate cost-based DRG weights using
California discharge data are:
1) Calculate the hospital-level cost-to-charge ratio (CCR)9 by dividing total
operating expenses (from quarterly data) by total charges (from hospital
disclosure data).
2) Using the discharge data, calculate cost for each DRG using CCR
calculated in the previous step, i.e., charge*RCC=cost. The costs for each
discharge in the patient discharge data can be estimated using CCR.
3) Calculate average cost for each DRG and grand average cost for all
discharges.
4) Calculate each DRG weight by dividing “average cost for each DRG” by
“grand average cost”.
To estimate hospital case mix index (CMI) by all-payer, Medicare, Medi-
Cal and Others,
9 Total costs can be estimated using hospital-level cost-to-charge ratios constructed to correspond
to each CY because hospital FYs do not necessarily overlap with the CY. Thus, it is necessary to
calculate a blended cost-to-charge ratio (CCR) for each hospital from the 2 fiscal years that
overlap each CY. This method is subject to several limitations. It is less accurate than using
departmental CCRs. Yet, it is difficult to employ department-level CCRs because California
discharge files include only total patient charges, and not departmental charges (Kominski and
Rice, 1994, HCFR 16(2): 180). Nevertheless, hosptial-level CCRs can adjust for certain
discrepancies between costs and charges (Finkler, 1982), and have been found to perform
somewhat better than charges as proxies for costs (Newhouse, Cretin and Witsberger, 1989;
Shwartz, Young and Siegrist, 1995).
40
Reproduced with permission of the copyright owner. Further reproduction prohibited without permission.
1) For each hospital, multiply DRG weight for each DRG by the number of
discharges in each DRG by hospital.
2) Calculate CMI by dividing “the sum of (DRG weight*discharge) for all
DRG” by “total number of discharges”.
3) For Medicare (Medi-Cal or Others) CMI, use only for Medicare (Medi-
Cal or Others) discharges.
Medicare pressure index: The Medicare pressure index was intended to capture
the pressure hospitals have been facing to cut costs due to the Medicare
Prospective payment System (PPS). The pressure index for each hospital was
constructed as the product of two ratios:
Pressure Index = (Cg4/Rg4- 1) * (D m /D t), where
C jj4 is average Medicare cost per discharge in 1984, standardized for case
mix and teaching intensity;
Rg4 is the Medicare reimbursement per discharge in 1984 based on the
national Medicare rate;
Dm is the number of Medicare discharges in 1984; and
Dt is the total of all discharges in 1984.
In order to measure the degree to which the change in Medicare
reimbursement to the PPS could affect the hospital’s profitability, the Medicare
pressure index was used. A high-pressure index means that hospital costs were
high relative to the PPS reimbursement and that Medicare business was a
substantial fraction of total business at the hospital. The pressure index captures
41
Reproduced with permission of the copyright owner. Further reproduction prohibited without permission.
the full impact that hospitals would have faced had the PPS national rate kicked in
immediately (i.e., the program’s full financial impact in the absence of change).
The pressure index for a hospital was kept constant in magnitude over time, but
was interacted with time in the cost models to allow for behavioral adjustments
leading to changes in the impact of Medicare pressure on hospital costs over time.
Hospital market competition: The Hirschman-Herfindahl Index (HHI)1 0
will be used as a basic measure of competition. In many early studies of hospital
competition, a hospital market was defined arbitrarily using either geopolitical
boundaries (e.g., the county) or fixed distances (e.g., all hospitals within an X
mile radius of a given point). Such definitions biased the estimated effect of
competition on hospital behavior downward by introducing error in the
measurement of competition. Hence, in this study, I used the work from
10 Most studies o f hospital competition rely on one o f two standard competition measures that
capture the classic determinants o f competition: the number and relative sizes o f firms. In
economic theory, producer markets for a homogeneous product sold to many informed buyers will
be competitive if there are many small sellers. The presence o f more firms is thus typically
associated with more competition. The classic economic definition of competition also requires
that all of the forms be small in the sense that none o f them is large enough to dictate the price in
the market. More generally, markets with more evenly balanced firms are apt to be more
competitive than markets in which there are some firms that are larger and more powerful than
their neighbors. (Baker 2001 HSR p.232)
The simplest type of measure counts the number o f competing firms. This is an
appealing, frequently easily implementable, and intuitive approach. Yet, this approach does not
capture the relative sizes o f firms, which can play an important role in competition. Both the
number and relative size o f firms are better captured in the Herfindahl-Hirschman index (HHI),
perhaps the most common measure o f competition. The HHI for a market is the sum o f the
squared market shares o f all o f the firms competing in the market. HHIs are the standard measure
used by the Department o f Justice and Federal Trade Commission in evaluating the degree of
concentration in markets for antitrust policy and are frequently used in empirical work in health
services research.
42
Reproduced with permission of the copyright owner. Further reproduction prohibited without permission.
Zwanziger and Melnick who collected patient-origins data by zip code to
determine the extent of each hospital’s market.
Three steps were required to calculate each hospital’s HHI. First, service-
specific market areas for each short-term general hospital in California should be
defined. Categories were developed for hospital services by combining all of the
Diagnosis-Related Groups (DRGs) that would be provided by the same type of
physician (Zwanziger, Melnick, Mann and Eyre, 1994). All of the discharges
from a given hospital that fell within a single service category were combined and
used to calculate service-specific market areas. Thus, this approach can correctly
account for the fact that simple services (e.g., checkups) are usually provided for
only the local population, while complex “tertiary” care (e.g., neurosurgery)
draws from a much wider area. Second, each hospital’s competitors were
identified and their share of the service-specific market was calculated. Third,
each hospital’s service-specific HHIs were derived from the market share values.
Then, the HHIs were averaged across services and zip code areas using discharge
weights, resulting in an overall measure of the competitiveness for each hospital.
The primary data sources for this task were the California Discharge Data Set for
1986, 1989, 1992, 1995 and 1998.
HMO penetration rate: HMO penetration data by county in California were
used for years 1989-1995. For 1986, 1989 HMO penetration data were used due
43
Reproduced with permission of the copyright owner. Further reproduction prohibited without permission.
to unavailability of 1986 HMO penetration data. For 1998, 1995 HMO
penetration data were used for the same reason.
Medicare wage index: The Medicare area wage index was used to control for the
relative cost of labor in each hospital’s geographic area. The Medicare wage
index is available from HCFA.
Ownership: A dummy variable of “For-profit” was created. If hospitals are
investor owned, it is coded as “For-profit=l”, otherwise “For-profit=0”.
Reproduced with permission of the copyright owner. Further reproduction prohibited without permission.
5. RESULTS
Based on the data, hospital expense and revenue patterns for selected years
from 1986 to 1998 were examined. Table 2 shows characteristics of hospitals
included in this study. In general, the average total operating expenses increased
from $38.3 million in 1986 to $82.0 million in 1998. The total net revenues also
increased gradually over time, from $38.5 million to $80.3 million in 1998. Just
looking at these two trends, hospitals made profits in earlier years but that trend
has been changed. Both the number of inpatient discharges and the number of
outpatient visits have increased significantly, especially for the number of
outpatient visits increased from 56,023 in 1986 to 111,438 in 1998. The
proportion of non-profit hospitals increased over the period time. Also, the
proportion of hospitals with more than 100 beds increased.
Table 2: Hospital Characteristics
Variable 1986 1989 1992 1995 1998
Number of hospitals 353 342 330 359 333
Operating expense (mil $) 38.3 51.9 67.5 69.4 82.0
Net revenue (mil $) 38.5 50.7 68.8 68.0 80.3
Discharge 6,800 7,271 7,414 7,168 7,892
Visits 56,023 69,748 83,078 91,863 111,483
% Profit hospitals 37.15% 30.09% 32.46% 36.30% 35.25%
% Non-profit hospitals 44.52% 48.16% 48.05% 47.89% 50.27%
% Bed<100 38.42% 35.14% 30.85% 31.15% 29.39%
% Bed 100-249 35.96% 36.69% 39.12% 37.98% 38.62%
% Bed 250-399 14.78% 15.76% 16.80% 17.49% 17.58%
% Bed 400+ 10.84% 12.40% 13.22% 13.39% 14.41%
45
Reproduced with permission of the copyright owner. Further reproduction prohibited without permission.
Table 3. Growth Rates in Total Operating Expenses for Hospital in Low, Medium
and High Competition Categories1 1
Period Low
Competition
Medium
Competition
High
Competition
1986-1989 .389 .284 .174
1989-1992 .332 .323 .174
1992-1995 .184 .087 .036
1995-1998 .161 .133 .079
Table 4. Growth Rates in Total Net Revenues for Hospital in Low, Medium and
High Competition Categories
Period Low
Competition
Medium
Competition
High
Competition
1986-1989 .325 .256 .130
1989-1992 .329 .345 .202
1992-1995 .171 .076 .032
1995-1998 .117 .123 .055
Tables 3 and 4 show the growth rates in total operating expenses and in
total net revenues in markets with different level of hospital competition. These
two tables show that the growth rates in total operating expenses and in total net
revenues are higher for hospitals in markets with low competition compared with
those in high competition markets. Hospital competition has consistently
contributed to the slower rate of growth in hospital operating costs. Also, the
1 1 These categories are defined using the HHI values. If an HHI value o f hospital falls in a low
quartile, it categorized into “High Competition” since the lower the HHI, the hospital market is
more competitive. If an HHI value o f hospital falls in a high quartile, it categorized into “Low
Competition” since the higher the HHI, the hospital market is less competitive.
Reproduced with permission of the copyright owner. Further reproduction prohibited without permission.
differences of the growth rate of increase (for both total operating expenses and
total net revenues) in low and high competition have decreased over time.
Table 5. Growth Rates in Total Operating Expenses for Hospital in Low, Medium
and High HMO Penetration Categories1 2
Period Low
Penetration
Medium
Penetration
High
Penetration
1986-1989 .338 .255 .256
1989-1992 .294 .221 .329
1992-1995 .159 .103 .058
1995-1998 .120 .138 .115
Table 6. Growth Rates in Total Net Revenues for Hospital in Low, Medium and
High HMO Penetration Categories
Period Low
Penetration
Medium
Penetration
High
Penetration
1986-1989 .302 .219 .239
1989-1992 .308 .239 .297
1992-1995 .131 .141 .025
1995-1998 .090 .089 .089
Tables 5 and 6 show the growth rates in total operating expenses and total
net revenues in markets with different level of HMO penetration categories. The
growth rates in total operating expenses (Table 5) are higher (except for 1989-
1992 period) for hospitals in markets with low HMO penetration compared with
those in high HMO penetrated markets. The growth rates in total net revenues
1 2 These categories are defined using the HMO penetration rates in the county. If a hospital located
in a county whose HMO penetration rate is in a low quartile, it categorized into “Low
Penetration”, vice versa.
47
Reproduced with permission of the copyright owner. Further reproduction prohibited without permission.
(Table 6) are higher for hospitals in markets with low HMO penetration compared
with those in high HMO penetrated markets.
Table 7. Average Hospital CMI in Low, Medium and High Hospital Competition
Categories
Low Competition Med Competition High Competition
AP
CMI
using
HCFA
wt
AP
CMI
using
Cost
wt
Medi
care
CMI
AP
CMI
using
HCFA
wt
AP
CMI
using
Cost
wt
Medi
care
CMI
AP
CMI
using
HCFA
wt
AP
CMI
using
Cost
wt
Medi
care
CMI
1986 .886 .946 1.177 .961 1.047 1.245 .921 .997 1.221
1989 .941 .945 1.241 1.023 1.056 1.315 .975 .979 1.322
1992 .969 .950 1.308 1.048 1.041 1.392 .948 .925 1.367
1995 .982 .919 1.316 1.094 1.050 1.420 1.020 .971 1.370
1998 1.027 .892 1.313 1.151 1.024 1.446 .971 1.093 1.421
Table 7 shows three hospital CMIs in relation to the different level of
hospital competition. Using the all-patient CMI with HCFA DRG weights, the
average CMI has increased over time for hospitals in markets with low hospital
competition. Yet, when applying all-patient CMI with cost-based DRG weights,
the average CMI has increased during the beginning periods and decreased during
the later years for the same hospitals in markets with low competition. The
average CMI has fluctuated over time for hospitals in markets with high hospital
competition. When using only the Medicare hospital CMI, the average Medicare
CMI has increased over time for all hospital competition categories. This table
shows that depending on which DRG weights to be used, there may be an
increasing or decreasing trend in case mix index.
48
Reproduced with permission of the copyright owner. Further reproduction prohibited without permission.
Table 8. Average Hospital CMI in Low, Medium and High HMO Penetration
Categories
Low Penetration Med Penetration High Penetration
AP
CMI
using
HCFA
wt
AP
CMI
using
Cost
wt
Medi
care
CMI
AP
CMI
using
HCFA
wt
AP
CMI
using
Cost
wt
Medi
care
CMI
AP
CMI
using
HCFA
wt
AP
CMI
using
Cost
wt
Medi
care
CMI
1986 .889 .952 1.163 .936 1.015 1.239 .963 1.050 1.240
1989 .952 .961 1.232 .996 1.014 1.328 1.016 1.046 1.300
1992 .966 .948 1.324 .995 .982 1.363 1.077 1.068 1.422
1995 .983 .924 1.320 .994 .939 1.356 1.110 1.067 1.429
1998 1.038 .907 1.347 1.046 .915 1.361 1.175 1.050 1.462
Table 8 shows the average hospital CMI in markets with the different level
of HMO penetration. As it can be noticed, using the all-patient CMI with HCFA
DRG weights, the average CMI has increased over time for hospitals in markets
with low HMO penetration. Yet, when applying the all-patient CMI with cost-
based DRG weights, the average CMI has decreased for hospitals in markets with
low HMO penetration. The average Medicare CMI has increased over time for
hospitals in all HMO penetration categories. Yet, when using the CMI with cost-
based DRG weights, the average CMI for hospitals in markets with high HMO
penetration has fluctuated and not changed significantly over time. When using
only the Medicare hospital CMI, the average Medicare CMI has increased over
time for all HMO penetration categories. This table also shows that case mix
index can be either increased or decreased depending on which DRG weights to
be used.
49
Reproduced with permission of the copyright owner. Further reproduction prohibited without permission.
Table 9: Means and Standard Errors of Covariates in Multivariate Regression
Models
Variables Mean Std. Dev.
Total operating expenses $61.6 mil $77.1 mil
Net revenues $61.0 mil $74.2 mil
Total inpatient operating expenses $47.0 mil $59.5 mil
Total inpatient net revenues $46.7 mil $57.4 mil
Total inpatient discharge 7,300.70 6,851.76
Total outpatient visits 82,206.52 106,779.50
All patient CMI using HCFA weight 1.015 0.246
All patient CMI using cost weight 0.995 0.272
Medicare CMI using HCFA weight 1.333 0.222
Medicare CMI using cost weight 1.380 0.237
Medi-Cal CMI using HCFA weight 0.869 0.647
Medi-Cal CMI using cost weight 0.823 0.739
Third-party and Others CMI using HCFA wt 0.932 0.320
Third-party and Others CMI using cost wt 0.880 0.343
Medicare area wage index 1.191 0.141
Medicare pressure index 0.0088 0.711
For-profit ownership 0.349 0.477
Hirshman-Herfindahl index 0.299 0.151
% HMO penetration 0.309 0.123
Results for All Patients
Using these variables, by focusing on case mix index as my primary
policy variable, I will examine expense and revenue patterns for selected year
from 1986 to 1998. For all patients, there are eight regression analyses. The first
four analyses (Tables 10-13) are using all-patient CMI with HCFA DRG weights,
the second four analyses (Tables 14-17) are using all patient CMI with cost-based
DRG weights, the third four analyses are using HCFA’s Medicare CMI.
50
Reproduced with permission of the copyright owner. Further reproduction prohibited without permission.
For the regression model for total hospital operating expenses (Table 10),
the coefficients of CMI, interacted with year dummy variable, were positive and
statistically significant for all years. This indicates that hospital operating
expenses increased as CMI increased. The coefficients of HHI were negative
(although they are statistically insignificant) for years 1986 and 1989. Yet, since
1992, the coefficients of HHI were positive and statistically significant. This
indicates that hospitals in a more competitive market had lowered their operating
expenses.
For total hospital net revenues (Table 11), the coefficients of CMI,
interacted with year dummy variable, were positive and statistically significant for
all years. This indicates that hospital net revenues increased as CMI increased.
The coefficients of HHI were positive and statistically significant for 1995 and
1998. This indicates that hospitals in a more competitive market had lowered their
net revenues in the later study years. The coefficients of for-profit, interacted with
year dummy variable, were positive and statistically significant for all years. This
indicates that for-profit hospitals tend to increase their revenues compared to non
profit or government hospitals.
For hospital inpatient operating expenses (Table 12), the coefficients of
CMI and HHI, interacted with year dummy variable, were all positive and
statistically significant. In addition, the magnitude of the coefficients of CMI is
greater when compared those in Table 10 (total hospital services) because CMI is
51
Reproduced with permission of the copyright owner. Further reproduction prohibited without permission.
mainly linked to inpatients. The coefficients of HMO penetration were negative
and statistically significant for yearsl989 (marginal using 95%), 1992 and 1995.
This indicates that HMO penetration negatively affected on hospital inpatient
operating expenses for those years.
For hospital inpatient net revenues (Table 13), the coefficients of CMI,
interacted with year dummy variable, were positive and statistically significant for
all years. In addition, the magnitude of the coefficients of CMI is greater when
