USF Libraries
USF Digital Collections

Certificate of need regulation in the nursing home industry

MISSING IMAGE

Material Information

Title:
Certificate of need regulation in the nursing home industry has it outlived its usefulness?
Physical Description:
Book
Language:
English
Creator:
Caldwell, Barbara J
Publisher:
University of South Florida
Place of Publication:
Tampa, Fla
Publication Date:

Subjects

Subjects / Keywords:
Access
Quality
Excess demand
Reimbursement
Medicaid
Dissertations, Academic -- Business Administration -- Doctoral -- USF
Genre:
bibliography   ( marcgt )
theses   ( marcgt )
non-fiction   ( marcgt )

Notes

Abstract:
ABSTRACT: The primary goals of the National Health Planning and Resources Development Act (P.L. 93-641) of 1974 were to (1) contain health care costs and (2) increase the accessibility and quality of health services. Certificate of need (CON) regulation is one attempt to constrain health care costs by limiting the supply of certain medical care facilities. With respect to the nursing home industry, prospective nursing home owners/operators are required to demonstrate that a "need" exists for more nursing home beds. Some States also imposed a construction moratorium that prevented any expansion of existing facilities or construction of new facilities regardless of whether or not a "need" existed. These CON/moratorium programs impose a supply side constraint that creates a potential barrier to entry and in the presence of excess demand may cause a nursing home bed shortage for those residents covered by Medicaid. Even though the Federal CON requirement lapsed in 1986, forty-two St ates and the District of Columbia continue to have a CON, a construction moratorium, or both for nursing home facilities. Yet maintaining these regulations comes with a cost.This paper investigates if differences exist in the quality of care and the access to care between nursing homes in those States without CON and/or construction moratorium and those States that still have such policies. Using data for the years 1991 through 2003 for all freestanding Medicaid-/Medicare-certified nursing home facilities in the United States and employing state and facility fixed effects models we find that Medicaid-eligible residents in those states without CON and/or construction moratorium policies have more access to a nursing home bed than those individuals in states with these policies. With respect to quality of care the results are mixed depending on the measure of quality that is employed. With the risk of becoming a nursing home resident at the age of 65 at 44 percent and at the age of 8 5 at 53 percent (Spillman and Lubitz 2002) coupled with the aging of the current population, the areas of quality of care and access to care remain important policy issues in the nursing home industry.
Thesis:
Dissertation (Ph.D.)--University of South Florida, 2006.
Bibliography:
Includes bibliographical references.
System Details:
System requirements: World Wide Web browser and PDF reader.
System Details:
Mode of access: World Wide Web.
Statement of Responsibility:
by Barbara J. Caldwell.
General Note:
Title from PDF of title page.
General Note:
Document formatted into pages; contains 183 pages.
General Note:
Includes vita.

Record Information

Source Institution:
University of South Florida Library
Holding Location:
University of South Florida
Rights Management:
All applicable rights reserved by the source institution and holding location.
Resource Identifier:
aleph - 001796878
oclc - 156942020
usfldc doi - E14-SFE0001621
usfldc handle - e14.1621
System ID:
SFS0025939:00001


This item is only available as the following downloads:


Full Text
xml version 1.0 encoding UTF-8 standalone no
record xmlns http:www.loc.govMARC21slim xmlns:xsi http:www.w3.org2001XMLSchema-instance xsi:schemaLocation http:www.loc.govstandardsmarcxmlschemaMARC21slim.xsd
leader nam Ka
controlfield tag 001 001796878
003 fts
005 20070724142829.0
006 m||||e|||d||||||||
007 cr mnu|||uuuuu
008 070724s2006 flu sbm 000 0 eng d
datafield ind1 8 ind2 024
subfield code a E14-SFE0001621
040
FHM
c FHM
035
(OCoLC)156942020
049
FHMM
090
HF71 (ONLINE)
1 100
Caldwell, Barbara J.
0 245
Certificate of need regulation in the nursing home industry :
b has it outlived its usefulness?
h [electronic resource] /
by Barbara J. Caldwell.
260
[Tampa, Fla] :
University of South Florida,
2006.
3 520
ABSTRACT: The primary goals of the National Health Planning and Resources Development Act (P.L. 93-641) of 1974 were to (1) contain health care costs and (2) increase the accessibility and quality of health services. Certificate of need (CON) regulation is one attempt to constrain health care costs by limiting the supply of certain medical care facilities. With respect to the nursing home industry, prospective nursing home owners/operators are required to demonstrate that a "need" exists for more nursing home beds. Some States also imposed a construction moratorium that prevented any expansion of existing facilities or construction of new facilities regardless of whether or not a "need" existed. These CON/moratorium programs impose a supply side constraint that creates a potential barrier to entry and in the presence of excess demand may cause a nursing home bed shortage for those residents covered by Medicaid. Even though the Federal CON requirement lapsed in 1986, forty-two St ates and the District of Columbia continue to have a CON, a construction moratorium, or both for nursing home facilities. Yet maintaining these regulations comes with a cost.This paper investigates if differences exist in the quality of care and the access to care between nursing homes in those States without CON and/or construction moratorium and those States that still have such policies. Using data for the years 1991 through 2003 for all freestanding Medicaid-/Medicare-certified nursing home facilities in the United States and employing state and facility fixed effects models we find that Medicaid-eligible residents in those states without CON and/or construction moratorium policies have more access to a nursing home bed than those individuals in states with these policies. With respect to quality of care the results are mixed depending on the measure of quality that is employed. With the risk of becoming a nursing home resident at the age of 65 at 44 percent and at the age of 8 5 at 53 percent (Spillman and Lubitz 2002) coupled with the aging of the current population, the areas of quality of care and access to care remain important policy issues in the nursing home industry.
502
Dissertation (Ph.D.)--University of South Florida, 2006.
504
Includes bibliographical references.
516
Text (Electronic dissertation) in PDF format.
538
System requirements: World Wide Web browser and PDF reader.
Mode of access: World Wide Web.
500
Title from PDF of title page.
Document formatted into pages; contains 183 pages.
Includes vita.
590
Adviser: Gabriel A. Picone, Ph.D.
653
Access.
Quality.
Excess demand.
Reimbursement.
Medicaid.
690
Dissertations, Academic
z USF
x Business Administration
Doctoral.
773
t USF Electronic Theses and Dissertations.
4 856
u http://digital.lib.usf.edu/?e14.1621



PAGE 1

Certificate of Need Regulation in the Nursing Home Industry: Has it Outlived its Usefulness? by Barbara J. Caldwell A dissertation submitted in partial fulfillment of the requirements for the degree of Doctor of Philosophy Department of Economics College of Business Administration University of South Florida Major Professor: Gabr iel A. Picone, Ph.D. Bradley P. Kamp, Ph.D. Philip K. Porter, Ph.D. R. Mark Wilson, Ph.D. Date of Approval: July 14, 2006 Keywords: access, quality, excess demand, reimbursement, Medicaid Copyright 2006, Barbara J. Caldwell

PAGE 2

Dedication This dissertation is dedicated to a very special friend and most especially to my family, without whose love, support, and pa tience I would not be where I am today. To Josefina Ramoni: Even though we are separated by many miles you have been a constant source of inspiration to me Your friendship is irreplaceable. To my parents, William and Rosalie Johnson: Mom and Dad you taught me the value of hard work, perseverance, and dedication. Because of you I have been able to forge through whatever obstacles have come my way. Thank you for loving and supporting me along my life’s journey. To my children Heather and Dani el: You are my prid e and joy. My hope is that traveling through this odyssey with me you have learned that through hard work and determination you can achieve whatever goals you set for your self. Thank you for se ttling for less than the perfect mom while I achieved my goal. And finally to my husband Bobby: Your devot ion to my accomplishment is nothing less than amazing. Thank you for keeping me on the right path and for having unfailing faith in me. You are the rock on which I find my foothold every day and without you I would not have finished this dissertation. Thank you for loving me unconditionally and supporting my dream.

PAGE 3

Acknowledgments I am indebted to many individuals who s upported my work on this dissertation. First and foremost I would lik e to thank my dissertation committee chair, Gabriel A. Picone, whose guidance and confidence was inva luable. I also want to thank my other committee members, Bradley P. Kamp, Philip K. Porter, and R. Mark Wilson for their helpful comments and advice. I know it w ould not have been possible for me to complete this dissertation without their support. I w ould also like to than k the rest of the Economics department faculty who provided me with an incredible experience through their academic instruction, advice, and support. I gratefully acknowledg e financial support from the National Science Foundation’s Doctoral Dissertation Research Grant (SES-0519001) as well as the Gaiennie Foundation at the University of South Florida, College of Business Administration. These two grants provided me with the necessary funding to purchase the data utilized in this dissertation. To both organizations I am deeply grateful. Additionally, I would lik e to thank David C. Grabowski, James H. Swan, and Elizabeth Cornelius for sharing their data with me. Finally, I would like to thank all of th e individuals from each of the state Medicaid and regulation agencies that so gr aciously provided me w ith the information on Medicaid reimbursement methods and rates and the status of certificate of need and construction moratorium policies. Without th eir kindness I would not have been able to create such an extensive data set.

PAGE 4

i Table of Contents List of Tables................................................................................................................. ....iv Abstract....................................................................................................................... .....viii Chapter One Introduction....................................................................................................1 1.1 What is Certificate of Need Regulation?...........................................................1 1.2 The Nursing Home Industry and CON..............................................................3 1.3 The Period After Federal CON Elimination......................................................6 Chapter Two The Background of the Nursing Home Industry............................................8 2.1 Legislative History.............................................................................................8 2.2 The Market for Nursing Home Care................................................................14 2.3 Payment for Nursing Home Care.....................................................................16 2.3.1 Reimbursement Methods..................................................................17 Chapter Three Literature Review.......................................................................................22 3.1 Nursing Homes and Quality ............................................................................22 3.1.1 Measuring Nursing Home Quality....................................................25 3.1.2 Nursing Home Quality and CON......................................................26 3.1.2.1 Nursing Home Quality and Medicaid Reimbursement Rates....................................................................27 3.1.3 Other Predictors of Nursing Home Quality......................................33 3.1.3.1 Nursing Home Ownership.................................................33 3.1.3.2 Nursing Home Size............................................................34 3.1.3.3 Nursing Home Staffing......................................................36 3.1.3.4 Nursing Home Resident Mix.............................................37 3.1.3.5 Nursing Home Case-Mix...................................................39 3.1.3.6 Nursing Home Reimbursement Methods...........................41 3.2 Nursing Homes and Access to Care.................................................................45 3.2.1 Nursing Home Access and CON......................................................45 3.2.2 Other Predictors of Nursing Home Access.......................................50 3.2.2.1 Nursing Home Reimbursement Rates and Methods..............................................................................50 3.2.2.2 Nursing Home Resident Mix.............................................53 3.2.2.3 Nursing Home Ownership.................................................54 3.3 Nursing Homes and Costs................................................................................55 3.3.1 Nursing Home Costs and CON.........................................................55

PAGE 5

ii Chapter Four Research Design..........................................................................................59 4.1 Objectives and Hypotheses..............................................................................59 4.2 Description of Data..........................................................................................61 4.2.1 The OSCAR Data System.................................................................62 4.2.2 State-level Medicaid Reimbur sement Data and Certificate of Need and C onstruction Moratorium Policies...............................64 4.2.3 The Area Resource File....................................................................66 4.2.4 The Regional Economic Information System...................................67 4.2.5 Sample Construction.........................................................................68 4.2.5.1 Elimination of Duplicate Records and Data Cleaning.............................................................................68 4.2.5.2 Data Errors.........................................................................70 4.2.5.3 Merging of the Various Data Files.....................................71 4.3 Description of Variables..................................................................................72 4.3.1 Nursing Home Quality......................................................................72 4.3.2 Nursing Home Access.......................................................................75 4.3.3 Certificate of Need and C onstruction Moratorium Policy................76 4.3.4 Other Independent Variables............................................................77 4.3.4.1 Facility-level Characteristics.............................................77 4.3.4.2 Market-level Characteristics..............................................79 4.3.4.3 State-level Characteristics..................................................80 4.4 Methodology....................................................................................................82 4.4.1 State Fixed Effects and Model Specification....................................82 4.4.2 Facility Fixed Effect s and Model Specification................................84 Chapter Five Research Results..........................................................................................88 5.1 Nursing Home, Count y, and State-level Char acteristics (1991-2003)............88 5.2 The Effect of Certificate of Need and Construction Moratorium Policies on the Quality of Care in Nursing Homes..........................................90 5.2.1 Process Measures of Quality.............................................................91 5.2.1.1 Other Findings...................................................................91 5.2.2 Outcome and Composite Measures of Quality.................................95 5.2.2.1 Other Findings...................................................................95 5.2.3 Structure Measures of Quality..........................................................98 5.2.3.1 Other Findings.................................................................101 5.3 The Effect of Certificate of Need and Construction Moratorium Policies on the Access to Care in Nursing Homes.........................................102 5.3.1 Other Findings................................................................................104 5.4 Robustness Checks.........................................................................................107 Chapter Six Conclusions..................................................................................................110 6.1 Main Findings................................................................................................110 6.2 Policy Implications........................................................................................112 6.3 Limitations.....................................................................................................114 6.4 Future Research.............................................................................................115

PAGE 6

iii References..................................................................................................................... ...117 Bibliography................................................................................................................... .131 Appendix A: Tables.........................................................................................................132 About the Author...................................................................................................End Page

PAGE 7

iv List of Tables Table 1 Nursing Home, County, and State-level Charac teristics (1991-2003)..........89 Table 2 Main Regression Re sults for Process-Based Quality Models.......................92 Table 3 Main Regression Results for Outcome and Process-Based Quality Models............................................................................................................96 Table 4 Main Regression Resu lts for Structure-Based Quality Models.....................99 Table 5 Main Regression Resu lts for Structure-Based Quality Models...................100 Table 6 Main Regression Re sults for Access and Heavy-Care Models...................103 Table A.1 CMS’ Deficiency Classification System.....................................................133 Table A.2 Illustrative Measures of Quality of Care in Nursing Homes.......................134 Table A.3 Certificate of Need and Moratorium Policies for the Years 19912003..............................................................................................................135 Table A.4 Medicaid Reimbursement Method for the Years 1991-2003......................137 Table A.5 Case-mix Reimbursement for the Years 1991-2003....................................139 Table A.6 Average Per Diem Medicaid Reimbursement Rates for the Years 1991-2003....................................................................................................140 Table A.7 Nursing Home, County, and State-level Characte ristics (1991-2003) by CON_MORT...........................................................................................142 Table A.8 Regression Results for Process-Based Quality Models...............................144 Table A.9 Regression Results for Process-Based Quality Models...............................145 Table A.10 Regression Results for Ou tcome and Composite-Based Quality Models..........................................................................................................146

PAGE 8

v Table A.11 Regression Results for Ou tcome and Composite-Based Quality Models..........................................................................................................147 Table A.12 Regression Results for Structure-Based Quality Models.............................148 Table A.13 Regression Results for Structure-Based Quality Models.............................149 Table A.14 Regression Results for Structure-Based Quality Models.............................150 Table A.15 Regression Results for Structure-Based Quality Models.............................151 Table A.16 Regression Re sults for Access Model..........................................................152 Table A.17 Regression Results for Heavy-Care Access Model.....................................153 Table A.18 Regression Results for Pr ocess-Based Quality Models without Method.........................................................................................................154 Table A.19 Regression Results for Pr ocess-Based Quality Models without Method.........................................................................................................155 Table A. 20 Regression Results for Outcome and Composite-Based Quality Models without Method..............................................................................156 Table A.21 Regression Results for Ou tcome and Composite-Based Quality Models without Method...............................................................................157 Table A.22 Regression Results for Stru cture-Based Quality Models without Method.........................................................................................................158 Table A.23 Regression Results for Stru cture-Based Quality Models without Method........................................................................................................159 Table A.24 Regression Results for Stru cture-Based Quality Models without Method........................................................................................................160 Table A.25 Regression Results for Stru cture-Based Quality Models without Method........................................................................................................161 Table A.26 Regression Results fo r Access Model without Method...............................162 Table A.27 Regression Results for Hea vy-Care Access Model without Method...........163 Table A.28 Regression Results for Pr ocess-Based Quality Models Most Restrictive Sample.......................................................................................164

PAGE 9

vi Table A. 29 Regression Results for Process-Based Quality Models Most Restrictive Sample......................................................................................165 Table A.30 Regression Results for Ou tcome and Composite-Based Quality Models Most Restrictive Sample.................................................................166 Table A.31 Regression Results for Ou tcome and Composite-Based Quality Models Most Restrictive Sample.................................................................167 Table A.32 Regression Results for St ructure-Based Quality Models Most Restrictive Sample.......................................................................................168 Table A.33 Regression Results for St ructure-Based Quality Models Most Restrictive Sample.......................................................................................169 Table A.34 Regression Results for St ructure-Based Quality Models Most Restrictive Sample.......................................................................................170 Table A.35 Regression Results for St ructure-Based Quality Models Most Restrictive Sample.......................................................................................171 Table A.36 Regression Results for A ccess Model Most Restrictive Sample.................172 Table A.37 Regression Results for Hea vy-Care Access Model Most Restrictive Sample..........................................................................................................173 Table A.38 Regression Results for Pr ocess-Based Quality Models Least Restrictive Sample.......................................................................................174 Table A. 39 Regression Results for Pr ocess-Based Quality Models Least Restrictive Sample.......................................................................................175 Table A.40 Regression Results for Ou tcome and Composite-Based Quality Models Least Restrictive Sample.................................................................176 Table A.41 Regression Results for Ou tcome and Composite-Based Quality Models Least Restrictive Sample.................................................................177 Table A.42 Regression Results for St ructure-Based Quality Models Least Restrictive Sample.......................................................................................178 Table A.43 Regression Results for St ructure-Based Quality Models Least Restrictive Sample.......................................................................................179

PAGE 10

vii Table A. 44 Regression Results for St ructure-Based Quality Models Least Restrictive Sample......................................................................................180 Table A.45 Regression Results for St ructure-Based Quality Models Least Sample..........................................................................................................181 Table A.46 Regression Results for A ccess Model Least Restrictive Sample.................182 Table A.47 Regression Results for Hea vy-Care Access Model Least Restrictive Sample..........................................................................................................183

PAGE 11

viii Certificate of Need Regulation in the Nursing Home Industry: Has it Outlived its Usefulness? Barbara J. Caldwell ABSTRACT The primary goals of the National Health Planning and Resources Development Act (P.L. 93-641) of 1974 were to (1) contai n health care costs and (2) increase the accessibility and quality of health services. Certificate of need (CON) regulation is one attempt to constrain health care costs by limiting the supply of certain medical care facilities. With respect to the nursing home industr y, prospective nursing home owners/operators are required to demonstrate that a “need” exists for more nursing home beds. Some States also imposed a construc tion moratorium that prevented any expansion of existing facilities or construction of new facilities regardless of whether or not a “need” existed. These CON/moratorium progr ams impose a supply side constraint that creates a potential barrier to entry and in the presence of excess demand may cause a nursing home bed shortage for thos e residents covered by Medicaid. Even though the Federal CON requirement lapsed in 1986, forty-two States and the District of Columbia con tinue to have a CON, a constr uction moratorium, or both for nursing home facilities. Yet maintaining these regulations comes with a cost. This paper investigates if differences exis t in the quality of care and the access to care between nursing homes in those Stat es without CON and/or construction

PAGE 12

ix moratorium and those States th at still have such policies. Using data for the years 1991 through 2003 for all freestanding Medicaid-/Med icare-certified nursing home facilities in the United States and employing state and faci lity fixed effects models we find that Medicaid-eligible residents in those states without CON a nd/or construction moratorium policies have more access to a nursing home be d than those individuals in states with these policies. With respect to quality of care the results are mixed depending on the measure of quality that is employed. With th e risk of becoming a nursing home resident at the age of 65 at 44 percent and at the ag e of 85 at 53 percent (Spillman and Lubitz 2002) coupled with the aging of the current population, the areas of quality of care and access to care remain important policy issues in the nursing home industry.

PAGE 13

1 Chapter One Introduction 1.1 What is Certificate of Need Regulation? In 1974, the United States Congress passe d The National Health Planning and Resources Development Act (P.L. 93-641). The pr imary goals of this legislation were to (1) contain health care costs and (2) incr ease the accessibility and quality of health services (Harrington, et al. 1997a; Spillman a nd Lubitz 2002). This federal regulation was passed at a time when health care cost s were escalating at an alarming rate and accessibility to services for those who “needed it the most” as well as the quality of health services were consider ed poor. Most importantly, th e escalating health care costs, and the federal government’s growing proportion of these costs, were the driving force behind the certificate of need (CON) regul ation requirement imposed by this Act. CON was one approach to containing hea lth care costs by limiting the supply of certain medical care facilities. In general, a CON program involves the regulation of the building, expansion, and modernization of hea lth care facilities and capital equipment on the part of institutional health care providers, including hospitals, nursing homes, and home health agencies. Oversight of these pr ograms is provided at the state level through designated agencies who at their discretion set the specific criteria within the program for their State (Blumstein and Sloan 1978). Du ring the mid-1960’s and early 70’s, some

PAGE 14

2 states began establishing CON programs. The National Health Planning and Resources Development Act required that all states a dopt a certificate of need program and by 1980 all states had some form of CON regulation. The impetus for this Act was the escalati on of public medical care expenditures following the passage of the 1965 amendments to the Social Security Act. These amendments established the Medicare and Me dicaid programs that are responsible for certain medical care requirements of the elde rly and the needy. The primary explanation for this rapid increase in costs is the pres ence of third-party payment for medical care expenses through Medicare and Medicaid, whic h enabled what has become known as the Medical Arms Race (MAR). Even though Medicare and Medicaid require deductibles or co-payments, a resident does not pay the full marginal cost of the last unit of care received. CON regulation imposed a supply side constraint that limited capac ity, with the intention of reducing the growth of fee-for-service reimbursements and combating the classic problem of moral hazard. Implicit in this regulation, (as noted by Norton (2000)), is the assumption that an unregulated market result s in an excess of capital expenditure and capacity. The MAR hypothesis is considered a speci al case of quality competition in which it is asserted that quality is overproduced by competitors in medical care markets. This hypothesis supports the idea of increased expenditures due to excessive capital investment. It begins by noti ng that physicians, acting as agents for residents in search of hospital care, influence their residents’ choice of hospital. Hospita ls, wanting to increase their admissions, attempt to a ttract physicians by offering the latest technology. Under

PAGE 15

3 this scenario, duplicate services, i.e., services that are in excess of what the market would demand, become the norm (Dranove, et al. 1992). During the tim e of the MAR, a retrospective payment system was the pr evailing method of cost reimbursement by Medicare and Medicaid. This type of syst em pays providers on the basis of incurred costs. Hence, the costs of these additional services were passed through to insurers who then cost-reimbursed the hospitals (Dranove and Satterthwaite 2000). With widespread coverage of medical care services for new, t echnically sophisticated services as well as the increase in the proporti on of elderly and indigent individuals qualified for care through Medicare and Medicaid, CON regulation was considered to be the countervailing force needed to offset this rapidly expa nding segment of governmental expenditures. 1.2 The Nursing Home Industry and CON In the decade following the establishmen t of Medicare and Medicaid, public medical care expenditures on nursi ng homes alone, just one part of what is termed longterm care, exceeded that for any other medical care service, accoun ting for 25 percent of total health expenditures in 1975 as compared to 15 percent in 1966 (Feder and Scanlon 1980). Additionally, between 1965 and 1973, the number of nursing home beds increased by 650,000, an increase of 139 per cent in less than a decade (Hawes and Phillips 1986). By 1977, the U.S. government, through Medicare, Medicaid, and other smaller government-funded programs, had purch ased approximately 62 percent of all nonprofit nursing home services and 60 percent of all proprie tary nursing home services (Gertler and Andreano 1982).

PAGE 16

4 With the passage of P.L. 93-641, in order to be eligible for federal funds available through the U.S. Public Health Service prospective nursing home owners/operators are required to demonstrate that a “need” exists for more nursing home beds. Additionally during the same time frame, some states imposed a construction moratorium that prevented any expansion of existing facilities or construction of new facilities regardless of whether or not a “need” existed. These CON/moratori um programs protect existing nursing homes from potential en try by new competitors. Most of the economic research studyi ng nursing homes employs the work of Scanlon (1980b) which views the industry as monopolistically competitive with nursing homes providing the same level of quality to both Medicaid and priv ate-pay residents. Scanlon hypothesizes that CON and construc tion moratorium policies act as a binding bed constraint, leaving certain individuals who demand nursing home services unable to gain access to care. Since nursing homes typi cally charge a higher rate for private-pay residents than the Medicaid reimbursement ra te, private-pay residents will be admitted first and then Medicaid residents will fill any remaining empty beds1. Since the privatepay demand is still met under a binding bed cons traint, the literature has referred to this unmet demand as “excess Medicaid demand” (Nyman 1985). During the period of federal CON, the issues of quality, access, and cost came under deeper scrutiny. Many proponents of the nursing home industry declared that quality was “low” because Medicaid reimburse ment rates were too low. Raising the 1 According to the MetLife Market Survey of Nursin g Home & Home Care Costs, the statewide average private-pay per diem rate in 2005 was $191in 2003 dollars (MetLife 2005). The statewide average Medicaid per diem rate for 20 03 in this study was $122. The averag e per diem rate for 2003 for each state in this sample was less than the statewide average priv ate pay per diem rate (Alaska’s private pay per diem was $500 in 2003 dollars and its Medicaid per diem rate in this sample was $321).