compared those in Table 11 (total hospital services) because CMI is mainly linked
to inpatients. This indicates that hospital inpatient net revenues increased as CMI
increased. The coefficients of HHI were positive and statistically significant for
1995 and 1998. This indicates that hospitals in a more competitive market had
lowered their inpatient net revenues. The coefficients of for-profit, interacted with
year dummy variable, were positive and statistically significant. This indicates
that for-profit hospitals tend to increase their revenues compared to non-profit or
government hospitals. The coefficients for HMO penetration were negative and
statistically significant (except for 1998). This indicates that HMO penetration
negatively affected hospital inpatient net revenues.
It is possible to observe similar patterns of results in Tables 14-17 (using
cost-CMI for all patients) and Tables 17-20 (using Medicare CMI). However, the
magnitude of coefficients of CMI is greater when applying HCFA DRG weights
52
Reproduced with permission of the copyright owner. Further reproduction prohibited without permission.
to all patients compared to using HCFA’s Medicare DRG weights or cost-based
all-payer DRG weights.
53
Reproduced with permission of the copyright owner. Further reproduction prohibited without permission.
Table 10: The Estimated Regression Model for Hospital Operating Expenses
(using HCFA-weight Hospital Case Mix Index for all patients)
ln(Tota! Operating Expenses) Coefficient Std. Error p-value
ln(discharges) 0.4769 0.0175 0.0000
ln(visits) 0.1590 0.0120 0.0000
ln(Medicare Wage Index) 0.4656 0.1156 0.0000
1989 dummy 0.2959 0.0344 0.0000
1992 dummy 0.5729 0.0371 0.0000
1995 dummy 0.7478 0.0465 0.0000
1998 dummy 0.7141 0.0498 0.0000
ln(CMI using HCFA weights) * 1986 0.5800 0.0670 0.0000
ln(CMI using HCFA weights) * 1989 0.5700 0.0644 0.0000
ln(CMI using HCFA weights) * 1992 0.4830 0.0536 0.0000
ln(CMI using HCFA weights) * 1995 0.4358 0.0522 0.0000
ln(CMI using HCFA weights) * 1998 0.5012 0.0531 0.0000
For-profit * 1986 -0.0125 0.0283 0.6580
For-profit * 1989 -0.0187 0.0287 0.5150
For-profit * 1992 0.0418 ' 0.0285 0.1430
For-profit * 1995 -0.0106 0.0272 0.6960
For-profit * 1998 -0.0324 0.0263 0.2180
Medicare pressure index * 1989 -0.4401 0.1564 0.0050
Medicare pressure index * 1992 -0.5537 0.1586 0.0000
Medicare pressure index * 1995 -0.4863 0.1611 0.0030
Medicare pressure index * 1998 -0.5936 0.1586 0.0000
% MediCal days * 1986 0.3249 0.0901 0.0000
% MediCal days * 1989 0.1423 0.0708 0.0450
% MediCal days * 1992 0.1777 0.0564 0.0020
% MediCal days * 1995 0.2471 0.0482 0.0000
% MediCal days * 1998 0.3821 0.0524 0.0000
ln(Market HHI) * 1986 -0.0175 0.0367 0.6340
ln(Market HHI) * 1989 -0.0029 0.0360 0.9360
ln(Market HHI) * 1992 0.0647 0.0321 0.0440
ln(Market HHI) * 1995 0.1385 0.0374 0.0000
ln(Market HHI) * 1998 0.1590 0.0383 0.0000
ln(HMO penetration) * 1986 0.1621 0.0984 0.1000
ln(HMO penetration) * 1989 0.0046 0.0995 0.9630
ln(HMO penetration) * 1992 0.0117 0.1061 0.9120
ln(HMO penetration) * 1995 -0.1175 0.1155 0.3090
ln(HMO penetration) * 1998 0.1192 0.1217 0.3270
Constant 11.1510 0.1731 0.0000
R-square: 0.9268
Reproduced with permission of the copyright owner. Further reproduction prohibited without permission.
Table 11: The Estimated Regression Model for Hospital Net Revenues (using
HCFA-weight Hospital Case Mix Index for all patients)
ln(Total Net Revenues)_________ Coefficient_____Std. Error______ p-value
ln(discharges)
ln(visits)
ln(Medicare Wage Index)
1989 dummy
1992 dummy
1995 dummy
1998 dummy
ln(CMI using HCFA weights) * 1986
ln(CMI using HCFA weights) * 1989
ln(CMI using HCFA weights) * 1992
ln(CMI using HCFA weights) * 1995
ln(CMI using HCFA weights) * 1998
For-profit * 1986
For-profit * 1989
For-profit * 1992
For-profit * 1995
For-profit * 1998
Medicare pressure index * 1989
Medicare pressure index * 1992
Medicare pressure index * 1995
Medicare pressure index * 1998
% MediCal days * 1986
% MediCal days * 1989
% MediCal days * 1992
% MediCal days * 1995
% MediCal days * 1998
ln(MarketHHI) * 1986
ln(Market HHI) * 1989
ln(Market HHI) * 1992
ln(Market HHI) * 1995
ln(Market HHI) * 1998
ln(HMO penetration) * 1986
ln(HMO penetration) * 1989
ln(HMO penetration) * 1992
ln(HMO penetration) * 1995
ln(HMO penetration) * 1998
Constant
0.5686 0.0189 0.0000
0.1638 0.0129 0.0000
0.4456 0.1247 0.0000
0.2685 0.0372 0.0000
0.4956 0.0400 0.0000
0.7181 0.0502 0.0000
0.6307 0.0537 0.0000
0.6791 0.0723 0.0000
0.6813 0.0695 0.0000
0.5766 0.0579 0.0000
0.5145 0.0563 0.0000
0.5781 0.0573 0.0000
0.1319 0.0306 0.0000
0.0976 0.0310 0.0020
0.1013 0.0308 0.0010
0.0639 0.0294 0.0300
0.0719 0.0284 0.0110
-0.4530 0.1688 0.0070
-0.3542 0.1712 0.0390
-0.1509 0.1739 0.3860
-0.3510 0.1712 0.0400
0.0100 0.0972 0.9180
-0.0942 0.0764 0.2180
0.3565 0.0609 0.0000
0.3713 0.0520 0.0000
0.5388 0.0566 0.0000
-0.0694 0.0396 0.0800
-0.0428 0.0389 0.2700
0.0146 0.0346 0.6730
0.1300 0.0403 0.0010
0.1608 0.0414 0.0000
-0.0542 0.1061 0.6090
-0.1899 0.1074 0.0770
-0.2479 0.1145 0.0310
-0.3073 0.1247 0.0140
0.0086 0.1313 0.9480
10.3549 0.1868 0.0000
R-square: 0.9478
55
Reproduced with permission of the copyright owner. Further reproduction prohibited without permission.
Table 12: The Estimated Regression Model for Hospital Inpatient Operating
Expenses (using HCFA-weight Hospital Case Mix Index for all patients)
ln(Inpatient Operating Expenses) Coefficient Std. Error p-value
ln(discharges)
ln(Medicare Wage Index)
1989 dummy
1992 dummy
1995 dummy
1998 dummy
ln(CMI using HCFA weights) * 1986
ln(CMI using HCFA weights) * 1989
ln(CMI using HCFA weights) * 1992
ln(CMI using HCFA weights) * 1995
ln(CMI using HCFA weights) * 1998
For-profit * 1986
For-profit * 1989
For-profit * 1992
For-profit * 1995
For-profit * 1998
Medicare pressure index * 1989
Medicare pressure index * 1992
Medicare pressure index * 1995
Medicare pressure index * 1998
% MediCal days * 1986
% MediCal days * 1989
% MediCal days * 1992
% MediCal days * 1995
% MediCal days * 1998
ln(Market HHI) * 1986
ln(Market HHI) * 1989
ln(Market HHI) * 1992
ln(Market HHI) * 1995
ln(Market HHI) * 1998
ln(HMO penetration) * 1986
ln(HMO penetration) * 1989
ln(HMO penetration) * 1992
ln(HMO penetration) * 1995
ln(HMO penetration) * 1998
Constant
0.6571 0.0190 0.0000
0.5352 0.1338 0.0000
0.2107 0.0399 0.0000
0.4565 0.0424 0.0000
0.5920 0.0531 0.0000
0.4970 0.0567 0.0000
0.8796 0.0761 0.0000
0.8475 0.0732 0.0000
0.8112 0.0613 0.0000
0.8317 0.0575 0.0000
0.9599 0.0572 0.0000
0.0342 0.0329 0.3000
-0.0084 0.0334 0.8020
0.0310 0.0331 0.3500
0.0189 0.0316 0.5500
-0.0251 0.0305 0.4110
-0.5448 0.1823 0.0030
-0.5927 0.1848 0.0010
-0.4461 0.1874 0.0170
-0.6697 0.1844 0.0000
0.4495 0.1048 0.0000
0.2667 0.0824 0.0010
0.2954 0.0656 0.0000
0.4139 0.0560 0.0000
0.6074 0.0608 0.0000
0.1113 0.0411 0.0070
0.1094 0.0403 0.0070
0.1687 0.0353 0.0000
0.2604 0.0415 0.0000
0.2515 0.0426 0.0000
-0.1830 0.1137 0.1080
-0.2234 0.1149 0.0520
-0.2455 0.1227 0.0460
-0.2902 0.1324 0.0290
-0.0956 0.1408 0.4970
11.2985 0.1737 0.0000
R-square: 0.9191
56
Reproduced with permission of the copyright owner. Further reproduction prohibited without permission.
Table 13: The Estimated Regression Model for Hospital Inpatient Operating
Revenues (using HCFA-weight Hospital Case Mix Index for all patients)
ln(Inpatient Operating Revenues) Coefficient Std. Error______ p-value
ln(discharges)
ln(Medicare Wage Index)
1989 dummy
1992 dummy
1995 dummy
1998 dummy
ln(CMI using HCFA weights) * 1986
ln(CMI using HCFA weights) * 1989
ln(CMI using HCFA weights) * 1992
ln(CMl using HCFA weights) * 1995
ln(CMI using HCFA weights) * 1998
For-profit * 1986
For-profit * 1989
For-profit * 1992
For-profit * 1995
For-profit * 1998
Medicare pressure index * 1989
Medicare pressure index * 1992
Medicare pressure index * 1995
Medicare pressure index * 1998
% MediCal days * 1986
% MediCal days * 1989
% MediCal days * 1992
% MediCal days * 1995
% MediCal days * 1998
ln(Market HHI) * 1986
ln(Market HHI) * 1989
ln(Market HHI) * 1992
ln(Market HHI) * 1995
ln(Market HHI) * 1998
ln(HMO penetration) * 1986
ln(HMO penetration) * 1989
ln(HMO penetration) * 1992
ln(HMO penetration) * 1995
ln(HMO penetration) * 1998
Constant
0.7404 0.0208 0.0000
0.5201 0.1464 0.0000
0.1907 0.0437 0.0000
0.3748 0.0464 0.0000
0.5663 0.0581 0.0000
0.4114 0.0621 0.0000
1.0136 0.0832 0.0000
0.9906 0.0801 0.0000
0.9197 0.0670 0.0000
0.9488 0.0629 0.0000
1.0769 0.0625 0.0000
0.1749 0.0360 0.0000
0.1017 0.0365 0.0050
0.0904 0.0362 0.0130
0.0920 0.0346 0.0080
0.0837 0.0334 0.0120
-0.5604 0.1994 0.0050
-0.4323 0.2022 0.0330
-0.1468 0.2050 0.4740
-0.4459 0.2017 0.0270
0.1287 0.1147 0.2620
0.0169 0.0901 0.8510
0.4531 0.0718 0.0000
0.5128 0.0613 0.0000
0.7406 0.0666 0.0000
0.0138 0.0450 0.7580
0.0279 0.0441 0.5270
0.0685 0.0386 0.0760
0.2000 0.0455 0.0000
0.2054 0.0466 0.0000
-0.3822 0.1244 0.0020
-0.4050 0.1258 0.0010
-0.4883 0.1342 0.0000
-0.4809 0.1449 0.0010
-0.1811 0.1540 0.2400
10.5641 0.1900 0.0000
R-square: 0.9420
57
Reproduced with permission of the copyright owner. Further reproduction prohibited without permission.
Table 14: The Estimated Regression Model for Hospital Operating Expenses
(using All-Patient Cost-weight Hospital Case Mix Index)
ln(Total Operating Expenses)_______ Coefficient____ Std. Error______ p-value
ln(discharges)
ln(visits)
ln(Medicare Wage Index)
1989 dummy
1992 dummy
1995 dummy
1998 dummy
ln(CMI using Cost weights) * 1986
ln(CMI using Cost weights) * 1989
ln(CMI using Cost weights) * 1992
ln(CMI using Cost weights) * 1995
ln(CMI using Cost weights) * 1998
For-profit * 1986
For-profit * 1989
For-profit * 1992
For-profit * 1995
For-profit * 1998
Medicare pressure index * 1989
Medicare pressure index * 1992
Medicare pressure index * 1995
Medicare pressure index * 1998
% MediCal days * 1986
% MediCal days * 1989
% MediCal days * 1992
% MediCal days * 1995
% MediCal days * 1998
ln(Market HHI) * 1986
ln(Market HHI) * 1989
ln(Market HHI) * 1992
ln(Market HHI) * 1995
ln(Market HHI) * 1998
ln(HMO penetration) * 1986
ln(HMO penetration) * 1989
ln(HMO penetration) * 1992
ln(HMO penetration) * 1995
ln(HMO penetration) * 1998
Constant
0.4755 0.0175 0.0000
0.1609 0.0120 0.0000
0.4702 0.1156 0.0000
0.3206 0.0343 0.0000
0.6168 0.0369 0.0000
0.8099 0.0466 0.0000
0.8162 0.0497 0.0000
0.4746 0.0551 0.0000
0.4461 0.0500 0.0000
0.4052 0.0454 0.0000
0.3863 0.0458 0.0000
0.4295 0.0454 0.0000
-0.0098 0.0283 0.7310
-0.0141 0.0287 0.6240
0.0420 0.0285 0.1410
-0.0067 0.0272 0.8060
-0.0311 0.0263 0.2380
-0.4090 0.1563 0.0090
-0.5334 0.1587 0.0010
-0.4705 0.1612 0.0040
-0.5882 0.1587 0.0000
0.3237 0.0903 0.0000
0.1792 0.0718 0.0130
0.1898 0.0566 0.0010
0.2586 0.0482 0.0000
0.3800 0.0521 0.0000
-0.0150 0.0367 0.6830
-0.0054 0.0360 0.8810
0.0644 0.0321 0.0450
0.1409 0.0374 0.0000
0.1580 0.0383 0.0000
0.1669 0.0984 0.0900
0.0036 0.0995 0.9710
0.0177 0.1061 0.8680
-0.1160 0.1155 0.3150
0.1176 0.1218 0.3350
11.1003 0.1743 0.0000
R-square: 0.9280
58
Reproduced with permission of the copyright owner. Further reproduction prohibited without permission.
Table 15: The Estimated Regression Model for Hospital Net Revenues (using All-
Patient Cost-weight Hospital Case Mix Index)
______ ln(Total Net Revenues) Coefficient____ Std. Error p-value
ln(discharges)
In(visits)
ln(Medicare Wage Index)
1989 dummy
1992 dummy
1995 dummy
1998 dummy
ln(CMI using Cost weights) * 1986
ln(CMI using Cost weights) * 1989
ln(CMI using Cost weights) * 1992
ln(CMI using Cost weights) * 1995
ln(CMI using Cost weights) * 1998
For-profit * 1986
For-profit * 1989
For-profit * 1992
For-profit * 1995
For-profit * 1998
Medicare pressure index * 1989
Medicare pressure index * 1992
Medicare pressure index * 1995
Medicare pressure index * 1998
% MediCal days * 1986
% MediCal days * 1989
% MediCal days * 1992
% MediCal days * 1995
% MediCal days * 1998
ln(Market HHI) * 1986
ln(Market HHI) * 1989
ln(Market HHI) * 1992
ln(Market HHI) * 1995
ln(Market HHI) * 1998
ln(HMO penetration) * 1986
ln(HMO penetration) * 1989
ln(HMO penetration) * 1992
ln(HMO penetration) * 1995
ln(HMO penetration) * 1998
Constant
0.5646 0.0189 0.0000
0.1667 0.0129 0.0000
0.4530 0.1252 0.0000
0.2970 0.0372 0.0000
0.5466 0.0400 0.0000
0.7898 0.0504 0.0000
0.7481 0.0538 0.0000
0.5431 0.0596 0.0000
0.5174 0.0541 0.0000
0.4759 0.0491 0.0000
0.4442 0.0496 0.0000
0.4841 0.0491 0.0000
0.1349 0.0307 0.0000
0.1033 0.0311 0.0010
0.1014 0.0309 0.0010
0.0684 0.0295 0.0210
0.0737 0.0285 0.0100
-0.4131 0.1692 0.0150
-0.3302 0.1717 0.0550
-0.1321 0.1745 0.4490
-0.3434 0.1718 0.0460
0.0042 0.0977 0.9660
-0.0561 0.0777 0.4710
0.3697 0.0613 0.0000
0.3823 0.0522 0.0000
0.5338 0.0564 0.0000
-0.0659 0.0397 0.0970
-0.0452 0.0390 0.2460
0.0148 0.0347 0.6700
0.1335 0.0405 0.0010
0.1601 0.0415 0.0000
-0.0503 0.1065 0.6360
-0.1917 0.1077 0.0750
-0.2426 0.1148 0.0350
-0.3046 0.1250 0.0150
0.0057 0.1318 0.9660
10.3097 0.1886 0.0000
R-square: 0.9484
59
Reproduced with permission of the copyright owner. Further reproduction prohibited without permission.
Table 16: The Estimated Regression Model for Hospital Inpatient Operating
Expenses (using All-Patient Cost-weight Hospital Case Mix Index)
ln(Inpatient Operating Expenses) Coefficient Std. Error p-value
ln(discharges) 0.6602 0.0189 0.0000
ln(Medicare Wage Index) 0.5405 0.1332 0.0000
1989 dummy 0.2495 0.0396 0.0000
1992 dummy 0.5266 0.0419 0.0000
1995 dummy 0.6998 0.0526 0.0000
1998 dummy 0.6800 0.0559 0.0000
ln(CMI using Cost weights) * 1986 0.7503 0.0623 0.0000
ln(CMI using Cost weights) * 1989 0.7019 0.0567 0.0000
ln(CMI using Cost weights) * 1992 0.7033 0.0516 0.0000
ln(CMI using Cost weights) * 1995 0.7589 0.0505 0.0000
ln(CMI using Cost weights) * 1998 0.8354 0.0489 0.0000
For-profit * 1986 0.0384 0.0328 0.2420
For-profit * 1989 -0.0032 0.0332 0.9240
For-profit * 1992 0.0304 0.0330 0.3560
For-profit * 1995 0.0255 0.0314 0.4170
For-profit * 1998 -0.0233 0.0304 0.4420
Medicare pressure index * 1989 -0.5058 0.1812 0.0050
Medicare pressure index * 1992 -0.5609 0.1839 0.0020
Medicare pressure index * 1995 -0.4218 0.1866 0.0240
Medicare pressure index * 1998 -0.6666 0.1835 0.0000
% MediCal days * 1986 0.4569 0.1045 0.0000
% MediCal days * 1989 0.3343 0.0831 0.0000
% MediCal days * 1992 0.3175 0.0655 0.0000
% MediCal days * 1995 0.4352 0.0558 0.0000
% MediCal days * 1998 0.5972 0.0601 0.0000
ln(Market HHI) * 1986 0.1156 0.0409 0.0050
ln(Market HHI) * 1989 0.1059 0.0401 0.0080
ln(Market HHI) * 1992 0.1691 0.0351 0.0000
ln(Market HHI) * 1995 0.2660 0.0413 0.0000
ln(Market HHI) * 1998 0.2505 0.0424 0.0000
ln(HMO penetration) * 1986 -0.1708 0.1131 0.1310
ln(HMO penetration) * 1989 -0.2231 0.1143 0.0510
ln(HMO penetration) * 1992 -0.2317 0.1221 0.0580
ln(HMO penetration) * 1995 -0.2856 0.1317 0.0300
ln(HMO penetration) * 1998 -0.0904 0.1402 0.5190
Constant 11.2087 0.1741 0.0000
R-square: 0.9206
Reproduced with permission of the copyright owner. Further reproduction prohibited without permission.
Table 17: The Estimated Regression Model for Hospital Inpatient Operating
Revenues (using All-Patient Cost-weight Hospital Case Mix Index)
ln(Inpatient Operating Revenues) Coefficient Std. Error______p-value
ln(discharges)
ln(M edicare W age Index)
1989 dummy
1992 dummy
1995 dummy
1998 dummy
ln(CMI using Cost weights) * 1986
ln(CMI using Cost weights) * 1989
ln(CMI using Cost weights) * 1992
ln(CMI using Cost weights) * 1995
ln(CMI using Cost weights) * 1998
For-profit * 1986
For-profit * 1989
For-profit * 1992
For-profit * 1995
For-profit * 1998
Medicare pressure index * 1989
Medicare pressure index * 1992
Medicare pressure index * 1995
Medicare pressure index * 1998
% MediCal days * 1986
% MediCal days * 1989
% MediCal days * 1992
% MediCal days * 1995
% MediCal days * 1998
ln(Market HHI) * 1986
ln(Market HHI) * 1989
ln(Market HHI) * 1992
ln(Market HHI) * 1995
ln(Market HHI) * 1998
ln(HMO penetration) * 1986
ln(HMO penetration) * 1989
ln(HMO penetration) * 1992
ln(HMO penetration) * 1995
Ln(HMO penetration) * 1998
Constant
0.7411 0.0207 0.0000
0.5285 0.1462 0.0000
0.2351 0.0434 0.0000
0.4552 0.0460 0.0000
0.6896 0.0577 0.0000
0.6190 0.0614 0.0000
0.8494 0.0684 0.0000
0.8008 0.0622 0.0000
0.7874 0.0567 0.0000
0.8539 0.0554 0.0000
0.9285 0.0537 0.0000
0.1795 0.0360 0.0000
0.1079 0.0364 0.0030
0.0894 0.0362 0.0140
0.0995 0.0345 0.0040
0.0859 0.0333 0.0100
-0.5111 0.1989 0.0100
-0.3958 0.2018 0.0500
-0.1195 0.2047 0.5600
-0.4421 0.2014 0.0280
0.1317 0.1147 0.2510
0.0886 0.0912 0.3310
0.4763 0.0719 0.0000
0.5350 0.0612 0.0000
0.7283 0.0660 0.0000
0.0194 0.0449 0.6660
0.0246 0.0440 0.5760
0.0696 0.0385 0.0710
0.2073 0.0453 0.0000
0.2052 0.0465 0.0000
-0.3709 0.1241 0.0030
-0.4059 0.1255 0.0010
-0.4752 0.1340 0.0000
-0.4767 0.1445 0.0010
-0.1789 0.1539 0.2450
10.4850 0.1911 0.0000
R-square: 0.9425
61
Reproduced with permission of the copyright owner. Further reproduction prohibited without permission.
Table 18: The Estimated Regression Model for Hospital Operating Expenses
(using HCFA’s Medicare Case Mix Index)
ln(Total Operating Expenses) Coefficient Std. Error______ p-value
ln(discharges)
ln(visits)
ln(Medicare Wage Index)
1989 dummy
1992 dummy
1995 dummy
1998 dummy
ln(HCFA’s Medicare CMI) * 1986
ln(HCFA’s Medicare CMI) * 1989
ln(HCFA’s Medicare CMI) * 1992
ln(HCFA’s Medicare CMI) * 1995
ln(HCFA’s Medicare CMI) * 1998
For-profit * 1986
For-profit * 1989
For-profit * 1992
For-profit * 1995
For-profit * 1998
Medicare pressure index * 1989
Medicare pressure index * 1992
Medicare pressure index * 1995
Medicare pressure index * 1998
% MediCal days * 1986
% MediCal days * 1989
% MediCal days * 1992
% MediCal days * 1995
% MediCal days * 1998
ln(Market HHI) * 1986
ln(Market HHI) * 1989
ln(Market HHI) * 1992
ln(Market HHI) * 1995
ln(MarketHHI)* 1998
ln(HMO penetration) * 1986
ln(HMO penetration) * 1989
ln(HMO penetration) * 1992
ln(HMO penetration) * 1995
ln(HMO penetration) * 1998
Constant
0.4196 0.0165 0.0000
0.1664 0.0122 0.0000
0.5064 0.1177 0.0000
0.2727 0.0384 0.0000
0.5741 0.0419 0.0000
0.8014 0.0496 0.0000
0.7551 0.0527 0.0000
0.5211 0.0932 0.0000
0.5934 0.0829 0.0000
0.4354 0.0713 0.0000
0.2525 0.0572 0.0000
0.4510 0.0627 0.0000
-0.0226 0.0290 0.4370
-0.0298 0.0294 0.3110
0.0228 0.0291 0.4330
-0.0243 0.0279 0.3850
-0.0367 0.0269 0.1720
-0.4403 0.1587 0.0060
-0.5569 0.1616 0.0010
-0.5079 0.1641 0.0020
-0.5916 0.1614 0.0000
0.1702 0.0878 0.0530
0.0682 0.0706 0.3340
0.0984 0.0569 0.0840
0.1529 0.0491 0.0020
0.3162 0.0536 0.0000
0.0201 0.0375 0.5920
0.0334 0.0368 0.3640
0.0960 0.0327 0.0030
0.1671 0.0382 0.0000
0.2010 0.0394 0.0000
0.0808 0.1000 0.4190
-0.0731 0.1011 0.4700
-0.0524 0.1078 0.6270
-0.1382 0.1154 0.2310
0.0617 0.1214 0.6120
11.4907 0.1724 0.0000
R-square: 0.9043
62
Reproduced with permission of the copyright owner. Further reproduction prohibited without permission.
Table 19: The Estimated Regression Model for Hospital Net Revenues (using
HCFA’s Medicare Case Mix Index)
Coefficient Std. Error______ p-value ln(Total Net Revenues)
In(discharges)
ln(visits)
ln(Medicare W age Index)
1989 dummy
1992 dummy
1995 dummy
1998 dummy
ln(HCFA’s Medicare CMI) * 1986
ln(HCFA’s Medicare CMI) * 1989
ln(HCFA’s Medicare CMI) * 1992
ln(HCFA’s Medicare CMI) * 1995
ln(HCFA’s Medicare CMI) * 1998
For-profit * 1986
For-profit * 1989
For-profit * 1992
For-profit * 1995
For-profit * 1998
Medicare pressure index * 1989
Medicare pressure index * 1992
Medicare pressure index * 1995
Medicare pressure index * 1998
% MediCal days * 1986
% MediCal days * 1989
% MediCal days * 1992
% MediCal days * 1995
% MediCal days * 1998
ln(Market HHI) * 1986
ln(Market HHI) * 1989
ln(Market HHI) * 1992
ln(Market HHI) * 1995
ln(Market HHI) * 1998
ln(HMO penetration) * 1986
ln(HMO penetration) * 1989
ln(HMO penetration) * 1992
ln(HMO penetration) * 1995
ln(HMO penetration) * 1998
Constant
0.5011 0.0179 0.0000
0.1726 0.0132 0.0000
0.5031 0.1278 0.0000
0.2445 0.0417 0.0000
0.5114 0.0455 0.0000
0.7899 0.0538 0.0000
0.6936 0.0572 0.0000
0.6306 0.1012 0.0000
0.6967 0.0900 0.0000
0.4710 0.0774 0.0000
0.2784 0.0621 0.0000
0.4916 0.0680 0.0000
0.1234 0.0315 0.0000
0.0864 0.0319 0.0070
0.0788 0.0316 0.0130
0.0489 0.0303 0.1070
0.0680 0.0292 0.0200
-0.4468 0.1723 0.0100
-0.3476 0.1754 0.0480
-0.1698 0.1781 0.3410
-0.3444 0.1752 0.0500
-0.1706 0.0953 0.0740
-0.1900 0.0766 0.0130
0.2466 0.0618 0.0000
0.2543 0.0533 0.0000
0.4539 0.0581 0.0000
-0.0236 0.0407 0.5620
0.0005 0.0399 0.9910
0.0504 0.0355 0.1560
0.1638 0.0415 0.0000
0.2081 0.0427 0.0000
-0.1568 0.1085 0.1490
-0.2840 0.1097 0.0100
-0.3197 0.1170 0.0060
-0.3329 0.1253 0.0080
-0.0665 0.1318 0.6140
10.7507 0.1872 0.0000
R-square: 0.9287
63
Reproduced with permission of the copyright owner. Further reproduction prohibited without permission.
Table 20: The Estimated Regression Model for Hospital Inpatient Operating
Expenses (using HCFA’s Medicare Case Mix Index)
ln(Inpatient Operating Expenses) Coefficient Std. Error p-value
ln(discharges)
!n(Medicare Wage Index)
1989 dummy
1992 dummy
1995 dummy
1998 dummy
ln(HCFA’s Medicare CMI) * 1986
ln(HCFA’s Medicare CMI) * 1989
ln(HCFA’s Medicare CMI) * 1992
ln(HCFA’s Medicare CMI) * 1995
ln(HCFA’s Medicare CMI) * 1998
For-profit * 1986
For-profit * 1989
For-profit * 1992
For-profit * 1995
For-profit * 1998
Medicare pressure index *1989
Medicare pressure index * 1992
Medicare pressure index * 1995
Medicare pressure index * 1998
% MediCal days * 1986
% MediCal days * 1989
% MediCal days * 1992
% MediCal days * 1995
% MediCal days * 1998
ln(MarketHHI)* 1986
ln(Market HHI) * 1989
ln(Market HHI) * 1992
In(MarketHHI)* 1995
ln(MarketHHI)* 1998
ln(HMO penetration) * 1986
ln(HMO penetration) * 1989
ln(HMO penetration) * 1992
ln(HMO penetration) * 1995
ln(HMO penetration) * 1998
Constant
0.5368 0.0182 0.0000
0.5444 0.1413 0.0000
0.1958 0.0462 0.0000
0.4228 0.0497 0.0000
0.5861 0.0585 0.0000
0.4745 0.0618 0.0000
0.5498 0.1115 0.0000
0.6343 0.0984 0.0000
0.6555 0.0853 0.0000
0.5545 0.0671 0.0000
0.9009 0.0719 0.0000
0.0001 0.0350 0.9980
-0.0363 0.0353 0.3050
-0.0100 0.0350 0.7750
-0.0001 0.0335 0.9970
-0.0340 0.0323 0.2920
-0.5046 0.1916 0.0090
-0.5961 0.1951 0.0020
-0.4946 0.1978 0.0130
-0.6711 0.1945 0.0010
0.2430 0.1059 0.0220
0.1798 0.0850 0.0350
0.1816 0.0686 0.0080
0.2970 0.0591 0.0000
0.5268 0.0643 0.0000
0.1725 0.0435 0.0000
0.1714 0.0425 0.0000
0.2270 0.0373 0.0000
0.3271 0.0438 0.0000
0.3455 0.0450 0.0000
-0.2662 0.1199 0.0270
-0.2986 0.1211 0.0140
-0.3112 0.1293 0.0160
-0.2741 0.1374 0.0460
-0.1195 0.1457 0.4120
12.2703 0.1714 0.0000
R-square: 0.8653
64
Reproduced with permission of the copyright owner. Further reproduction prohibited without permission.
Table 21: The Estimated Regression Model for Hospital Inpatient Operating
Revenues (using HCFA’s Medicare Case Mix Index)
ln(Inpatient Operating Revenues)______Coefficient Std. Error______p-value
ln(discharges)
ln(Medicare Wage Index)
1989 dummy
1992 dummy
1995 dummy
1998 dummy
ln(HCFA’s Medicare CMI) * 1986
ln(HCFA’s Medicare CMI) * 1989
ln(HCFA’s Medicare CMI) * 1992
ln(HCFA’s Medicare CMI) * 1995
ln(HCFA’s Medicare CMI) * 1998
For-profit * 1986
For-profit * 1989
For-profit * 1992
For-profit * 1995
For-profit * 1998
Medicare pressure index * 1989
Medicare pressure index * 1992
Medicare pressure index * 1995
Medicare pressure index * 1998
% MediCal days * 1986
% MediCal days * 1989
% MediCal days * 1992
% MediCal days * 1995
% MediCal days * 1998
ln(Market HM) * 1986
In(MarketHHI)* 1989
ln(Market HHI) * 1992
ln(MarketHHI)* 1995
ln(Market HHI) * 1998
ln(HMO penetration) * 1986
ln(HMO penetration) * 1989
ln(HMO penetration) * 1992
ln(HMO penetration) * 1995
ln(HMO penetration) * 1998
Constant
0.6032 0.0201 0.0000
0.5462 0.1565 0.0000
0.1789 0.0512 0.0000
0.3547 0.0550 0.0000
0.5800 0.0648 0.0000
0.4161 0.0685 0.0000
0.6795 0.1235 0.0000
0.7424 0.1090 0.0000
0.7152 0.0945 0.0000
0.5927 0.0743 0.0000
0.9523 0.0796 0.0000
0.1413 0.0388 0.0000
0.0722 0.0391 0.0650
0.0448 0.0388 0.2480
0.0710 0.0371 0.0560
0.0753 0.0358 0.0350
-0.5084 0.2123 0.0170
-0.4292 0.2161 0.0470
-0.1950 0.2190 0.3730
-0.4432 0.2154 0.0400
-0.1144 0.1172 0.3290
-0.0996 0.0942 0.2910
0.3099 0.0760 0.0000
0.3622 0.0655 0.0000
0.6253 0.0712 0.0000
0.0870 0.0481 0.0710
0.1003 0.0471 0.0330
0.1347 0.0413 0.0010
0.2752 0.0485 0.0000
0.3078 0.0499 0.0000
-0.4867 0.1328 0.0000
-0.4940 0.1341 0.0000
-0.5650 0.1432 0.0000
-0.4606 0.1522 0.0030
-0.2138 0.1614 0.1850
11.6633 0.1899 0.0000
R-square: 0.9017
65
Reproduced with permission of the copyright owner. Further reproduction prohibited without permission.
Table 22: Coefficient Differences between Total Operating Expenses and Total
Net Revenues for Models with CMIs Using HCFA DRG weights Applied to All
Patients (from Tables 10 and 11)
ln(revenues) ln(expenses) Difference t-test
ln(CMI)*1986 0.6791
(0.0723)
0.5800
(0.0670)
0.0991 1.0054
ln(CMI)*1989 0.6813
(0.0695)
0.5700
(0.0644)
0.1113 1.1747
ln(CMI)*1992 0.5766
(0.0579)
0.4830
(0.0536)
0.0936 1.1863
ln(CMI)*1995 0.5145
(0.0563)
0.4358
(0.0522)
0.0787 1.0251
ln(CMI)*1998 0.5781
(0.0573)
0.5012
(0.0531)
0.0769 0.9844
Table 23: Coefficient Differences between Inpatient Operating Expenses and
Inpatient Net Revenues for Models with CMIs Using HCFA DRG Weights
Applied to All Patients (from Tables 12 and 13)
ln(revenues) ln(expenses) Difference t-test
ln(CMI)*1986 1.0136
(0.0832)
0.8796
(0.0761)
0.1340 1.1884
ln(CMI)*1989 0.9906
(0.0800)
0.8475
(0.0731)
0.1431 1.3188
ln(CMI)*1992 0.9197
(0.0670)
0.8112
(0.0613)
0.1085 1.1948
ln(CMI)*1995 0.9488
(0.0629)
0.8317
(0.0575)
0.1171 1.3741
ln(CMI)*1998 1.0769
(0.0625)
0.9599
(0.0572)
0.1170 1.3810
1 3 To evaluate the coefficient differences in 1986, for example, (0.6791-
0.5800)/V(0.0723)2 +(0.0670)2 = 1.0054 < 1.96 so the coefficient difference between total
operating expenses and total net revenues is statistically insignificant.
66
Reproduced with permission of the copyright owner. Further reproduction prohibited without permission.
Table 24: Coefficient Differences between Total Operating Expenses and Total
Net Revenues for Models with CMIs Using Cost-DRG weights Applied to All
Patients (from Tables 14 and 15)
ln(revenues) ln(expenses) Difference t-test
ln(CMI)*1986 0.5431
(0.0596)
0.4746
(0.0551)
0.0685 0.8439
ln(CMI)*1989 0.5174
(0.0541)
0.4461
(0.0500)
0.0713 0.9679
ln(CMI)*1992 0.4759
(0.0491)
0.4052
(0.0454)
0.0707 1.0572
ln(CMI)*1995 0.4442
(0.0496)
0.3863
(0.0458)
0.0579 0.8576
ln(CMI)*1998 0.4841
(0.0491)
0.4295
(0.0454)
0.0546 0.8165
Table 25: Coefficient Differences between Inpatient Operating Expenses and
Inpatient Net Revenues for Models with CMIs Using Cost-DRG weights Applied
to All Patients (from Tables 16 and 17)
ln(revenues) ln(expenses) Difference t-test
ln(CMI)*1986 0.8494
(0.0684)
0.7503
(0.0623)
0.0991 1.0711
ln(CMI)*1989 0.8008
(0.0622)
0.7019
(0.0567)
0.0989 1.1751
ln(CMI)*1992 0.7874
(0.0567)
0.7033
(0.0516)
0.0841 1.0970
ln(CMI)*1995 0.8539
(0.0554)
0.7589
(0.0505)
0.0950 1.2673
ln(CMI)*1998 0.9285
(0.0537)
0.8354
(0.0489)
0.0931 1.2819
67
Reproduced with permission of the copyright owner. Further reproduction prohibited without permission.
Table 26: Coefficient Differences between Total Operating Expenses and Total
Net Revenues for Models with HCFA’s Medicare CMIs (from Tables 18 and 19)
ln(revenues) ln(expenses) Difference t-test
ln(CMI)*1986 0.6306
(0.1012)
0.5211
(0.0932)
0.1095 0.7959
ln(CMI)*1989 0.6967
(0.0900)
0.5934
(0.0829)
0.1033 0.8442
ln(CMI)*1992 0.4710
(0.0774)
0.4354
(0.0713)
0.0356 0.3383
ln(CMI)*1995 0.2784
(0.0621)
0.2525
(0.0572)
0.0259 0.3068
ln(CMI)*1998 0.4916
(0.0680)
0.4510
(0.0627)
0.0406 0.4389
Table 27: Coefficient Differences between Inpatient Operating Expenses and
Inpatient Net Revenues for Models with HCFA’s Medicare CMIs (Tables 20 and
21)
ln(revenues) ln(expenses) Difference t-test
ln(CMI)*1986 0.6795
(0.1235)
0.5498
(0.1115)
0.1297 0.7795
ln(CMI)*1989 0.7424
(0.1090)
0.6343
(0.0984)
0.1081 0.7361
ln(CMI)*1992 0.7152
(0.0945)
0.6555
(0.0853)
0.0597 0.4690
ln(CMI)*1995 0.5927
(0.0743)
0.5545
(0.0671)
0.0382 0.3816
ln(CMI)*1998 0.9523
(0.0796)
0.9009
(0.0719)
0.0514 0.4792
Table 22 to Table 27 show the relationship between hospital revenues and
expenses by focusing on CMI. As shown, hospital CMIs are an important
predictor to understand the hospital costs and revenues behavior (since all CMI
coefficients are statistically significant for all periods). The coefficients for CMI
Reproduced with permission of the copyright owner. Further reproduction prohibited without permission.
(both total services and inpatient services) in revenue models were higher than
those in expense models, regardless of different CMIs applied. This means that
when there is an increase in hospital CMI, hospital’s revenues were higher than
hospital’s expenses.
By taking the difference between coefficients for CMI in revenue models
and those in expense models, I estimate the differential effects of CMI on hospital
costs and revenues over time. For hospital total operating expenses and revenues,
when using CMIs with HCFA DRG weights applied to all patients, HCFA’s
Medicare CMIs, or CMIs with cost weights based on all patients in California, the
differences in the coefficient for CMIs in revenue and expense models become
smaller over time, meaning that the rate of increase in hospital costs due to CMIs
is faster than that in hospital revenues. However, by closely looking at the
hospital total inpatient expenses and revenues (Table 25), when using CMIs with
cost weights based on all patients in California, the differences in the coefficients
for CMIs in revenue and expense models stayed relatively stable (except for
1992).
To investigate the performance of different CMIs to predict hospital costs
and revenues, I will compare the R-square of regression models using 1) All-
Patient CMI applying HCFA’s Medicare DRG weights, 2) All-Patient CMI
applying cost-based DRG weights, and 3) HCFA’s Medicare CMI. R-square is a
measure of “goodness of fit” and tells how well the sample regression line fits the
69
Reproduced with permission of the copyright owner. Further reproduction prohibited without permission.
data. R-square informs the proportion of variation in the dependent variable
explained by the explanatory variables and therefore provides an overall measure
of the extent to which the variation in one variable determines the variation in the
other.1 4
Table 28: Comparison on R-squares of Total Operating Expenses and Total Net
Revenues Models
Total Expense
Model
Total Revenue
Model
All-Patient CMI applying HCFA’s
Medicare DRG weights
0.9268 0.9478
All-Patient CMI applying cost-
based DRG weights
0.9280 0.9484
HCFA’s Medicare CMI 0.9043 0.9287
Table 29: Comparison on R-squares of Inpatient Operating Expenses and
Inpatient Net Revenues Models
Inpatient
Expense Model
Inpatient
Revenue Model
All-Patient CMI applying HCFA’s
Medicare DRG weights
0.9191 0.9420
All-Patient CMI applying cost-
based DRG weights
0.9206 0.9425
HCFA’s Medicare CMI 0.8653 0.9017
Tables 28 and 29 show R-squares of hospital expenses and revenues
models, both total and inpatient operation. When applying either HCFA’s
Medicare DRG weights or cost-based DRG weights to all patients, R-squares are
1 4 Gujarati DN (1995) “Basic Econometrics”, pp74-80.
70
Reproduced with permission of the copyright owner. Further reproduction prohibited without permission.
similar. Yet, when using HCFA’s Medicare CMI, the R-square is lower than
using all patient CMIs. A similar pattern can be observed when comparing R-
squares of inpatient expenses and revenues models. Thus, for total hospital costs
and revenues (inpatient costs and revenues as well), it would be possible to
conclude that the CMI based on based on all patient weights in California or the
CMI based on HCFA’s Medicare weights applied to all patients performs better
than Medicare CMI to predict hospital costs and revenues.
Results for Medicare Patients
The first four analyses (Appendices A1-A4) are using CMI with HCFA
DRG weights for Medicare beneficiaries, the second four analyses (Appendices
A5-A8) are using CMI with cost-based DRG weights.
For the regression model for Medicare hospital operating expenses
(Appendix Al), the coefficients of CMI, interacted with year dummy variable,
were positive and statistically significant for all years. This indicates that
Medicare operating expenses increased as CMI increased. The coefficients of for-
profit, interacted with year dummy variable, were negative and statistically
significant for all periods. This indicates that for-profit hospitals tend to decrease
their Medicare operating expenses compared to non-profit or government
hospitals. The coefficients of HHI were positive and statistically significant for
71
Reproduced with permission of the copyright owner. Further reproduction prohibited without permission.
years 1992, 1995 and 1998. This indicates that hospitals in a more competitive
market had lowered their operating expenses.
For hospital Medicare net revenues (Appendix A2), the coefficients of
CMI, interacted with year dummy variable, were positive and statistically
significant for all years. This indicates that hospital Medicare net revenues
increased as CMI increased. The Medicare pressure index negatively affected
hospital Medicare net revenues.
For hospital Medicare inpatient operating expenses (Appendix A3), the
coefficients of CMI and HHI, interacted with year dummy variable, were all
positive and statistically significant. The coefficients for Medicare pressure index
were negative and statistically significant. As Medicare pressure increased,
hospitals tend to lower their operating expenses. The coefficients of HMO
penetration were negative and statistically significant for years 1986 and 1995.
This indicates that HMO penetration negatively affected on hospital Medicare
inpatient operating expenses for those years.
For hospital Medicare inpatient net revenues (Appendix A4), the
coefficients of CMI, interacted with year dummy variable, were positive and
statistically significant for all years. This indicates that hospital Medicare
inpatient net revenues increased as CMI increased. The coefficients of HHI were
positive and statistically significant for all years except 1986 (p=0.066). This
indicates that hospitals in a more competitive market had lowered their inpatient
72
Reproduced with permission of the copyright owner. Further reproduction prohibited without permission.
net revenues. Medicare pressure negatively affected hospital Medicare inpatient
net revenues for all years.
It is possible to observe similar patterns of results using CMI with cost-
based DRG weights (Appendices A5-A8). The magnitude of coefficients of CMI
is similar when using HCFA DRG weights for Medicare patients compared to
using cost-based DRG weights for Medicare patients. It is possible to assume that
since HCFA’s Medicare DRG payment system was primarily designed to pay
hospital inpatient services for Medicare patients, hospitals have had more
experience to control their costs for Medicare patients within their expected
payment levels.
73
Reproduced with permission of the copyright owner. Further reproduction prohibited without permission.
Table 30: Coefficient Differences between Medicare Operating Expenses and
Medicare Net Revenues for Models with HCFA’s Medicare CMIs (from
Appendices A1 and A2)
ln(revenues) ln(expenses) Difference t-test
ln(CMI)*1986 0.9755
(0.1212)
0.9207
(0.1279)
0.0548 0.3110
ln(CMI)*1989 0.7569
(0.1085)
0.6921
(0.1145)
0.0648 0.4108
ln(CMI)*1992 0.5814
(0.0921)
0.7810
(0.0972)
-0.1996 -1.4906
ln(CMI)*1995 0.5607
(0.0800)
0.6205
(0.0844)
-0.0598 -0.5142
ln(CMI)*1998 0.4854
(0.0836)
0.4607
(0.0882)
0.0247 0.2033
Table 31: Coefficient Differences between Medicare Inpatient Operating
Expenses and Medicare Inpatient Net Revenues for Models with HCFA’s
Medicare CMIs (from Appendices A3 and A4)
ln(revenues) ln(expenses) Difference t-test
ln(CMI)*1986 1.4065
(0.1370)
1.3759
(0.1378)
0.0306 0.1575
ln(CMI)*1989 1.1837
(0.1214)
1.1403
(0.1221)
0.0434 0.2521
ln(CMI)*1992 0.8758
(0.1034)
1.0799
(0.1040)
-0.2041 -1.3917
ln(CMI)*1995 1.0174
(0.0852)
1.0715
(0.0857)
-0.0541 -0.4477
ln(CMI)*1998 1.0562
(0.0852)
1.0556
(0.0857)
0.0006 0.0050
74
Reproduced with permission of the copyright owner. Further reproduction prohibited without permission.
Table 32: Coefficient Differences between Medicare Operating Expenses and
Medicare Net Revenues for Models with Cost-Weight CMI for Medicare Patients
(from Appendices A5 and A6)
ln(revenues) ln(expenses) Difference t-test
ln(CMI)*1986 0.9322
(0.1184)
0.9005
(0.1248)
0.0317 0.1843
ln(CMI)*1989 0.7037
(0.0979)
0.6756
(0.1032)
0.0281 0.1975
ln(CMI)*1992 0.5522
(0.0879)
0.7583
(0.0927)
-0.2061 -1.6133
ln(CMI)*1995 0.5464
(0.0758)
0.6079
(0.0799)
-0.0615 -0.5584
ln(CMI)*1998 0.4911
(0.0797)
0.4718
(0.0840)
0.0193 0.1667
Table 33: Coefficient Differences between Medicare Inpatient Operating
Expenses and Medicare Inpatient Net Revenues for Models with Cost-Weight
CMI for Medicare Patients (from Appendices A7 and A8)
ln(revenues) ln(expenses) Difference t-test
ln(CMI)*1986 1.3834
(0.1323)
1.3760
(0.1330)
0.0074 0.0394
ln(CMI)*1989 1.1233
(0.1082)
1.1146
(0.1088)
0.0087 0.0567
ln(CMI)*1992 0.8413
(0.0974)
1.0514
(0.0979)
-0.2101 -1.5214
ln(CMI)*1995 1.0126
(0.0794)
1.0653
(0.0798)
-0.0527 -0.4681
ln(CMI)*1998 1.0489
(0.0792)
1.0523
(0.0796)
-0.0034 -0.0303
Table 30 to Table 33 show the relationship between hospital revenues and
expenses by focusing on Medicare CMIs. Although the coefficient difference is
statistically insignificant, the coefficients for Medicare CMI (both total services
75
Reproduced with permission of the copyright owner. Further reproduction prohibited without permission.
and inpatient services) in revenue models were higher than those in expense
models for years of 1986 and 1989. Then, the coefficients for Medicare CMI in
revenue models were lower than those in expense models for years of 1992 and
1995. That is, when there was an increase in hospital Medicare CMI (using both
HCFA weights and cost-weights applied to Medicare patients), hospital’s
revenues were lower than hospital’s expenses. For 1998, using HCFA-weight
Medicare CMI, the difference between coefficients for CMI in revenues and
expenses models was a positive value. Yet, when using cost-weight Medicare
CMI, the difference turned into a negative value.
Table 34: Comparison on R-squares of Medicare Operating Expenses and
edicare Net Revenues Models
Total Expense Total Revenue
Model Model
Medicare CMI applying cost-
based DRG weights
0.9026 0.9285
HCFA’s Medicare CMI 0.9341 0.9513
Table 35: Comparison on R-squares of Medicare Inpatient Operating Expenses