PAGE 17

5 reimbursement rate seems to be the obvious so lution. However, early studies find that an increase in the reimbursement rate, under c onditions of excess dema nd, actually leads to the counterintuitive result of lower quality. When the number of prospective residents is greater than the supply of nursing home beds the nursing home provides care to privatepay residents first since the private-pay pric e is typically greater than the Medicaid payment rate. Therefore, raising quality only benefits the nursing home by attracting additional private-pay residents since at any quality level a sufficient number of Medicaid residents are available to fill an empty nursing home bed. However, in order to attract an additional private-pay resident a Medicaid resident must be displaced. The foregone Medicaid payment associated with that displ aced resident becomes a cost to the nursing home attributed to increasing quality. Furthe rmore, an increase in the reimbursement rate increases this opportunity cost of providing hi gher quality. Therefor e an increase in the reimbursement rate may actually lower quality (Nyman 1985). With respect to the goal of supply-si de constraint, between 1976 and 1980 the supply of nursing home beds increased at an an nual rate of approximately 3 percent, but between 1981 and 1983 the growth rate was onl y about 1.75 percent (Hawes and Phillips 1986). However, one result of this supply cons traint is that under conditions of excess demand access to nursing home care is constr ained for Medicaid residents but not for private-pay residents (Ettner 1993; Hola han and Cohen 1987; Nyman 1985; Scanlon 1980b). Again, this is due to the fact that with a limited supply of beds nursing homes prefer to admit the higher paying private resident than the Medicaid resident. With respect to cost containment, between 1960 and 1983 nursing home expenditures increased from $480 million to $28.8 billion, with Medicaid’s share

PAGE 18

6 increasing from 28 to 55 percent over the same time period (Holahan and Cohen 1987). One explanation given for this increase was th at the certificate of n eed cost containment initiative led to higher private-pay prices. These higher prices then resulted in privatepay residents becoming Medicaid-eligible soon er than they would otherwise. As a consequence, any cost saving from CON will in actuality be net of an increase in the proportion of nursing home days-of-care paid for by Medicaid (Nyman 1994). 1.3 The Period After Federal CON Elimination In 1986, during the deregulat ion era of the Reagan administration, Congress allowed the federal CON requirement to lapse due to its perceived anticompetitive and excessively regulatory nature. However, even after its removal fort y-five States and the District of Columbia continued to have a CON, a construction moratorium, or both a CON and construction moratorium for nursing ho me facilities (Harrington, et al. 1997a). Yet maintaining these regulations comes at a cost. Since quality remains a controversial subject, access to care for Medicaid-eligible individuals remains an issue in many States, and medical care expenses for nursing homes continue to climb, one might question the effectiveness of retaining these regulations. Today eight states no longer have certific ate of need or construction moratorium regulations in effect fo r nursing home facilities.2 The existing literature concerning the nursing home industry does not tell us if any di fferences in quality of care or access to care exist between those states without CON policies and those States that have the 2 The eight states are Arizona, California, Colorado, Idaho, Indiana, Kansas, New Mexico, and Pennsylvania. Nevada has no regu lation in its two largest counties.

PAGE 19

7 regulation. With the risk of becoming a nursi ng home resident at the age of 65 at 44 percent and at the age of 85 at 53 percent (Spillman and Lubitz 2002), coupled with the aging of the current population, the areas of quality of care and access to care remain important policy issues in the nursing home industry. The results from this study show that Medi caid-eligible individua ls in those states without have some type of CON regulation ha ve more access to a nursing home bed than in those states with this type of supply cons traint. The results of the models measuring the quality of care are not as consistent; each result is dependent on the choice of quality measure which supports the idea that there is no real standard by which to measure quality of care. In conjunction with the wo rk of Grabowski et al.(2003) that shows the repeal of CON regulations ha s not significantly increased th e Medicaid cost of nursing home care, the results of this research furthe r suggest that perhaps the use of supply-side constraints are not as effective in increasi ng the accessibility of nursing home care for Medicaid-eligible individuals and that there are still some aspects of the quality of nursing home care that remain questionable. The remainder of this dissertation is or ganized as follows: Chapter 2 provides a detailed look at the background and legisl ative history of the nursing home industry; Chapter 3 reviews the relevant literature with regard to th e issues of quality of care and access to care in nursing homes; Chapter 4 states the objectives of this study and describes the data and methodology, including th e specification of the models utilized in this research; Chapter 5 presents and describe s the results of the analysis; and Chapter 6 summarizes the main findings of this study, potential policy implications, limitations of the study, and areas fo r further research.

PAGE 20

8 Chapter Two The Background of the Nursing Home Industry Today’s nursing home industry originates from the development during the late 19th and early 20th centuries of five types of fac ilities classified as county poorhouses, state mental hospitals, voluntary homes for the aged typically run by religious organizations, proprietary boarding houses and hospital-affiliated nursing homes (Waldman 1983). The county poorhouse, or the county “almshouse,” was operated and financed by local governments for those children and adults who were poor, old, disabled, and mentally challenge d. Later in the nineteenth ce ntury reformers were able to segregate the occupants of these almshouses into specialized inst itutions such as orphanages and mental hospitals. The rema ining occupants of these poorhouses were primarily the poor and the aged. 2.1 Legislative History The initial, most significant influen ce on the nursing home industry was the passage of the Social Security Act in 1935, which provided income maintenance to the disabled and aged. The Old Age Assistance (OAA) program was a means-tested old age pension fund financed by states on a matc hing basis with the federal government for persons age 65 and over. Through this inco me assistance, these residents of the

PAGE 21

9 poorhouses were able to afford alterna tive living arrangements (Waldman 1983). However, this legislation specifically disa llowed funding to public institutions, thus encouraging the development of the propr ietary nursing home sector (Vladeck 1980). The 1950 amendments to the Social Securi ty Act eliminated the restriction on OAA payments to residents of public medical fa cilities and allowed payments to be made directly to the vendors of me dical care. These direct me dical vendor payments were limited under a sharing formula used to determ ine the federal portion of the cost of the assistance program. These payments were av ailable only to persons whose income was at or below the eligibility level for the a ssistance program. Many states used a “spenddown” method to determine who was eligible for the paymen ts. Persons whose income was above the specified level were ineligib le for money payments but could become eligible for medical vendor payments if the am ount they spent for me dical care from their own funds brought their net income down to the eligibility levels The methods of reimbursing nursing homes under this program were left to the states Additionally, the amendments also mandated that states esta blish licensing requirem ents for nursing homes (Giacalone 2001; Waldman 1983). Enacted in 1960, the Medical Assistance for the Aged (MAA) program began as a possible substitute for the Medicare program that was under consideration at the time. The MAA was an expansion of the direct vendor payment program that included two important differences. The first allowed stat es to provide assist ance to those whose income was above the OAA standard but belo w what was considered insufficient to pay for medical care. The second difference was that the federal matching contributions had no ceiling. Between 1960 and 1965, vendor payments associated with nursing home care

PAGE 22

10 had risen tenfold and the number of persons receiving nursing ho me care under MAA had reached 300,000 (Vladeck 1980; Waldman 1983). Other important legislation of the 1940s and 1950s impacted the supply side of the nursing home industry. Se veral pieces of important le gislation were enacted to encourage the construction and moderniza tion of nursing homes in the proprietary, nonprofit and public sectors. First, the 1954 amendment to the Hill-Burton Act provided financial support for the construction a nd renovation of government and nonprofit nursing homes based on the ability to show n eed. Additionally, this amendment required that certain construction and operational standards be met in order to provide “quality” care (Vladeck 1980). In 1956, under the auspices of the Sm all Business Administration (SBA), proprietary nursing homes became eligible for government loans (Giacalone 2001). Probably the most important pi ece of legislation for the proprietary sector was the passage of Section 232 of the Housing Ac t of 1959, which provides a program of mortgage insurance for proprietary nursing homes. The Federal Housing Administration (FHA) insures lenders against losses for loan s for construction or renovation of nursing homes. However, a condition of the loan insu rance is that the propos ed project receives a certificate of need from the appropriate state agency indicating that a need exists for the proposed beds (Waldman 1983). Perhaps the most influential impact on th e nursing home industry was the passage of the 1965 amendments to the Social S ecurity Act establishing the Medicare and Medicaid programs. Medicare is a federal health insurance progr am for people 65 years of age and older, some people with disabi lities under age 65, and people with End-Stage

PAGE 23

11 Renal Disease. Specific to nursing home care, Part A of Medicare provides skilled nursing care on a limited basis to those individua ls recently discharged from a hospital. The intent is to provide continued care to those who still require assistance during their recovery but not at the same level provided in a hospital. In order to participate as a Medicare provider, a nursing home must be certi fied as a Skilled Nursing Facility (SNF). SNFs typically serve post-operative residents requiring a considerable amount of therapy and nursing assistance to facilita te the recovery of an acute illness. Certified nursing homes are paid the reimbursement ra te set by the Federal government. Replacing the medical vendor payment and MAA programs, Medicaid is a federal/state vendor payment program that pays for medical assistance for certain individuals and families with low incomes and limited resources. While restricted by guidelines established by federal statutes, each state establishes its own eligibility standards, determines the type, amount, duratio n and scope of services, sets the payment rate for services, and administers its own pr ogram. Additionally, states have the option of providing care to those cons idered “medically needy.” Th ese individuals are eligible for Medicaid except that their income and/or resources are above the eligibility level set by their state. They may qualify immediat ely or may “spend down” by incurring medical expenses that reduce their income to or below their state’s medically needy income level. Under Medicaid nursing home care must be provided for individuals aged 21 or older who qualify as “categorically needy.” For those who qualify as “medically needy,” residents must pay all of their income except for a small spending allowance. A Medicaid-certified nursing facility may be classified as either a SNF, as an Intermediate Care Facility (ICF), or as dually certified. An ICF typically provide s less rehabilitative

PAGE 24

12 care and assistance while providing basic nursi ng care, therapy, and social activities to those considered chronically ill. In 1972 Public Law 92-603 established the Supplementary Security Income (SSI) program, replacing the OAA program. Similar to OAA, SSI is an income maintenance program which provides mont hly payments to aged and disabled persons but, unlike OAA, it is a federally financed and administ ered program (Waldman 1983). Eligibility for SSI automatically assures Medicaid elig ibility. Additionally, Section 249 of the law required states to reimburse nursing home care under Medicaid on a “reasonable costrelated basis,” as was required by Medi care (Vladeck 1980; Waldman 1983). This reimbursement requirement was later repealed an d states were then given the authority to set their own reimbursement method with the st ipulation that the rate s be “reasonable and adequate to cover the costs of an efficiently operated facility” (Waldman 1983). Although not originally intended to be the mechanism for paying for long-term care, Medicaid has become the major public method for funding nursing home care. Between 1965 and 1973, the number of nursi ng home beds increased by 650,000, an increase of 139 percent (Hawes and Phillip s 1986). Expenditures grew even more rapidly. Just prior to the implementati on of Medicare and Me dicaid, expenditures on nursing home care totaled $1.328 billion. By 1978, the nation spent $15.8 billion on nursing homes, with the government paying 53 pe rcent of the total (Hawes and Phillips 1986). In an attempt to control the escalati ng health care costs, and the federal government’s growing proportion of these cost s, the United States Congress passed the National Health Planning and Resources De velopment Act (P.L. 93-641) of 1974. As a

PAGE 25

13 part of this Act, states required potentia l nursing home operators as well as existing nursing homes that were interested in expans ion to acquire a certif icate of need (CON) prior to new construction. This legislati on has a supply side effect on the industry by limiting the number of nursing home beds avai lable to prospective residents. Even though this federal legislation was repealed in 1986, today 42 States and the District of Columbia still maintain CON and/or construction moratorium policies. The Nursing Home Reform Act of the Omnibus Budget Reconciliation Act of 1987 (OBRA 87) mandated extensiv e legislative requirements a ffecting the delivery of care to residents in nursing homes (Marek, et al. 1996). This Act repealed the SNF and ICF benefits (except for the ICF benefits for persons with mental retardation or related conditions) under Medicaid and replaced them with a mandatory nursing facility (NF) benefit that combines the total services previously covered under the ICF and SNF benefits. This change became effective in 1993. This legislation also imposed the standards of care that facili ties must meet in order to receive both federal and state funding. It also defined the state survey and certification process that determines compliance with the federal standards. Additionally, sanctions were established for facilities that fail to meet the established sta ndards (Harrington and Carrillo 1999). Many of the requirements under OBRA 87, most signi ficantly the survey inspection procedures and the required nurse’s aide training, in creased the cost structures of nursing home facilities (Giacalone 2001).

PAGE 26

14 2.2 The Market for Nursing Home Care The nursing home industry has many propert ies of a perfectly competitive market. Information failures common to other sector s of the medical care industry (Arrow 1963) seem rare in nursing home care. Consumer s have the ability to observe many of the features of a nursing home’s services, such as room conditions, staff and resident interaction, organized activitie s and other available services Start-up costs are not a significant barrier to entry, as plant and equipment costs ar e significantly below those of a hospital (Bishop 1988). In addition, substitu tes such as family or in-home care are readily available for many individuals. However, other features of the nursing home industry deviate from the perfectly competitive standards. The decision to en ter a nursing home often occurs at a time of crisis, thus dramatically reduc ing the ability to gather and process existing information. Location quite often is an important criteri on when selecting a nursing home. Residents want to remain close to family and friends making it unlikely that they will “vote with their feet” and move to a new nursing home. Additionally there are also switching costs, termed the “transfer trauma,” resulting in much less movement between homes than expected (Bishop 1988; Nyman 198 5; Weisbrod and Schlesinger 1986). It follows then that most researchers have modeled th e nursing home industry as monopolistically competitive. Homes differentiate themselves with respect to location and types of services offered. Another important feature of the nursing home industry is the major presence of the government, both as the largest purchaser of services (through Medicaid) and as the industry’s regulator. As regul ator, the government imposes the reimbursement rate on

PAGE 27

15 those homes that meet the certification require ments. This rate becomes a price ceiling for the home for its public-pay residents. When this rate is below the market-clearing price, a shortage of nursing home beds will exist. Moral hazard resulting from the Medicaid subsidy increases demand for nursi ng home care, while CON and moratorium policies restrict supply. It is very common for the nursing home market not to clear, with a shortage of beds the norm. Nursing homes typically have a mixture of private-pay and public-pay (Medicaid) residents. With two types of customers, pr ice discrimination arises. The firm (a nursing home) faces a downward-sloping demand curv e for its private-pay residents and a perfectly elastic demand at the governmen t reimbursement rate for its public-pay residents. The nursing home then chooses it s total output, how to allocate this output between private-pay and public-pay residents, and what price to charge its private-pay residents (Palmer and Vogel 1983). For-profit homes are assumed to maximi ze profits by equating marginal revenue and marginal cost while equalizing the leve l of marginal revenue of private-pay and public-pay residents. Nonprofit homes are assumed to maximize output subject to a break-even constraint and a qua lity constraint (Palmer a nd Vogel 1983; Scanlon 1980b). This is accomplished by equa ting average revenue and average cost and allocating the total output between private-pa y and public-pay residents such that the marginal revenue from each is equal. For both types of homes, the result is typically the private-pay price being higher than the Medicaid reimbursement rate. Scanlon (1980b) hypothesizes that CON a nd construction moratorium policies act as a binding bed constraint where there exist certain individuals who demand nursing

PAGE 28

16 home services yet are unable to gain acce ss to care. Since nursing homes typically charge a higher rate for pr ivate-pay residents than the Medicaid reimbursement rate, private-pay residents will be admitted first a nd then Medicaid-eligible individuals will fill any remaining empty beds. Since the privat e-pay demand is still met under a binding bed constraint, the literature has referred to th is unmet Medicaid demand as “excess Medicaid demand” (Nyman 1985). The nursing home industry is comprised of for-profit, nonprofit and government facilities. As of June 2005, there were 16,023 total facilities, of which 65.86 percent were for-profit, 28.04 percent were nonprofit, and 6.1 percen t were government owned; 5.31 percent were Medicare-certified only, 6.72 percent were Medica id-certified only, and 87.97 percent were both Medicareand Medi caid-certified. These facilities provided 1,748,001 beds of which 71,960 were Medicare -certified only, 225,710 were Medicaidcertified only, 1,382,395 were dually certified, and 67,936 were non-certified. The mean occupancy rate of these facilities was 85.58 percent while the median rate was 88.61 percent (AHCA 2005). 2.3 Payment for Nursing Home Care National health expenditures for nur sing home care in 2003 were $110.8 billion. Private payment accounted for 39 percent of these expenditures, Medicare accounted for 12 percent while both federal and state Medica id payments were 46 percent of the total (CMS 2003). Nursing homes typically do not ch arge the same rate for all payer types (private, Medicare, or Medicaid residents). Traditionally, private-pay rates are based upon charges determined by the nursing homes themselves while the federal government

PAGE 29

17 determines the Medicare rate and each state determines its own Medicaid rate; each rate is dependent on the type of reimbursement method utilized. Due to the differences between Medicare and Medicaid in the determination of rates, the difference between the care needs of acute and chronic care residents, as well as th e differences in the methods of defining reimbursable costs, most often th e private-pay rate is higher than both the Medicare and Medicaid rates and the Medicare rate is hi gher than the Medicaid rate. 2.3.1 Reimbursement Methods Reimbursement methods refer to the wa y in which Medicare and Medicaid programs pay for nursing home care. Medicare reimbursement for nursing home services was originally based upon the “reasonable co st” formula that originated for hospital payments. This methodology covers retrospe ctively the actual cost s of providing care and places no ceilings on reimbursement rates (Hawes and Ph illips 1986; Swan and Harrington 1985). Escalating health care co sts during the 1970s, fueled by the MAR and moral hazard caused the government to reassess its Medicare payment system. In 1983, a Prospective Payment System (PPS) was put in place for hospitals. This type of reimbursement method sets the rate to be paid for each type of service prospectively with the intent of providing an in centive to reduce costs. Th e Balanced Budget Act of 1997 implemented this same method for nursing homes such that today the Medicare reimbursement system is a prospective me thod of payment (Dranove and Satterthwaite 2000).

PAGE 30

18 When Medicaid was first established, states were given much discretion in setting nursing home rates while remaining within the federal guidelines. In 1972, the Social Security Amendments required states to implement “reasonable cost-related” reimbursement plans for nursing homes (S ocial Security Amendments of 1972). However, there was no specific requirement for this method to be a retrospective reimbursement method (Cotterill 1983). Pa rt of the reason these changes were implemented was because providers complained that states were too restrictive in their policies (Harrington and Swan 1984; Swan and Harrington 1985) and that nursing homes were not able to cover the cost of providing care to their public-pay residents. While the method of reimbursement was not clearly defi ned in order to meet the reasonable costrelated criteria, most states chose their ow n reimbursement policies while some states used the Medicare reimbursement formula to meet the Medicaid requirement. With the ensuing escalation in health care co sts as well as pressure from states to allow more discretion in interpreting reasona ble costs, the Federal Boren Amendment of the Omnibus Reconciliation Act of 1980 gave st ates the authority to set rate methods and standards that are “reasonable and adequate to meet the co sts which must be incurred by efficiently and economically operated facilities in or der to provide care and services in conformity with applicable state and fe deral laws, regulations, and quality and safety standards” (Omnibus Budget Reconciliation Act of 1980). This reimbursement policy increased the authority as well as the flexibility of states by giving them greater discretion in setting rates. The intent was to provide the incentive to constrain costs while still rec ognizing the concept of cost-related rates (Cotterill 1983).

PAGE 31

19 States typically used this discretion by setting their Me dicaid rates below those of Medicare and private payers (Swan and Harrington 1985). Medicaid expenditures for nursing home serv ices are usually described on a cost per diem basis. The rates that are paid are determined in part by the reimbursement method chosen. States attempt to contro l these rates by developing reimbursement methods that restrain their overall ex penditures on nursing homes. Several reimbursement methods exist under Medica id. While state Medicaid programs are required to set reimbursement policies with some relationship to cost which requires a periodic adjustment for inflation, much discre tion is allowed in what other options they include in the per diem rate that results from the reimbursement method of choice. A retrospective system util izes a reimbursement formula that pays after the services have been provided, based on the actual costs incurred by the nursing home. This type of system tends to encourage provi ders to increase their expenditures in order to increase their revenues (Swan and Harringt on 1985). Additionally, states that have this type of reimbursement system f ace little incentive to minimize costs. Prospective reimbursement systems utilize a formula, usually based on past costs, that sets the payment rate for services prio r to the provision of care and before the costs are actually incurred. This type of methodology gives prov iders an incentive to keep their costs aligned with or lower than the reimbursement allowances. However, one criticism of this type of reimbursement met hod is that it may also provide incentives for services and quality to be lowered in order to stay within the allowable costs. Prospective methods have been shown to lower the per diem rate as compared to retrospective methods (Harrington and Sw an 1984; Swan, et al. 1993).

PAGE 32

20 Within the prospective methods of reimbur sement, a distinction may be made as to whether facilities are reimbursed based on an individual basis (fac ility-specific) or on a class or group basis (flat-rate). Stat es using a facility-specific method apply a reimbursement formula based on historical cost reports that results in a reimbursement rate for each facility in the state. States using flat-rate methods also use a formula usually based on past expenditures to set the reimbursement rate. States with this type of system then pay each facility a per diem rate that is based on the median per diem rate of the group of facilities in the class. Consequently, each indivi dual facility can not significantly adjust the rate it receives by adjusting its own costs (Cotterill 1983). An “adjusted” prospective method allows the rate to be adjusted upward during the rate period (Harrington, et al. 2000b). Some states have implemented what are called combination methods. This method of reimbursement is often based on co st centers, some of which are reimbursed prospectively and some of which are paid retrospectively (Ha rrington, et al. 2000b; Swan, et al. 1993). In addition, some states have moved to a “case-mix” reimbursement method. This type of system allows a st ate to adjust their rates for resident characteristics. The intent of this type of method is to recognize the level of care that is required for different types of residents a nd to compensate provi ders accordingly. Regardless of the method of reimbursement, the intent is to provide an incentive to nursing home owners or operators to lower their cost of providing care to public-pay residents. The Balanced Budget Act (BBA ) of 1997 repealed the Boren Amendment giving states even more flex ibility in setting nursing home payment rates and causing the

PAGE 33

21 nursing home industry to raise concerns that lower Medicaid reimbursement rates would adversely affect quality of care (Wiener and Stevenson 1998). In 2003, no states were using a retrosp ective reimbursement method, twenty-one states were using a facility-s pecific method, four states we re using a prospective class (flat-rate) method, twenty-four states were using a prospective-adjusted method, and one state was using a combination method. The dilemma for most stat es in deciding upon which type of reimbursement method to employ and the subsequent rate setting is how to ensure quality and access while cont rolling costs (Swan and Harrington 1985).

PAGE 34

22 Chapter Three Literature Review This chapter is intended to provide an overview of the literature relating to two distinct characteristics of th e nursing home industry investig ated in this dissertation; quality of care and access to care. Each se ction summarizes thos e studies that have analyzed the effects of CON and/or moratori um policies with separate subsections that describe some of the other variables cons idered important pred ictors of nursing home quality and access within the relevant literatu re. Additionally, a brief description of the impact of supply regulations on the cost of nursing home care is included to support the findings of this research that CON/moratori um policies may no longer be achieving their original intent. 3.1 Nursing Homes and Quality Quality of nursing home care has been a concern for the public as well as policymakers for the last thirty years and still remains a concern today. Providing the first real insight into the problems that we re occurring throughout the industry, Vladeck (1980) documents instances of abuse, neglect, and a general “lack of concern” in the care of nursing home residents. During the 1980s, reports still found the quality of nursing home care to be low (IOM 1986; GAO 1987). As of November 1985, 25 percent of

PAGE 35

23 SNFs and 16 percent of ICFs were noncompliant for two or more of the requirements to be Medicare and Medicaid certified (GAO 1987) As a result of the Institute of Medicine’s (IOM) report in 1986 (IOM 1986), the Nursing Home Reform Act of OBRA 87 altered the focus of the standards that nursi ng homes were required to meet in order to participate in Medicare and Medicaid. Not only were the standards for the delivery of care emphasized but also the results of that car e. Stricter requirem ents were established to maintain compliance along with an expansi on of enforcement sanctions that could be placed on those homes found to be noncompliant. Although improvement has been documented in some instances after the passage of OBRA 87 (IOM 1996), throughout the last de cade many reports continue to document what is considered unaccep table treatment of nursing hom e residents (CMS 2001; U.S. House of Representatives 2002; GAO 1998; GAO 1999; GAO 2002). Between July 1995 and February 1998, 407 of 1,370 nursing home s in California were cited for care violations considered as “s erious” under federal or stat e deficiency categories (GAO 1998).3 The number of abuse violations of all homes increased every year from 5.9 percent in 1996 to 16 per cent in 2000 (CMS 2001). Be tween March 2001 and August 2002, 39 percent of nursing homes in Texas were cited for “potential-to-harm” violations and 47 percent had “actual harm” violations or worse (U.S. House of Representatives 2002). A recently issued report states that for the 18-month period ending January 2002, 20 percent of nursing homes “were cited for deficiencies involving actual harm or 3 Prior to 2001, CMS was known as the Health Care Financing Administration (HCFA). The Centers for Medicare and Medicaid Services (CMS) has regulatory authority overseeing nursing homes’ compliance with Medicare and Medicaid participation requirements. See Table A.1 in Appendix A for the list of CMS’s Deficiency Cl assification System.

PAGE 36

24 immediate jeopardy to residents” (GAO 2003). Although this percentage is down from 29 percent in the previous survey period, viol ations continue to occur and residents are still being harmed. Nursing home quality cont inues to be an important issue of concern worthy of continued assessment and analysis. Often considered a value-ba sed construct, a universally accepted definition of health care quality does not exist. Policymakers, health care professionals, administrators, owners, investors, thirdparty insurers, and consumers base their definition of quality on their own subjective criteria resulting from their preferences for desired outcomes from such care (Davis 1991). It is often difficult to reach a consensus on what defines technical medical quality. It is even harder to define the less tangible characteristics of what is considered “caring and decent” treatment of residents in nursing homes (Nyman 1987). Consequently the lite rature on nursing home quality is widely disparate in its measurement of quality as well as in the results of measuring the effects of different policies and facility charac teristics on nursing home quality. The following subsections discuss the m easurement of nursing home quality, the effect of CON and/or construction moratorium on the quality of car e, the effect of a change in the reimbursement rate under ex cess demand on the quality of care, and the effect of other predic tor variables comm only used in the literatur e on the quality of care in nursing homes.