and Medicare Inpatient Net Revenues Models__________ _________________
Inpatient
Expense Model
Inpatient
Revenue Model
Medicare CMI applying cost-
based DRG weights
0.9100 0.9338
HCFA’s Medicare CMI 0.9370 0.9532
76
Reproduced with permission of the copyright owner. Further reproduction prohibited without permission.
Tables 34 and 35 show R-squares of hospital’s Medicare expenses and
revenues models. When evaluating the performance of two different CMIs to
predict Medicare expenses and revenues (both total and inpatient), the HCFA’s
Medicare CMI performs slightly better than cost-based DRG weights applied to
Medicare patients.
Results for Medi-Cal Patients
The first four analyses (Appendices B1-B4) are using CMIs with HCFA
DRG weights applied to Medi-Cal patients, the second four analyses (Appendices
B5-B8) are using CMIs with cost-based DRG weights applied to Medi-Cal
patients.
For the regression model for Medi-Cal hospital operating expenses
(Appendix Bl), the coefficients of CMI, interacted with year dummy variable,
were positive and statistically significant for all years. The coefficients of %
Medi-Cal days were positive and statistically significant. This indicates that as
Medi-Cal days increased, Medi-Cal operating expenses increased.
For hospital Medi-Cal net revenues (Appendix B2), the coefficients of
CMI, interacted with the year dummy variable, were positive statistically
significant for only two years (1986 and 1992). It can mean that hospital Medi
Cal net revenues are not closely related with the Medi-Cal case mix index. The
coefficients of % Medi-Cal days were positive and statistically significant. This
77
Reproduced with permission of the copyright owner. Further reproduction prohibited without permission.
indicates that as Medi-Cal days increased, Medi-Cal net revenues increased as
well.
For hospital Medi-Cal inpatient operating expenses (Appendix B3), the
coefficients of CMI, interacted with the year dummy variable, were all positive
and statistically significant. The coefficients of % Medi-Cal days were positive
and statistically significant. This indicates that as Medi-Cal days increased, Medi
Cal net revenues increased as well.
For hospital Medi-Cal inpatient net revenues (Appendix B4), the
coefficients of CMI, interacted with the year dummy variable, were positive and
statistically significant (except for 1995) for all years. This indicates that hospital
Medi-Cal inpatient net revenues increased as CMI increased.
It is possible to observe similar patterns of results using CMI with cost-
based DRG weights (Appendices B5-B8). The magnitude of coefficients of CMI
is slightly higher when using HCFA DRG weights for Medicare patients
compared to using cost-based DRG weights for Medicare patients.
78
Reproduced with permission of the copyright owner. Further reproduction prohibited without permission.
Table 36: Coefficient Differences between Medi-Cal Operating Expenses and
Medi-Cal Net Revenues for Models with CMIs Using HCFA’s DRG Weights
Applied to Medi-Cal Patients (Appendices B1 and B2)
ln(revenues) In(expenses) Difference t-test
ln(CMI)*1986 0.4301
(0.1689)
0.4556
(0.1210)
-0.0255 -0.1227
ln(CMI)*1989 0.1918*
(0.1638)
0.2487
(0.1172)
-0.0569 -0.2825
ln(CMI)*1992 0.3018
(0.1222)
0.2982
(0.0868)
0.0036 0.0240
ln(CMI)*1995 0.1935’
(0.1326)
0.2824
(0.0942)
-0.0889 -0.5466
ln(CMI)*1998 0.2835*
(0.1484)
0.3533
(0.1030)
-0.0698 -0.3864
* means statistically insignificant (p-value> 0.05)
Table 37: Coefficient Differences between Medi-Cal Inpatient Operating
Expenses and Inpatient Net Revenues for Models with CMIs Using HCFA’s DRG
Weights Applied to Medi-Cal Patients (from Appendices B3 and B4)
ln(revenues) ln(expenses) Difference t-test
ln(CMI)*1986 0.5445
(0.1759)
0.5232
(0.1280)
0.0213 0.0979
ln(CMI)*1989 0.3831
(0.1683)
0.3650
(0.1223)
0.0181 0.0870
ln(CMI)*1992 0.4005
(0.1279)
0.3516
(0.0922)
0.0489 0.3101
ln(CMI)*1995 0.2097*
(0.1292)
0.1829
(0.0921)
0.0268 0.1689
ln(CMI)*1998 0.3676
(0.1375)
0.4040
(0.0922)
-0.0364 -0.2199
* means statistically insignificant (p-value> 0.05)
79
Reproduced with permission of the copyright owner. Further reproduction prohibited without permission.
Table 38: Coefficient Differences between Medi-Cal Operating Expenses and
Medi-Cal Net Revenues for Models with CMIs Using Cost Weights Applied to
Medi-Cal Patients (from Appendices B5 and B6)
ln(revenues) ln(expenses) Difference t-test
ln(CMI)*1986 0.3523
(0.1321)
0.3829
(0.0943)
-0.0306 -0.1885
ln(CMI)*1989 0.1780’
(0.1205)
0.2271
(0.0861)
-0.0491 -0.3315
ln(CMI)*1992 0.2661
(0.0984)
0.2660
(0.0700)
0.0001 0.0008
ln(CMI)*1995 0.1990’
(0.1106)
0.2650
(0.0785)
-0.0660 -0.4866
ln(CMI)*1998 0.2617
(0.1172)
0.3048
(0.0811)
-0.0431 -0.3024
* means statistically insignificant (p-value> 0.05)
Table 39: Coefficient Differences between Medi-Cal Inpatient Operating
Expenses and Medi-Cal Inpatient Net Revenues for Models with CMIs Using
Cost Weights Applied to Medi-Cal Patients (from Appendices B7 and B8)
ln(revenues) ln(expenses) Difference t-test
ln(CMI)*1986 0.4460
(0.1376)
0.4337
(0.0999)
0.0123 0.0723
ln(CMI)*1989 0.3336
(0.1239)
0.3065
(0.0900)
0.0271 0.1770
ln(CMI)*1992 0.3533
(0.1031)
0.3114
(0.0745)
0.0419 0.3294
ln(CMI)*1995 0.2210
(0.1091)
0.1913
(0.0780)
0.0297 0.2215
ln(CMI)*1998 0.3318
(0.1114)
0.3499
(0.0757)
-0.0181 -0.1344
Table 36 to Table 39 show the relationship between hospital revenues and
expenses by focusing on Medi-Cal CMIs. For total Medi-Cal services, the
coefficients for Medi-Cal CMI in revenue models were lower than those in
80
Reproduced with permission of the copyright owner. Further reproduction prohibited without permission.
expense models for all years except 1992. For Medi-Cal inpatient services, the
coefficients for Medi-Cal CMI in revenue models were higher than those in
expense models except 1998. Yet, it is important to note that there is a downward
trend in the coefficient differences from 1992 and it turned to a negative value in
1998.
Table 40: Comparison on R-squares of Medi-Cal Operating Expenses and Medi-
M et Revenues Models
Total Expense
Model
Total Revenue
Model
Medi-Cal CMI applying cost-based
DRG weights
0.8905 0.8377
Medi-Cal CMI applying HCFA’s
DRG weights
0.8877 0.8351
e 41: Comparison on R-squares of Me
VIedi-Cal Inpatient Net Revenues Moc
di-Cal Inpatient Operating Expenses
els
Inpatient
Expense Model
Inpatient
Revenue Model
Medi-Cal CMI applying cost-based
DRG weights
0.8554 0.8178
Medi-Cal CMI applying HCFA’s
DRG weights
0.8523 0.8141
Tables 40 and 41 show R-squares of hospital’s Medi-Cal expenses and
revenues models. When evaluating the performance of two different CMIs to
predict Medi-Cal expenses and revenues (both total and inpatient), there is almost
no difference in R-squares.
81
Reproduced with permission of the copyright owner. Further reproduction prohibited without permission.
Results for Third-Party Patients and All Other Patients
The first four analyses (Appendices C1-C4) are using CMI with HCFA
DRG weights applied to third-party and all other patients, the second four
analyses (Appendices C5-C8) are using CMI with cost-based DRG weights
applied to third-party and all other patients.
For the regression model for hospital operating expenses for third-party
and other patients (Appendix Cl), the coefficients of CMI, interacted with the
year dummy variable, were positive and statistically significant for all years. This
indicates that hospital operating expenses for third-party and other patients
increased as CMI increased. The coefficients of HHI were positive and
statistically significant for years 1995 and 1998. This indicates that hospitals in a
more competitive market had lowered their operating expenses.
For hospital net revenues for third-party and other patients (Appendix C2),
the coefficients of CMI, interacted with the year dummy variable, were positive
and statistically significant for all years. This indicates that hospital net revenues
increased as CMI increased. The coefficients of for-profit, interacted with year
dummy variable, were positive and statistically significant. This indicates that for-
profit hospitals tend to increase their revenues compared to non-profit or
government hospitals. The coefficients of HHI were positive and statistically
significant for 1995 and 1998. This indicates that hospitals in a more competitive
82
Reproduced with permission of the copyright owner. Further reproduction prohibited without permission.
market had lowered their net revenues. The coefficients of HMO penetration were
negative and statistically significant for years 1989, 1992 and 1995. This indicates
that HMO penetration negatively affected on hospital net revenues for third-party
and others patients for those years.
For hospital inpatient operating expenses (Appendix C3), the coefficients
of CMI, interacted with the year dummy variable, were all positive and
statistically significant. The coefficients of HHI were positive and statistically
significant for years 1992, 1995 and 1998. This indicates that hospitals in a more
competitive market had lowered their inpatient operating expenses. The
coefficients of HMO penetration were negative and statistically significant for
years 1986, 1989 and 1992. This indicates that HMO penetration negatively
affected on hospital inpatient operating expenses for those years.
For hospital inpatient net revenues (Appendix C4), the coefficients of
CMI, interacted with the year dummy variable, were positive and statistically
significant for all years. This indicates that hospital inpatient net revenues
increased as CMI increased. The coefficients of for-profit, interacted with year
dummy variable, were positive and statistically significant. This indicates that for-
profit hospitals tend to increase their inpatient revenues compared to non-profit or
government hospitals. The coefficients of HHI were positive and statistically
significant for 1995 and 1998. This indicates that hospitals in a more competitive
market had lowered their inpatient net revenues. The coefficients for HMO
83
Reproduced with permission of the copyright owner. Further reproduction prohibited without permission.
penetration were ail negative and statistically significant. This indicates that HMO
penetration negatively affected hospital inpatient net revenues for third-party and
other patients.
It is possible to observe similar patterns of results in Appendices C5-C8.
However, the magnitude of coefficients of CMI is greater when using HCFA
DRG weights compared to using cost-based DRG weights.
84
Reproduced with permission of the copyright owner. Further reproduction prohibited without permission.
Table 42: Coefficient Differences between Third-Party Payer/Others Operating
Expenses and Net Revenues for Models with CMIs Using HCFA’s DRG Weights
Applied to Third-Party and All Other Patients (Appendices Cl and C2)
ln(revenues) ln(expenses) Difference t-test
ln(CMI)*1986 0.2742
(0.0853)
0.2761
(0.0715)
-0.0019 -0.0171
ln(CMI)*1989 0.4920
(0.0859)
0.4848
(0.0720)
0.0072 0.0642
ln(CMI)*1992 0.4234
(0.0718)
0.3995
(0.0602)
0.0239 0.2551
ln(CMI)*1995 0.3412
(0.0658)
0.3772
(0.0552)
-0.0360 -0.4192
ln(CMI)*1998 0.3743
(0.0642)
0.3806
(0.0538)
-0.0063 -0.0752
Table 43: Coefficient Differences between Third-Party Payer/Others Inpatient
Operating Expenses and Inpatient Net Revenues for Models with CMIs Using
HCFA’s DRG Weights Applied to Third-Party and All Other Patients
(Appendices C3 and C4)
ln(revenues) ln(expenses) Difference t-test
ln(CMI)*1986 0.6137
(0.0933)
0.5820
(0.0780)
0.0317 0.2607
ln(CMI)*1989 0.8074
(0.0938)
0.7681
(0.0784)
0.0393 0.3215
ln(CMI)* 1992 0.8142
(0.0781)
0.7784
(0.0653)
0.0358 0.3517
ln(CMI)*1995 0.7610
(0.0662)
0.7776
(0.0553)
-0.0166 -0.1924
ln(CMI)*1998 0.9541
(0.0617)
0.8455
(0.0516)
0.1086 1.3502
Reproduced with permission of the copyright owner. Further reproduction prohibited without permission.
Table 44: Coefficient Differences between Third-Party Payer/Others Operating
Expenses and Medi-Cal Net Revenues for Models with CMIs Using Cost Weights
Applied to Third-Party and All Other Patients (Appendices C5 and C6)
ln(revenues) ln(expenses) Difference t-test
ln(CMI)*1986 0.2203
(0.0683)
0.2155
(0.0573)
0.0048 0.0538
ln(CMI)*1989 0.3560
(0.0638)
0.3520
(0.0534)
0.0040 0.0481
ln(CMI)*1992 0.3441
(0.0603)
0.3284
(0.0505)
0.0157 0.1996
ln(CMI)*1995 0.2953
(0.0580)
0.3298
(0.0486)
-0.0345 -0.4559
ln(CMI)*1998 0.3139
(0.0547)
0.3228
(0.0458)
-0.0089 -0.1248
Table 45: Coefficient Differences between Third-Party Payer/Others Inpatient
Operating Expenses and Inpatient Net Revenues for Models with CMIs Using
Cost Weights Applied to Third-Party and All Other Patients (Appendices C7 and
C8)
ln(revenues) ln(expenses) Difference t-test
ln(CMI)*1986 0.5403
(0.0750)
0.5112
(0.0624)
0.0291 0.2983
ln(CMI)*1989 0.6520
(0.0699)
0.6271
(0.0583)
0.0249 0.2736
ln(CMI)*1992 0.7028
(0.0659)
0.6781
(0.0549)
0.0247 0.2880
ln(CMI)*1995 0.6963
(0.0592)
0.7152
(0.0493)
-0.0189 -0.2453
ln(CMI)*1998 0.8302
(0.0540)
0.7454
(0.0450)
0.0848 1.2064
Table 42 to Table 45 show the relationship between hospital revenues and
expenses by focusing on third-party payers and all other payers CMI. For hospital
total and inpatient services, there were mixed results in differences between
86
Reproduced with permission of the copyright owner. Further reproduction prohibited without permission.
coefficients for CMI in revenue and expense models. Yet, it is possible to say that
the general trend is a smaller difference between revenues and costs, expect for
1998.
Table 46: Comparison on R-squares of Third-Party Payers and All Others
Operating Expenses and Third-Party Payers and All Others Net Revenues Models
Total Expense
Model
Total Revenue
Model
Third-Party Payer/Others CMI
applying cost-based DRG weights
0.9392 0.9262
Third-Party Payer/Others CMI
applying HCFA’s DRG weights
0.9382 0.9263
Table 47: Comparison on R-squares of Third-Party Payers and All Others
Inpatient Operating Expenses and Third-Party Payers and All Others Inpatient Net
Revenues Models
Inpatient
Expense Model
Inpatient
Revenue Model
Third-Party Payer/Others CMI
applying cost-based DRG weights
0.9495 0.9425
Third-Party Payer/Others CMI
applying HCFA’s DRG weights
0.9486 0.9432
Tables 46 and 47 show R-squares of hospital’s third-party and all other
payers expenses and revenues models. When evaluating the performance of two
different CMIs to predict third-party payers and all other payers expenses and
revenues (both total and inpatient), there is almost no difference in R-squares.
87
Reproduced with permission of the copyright owner. Further reproduction prohibited without permission.
6. DISCUSSIONS AND IMPLICATIONS
Hospital operating expenses and revenues are affected by several factors
such as case mix index, hospital competition, managed care penetration, etc. By
focusing on hospital case mix index as a primary policy variable, this study
attempts to address both policy and methodological issues.
For policy issues, there are several ways to measure hospital case mix
index. One of the most widely used CMIs is the Medicare DRG system. There
have been several studies to test the DRG system’s performance to predict
hospital costs. Due to the clinical heterogeneity within DRGs, HCFA refined the
current Medicare DRG system and evaluated the refined DRG system. However,
the ability of these systems to explain additional variation in resource use over the
DRG classification was limited and varied considerably across DRGs (Thomas
and Ashcraft, Inquiry, 1989; MacKenzie et al., 1991) and did not provide much
benefits.
For methodological issues, when predicting hospital costs and revenues
for all patients, CMIs using all-patient weights predict better compared to
HCFA’s Medicare CMI. However, this study also showed that there is no
difference in using the all-payer cost weights or the HCFA’s Medicare weights
when predicting hospital expenses (for Medicare, Medi-Cal or others).
88
Reproduced with permission of the copyright owner. Further reproduction prohibited without permission.
Many studies employed Medicare CMI as a control variable in their
studies due to the easy data accessibility. Yet, it is important to develop CMIs
using the study population. That is, when analyzing the hospital’s cost behavior in
California, controlling for case mix index, it is necessary to develop CMIs using
all patients in California. For example, for total hospital impatient expenses and
revenues (Tables 23, 25 and 27), the differences between coefficients for CMIs in
revenue models and expense models provide the exactly opposite results when
using different case mix indices.
Managed care organizations are competing with each other for covered
lives. One of MCOs’ goals will be to minimize expenses per enrollee. This can be
done by reducing payments to hospitals so hospitals may receive a certain amount
of money (usually much lower compared to FFS) from MCOs regardless of
patient’s severity of illness. Due to hospital competition and managed care, it can
be possible to assume that hospital costs would stay stable (or increase) over time
but hospital revenues would decrease because of insufficient payment. Thus the
gap between hospital costs and revenues would be smaller. This study showed
that the differences between coefficients for CMIs in hospital costs and revenues
are getting smaller and smaller (sometimes coefficients for CMIs in hospital costs
are greater than those in hospital revenues such as for Medicare, Medi-Cal and
third-party payers). For example, Medi-Cal patients may be less likely to have a
primary care physician and more like to wait longer before seeking care.
89
Reproduced with permission of the copyright owner. Further reproduction prohibited without permission.
Therefore, they may be more likely to be more severely ill when hospitalized. In
order to guarantee the adequate quality of care provided by hospitals, it is
important to pay reasonable amounts to hospitals. Without considering the
expected utilization of resources, it could penalize some hospitals who treat more
severe patients. California state government has transferred their Medi-Cal
beneficiaries from the FFS-based system to the Medicaid managed care system.
The results in this study may imply the importance of the reasonable hospital
payment. When the California state government considers the Medicaid managed
care, it is important to
The coefficients of the case mix index were positive and significant as
expected. Yet, their values were lower than 1.00 (except for Medicare patients)
that implies a relationship between hospital case mix index and hospital costs is
not proportional.
There are several limitations in this study. The implications of this study
should be tempered by the limitations of the methods used. For example, this
study only used the HMO penetration rates as a proxy to managed care
penetration. Yet, PPO is also a popular form of managed care so it would be
better to include the PPO penetration as well as HMO penetration rates. This
study was restricted to only one state, California, which may preclude valid
generalizations to other states. However, the hospital industry in California has