PAGE 37

25 3.1.1 Measuring Nursing Home Quality Perhaps the most widely accepted paradigm for measuring health care quality was developed by Donabedian and describes thre e distinct categories of quality assessment: structure, process, and outcome (Donabedian 1966, 1988). This framework was originally developed for the study of medical care delivery and has been widely used in the literature on nursing homes. Structural ev aluation looks at the at tributes of the setting in which the care is being provided. Examples of this type of measurement include the physical amenities of the fac ility, qualifications of the staff, and the administrative organization of the facility (Ullmann 1981). Process evaluation meas ures the types and quantities of services actually provided to residents against the professionally accepted standards of appropriate care fo r specific problems or conditions. Included in this type of measurement are the therapy se rvices offered, the use of phys ical restraints, and the use of urethral catheterization. Outcome evalua tion assesses the actual health and well being of the resident. Measurements of this type include mortality, change in functional status, and facility-acquired pressure sores.4 According to Donabedian, his three-part approach to quality measurement does not depend more heavily on one aspect than a nother but that its eff ectiveness is possible “only because good structure increases the likelihood of good process, and good process increases the likelihood of a good outcome” (Donabedian 1988). Most of the nursing home literature focuses on structural and proces s measures of quality. Ideally the use of outcome measures of quality is considered th e “gold standard.” However, data on these types of measurements have been difficult to obtain. Furthermore, care must be given in 4 Table A.2 in Appendix A details common measurements of structure, process, and outcome as referenced by the Institute of Medicine (1996).

PAGE 38

26 recognizing that measures such as mortality and a decline in health status are often natural occurrences with the elderly and frail residents in nursing homes. Additionally, often times the same variable is used as a predictor of quality in one study and as an indicator of quality in anothe r (Davis 1991). Staff-to-resident ratios and expenditures are two examples of variables that have served both as predictors of quality as well as indicators of quality. Higher staff-to-resident ratios are often considered an indicator of higher quality in some studies (Birnbaum, et al. 1981; Bishop 1980; Elwell 1984; Fottler, et al. 1981; Nyman 1988b) while used as a predictor of higher quality (e.g., lower mortality) in another (Linn, et al. 1977). Likewise, expenditures on nursing home care is considered an indicator of higher quality, i. e. higher expenditures implies higher quality, in some studies (Meiners 1982; Nyman 1988a; Ullmann 1984) while used as a predictor of higher quality (e.g., use of rehabilitation services) in others (Birnbaum, et al. 1981; Ullmann 1985). As a result, no consensus really exists on a comprehensive set of outcome measures of quality of care in nursing homes to be used as a substitute for structure or process measures (Shaughnessy, et al. 1990). While not necessarily true measures of health status, process measures are often considered measures of substandard care (IOM 1986; Spector and Takada 1991) and therefore, most of the early studies on quality in nursing homes utilize structural and process measures of quality while more recent work includes several outcome measures of quality. 3.1.2 Nursing Home Quality and CON Most of the early empirical research on nursing home quality utilizes Scanlon’s model of a monopolistically comp etitive market providing the same level of care to both

PAGE 39

27 private-pay and public-pay resi dents. When the number of prospective residents is greater than the supply of nursing home beds the nursing home provides care to privatepay residents first since the private-pay pric e is typically greater than the Medicaid payment rate. Therefore, under the conditi ons of a binding CON policy, a nursing home has no real incentive to compete for Medicaid residents on the basis of quality since at any quality level a sufficient number of Medicai d residents are available to fill an empty nursing home bed. Currently, the literature does not include any investigati on into the differences in quality, if any, between those states without CON and/or mo ratorium policies and those states that have such a policy in place for nur sing homes. In one study that analyzes the effects of competition on nursing home quality, the presence of a statewide nursing home construction moratorium results in a lower le vel of quality of care (Zinn 1994). Viewed as a barrier to entry, a moratorium redu ces competition and provides no incentive to provide higher quality. Us ing data from the 1987 Medica re and Medicaid Automated Certification Survey (MMACS) and the met hod of two-stage least squares (2SLS), Zinn finds that a moratorium on nursing home cons truction leads to lower RN staffing and a higher percentage of residents physically re strained thus implying lower quality while controlling for market concentration, area dem ographics, and Medicaid policies as well as resident and facility ch aracteristics (Zinn 1994). 3.1.2.1 Nursing Home Quality and Me dicaid Reimbursement Rates During the period of Federal CON many proponents of the nursing home industry declared that quality was “low” because Medi caid reimbursement rates were too low.

PAGE 40

28 One part of the literature on nursing home quality investigates the effects of Medicaid reimbursement rates and CON on nursing hom e quality. Although not employing a direct measure of the presence of a CON and/or mo ratorium policy, but rather a measure of excess demand (indicating the pres ence of a binding CON policy), an influential series of papers assess the impact of a change in the Medicaid reimbursement rate on nursing home quality. Because a binding CON policy provides no incentive to nursing homes to compete for Medicaid resident s on the basis of quality, the theory argues that under such a binding constraint a higher payment level actu ally leads to lower quality (Gertler 1989, 1992; Nyman 1985, 1988a, 1988b, 1989b). Under conditions of excess demand, the pr oportion of Medicaid and private-pay residents in a nursing home ar e found to be endogenous w ith quality. Recognizing the joint determination of quality and the number of private-pay residents and the bias that would result from using ordinary least s quares (OLS), two studies using 1978-79 data from the state of Wisconsin employ th e method of 2SLS (Nyman 1985, 1988b). Additionally, Nyman uses OLS for compara tive purposes since most of the early literature studying the nursing home industry us es OLS. The main equation of interest, the quality equation, estimates the weighted number of Medicaid cert ification violations in a home (a negative measure of quality) as a function of the number of private-pay residents in the home, the average number of unfilled beds in the county in which the home is located (represents th e likelihood that a home is not operating in a market with excess demand), a measure for the Medicaid reimbursement rate, as well as several demographic and facility characteristics as c ontrol variables. Nyman’s results show that, under excess demand, an increase in the Medicai d reimbursement rate decreases quality.

PAGE 41

29 Using 1983 data from Wisconsin, Nyman (1988a) uses OLS to estimate the relationship between excess demand and nursi ng home expenditures. In this study, nursing home expenditures serve as a proxy for quality. When excess demand exists, nursing homes are able to lower costs by lo wering quality without reprisal since prospective residents, especially Medicaid-el igible individuals, are usually forced to accept the first bed that becomes available regardless of the quality of the home. Separating the data into two groups, those homes in a county with a tight bed supply and those homes in a county with a surplus bed s upply, the results show that nursing homes in counties with a tight bed supply spend significan tly less on resident car e than those with a surplus bed supply. Utilizing the same data set and employing OLS as well as 2SLS, Nyman (1989b) finds that nursing homes in coun ties with a tighter bed supply have more Medicaid violations than those counties with a surplus bed supply. To account for the same endogeneity issu e but using 1980 New York state-level data, two studies estimate a reduced-form e quation of the effect of a change in reimbursement rate on quality (Gertler 1989, 1992). At the time of the studies, New York utilized a cost-plus method of reimbur sement. Additionally, even though no direct or proxy measure for CON was utilized, the da ta were from a time when New York was under CON regulation and was considered as facing excess demand conditions. Gertler (1989) uses three input measures of quality : hundreds of nursing la bor hours, hundreds of other labor hours, and a supplies quantity index. The two labor hours measures are adjusted for productivity differences across the nursing homes to eliminate the concern that a nursing home spending more labor hours on care may not necessarily mean the home is of higher quality; perh aps the nursing home is simply more inefficient. While

PAGE 42

30 controlling for resident, facility, economic, de mographic, and market characteristics, the results show that an increase in the Medi caid reimbursement rate improves access for Medicaid residents but lowers quality. Sim ilarly, Gertler (1992) finds the same result using total Medicaid expenditures as the measure of quality. In later work, Cohen and Spector (1996) use the 1987 Institutional Population Component (IPC) of the Nati onal Medical Expenditure Survey (NMES) to assess the effect of the Medicaid reimbursement rate on quality. The NMES IPC is a nationally representative sample of residents in or admitted to nursing and personal care homes as well as homes for the mentally retarded. In th is study, two different st rategies are used to assess the effects of the Medica id reimbursement rate on qualit y. First, the effect of the Medicaid reimbursement rate on staff intensity is estimated, and then the effect of staff intensity on resident outcomes is examined to s ee if more intense staffing results in better outcomes. Second, the authors estimate a “g lobal effect” of reimbursement on outcomes by measuring the direct effect of the Medicaid reimburseme nt rate on outcomes without using staff intensity as a right-hand side variable. Quality is first proxied by three structural measures adjusted for case-mix: RNs per 100 residents, licensed practical nurses (L PNs) per 100 residents, and total nursing staff per 100 residents. Then three outcom e measures are estimated: mortality, the presence of a pressure sore, and a change in functional status. The three staffing equations are estimated with 2SLS to acc ount for the endogeneity of the Medicaid reimbursement rate. The other explanatory variables include the number of empty beds per 1,000 aged 75 and over by county (used as a measure of market tightness to represent excess demand), the statewide average level of Medicaid reimbursement, a vector of

PAGE 43

31 facility characteristics, a vector of su pply and demand factors, and several policy variables. The mortality and pressure sore models are estimated using logistic regression while the functional status model is estimat ed using OLS with resident health and demographic characteristics, RN, LPN, and ai de staffing levels, st ate quality regulation policies, and ownership type serving as cont rol variables. The “global” effect is estimated for each of the outcome measures including as right-hand side variables the reimbursement variables and the exogenous variables from the staffing and outcome equations. The results indicate that an increase in the Medicaid reimbursement rate has a positive and significant impact on the number of LPNs per 100 residents but not RNs or total staffing. The results of the outcome e quations indicate that a higher RN staffing intensity leads to lower mortality, fewer pre ssure sores, and improvements in functional status, all indicating higher qual ity. However, in the “global” effects model, an increase in the reimbursement rate does not ha ve a significant impact on outcomes. More recent work employs a three-part methodology to empi rically test the effect of a change in the Medicaid reimbursement rate on nursing home quality (Grabowski 1999, 2001a, 2001b). Noting that the previous ly cited work by Scanlon, Nyman, and Gertler assumes that nursing homes are integrated facilities, i.e., they serve both privatepay and public-pay residents, the first st age of Grabowski’s methodology models the choice of payer mix regime (public-only, in tegrated, private-only); the second stage models the choice of payer mix conditional on the home choosing to be an integrated facility, and finally, the third st age models the qualit y decision. In these studies quality is proxied by an outcome measure: the proporti on of residents with facility-acquired pressure sores (Grabowski 1999, 2001b); structural measures: the number of RNs

PAGE 44

32 (Grabowski 2001b) and professional and non-pr ofessional staffing levels (Grabowski 2001a); process measures: medication error ra tes, use of physical restraints, use of catheters, and use of feedi ng tubes (Grabowski 2001a) and a composite of process and outcome measures: the number of nursi ng home deficiency c itations (Grabowski 1999, 2001a). The explanatory variables include meas ures of resident case-mix, various facility characteristics, state-level Medicaid reimbur sement policies, and market-based (county) characteristics including the number of empty beds per elderly population to represent the measure of excess demand. Using 1995-96 data on all U.S. Medicaidcertified nursing home s, Grabowski’s results show that an increase in the Medicaid reimbursement rate leads to a small, but significant increase in nursing home quality. Additionally, replicating Gertler’s reducedform model using all U.S. nursing homes in 1981 and all nursing homes in the state of New York for the 1995-96 time period, the OLS re sults also indicate that an increase in the Medicaid reimbursement rate increases qua lity. These results are contrary to the results obtained by Nyman and Gertler. Grabowski attributes this result in part to the fact that occupancy rates of nursing homes, an i ndirect measure of excess demand, have been declining over the time period between the earl ier studies and his studies. Using multiple waves of the National Nursing Home Survey (NNHS), the national occupancy rate was 92.9 percent in 1977, 91.8 percent in 1985 and 87.4 percent in 1995 (Strahan 1997). This shift in the tightness of the market, which ma y represent a change in the prevalence of excess demand conditions, may help to explai n the differences in the results between earlier studies (Gertler 1989, 1992; Nyman 1985, 1988a, 1988b, 1989b) and Grabowski’s more recent studies.

PAGE 45

33 3.1.3 Other Predictors of Nursing Home Quality The literature concerning the nursing ho me industry is quite extensive beyond the use of CON and the Medicaid reimbursement ra te as predictors of nursing home quality. Among those reviewed in this section are th e ownership, size, and staffing levels of a facility, the resident mix and case-mix of a facility’s residents, and the method of Medicaid reimbursement. While recognizing that each of the studies included in this portion of the literature review utilizes different data se ts and methods of analysis, this section provides a summary of the results ob tained when utilizing several of the other more important predictors of nursing home quality. 3.1.3.1 Nursing Home Ownership The ownership of a facility refers to whether the nursing home is established as for-profit, nonprofit, or government-owned. A subset of the for-pro fit classification is corporate nursing home chains. Economic th eory provides a framework for analyzing the effect of ownership on nursing home behavior For-profit nursing homes’ motivation is assumed to be profit maximization, subject to various regulatory constraints, while a common assumption is for nonprofit homes to ma ximize their size, subject to quality and break-even constraints (Scanlon 1980b). The ti e-in of ownership with quality is often analyzed through the costs that nursing homes experience. With hi gher costs indicating higher quality (and not less efficiency), cost minimization on the part of for-profit nursing homes may render profits and quality as c onflicting objectives. Much of the early empirical evidence using expend itures as a measure of quality supports a pattern of lower

PAGE 46

34 expenditures in for-profit nursing homes thus indicating lower quality (Birnbaum, et al. 1981; Frech III 1985; Frech III and Ginsbur g 1981; Meiners 1982; Ullmann 1984). Utilizing various structural, process, and outcome measures of quality have led to inconclusive results concerni ng the effect of ownership on nursing home quality. Using various structural indicators of quality, as well as various data sets and methods, some studies have found no significant differen ce between nonprofit and for-profit nursing homes (Cohen and Dubay 1990; Cohen and Spect or 1996; Harrington, et al. 1998; Lee, et al. 1983; Winn 1974), while others have found that nonprofits have higher quality (Aaronson, et al. 1994; Elwell 1984; Grabowski and Hirth 2 002; Zinn 1994). Likewise, the use of process measures of quality (Grabowski and Hirth 2002; Lee, et al. 1983; Nyman 1988c; Zinn 1994) as well as outcome m easures of quality (Bliesmer, et al. 1998; Chou 2001; Cohen and Spector 1996; Harrington, et al. 2002; Harri ngton, et al. 2000c; Mukamel 1997; Nyman 1985, 1988c, 1989b; Porell, et al. 1998; Spector, et al. 1998; Spector and Takada 1991) have led to the same inconclusive evidence. Finally, two studies utilizing composite measures of quality, one that includes structural and expenditure measures (Greene and Monahan 1981) and one that incl udes structural and outcome measures (Davis 1993), find that nonprof its have higher quality than for-profits. 3.1.3.2 Nursing Home Size The literature on the relationship between quality of care and the size of a nursing home facility also provides conflicting results. Quality of care and the size of a nursing home facility are often found to have a positive relationship. This relationship is usually justified by one of two different explanations. The first argues that larger homes tend to

PAGE 47

35 have more highly trained and professional ad ministrators, who in turn maintain higher resident standards. In this case, si ze is seen as a proxy for administrative professionalism. Secondly, it has been suggest ed that certain economies of scale exist in nursing home operations. With a greater numbe r of beds, nursing homes should be able to attract better personnel and provide a broader selection of services and hence provide higher quality. With either explanation, size is expected to be related positively to the quality of care provided (Greene and Monahan 1981). On the other hand, quality of care and the size of a nursing home facility are often found to have a negative relationship. Larger homes may provide less personal care experience more problems, and experience management inefficiencies (Harrington, et al. 2000c; Nyman 1985). Most analyses of the relationship between quality and the size of nursing home facilities use the number of beds in the facili ty as the predictor of quality. Inconsistent results occur regardless of whether a structural process, or outcome measure of quality is used in the analysis. Studies that have us ed a structural measur e of quality, such as nursing hours per resident day, find that either a significant relations hip does not exist (Cohen and Dubay 1990; Cohen and Spector 1996; Fottler, et al. 1981; Lee, et al. 1983) or that a negative relationship exists indicati ng that larger homes pr ovide a lower level of quality (Aaronson, et al. 1994; Harrington, et al. 1998). Nyman (1988c) shows that size does not have a significant relationship with two process measures of quality, resident care plans and diet plans, while Ullmann ( 1981) and Lee et al. (1983) show size has a positive relationship with an index of rehabilitation services indicating that larger homes provide a higher level of quality. Size and outcome measures of quality, such as mortality and improvement in functional status are not significant in some studies (Linn,

PAGE 48

36 et al. 1977; Porell, et al. 1998; Spector and Takada 1991) while size and deficiency citations or certificati on violations exhibit a positive rela tionship in other studies (Graber and Sloane 1995; Harrington, et al. 2000c; Nyman 1985, 1988b, 1989b). 3.1.3.3 Nursing Home Staffing As a result of the IOM’s 1986 report on th e poor quality of care in nursing homes and the subsequent passage of the Nursing Home Reform Act of OBRA 87, increased nurse staffing became a requirement for al l Medicareand Medicaid-certified nursing home facilities. RNs, LPNs, licensed vocational nurses (LVNs), and NAs make up approximately 60 percent of total nursing home personnel (Harring ton, et al. 1999) and provide the majority of care that nursi ng home residents receive. Many studies consistently find a positive relationship betw een higher nurse staffi ng levels, especially RNs, and outcome measures of quality of care. One of the first studies to investigate the relationship between nurse staffing and outcome measures of quality finds that more RNs per resident are associated with lower mortality rates, improved resident health, and higher discharge rates (Linn, et al. 1977). More RN hours (Braun 1991), more RN hour s per 100 residents (Cohen and Spector 1996), more licensed nursing hours (Bliesmer, et al. 1998) and more LPN staffing levels (Porell, et al. 1998) result in lower morta lity rates and less likelihood of death. Higher nursing staff levels per reside nt (Spector and Takada 1991) more RNs and LPNs per 100 residents (Cohen and Spector 1996), and more licensed nursing hours (Bliesmer, et al. 1998) also result in improved f unctional status. Harrington et al. (2000c) show that more

PAGE 49

37 RN hours per resident day lead to lower le vels of various measures of nursing home deficiencies. While recognizing that health is not the only desirable outcome to be produced by nursing homes, Nyman (1988c) uses a measure of “quality of life” to represent good quality and finds that more nursing hours per resident day increases the quality of life. Recognizing the importance of these findings, the IOM (1996) issued a committee report recognizing the importance of nurse staffing levels and recommended adding more RNs to the staff of nursing homes. 3.1.3.4 Nursing Home Resident Mix Resident mix refers to the proportions of public-pay (Medicaid or Medicare) and private-pay residents in a nursing home f acility. Policymakers, nursing home industry proponents, and researchers of ten assert that nursing ho mes with higher numbers of public-pay residents are constrained by lower pe r diem rates. As a consequence, it is reasoned that these nursing homes provide a lo wer level of quality since the public-pay reimbursement rates are usually lower than th e rate charged to priv ate-pay residents. Many studies investigate the relationship betw een resident mix and nursing home quality. Again many early studies use expenditure s to represent nursing home quality. With the idea that higher expenditures impl y higher quality, many of these studies find that the proportion of public -pay (Medicaid) residents re sults in nursing homes providing a lower level of quality (Birnbaum, et al 1981; Elwell 1984; Nyman 1988a; Ruchlin and Levey 1972; Schlenker and Shaughnessy 1984). Ho wever, as Davis (1991) notes, a very

PAGE 50

38 important question to answer is whether lo wer expenditures impede structural, process and outcome quality. Assessing this relationship between qualit y and resident mix, some studies use various staffing measures to represent a structural measure of quality w ith the percentage of Medicaid, Medicare, or private-pay residents as the predictor variable. Fottler et al. (1981) and Zinn (1994) find that an increase in the percen tage of Medicaid residents leads to a lower level of RN staffing, thus implying lower quality. Conversely, Harrington et al. (1998) find that an increase in the per centage of Medicaid residents leads to an increase in th e average number of LVN a nd LPN hours per resident day implying that these two types of nursing se rvices improve quality. Alternatively, Zinn (1993) finds that an increase in the percentage of Medica re residents as well as an increase in the percentage of private-pay residents increases the number of LPNs per resident and the number of RN s per resident, respectively. Fewer studies have investigated the re lationship between process measures of quality and resident mix. Nyman (1988c) fi nds no significant rela tionship between the percentage of Medicaid reside nts and ratings of resident care plans, diet plans or medication plans. Likewise, Zinn (1994) fi nds a nonsignificant relationship between the percentage of Medicaid resident s and the percentage of reside nts catheterized as well as the percentage restrained but finds a positive relationship between the percentage of Medicaid residents and the percentage not toileted. The results using outcome measures of qua lity, such as facility deficiencies and changes in functional status tend to be more consistent across studies. Several studies

PAGE 51

39 find that the percentage of Medicaid residents is asso ciated with worse outcomes (Harrington, et al. 2002; Harrington, et al. 2000c; Nyman 1988b) while the percentage of Medicare and/or the percentage of private-pay residents is typi cally associated with better outcomes (Nyman 1985, 1989b; Porell, et al. 1998; Spector and Takada 1991). 3.1.3.5 Nursing Home Case-Mix Nursing homes serve different types of resi dents requiring different levels of care. Studies usually include measures of case-mix to serve as an indicator of the severity of a resident’s functional condition. One of the most widely used measures of resident casemix is the activities of daily living (ADL) index developed by Katz (1963). This index summarizes a resident’s over-all performa nce in six functions: bathing, dressing, toileting, transferring, contin ence, and eating. The higher the index value the more severe is the resident’s functional conditi on. Another measure of case-mix frequently used in the literature is the Resource Utilization Groups (RUG). The RUGs approach categorizes residents according to the amount of resources re quired for their care. By classifying residents into hom ogeneous categories based on their resource utilization, this type of case-mix index represents, at least relatively, the time or cost of the average resident in the group (Fries 1990). Now in its third version, a higher RUG index indicates a greater degree of complexity and, consequently, a gr eater need for input resources (Fries, et al. 1994). Additionall y, certain process measures as well as the percentage of SNF residents in the nursing home facility ar e used in some studies as measures of resident case-mix.

PAGE 52

40 Various measures of nursing home sta ffing are the predominant structural measure of quality within the nursing home literature when examining the effect of casemix. One might conclude that a more severe resident case-mix require s greater staffing. However, once again the literature gives c onflicting evidence. Using a long-term care index representing functional severity of nursi ng home residents as th e indicator of casemix, Cohen and Dubay (1990) find a positive relationship between case-mix and nurses per bed while Aaronson et al. ( 1994) find a negative relationshi p with care staff per bed. However, two studies employing a modified ADL index both find that a more severe case-mix leads to higher staffing levels (Grabowski 2001b; Harrington, et al. 1998). Using process measures of quality, the ev idence is also mixed concerning the relationship between quality of care and case-mix. Nyman finds that more residents with special needs is not a significant predicto r of quality (Nyman 1988c) while Zinn (1994) finds higher functional severity has a positive relationship with the percentage of residents catheterized but a negative relationship with the percentage restrained. Two other studies find a positive relationship be tween a more severe case-mix and several process measures representing “poor” quality of care indicating that homes with a more severe case-mix provide lower levels of quality (Aaronson, et al. 1994; Mukamel 1997). The use of deficiencies as an outcome m easure of quality also presents opposing results. In several studies, a more severe case-mix leads to more nursing home violations (“lower” quality) (Graber and Sloane 1995; Grabowski 2001b; Harrington, et al. 2002) while leading to less deficiencies (“higher” qualit y) in another (Harrington, et al. 2000c).

PAGE 53

41 3.1.3.6 Nursing Home Reimbursement Methods Reimbursement methods refer to the way in which Medicare and state Medicaid programs pay for nursing home care. Setti ng Medicaid reimbursement rates for nursing home care is one way that states attempt to control expenditures. However, states focus not only on the overall level of reimbursement but also on the payment method by which they pay nursing home providers. As describe d previously in Chap ter 2, section 2.3.1, reimbursement policies differ most fundament ally on two issues: whether rates are set retrospectively or prospectivel y and if the resulting prospec tive rate is based on facilityspecific costs or on a flat-rate set independently of an individual f acility’s costs (Wiener and Stevenson 1998). A retrospective reimbursement method provides the least cost containment incentives because the resulting reimbursement rate is based on actual costs incurred. In contrast facility-specific and flat-rate systems provide the greatest incentive for nursing home facilities to be efficient because w ith these methods a nursing home’s profits are the difference between its payment and its expenses. However, if facilities limit their services in order to increase their profits, quality of care may be adversely affected. Additionally, with a retrosp ective reimbursement method Medicaid revenue increases when more quality is provided because costs are assumed to increase with quality, an incentive may exist to provide more quality. In contrast, a facility reimbursed by a facility-specific or flat-rate method does not receive an increa se in the Medicaid reimbursement rate when more quality is provided and therefore has no financial incentive to increase its quality of care.

PAGE 54

42 Most of the empirical studies using struct ural measures of quality lend support to the theoretical expectations of lower nursing home quality in states that employ a facilityspecific or flat-rate reimbursement system. Cohen and Dubay (1990) find that nursing homes in states using a flat-rate reimburse ment method have fewer nurses per bed than those homes in states using a retrospec tive reimbursement met hod. However, these authors find no significa nt difference between those homes in states using a facilityspecific reimbursement method and those homes in states using a retrospective method. Compared to a retrospective method, Zinn ( 1994) finds this same negative relationship for both a facility-specific and flat-rate re imbursement method and the number of RNs per resident. Cohen and Spector (1996) fi nd the same negative relationship between a flat-rate reimbursement method compared to a retrospective method for the number of RNs but a positive relationship for the num ber of LPNs. In a more recent study, Grabowski (2001b) finds the same negative relationship between a facility-specific reimbursement method and the number of RNs compared to a retrospective method and an even lower level of quality associated with a flat-rate method co mpared to a facilityspecific method. However, under conditions of excess demand, Grabowski finds that the method of reimbursement becomes nonsignificant. Studies using process measures of quality pr ovide inconclusive results. Lee et al. find that retrospective and flat-rate methods, co mpared to facility-specific, lead to lower levels of rehabilitation services offered to re sidents. Zinn finds th at a prospective method leads to a greater percentage of residents restrained while a flat-rate method leads to a greater percentage of re sidents not toileted.

PAGE 55

43 Unlike the structural and proce ss measures of quality, outcome measures are typically not significant diffe rent across reimbursement methods. Cohen and Spector (1996) find no significant di fference between reimburseme nt methods and mortality, presence of a pressure sore, or a change in functional status. Likewise, Grabowski (2001b) finds similar results using the percen tage of residents w ith facility-acquired pressure sores as the measure of quality. Case-mix reimbursement systems are designed to mitigate the disincentives of flat-rate and facility-specific methods of re imbursement to limit nursing home services in order to lower operating costs and to limit the temptation of only admitting lighter-care nursing home residents. Under case-mix reim bursement, nursing homes receive a higher reimbursement rate when individuals require more services. The major theoretical strength of case-mix reimbursement is that it should make nursing homes indifferent to the relative care needs of the individuals they admit. One major criticism of this type of reimbursement system however is that it cr eates a disincentive for nursing homes to provide rehabilitation to its more disabl ed residents (Wiener and Stevenson 1998). With respect to structural measures of quality, Cohen and Dubay (1990) find that use of case-mix reimbursement does not have a significant effect on the number of nurses per bed. In contrast, Zinn ( 1994) finds that use of case-mix reimbursement leads to an increase in the number of RNs per resident. In two separate studies, Grabowski finds that case-mix does not show a significant relations hip with the number of RNs in one study (Grabowski 2001b) and finds a nonsignificant relationship for the number of RNs, a negative relationship with the number of LP Ns, and a positive relationship with the number of NAs in the second study (Grabowski 2002).

PAGE 56

44 With process measures of quality, Zinn (1994) finds that use of case-mix reimbursement leads to a decrease in the per centage of residents not toileted. However, Grabowski (Grabowski 2002) finds that a case-mix reimbursement method has a nonsignificant relationship with both the percentage of resi dents with catheters and the percentage of residents with feeding tubes but a positive relationship with the percentage of residents with physical rest raints. However, when separating the data into the most and least tight markets, the number of RNs has a positive relationship with case-mix reimbursement in the most and least tight markets (the measure of excess demand), the number of LPNs is not signifi cant in either market and the number of NAs is positive in the tightest market and negative in the least tight market. Also in the tightest market, there is a decline in the percentage of resident s with catheters as well as the percentage of residents with feeding tubes when case-mix reimbursement is used (Grabowski 2002). Few studies have investig ated the relationship of case-mix reimbursement and outcome measures of quality. Two studi es by Grabowski show that, although the resulting coefficient is negative, there is no significant relationship between case-mix reimbursement and the percentage of residents with facility acquired pressure sores for the full sample as well as the most and least tight markets (Grabowski 2001b, 2002). One possible interpretation of these re sults between reimbursement method and quality of care is that prospective and flat -rate methods of reimbursement may result in more adverse quality of care than a retros pective method while case-mix reimbursement may not adversely affect the qua lity of care and in some cases improve the quality of care provided to nursing home residents.

PAGE 57

45 3.2 Nursing Homes and Access to Care Access to nursing home care is typically defi ned in the literature as the ability of an individual seeking nursing home care to obtain admission to a nursing home facility. As quoted by Blumstein and Sloan (1978) one of the major goals of the 1974 National Health Planning and Resources Development Act include achieving “equal access to quality health care at a reasonable cost” and improving the “maldistribution of health care facilities and manpower” and one of the Act’s first priorities is the “provision of primary care services for medically underserved popul ations”. While the main goal of CON regulation is the containment of the share of medical care expenditures paid for by the federal and state governments one of the perhap s unintended consequences of this type of regulation in the nursing home industry is reduced access to care for public-pay and “heavy-care” residents (Cotterill 1983; Feder and Scanlon 1980; Greenlees, et al. 1982; Scanlon 1980a, 1980b; Schlenker 1986). The two reasons most often cited by researchers of the nursing home industry are (1) the excess demand for nursing home beds caused by CON and/or construction morato rium and (2) the low level of the publicpay reimbursement rate. 3.2.1 Nursing Home Access and CON Although CON and construction moratorium policies are designed to control government expenditures by limiting the number of nursing home beds, they also create potential barriers to entry for new providers. As a consequence, when a constraint such as CON is binding, excess demand for nursing home care may result in many markets. As described by Scanlon (1980b) in his excess demand model of the nursing home

PAGE 58

46 market, a profit-maximizing facility first ad mits the higher-paying private residents and then fills the remaining empty beds with lo wer-paying Medicaid residents. Thus, when the number of nursing home beds is constraine d the private-pay demand is satisfied first and any remaining demand becomes “excess Medicaid demand.” As a result, Medicaid residents will have access to those homes l east capable of competing successfully for private-pay residents, perhaps due to lowe r quality of care; and with such a limited choice, public-pay residents of ten have no option but to ente r whatever facility will accept them, even though it may provide undesi rable quality of care (Hawes and Phillips 1986). Additionally, there is what is referred to in the literature as “cream-skimming”; when a Medicaid-eligible individual is admitte d to a nursing home, it will be the lightercare individual that will be admitted befo re a heavy-care individual since the former represents a more profitable resident (Cohen and Dubay 1990; Frech III 1985; Hawes and Phillips 1986). As with the relationship between quality of care and CON and/or construction moratorium policies, the existing literature do es not provide insight into any differences in access to nursing home care th at may exist between those states without CON policies and those states that have such policies. Nor does the existing literature shed much insight into the direct effect of these types of policies on nursing home access for publicpay residents. Feder and Scanlon (1980) attribute the access problem for Medicaid residents directly to the excess dema nd caused by CON regulation. Conducting interviews in eight states in 1978, many of the states they visited reported access problems for their Medicaid-eligib le individuals, especially t hose with heavy-care needs.

PAGE 59

47 The authors attribute the problem to a bed s hortage that enables nursing home operators to discriminate in favor of light-care residents. Early work by Lee et al. (1983) estimate s a multi-equation simultaneous model of nursing home behavior using data from the Na tional Center for Health Statistics’ 19731974 Nursing Home Survey. One of the equations estimates the effect of the presence of CON on a nursing home’s occupancy rate and th en this result is used as one of the explanatory variables to estimate the percenta ge of total resident days provided to private-pay residents. The study finds a positive relationship between CON and occupancy rates as well as a positive rela tionship between occupancy rate and the percentage of total resident days provided to private-pay residents. This result indicates that, at the time, public-pay residents had less access to nursing home care relative to private-pay residents. Again, although not a direct measure of CO N, many studies estimate the effect of excess demand on access to nursing home care for public-pay residents. Many of these studies analyze data from the time of Fe deral CON regulation and when nursing homes were experiencing high occupancy rates (C ohen and Dubay 1990; Ettner 1993; Gertler 1992; Harrington and Swan 1987; Nyman 198 9a; Reschovsky 1996). While each of these studies employs a different data se t, methodology, measure of excess demand and measure of access, the results consistently s how that public-pay resi dents have less access to nursing home care than pr ivate-pay residents. Nyman (1989a) utilizes county-level data for nursing homes in the state of Wisconsin from 1983 to assess the impact of excess demand on access to nursing home care for Medicaid recipients. While contro lling for demographic and Medicaid policy

PAGE 60

48 variables as well as a measur e of quality, the OLS result for the effect of excess demand (the number of beds per thousand elderly) on the number of Medi caid residents (the number of Medicaid residents per thousand elderly) is negativ e indicating that an increase in excess demand decreases Medicaid-eligible individuals’ access to care. Gertler (1992), using data from New York State in 1980 which at the time had a high occupancy rate and what was considered a binding CON, estimates a reduced-form equation of the effect of the number of beds in a facility on the number of Medi caid residents. While controlling for several demand, supply, and po licy variables as well as a measure of resident case-mix, the results indicate that a decrease in the number of beds, i.e., an increase in excess demand, leads to a decrea se in the number of Medicaid residents. Utilizing a panel data set from 42 st ates for the entire period of 1978-1983, Harrington and Swan (1987) estimate the e ffect of nursing home beds per 1,000 aged (their measure of excess demand) on the num ber of Medicaid nursing home recipients per 1,000 aged (their measure of access) at the st ate level using a fixed effects estimator. While controlling for several Medicaid policies and demographic variables, an increase in excess demand causes a decrease in nur sing home access for Medicaid-eligible individuals. Performing a cr oss-sectional analysis, Cohen and Dubay (1990) estimate a reduced-form regression equation of the effect of the number of Medicare-certified beds per elderly population aged 65+ (a measur e of the tightness of the market) on the percentage of Medicaid residents in a nursi ng home facility. Using data from the 1981 Medicare cost reports an d Medicare/Medicaid Automated Certification System (MMACS) files for 1,020 nursing homes th roughout the United States, the results indicate a negative but n onsignificant relationship.

PAGE 61

49 Pursuing a different technique, Ettner (1993) uses a probit model to estimate the impact of Medicaid status on the probability of being on a waiting list for a nursing home bed while controlling for resident mix, casemix, and bed supply. Using data from the 1982-1984 National Long Term Care Survey (N LTCS) Ettner finds that Medicaideligible individuals have a greater probability of being on a waiting list and that these individuals face greater access problems than non-Medicaid individuals in areas where bed supply is constrained and that on the ma rgin, Medicaid-eligible individuals benefit the most from an increase in the bed supply. Utilizing data from the National Long -Term Care Channeling Demonstration, Reschovsky (1996) uses a probit model to estim ate the effect of the number of empty beds per 1,000 population aged 75+ on the probability of a Me dicaid-eligible individual entering a nursing home, specified as the pr oduct of the probability of demand for such care and the probability that a nursing home admits the person conditional on demand. Using the interaction between the measure of bed availability and the expected net revenue from the sample person as a measur e of market tightness, the results again indicate that the tighter the market conditi on the lower the probability of admission for Medicaid-eligible individuals. Furthermore, by limiting the number of nursing home beds, CON and construction moratorium not onl y provide nursing homes with the potential ability to choose which payer type (Medicaid or private) to admit but also with the potential ability to choose which type of Medicaid resident to admit. Because every Medicaid-eligible individual will not be able to find an empty bed in the presence of excess demand, “heavy-care” Medicaid-eligible individuals ma y experience the most difficulty in gaining

PAGE 62

50 access to nursing home care (Grabowski 2002). Using a health status index representing severity of illness and the n eed for heavy-care, studies show that excess demand leads to fewer heavy-care residents being admitted to nursing homes while controlling for Medicaid policies as well as facility and area character istics (Cohen and Dubay 1990; Nyman, et al. 1987). 3.2.2 Other Predictors of Nursing Home Access In the existing literature, se veral variables are consisten tly used as predictors of nursing home access. Among those reviewed in this section are the reimbursement rate, reimbursement method, case-mix, resident mix, and ownership of a nursing home facility. While recognizing th at each of the studies incl uded in this portion of the literature review utilizes different data sets and methods of analysis, this section provides a summary of the results obtained when uti lizing several of the other more important predictors of nursing home access. 3.2.2.1 Nursing Home Reimbursement Rates and Methods In addition to the excess demand theory many researchers have attributed the difficulty of gaining access to nursing home care for Medicaid re sidents to lower-thancost Medicaid reimbursement rates. Beside s providing an incentive for nursing homes to admit private-pay residents before public-pay re sidents, a Medicaid rate that is lower than the private-pay rate also pot entially provides nursing homes with a possible incentive to selectively admit those individuals with the fe west functional disabilities. This incentive is particularly strong under reimbursement sy stems that provide the same Medicaid rate

PAGE 63

51 for every resident regardless of the resident ’s degree of physical impairment (Hawes and Phillips 1986). Therefore, if all Medicaid residents in a give n home bring in the same per diem reimbursement, nursing homes will be relatively reluctant to admit the most severely impaired residents, i.e., those needing “heavy care” (Hol ahan and Cohen 1987; Nyman, et al. 1987). Even though at least one study finds that the marginal cost of the most dependent SNF resident is lower than the average Medicaid reimbursement rate for SNF residents, most studies of nursing ho me access include a measure of the Medicaid reimbursement rate (Mukamel and Spector 2002; Nyman 1988d).5 Using various measures to represent demand for nursing home care as well as measures of demographic, resident, and facility characteri stics as control variab les, the results from various studies suggest that an increase in the Medicaid reimbursement rate improves access to nursing home care for public-pay re sidents (Aaronson, et al. 1994; Gertler 1992; Greenlees, et al. 1982; Reschovsky 1996). Other papers have looked at the effect of the reimbursement method on access to nursing home care for public-pay residents and in particular those residents considered heavy-care.6 Many studies focus on four methods of reimbursement, retrospective, prospective-class (flat-rate), prospective-facility-specifi c, and prospective-case-mix adjusted, and their effect on access to nursing home care for Medicaid-eligible individuals. In contrast to a retrospective reimbursement, which encourages nursing homes to increase expenditures, flat-rate and facility-specific reimbursement methods encourage nursing homes to minimize expenditures and produce efficiently while the 5 Nyman’s study uses data from New York State in 1983, which was at the time of Federal CON while Mukamel and Spector’s study uses data from Ne w York State in 1991, which was still under CON regulation. 6 The various Medicaid reimbursement methods are described in detail in section 2.3.1.

PAGE 64

52 intended purpose of a case-mix adjusted re imbursement method is to make facilities indifferent to resident s’ care needs when they seek admission. Since each state’s Medicaid reimbursement rate actually results from the choice of reimbursement method, Medicaid rate s tend to be higher under retrospective reimbursement and lower as cost containment incentives increase under facility-specific and flat-rate methods (Cohen and Dubay 1990; Harrington and Swan 1984). Therefore, it is expected that states with a retrospec tive reimbursement method will provide greater access to nursing home care than those states with a facility-specific or flat-rate reimbursement method. Furthermore, those stat es with a facility-specific reimbursement method are expected to provide greater access for Medicaid residents than those with a flat-rate reimbursement method. Nyman (1990) suggests that simply cove ring the costs of each resident is insufficient to ensure access for heavy-care residents, particularly under conditions of excess demand. The purpose of a prospectiv e-case-mix-adjusted reimbursement method is that the use of case-adjusted rate s would permit governments to pay higher reimbursement rates for heavier-care resident s, thus creating an incentive for nursing homes to admit these types of residents. Thus the main po licy designed to insure that heavy-care residents gain access to nursing homes is represented by these case-adjusted prospective reimbursement systems. The exp ected result of case-mix reimbursement is improved access to nursing homes for heavycare Medicaid residents (Nyman, et al. 1987; Weissert and Musliner 1992). The results in the literature however, are inconclusive. Several studies find that states using a retrospective method actually ha ve fewer Medicaid re sidents compared to

PAGE 65

53 those that use a facility-s pecific method (Cohen and Duba y 1990; Lee, et al. 1983) but that those states that use a flat-rate method admit fewer Me dicaid residents compared to those that use a retrospective method as th e theory suggests (Cohen and Dubay 1990). Several studies have researched the e ffect of a case-mix reimbursement method on access to nursing homes for heavy-care Me dicaid residents (Cohen and Dubay 1990; Grabowski 2002; Holahan and Cohen 1987; No rton 1992; Nyman, et al. 1987; Thorpe, et al. 1991). Using various indexe s to represent a measure of the severity of residents’ disability levels to represent access for hea vy-care Medicaid residents, most studies find that the use of a case-mix reimbursement meth od leads to an increase in access for heavycare Medicaid residents (Cohen and Duba y 1990; Grabowski 2002; Holahan and Cohen 1987; Norton 1992; Thorpe, et al. 1991). However, Grabow ski (2002) finds that the increase is not as large under conditions of excess demand while Nyman et al. (1987) find that under conditions of excess demand heavy-ca re residents have less access to nursing home care which supports the idea that hom es located where they can choose among residents will select the lighter -care residents. Overall, th e literature seems to provide support to the theory that reimbursement met hods that pay higher rates provide greater access to Medicaid residents and in particular to those residents requiring heavy-care. 3.2.2.2 Nursing Home Resident Mix Studies that include a measure of a nursi ng home’s resident mix primarily use this variable to serve as an indicator of access for heavy-care residents rather than as an indicator of access for Medicaid residents in ge neral. The idea is to test whether or not nursing homes are able to “cream-skim.” If an increase in the number of Medicaid

PAGE 66

54 residents leads to an increase in a measure of di sability or severity of illness, for example, then it could be said th at the nursing home was not practicing “cream-skimming” by admitting only lighter-care residents. On th e other hand, if an increase leads to a decrease in the severity of illness or disabi lity level then this result would support the notion that nursing homes prefer admitting lighter-care Medicaid residents for all the aforementioned reasons. The few studies that have investigated this relationship do indeed find that an increase in the percentage of Medicaid residents results in a decrease in the severity of illness of nursing homes’ residents suggesting that nursing homes practice “cream-skimming” (Cohen and Dubay 1990; Nyman, et al. 1987). 3.2.2.3 Nursing Home Ownership Studies also examine whethe r or not any differences exist between for-profit and nonprofit nursing home facilities with resp ect to the number of Medicaid residents admitted. One theory suggests that: (1) since nonprofit nursing homes tend to have higher costs than for-profit nursing homes, (2) since Medicare residents tend to cost more relative to Medicaid and privat e-pay residents due to the “s ubacute” rather than the longterm nature of their required care, and (3 ) since private-pay residents’ demand is a function of price and quality and increasing quality increases cost s, nonprofit homes will be more oriented toward private-pay and Medi care residents than Me dicaid residents and that for-profit homes will be more inclined to admit Medicaid residents. Several studies support this result by finding for-profit nur sing homes have more Medicaid residents compared to nonprofit nursing homes (Cohen and Dubay 1990; Lee, et al. 1983; Vladeck 1980).

PAGE 67

55 Indicating a different result, another theory suggests that nonprofits may rely more heavily on Medicaid residents as a s ource of revenue and therefore have more Medicaid residents than private-pay. Suppor ting this theory, Ger tler (1989) and Davis (1993) find that nonprofit nursing homes have a greater percentage of Medicaid residents compared to for-profit nursing homes while specifically accounting for nursing home case-mix. 3.3 Nursing Homes and Costs The containment of public expenditures on me dical care, due to their rapid growth after the establishment of Me dicare and Medicaid, was one of the main goals of CON regulation. Specific to the nursing home industry, although st udies have shown that the rate of growth of nursing home beds has sl owed with the use of CON and/or construction moratorium (Harrington, et al. 1997b; Sw an and Harrington 1990), the evidence is not conclusive that CON and/or construction mo ratorium policies have been effective in reducing public expenditures for nursing home car e or for long-term care in general. 3.3.1 Nursing Home Costs and CON Early work typically examines the effects of CON policies on bed capacity (Feder and Scanlon 1980; Harrington, et al. 1997a; Ha rrington, et al. 1992; Harrington, et al. 1997b) or the effects of bed capacity on Medicaid expend itures (Harrington and Swan 1987; Swan 1990). Other studies have l ooked at the effect of excess demand on Medicaid nursing home expenditures with most results indicating that under conditions of

PAGE 68

56 excess demand nursing homes are able to lower their costs with impunity due to the lack of competition for residents (Davis and Freeman 1994; Nyman 1988a, 1988b). Although limiting the supply of nursing home beds is intended to constrain nursing home expenditures, Medicaid expend itures for all long-term care may not decrease due to the availability of community -based care, including home health care, as a substitute for nursing home care. Two recent studies address the direct impact of nursing home supply regulations on Medicaid nursing home expenditures as well as Medicaid long-term care expenditures. The fi rst study investigates th e effect of nursing home CON and/or moratorium policies on Medicaid nursing home per capita expenditures, as well as Medi caid long-term care per capita expenditures, using a random effects model for 1991 through 1997 with the state as the unit of analysis (Miller, et al. 2002). Data on nursing home and long-term car e expenditures are from annual Medicaid Financial Management Reports, CMS Form 64, and include all states except Arizona. Included in the model are several variables to control for the demand for and the supply of nursing home care as well as state polic ies, including the type of reimbursement method, and state political fact ors. The results indicate th at the presence of either a nursing home CON, a nursing home moratorium or both a CON and moratorium has a positive, although not significant, effect on nur sing home care per capita expenditures. Additionally, using Medicaid long -term per capita expenditures as the dependent variable and the same explanatory variables, this sa me study finds that the presence of a nursing home CON, a nursing home moratorium and both a CON and moratorium has a positive and significant effect on Medicaid long-term per capita expenditures.

PAGE 69

57 The second study investigates the repeal of CON or moratorium policies on Medicaid nursing home expenditures as well as Medicaid long-te rm care expenditures (Grabowski, et al. 2003). Un like the study by Miller et al. (2002) that does not control for state or year fixed effects, Grabowski’ s study uses a fixed effects model, for 1981 through 1998, to control for the presence of unobserved state-specific as well as unobserved time-specific attributes that may influence both the elimination of CON or moratorium regulation and the level of nursi ng home and long-term care expenditures. The expenditure data include information on all states except Arizona from CMS’ Office of the Actuary. Controlling for demographic and economic variables as well, the results indicate that those states w ithout CON or moratorium policies have a very small, but statistically insignificant, increase in Me dicaid nursing home expenditures as well as Medicaid long-term care expenditures. Additionally, when the nursing home Medicaid expenditure data are decomposed into the pe r diem Medicaid rate and Medicaid recipient days, the results indicate that the repeal of CON did not statistically increase the Medicaid per diem rate or the number of Medicaid recipient days. These two studies are the first to actually investigate the effe ct of nursing home supply regulations on nursing home costs during the time period during which the Federal CON regulation was eliminated and in which some states eliminated their CON policy while other states kept their policy in place. These studies provide supporting evidence to early studies that suggest that CON and/or moratorium policies are not having the intended effect of control ling Medicaid nursing home expenditures (Birnbaum, et al. 1981; Lee, et al. 1983). Thes e results also suggest that this type of regulation may not be having the intended e ffect of reducing Medicaid long-term care

PAGE 70

58 expenditures. One possible explanation is that market conditions have changed with respect to the lessening of excess demand and th e increased availability of substitutes for nursing home care (Grabo wski, et al. 2003).

PAGE 71

59 Chapter Four Research Design This chapter focuses on the data and met hods utilized in this dissertation. The first section describes the obj ectives and hypotheses. The second section describes the various sources of data as well as the proce dures utilized to clean and merge the various data files. The third section describes the variables included in each of the models. The last section describes the methodologies app lied and includes the specification of the models in the analysis of th e differences in the quality of nursing home care and the access to nursing home care for those states w ith and those states without CON and/or construction moratorium regulation. 4.1 Objectives and Hypotheses Since the elimination of Federal CON re gulation in the nursing home industry, little empirical analysis has been performed to see if the quality of care and the access to care is any different between those states without CON and/ or construction moratorium policies and those states that still have these policies in pl ace. The question remains if CON is effectively meeting its original inte nt of improving quality of care for nursing home residents, increasing the accessibility of nurs ing home care for public-pay residents, and containing public expend itures for nursing home care. The specific

PAGE 72

60 objectives of this dissertation are to investigate if CON re gulation is achieving the goals of improving quality and increasing accessi bility by analyzing whether or not any differences exist between those states w ithout and those states with CON and/or moratorium policies in place. These results, taken together with the results of Miller et al. (2002) and Grabowski et al. (2003) and, ma y shed some light on the effectiveness of retaining state-level CON and/or co nstruction moratorium policies. The hypotheses to be tested in this di ssertation are based on Scanlon’s (1980b) model of a monopolistically comp etitive market which provides the same level of care to both private-pay and public -pay residents and Nyman’s excess demand paradigm (Nyman 1985, 1988b, 1989b). When the number of pr ospective residents is greater than the supply of nursing home beds, the nursing home provides care to priv ate residents first since the private-pay price is t ypically greater than the Medica id payment rate. If CON is an effective barrier to entry and excess demand exists, and si nce the private-pay price is typically higher than the public-pay reimbur sement rate, nursing homes will not have to compete for Medicaid-eligible individuals on the basis of quality since at any quality level a sufficient number of these individuals are available to fill an empty nursing home bed. As a result, nursing homes are able to provide minimal quality of care and still attract as many Medicaid-eligible individuals as they desire. By eliminating CON and/or moratorium policies, nursing homes may have to compete on quality in order to attract privat e-pay as well as pub lic-pay individuals. Additionally, elimination of these policies will le ad to an increase in supply which will increase access for public-pay individuals since these are the one s who face the excess demand condition. And since nursing homes are also able to discriminate between light-

PAGE 73

61 care and heavy-care individuals, elimination of a supply constraint may also increase access for heavy-care individuals. Formally stated, the following hypotheses wi ll be tested in this dissertation: Hypothesis 1: Quality of care is higher in nursing homes in states without CON and/or moratorium policies. Hypothesis 2: Access to care for Medicaid-el igible individuals is greater in nursing homes in states without CON a nd/or moratorium policies. 4.2 Description of Data The data for this dissertation comes fro m five main sources: (1) the On-Line Survey, Certification, and Reporting (OSCAR) system,7 (2) state-level Medicaid reimbursement data from the 1998 State Data Book on Long Term Care Program and Market Characteristics (Harrington, et al. 2000b), (3) su rveys of state Medicaid offices and regulation departments conducted during this study, (4) the Ar ea Resource File (ARF), and (5) the Bureau of Economic Analysis’ (BEA) Regional Economic Information System (REIS). Sections 4.2.1 th rough 4.2.4 describe in detail each of these data sources while section 4.2.5 de scribes the construction of the data set utilized in this research. 7 Prior to OSCAR, the reporting mechanism was the Medicare/Medicaid Automated Certification System (MMACS).