some characteristics that are relevant to the hospital industry as a whole since as
90
Reproduced with permission of the copyright owner. Further reproduction prohibited without permission.
other states both Medi-Cal and third-party payers (e.g., Blue Cross) use case-
based payment for inpatient services, Medicaid managed care, etc.
Another limitation is using data for selected years instead of using the all
year data for the study period. However, it is still better than using just a one year
data (only cross-sectional). Although this study does not cover the whole years,
the variables used in this study are likely to be correlated from one period to the
next and probably do not represent a serious weakness.
Payer-specific hospital operating expenses and revenues were derived
from the proportional allocation. It is difficult to know the exact hospital costs by
different payers since there is no cost accounting available to figure out this issue.
It is important to keep developing a better hospital cost accounting method.
Despite the potential limitations, this study enhances our understanding of
the relationship between hospital case mix and hospital costs and revenues. It
contributes to evidence suggesting that 1) when there was an increase in hospital
CMIs, the differences in coefficients for CMIs in revenues and costs models
become smaller over time (especially for Medicare inpatient services, Medi-Cal
and third-party and other payers); and 2) hospital case mix index based on all
patient cost weights or HCFA’s Medicare weights perform similarly when
predicting hospital costs and revenues.
91
Reproduced with permission of the copyright owner. Further reproduction prohibited without permission.
BIBLIOGRAPHY
Anderson GF, Zhang N, Worzala C. (1999) “Hospital Expenditures and
Utilization: The Impact of HMOs,” American Journal o f Managed Care,
5(7):853-64.
Averill RF, McGuire TE, Manning BE, Fowler DA, Horn SD, Dickson PS, Coye
MJ, Knowlton DL, Bender JA. (1992) “A Study of the Relationship between
Severity of Illness and Hospital Cost in New Jersey Hospitals,” Health Services
Research, 27(5):587-606; discussion 607-12.
Bamezai A, Zwanziger J, Melnick GA, Mann JM. (1999) “Price Competition and
Hospital Cost Growth in the United States (1989-1994),” Health Economics,
8(3):233-43.
Butler PW, Bone RC, Field T. (1985) “Technology under Medicare Diagnosis-
Related Groups Prospective Payment. Implications for Medical Intensive Care,”
Chest, 87(2):229-34.
Carpenter CE, Rosko MD, Louis DZ, Yuen EJ. (1999) “Severity of Illness and
Profitability: A Patient Level Analysis,” Health Services Management Research,
12(4):217-26.
Carter GM, Rogowski JA. (1992) “How Recalibration Method, Pricing, and
Coding Affect DRG Weights,” Health Care Financing Review, 14(2):83-96.
Carter GM, Jacobson PD, Kominski GF, Perry MJ. (1994) “Use of Diagnosis-
Related Groups by Non-Medicare Payers,” Health Care Financing Review,
16(2):127-58.
Chulis GS. (1991) “Assessing Medicare's Prospective Payment System for
Hospitalsf Medical Care Review, 48(2): 167-206.
Chung J, Meltzer D. (2000) “Effects of Competition under Prospective Payment
on Hospital Costs among High and Low Cost Admissions: Evidence from
California, 1983-1993” Unpublished Paper.
Cody M, Friss L, Hawkinson ZC. (1995) “Predicting Hospital Profitability in
Short-Term General Community Hospitals,” Health Care Management Review,
20(3):77-87.
92
Reproduced with permission of the copyright owner. Further reproduction prohibited without permission.
Cotterill P, Bobular J, Connerton R. (1986) “Comparison of Alternative Relative
Weights for Diagnosis-Related Groups,” Health Care Financing Review, 7(3):37-
51.
Coulton CJ, McClish D, Doremus H, Powell S, Smookler S, Jackson DL.(1985)
“Implications of DRG payments for Medical Intensive Care,” Medical Care,
23(8):977-85.
Dor A, Farley DE. (1996) “Payment Source and the Cost of Hospital Care:
Evidence from a Multiproduct Cost Function with Multiple Payers’” Journal o f
Health Economics, 15(1): 1 -21
Eisenberg BS. (1984) “Diagnosis-Related Groups, Severity of Illness, and
Equitable Reimbursement Under Medicare,” JAMA, 251(5):645-6.
Eisenberg JM. (1985) “Physician Utilization. The State of Research about
Physicians’ Practice Patterns,” Medical Care, 23(5):461-83.
Epstein AM, Stem RS, Tognetti J, Begg CB, Hartley RM, Cumella E Jr, Ayanian
JZ. (1988) “The Association of Patients' Socioeconomic Characteristics with the
Length of Hospital Stay and Hospital Charges within Diagnosis-Related Groups,”
New England Journal o f Medicine, 318(24): 1579-85.
Farley DO. (1993) “Effects of Competition on Dialysis Facility Service Levels
and Patient Selection,” RAND Graduate School Ph.D. Dissertation.
Forgione DA, D'Annunzio CM. (1999) “The Use of DRGs in Health Care
Payment Systems Around the World,” Journal o f Health Care Finance, 26(2):66-
78.
Foster Higgins National Survey of Employer-Sponsored Health Plans, 1996.
Frank RG, Lave JR. (1985) “The psychiatric DRGs. Are they different?” Medical
Care, 23(10): 1148-55.
Gaskin DJ, Hadley J. (1997) “The Impact on HMO Penetration on the Rate of
Hospital Cost Inflation, 1985-1993,” Inquiry, 34:205-216.
Greenlick MR, Freeborn DK, Colombo TJ, Prussin JA, Saward EW. (1972)
“Comparing the Use of Medical Care Services by A Medically Indigent and A
General Membership Population in A Comprehensive Prepaid Group Practice
Program,” Medical Care, 10(3): 187-200.
93
Reproduced with permission of the copyright owner. Further reproduction prohibited without permission.
Hadley J, Swartz K. (1989) “The Impact on Hospital Costs between 1980 and
1984 of Hospital Rate Regulation, Competition, and Changes in Health Insurance
Coverage,” Inquiry, 26:35-47.
Hadley J, Zuckerman S, Feder J. (1989) “Profits and Fiscal Pressure in the
Prospective Payment System: Their Impacts on Hospitals,” Inquiry, 26(3):354-65.
Hadley J, Zuckerman S, Iezzoni LI. (1996) “Financial Pressure and Competition:
Changes in Hospital Efficiency and Cost-Shifting Behavior,” Medical Care,
34(3):205-219.
Jeffers JR, Siebert CD. (1974) “Measurement of Hospital Cost Variation: Case
Mix, Service Intensity, and Input Productivity Factors,” Health Services
Research, 9(4):293-307.
Keeler EB. (1990) “Changes in Sickness at Admission Following the Introduction
of the Prospective Payment System,” JAMA, 264( 15): 1962-1968
Kominski GF, Rice T. (1994) “Should Insurers Pay the Same Fees under An All-
Payer System?” Health Care Financing Review, 16(2): 175-89.
Lave JR, Lave LB. (1971) “The Extent of Role Differentiation Among Hospitals,”
Health Services Research, 6(1): 15-38.
Lave JR, Lave LB, Silverman LP. (1973) “A Proposal for Incentive
Reimbursement for Hospitals,” M edical Care, 11(2):79-80.
Levit KR, Lazenby HC, Braden BR. (1998) “National Health Spending Trends in
1996. National Health Accounts Team,” Health Affairs, 17(1):35-51.
Luft HS, Robinson JC, Gamick DW, Maerki SC, McPhee SJ. (1986) “The Role of
Specialized Clinical Services in Competition Among Hospitals,” Inquiry,
23(l):83-94.
Managed Care On-Line Information Exchange, 2000
Mann J. (1989) “Access to Hospital Care Under the New Jersey All-Payer
Hospital Rate-Setting System” UCLA Ph.D. Dissertation.
94
Reproduced with permission of the copyright owner. Further reproduction prohibited without permission.
Manning WG, Leibowitz A, Goldberg GA, Rogers WH, Newhouse JP. (1984) “A
Controlled Trial of the Effect of a Prepaid Group Practice on Use of Services,”
New England Journal o f Medicine, 310(23): 1505-10.
Martin SG, Frick AP, Shwartz M. (1984) “An Analysis of Hospital Case Mix,
Cost, and Payment Differences for Medicare, Medicaid, and Blue Cross Plan
Patients Using DRGs,” Inquiry, 21(4):369-79.
Melnick GA, Mann JM, Serrato CA (1988) “Hospital Costs and Patient Access
Under the New Jersey Hospital Rate-Setting System, RAND Publication.
Melnick GA, Zwanziger J. (1988) “Hospital Behavior under Competition and
Cost-Containment Policies: The California Experience, 1980 to 1985, JAMA,
260(18):2669-2675.
Melnick GA, Zwanziger J. (1995) “State Health Care Expenditures under
Competition and Regulation, 1980 through 1991,” American Journal o f Public
Health, 85(10): 1391-1396.
Morrisey MA. (2001) “Competition in Hospital and Health Insurance Markets: A
Review and Research Agenda,” Health Services Research, 36(1 Pt 2): 191-221
Munoz E, Laughlin A, Regan DM, (1985) “The Financial Effects of Emergency
Department-Generated Admissions under Prospective Payment Systems,” JAMA,
254(13):1763-1771.
Pettengill J, Vertrees J. (1982) “Reliability and Validity in Hospital Case-Mix
Measurement,” Health Care Financing Review, 4(2):101-128.
Reinhardt UE. (1996) “Spending More Through ‘Cost Control:’ Our Obsessive
Quest to Gut the Hospital,” Health Affairs, 15(2): 145-54.
Robinson JC. (1996) “Decline in Hospital Utilization and Cost Inflation Under
Managed Care in California,” JAMA, 276(13): 1060-1064.
Robinson JC. (1991) “HMO Market Penetration and Hospital Cost Inflation in
California,” JAMA, 266(19):2719-2723.
Robinson JC, Luft HS. (1987) “Competition and the Cost of Hospital Care, 1972
to 1982,” JAMA, 257(23):3241-3245.
95
Reproduced with permission of the copyright owner. Further reproduction prohibited without permission.
Robinson JC, Luft HS. (1985) “The Impact of Hospital Market Structure on
Patient Volume, Average Length of Stay, and the Cost of Care,” Journal of
Health Economics, 4:333-356.
Salkever DS, Seidman RL. (1978) “On Viewing the Medical Sector as a Market,”
Journal o f Health Politics, Policy, and Law, 3(3):361-3
Shwartz M, Young DW, Siegrist R. (1995) “The Ratio of Costs to Charges: How
Good a Basis for Estimating Costs?” Inquiry, 32:476-481.
Smits HL, Fetter RB, McMahon LF Jr. (1984) “Variation in Resource Use Within
Diagnosis-Related Groups: The Severity Issue,” Health Care Financing Review,
Suppl:71-8.
Studnicki J. (1979) “Differences in Length of Hospital Stay for Medicaid and
Blue Cross Patients and the Effect of Intensity of Services,” Public Health Report
94(5):438-45.
Thomas JW, Ashcraft ML. (1991) “Measuring Severity of Illness: Six Severity
Systems and Their Ability to Explain Costs,” Inquiry, 28:39-55.
Thomas JW, Ashcraft ML. (1989) “Measuring Severity of Illness: A Comparison
of Interrater Reliability among Severity Methodologies,” Inquiry, 26(4):483-92.
Thorpe KE. (1987) “Does All-Payer Rate Setting Work? The Case of the New
York Prospective Hospital Reimbursement Methodology, ” Journal o f Health
Politics, Policy, and Law, 12(3):391-408.
Thorpe KE, Cretin S, Keeler EB. (1988) “Are the Diagnosis-Related Group Case
Weights Compressed?” Health Care Financing Review, 10(2):37-46.
Whetsell GW. (1999) “The History and Evolution of Hospital Payment Systems:
How Did We Get Here?” Nursing Administration Quarterly, 23(4): 1-15.
Wholey D, Feldman R, Christianson JB. (1995) “The Effect of Market Structure
on HMO Premiums, ” Journal o f Health Economic, 14:81-105.
Zwanziger J, Melnick GA. (1988) “The Effects of Hospital Competition and the
Medicare PPS Program on Hospital Cost Behavior in California,” Journal o f
Health Economics, 7(4):301-320.
96
Reproduced with permission of the copyright owner. Further reproduction prohibited without permission.
Zwanziger J, Melnick G, Mann JM, Simonson L. (1994) “How Hospitals Practice
Cost Containment with Selective Contracting and the Medicare Prospective
Payment System, Medical Care, 32(11): 1153-1162.
Zwanziger J, Melnick GA, Bamezai A. (2000) “Can Cost Shifting Continue in a
Price Competitive Environment,” Health Economics, 9:211-225
Zwanziger J, Melnick GA, Bamezai A. (1994) “Cost and Price Competition in
California Hospitals, 1980-1990, ” Health Affairs, Fall:l 18-126
Zwanziger J, Melnick GA, Bamezai A. (2000) “The Effect of Selective
Contracting on Hospital Costs and Revenues,” Health Services Research,
3 5(4): 849-867
Zwanziger J, Melnick GA. (1996) “Can Managed Care Plans Control Health Care
Costs?” Health Affairs, 15(2):185-199.
97
Reproduced with permission of the copyright owner. Further reproduction prohibited without permission.
APPENDICES
Appendices A1-A8: Regression Results for Medicare Patients
A1: The Estimated Regression Model for Hospital Operating Expenses for
Medicare Patients (using HCFA-weight Hospital Case Mix Index for Medicare
Patients)
ln(Medicare Operating Expenses) Coefficient Std. Error p-value
ln(discharges) 0.6892 0.0221 0.0000
ln(visits) 0.1093 0.0114 0.0000
ln(Medicare Wage Index) 0.6563 0.1620 0.0000
1989 dummy 0.2087 0.0483 0.0000
1992 dummy 0.5811 0.0511 0.0000
1995 dummy 0.7506 0.0630 0.0000
1998 dummy 0.8219 0.0675 0.0000
ln(CMI using HCFA weights * 1986) 0.9207 0.1279 0.0000
ln(CMI using HCFA weights * 1989) 0.6921 0.1145 0.0000
ln(CMI using HCFA weights * 1992) 0.7810 0.0972 0.0000
ln(CMI using HCFA weights * 1995) 0.6205 0.0844 0.0000
ln(CMI using HCFA weights * 1998) 0.4607 0.0882 0.0000
For-profit * 1986 -0.1827 0.0398 0.0000
For-profit * 1989 -0.2080 0.0403 0.0000
For-profit * 1992 -0.0379 0.0398 0.3410
For-profit * 1995 -0.0920 0.0386 0.0170
For-profit * 1998 -0.1506 0.0370 0.0000
Medicare pressure index * 1989 -0.6717 0.2143 0.0020
Medicare pressure index * 1992 -0.6897 0.2178 0.0020
Medicare pressure index * 1995 -0.6214 0.2219 0.0050
Medicare pressure index * 1998 -0.5974 0.2195 0.0070
ln(Market HHI) * 1986 -0.0330 0.0519 0.5250
ln(MarketHHI)* 1989 0.0015 0.0508 0.9760
ln(Market HHI) * 1992 0.1008 0.0450 0.0250
ln(Market HHI) * 1995 0.1568 0.0524 0.0030
ln(Market HHI) * 1998 0.1547 0.0533 0.0040
ln(HMO penetration) * 1986 0.0832 0.1360 0.5410
ln(HMO penetration) * 1989 0.2311 0.1372 0.0920
ln(HMO penetration) * 1992 0.1033 0.1460 0.4790
ln(HMO penetration) * 1995 -0.0588 0.1572 0.7080
ln(HMO penetration) * 1998 0.0527 0.1667 0.7520
Constant 9.6289 0.1769 0.0000
R-square: 0.9341
Reproduced with permission of the copyright owner. Further reproduction prohibited without permission.
A 2 : The Estimated Regression Model for Hospital Net Revenues for Medicare
Patients (using HCFA-weight Hospital Case Mix Index for Medicare Patients)
ln(Medicare Operating Revenues) Coefficient Std. Error p-value
ln(discharges) 0.7368 0.0210 0.0000
ln(visits) 0.0999 0.0108 0.0000
ln(Medicare Wage Index) 0.4505 0.1535 0.0030
1989 dummy 0.1841 0.0458 0.0000
1992 dummy 0.5102 0.0485 0.0000
1995 dummy 0.5821 0.0597 0.0000
1998 dummy 0.6906 0.0639 0.0000
ln(CMI using HCFA weights * 1986) 0.9755 0.1212 0.0000
ln(CMI using HCFA weights * 1989) 0.7569 0.1085 0.0000
ln(CMI using HCFA weights * 1992) 0.5814 0.0921 0.0000
ln(CMI using HCFA weights * 1995) 0.5607 0.0800 0.0000
ln(CMI using HCFA weights * 1998) 0.4854 0.0836 0.0000
For-profit * 1986 0.0247 0.0377 0.5120
For-profit * 1989 -0.0266 0.0381 0.4850
For-profit * 1992 -0.0143 0.0377 0.7050
For-profit * 1995 0.0051 0.0366 0.8900
For-profit * 1998 -0.0185 0.0351 0.5970
Medicare pressure index * 1989 -0.6732 0.2031 0.0010
Medicare pressure index * 1992 -0.5382 0.2064 0.0090
Medicare pressure index * 1995 -0.5887 0.2102 0.0050
Medicare pressure index * 1998 -0.4814 0.2080 0.0210
ln(Market HHI) * 1986 -0.0653 0.0492 0.1850
ln(Market HHI) * 1989 -0.0111 0.0481 0.8170
ln(Market HHI) * 1992 0.0634 0.0426 0.1370
ln(Market HHI) * 1995 0.0248 0.0496 0.6170
ln(Market HHI) * 1998 0.0457 0.0505 0.3660
ln(HMO penetration) * 1986 -0.0622 0.1289 0.6300
ln(HMO penetration) * 1989 0.1317 0.1300 0.3110
ln(HMO penetration) * 1992 0.0587 0.1383 0.6710
ln(HMO penetration) * 1995 0.0617 0.1490 0.6790
ln(HMO penetration) * 1998 0.0546 0.1580 0.7300
Constant 9.3651 0.1677 0.0000
R-square: 0.9513
99
Reproduced with permission of the copyright owner. Further reproduction prohibited without permission.
A 3 : The Estimated Regression Model for Hospital Inpatient Operating Expenses
for Medicare Patients (using HCFA-weight Hospital Case Mix Index for
Medicare Patients)
ln(Medicare Inpatient Oper Expenses) Coefficient Std. Error_____ p-value
ln(discharges)
ln(Medicare Wage Index)
1989 dummy
1992 dummy
1995 dummy
1998 dummy
ln(CMI using HCFA weights * 1986)
ln(CMI using HCFA weights * 1989)
ln(CMI using HCFA weights * 1992)
ln(CMI using HCFA weights * 1995)
ln(CMI using HCFA weights * 1998)
For-profit * 1986
For-profit * 1989
For-profit * 1992
For-profit * 1995
For-profit * 1998
Medicare pressure index * 1989
Medicare pressure index * 1992
Medicare pressure index * 1995
Medicare pressure index * 1998
ln(Market HHI) * 1986
ln(Market HHI) * 1989
ln(Market HHI) * 1992
ln(Market HHI) * 1995
ln(Market HHI) * 1998
ln(HMO penetration) * 1986
ln(HMO penetration) * 1989
ln(HMO penetration) * 1992
ln(HMO penetration) * 1995
ln(HMO penetration) * 1998
Constant
0.7930 0.0218 0.0000
0.6193 0.1800 0.0010
0.1282 0.0542 0.0180
0.6168 0.0569 0.0000
0.7509 0.0699 0.0000
0.7460 0.0742 0.0000
1.3759 0.1378 0.0000
1.1403 0.1221 0.0000
1.0799 0.1040 0.0000
1.0715 0.0857 0.0000
1.0556 0.0857 0.0000
-0.1621 0.0440 0.0000
-0.2268 0.0444 0.0000
-0.0488 0.0442 0.2710
-0.0379 0.0424 0.3720
-0.0920 0.0410 0.0250
-0.7869 0.2410 0.0010
-0.6647 0.2448 0.0070
-0.5652 0.2490 0.0230
-0.6330 0.2461 0.0100
0.1723 0.0551 0.0020
0.1971 0.0538 0.0000
0.2917 0.0470 0.0000
0.3831 0.0549 0.0000
0.3826 0.0562 0.0000
-0.2570 0.1516 0.0900
-0.0171 0.1528 0.9110
-0.2188 0.1627 0.1790
-0.3513 0.1738 0.0430
-0.1787 0.1848 0.3340
9.8846 0.1792 0.0000
R-square: 0.9026
100
Reproduced with permission of the copyright owner. Further reproduction prohibited without permission.
A4: The Estimated Regression Model for Hospital Inpatient Net Revenues for
Medicare Patients (using HCFA-weight Hospital Case Mix Index for Medicare
Patients)
ln(Medicare Inpatient Oper Revenues) Coefficient Std. Error p-value
In(discharges) 0.8357 0.0217 0.0000
ln(Medicare Wage Index) 0.4635 0.1789 0.0100
1989 dummy 0.1048 0.0539 0.0520
1992 dummy 0.5349 0.0565 0.0000
1995 dummy 0.5701 0.0695 0.0000
1998 dummy 0.6053 0.0738 0.0000
ln(CMI using HCFA weights * 1986) 1.4065 0.1370 0.0000
ln(CMI using HCFA weights * 1989) 1.1837 0.1214 0.0000
ln(CMI using HCFA weights * 1992) 0.8758 0.1034 0.0000
ln(CMI using HCFA weights * 1995) 1.0174 0.0852 0.0000
ln(CMI using HCFA weights * 1998) 1.0562 0.0852 0.0000
For-profit * 1986 0.0343 0.0437 0.4330
For-profit * 1989 -0.0545 0.0442 0.2170
For-profit * 1992 -0.0280 0.0440 0.5250
For-profit * 1995 0.0579 0.0422 0.1700
For-profit * 1998 0.0302 0.0407 0.4590
Medicare pressure index * 1989 -0.7831 0.2396 0.0010
Medicare pressure index * 1992 -0.5296 0.2434 0.0300
Medicare pressure index * 1995 -0.5371 0.2475 0.0300
Medicare pressure index * 1998 -0.5208 0.2447 0.0330
ln(Market HHI) * 1986 0.1008 0.0547 0.0660
ln(Market HHI) * 1989 0.1450 0.0534 0.0070
ln(Market HHI) * 1992 0.2121 0.0467 0.0000
ln(Market HHI) * 1995 0.2093 0.0546 0.0000
ln(Market HHI) * 1998 0.2290 0.0558 0.0000
ln(HMO penetration) * 1986 -0.4020 0.1507 0.0080
ln(HMO penetration) * 1989 -0.1205 0.1518 0.4270
ln(HMO penetration) * 1992 -0.2639 0.1617 0.1030
ln(HMO penetration) * 1995 -0.2305 0.1728 0.1830
ln(HMO penetration) * 1998 -0.1637 0.1837 0.3730
Constant 9.5175 0.1781 0.0000
R-square: 0.9285
101
Reproduced with permission of the copyright owner. Further reproduction prohibited without permission.
A5: The Estimated Regression Model for Hospital Operating Expenses for
Medicare Patients (using Cost-weight Hospital Case Mix Index for Medicare