PAGE 74

62 4.2.1 The OSCAR Data System The OSCAR data system is a repository for data that is collected by state surveyors for all federally certified Medicare and Medicaid nursing hom e facilities in the United States. These data are collected a nd maintained by CMS in order to determine whether or not nursing homes are in compliance with federal regulatory requirements. In order to become certified every facility must ha ve an initial survey to verify compliance. After the initial cer tification, states are surveyed no le ss than every 15 months to ensure continued compliance as well as verificat ion of the correction of any previous deficiencies. A nursing home is also required to be surv eyed when there is a change in management or organization. Finally, a hom e may be surveyed as part of a follow-up when a complaint has been filed that alleges substandard care (Harri ngton, et al. 2000a). The OSCAR data are collected in three sepa rate files: (1) facility characteristics and staffing data, (2) resident characteristics, and (3) surv ey deficiencies. Facility characteristics include measures such as size, ownership, certificati on status, and resident mix while staffing data consists of measures such as the number of full-time equivalent (FTE) RN, LPN, and aide nursing hours. Resi dent characteristics include measures such as the average number of ADLs of the nur sing home’s residents and the number of residents with physical restraints. Survey deficiencies are classi fied into 17 major categories which include resident rights, qual ity of life, nursing serv ices, and quality of care (Harrington, et al. 2000c). Recognizing the importance of the accuracy and reliability of survey data used in research, OSCAR data are collected in a twopart process. First, nursing home personnel use standardized forms to reco rd the facility characteristics, resident characteristics and

PAGE 75

63 staffing levels at the beginning of each surve y. These reports are then certified by the facility as being correct. Th is information is given to th e state surveyors who audit the data by comparing the facility report with indi vidual resident medical records, staffing records, and observations of residents. Once th e review is complete, the state staff enters the data into the OSCAR system. In the sec ond part of the process, the state surveyors decide if the facility has or has not met each federal standard based on information from several sources including, but not limited to, inte rviews with a sample of residents, family members, and staff as well as a review of resi dent and facility records. Once a judgment concerning compliance is made, the state surveyor enters the data for each standard into OSCAR. By using standard forms as well as sampling and survey procedures to ensure accuracy, this two-part process assures that state surveyors are determining deficiencies independently of the facility’s staff. Furthermore, team members and state supervisors subsequently review the decisions of the stat e surveyor. Additionall y, facilities have the option to challenge and appeal decisions through an admi nistrative review process (Harrington, et al. 2000a). The use of a database such as OSCAR always brings concerns about survey procedures and the re liability of surveys both across a nd within states in judging the quality of nursing home facilities. In order to mitigate some of this concern, federal regulations were implemented in 1990 to im prove the sampling procedures and survey methods used by the survey teams. The federa l procedures require state surveyors to use a stratified random sample of residents to c onduct face-to-face interv iews, closed record reviews, and individual as well as group st ructured interviews. CMS also implemented new federal training for state surveyors. Additionally since July 1995, federal surveyors

PAGE 76

64 accompany and observe state surveyors on a se lected number of surveys. Through the Health Care Standards and Quality Bureau of CMS, federal survey teams resurvey a sample of facilities within si xty days of the state survey. A “survey concurrence index” is created for what are considered key co mponents from OSCAR for each state. Any state that falls below the c oncurrence standards are then evaluated and monitored by CMS (Harrington, et al. 2000a). Although OSCAR is not considered an “ideal ” database for research it is currently the most comprehensive, longitudinal information available for all Medicare/Medicaid certified nursing homes in the United States.8 Section 4.2.5 and its subsections discuss the construction of the sample used in this research, the eliminati on and cleaning of the OSCAR data, and the merging of the four data files. 4.2.2 State-level Medicaid Reimbursement Data and Certificate of Need and Construction Moratorium Policies The 1998 State Data Book on Long Term Care Program and Market Characteristics (Harrington, et al. 2000b) is a bo ok summarizing the findings from a project on state long-term care program a nd market characteristics conducted by researchers at the University of California San Francisco, and Wichita State University under a cooperative agreement with CMS and the U.S. Department of Housing and Urban Development. This 1998 version is an update of earlier releases which builds upon a cross-sectional longitudinal data set on long-term care program characteristics for 8 The Minimum Data Set (MDS) provides a compre hensive assessment of each resident's functional capabilities in each Medicare or Medicaid certified nursing home. A Quality Indicator Report (QI) presents data on 24 “indicators” of quality at the state level. However, this information is only available beginning in 2000.

PAGE 77

65 each state that includes the years 1978-1998. Collected by telephone survey, the data consist of information from three sources: (1) state long-term car e providers, (2) state CON and moratorium programs, and (3) Medi caid reimbursement agencies. This book provides the state reimbursement method as well the status of any CON and/or moratorium policy by year. This disserta tion uses information for the years 1991-1998 from this book. For the years 1999-2003 I conducted a surv ey through telephone and e-mail of each state’s Medicaid reimbursement agency and CON/moratorium regulatory agency. For every state I obtained information for each year on whether or not the state had a CON, a moratorium, both, or neither type of regulation. Additionally, I obtained the annual, average Medicaid per diem reimbursement rate for freestanding nursing facilities for every state except New Mexico.9 Lastly, I obtained the type of reimbursement method utilized for every state for each year which included whether or not the state adjusted the rate based on the case-mix of the residents. Appendix A contains Tables A.3, A.4, A.5, and A6 which display the info rmation on states’ CON policies, Medicaid reimbursement methods, whether or not case-mi x adjustment is utilized, and Medicaid reimbursement rates for th e years 1991-2003, respectively. The data utilized from these first two sources are not without drawbacks. The primary issue is that the unit of observation in the empirical analysis is the individual nursing home while the reimbursement method, reimbursement rate, and CON policy are measured at the state-level. While th e reimbursement method and CON policy are typically assigned at the st ate-level, using the state-specific average annual 9 The econometric strategy used to control for the unavailability of New Mexico’s rate is discussed in section 4.3.4.3 of this chapter.

PAGE 78

66 reimbursement rate for each nursing home in a state is not ideal. However, using the average, annual per diem rate for the state el iminates the concern of potential endogeneity with a facility’s quality level, resident mi x and case-mix that woul d exist if a facilitylevel reimbursement rate was utilized. Th e other issue associated with the rate information is that the rates are not necessari ly measured with the same precision across states. 4.2.3 The Area Resource File The Area Resource File (ARF) is a national, county-level health resources information database maintained by Quality Resource Systems, Inc. (QRS) under contract to the National Center for Health Workforce Analysis (NCHWA), Bureau of Health Professions within the Health Res ources and Services Ad ministration (HRSA), which is a part of the Department of H ealth and Human Services (DHHS) (QRS 2006). The file contains statistics on categories of health resources such as: health professions, health facilities, health prof essions training, and utilization. It also contains specific geographic codes and descri ptors as well as information on economic activity, population, and environmental characteristics. These variables come from over 50 primary sources of data including the Ameri can Medical Associati on, the National Center for Health Statistics, and the Bureau of Labor Statistics. ARF is updated annually and provides data on over 6,000 variables for all counties in the United States. Some variable s are carried “historically” while others are “updated” and appear for only a few years. The 2003 version of AR F was used in this study. Previous to the 2001 version, data for Alaska and some independent cities in

PAGE 79

67 Virginia were not assigned to a county. However beginning with the 2001 version, data are broken out for all Virginia independent cities and Alaska boroughs/census areas for all data from 1992 through the current year. Therefore as explained in section 4.2.5.3, I was able to match every OSCAR observation in the final data set with ARF data. The population aged 65 and up per county was used to construct the variable that serves as a proxy for market tightness disc ussed below in section 4.3.4.2.10 4.2.4 The Regional Economic Information System The Regional Economic Information Syst em (REIS) contains estimates of personal income and employment for local areas that is prepared by the Regional Economic Measurement Division of the Bure au of Economic Analysis (BEA). The variable I obtained is the county per capita person al income for the years 1991-2003 which I then adjusted using the 2003 CPI as the base. Personal income is defined as “the income received by, or on behalf of, all the resi dents of an area (nat ion, state, or county) from all sources” (REIS 2005). It consists of the income received by persons from participation in production, government and bus iness transfer payments to persons, and government interest payments to persons. Personal income is the “sum of wage and salary disbursements, supplements to wages and salaries, proprietor s’ income, dividends, interest, and personal current transfer receipts, less contributions for government social insurance” (REIS 2005). County per capita person al income is then calculated as the 10 The population data for 2003 was obtained from the U.S. Census Bureau’s American Fact Finder at www.census.gov since the 2003 ARF file only contains data through 2002.

PAGE 80

68 personal income of the residents of the c ounty divided by the population of the county as of July 1 of the respective year. The estimates of county per capita persona l income used in this dissertation incorporate the results of a comprehensive revision to the national income and product accounts (NIPAs) released in December 2003 and the annual revision released in July 2004 (REIS 2005). The data come from a new table, CA04 County income and employment summary, 1969-2003, which provides one table with the entire time series of the summary estimates including per capita personal income. 4.2.5 Sample Construction As described in section 4.2.1, the accuracy of the OSCAR data is dependent upon the accuracy of the facility personnel as well as the state survey team and data entry personnel. Even though CMS attempts to en sure accurate data co llection as well as accurate data entry, there was some additiona l data cleaning required for the final data set. Section 4.2.5.1 describes the procedure used to identify and eliminate duplicate records in the OSCAR data and other data cleaning steps; sec tion 4.2.5.2 describes the procedure used to examine and eliminate obser vations for obvious repo rting errors in the OSCAR data; and section 4.2.5.3 describes the merging of the various data files. 4.2.5.1 Elimination of Duplicate Records and Data Cleaning The OSCAR data used in this study encompass the years 1991-2003 where each yearly file is a result of a download of the data base at a particular po int in time. The final data set is a result of appendi ng each of the individual years to gether into one file. Since

PAGE 81

69 the federal requirement is for each facility to be surveyed every nine to fifteen months it is possible for the same survey date to be in more than one OSCAR file dependent on the point in time that the file download was created. In order to identify duplicate observations the appended data set was sorted by provider number, facility name, facility addr ess, city, state, and survey date. The observations that had matching values for each of these six variables were identified and marked for further analysis of the actual valu es of the variables. The record that came from the most recent survey download was kept in the final data set. Additionally, if the values of data were the same for those observations with the same provider number, facility name, and survey date the latest su rvey date was retained. Otherwise neither observation was eliminated. This pr ocess eliminated 16,223 observations. The data were then analyzed for observa tions that came from different survey dates but were performed in the same calenda r year. Since it is possible for a nursing home to be surveyed anywhere between 9 a nd 15 months after the initial certification (Harrington, et al. 2000c), rath er than simply keeping the observation that came from the latest survey I calculated the number of months between the two surveys in the same year as well as the difference between the last survey in the current year and the last survey in the previous calendar year. If the number of months was greater than 9, both observations were retained. For those observa tions where the number of months between the surveys in the same year was less than 9 but the difference between the last survey of the current year and the last su rvey of the prior year was grea ter than or equal to 16, both observations were also retained. Otherwise only the latest survey in the calendar year was retained. This process eliminated 577 observations from the data set.

PAGE 82

70 4.2.5.2 Data Errors The OSCAR data were examined for missi ng values as well as values that appeared to be obvious errors. The means a nd standard deviations of the data were computed and examined for the main variables of interest. Most of the missing data and errors were found primarily in the reporting of the total number of beds, the total number of residents, and the va rious staffing variables. Two problems were associated with the tota l number of beds. Since the focus of this study is on the Medicaid population, if the total number of beds or the total number of certified beds was equal to the number of Medicar e beds the observation was eliminated. Additionally an observation was eliminated if the total number of beds was either missing or less than five. Several problems were detected with re spect to resident data. First, some facilities had missing or zero observations for their total number of residents. These observations were eliminated from the data. Second, some facilitie s reported extremely low numbers of residents. Observations with less than a 10 percent occupancy rate were considered to be erroneous and were eliminated from the data. Third, some facilities reported more residents than the number of beds, which suggests more than 100 percent occupancy. These observations were eliminated from the data. Finally if the percentage of Medicaid residents was zero or if the percentage of Me dicare residents was greater than 70, the observation was eliminated. The latte r implies that the facility is considered to be of a primarily rehabilitati ve care nature (Grabowski 2001a, 2001b). Several problems were identified with facility staffing data. To create the various staffing measures, the total number of sta ff hours reported in a two week period was

PAGE 83

71 divided by the total number of residents and by the 14 days in the reporting period to create hours per resident day for each type of staffing measure RNS, LPNS, AIDES, LICNUR (RNs and LPNs), and NURSTAFF (R Ns, LPNs, and aides). Some facilities reported very high levels of staffing hours while others repor ted very low or no staffing hours. The following decision rules were used to eliminate observations with values that appeared to be either too high or too low. First, observations with values that exceeded 24 hours per resident day for the staffing variables RNS, LPNS, or AIDES were eliminated. Second, since all facilities are required to have some licensed nurses, observa tions with values of zero for the variables LICNUR or NURSTAFF were also eliminated because they we re thought to be erroneous inputs. Lastly, since current minimum federa l standards require that all certified nursing homes with 60 or more beds have an RN on duty for 8 hours a day seven days a week and a licensed nurse (either an RN or LPN) on duty evening and nights (Omnibus Budget Reconciliation Act of 1987), obser vations with total beds gr eater than 60 and RNS equal to zero were eliminated as well. The final adjustment to the OSCAR data was the elimination of hospital-based nursing homes. Since the reimburse ment rate used in this stu dy is the annual, average per diem rate for freestanding nursing homes, 24,217 observations on hospital-based nursing homes were eliminated from the final data set. 4.2.5.3 Merging of the Various Data Files The OSCAR, ARF, and REIS data files were merged together based on the Federal Information Processing Code (FIPS) county code and state code for each

PAGE 84

72 observation. Since the OSCAR da ta contains a county identifi er, it was possible to match the data files together based on the county na me, zip code, and FIPS identifiers. The final merged panel data set contains 150,705 observations fo r the years 1991-2003. Panel data corrects for several of the i ssues involved with working with only cross-sectional or time-series data. As noted by Kennedy (Kennedy 2003), two of the most beneficial aspects of panel data are th at researchers are able to deal with omitted variable bias and that more variability is created through combining variation across units with variation across time helping to alleviate multicollinearity problems. 4.3. Description of the Variables The following sections describe the depende nt and independent variables selected for the models of quality of care and access to care for nursing homes in the United States. 4.3.1 Nursing Home Quality Since there is no universally accepted meas ure of health care quality, this study follows the paradigm of Donabedian and ut ilizes process, structure, and outcome measures of quality. The process-based measures of quality are the proportion of residents with catheters (P ROPCATHETER), the proportion of residents with feeding tubes (PROPPARENTERAL), and the proportion of residents with physical restraints (PROPMOBLREST). While these measures ar e not actual measures of health status, they are widely used as indicators of negative quality because they suggest that substandard care is being provided to nur sing home residents (I OM 1986; Spector and

PAGE 85

73 Takada 1991). For instance, urethral catheteri zation places residents at greater risk for urinary tract infections ofte n leading to hospitalization wh ile longer-term complications are associated with bladder and renal stones, abscesses, and rena l failure (Zinn 1993). Spector and Takada (1991) found that residents in facilities wi th moderate to high use of urethral catheterization had twice the proba bility of functional decline compared to residents in low use facilities. The use of feeding tubes often results in complications including self extubation, infections, aspiration, cloggi ng, and pain (Galindo-Ciocon 1993; Zinn 1993). The Institute of Medicine (2 001) suggests that as little feeding tube use as possible is beneficial to residents. Finally, the immobility associated with the use of physical restraints may increase the like lihood of pressure ulcers, incontinence, and depression (Zinn 1993). One study fi nds that physical restraint use is associated with an increased risk of morbidity and mortality in nur sing home residents (Phillips, et al. 1993). Besides physical consequences associated w ith restraints there are often psychological consequences as well (Castle and Mor 1998). Th erefore lower levels of physical restraint use indicate higher quality of care. The structure-based measures of quality are the number of registered nurse hours per resident day (RNS), the number of licen sed practical nurse hour s per resident day (LPNS), the number of aide hours per resi dent day (AIDES), the number of licensed nurse hours per resident day (LICNUR), a nd the number of nursing staff hours per resident day (NURSTAF F). The first three staffing measures are not considered substitutes for one another since they require different levels of tr aining and certification and each type actually has various levels of care that they are permitted to administer. The second two measures of staffing are include d to give a more general measure of the

PAGE 86

74 types of care available to reside nts. Staffing intensity is of ten used as an indicator of positive nursing home quality since more staffing is likely to be associated with an improved quality of life for residents since th ey are receiving more individual attention. As is the case of most of the nursing home quality literature, the use of these structurebased measures assumes that more staffing implies higher quality rather than more inefficiency since the majority of nursing hom es are for-profit facilities which focus on cost minimization (Zinn 1994). The outcome measure of quality is the propor tion of residents with pressure sores (PROPPRESSORE). Pressure sores (decubitis ulcers) are an injury to the skin and nearby tissue. They occur most often in bony areas such as the hips, heels, or tailbone and are caused by constant pressure on the sk in. People confined to a bed or chair and unable to move are at greatest risk for develo ping pressure sores, which makes the elderly population in a nursing home vulne rable to such a condition. Pr essure sores are found to be associated with an increas ed rate of morbidity and mo rtality (Allman 1989; Brandeis, et al. 1990). Pressure sores are often used as a measure of negative nursing home quality since they are treatable and preventable c onditions even though th ey occur frequently (Grabowski 2001b; Harrington, et al. 2000c; Smith 1995). The final measure of quality is the total number of facility deficiencies (DEFS1TOT). This measure is considered composite-based because it is based on 180 survey items that include structural, proce dural, and outcome measur es of quality that represent the standards of nursing home qua lity (Harrington, et al. 2000a). When a facility fails to meet one of th e standards, a deficiency or cita tion is given to the facility.

PAGE 87

75 Penalties are often severe depending on the type and number of deficiencies and can result in civil penalties as well as th e extreme outcome of facility closure. Unfortunately the OSCAR data do not allo w a determination of how much of the proportion of residents with a urethral cathete rization, with tube feeding, with a physical restraint, with pressure sores, or how many deficiencies or staffing hours are attributable to Medicaid residents and which are attributable to private-pa y residents. It is assumed therefore that the measures are distribute d proportionally among public and private-pay residents in a facility. This seems to be a plausible assumption since there is a legal restriction that facilities provide the same level of quality to all nursing home residents.11 4.3.2 Nursing Home Access The percentage of Medicaid residents in a facility (PCTMCAID) is the measure of access in this dissertation. Based on Scanl on’s model of a monopolistic competitor, a nursing home will admit a private-pay individu al before a Medicaid-eligible individual due to the higher private-pay price. If a CON or moratorium is binding, excess demand will exist and it will be Medicaid-eligible i ndividuals who are unable to obtain a nursing home bed (Scanlon 1980b). Therefore it is expe cted that the elimination of CON and/or moratorium policies will lead to a highe r percentage of Medicaid residents. Additionally, I use the variab le ADLINDEX which is a measure of resident acuity (described in detail in Section 4.3.4.1 below) as a measure of access to care for heavycare residents. I expect to find that if C ON and/or moratorium policies are effective in 11The State Operations Manual Guidelin e §483.12(c) states “Identical policies and practices concerning services means that facilities must not distinguish between residents based on their source of payment when providing services that are required to be provided under the law” (Health Care Financing Agency 1995)

PAGE 88

76 constraining supply the ADLINDEX will be lower in those states retaining these types of policies. 4.3.3 Certificate of Need and Co nstruction Moratorium Policy The main variable of interest in this disse rtation is the presence of a CON and/or a moratorium policy. The primary goal of CON and construction moratorium policies is to retard the growth of health care costs by preventing the “unnecessary” expansion of nursing home beds. In effect these types of policies imply that fewer nursing home beds will lead to fewer Medicaid residents in the nursing homes which in turn lead to lower Medicaid expenditures. However a recent st udy has shown that these types of policies do not have a large impact on constraining Medica id expenditures (Grabowsk i, et al. 2003). Although these policies theore tically restrict or prohi bit growth in the nursing home market it is not always the case in prac tice. Many states have exceptions for their moratorium policy that allows for additional beds or expansion if it is deemed a critical need. Additionally, my survey of states’ regul atory agencies indicates that states vary on the restrictiveness of their CON policies. In some states the CON actually acts as a moratorium while in other states it appears that the CON is simply a formality and most applications for additional or new beds get approved. For this dissertation I use a dummy variable to represent whethe r or not a state does not have a CON and/or moratorium policy (CON_MORT) for each year since the time it would take to ascertain the degree of complexity of each state’s policies was prohi bitive. Table A.3 in Appendix A shows the compilation of information on states’ regul atory policies for the years 1991-2003. I expect to find higher quality and more Medicaid residents in nursing homes in states

PAGE 89

77 without a CON and/or moratorium policy relati ve to those states with this type of regulation. 4.3.4 Other Independent Variables Several facility, market, and state-le vel exogenous variables are included to control for economic and demographic conditions that may influence the quality of care and access to care in nursing homes. The followi ng sections describe these variables in detail. 4.3.4.1 Facility-level Characteristics The first facility-level characteristic is the total number of beds in a facility (TOTBEDS). This variable represents the size of the facility. As described in the literature review on quality, size has been shown to be both a positive and negative indicator of quality. With resp ect to access, previous research has not directly controlled for the size of the facility. However, one might expect that a facility with a greater number of beds will have more Medicaid residents relative to a facility with fewer beds if excess demand is not an issue. Therefore I control for the total num ber of beds in the facility since the facilities in the data set are of varying sizes. The second facility-level ch aracteristic represents th e nursing home’s ownership type. Two dummy variables represent if the facility is nonpr ofit (CONTROL2) or government owned (CONTROL3) w ith for-profit the omitted cat egory. With respect to access, previous research either supports the theory that nonprofit homes will be more oriented toward private-pay and Medicare resi dents than Medicaid re sidents and that for-

PAGE 90

78 profit homes will be more inclined to adm it Medicaid residents or that that nonprofits may rely more heavily on Medicaid resident s as a source of revenue and therefore have more Medicaid residents than private-pay as discussed in section 3.2.2.3 of the literature review. In conjunction with ow nership type, I include a dummy variable that controls for whether or not the nursing home is part of a multi-chain facility (MULTI). While there is not much literature that addresses the impact of chain ownership on quality of care or access to care, the industry has moved in the di rection of mergers in the last several years and therefore I control for this feature. The final variable to control for facility-l evel characteristics is really a residentlevel characteristic. Nursing homes serve diffe rent types of resident s requiring different levels of care. Studies usually include a m easure of “resident acuity” to serve as an indicator of the severity of a resident’s functional conditi on. One of the most widely used measures of resident acuity is the activiti es of daily living (ADL) index developed by Katz (1963). This index summa rizes a resident’s over-all pe rformance in six functions: bathing, dressing, toileting, transf erring, continence, and eating. The measure of resident acuity used in this dissertation (ADLINDEX) is calculated by summi ng various levels of dependencies in eating, toileting, transferri ng, and mobility which are weighted by the respective proportion of residents.12 The result is an index of the average functioning level for the residents in each f acility for each year. The high er the value of the index the more dependent the resident is in the functions mentioned above. 12 This index is calculated based on the formula used in the 2002 Nursing Home Statistical Yearbook published by Cowles Research Group.

PAGE 91

79 4.3.4.2 Market-level Characteristics The county in which the nursing home is located is used to approximate the market in this dissertation. Most economic studies have used the county as a proxy for the nursing home market (Cohen and Spect or 1996; Gertler 1992; Nyman 1985; Zinn 1994). Nyman found that 80 percent of reside nts in facilities in Wisconsin chose a nursing home in the same county in which the resident had previous ly lived. Gertler found that 75 percent of resident s in New York facilities had previously lived in the same county as the nursing home. As noted by Bana szak-Holl, et al. ( 1996) the county can be considered a good approximation of the ma rket for nursing home care based on the patterns of funding and resident origin. For instance, federal block grants for long-term care services are distributed at the county level. While it ha s been argued that residentorigin data are preferable to county b oundaries in delineating nursing home markets (Zwanziger, et al. 2002), the OSCAR data do not provide resi dent-origin data. The first market-level characteristic, EMPTYELDERLY, is the number of empty nursing homes beds per 1,000 noninstitutionalized elderly (aged 65+) in a county. This variable serves as a proxy for the presence of excess demand and attempts to account for the tightness of the market. A bed constraint is assumed to be more restrictive in those markets with fewer empty beds and less rest rictive in those markets with more empty beds. While it would be desirable to c ontrol for the number of Medicaid-eligible individuals in a given market who are wa iting to obtain a nursing home bed due to a CON or construction moratorium law, this inform ation is not available in the data used in this dissertation and other economic analys es. Since it is not clear that CON and moratorium policies are always binding and that the occupancy rates in many areas are

PAGE 92

80 declining, controlling for ex cess demand helps alleviate th e concern that nursing home occupancy rates fluctuate from state to state and from year to year In 2003, the three states with the highest occupancy rate were Hawaii (97.6%), Vermont (98.8%), and Rhode Island (92.3%) while the three with the lowest rate were Oregon (70.1%), Montana (70.1%), and Oklahoma (67.1%). It is also true that each of these states had CON or moratorium policies in place duri ng the entire study period demonstrating that these policies do not impl y a condition of excess demand for nursing home beds. The other two market charac teristics included in this analysis represent exogenous demand variables. The first is the county pe r capita personal inco me, adjusted for 2003 dollars (INCOME). The average per capita pers onal income during the time of this study was $27, 280. The second characteristic is a Herfindahl index to measure the concentration of the market (HHI). This index is constructed by summing the squared market shares of all nursing home facilities in the county. The index ranges from 0 to 1 with higher values signifying a more concentrated market. 4.3.4.3 State-level Characteristics Two variables represent state-level charac teristics. The first is the type of Medicaid reimbursement method utilized by the state in which the nursing home is located. The five basic methods of reimburse ment are described in detail in Chapter 2, section 2.3.1. In this study a retrosp ective method is represented by METHOD0; prospective facility-specific is METHOD1; prospective class (flat-rate) is METHOD2; adjusted is METHOD3 (the omitted categor y); and combination is METHOD4. Since the retrospective method is the most gener ous type of reimbursement and prospective

PAGE 93

81 class the least generous, these dummy vari ables provide a test of the effect of reimbursement on quality of care and access to care. Additionally, a dummy variable is included for those states that employ cas e-mix reimbursement methods which pay different rates based on a nursing home’s mix of resident acuity a nd the costs of caring for those residents’ needs. While each state’s reimbursement method can be quite complex, these variables attempt to capture the main differences in reimbursement methodologies. Tables A.4 and A.5, re spectively, in Appendix A display the reimbursement method and whether or not case-mix is utilized for the years 1991-2003. I expect to find that those nursing homes in states that employ a more restrictive reimbursement method have fewer Medicaid residents and lower quality than those nursing homes in states with a more generous method. The second state-level characteristic is the annual, average per diem Medicaid reimbursement rate (RATE). The use of each facility’s specific reimbursement rate would be endogenous to that facility’s quality level. Since no one nursing home can influence the state’s reimbursement rate, us ing the average state Medicaid rate is exogenous at the facility level. In order to a ccount for the fact that I was unable to obtain 1999-2003 rates for the state of New Me xico, I created a dummy variable (RATEMISSING) that equals one for each of the observations with a missing reimbursement rate. For those same observati ons the variable RATE is set equal to zero. I expect to find that an increase in the re imbursement rate leads to an increase in the access to care for Medicaid residents. With respect to quality, ea rly literature showed that an increase in the reimbursement rate lead to a decrease in quality (Gertler 1989, 1992; Nyman 1985, 1988a, 1988b, 1989b) while more recent evidence suggests that an

PAGE 94

82 increase in the payment rate leads to an increase in quality (Grabowski 2001a, 2001b, 2004). This later result has been attributed to changes over tim e in the market for nursing home care related to the decline in nurs ing home utilization (Bishop 1999) possibly due to the availability of nursing home substitutes. I expect to find that an increase in the reimbursement rate has a positive effect on quality of care. Table A.6 in Appendix A displays the average, annual Medicaid per diem rate for each state for the years 19912003. 4.4 Methodology The first part of the analysis conducted is the summary statistics of the final sample. This information provides a gene ral idea of the popula tion under study through the means and standard deviations of the vari ables of the access to care and quality of care models. The next two sections describe the methods as well as the specification of the models used in this study. 4.4.1 State Fixed Effects and Model Specification In order to examine the effect of CON and/or construction moratorium policies on quality of care and access to care, I first estimate a state fixed effects model of the following form: yist = + CON_MORTist + xist + dt + s + it, (1) where yist is the outcome (quality or access) of nursing home i in state s at time t CON_MORTist is a binary variable indicating if the nursing home is in a state s without a CON and/or moratorium policy at time t ; xist is the vector of facility, market, and state-

PAGE 95

83 level characteristics discussed above; dt is a vector of year dummy variables, s is a vector of state dummy variables; and it ci + it which is often referred to as the composite error term. In panel data anal ysis and in this model in particular ci represents an unobserved, time-constant variable fo r the individual nursing home such as administrative ability or location. For each t it is the sum of the unobserved effect (ci) and the idiosyncratic error ( it). The year dummies control for factors that are common across all states in a particular year such as federal nursing home policies and technological advances in health care. The st ate fixed effects contro l for any factors that are specific to a state that remain invarian t over time such as political sentiments and geographic characteristics. This st rategy purges the unobserved and potentially confounded cross-sectional hete rogeneity by relying on within state variations in CON and moratorium policies over time, and by using those states that did not change policy as a control for unrelated time-series variations (Grabow ski, et al. 2003). Estimation of equation (1) is achieved by using pooled OLS which ignores the panel structure of the data and treats the obs ervations as being seri ally uncorrelated for a given nursing home with homoskedastic e rrors across nursing homes and years. The restrictive assumption of OLS when estima ting equation (1) is th at not only are the idiosyncratic error terms uncorrelated with th e explanatory variables in each time period but the unobserved effect, ci, is also uncorrelated with th e explanatory variables. The resulting estimated coefficient of will be the difference over time in the average of the outcome of interest between those states without CON and/or moratorium policies and those states that still have these policies. An alternative interpretation is that measures the change in the outcome of interest when a state eliminates its CON and/or moratorium

PAGE 96

84 policy. Based on the underlying identification of the models, the resulting coefficient is the effect of CON_MORT. 4.4.2 Facility Fixed Effects and Model Specification Quite often the point of using panel da ta is to allow the unobserved effect (ci) to be arbitrarily correlated with the xist. The second method of es timating the effect of CON and/or construction moratorium policies on qua lity of care and access to care is based on this concept. This method star ts with equation (1) written as: yist = + CON_MORTist + xist + dt + ci + it (2) The fixed effects transformation, also calle d the within transformation, is obtained by first averaging equation (2) over t = 1,…., T to get the cross section equation (3) where x T 1 x CON_MORT T 1 CON_MORT y T 1 yT 1 T 1 T 1 s t ist t is t ist is ist i T 1 T 1 T 1 and d T 1 dt it i t t. When equation (3) is subtracted from equation (2) for each t, the result is: ) ( ) d d ( ) x x ( ) MORT CON CON_MORT ( ) y y (is ist t is ist is ist is ist or it t ist ist ist d x ORT M CON_ y (4) where ) y y ( yis ist ist is the time-demeaned data on y, and similarly for, ORT M CON_ist and d x it t ist This fixed effects equation, referr ed to as the facility fixed effects i i is is is c d x CON_MORT y

PAGE 97

85 model, is estimated using the panel data fixed effects command (xtreg, fe) in Stata S/E 9. The resulting fixed effects estimator is the pooled OLS estimator from equation (4). The same results are obtained as t hose that would be returned if equation (2) was modified to include a vector of facility dummy va riables and then estimated by OLS. This model assumes strict exogeneity of the explanatory variables conditional on ci: E( it|xis,ci) = 0, t=1,…T. In other words, the explan atory variables in each time period are uncorrelated with the idio syncratic error term in each time period. The difference between the state fixed effects model and the f acility fixed effects model is that the fixed effects estimator allows for arbitrary correlation between ci and the explan atory variables in any time period. However, one drawback of this type of model is that time-constant explanatory variables cannot be included in xist. The strategy for estimating (and ) is to transform the original equation to eliminate the unobserved effect ci. However, this also results in the elimination of any other time-invariant variables. This is the reason why equation (2) does not include the vector of state dummy variables. This type of transformation is commonly referred to as the within transformation because the unobserved effect is differenced out of the equation and explo its the panel nature of the data set and relies on variati on within facilities. While equation (4) is the estimating equation, “the in terpretation of comes from the (structura l) conditional expectation E(yit|xi,ci) = E(yit|xit,ci) = xit + ci” (Wooldridge 2002). This im plies that the resulting estimated coefficient of is interpreted based on equation (2).13 This fixed effects estimator, referred to as the within estimato r, will be the difference over time in the average of the outcome of interest between t hose nursing homes that ar e in states that do 13 It is important to recognize that the interpretation of is the same for equations (1) and (2).

PAGE 98

86 not have a CON and/or moratorium policy and those nursing homes in states that have such a policy. The estimation strategy in both the state fixed effects and facility fixed effects models as described above are variants of th e difference-in-differences model. A study that uses a difference-in-differences model examines an outcome measure for observations in treatment groups and comparis on groups that are not randomly assigned. The treatment group is the one that experien ces an exogenous event, such as a policy change, and the control group is the one that does not experience the event. This method allows a researcher to examine the differ ence before and after the treatment between those in the treatment groups and those in the comparison groups for the outcome variable of interest. In this dissertation, th e outcome variable is either quality of care or access to care and the treatment is the rem oval of a state’s CON and/or construction moratorium policy. As a result of this defi nition, the treatment gr oups are states without CON and/or moratorium policies and the comparison groups are states with a CON and/or construction moratorium policy in place. Using the specification in Meyer (1995), the following equation represents the general form of the difference-in-differen ces estimation with a treatment group and a comparison group before and after the treatment: yit j = + 1dt + 1dj + dt j + zit j + it j, (5) where yit j is the outcome of individual i at time t indexed by j for the group, dt = 1 if t =1 (after treatment) and 0 otherwise, dj = 1 if j = 1 (in the treatment group) and 0 otherwise, dt j = 1 if t = 1 and j = 1 and 0 otherwise, zit j is a vector of characte ristics of the unit under study, and it j is the random error term. The resulting estimated coefficient of is the

PAGE 99

87 difference-in-differences estimator, i.e., ˆis the difference over time in the average difference of the outcome of interest between the treatment group and th e control group. For the models in this study the term CON_MORTist captures the intended effect of 1dt + 1dj + dt j from equation (5) since it represen ts whether or not the nursing home is in a state without a CON and/ or moratorium policy at time t (whether or not the nursing home is in the treatment group at time t). The resulting coefficient of represents the difference over time in the av erage of the outcome of interest between those nursing homes that are in states w ithout a CON and/or construction moratorium policy and those states still retaining such policies. One last point concerning the resulting in ference statistics must be mentioned. The “grouped” nature of the explanatory va riable of interest may have introduced heteroskedasticity and biased the estimates of the standard error. Moulton (Moulton 1990) shows that when aggregate variables are regressed on micro units (a “grouped” structure) the estimates of the standard erro rs will be biased downward. For both the state and facility fixed effects models, I use the cluster option which adjusts the standard errors using the Hubert-White robust esti mator and corrects for intra-home cluster correlations. The cluster option relaxes the assumption of independence of the observations and requires that th e observations only be indepe ndent across the clusters (in this study the individual nursing homes).

PAGE 100

88Chapter Five Research Results This chapter describes the estimation resu lts of the quality of care and access to care models. The first section of this chapter describes the summary statistics of the final data set. Sections 5.2 an d 5.3 describe the results of the effect of CON and/or construction moratorium policies on the quality of care and access to care in the nursing home market, respectively. For brevity, the partial results of the models are shown in Tables 2 through 6 in the resp ective sections while the full results of the models are included in Appendix A. Subsections 5.2.1.1, 5.2.2.1, 5.2.3.1 and 5.3.1 discuss other interesting findings while section 5.4 discusses the results of an alternative model and an alternative method of iden tifying excess demand to check for robustness. 5.1 Nursing Home, County, and Statelevel Characteristics (1991-2003) Table 1 on the following page shows the de scriptive statistics for the final data set. Between 1991 and 2003 there were 150, 705 surveys of 15,892 free-standing nursing homes in the United States. Each facility wa s surveyed an average of 9.48 times between 1991 and 2003. The mean number of surveys in each year was 11,593, with a high of 12,817 in 1996 and a low of 8,623 in 200314. The average number of nursing home 14 The lower number of surveys in 2003 is due to the timing of the 2003 OSCAR file download.

PAGE 101

89Table 1 Nursing Home, County, and State-level Characteristics (1991-2003)a Variable Description N Mean Overall SD Within SD Access Measure PCTMCAID Percent of Medicaid residents 150,705 68.94 18.59 8.45 Quality Measures PROPPRESSORE Proportion of residents w ith decubitus ulcer 150,571 0.07 0.05 0.04 PROPPARENTERAL Proportion of residents with feeding tubes 150,571 0.06 0.06 0.03 PROPCATHETER Proportion of residents with catheters 150,571 0.07 0.05 0.04 PROPMOBLREST Proportion of residents with physical restraints 150,571 0.16 0.15 0.12 DEFS1TOT Number of health deficiencies 150,670 6.81 7.24 5.81 RNS Registered nurse (RN) hours per resident day 150,705 0.34 0.31 0.19 LPNS Licensed practical nurse (LPN) hours per resident day 150,705 0.68 0.41 0.28 AIDES Nurses’ aides hours per resident day 150,705 2.14 1.01 0.73 LICNUR RN and LPN hours per re sident day 150,705 0.94 0.43 0.29 NURSTAFF RN, LPN, and aide hours pe r resident day 150,705 2.91 1.08 0.77 Facility Characteristics TOTRES Number of residents 150,705 94.17 49.69 11.50 CENMCAID Number of Medicaid residents 150,705 65.50 41.16 11.25 CENMCARE Number of Medicare residents 150,705 6.48 7.80 4.74 CENOTHER Number of private-pay residents 150,705 22.19 19.49 8.62 TOTBEDS Total number of beds 150,705 109.59 55.64 9.42 CONTROL1 =1 if for-p rofit facility (base) 150,705 0.74 0.44 0.12 CONTROL2 =1 if nonprofit f acility 150,705 0.22 0.41 0.12 CONTROL3 =1 if government facility 150,705 0.04 0.20 0.05 MULTI =1 if chain fac ility 150,705 0.56 0.50 0.24 ADLINDEX ADL index 150,571 9.80 1.52 0.89 Market (County) Characteristics POP65UP Elderly population (aged 65+)b 150,705 77,761 165,169 5,937 EMPTYELDERLY Number of empty beds per 1000 noninstitutionalized elderlyd 150,705 8.39 10.00 4.57 INCOME Per capita personal income (2003 $)c 150,705 27,280 7,114 2,193 HHI Herfindahl-Hirschman index 150,705 0.25 0.27 0.08 State-level Characteristics CON_MORT State does not have CON and/or moratorium policy (=1 if yes)e 150,705 0.15 0.36 0.14 RATE Average, annual Medicaid per diem reimbursement rate (2003 $)e 150,705 104.00 25.70 11.42 RATEMISSING =1 when New Mexico’s rate is missing 150,705 0.002 0.04 0.03 METHOD0 =1 if state uses retrospective reimbursemente 150,705 0.03 0.16 0.11 METHOD1 =1 if state uses prosp ective, facility-specific reimbursemente 150,705 0.34 0.47 0.24 METHOD2 =1 if state uses prospective, class reimbursemente 150,705 0.14 0.35 0.13 METHOD3 =1 if state uses pr ospective, adjusted reimbursemente (base) 150,705 0.43 0.49 0.20 METHOD4 =1 if states combines retrospective and prospective reimbursemente 150,705 0.06 0.25 0.12 CASEMIX =1 if state uses case-mix Medicaid reimbursemente 150,705 0.53 0.50 0.25 aThe data are from the Online, Surv ey, Certification, and Reporting (OSCAR) system unless otherwise noted. bThis variable is from the Area Resource File (ARF). cThis variable is from the Bureau of Economic Analysis’ (BEA) Regional Economic Information System (REIS). dThis variable is constructed using OSCAR and ARF files. eThese variables are from the 1998 State Data Book on Long-Term Care Program and Market Characteristics (Harrington, et al. 2000b) and the author’s state surveys.

PAGE 102

90 residents is 94, with 25 percen t private-pay residents, 69 per cent Medicaid residents, and 6 percent Medicare residents. The percentage of for-profit, nonprofit, and government nursing homes in the sample is 74 percent, 22 percent, and 4 percen t respectively. In re cent years there has been quite a bit of merger activity in the nursing home industry. In this sample 56 percent of the nursing faciliti es were owned by a chain. The average, annual per diem Medicaid reimbursement rate between 1991 and 2003, in 2003 dollars, was $104.00. With resp ect to the reimbursement method utilized to set rates, 3 percent of the sample used a retrospective method, 34 percent used a prospective, facility-specific method, 14 pe rcent used a prospec tive, class method, 43 percent used a prospec tive, adjusted method, and 6 per cent used a combination method. Additionally, 53 percent of the states in the sample adjusted for the case-mix of the facility’s residents.15 5.2 The Effect of Certificate of Need and Construction Moratorium Policies on the Quality of Care in Nursing Homes The results for the process, structure, and outcome measures of quality are discussed separately in the following sections Section 5.2.1 discusses the results of the process-based measures of quality, secti on 5.2.2 the outcome and composite-based measures, and section 5.2.3 the structure-based measures. 15 Descriptive statistics by CON_MORT are in table A.7 in Appendix A.

PAGE 103

915.2.1 Process Measures of Quality The results for each of the estimations fo r the process-based measures of quality in the state and facility fixed effects mode ls show that nursing homes in those states without a CON and/or construction mora torium policy (CON_MORT) have a lower proportion of residents that have urethral catheters (PROPCATHETER), a lower proportion of residents with feeding tubes (PROPPARENTERAL), and a lower proportion of residents with physical restra ints (PROPMOBLREST). Alternatively the resulting coefficient of CON_MORT in both m odels can be interpre ted as those nursing homes in states eliminating a CON and/or moratorium policy have an improvement in quality. As shown in Table 2 on the following pa ge, all of the results (except for the state fixed effect model for PROPPARENTERAL) are statistically significant at the 5 percent level or better. Using this type of qualit y measure, these results suggest that nursing homes in states without CON and/or moratori um policies have higher quality than those states with this type of regulation. Thes e results hold while cont rolling for the excess demand of the nursing home’s market and suppo rt the hypothesis that the elimination of CON leads to higher quality of care. 5.2.1.1 Other Findings The annual, average per diem Medicaid re imbursement rate (RATE) is negative for all models. However, the coefficient is only statistically significant for the PROPARENTERAL facility fixe d effects model and both the state and facility fixed effects for the PROPMOBLREST models. While early studies investigating the

PAGE 104

92 Table 2 Main Regression Results for Process-Based Quality Models Dependent Variable PROPCATHETER PROPPARENTERAL PROPMOBLREST State Fixed Effects Facility Fixed Effects State Fixed Effects Facility Fixed Effects State Fixed Effects Facility Fixed Effects Explanatory Variables Coefficient (Std Error)a Coefficient (Std Error)a Coefficient (Std Error)a Coefficient (Std Error)a Coefficient (Std Error)a Coefficient (Std Error)a CON_MORT -0.00957 (0.00132)*** -0.01037 (0.00133)*** -0.00096 (0.00111) -0.00194 (0.00099)** -0.05028 (0.00321)*** -0.04827 (0.00325)*** RATE -0.00002 (0.00003) -0.00002 (0.00003) -0.00004 (0.00003) -0.00004 (0.00002)* -0.00061 (0.00008)*** -0.00063 (0.00008)*** CONTROL2 -0.01008 (0.00072)*** -0.00278 (0.00111)** -0.01040 (0.00093)*** -0.00091 (0.00093) 0.00309 (0.00151)* 0.01006 (0.00324)*** CONTROL3 -0.00772 (0.00141)*** -0.00069 (0.00282) -0.00241 (0.00252) -0.00098 (0.00165) 0.01414 (0.00339)*** 0.01138 (0.00791) METHOD0 -0.00122 (0.00172) -0.00133 (0.00171) -0.00119 (0.00149) -0.00034 (0.00127) 0.01015 (0.00451)** 0.013562 (0.00457)*** METHOD1 0.00154 (0.00080)* 0.00094 (0.00077) -0.00275 (0.00067)*** -0.00274 (0.00057)*** 0.01518 (0.00183)*** 0.01701 (0.00183)*** METHOD2 -0.00370 (0.00158)** -0.00567 (0.00154)*** 0.00236 (0.00135)* 0.00121 (0.00126) 0.04824 (0.00357)*** 0.04961 (0.00360)*** METHOD4 -0.00177 (0.00132) -0.00214 (0.00129)* 0.00114 (0.00120) 0.00079 (0.00105) 0.00679 (0.00337)** 0.00772 (0.00339)** CASEMIX 0.00195 (0.00064)*** 0.00252 (0.00061)*** -0.00380 (0.00059)*** -0.00181 (0.00049)*** 0.01002 (0.00166)*** 0.01198 (0.00168)*** N 150,571 150,571 150,571 150,571 150,571 150,571 R2 0.1449 0.0535b 0.2941 0.0785b 0.2351 0.1571b Prob > F 0.0000 0.0000 0.0000 0.0000 0.0000 0.0000 aThe standard errors represent the Huber-White robust standard errors corrected for intra-home correlation using the cluster opt ion in Stata 9 S/E. Each regression contains 15892 clusters. bThis value is the R2 within. *p < .10 **p < .05 ***p < .02

PAGE 105

93 relationship between reimbursement rate and quality found that under conditions of excess demand an increase in the rate actually decreased quality (Gertler 1989, 1992; Nyman 1985, 1988b, 1989b), these results support the findings of more recent work that uses panel data and finds that an increase in the reimbursement rate actually improves quality (Cohen and Spector 1996; Grabowski 2001a, 2001b). While the absolute magnitude of the results is relatively small, the relative size of the estimates is fairly substantial. The Medicaid rate elasticity of quality implied by the estimate from the PROPMOBLREST model is -0.40 which says that a 10 percent increase in the Medicaid rate is associated with a 4 percent decrease in the proporti on of residents with physical restraints.16 Ownership appears to be a significant de terminant of process-based quality of care. The results indicate that nonpro fit nursing homes (CONTROL2) have fewer residents with catheters and f eeding tubes compared to for-profit nursing homes (the base group). However, nonprofit facilities have mo re residents with physical restraints. Government-owned nursing homes (CONTROL3 ) follow the same pattern as nonprofit facilities. Perhaps these results support the belief that sin ce nonprofit nursing homes experience higher costs and higher costs indica te higher quality (and not less efficiency) money is being spent on providing more personal care to residents rath er than the use of more mechanical methods of care. The pos itive result on PROPMOBLREST is perhaps attributed to a higher leve l of resident dependency. In the PROPCATHETER model, a prospect ive, facility-specific reimbursement method (METHOD1) leads to a higher proportion of residents with urethral catheters 16 The elasticity is calculated at the mean of the Medicaid reimbursement rate.

PAGE 106

94 compared to an adjusted method (the base group) but only in th e state fixed effects model. A prospective, class reimburse ment method (METHOD2) leads to a lower proportion in both the state and facility fixed effects models. Having a retrospective or combination method has no significant imp act on the proportion of residents with catheters compared to an adjusted method. In the PROPPARENTERAL mode l, states that use a prospective facility-specific reimbursement method (METHOD1) have a lowe r proportion of reside nts with a feeding tube compared to an adjusted method in both the state and facility fixed effects models. Having a prospective class method leads to a higher proportion of residents with feeding tubes but is only signi ficant in the state fixed eff ects model. Once again having a retrospective or combination method has no significant impact on the proportion of residents with catheters compar ed to an adjusted method. In the PROPMOBLREST model all the me thods have a statistically significant higher proportion of residents phys ically restrained than comp ared to those states using an adjusted method. These re sults indicate that all methods other than adjusted lead to lower quality. Additionally, the coefficients on CASEMIX tell us that states that use a case-mix adjustment in setting their rates have a lowe r proportion of resident s with feeding tubes but a higher proportion of residents with cathet ers and physical restraints. The results are statistically significant at the 1 percent level in each of the state and facility fixed effects models. Based on the process-based measures of quality used in this study, the evidence is not conclusive that one method of reimbursement encourages higher quality than another or that having case-mix adjust ment always improves quality of care.

PAGE 107

955.2.2 Outcome and Composite Measures of Quality The results for both the outcome and composite-based models are displayed in Table 3 on the following page. The results of the outcome-based m odel of quality show that there is no significant effect of CON and/or constr uction moratorium policy on the proportion of residents with pressure sores (PROPPRESSORE). Thes e results hold for both the state and fixed effects models. This result is not consiste nt with the hypothesis that there will be higher quality of care in those states without CON policies. One possible explanation is that the emphasis placed on the importance of the prevention and treatment of pressure sores (IOM 2001; OEI 1999; Smith 1995) l eads to more attention to this condition by nursing staff in all states. For the composite-based facility fixed effects model, nursing homes in those states that do not have a CON and/or mora torium policy have more total deficiencies (DEFS1TOT), which indicates lower quality, than those states with these policies. This result does not support the hypothesis of higher quality in nursing homes in states without CON and/or construction moratorium policies. Perhaps those states that have CON and/or moratorium policies place greater empha sis on training nursing home staff or have strong local attitudes toward the importance of maintaining quality of care in nursing homes, such as ombudsman programs, than those states without such policies. 5.2.2.1 Other Findings The annual, average per diem Medicaid reim bursement rate is negative in both the outcome and composite-based models. An incr ease in the rate d ecreases the proportion

PAGE 108

96 Table 3 Main Regression Results for Outcome and Composite-Based Quality Models Dependent Variable Outcome-Based PROPPRESSORE Composite-Based DEFSITOT State Fixed Effects Facility Fixed Effects State Fixed Effects Facility Fixed Effects Explanatory Variables Coefficient (Std Error)a Coefficient (Std Error)a Coefficient (Std Error)a Coefficient (Std Error)a CON_MORT 0.00081 (0.00109) 0.00032 (0.00104) 0.11049 (0.15330) 0.34217 (0.14785)** RATE -0.00018 (0.00002)*** -0.00018 (0.00002)*** -0.00801 (0.00370)** -0.11655 (0.00366)* CONTROL2 -0.00752 (0.00048)*** 0.00036 (0.00106) -1.33765 (0.06161)*** 0.10461 (0.16863) CONTROL3 -0.00701 (0.00102)*** 0.00244 (0.00201) -1.46497 (0.11455)*** -0.74093 (0.44590)* METHOD0 -0.00213 (0.00144) -0.00191 (0.00143) -0.34340 (0.19223)* -0.09740 (0.19073) METHOD1 9.75e-06 (0.00073) 0.00015 (0.00072) 1.04717 (0.11105)*** 1.18294 (0.11003)*** METHOD2 -0.00112 (0.00123) -0.00165 (0.00123) 1.11955 (0.17030)*** 1.32699 (0.17023)*** METHOD4 -0.00085 (0.00110) -0.00094 (0.00111) -0.01786 (0.16576) 0.01779 (0.16702) CASEMIX 0.00008 (0.00058) 0.00090 (0.00057) 0.46302 (0.08352)*** 0.49594 (0.08272)*** N 150,571 150,571 150,537 150,537 R2 0.1209 0.0142b 0.1830 0.0441b Prob > F 0.0000 0.0000 0.0000 0.0000 of residents with pressure sores and the to tal number of facility deficiencies. This negative relationship is statistically significant at the 10 percent level or better in each of the state and facility fixed effects models. These results provide further evidence that increasing the reimbursement rate improves qua lity of care in nursing homes. Once again aThe standard errors represent the Hube r-White robust standard errors correct ed for intra-home correlation using the cluster option in Stata 9 S/E. Each regression contains 15892 clusters. *p < .10 **p < .05 ***p < .01 bThis value is the R2 within.