Patients)
ln(Medicare Operating Expenses) Coefficient Std. Error p-value
In(discharges)
ln(visits)
ln(Medicare Wage Index)
1989 dummy
1992 dummy
1995 dummy
1998 dummy
ln(CMI using Cost weights * 1986)
ln(CMI using Cost weights * 1989)
ln(CMI using Cost weights * 1992)
ln(CMI using Cost weights * 1995)
ln(CMI using Cost weights * 1998)
For-profit * 1986
For-profit * 1989
For-profit * 1992
For-profit * 1995
For-profit * 1998
Medicare pressure index * 1989
Medicare pressure index * 1992
Medicare pressure index * 1995
Medicare pressure index * 1998
ln(Market HHI) * 1986
ln(Market HHI) * 1989
ln(Market HHI) * 1992
ln(Market HHI) * 1995
ln(Market HHI) * 1998
ln(HMO penetration) * 1986
ln(HMO penetration) * 1989
ln(HMO penetration) * 1992
ln(HMO penetration) * 1995
ln(HMO penetration) * 1998
Constant
0.6897 0.0221 0.0000
0.1093 0.0113 0.0000
0.6534 0.1617 0.0000
0.2857 0.0543 0.0000
0.6939 0.0560 0.0000
0.9089 0.0675 0.0000
0.9999 0.0717 0.0000
0.9005 0.1248 0.0000
0.6756 0.1032 0.0000
0.7583 0.0927 0.0000
0.6079 0.0799 0.0000
0.4718 0.0840 0.0000
-0.1841 0.0398 0.0000
-0.2046 0.0401 0.0000
-0.0396 0.0397 0.3190
-0.0925 0.0385 0.0160
-0.1499 0.0370 0.0000
-0.6738 0.2142 0.0020
-0.6852 0.2176 0.0020
-0.6197 0.2217 0.0050
-0.5899 0.2192 0.0070
-0.0378 0.0518 0.4650
-0.0053 0.0506 0.9170
0.0949 0.0448 0.0340
0.1475 0.0522 0.0050
0.1488 0.0532 0.0050
0.0872 0.1359 0.5210
0.2180 0.1370 0.1120
0.1137 0.1457 0.4350
-0.0649 0.1571 0.6790
0.0452 0.1666 0.7860
9.4835 0.1780 0.0000
R-square: 0.9370
102
Reproduced with permission of the copyright owner. Further reproduction prohibited without permission.
A6: The Estimated Regression Model for Hospital Operating Revenues for
Medicare Patients (using Cost-weight Hospital Case Mix Index for Medicare
Patients)
In(Medicare Operating Revenues) Coefficient Std. Error p-value
ln(discharges) 0.7365 0.0209 0.0000
ln(visits) 0.1002 0.0108 0.0000
ln(Medicare Wage Index) 0.4475 0.1534 0.0040
1989 dummy 0.2668 0.0515 0.0000
1992 dummy 0.6310 0.0531 0.0000
1995 dummy 0.7394 0.0640 0.0000
1998 dummy 0.8737 0.0680 0.0000
ln(CMI using Cost weights * 1986) 0.9322 0.1184 0.0000
ln(CMI using Cost weights * 1989) 0.7037 0.0979 0.0000
ln(CMI using Cost weights * 1992) 0.5522 0.0879 0.0000
ln(CMI using Cost weights * 1995) 0.5464 0.0758 0.0000
ln(CMI using Cost weights * 1998) 0.4911 0.0797 0.0000
For-profit * 1986 0.0235 0.0377 0.5340
For-profit * 1989 -0.0218 0.0381 0.5670
For-profit * 1992 -0.0162 0.0377 0.6670
For-profit * 1995 0.0047 0.0365 0.8980
For-profit * 1998 -0.0178 0.0351 0.6120
Medicare pressure index * 1989 -0.6695 0.2032 0.0010
Medicare pressure index * 1992 -0.5311 0.2065 0.0100
Medicare pressure index * 1995 -0.5875 0.2103 0.0050
Medicare pressure index * 1998 -0.4739 0.2080 0.0230
ln(Market HHI) * 1986 -0.0688 0.0492 0.1620
In(MarketHffl)* 1989 -0.0170 0.0480 0.7240
ln(Market HHI) * 1992 0.0574 0.0425 0.1770
ln(Market HHI) * 1995 0.0169 0.0495 0.7320
ln(Market HHI) * 1998 0.0416 0.0505 0.4100
ln(HMO penetration) * 1986 -0.0570 0.1289 0.6580
ln(HMO penetration) * 1989 0.1211 0.1300 0.3520
ln(HMO penetration) * 1992 0.0686 0.1382 0.6190
ln(HMO penetration) * 1995 0.0550 0.1490 0.7120
ln(HMO penetration) * 1998 0.0451 0.1580 0.7750
Constant 9.2233 0.1689 0.0000
R-square: 0.9532
103
Reproduced with permission of the copyright owner. Further reproduction prohibited without permission.
A7: The Estimated Regression Model for Hospital Inpatient Operating Expenses
for Medicare Patients (using Cost-weight Hospital Case Mix Index for Medicare
Patients)
ln(Medicare Inpatient Oper Expenses) Coefficient Std. Error p-value
ln(discharges) 0.7964 0.0216 0.0000
ln(Medicare Wage Index) 0.6155 0.1782 0.0010
1989 dummy 0.2425 0.0604 0.0000
1992 dummy 0.7955 0.0619 0.0000
1995 dummy 1.0007 0.0739 0.0000
1998 dummy 1.0639 0.0781 0.0000
ln(CMI using Cost weights * 1986) 1.3760 0.1330 0.0000
ln(CMI using Cost weights * 1989) 1.1146 0.1088 0.0000
ln(CMI using Cost weights * 1992) 1.0514 0.0979 0.0000
ln(CMI using Cost weights * 1995) 1.0653 0.0798 0.0000
ln(CMI using Cost weights * 1998) 1.0523 0.0796 0.0000
For-profit * 1986 -0.1631 0.0436 0.0000
For-profit * 1989 -0.2196 0.0440 0.0000
For-profit * 1992 -0.0510 0.0438 0.2450
For-profit * 1995 -0.0378 0.0420 0.3690
For-profit * 1998 -0.0939 0.0406 0.0210
Medicare pressure index * 1989 -0.7924 0.2390 0.0010
Medicare pressure index * 1992 -0.6523 0.2426 0.0070
Medicare pressure index * 1995 -0.5630 0.2467 0.0230
Medicare pressure index * 1998 -0.6179 0.2439 0.0110
ln(Market HHI) * 1986 0.1634 0.0546 0.0030
ln(Market HHI) * 1989 0.1843 0.0532 0.0010
ln(Market HHI) * 1992 0.2805 0.0465 0.0000
ln(Market HHI) * 1995 0.3683 0.0544 0.0000
ln(Market HHI) * 1998 0.3734 0.0556 0.0000
ln(HMO penetration) * 1986 -0.2516 0.1501 0.0940
ln(HMO penetration) * 1989 -0.0357 0.1514 0.8140
ln(HMO penetration) * 1992 -0.1994 0.1610 0.2160
ln(HMO penetration) * 1995 -0.3609 0.1722 0.0360
ln(HMO penetration) * 1998 -0.1917 0.1831 0.2950
Constant 9.6334 0.1814 0.0000
R-square: 0.9100
104
Reproduced with permission of the copyright owner. Further reproduction prohibited without permission.
A8: The Estimated Regression Model for Hospital Inpatient Operating Revenues
for Medicare Patients (using Cost-weight Hospital Case Mix Index for Medicare
Patients)
ln(Medicare Inpatient Oper Revenues) Coefficient Std. Error p-value
ln(discharges) 0.8383 0.0215 0.0000
ln(Medicare Wage Index) 0.4594 0.1773 0.0100
1989 dummy 0.2227 0.0600 0.0000
1992 dummy 0.7182 0.0615 0.0000
1995 dummy 0.8156 0.0735 0.0000
1998 dummy 0.9222 0.0777 0.0000
ln(CMI using Cost weights * 1986) 1.3834 0.1323 0.0000
ln(CMI using Cost weights * 1989) 1.1233 0.1082 0.0000
ln(CMI using Cost weights * 1992) 0.8413 0.0974 0.0000
Jn(CMI using Cost weights * 1995) 1.0126 0.0794 0.0000
ln(CMI using Cost weights * 1998) 1.0489 0.0792 0.0000
For-profit * 1986 0.0333 0.0434 0.4430
For-profit * 1989 -0.0461 0.0437 0.2920
For-profit * 1992 -0.0304 0.0436 0.4850
For-profit * 1995 0.0584 0.0418 0.1620
For-profit * 1998 0.0285 0.0404 0.4810
Medicare pressure index * 1989 -0.7829 0.2377 0.0010
Medicare pressure index * 1992 -0.5150 0.2414 0.0330
Medicare pressure index * 1995 -0.5371 0.2454 0.0290
Medicare pressure index * 1998 -0.5063 0.2426 0.0370
In(MarketHHI)* 1986 0.0936 0.0544 0.0850
ln(Market HHI) * 1989 0.1336 0.0530 0.0120
ln(Market HHI) * 1992 0.2014 0.0462 0.0000
ln(Market HHI) * 1995 0.1966 0.0541 0.0000
ln(Market HHI) * 1998 0.2221 0.0553 0.0000
ln(HMO penetration) * 1986 -0.3953 0.1494 0.0080
ln(HMO penetration) * 1989 -0.1363 0.1506 0.3660
ln(HMO penetration) * 1992 -0.2449 0.1602 0.1270
ln(HMO penetration) * 1995 -0.2420 0.1713 0.1580
ln(HMO penetration) * 1998 -0.1783 0.1822 0.3280
Constant 9.2774 0.1804 0.0000
R-square: 0.9338
105
Reproduced with permission of the copyright owner. Further reproduction prohibited without permission.
Appendices B1-B8: Regression Results for Medi-Cal Patients
Bl: The Estimated Regression Model for Hospital Operating Expenses for Medi-
Cal Patients (using HCFA-weight Hospital CMI for Medi-Cal Patients)
ln(Medi-Cal Operating Expenses) Coefficient Std. Error p-value
ln(discharges) 0.4661 0.0219 0.0000
ln( visits) 0.2148 0.0264 0.0000
ln(Medicare Wage Index) 0.6175 0.3526 0.0800
1989 dummy 0.1356 0.1127 0.2290
1992 dummy 1.0789 0.1185 0.0000
1995 dummy 1.2848 0.1525 0.0000
1998 dummy 1.1899 0.1612 0.0000
ln(CMI using HCFA weights * 1986) 0.4556 0.1210 0.0000
ln(CMI using HCFA weights * 1989) 0.2487 0.1172 0.0340
ln(CMI using HCFA weights * 1992) 0.2982 0.0868 0.0010
ln(CMI using HCFA weights * 1995) 0.2824 0.0942 0.0030
ln(CMI using HCFA weights * 1998) 0.3533 0.1030 0.0010
For-profit * 1986 -0.1962 0.0874 0.0250
For-profit * 1989 -0.0907 0.0894 0.3100
For-profit * 1992 -0.0843 0.0873 0.3340
For-profit * 1995 -0.1038 0.0843 0.2180
For-profit * 1998 -0.0305 0.0813 0.7080
Medicare pressure index * 1989 0.0126 0.4853 0.9790
Medicare pressure index * 1992 -0.5327 0.4887 0.2760
Medicare pressure index * 1995 -0.3244 0.4988 0.5160
Medicare pressure index * 1998 -0.4118 0.4871 0.3980
% Medi-Cal days * 1986 3.6616 0.2879 0.0000
% Medi-Cal days * 1989 3.2404 0.2329 0.0000
% Medi-Cal days * 1992 1.5486 0.1701 0.0000
% Medi-Cal days * 1995 1.5463 0.1458 0.0000
% Medi-Cal days * 1998 1.6449 0.1548 0.0000
ln(Market HFH) * 1986 0.1785 0.1168 0.1270
ln(Market HHI) * 1989 0.1767 0.1133 0.1190
ln(MarketHHI)* 1992 0.0407 0.0998 0.6840
ln(Market HHI) * 1995 0.1543 0.1171 0.1880
ln(Market HFfi) * 1998 0.0898 0.1202 0.4550
ln(HMO penetration) * 1986 0.4579 0.2999 0.1270
ln(HMO penetration) * 1989 0.1992 0.3039 0.5120
ln(HMO penetration) * 1992 -0.0870 0.3234 0.7880
ln(HMO penetration) * 1995 -0.0254 0.3509 0.9420
ln(HMO penetration) * 1998 0.1274 0.3735 0.7330
Constant 9.1971 0.2694 0.0000
R-square: 0.8877
Reproduced with permission of the copyright owner. Further reproduction prohibited without permission.
B2: The Estimated Regression Model for Hospital Net Revenues for Medi-Cal
Patients (using HCFA-weight Hospital Case Mix Index for Medi-Cal Patients)
ln(Medi-Cal Operating Revenues) Coefficient Std. Error p-value
ln(discharges) 0.4882 0.0312 0.0000
ln(visits) 0.2113 0.0373 0.0000
ln(Medicare Wage Index) 0.8441 0.4911 0.0860
1989 dummy 0.1126 0.1564 0.4720
1992 dummy 0.3394 0.1654 0.0400
1995 dummy 0.2748 0.2143 0.2000
1998 dummy 0.0486 0.2269 0.8300
ln(CMI using HCFA weights * 1986) 0.4301 0.1689 0.0110
ln(CMI using HCFA weights * 1989) 0.1918 0.1638 0.2420
ln(CMI using HCFA weights * 1992) 0.3018 0.1222 0.0140
ln(CMI using HCFA weights * 1995) 0.1935 0.1326 0.1450
ln(CMI using HCFA weights * 1998) 0.2835 0.1484 0.0560
For-profit * 1986 0.0814 0.1226 0.5070
For-profit * 1989 0.1702 0.1256 0.1750
For-profit * 1992 0.0154 0.1238 0.9010
For-profit * 1995 -0.0987 0.1189 0.4070
For-profit * 1998 0.0039 0.1167 0.9730
Medicare pressure index * 1989 0.1549 0.6785 0.8190
Medicare pressure index * 1992 0.4237 0.6865 0.5370
Medicare pressure index * 1995 0.0412 0.7150 0.9540
Medicare pressure index * 1998 -0.1422 0.6962 0.8380
% Medi-Cal days * 1986 2.2312 0.4010 0.0000
% Medi-Cal days * 1989 2.0825 0.3241 0.0000
% Medi-Cal days * 1992 2.5774 0.2372 0.0000
% Medi-Cal days * 1995 2.6976 0.2031 0.0000
% Medi-Cal days * 1998 2.8869 0.2183 0.0000
ln(Market HHI) * 1986 0.2071 0.1643 0.2080
ln(Market HHI) * 1989 0.2211 0.1594 0.1660
ln(MarketHM)* 1992 0.2189 0.1412 0.1210
ln(Market HHI) * 1995 0.2273 0.1648 0.1680
ln(Market HHI) * 1998 0.1987 0.1694 0.2410
ln(HMO penetration) * 1986 0.5495 0.4181 0.1890
ln(HMO penetration) * 1989 0.3474 0.4239 0.4130
ln(HMO penetration) * 1992 0.0868 0.4522 0.8480
ln(HMO penetration) * 1995 0.1737 0.4919 0.7240
ln(HMO penetration) * 1998 0.9504 0.5307 0.0740
Constant 9.1919 0.3817 0.0000
R-square: 0.8351
107
Reproduced with permission of the copyright owner. Further reproduction prohibited without permission.
B3: The Estimated Regression Model for Hospital Inpatient Operating Expenses
for Medi-Cal Patients (using HCFA-weight Hospital Case Mix Index for Medi-
Cal Patients)
ln(Medi-Cal Inpatient Oper Expenses) Coefficient Std. Error p-value
ln(discharges) 0.6287 0.0216 0.0000
ln(Medicare Wage Index) 0.7897 0.3785 0.0370
1989 dummy 0.1323 0.1212 0.2750
1992 dummy 0.9767 0.1260 0.0000
1995 dummy 0.9941 0.1606 0.0000
1998 dummy 1.0350 0.1684 0.0000
ln(CMI using HCFA weights * 1986) 0.5232 0.1280 0.0000
ln(CMI using HCFA weights * 1989) 0.3650 0.1223 0.0030
ln(CMI using HCFA weights * 1992) 0.3516 0.0922 0.0000
ln(CMI using HCFA weights * 1995) 0.1829 0.0921 0.0470
ln(CMI using HCFA weights * 1998) 0.4040 0.0922 0.0000
For-profit * 1986 -0.1786 0.0945 0.0590
For-profit * 1989 -0.0868 0.0963 0.3680
For-profit * 1992 -0.0945 0.0942 0.3160
For-profit * 1995 -0.1235 0.0909 0.1740
For-profit * 1998 -0.0478 0.0875 0.5850
Medicare pressure index * 1989 -0.4292 0.5247 0.4130
Medicare pressure index * 1992 -0.4271 0.5285 0.4190
Medicare pressure index * 1995 -0.4900 0.5378 0.3620
Medicare pressure index * 1998 -0.3590 0.5255 0.4950
% Medi-Cal days * 1986 3.7611 0.3099 0.0000
% Medi-Cal days * 1989 3.3923 0.2497 0.0000
% Medi-Cal days * 1992 2.1073 0.1837 0.0000
% Medi-Cal days * 1995 2.1521 0.1570 0.0000
% Medi-Cal days * 1998 2.2343 0.1651 0.0000
ln(Market HHI) * 1986 0.1940 0.1207 0.1080
ln(Market HHI) * 1989 0.2008 0.1171 0.0870
ln(Market HHI) * 1992 0.1133 0.1018 0.2660
ln(Market HHI) * 1995 0.2098 0.1206 0.0820
ln(Market HHI) * 1998 0.1438 0.1237 0.2450
ln(HMO penetration) * 1986 0.1251 0.3210 0.6970
ln(HMO penetration) * 1989 -0.1712 0.3261 0.6000
ln(HMO penetration) * 1992 -0.2648 0.3462 0.4440
ln(HMO penetration) * 1995 0.0722 0.3744 0.8470
ln(HMO penetration) * 1998 0.0466 0.3962 0.9060
Constant 9.8010 0.2309 0.0000
R-square: 0.8523
108
Reproduced with permission of the copyright owner. Further reproduction prohibited without permission.
B4: The Estimated Regression Model for Hospital Inpatient Net Revenues for
Medi-Cal Patients (using HCFA-weight Hospital Case Mix Index for Medi-Cal
Patients)
ln(Medi-Cal Inpatient Oper Revenues) Coefficient Std. Error p-value
ln(discharges) 0.6531 0.0305 0.0000
ln(Medicare Wage Index) 0.7221 0.5184 0.1640
1989 dummy 0.1266 0.1652 0.4440
1992 dummy 0.2268 0.1726 0.1890
1995 dummy -0.0030 0.2216 0.9890
1998 dummy -0.1306 0.2341 0.5770
ln(CMI using HCFA weights * 1986) 0.5445 0.1759 0.0020
ln(CMI using HCFA weights * 1989) 0.3831 0.1683 0.0230
ln(CMI using HCFA weights * 1992) 0.4005 0.1279 0.0020
ln(CMI using HCFA weights * 1995) 0.2097 0.1292 0.1050
ln(CMI using HCFA weights * 1998) 0.3676 0.1375 0.0080
For-profit * 1986 0.1069 0.1301 0.4110
For-profit * 1989 0.1784 0.1330 0.1800
For-profit * 1992 0.0148 0.1314 0.9100
For-profit * 1995 -0.1194 0.1261 0.3440
For-profit * 1998 -0.0501 0.1234 0.6850
Medicare pressure index * 1989 -0.2610 0.7209 0.7170
Medicare pressure index * 1992 0.5186 0.7294 0.4770
Medicare pressure index * 1995 -0.2380 0.7575 0.7530
Medicare pressure index * 1998 -0.1866 0.7379 0.8000
% Medi-Cal days * 1986 2.3553 0.4240 0.0000
% Medi-Cal days * 1989 2.2015 0.3415 0.0000
% Medi-Cal days * 1992 3.0601 0.2518 0.0000
% Medi-Cal days * 1995 3.2604 0.2149 0.0000
% Medi-Cal days * 1998 3.4389 0.2284 0.0000
ln(Market HHI) * 1986 0.0991 0.1666 0.5520
ln(Market HHI) * 1989 0.1421 0.1616 0.3790
ln(Market HHI) * 1992 0.1475 0.1413 0.2970
ln(Market HHI) * 1995 0.1290 0.1669 0.4400
ln(Market HHI) * 1998 0.0833 0.1714 0.6270
ln(HMO penetration) * 1986 0.4178 0.4401 0.3430
ln(HMO penetration) * 1989 0.1907 0.4473 0.6700
ln(HMO penetration) * 1992 0.0420 0.4764 0.9300
tn(HMO penetration) * 1995 0.3309 0.5169 0.5220
ln(HMO penetration) * 1998 0.8836 0.5571 0.1130
Constant 9.6112 0.3208 0.0000
R-square: 0.8141
109
Reproduced with permission of the copyright owner. Further reproduction prohibited without permission.
B5: The Estimated Regression Model for Hospital Operating Expenses for Medi-
Cal Patients (using Cost-weight Hospital Case Mix Index for Medi-Cal Patients)
ln(Medi-Cal Operating Expenses) Coefficient Std. Error p-value
ln(discharges) 0.4729 0.0221 0.0000
ln(visits) 0.2151 0.0263 0.0000
ln(Medicare Wage Index) 0.6353 0.3520 0.0710
1989 dummy 0.1612 0.1107 0.1450
1992 dummy 1.1100 0.1170 0.0000
1995 dummy 1.3264 0.1524 0.0000
1998 dummy 1.2615 0.1635 0.0000
ln(CMI using Cost weights * 1986) 0.3829 0.0943 0.0000
ln(CMI using Cost weights * 1989) 0.2271 0.0861 0.0080
ln(CMI using Cost weights * 1992) 0.2660 0.0700 0.0000
ln(CMI using Cost weights * 1995) 0.2650 0.0785 0.0010
ln(CMI using Cost weights * 1998) 0.3048 0.0811 0.0000
For-profit * 1986 -0.1909 0.0874 0.0290
For-profit * 1989 -0.0913 0.0892 0.3060
For-profit * 1992 -0.0840 0.0871 0.3350
For-profit * 1995 -0.0992 0.0842 0.2390
For-profit * 1998 -0.0325 0.0812 0.6890
Medicare pressure index * 1989 0.0103 0.4842 0.9830
Medicare pressure index * 1992 -0.5413 0.4882 0.2680
Medicare pressure index * 1995 -0.3288 0.4982 0.5090
Medicare pressure index * 1998 -0.4174 0.4866 0.3910
% Medi-Cal days * 1986 3.6387 0.2881 0.0000
% Medi-Cal days * 1989 3.2590 0.2349 0.0000
% Medi-Cal days * 1992 1.5368 0.1699 0.0000
% Medi-Cal days * 1995 1.5262 0.1456 0.0000
% Medi-Cal days * 1998 1.6179 0.1544 0.0000
ln(Market HHI) * 1986 0.1870 0.1167 0.1090
ln(Market HHI) * 1989 0.1805 0.1130 0.1100
ln(Market HHI) * 1992 0.0412 0.0995 0.6790
ln(Market HHI) * 1995 0.1574 0.1168 0.1780
ln(Market HHI) * 1998 0.0898 0.1198 0.4540
ln(HMO penetration) * 1986 0.4638 0.2994 0.1220
ln(HMO penetration) * 1989 0.2032 0.3033 0.5030
ln(HMO penetration) * 1992 -0.0903 0.3228 0.7800
ln(HMO penetration) * 1995 -0.0226 0.3502 0.9480
ln(HMO penetration) * 1998 0.1246 0.3733 0.7390
Constant 9.1374 0.2694 0.0000
R-square: 0.8905
110
Reproduced with permission of the copyright owner. Further reproduction prohibited without permission.
B6: The Estimated Regression Model for Hospital Net Revenues for Medi-Cal
Patients (using Cost-weight Hospital Case Mix Index for Medi-Cal Patients)
ln(Medi-Cal Operating Revenues) Coefficient Std. Error p-value
ln(discharges) 0.4949 0.0316 0 0000
ln(visits) 0.2116 0.0372 0 0000
ln(Medicare Wage Index) 0.8618 0.4908 0 0790
1989 dummy 0.1390 0.1538 0 3660
1992 dummy 0.3723 0.1635 0 0230
1995 dummy 0.3211 0.2143 0 1340
1998 dummy 0.1234 0.2302 0 5920
ln(CMI using Cost weights * 1986) 0.3523 0.1321 0 0080
ln(CMI using Cost weights * 1989) 0.1780 0.1205 0 1400
ln(CMI using Cost weights * 1992) 0.2661 0.0984 0 0070
ln(CMI using Cost weights * 1995) 0.1990 0.1106 0 0720
ln(CMI using Cost weights * 1998) 0.2617 0.1172 0 0260
For-profit * 1986 0.0857 0.1227 0 4850
For-profit * 1989 0.1702 0.1255 0 1750
For-profit * 1992 0.0160 0.1237 0 8970
For-profit * 1995 -0.0949 0.1189 0 4250
For-profit * 1998 0.0041 0.1166 0 9720
Medicare pressure index * 1989 0.1524 0.6776 0 8220
Medicare pressure index * 1992 0.4120 0.6863 0 5480
Medicare pressure index * 1995 0.0205 0.7149 0 9770
Medicare pressure index * 1998 -0.1675 0.6964 0 8100
% Medi-Cal days * 1986 2.2030 0.4018 0 0000
% Medi-Cal days * 1989 2.0913 0.3272 0 0000
% Medi-Cal days * 1992 2.5677 0.2371 0 0000
% Medi-Cal days * 1995 2.6852 0.2031 0 0000
% Medi-Cal days * 1998 2.8699 0.2181 0 0000
ln(Market HHI) * 1986 0.2148 0.1642 0 1910
ln(Market HHI) * 1989 0.2251 0.1591 0 1570
ln(Market HHI) * 1992 0.2208 0.1409 0 1170
ln(Market HHI) * 1995 0.2323 0.1645 0 1580
ln(Market HHI) * 1998 0.2021 0.1690 0 2320
ln(HMO penetration) * 1986 0.5484 0.4177 0 1900
ln(HMO penetration) * 1989 0.3444 0.4235 0 4160
ln(HMO penetration) * 1992 0.0766 0.4518 0 8650
ln(HMO penetration) * 1995 0.1605 0.4913 0 7440
ln(HMO penetration) * 1998 0.9311 0.5308 0 0800
Constant 9.1336 0.3821 0 0000
R-square: 0.8377
Reproduced with permission of the copyright owner. Further reproduction prohibited without permission.
B7: The Estimated Regression Model for Hospital Inpatient Operating Expenses
for Medi-Cal Patients (using Cost-weight Hospital Case Mix Index for Medi-Cal
Patients)
ln(Medi-Cal Inpatient Oper Expenses) Coefficient Std. Error p-value
ln(discharges) 0.6360 0.0218 0.0000
ln(Medicare Wage Index) 0.8120 0.3780 0.0320
1989 dummy 0.1592 0.1191 0.1820
1992 dummy 1.0149 0.1244 0.0000
1995 dummy 1.0419 0.1604 0.0000
1998 dummy 1.1239 0.1698 0.0000
ln(CMI using Cost weights * 1986) 0.4337 0.0999 0.0000
ln(CMI using Cost weights * 1989) 0.3065 0.0900 0.0010
ln(CMI using Cost weights * 1992) 0.3114 0.0745 0.0000