PAGE 109

97 while the absolute magnitude of the results is relatively small, the relative size of the estimates is significant. The Medicaid rate elasticity of quality implied by the estimate from the PROPPRESSORE model is -0.29 wh ile the elasticity for the DEFS1TOT model is -.12. Nonprofit and government-owned nursing homes have a lower proportion of residents with pressure sores than for-profit facilities. The results of the state fixed effects model are significant at the 1 percent level. The coefficients in the facility fixed effects model are positive but not statistic ally significant. Nonprofit and governmentowned facilities also have fewer deficiencies in the state fixed effects model. Only the coefficient for government-owned facilities is negative and signifi cant in the facility fixed effects model. Once again these re sults indicate that n onprofit and governmentowned nursing homes provide higher quali ty than for-profit nursing homes. Unlike the process-based models, in th e PROPPRESSORE model the type of reimbursement method and whether or not a state uses case-mix adjustment are not statistically significant. However, in th e DEFS1TOT model three of the reimbursement methods and CASEMIX are significant. Use of a retrospective method results in fewer deficiencies in the state fi xed effects model and is stat istically significant at the 10 percent level. The relationship is also negativ e in the facility fixed effects model but is not statistically significant. Both prospect ive, facility-specific and class reimbursement methods lead to more deficiencies than havi ng an adjusted reimbursement method in both the state and fixed effects models. The co efficient on the combination method is not significant in either the state or facility fixed effects model.

PAGE 110

98 The coefficient on CASEMIX is positive a nd statistically significant at the 1 percent level in both the stat e and fixed effects models. Since more costs are covered when a facility is reimbursed based on a re sident’s level of seve rity, one would expect fewer residents with pressure sores and fewer facility deficiencies. While not expected, perhaps this result can be explained by the baseline level of residents’ dependencies. Nursing homes in states without CON and/ or construction moratorium polices have residents with a higher level of dependency as measured by ADLINDEX (see Table A.7). Due to the increased care needs of their mo re functionally disabled population, perhaps these homes face an increased opportunity for be ing cite for a violation and this in turn causes more deficiencies. 5.2.3 Structure Measures of Quality The five structure-based measures of qua lity used in this study, RNS, LPNS, AIDES, LICNUR, and NURSTAFF, are all m easured in hours per resident day and represent “good” quality. As seen in Tables 4 and 5 on the following pages, for each of the models in which CON_MORT is statisti cally significant a pattern emerges among the five measures of structure-based quality; nur sing homes in those states without a CON and/or construction moratorium policy have mo re registered and licensed practical nurse hours per resident day and less ai de hours per resident day than those states that still have a policy. These results suppor t the hypothesis that those states without CON policies have a higher level of quality. These results indicate a possible substitution effect is occurring. For those states continuing with these supply restrictions, while maintaining the minimum requirement of RNs and LPNs, th e presence of a CON restriction allows

PAGE 111

99 Table 4 Main Regression Results for St ructure-Based Quality Models Dependent Variable RNS LPNS AIDES State Fixed Effects Facility Fixed Effects State Fixed Effects Facility Fixed Effects State Fixed Effects Facility Fixed Effects Explanatory Variables Coefficient (Std Error)a Coefficient (Std Error)a Coefficient (Std Error)a Coefficient (Std Error)a Coefficient (Std Error)a Coefficient (Std Error)a CON_MORT 0.00638 (0.00548) 0.00998 (0.00489)* 0.02957 (0.00844)*** 0.03114 (0.00705)*** -0.05294 (0.01833)*** -0.04119 (0.01646)** RATE 0.00028 (0.00015)* 0.00050 (0.00013)*** 0.00081 (0.00020)*** 0.00093 (0.00018)*** -0.00203 (0.00051)*** -0.00200 (0.00049)*** CONTROL2 0.05902 (0.00472)*** -0.00428 (0.00481) 0.02486 (0.00587)*** 0.00214 (0.00703) 0.24987 (0.01392)*** -0.00376 (0.01855) CONTROL3 0.03751 (0.00909)*** 0.00289 (0.00805) 0.02317 (0.01017)** -0.00374 (0.01696) 0.25495 (0.02299)*** 0.07022 (0.04590) METHOD0 0.00055 (0.00915) 0.00417 (0.00790) 0.00711 (0.01065) 0.00382 (0.00952) -0.21510 (0.02744)*** -0.20872 (0.02595)*** METHOD1 0.01893 (0.00407)*** 0.01398 (0.00348)*** -0.00109 (0.00506) 0.00132 (0.00457) -0.03343 (0.01333)*** -0.03751 (0.01214)*** METHOD2 0.02780 (0.00548)*** 0.02154 (0.00503)*** 0.03998 (0.01221)*** 0.01061 (0.00933) -0.05716 (0.02830)** -0.12369 (0.02445)*** METHOD4 0.00961 (0.00664) 0.00950 (0.00572)* -0.00171 (0.00865) -0.00374 (0.00777) -0.04944 (0.02158)*** -0.05187 (0.01879)*** CASEMIX 0.00019 (0.00391) -0.00016 (0.00357) -0.01649 (0.00453)*** -0.01119 (0.00427)*** -0.08810 (0.01114)*** -0.07188 (0.01076)*** N 150,571 150,571 150,571 150,571 150,571 150,571 R2 0.1954 0.0130b 0.1266 0.0273b 0.0944 0.0105b Prob > F 0.0000 0.0000 0.0000 0.0000 0.0000 0.0000 aThe standard errors represent the Huber-White robust standard errors corrected for intra-home correlation using the cluster opt ion in Stata 9 S/E. Each regression contains 15892 clusters. bThis value is the R2 within. *p < .10 **p < .05 ***p < .01

PAGE 112

100 Table 5 Main Regression Results for St ructure-Based Quality Models Dependent Variable LICNUR NURSTAFF State Fixed Effects Facility Fixed Effects State Fixed Effects Facility Fixed Effects Explanatory Variables Coefficient (Std Error)a Coefficient (Std Error)a Coefficient (Std Error)a Coefficient (Std Error)a CON_MORT 0.03481 (0.00933)*** 0.03915 (0.00788)*** -0.00769 (0.02114) 0.00704 (0.01845) RATE 0.00111 (0.00021)*** 0.00134 (0.00019)*** -0.00066 (0.00055) -0.00052 (0.00052) CONTROL2 0.08555 (0.00639)*** 0.00402 (0.00748) 0.34693 (0.01591)*** 0.00929 (0.02027) CONTROL3 0.08903 (0.01234)*** 0.01347 (0.01682) 0.39290 (0.02981)*** 0.10864 (0.05227)** METHOD0 -0.00234 (0.01300) 0.00311 (0.01167) -0.22118 (0.03186)*** -0.20223 (0.02967)*** METHOD1 0.00896 (0.00554) 0.00934 (0.00506)* -0.03159 (0.01429)** -0.03033 (0.01311)** METHOD2 0.04853 (0.01068)*** 0.02981 (0.00626)*** -0.02749 (0.02894) -0.08105 (0.02468)*** METHOD4 0.00476 (0.00962) 0.00582 (0.00880) -0.03836 (0.02250)* -0.03369 (0.02013)* CASEMIX -0.01297 (0.00288)*** -0.00810 (0.00149)*** -0.08851 (0.01241)*** -0.06977 (0.01193)*** N 150,571 150,571 150,571 150,571 R2 0.1127 0.0174b 0.1193 0.0109b Prob > F 0.0000 0.0000 0.0000 0.0000 facilities in these states to hire less costly inputs without sacrif icing market share. aThe standard errors represent the H uber-White robust standard errors corre cted for intra-home correlation using the cluster option in Stata 9 S/E. Each regression contains 15892 clusters. *p < .10 **p < .05 ***p < .01 bThis value is the R2 within.

PAGE 113

101 5.2.3.1 Other Findings The coefficient of the average, annual Medicaid per diem reimbursement rate is positive and statistically significant for the numbe r of registered nurse, licensed practical nurse, and licensed nurse hours per resident day indicating that an increase in the reimbursement rate provides high er quality. Again it appears that an increase in the rate causes a substitution effect since the coefficient for the AIDES model is negative and statistically significant at the 1 percent le vel; facilities are able to provide a more qualified level of staffing, which costs more, when there is an increase in the payment rate. The results of ownership on the structur e-based measures of quality are very consistent; nonprofit and government-owned faci lities provide a higher level of staffing than for-profit faciliti es. These results support the findings of earlier work (Aaronson, et al. 1994; Elwell 1984; Grabowski and Hirth 200 2; Zinn 1994) and lend credence to the theory that nonprofit facilit ies have higher costs because they have higher quality. The use of a retrospective reimbursement method is negative and statistically significant in the AIDES and NURSTAFF models If costs are pa ssed through to the insurer as is the case with retrospective re imbursement, then having fewer nurses’ aides which cost less than certified nurses implies that facilities are again substituting toward more highly trained and costly inputs. However, even though all of the coefficients of the other three models are positive, none are statistically significant. Both prospective, facility-specific and prospective class reimbursement methods lead to more RN hours and licensed nurse hours (LICNUR) and less aides and nursing staff (NURSTAFF) hours. Additionally, a clas s method results in more LPN hours. The

PAGE 114

102 coefficient on the dummy variable for a combination method (METHOD4) is positive and statistically significant in the facility fixed effects model for RNS and is negative and statistically significant for both the AIDES a nd NURSTAFF models in both the state and fixed effects models. Taken together the re sults of the reimburse ment method variables do not give a clear picture as to which met hod of reimbursement pr ovides higher quality of care. The use of case-mix reimbursement provides rather interesting results. Assuming that compensating for the case-mix of the re sident population should improve the quality of care because the costs of such care are ac counted for, it is surprising to find in the models for which the value is statistically significant that the coefficient is negative rather than positive. However these results are si milar to ones obtained in a panel study with data from 1991 to 1998 using state fixed effects (Grabowski 2002; 2004). Perhaps the increased rate that facilities receive for the case-mix of the residents is spent on other amenities rather than providing a more professional level of nursing staff. 5.3 The Effect of Certificate of Need and Construction Moratorium Policies on the Access to Care in Nursing Homes The results of the estimation for both the state and facility fixed effects models show that nursing homes in those states without CON and/or construction moratorium policies have a statistically significant higher percentage of Medicaid residents. As shown in Table 6 on the following page, the coefficient on CON_MORT is significant at less than the 1 percent level for both models. In the stat e fixed effects model, nursing

PAGE 115

103 Table 6 Main Regression Results for Access and Heavy-Care Models Dependent Variable PCTMCAID ADLINDEX State Fixed Effects Facility Fixed Effects State Fixed Effects Facility Fixed Effects Explanatory Variables Coefficient (Std Error)a Coefficient (Std Error)a Coefficient (Std Error)a Coefficient (Std Error)a CON_MORT 0.95788 (0.34786)*** 1.55143 (0.29731)*** -0.12390 (0.02821)*** -0.10086 (0.02636)*** RATE -0.00507 (0.00777) 0.00302 (0.00616) 0.00109 (0.00069) 0.00081 (0.00064) CONTROL2 -9.38626 (0.34208)*** -0.91782 (0.27432)*** 0.06312 (0.02116)*** 0.02638 (0.02532) CONTROL3 0.79328 (0.60256) 0.92467 (0.65133) 0.32173 (0.04918)*** 0.08470 (0.06216) METHOD0 -1.91870 (0.48414)*** -1.97860 (0.40190)*** 0.04236 (0.04044) 0.02111 (0.03785) METHOD1 -0.22563 (0.23044) -0.40020 (0.18285)** 0.06795 (0.01859)*** 0.03508 (0.01670)** METHOD2 -1.40048 (0.37974)*** -0.78225 (0.31524)** -0.04475 (0.03208) -0.05763 (0.03013)* METHOD4 0.21310 (0.36760) 0.62293 (0.29102)** -0.11659 (0.03047)*** -0.09409 (0.02806)*** CASEMIX -0.08600 (0.17625) -0.09821 (0.14293) 0.20506 (0.01521)*** 0.20339 (0.01400)*** N 150,571 150,571 150,571 150,571 R2 0.2016 0.0060b 0.2424 0.0556b Prob > F 0.0000 0.0000 0.0000 0.0000 homes in states without CON and/or constr uction moratorium polic ies have .96 percent more Medicaid residents than those nursing homes in states with some type of regulation. This percentage increases to 1.55 percent in the facility fixed effects model. These results hold while controlling for excess dema nd in the market and support the hypothesis aThe standard errors represent the Huber-White robust standard errors corrected for intra-home correlation using the cluster option in Stata 9 S/E. Each regression contains 15892 clusters. *p < .10 **p < .05 ***p < .01 bThis value is the R2 within.

PAGE 116

104 that Medicaid-eligible indivi duals in nursing homes in thos e states without CON and/or moratorium policies have better access to nur sing home beds than those individuals in states that still have these policies. While th e magnitude of these results is not large the results do indicate that there is a difference in the access to care for Medicaid-eligible individuals between those stat es without CON and/or moratorium policies and those with these policies. With respect to the access to care for heavycare residents in a f acility, facilities in those states without CON and/or moratorium policies actually have residents with a lower level of dependency in both the state and facility fixed effect s models. Medicaid residents are often believed to be more de pendent in their functional abilities than private-pay residents and theref ore are not the type of resi dents that nursing homes want to have fill their empty beds. This result in conjunction with the result of the access model indicates that Medicaid residents are not necessarily more dependent than privatepay residents. Or perhaps the facilities that are in states without these polices are “creamskimming.” 5.3.1 Other Findings Results of several of the other explanatory variables are worth mentioning. The annual, average per diem Medicaid reimbursement rate was not significant in either of the state or facility fixed effects models for the access to care (PCTMCAID) or heavy-care (ADLINDEX) models. While earlier work ha s shown that increasi ng the reimbursement rate has increased the access to care for Medicaid residents (Aaronson, et al. 1994; Gertler 1989, 1992) these results do not suppor t the same finding. Additionally, the

PAGE 117

105 results do not seem to support the theory that the Medicaid reimbursement rate is “too low” to provide nursing homes with the incentive to admit Medicaid residents. Furthermore, these results do not show th at increasing the reimbursement rate will improve access for heavy-care residents. In both the state and facility fixed eff ects models nonprofit nursing homes have a significantly lower percentage of Medicaid residents than for-pr ofit facilities. This result supports the theory that to the extent hi gher costs and quality are associated with increased demand by private-pay and Medicare residents, nonprofit facilities are more oriented toward those residents while for-pro fit facilities are more oriented toward Medicaid residents. It is relevant to also no te that the effect of inco me on access is statistically significant in both models. However, it is ne gative in the state fixed effects model and positive in the facility fixed effects model. While one expects that an increase in the per capita personal income in the county in whic h the nursing homes is located would lead to fewer Medicaid residents, perhaps this re sult helps to explai n why the effect of CON_MORT is rather small. With respect to the heavy-care model, nonprofit and government -owned facilities have residents with a statistically signifi cant higher level of dependency than for-profit facilities in the state fixed effects model. While the coefficients are positive for both nonprofit and government ownership in the fac ility fixed effects model they are not statistically significant. This result supports the idea that one of the reasons nonprofit and government nursing homes have higher costs th an for-profit faciliti es is because they serve a more impaired resident population.

PAGE 118

106 The effect of the reimbursement method in the access model was fairly consistent across the various measures for both the state an d facility fixed effects models. For both models, nursing homes in states that employed either a retrospective or prospective class reimbursement method had a statistically si gnificant lower percentage of Medicaid residents compared to those st ates that employed an adjusted method. Since retrospective reimbursement is less restrictiv e than an adjusted method (the base group) the result for retrospective is not of the e xpected sign. However the fact that those states with a prospective class method have a lower percenta ge of Medicaid residents is expected since that method is the most restrictive of all. Wh ile only statistically sign ificant in the facility fixed effects model, the negative sign on th e coefficient for the prospective facilityspecific method is also expected since this method is more stringent than an adjusted method. Lastly, while only statistically signif icant in the facility fixed effects model, the positive sign on the coefficient for the combin ation method is also expected since this method is also more stringent than an adju sted method. Finally, having a method that accounts for resident case-mix does not appear to have any effect on the percentage of Medicaid residents in a faci lity. This is the same result found by Cohen and Dubay (1990). With respect to the heavy-care model, one would expect that the more stringent the reimbursement method the less functiona lly dependent is th e resident population since caring for more disabled residents implie s higher costs for the facility. In contrast one would expect to see a more disabled re sident population with a retrospective method since it is the least restrictive payment method. As expected, the coefficient on the retrospective method is positive in both the state and facility fi xed effects model but

PAGE 119

107 neither value is statistically significant. As expected, the coefficient on METHOD2 is negative in both models but is only statistical ly significant tin the f acility fixed effects model. Again as expected, the coefficients on METHOD4 are negative and statistically significant in both the state and facility fi xed effects model indicating that states employing a combination method have residents th at have less functional disabilities than those residents in states usi ng an adjusted method. Finally, in both the state and fixed effects models those states that use a casemix adjustment have residents with more functional disabilities. This could be an important polic y implication since one could interpret the results to mean that nursing ho mes reimbursed in this way are provided with an incentive to accept he avy-care residents. 5.4 Robustness Checks In order to check for the robustness of the main findings the models were estimated without the reimbur sement method to alleviat e any concern over potential multicollinearity between the state fixed e ffects and the reimbursement method. The results of these models are consistent with the results of the main estimations for the main variable of interest, the effect of CON a nd/or construction moratorium policies on access to care and quality of care. Specifically, a ll the coefficients of CON_MORT are of the same sign and the same level or better of statistical significance in the quality, access, and heavy-care state and fixed effects models. The full results of these estimations are in Tables A.18-A.27 in Appendix A. An alternative method to control for excess demand is to separate the data into the least and most restrictive markets. The first step was to assign a unique identifier to each

PAGE 120

108 ARF state and county combination. Then the median of EMPTYELDERLY (the number of empty nursing homes beds per 1,000 noni nstitutionalized elderly (aged 65+) in a county) was calculated for each county over the period 1991-2003. Next the annual median level for the entire data set was cal culated for the entire period of 1991 through 2003. Those observations that were below the median value of EMPTYELDERLY for the entire data set were classified as the mo st restrictive markets while those observations above the median were classified as the least restrictive markets. This follows a similar method used in recent work to proxy for excess demand (Grabowski 2002, 2004). The results for the effect of CON_MORT in both the least and most restrictive markets show that nursing homes in those states without CON and/or construction moratorium policies have a higher percentage of Medicaid residents. The coefficients are statistically significant in all of the models except for the least restrictive state fixed effects model. In the most restrictive markets nursing homes in states without CON and/or moratorium policies have 2 percent a nd 3 percent more Medicaid residents in the state and facility fixed effects models, respectively. This suggests that regardless of the tightness of the market, CON policies may restri ct access to care for Medicaid residents. The results for the heavy-car e access model are the same as the results using the full sample. It seems logical that those nursing homes in least restrictive ma rkets would be apt to compete for residents on the basis of quality for both private-pay and Medicaid residents while there may not be such an incentive in the most restrictive markets. However, the results of the quality models in both the least and most restrictive markets are quite similar to those in the main findi ngs using the full sample. The two notable

PAGE 121

109 exceptions are that in the most restrictiv e markets the proportion of residents with catheters is higher and the number of defici encies is lower in nursing homes in those states without CON and/or construction morato rium policies. The full results of these estimations are in Tables A.28-A.47 in Appendix A.

PAGE 122

110Chapter Six Conclusions This chapter summarizes the main findings of the effects of certif icate of need and construction moratorium policies on quality of care and access to care in the nursing home industry and discusses their potentia l policy implications. Additionally, the limitations of this current research as we ll as opportunities for fu ture research are discussed. 6.1 Main Findings This dissertation is the first attempt to employ panel data to analyze the differences in the quality of care and access to care in nursing homes in those states without certificate of need and/or construc tion moratorium policies and nursing homes in those states still retaining such policies. This data set include s observations for all Medicaid and Medicare-certified freestanding nursing homes in the United States over the thirteen year period of 1991-2003. While controlling for excess demand in the market, the results of the quality of care estimations in this dissertation show th at nursing homes in those states without CON and/or construction moratorium policies have higher quality of care than nursing homes in those states without these policies when qua lity is a process-based measure. Nursing

PAGE 123

111 homes in states without these types of polic y have residents with a lower proportion of residents with urethral catheters, feeding t ubes, and physical restra ints. These results support the hypothesis that the elimination of CON leads to higher quality of care. Additionally, the results of the estimations us ing structure-based measures of quality also support this hypothesis. Nursing homes in states without CON policies have more registered and licensed practic al nurse hours per resident day while substituting away from aide hours per resident day. The composite-based measure of quality, facility deficiencies, was positive but only statistically significant in the facility fixed effects model. This result does not support the hypothesis of higher quality of car e in nursing homes in states without CON policies. The only outcome measure of quality us ed in this dissertation, the proportion of residents with pressure sores, was not statistical ly significant in either the state or facility fixed effects models. Unfortunately the data used in this dissertation for the years prior to 1997 from OSCAR do not include ma ny other outcome measures of quality. This is one area of research where perhaps the Minimum Data Set, which provides a comprehensive assessment of each resident's functional cap abilities in each Medi care or Medicaid certified nursing home, will be beneficial. The results of the estimation for both the st ate and facility fixed effects access to care models show that nursing homes in t hose states without CON and/or construction moratorium policies have a statistically significant higher percentage of Medicaid residents. These results hold while contro lling for excess demand in the market and support the hypothesis that Medicaid-eligible individuals in those states without CON and/or moratorium policies have better access to nursing home beds than those

PAGE 124

112 individuals in states that stil l have these policies. While it may be true that the occupancy rates of nursing homes have declined over recen t years, it appears that Medicaid-eligible individuals still face access re strictions in those states having CON and/or moratorium policies. With respect to the access to care for heavycare residents, facilities in those states without CON and/or moratorium policies actua lly have residents with a lower level of dependency in both the state and facility fixe d effects models. Medicaid residents are often believed to be more dependent in their functional abilities than private-pay residents and therefore are not the type of residents that nursing homes want to have fill their empty beds. This result in conjunction with the result of the access model indicates that Medicaid residents are not necessarily more dependent than private-pay residents. Another explanation is that facilities which ar e in states without th ese policies are better able to “cream-skim.” 6.2 Policy Implications While the primary goals of the Natio nal Health Planning and Resources Development Act (P.L. 93-641) of 1974 were to (1) contain health care costs and (2) increase the accessibility and quali ty of health services, it is not evident that today those goals are being achieved through the use of certificate of need and construction moratorium policies in the nursing home industr y. The results of this research with respect to quality of care and access to care, coupled with recent work on the effect of these policies on the cost of care in nursing homes (Grabowski, et al. 2003; Miller, et al.

PAGE 125

113 2002; Miller, et al. 2001), indica te that perhaps this type of supply regulation is no longer meeting its original purposes. While it is true that there has been a significant decline in nursing home utilization over the last tw o decades (Bishop 1999), it is al so true that the baby boomer generation is beginning to retir e resulting in an aging populati on that will need long term care. With the risk of becoming a nursing home resident at the age of 65 at 44 percent and at the age of 85 at 53 pe rcent (Spillman and Lubitz 2002 ) perhaps those states that still have these policies in place should consid er the costs and benefits of retaining such regulations. It is also true that quality of care continues to be a major concern with continued reports of nursing homes with seri ous deficiencies as well as survey and oversight shortcomings (GAO2005). From the re sults of this research on the effects of CON and/or moratorium policies on quality of care and access to care in the nursing home market, it is no longer clear that the be nefits of these policie s outweigh the costs. Substitutes to long term nursing home care ex ist in the market today that did not exist two decades ago. Medica id now provides for alternatives such as home health care and assisted living facilities. If it is ind eed true that changing market conditions have rendered CON and construction moratorium po licies less important in the market for nursing home care, one might then question th e relevancy of states maintaining these types of regulation. Perhaps the costs a ssociated with maintaining CON regulation agencies could be better spent on other aspect s of ensuring the dignity and quality of care for nursing home residents ra ther than on the administra tion of a policy that may no longer be relevant in toda y’s nursing home market.