ln(CMI using Cost weights * 1995) 0.1913 0.0780 0.0140
ln(CMI using Cost weights * 1998) 0.3499 0.0757 0.0000
For-profit * 1986 -0.1730 0.0944 0.0670
For-profit * 1989 -0.0875 0.0962 0.3630
For-profit * 1992 -0.0946 0.0941 0.3150
For-profit * 1995 -0.1197 0.0908 0.1880
For-profit * 1998 -0.0490 0.0875 0.5760
Medicare pressure index * 1989 -0.4221 0.5238 0.4200
Medicare pressure index * 1992 -0.4375 0.5281 0.4080
Medicare pressure index * 1995 -0.4944 0.5374 0.3580
Medicare pressure index * 1998 -0.3706 0.5252 0.4810
% Medi-Cal days * 1986 3.7340 0.3104 0.0000
% Medi-Cal days * 1989 3.4168 0.2520 0.0000
% Medi-Cal days * 1992 2.0961 0.1835 0.0000
% Medi-Cal days * 1995 2.1366 0.1568 0.0000
% Medi-Cal days * 1998 2.2059 0.1648 0.0000
ln(Market HHI) * 1986 0.2064 0.1204 0.0870
ln(Market HHI) * 1989 0.2079 0.1166 0.0750
ln(Market HHI) * 1992 0.1170 0.1014 0.2490
ln(Market HHI) * 1995 0.2175 0.1202 0.0710
ln(Market HHI) * 1998 0.1478 0.1234 0.2310
ln(HMO penetration) * 1986 0.1236 0.3206 0.7000
ln(HMO penetration) * 1989 -0.1720 0.3256 0.5970
ln(HMO penetration) * 1992 -0.2779 0.3458 0.4220
ln(HMO penetration) * 1995 0.0613 0.3740 0.8700
ln(HMO penetration) * 1998 0.0292 0.3967 0.9410
Constant 9.7424 0.2309 0.0000
R-square: 0.8554
112
Reproduced with permission of the copyright owner. Further reproduction prohibited without permission.
B8: The Estimated Regression Model for Hospital Inpatient Net Revenues for
Medi-Cal Patients (using HCFA-weight Hospital Case Mix Index for Medi-Cal
Patients)
ln(Medicare Inpatient Oper Revenues) Coefficient Std. Error p-value
ln(discharges) 0.6625 0.0309 0.0000
ln(Medicare Wage Index) 0.7459 0.5177 0.1500
1989 dummy 0.1601 0.1624 0.3250
1992 dummy 0.2696 0.1706 0.1140
1995 dummy 0.0510 0.2213 0.8180
1998 dummy -0.0406 0.2364 0.8640
ln(CMI using Cost weights * 1986) 0.4460 0.1376 0.0010
ln(CMI using Cost weights * 1989) 0.3336 0.1239 0.0070
ln(CMI using Cost weights * 1992) 0.3533 0.1031 0.0010
ln(CMI using Cost weights * 1995) 0.2210 0.1091 0.0430
ln(CMI using Cost weights * 1998) 0.3318 0.1114 0.0030
For-profit * 1986 0.1127 0.1301 0.3860
For-profit * 1989 0.1796 0.1329 0.1770
For-profit * 1992 0.0168 0.1313 0.8980
For-profit * 1995 -0.1137 0.1260 0.3670
For-profit * 1998 -0.0488 0.1233 0.6930
Medicare pressure index * 1989 -0.2662 0.7197 0.7120
Medicare pressure index * 1992 0.5002 0.7290 0.4930
Medicare pressure index * 1995 -0.2642 0.7572 0.7270
Medicare pressure index * 1998 -0.2200 0.7378 0.7660
% Medi-Cal days * 1986 2.3146 0.4249 0.0000
% Medi-Cal days * 1989 2.2340 0.3446 0.0000
% Medi-Cal days * 1992 3.0479 0.2516 0.0000
% Medi-Cal days * 1995 3.2452 0.2147 0.0000
% Medi-Cal days * 1998 3.4145 0.2279 0.0000
ln(Market HHI) * 1986 0.1083 0.1663 0.5150
ln(Market HHI) * 1989 0.1485 0.1611 0.3570
ln(Market HHI) * 1992 0.1502 0.1408 0.2860
ln(Market HHI) * 1995 0.1358 0.1664 0.4150
ln(Market HHI) * 1998 0.0878 0.1709 0.6070
ln(HMO penetration) * 1986 0.4170 0.4396 0.3430
ln(HMO penetration) * 1989 0.1887 0.4467 0.6730
Ln(HMO penetration) * 1992 0.0304 0.4759 0.9490
Ln(HMO penetration) * 1995 0.3206 0.5165 0.5350
Ln(HMO penetration) * 1998 0.8702 0.5574 0.1190
Constant 9.5338 0.3212 0.0000
R-square: 0.8178
113
Reproduced with permission of the copyright owner. Further reproduction prohibited without permission.
Appendices C1-C8: Regression Results for Third-Party and All Other Patients
C l: The Estimated Regression Model for Hospital Operating Expenses for Third-
Party Payers and All Others (using HCFA-weight Hospital CMI) _____
ln(Others Operating Expenses) Coefficient Std. Error p-value
ln(discharges) 0.4837 0.0182 0.0000
In(visits) 0.1817 0.0139 0.0000
ln(Medicare Wage Index) 0.2869 0.1452 0.0480
1989 dummy 0.3760 0.0533 0.0000
1992 dummy 0.3811 0.0551 0.0000
1995 dummy 0.6001 0.0668 0.0000
1998 dummy 0.6572 0.0719 0.0000
ln(CMI using HCFA weights * 1986) 0.2761 0.0715 0.0000
ln(CMI using HCFA weights * 1989) 0.4848 0.0720 0.0000
ln(CMI using HCFA weights * 1992) 0.3995 0.0602 0.0000
ln(CMI using HCFA weights * 1995) 0.3772 0.0552 0.0000
ln(CMI using HCFA weights * 1998) 0.3806 0.0538 0.0000
For-profit * 1986 -0.0057 0.0357 0.8720
For-profit * 1989 0.0031 0.0362 0.9320
For-profit * 1992 0.0052 0.0360 0.8850
For-profit * 1995 -0.0594 0.0346 0.0860
For-profit * 1998 -0.0832 0.0334 0.0130
Medicare pressure index * 1989 -0.3753 0.1961 0.0560
Medicare pressure index * 1992 -0.6931 0.1989 0.0010
Medicare pressure index * 1995 -0.3809 0.2018 0.0590
Medicare pressure index * 1998 -0.7146 0.1996 0.0000
% Others days * 1986 0.1624 0.0900 0.0710
% Others days * 1989 0.2280 0.0903 0.0120
% Others days * 1992 0.5603 0.0882 0.0000
% Others days * 1995 0.6482 0.0824 0.0000
% Others days * 1998 0.4196 0.0817 0.0000
ln(Market HHI) * 1986 -0.0468 0.0468 0.3170
ln(Market HHI) * 1989 -0.0369 0.0458 0.4200
ln(Market HHI) * 1992 0.0717 0.0403 0.0750
ln(Market HHI) * 1995 0.1709 0.0471 0.0000
ln(Market HHI) * 1998 0.1650 0.0481 0.0010
ln(HMO penetration) * 1986 0.0433 0.1246 0.7280
ln(HMO penetration) * 1989 -0.2463 0.1261 0.0510
ln(HMO penetration) * 1992 -0.0503 0.1337 0.7070
ln(HMO penetration) * 1995 -0.1672 0.1451 0.2490
ln(HMO penetration) * 1998 0.0703 0.1526 0.6450
Constant 10.6370 0.1776 0.0000
R-square: 0.9382
114
Reproduced with permission of the copyright owner. Further reproduction prohibited without permission.
C2: The Estimated Regression Model for Hospital Operating Revenues for Third-
Party Payers and All Others (using HCFA-weight Hospital Case Mix Index for
Third-Party and All Other Patients)
ln(Others Operating Revenues) Coefficient Std. Error p-value
ln(discharges) 0.5186 0.0217 0.0000
ln(visits) 0.1988 0.0166 0.0000
ln(Medicare Wage Index) 0.5209 0.1733 0.0030
1989 dummy 0.3787 0.0636 0.0000
1992 dummy 0.7056 0.0657 0.0000
1995 dummy 1.0255 0.0797 0.0000
1998 dummy 0.9988 0.0858 0.0000
ln(CMI using HCFA weights * 1986) 0.2742 0.0853 0.0010
ln(CMI using HCFA weights * 1989) 0.4920 0.0859 0.0000
ln(CMI using HCFA weights * 1992) 0.4234 0.0718 0.0000
ln(CMI using HCFA weights * 1995) 0.3412 0.0658 0.0000
ln(CMI using HCFA weights * 1998) 0.3743 0.0642 0.0000
For-profit * 1986 0.1268 0.0426 0.0030
For-profit * 1989 0.1045 0.0432 0.0160
For-profit * 1992 0.1668 0.0429 0.0000
For-profit * 1995 0.0897 0.0413 0.0300
For-profit * 1998 0.0780 0.0398 0.0510
Medicare pressure index * 1989 -0.4481 0.2341 0.0560
Medicare pressure index * 1992 -0.5704 0.2373 0.0160
Medicare pressure index * 1995 -0.0180 0.2408 0.9400
Medicare pressure index * 1998 -0.4395 0.2382 0.0650
% Others days * 1986 0.1086 0.1074 0.3120
% Others days * 1989 0.1587 0.1078 0.1410
% Others days * 1992 0.1153 0.1053 0.2730
% Others days * 1995 0.1447 0.0984 0.1420
% Others days * 1998 0.1557 0.0974 0.1100
ln(Market HHI) * 1986 -0.0501 0.0558 0.3690
ln(Market HHI) * 1989 -0.0345 0.0546 0.5280
ln(Market HHI) * 1992 0.0598 0.0481 0.2130
ln(Market HHI) * 1995 0.2405 0.0562 0.0000
ln(Market HHI) * 1998 0.3431 0.0574 0.0000
ln(HMO penetration) * 1986 -0.1724 0.1487 0.2470
ln(HMO penetration) * 1989 -0.4705 0.1504 0.0020
ln(HMO penetration) * 1992 -0.4275 0.1596 0.0070
ln(HMO penetration) * 1995 -0.6529 0.1731 0.0000
ln(HMO penetration) * 1998 -0.2596 0.1821 0.1540
Constant 10.2110 0.2119 0.0000
R-square: 0.9263
115
Reproduced with permission of the copyright owner. Further reproduction prohibited without permission.
C3: The Estimated Regression Model for Hospital Inpatient Operating Expenses
for Third-Party Payers and All Others (using HCFA-weight Hospital Case Mix
Index for Third-Party and All Other Patients)
ln(Others Inpatient Oper Expenses) Coefficient Std. Error p-value
ln(discharges) 0.7369 0.0190 0.0000
ln(Medicare Wage Index) 0.6141 0.1612 0.0000
1989 dummy 0.3653 0.0594 0.0000
1992 dummy 0.0626 0.0611 0.3060
1995 dummy 0.1600 0.0736 0.0300
1998 dummy 0.2556 0.0795 0.0010
ln(CMI using HCFA weights * 1986) 0.5820 0.0780 0.0000
ln(CMI using HCFA weights * 1989) 0.7681 0.0784 0.0000
ln(CMI using HCFA weights * 1992) 0.7784 0.0653 0.0000
ln(CMI using HCFA weights * 1995) 0.7776 0.0553 0.0000
ln(CMI using HCFA weights * 1998) 0.8455 0.0516 0.0000
For-profit * 1986 0.0459 0.0398 0.2480
For-profit * 1989 0.0339 0.0403 0.4010
For-profit * 1992 0.0257 0.0401 0.5220
For-profit * 1995 0.0056 0.0385 0.8840
For-profit * 1998 -0.0660 0.0372 0.0760
Medicare pressure index * 1989 -0.4744 0.2190 0.0300
Medicare pressure index * 1992 -0.5549 0.2220 0.0130
Medicare pressure index * 1995 0.0017 0.2250 0.9940
Medicare pressure index * 1998 -0.7312 0.2224 0.0010
% Others days * 1986 0.0004 0.1001 0.9970
% Others days * 1989 -0.0030 0.1005 0.9760
% Others days * 1992 0.6927 0.0981 0.0000
% Others days * 1995 0.8358 0.0915 0.0000
% Others days * 1998 0.5028 0.0902 0.0000
ln(Market HHI) * 1986 0.0511 0.0504 0.3110
ln(Market HHI) * 1989 0.0475 0.0493 0.3360
ln(Market HHI) * 1992 0.1460 0.0427 0.0010
ln(Market HHI) * 1995 0.2349 0.0502 0.0000
ln(Market HHI) * 1998 0.1942 0.0515 0.0000
ln(HMO penetration) * 1986 -0.3352 0.1382 0.0150
ln(HMO penetration) * 1989 -0.5967 0.1395 0.0000
ln(HMO penetration) * 1992 -0.3316 0.1484 0.0260
ln(HMO penetration) * 1995 -0.2692 0.1592 0.0910
ln(HMO penetration) * 1998 -0.2219 0.1692 0.1900
Constant 10.5034 0.1531 0.0000
R-square: 0.9486
116
Reproduced with permission of the copyright owner. Further reproduction prohibited without permission.
C4: The Estimated Regression Model for Hospital Inpatient Operating Revenues
for Third-Party Payers and All Others (using HCFA-weight Hospital Case Mix
Index for Third-Party and All Other Patients)
ln(Others Inpatient Oper Revenues) Coefficient Std. Error p-value
ln(discharges) 0.7826 0.0227 0.0000
ln(Medicare Wage Index) 0.8476 0.1928 0.0000
1989 dummy 0.3707 0.0710 0.0000
1992 dummy 0.3697 0.0730 0.0000
1995 dummy 0.5847 0.0880 0.0000
1998 dummy 0.6285 0.0951 0.0000
ln(CMI using HCFA weights * 1986) 0.6137 0.0933 0.0000
ln(CMI using HCFA weights * 1989) 0.8074 0.0938 0.0000
ln(CMI using HCFA weights * 1992) 0.8142 0.0781 0.0000
ln(CMI using HCFA weights * 1995) 0.7610 0.0662 0.0000
ln(CMI using HCFA weights * 1998) 0.9541 0.0617 0.0000
For-profit * 1986 0.1721 0.0476 0.0000
For-profit * 1989 0.1276 0.0482 0.0080
For-profit * 1992 0.1836 0.0480 0.0000
For-profit * 1995 0.1524 0.0460 0.0010
For-profit * 1998 0.0932 0.0445 0.0360
Medicare pressure index * 1989 -0.5349 0.2619 0.0410
Medicare pressure index * 1992 -0.4455 0.2655 0.0940
Medicare pressure index * 1995 0.3773 0.2692 0.1610
Medicare pressure index * 1998 -0.5100 0.2660 0.0550
% Others days * 1986 -0.0649 0.1198 0.5880
% Others days * 1989 -0.0760 0.1203 0.5280
% Others days * 1992 0.2758 0.1174 0.0190
% Others days * 1995 0.3291 0.1095 0.0030
% Others days * 1998 0.1851 0.1079 0.0860
In(MarketHHI)* 1986 -0.0208 0.0603 0.7310
ln(Market HHI) * 1989 -0.0141 0.0590 0.8110
ln(Market HHI) * 1992 0.0657 0.0510 0.1980
ln(Market HHI) * 1995 0.2398 0.0600 0.0000
ln(Market HHI) * 1998 0.3024 0.0616 0.0000
ln(HMO penetration) * 1986 -0.5280 0.1653 0.0010
ln(HMO penetration) * 1989 -0.7989 0.1669 0.0000
ln(HMO penetration) * 1992 -0.6858 0.1775 0.0000
ln(HMO penetration) * 1995 -0.7260 0.1905 0.0000
ln(HMO penetration) * 1998 -0.5533 0.2024 0.0060
Constant 10.0814 0.1832 0.0000
R-square: 0.9432
117
Reproduced with permission of the copyright owner. Further reproduction prohibited without permission.
C5: The Estimated Regression Model for Hospital Operating Expenses for Third-
Party Payers and All Others (using Cost-weight Hospital Case Mix Index for
Third-Party and All Other Patients)
ln(Others Operating Expenses) Coefficient Std. Error p-value
ln(discharges) 0.4833 0.0182 0.0000
ln(visits) 0.1848 0.0139 0.0000
ln(Medicare Wage Index) 0.2951 0.1454 0.0430
1989 dummy 0.3810 0.0533 0.0000
1992 dummy 0.3916 0.0552 0.0000
1995 dummy 0.6309 0.0672 0.0000
1998 dummy 0.7050 0.0726 0.0000
ln(CMI using Cost weights * 1986) 0.2155 0.0573 0.0000
ln(CMI using Cost weights * 1989) 0.3520 0.0534 0.0000
ln(CMI using Cost weights * 1992) 0.3284 0.0505 0.0000
ln(CMI using Cost weights * 1995) 0.3298 0.0486 0.0000
ln(CMI using Cost weights * 1998) 0.3228 0.0458 0.0000
For-profit * 1986 0.0004 0.0357 0.9910
For-profit * 1989 0.0081 0.0362 0.8220
For-profit * 1992 0.0082 0.0360 0.8200
For-profit * 1995 -0.0556 0.0346 0.1080
For-profit * 1998 -0.0793 0.0334 0.0180
Medicare pressure index * 1989 -0.3465 0.1965 0.0780
Medicare pressure index * 1992 -0.6748 0.1993 0.0010
Medicare pressure index * 1995 -0.3763 0.2024 0.0630
Medicare pressure index * 1998 -0.7118 0.2001 0.0000
% Others days * 1986 0.1458 0.0900 0.1050
% Others days * 1989 0.2126 0.0904 0.0190
% Others days * 1992 0.5567 0.0884 0.0000
% Others days * 1995 0.6219 0.0831 0.0000
% Others days * 1998 0.4141 0.0817 0.0000
ln(Market HHI) * 1986 -0.0436 0.0468 0.3520
ln(Market HHI) * 1989 -0.0404 0.0458 0.3780
ln(Market HHI) * 1992 0.0723 0.0403 0.0730
ln(Market HHI) * 1995 0.1715 0.0471 0.0000
ln(Market HHI) * 1998 0.1637 0.0482 0.0010
ln(HMO penetration) * 1986 0.0411 0.1247 0.7420
ln(HMO penetration) * 1989 -0.2522 0.1262 0.0460
ln(HMO penetration) * 1992 -0.0561 0.1338 0.6750
ln(HMO penetration) * 1995 -0.1786 0.1452 0.2190
ln(HMO penetration) * 1998 0.0642 0.1528 0.6740
Constant 10.6076 0.1783 0.0000
R-square: 0.9392
118
Reproduced with permission of the copyright owner. Further reproduction prohibited without permission.
C6: The Estimated Regression Model for Hospital Operating Revenues for Third-
Party Payers and All Others (using Cost-weight Hospital Case Mix Index for
Third-Party and All Other Patients)
ln(Others Operating Revenues) Coefficient Std. Error p-value
In(discharges) 0.5179 0.0217 0.0000
In(visits) 0.2020 0.0166 0.0000
ln(Medicare Wage Index) 0.5282 0.1735 0.0020
1989 dummy 0.3817 0.0637 0.0000
1992 dummy 0.7137 0.0658 0.0000
1995 dummy 1.0512 0.0802 0.0000
1998 dummy 1.0431 0.0866 0.0000
ln(CMI using Cost weights * 1986) 0.2203 0.0683 0.0010
ln(CMI using Cost weights * 1989) 0.3560 0.0638 0.0000
ln(CMI using Cost weights * 1992) 0.3441 0.0603 0.0000
ln(CMI using Cost weights * 1995) 0.2953 0.0580 0.0000
ln(CMI using Cost weights * 1998) 0.3139 0.0547 0.0000
For-profit * 1986 0.1333 0.0426 0.0020
For-profit * 1989 0.1096 0.0432 0.0110
For-profit * 1992 0.1695 0.0430 0.0000
For-profit * 1995 0.0935 0.0413 0.0240
For-profit * 1998 0.0818 0.0399 0.0400
Medicare pressure index * 1989 -0.4149 0.2346 0.0770
Medicare pressure index * 1992 -0.5471 0.2379 0.0220
Medicare pressure index * 1995 -0.0093 0.2416 0.9690
Medicare pressure index * 1998 -0.4325 0.2389 0.0700
% Others days * 1986 0.0943 0.1074 0.3800
% Others days * 1989 0.1440 0.1079 0.1820
% Others days * 1992 0.1130 0.1055 0.2840
% Others days * 1995 0.1204 0.0992 0.2250
% Others days * 1998 0.1506 0.0975 0.1230
ln(Market HHI) * 1986 -0.0455 0.0559 0.4160
ln(Market HHI) * 1989 -0.0369 0.0547 0.5000
ln(Market HHI) * 1992 0.0614 0.0481 0.2020
ln(Market HHI) * 1995 0.2421 0.0562 0.0000
ln(Market HHI) * 1998 0.3429 0.0575 0.0000
ln(HMO penetration) * 1986 -0.1759 0.1489 0.2380
ln(HMO penetration) * 1989 -0.4761 0.1506 0.0020
ln(HMO penetration) * 1992 -0.4319 0.1597 0.0070
ln(HMO penetration) * 1995 -0.6624 0.1733 0.0000
ln(HMO penetration) * 1998 -0.2639 0.1824 0.1480
Constant 10.1862 0.2129 0.0000
R-square: 0.9262
119
Reproduced with permission of the copyright owner. Further reproduction prohibited without permission.
Cl: The Estimated Regression Model for Hospital Inpatient Operating Expenses
for Third-Party Payers and All Others (using Cost-weight Hospital Case Mix
Index for Third-Party and All Other Patients)
ln(Others Inpatient Oper Expenses) Coefficient Std. Error p-value
ln(discharges) 0.7422 0.0190 0.0000
ln(Medicare Wage Index) 0.6283 0.1608 0.0000
1989 dummy 0.3777 0.0592 0.0000
1992 dummy 0.0812 0.0609 0.1830
1995 dummy 0.2221 0.0736 0.0030
1998 dummy 0.3549 0.0795 0.0000
ln(CMI using Cost weights * 1986) 0.5112 0.0624 0.0000
ln(CMI using Cost weights * 1989) 0.6271 0.0583 0.0000
ln(CMI using Cost weights * 1992) 0.6781 0.0549 0.0000
ln(CMI using Cost weights * 1995) 0.7152 0.0493 0.0000
ln(CMI using Cost weights * 1998) 0.7454 0.0450 0.0000
For-profit * 1986 0.0568 0.0397 0.1530
For-profit * 1989 0.0379 0.0402 0.3460
For-profit * 1992 0.0278 0.0400 0.4870
For-profit * 1995 0.0107 0.0384 0.7810
For-profit * 1998 -0.0605 0.0371 0.1030
Medicare pressure index * 1989 -0.4249 0.2185 0.0520
Medicare pressure index * 1992 -0.5029 0.2216 0.0230
Medicare pressure index * 1995 0.0327 0.2249 0.8840
Medicare pressure index * 1998 -0.7147 0.2221 0.0010
% Others days * 1986 -0.0347 0.0998 0.7280
% Others days * 1989 -0.0213 0.1002 0.8320
% Others days * 1992 0.6811 0.0980 0.0000
% Others days * 1995 0.7811 0.0918 0.0000
% Others days * 1998 0.4923 0.0899 0.0000
ln(MarketHHI)* 1986 0.0630 0.0502 0.2100
ln(Market HHI) * 1989 0.0485 0.0492 0.3240
ln(Market HHI) * 1992 0.1529 0.0425 0.0000
ln(Market HHI) * 1995 0.2423 0.0500 0.0000
ln(Market HHI) * 1998 0.1970 0.0513 0.0000
ln(HMO penetration) * 1986 -0.3403 0.1378 0.0140
ln(HMO penetration) * 1989 -0.6183 0.1391 0.0000
ln(HMO penetration) * 1992 -0.3442 0.1479 0.0200
ln(HMO penetration) * 1995 -0.2985 0.1588 0.0600
ln(HMO penetration) * 1998 -0.2272 0.1688 0.1790
Constant 10.4818 0.1528 0.0000
R-square: 0.9495
120
Reproduced with permission of the copyright owner. Further reproduction prohibited without permission.
C8: The Estimated Regression Model for Hospital Inpatient Operating Revenues
for Third-Party Payers and All Others (using Cost-weight Hospital Case Mix
Index for Third-Party and All Other Patients)
ln(Others Inpatient Oper Revenues) Coefficient Std. Error p-value
ln(discharges) 0.7866 0.0228 0.0000
ln(Medicare Wage Index) 0.8635 0.1930 0.0000
1989 dummy 0.3817 0.0710 0.0000
1992 dummy 0.3871 0.0731 0.0000
1995 dummy 0.6437 0.0883 0.0000
1998 dummy 0.7374 0.0954 0.0000
ln(CMI using Cost weights * 1986) 0.5403 0.0750 0.0000
ln(CMI using Cost weights * 1989) 0.6520 0.0699 0.0000
ln(CMI using Cost weights * 1992) 0.7028 0.0659 0.0000
ln(CMI using Cost weights * 1995) 0.6963 0.0592 0.0000
ln(CMI using Cost weights * 1998) 0.8302 0.0540 0.0000
For-profit * 1986 0.1838 0.0476 0.0000
For-profit * 1989 0.1315 0.0482 0.0070
For-profit * 1992 0.1852 0.0480 0.0000
For-profit * 1995 0.1572 0.0460 0.0010
For-profit * 1998 0.0994 0.0445 0.0260
Medicare pressure index * 1989 -0.4782 0.2623 0.0690
Medicare pressure index * 1992 -0.3878 0.2659 0.1450
Medicare pressure index * 1995 0.4121 0.2699 0.1270
Medicare pressure index * 1998 -0.4893 0.2665 0.0670
% Others days * 1986 -0.0981 0.1197 0.4130
% Others days * 1989 -0.0933 0.1202 0.4380
% Others days * 1992 0.2658 0.1176 0.0240
% Others days * 1995 0.2772 0.1102 0.0120
% Others days * 1998 0.1777 0.1079 0.1000
ln(Market HHI) * 1986 -0.0061 0.0603 0.9200
ln(Market HHI) * 1989 -0.0114 0.0590 0.8470
In(MarketHHI)* 1992 0.0742 0.0510 0.1460
ln(Market HHI) * 1995 0.2492 0.0600 0.0000
ln(Market HHI) * 1998 0.3070 0.0616 0.0000
ln(HMO penetration) * 1986 -0.5355 0.1653 0.0010
ln(HMO penetration) * 1989 -0.8215 0.1670 0.0000
ln(HMO penetration) * 1992 -0.6997 0.1775 0.0000
ln(FIMO penetration) * 1995 -0.7557 0.1906 0.0000
ln(HMO penetration) * 1998 -0.5558 0.2026 0.0060
Constant 10.0719 0.1834 0.0000
R-square: 0.9495
Reproduced with permission of the copyright owner. Further reproduction prohibited without permission.

Linked assets

University of Southern California Dissertations and Theses

Conceptually similar

PDF

A comparative study of costs and quality between case managed and non-case managed hospitals

PDF

Performance management in health care in Iceland

PDF

The effects of Medi -Cal managed care on access to mammography screening for women residing in Los Angeles County

PDF

The extent and effectiveness of nursing home regulation in the 50 states

PDF

Recruitment of family physicians into rural California: Predictors and possibilities

PDF

Does procedural fairness matter in standardization? An examination of a drug standardization process in hospitals and its impact on hospital effectiveness

PDF

Variations in physician practice patterns for eye care under the National Health Insurance of Taiwan

PDF

Physician profiling and clinical pathways: Combining the tools to change physician resource utilization

PDF

Nurse executive leadership and the relationship to organizational commitment among nurses

PDF

The effect of rate regulation and price competition on labor efficiency in ancillary and nursing service areas of hospitals in California, 1983-1991

PDF

An analysis of health risk selection and quality of care under Medicare fee -for -service and Medicare managed care health care systems

PDF

Toward a theory of networks and network effectiveness: An analytical framework for the study of multilevel networks in health care service delivery

PDF

Regionalization of fire protection and emergency medical aid services: A comparative case study analysis of economic and socio-political impacts

PDF

Dual eligible beneficiaries and managed care: Ensuring consumer protection through state Medicaid contracts

PDF

Total quality management and the South Korean ministry of defense

PDF

Evaluation of the Care Advocate Program: Bridging managed care and home community -based services

PDF

Health care finance reform and market failure at King/Drew Medical Center, Los Angeles: An analysis of pediatric social outcomes in an urban public safety -net hospital

PDF

A quantitative and qualitative evaluation of the least burdensome provisions of the Food and Drug Administration Modernization Act of 1997

PDF

Service evaluation and quality improvement of a geriatric evaluation and management program

PDF

Neighborhood network structure of social capital: A multilevel analysis of the Los Angeles experiment

Asset Metadata

Creator Lee, Keon-Hyung (author)

Core Title The effects of case mix, hospital competition, and managed care penetration on hospital costs and revenues

Degree Doctor of Philosophy

Degree Program Public Administration

Publisher University of Southern California (original), University of Southern California. Libraries (digital)

Tag health sciences, health care management,OAI-PMH Harvest,Political Science, public administration

Language English

Contributor Digitized by ProQuest (provenance)

Permanent Link (DOI) https://doi.org/10.25549/usctheses-c16-547308

Unique identifier UC11334919

Identifier 3094352.pdf (filename),usctheses-c16-547308 (legacy record id)

Legacy Identifier 3094352.pdf

Dmrecord 547308

Document Type Dissertation

Rights Lee, Keon-Hyung

Type texts

Source University of Southern California (contributing entity), University of Southern California Dissertations and Theses (collection)

Access Conditions The author retains rights to his/her dissertation, thesis or other graduate work according to U.S. copyright law. Electronic access is being provided by the USC Libraries in agreement with the au...

Repository Name University of Southern California Digital Library

Repository Location USC Digital Library, University of Southern California, University Park Campus, Los Angeles, California 90089, USA