PAGE 126

1146.3 Limitations There are several limitations to this research that should be noted. The first is that while the OSCAR system is one of the most comprehensive, longitudinal information sets available for all Medicare-/Medicaid-ce rtified nursing homes in the United States, it must be recognized that it was created fo r the purpose of provider certification. However, much research in th e nursing home industry has ut ilized this data since its implementation. Due to the acknowledgment of the usefulness of such data and recognizing the importance of the accuracy and reliability of survey data used in research, emphasis has been placed on state survey pr ocedures to ensure both the accuracy and reliability in the survey and data input proces ses. However, concerns still remain today with the reliability of the findings of nur sing home surveys (GAO2005). Therefore it is important to realize that the results obtained from using data from a system such as OSCAR must be viewed with the ac knowledgment of the data limitations. The second limitation is the use of a bina ry indicator for the main variable of interest, CON_MORT, which does not capture the complexity or effectiveness of a state’s policy. Although these policies theoretically restrict or prohibit growth in the nursing home market it is not always the case in practice. Many states have exceptions for their moratorium policy that allow for add itional beds or expansion if it is deemed a critical need. Additionally, my survey of st ates’ regulatory agencies indicated that states vary on the restrictiveness of their CON policies. In some states the CON actually acts as a moratorium while in other states it appears that the CON is simply a formality and most applications for additional or new beds get approved.

PAGE 127

115 The third limitation is the use of only one outcome measure of quality, the proportion of residents with pressure sores. Since the results using this measure of quality are not significant in e ither the state or facility fixed effects models, without another outcome measure to te st it is not readily evident whether or not there is no significant difference between states with a nd without CON and/or moratorium policies or whether there is a lack of precision in the estimation. The final limitation to mention concerns the reimbursement rate. While other work accounts for the fact that the rate utilized by some states covers various ancillaries (Harrington, et al. 2000b; Swan, et al. 1993) I did not co llect this information when I surveyed each of the states’ Medicaid reimbursement agencies. 6.4 Future Research Future research should address several of the limitations mentioned above. First, in order to eliminate any concern with th e use of a single binary indicator for the existence of a CON and/or construction morato rium policy, future work should include a measure of the effectiveness of the policy. Although these policies th eoretically restrict or prohibit growth in the nursi ng home market it is not always the case in practice. Many states have exceptions for their moratorium policy that allows for additional beds or expansion if it is deemed a cr itical need. My survey of states’ regulatory agencies indicates that states vary on the restrictivene ss of their CON policies. In some states the CON actually acts as a moratorium while in othe r states it appears that the CON is simply a formality and most applications for additional or new beds get approved. This

PAGE 128

116 difference in the effectiveness of the regulation should be account ed for in future research to try and eliminate any omitted variable bias that might influence the results. Second, follow on work from this disser tation should include an attempt to use other outcome-based measures of quality. While it may not be possible to achieve this with OSCAR data that cover the same time peri od as in this study, it should be possible to construct a panel with data from th e mid-1990s through 2005. Since reports of inadequate quality of care continue to be published it is important to look at more outcome-based measures that are more focused directly on residents’ “quality of life” (Nyman 1988c) along with their quality of care. Perhaps the use of the Minimum Data Set (MDS) will be an additional source of data to contribute toward this need to provide better outcome-based measures of quality at the resident level. Third, future work with this data should investigate the rela tionship of ownership and quality of care as well as access to care. Though not the ma in variable of interest in this dissertation, the results indicate that nonprofit nursi ng homes provide higher quality of care (for most of the quali ty measures) than for-profit hom es and that nonprofits have fewer Medicaid residents than for-profit facili ties but more heavy-care residents. An interesting model to inves tigate would be a matching of nonprofit and for-profit nursing homes in states with and wit hout CON and/or moratorium policies to see whether or not any behavioral patterns with resp ect to quality and access emerge. While much work has been done on th e quality of care in the nursing home industry and some work on the access to care, future work must continue and attempt to clarify for policymakers those policies that are effective in assuring adequate care to our aging population and those that may not be achieving thei r intended goals.

PAGE 129

117 References Aaronson, W. E., Zinn, J. S., and Rosko, M. D. 1994. "Do For-Profit and Not-for-Profit Nursing Homes Behave Differently?" The Gerontologist, 34:6, pp. 775-786. Allman, R. M. 1989. "Epidemiology of Pre ssure Sores in Different Populations." Decubitus, 2, pp. 30-33. American Health Care Association (AHCA). 2005. June 2005 Oscar Report. http://www.ahca.org/research/oscar_oper.htm Accessed: December 16, 2005. Arrow, K. J. 1963. "Uncertainty and th e Welfare Economics of Medical Care." The American Economic Review, 13:5, pp. 941-973. Banaszak-Holl, J., Zinn, J. S., and Mor, V. 1996. "The Impact of Market and Organizational Characterist ics on Nursing Care Facil ity Service Innovation: A Resource Dependency Perspective." Health Services Research, 31:1, pp. 91-117. Birnbaum, H., Bishop, C. E., Lee, J. A., and Jensen, G. 1981. "Why Do Nursing Home Costs Vary? The Determinants of Nursing Home Costs." Medical Care, XIX:11, pp. 1095-1107. Bishop, C. E. 1980. "Nursing Home Cost Studies and Reimbursement Issues." Health Care Financing Review, 1:4, pp. 47-64. ---. 1988. "Competition in the Market for Nursing Home Care." Journal of Health Politics, Policy and Law, 13:2, pp. 341-360. ---. 1999. "Where Are The Missing Elders? Th e Decline In Nursing Home Use, 1985 and 1995." Health Affairs, 18:4, pp. 146-155.

PAGE 130

118 Bliesmer, M. M., Smayling, M., Kane, R. L., and Shannon, I. 1998. "The Relationship Between Nursing Staffing Levels and Nursing Home Outcomes." Journal of Aging and Health, 10:3, pp. 351-371. Blumstein, J. F. and Sloan, F. A. 197 8. "Health Planning and Regulation Through Certificate of Need: An Overview." Utah Law Review, 3:1, pp. 3-38. Brandeis, G. H., Morris, J. N., Nash, D. J ., and Lipstiz, L. A. 1990. "The Epidemiology and Natural History of Pressure Ulcers in Elderly Nursing Home Residents." Journal of the American Medical Association, 264:22, pp. 2905-2909. Braun, B. I. 1991. "The Effect of Nursing Home Quality on Patient Outcome." Journal of the American Geriatrics Society, 39:4, pp. 329-338. Castle, N. G. and Mor, V. 1998. "Physical Rest raints in Nursing Homes: A Review of the Literature Since the Nursing Home Reform Act of 1987." Medical Care Research and Review, 55:2, pp. 139-170. Centers for Medicare and Medi caid Services (CMS). 2001. National Health Expenditures By Type of Service and Source of Funds: Calendar Years 1960-2001. http://cms.hhs.gov/statistics/nhe/#business Accessed: October 4, 2003. ---. 2003. Nursing Home Care Expenditures Ag gregate and per Capita Amounts and Percent Distribution, by Source of F unds: Selected years 1980-2003, Table 7. http://www.cms.hhs.gov/NationalHealth ExpendData/downloads/nhetables.pdf Accessed: December 16, 2005. Chou, S. 2001. "Asymmetric information, ow nership and quality of care: an empirical analysis of nursing homes." Journal of Health Economics, 21, pp. 293-311. Cohen, J. W. and Dubay, L. C. 1990. "The Effects of Medicaid Reimbursement Method and Ownership on Nursing Home Co sts, Case Mix, and Staffing." Inquiry, 27, pp. 183-200. Cohen, J. W. and Spector, W. D. 1996. "T he Effect of Medicaid Reimbursement on Quality of Care in Nursing Homes." Journal of Health Economics, 15, pp. 23-48.

PAGE 131

119 Cotterill, P. G. 1983. "Provider Incentives Under Alternative Reimbursement Systems," in Long-Term Care: Perspectives fr om Research and Demonstrations. Vogel, R. J. and Palmer, H. C. eds. Washington, D.C.: U.S. Department of Health and Human Services, Health Care Fina ncing Administration, pp. 625-664. Davis, M. A. 1991. "On Nursing Home Quality: A Review and Analysis." Medical Care Review, 48:2, pp. 129-166. ---. 1993. "Nursing Home Ownership Revi sited: Market, Cost and Quality Relationships." Medical Care, 31:11, pp. 1062-1068. Davis, M. A. and Freeman, J. W. 1994. "Excess Demand and Cost Relationships Among Kentucky Nursing Homes." Health Care Financing Review, 15:4, pp. 137-152. Donabedian, A. 1966. "Evaluating The Quality Of Medical Care." Milbank Memorial Fund Quarterly/Health and Society, 44, pp. 166-206. ---. 1988. "The Quality of Care: How Can It Be Assessed?" Journal of the American Medical Association, 260:12, pp. 1743-1748. Dranove, D. and Satterthwaite, M. A. 2000. "T he Industrial Organization of Health Care Markets," in Handbook of Health Economics. Culyer, A. J. and Newhouse, J. P. eds. Amsterdam: Elsevier Science B.V., pp. 1093-1139. Dranove, D., Shanley, M., and Simon, C. 1992. "Is Hospital Competition Wasteful?" The RAND Journal of Economics, 23:2, pp. 247-262. Elwell, F. 1984. "The Effects of Ownership on Institutional Services." The Gerontologist, 24:1, pp. 77-83. Ettner, S. L. 1993. "Do elderly Medicaid pa tients experience reduced access to nursing home care?" Journal of Health Economics, 11, pp. 259-280. Feder, J. and Scanlon, W. J. 1980. "Regulating the Bed Supply in Nursing Homes." Milbank Memorial Fund Quarterly/Health and Society, 58:1, pp. 54-88.

PAGE 132

120 Fottler, M. D., Smith, H. L., and James, W. L. 1981. "Profits and Patient Care Quality in Nursing Homes: Are They Compatible?" The Gerontologist, 21:5, pp. 532-538. Frech III, H. E. 1985. "The Property Rights Th eory of the Firm: Some Evidence from the U.S. Nursing Home Industry." Zeitschrift fur die gesamte Staatswissenschaft (ZgS) ((Journal of Institutional and Theore tical Economics)), 141:1, pp. 146-166. Frech III, H. E. and Ginsburg, P. B. 1981. "The Cost of Nursing Home Care in the United States: Government Financing, Ownership, and Efficiency," in Health, Economics, and Health Economics. van der Gaag, J. and Perlman, M. eds. Amsterdam: North-Holland Publishing Company, pp. 67-81. Fries, B. E. 1990. "Comparing case-mix systems for nursing home payment." Health Care Financing Review, 11:4, pp. 103-119. Fries, B. E., Schneider, D. P., Foley, W. J ., Gavazzi, M., MBA, Burke, R., and Cornelius, E. 1994. "Refining a Case-Mix Measure fo r Nursing Homes: Resource Utilization Groups (RUG-III)." Medical Care, 32:7, pp. 668-685. Galindo-Ciocon, P., RN. 1993. "Tube Feedi ng: Complications Among the Elderly." Journal of Gerontological Nursing, 19:6, pp. 17-22. Gertler, P. J. 1989. "Subsidies, Quality, and the Regulation of Nursing Homes." Journal of Public Economics, 38, pp. 33-52. ---. 1992. "Medicaid and the Cost of Improving Access to Nursing Home Care." The Review of Economics and Statistics, 74:2, pp. 338-345. Gertler, P. J. and Andreano, R. L. 1982. "A Dominant Buyer in a Mixed Industry: The Role of Medicaid in the L ong-Term Care Market," in Economics of Health Care. Gaag, J. V. d., Neenan, W. B., and Jr., T. T. eds. New York: Praeger, pp. 141-177. Giacalone, J. A. 2001. The U.S. Nursing Home Industry. Armonk and London: M.E. Sharpe.

PAGE 133

121 Graber, D. R. and Sloane, P. D. 1995. "Nursi ng Home Survey Deficiencies for Physical Restraint Use." Medical Care, 33:10, pp. 1051-1063. Grabowski, D. C. 1999. "Medicaid reimburse ment and the quality of nursing home care." DAI, 60:12A, pp. 184. ---. 2001a. "Does an Increase in the Medica id Reimbursement Rate Improve Nursing Home Quality." Journal of Gerontology:Social Sciences, 56B:2, pp. S84-S93. ---. 2001b. "Medicaid reimbursement a nd the quality of nursing home care." Journal of Health Economics, 20, pp. 549-569. ---. 2002. "The Economic Implications of Case-Mix Medicaid Reimbursement for Nursing Home Care." Inquiry, 39, pp. 258-278. ---. 2004. "A Longitudinal Study of Medicaid Payment, Private-Pay Price and Nursing Home Quality." International Journal of He alth Care Finance and Economics, 4, pp. 5-26. Grabowski, D. C. and Hirth, R. A. 2002. "Competitive spillovers across non-profit and for-profit nursing homes." Journal of Health Economics. Grabowski, D. C., Ohsfeldt, R. L., and Mo rrisey, M. A. 2003. "The Effects of CON Repeal on Medicaid Nursing Home and Long-Term Care Expenditures." Inquiry, 40:2, pp. 146-157. Greene, V. L. and Monahan, D. J. 1981. "S tructural and Operati onal Factors Affecting Quality of Patient Care In Nursing Homes." Public Policy, 29:4, pp. 399-415. Greenlees, J. S., Marshall, J. M., and Ye tt, D. E. 1982. "Nursing Home Admissions Policies under Reimbursement." The Bell Journal of Economics, 13:1, pp. 93-106. Harrington, C. and Carrillo, H. 1999. "The Regulation and Enforcement of Federal Nursing Home Standards, 1991-1997." Medical Care Research and Review, 56:4, pp. 471-494.

PAGE 134

122 Harrington, C., Carrillo, H., Mu llan, J., and Swan, J. H. 1998. "Nursing Facility Staffing in the States: The 1991 to 1995 Period." Medical Care Research and Review, 55:3, pp. 334-363. Harrington, C., Carrillo, H., Thollaug, S., and Summers, P. 1999. "Nursing facilities, staffing, residents, and facility defici encies, 1991-97." (Report prepared for the Health Care Financing Administration), Sa n Francisco:University of California. Harrington, C., Carrillo, H., Thollaug, S. C ., Summers, P. R., and Wellin, V. October 2000a. Nursing Facilities, Staffi ng, Residents, and Facility Deficiencies, 1993 Through 1999. Department of Social and Beha vioral Sciences, University of California. San Francisco, CA. Harrington, C., Curtis, M., Carrillo, H., Bedney, B., Swan, J. H., and Nyman, J. A. 1997a. "State Regulation of the Supply of Long Term Care Providers." Journal of Applied Gerontology, 16:1, pp. 5-26. Harrington, C., Preston, S., Gran t, L., and Swan, J. H. 1992. "Revised Trends in States' Nursing Home Capacity." Health Affairs, 11:2, pp. 170-180. Harrington, C. and Swan, J. H. 1984. "Medic aid nursing home reimbursement policies, rates, and expenditures." Health Care Financing Review, 6:1, pp. 39-49. ---. 1987. "The Impact of State Medicaid Nursing Home Policies on Utilization and Expenditures." Inquiry, 24, pp. 157-172. Harrington, C., Swan, J. H., Nyman, J. A., and Carrillo, H. 1997b. "The Effect of Certificate of Need and Moratoria Policy on Change in Nursing Home Beds in the United States." Medical Care, 35:6, pp. 574-588. Harrington, C., Swan, J. H ., Wellin, V., Clemea, W., and Carrillo, H. M. 2000b. 1998 State Data Book On Long Term Care Program and Market Characteristics. Department of Social and Behavioral Sc iences, University of California. San Francisco, CA.

PAGE 135

123 Harrington, C., Woolhandler, S., Mullan, J., Carrillo, H., and Himmelstein, D. U. 2002. "Does Investor Ownership of Nursing Ho mes Compromise the Quality of Care." International Journal of Health Sciences, 32:2, pp. 315-325. Harrington, C., Zimmerman, D. R., Karon, S. L., Robinson, J., and Beutel, P. 2000c. "Nursing Home Staffing and Its Relationship to Deficiencies." Journal of Gerontology:Social Sciences, 55B:5, pp. S278-S287. Hawes, C. and Phillips, C. D. 1986. "The Changing Structure of the Nursing Home Industry and the Impact of Ownership on Quality, Cost, and Access," in ForProfit Enterprise in Health Care. Gray, B. H. ed. Washington, D. C.: National Academy Press, pp. 492-541. Health Care Financing Agency. 1995. State Operations Manual: Provider Certification. Washington, D.C.: Department of Health and Human Services. Holahan, J. and Cohen, J. W. 1987. "Nur sing Home Reimbursement: Implications for Cost Containment, Access and Quality." The Milbank Quarterly, 65:1, pp. 112147. Institute of Medicine (IOM). 1986. Improving the Quality of Care in Nursing Homes. Washington, D.C.: National Academy Press. ---. 1996. Nursing Staff in Hospitals and Nu rsing Homes: Is It Adequate? Washington, D.C.: National Academy Press. ---. 2001. Improving the Quality of Long-Term Care. Washington, D.C.: National Academy Press. Katz, S., Ford, A. B., Moskowitz, R. W., Jackson, B. A., and Jaffee, M. W. 1963. "Studies of Illness in the Aged, The I ndex of ADL: A Standardized Measure of Biological and Psychosocial Function." Journal of the American Medical Association, 185:12, pp. 914-919. Kennedy, P. 2003. A Guide to Econometrics. Cambridge, MA.: The MIT Press.

PAGE 136

124 Lee, J. A., Birnbaum, H., and Bishop, C. E. 1983. "How Nursing Homes Behave: A Multi-Equation Model of Nursing Home Behavior." Social Science and Medicine, 17:23, pp. 1897-1906. Linn, M. W., Gurel, L., and Linn, B. S. 1977. "Patient Outcome as a Measure of Quality of Nursing Home Care." American Journal of Public Health, 67:4, pp. 337-342. Marek, K. D., Rantz, M., Fagin, C., and Kr ejci, J. W. 1996. "OBR A '87 Has it Resulted in Better Quality of Care?" Journal of Gerontological Nursing, 22:10, pp. 28-36. Meiners, M. R. 1982. "An econometric analys is of the major determinants of nursing home costs in the United States." Social Science & Medicine, 16:8, pp. 887-898. Meyer, B. D. 1995. "Natural and Quasi-Experiments in Economics." Journal of Business & Economic Statistics, 13:2, pp. 151-161. Miller, N. A., Harrington, C., Ramsland, S., and Goldstein, E. 2002. "State Policy Choices and Medicaid Long-Term Care Expenditures." Research on Aging, 24:4, pp. 413-444. Miller, N. A., Ramsland, S., Goldstein, E ., and Harrington, C. 2001. "Use of Medicaid 1915(c) Homeand Community-Based Care Waivers to Reconfigure State LongTerm Care Systems." Medical Care Research and Review, 58:1, pp. 100-119. Moulton, B. R. 1990. "An Illustration of a P itfall in Estimating the Effects of Aggregate Variables on Micro Units." The Review of Economics and Statistics, 72:2, pp. 334-338. Mukamel, D. B. 1997. "Risk-Adjusted Outcome Measures and Quality of Care in Nursing Homes." Medical Care, 35:4, pp. 367-385. Mukamel, D. B. and Spector, W. D. 2002. "The competitive nature of the nursing home industry: price mark ups and demand elasticities." Applied Economics, 34, pp. 413-420.

PAGE 137

125 Norton, E. C. 1992. "Incentive regulation of nursing homes." Journal of Health Economics, 11:2, pp. 105-128. ---. 2000. "Long-Term Care," in Handbook of Health Economics. Culyer, A. J. and Newhouse, J. P. eds. Amsterdam: Elsevier Science B.V., pp. 955-994. Nyman, J. A. 1985. "Prospective and 'Cos t-Plus' Medicaid Reimbursement, Excess Medicaid Demand, and the Qual ity of Nursing Home Care." Journal of Health Economics, 4, pp. 237-259. ---. 1987. "Improving the Quality of Nursi ng Homes: Regulation or Competition?" Journal of Policy Analysis and Management, 6:2, pp. 247-251. ---. 1988a. "The Effect of Competition on Nurs ing Home Expenditures under Prospective Payment." Health Services Research, 23:4, pp. 555-574. ---. 1988b. "Excess Demand, the Percentage of Medicaid Patients, and the Quality of Nursing Home Care." The Journal of Human Resources, 23:1, pp. 76-92. ---. 1988c. "Improving the Quality of Nursing Home Outcomes: Are Adequacyor Incentive-Oriented Policies More Effective?" Medical Care, 26:12, pp. 11581171. ---. 1988d. "The Marginal Cost of Nursing Home Care." Journal of Health Economics, 7, pp. 393-412. ---. 1989a. "Analysis of Nursing Home Use and Bed Supply: Wisconsin, 1983." Health Services Research, 24:4, pp. 511-537. ---. 1989b. "Excess Demand, Consumer Rationality and the Quality of Care in Regulated Nursing Homes." Health Services Research, 24:1, pp. 105-127.

PAGE 138

126 ---. 1990. "The Future of Nursing Home Policy: Should Policy Be Based on an Excess Demand Paradigm?," in Advances In Health Economics And Health Services Research. Scheffler, R. M. and Rossiter, L. F. eds. Greenwich: JAI Press Inc., pp. 229-250. ---. 1994. "The Effects of Market Concen tration and Excess Demand on the Price of Nursing Home Care." Journal of Indus trial Economics, 42:2, pp. 193-204. Nyman, J. A., Levey, S., and Rohrer, J. E. 1987. "RUGs and Equity of Access to Nursing Home Care." Medical Care, 25:5, pp. 361-372. Office of Inspector General (OEI). March 1999. Nursing Home Survey and Certification: Deficiency Trends. OEI-02-98-00331. Department of Health and Human Services. Washington, D.C. Omnibus Budget Reconciliati on Act of 1980. U.S. Public Law 96-499. Washington, D.C.: U.S. Government Printing Office, 1980. Omnibus Budget Reconciliati on Act of 1987. U.S. Public Law 100-203. Subtitle C: Nursing Home Reform: 1987. Palmer, H. C. and Vogel, R. J. 1983. "Models of the Nursing Home," in Long-Term Care: Perspectives from Research and Demonstrations. Palmer, H. C. and Vogel, R. J. eds. Washington, D.C.: U.S. Depa rtment of Health and Human Services, Health Care Financing Administration, pp. 537-578. Phillips, C. D., Hawes, C., and Fries, B. E. 1993. "Reducing the Use of Physical Restraints in Nursing Homes: Will It Increase Costs?" American Journal of Public Health, 83:3, pp. 342-348. Porell, F., Caro, F. G., Silva, A., and Monane, M. 1998. "A Longitudinal Analysis of Nursing Home Outcomes." Health Services Research, 33:4, pp. 835-865. Quality Resource Systems Inc. (QRS). 2006. Area Resource File. http://www.arfsys.com/ Accessed: January 19, 2006.

PAGE 139

127 Regional Economic Information System (REIS). 2005. Table CA04 County income and employment summary. Accessed: December 6, 2005. Reschovsky, J. D. 1996. "Demand for and A ccess to Institutional Long-Term Care: The Role of Medicaid in Nursing Home Markets." Inquiry, 33, pp. 15-29. Ruchlin, H. S. and Levey, S. 1972. "Nur sing Home Cost Analysis: A Case Study." Inquiry, 9:3, pp. 3-15. Scanlon, W. J. 1980a. "Nursing Home Utilizat ion Patterns: Implications for Policy." Journal of Health Politics, Policy and Law, 4:4, pp. 619-641. ---. 1980b. "A Theory of the Nursing Home Market." Inquiry, 17:Spring, pp. 25-41. Schlenker, R. E. 1986. "Case Mix Reimbursement For Nursing Homes." Journal of Health Politics, Policy and Law, 11:3, pp. 445-461. Schlenker, R. E. and Shaughne ssy, P. W. 1984. "Case mix, qua lity, and cost relationships in Colorado nursing homes." Health Care Financing Review, 6:2, pp. 61-71. Shaughnessy, P. W., Schlenker, R. E., and Kramer, A. M. 1990. "Quality of Long-Term Care in Nursing Homes and Swing-Bed Hospitals." Health Services Research, 25:1, pp. 65-96. Smith, D. M. 1995. "Pressure Ul cers in the Nursing Home." Annals of Internal Medicine, 123:6, pp. 433-438. Social Security Amendments of 1972. U.S. Public Law 92-603. Washington, D.C.: U.S. Government Printing Office, 1972. Spector, W. D., Selden, T. M., and Cohen, J. W. 1998. "The Impact of Ownership Type on Nursing Homes Outcomes." Health Economics, 7, pp. 639-653.

PAGE 140

128 Spector, W. D. and Takada, H. A. 1991. "Cha racteristics of Nursing Homes That Affect Nursing Home Outcomes." Journal of Aging and Health, 3:4, pp. 427-454. Spillman, B. C. and Lubitz, J. 2002. "New Es timates of Lifetime Nursing Home Use: Have Patterns of Use Changed?" Medical Care, 40:10, pp. 965-975. Strahan, G. W. January 23, 1997. An Overview of Nursing Homes and Their Current Residents: Data From the 1995 National Nursing Home Survey. Advance Data Number 280. U.S. Department of Health and Human Services, Centers of Disease Control and Prevention, Nati onal Center for Health Statistics, Division of Health Care Statistics. Swan, J. H. 1990. "The Share of Medicaid for Nursing Home Care." Journal of Health & Social Policy, 1:3, pp. 35-53. Swan, J. H. and Harrington, C. 1985. "Medicaid Nursing Home Reimbursement Policies," in Long Term Care of the Elde rly: Public Policy Issues. Harrington, C., Newcomer, R. J., and Estes, C. J. eds. Beverly Hills: Sage Publications, pp. 125151. ---. 1990. "Certificate of need and nu rsing home bed capacity in states." The Journal of Health & Social Policy, 2:2, pp. 87-105. Swan, J. H., Harrington, C., Grant, L., Lueh rs, J., and Preston, S. 1993. "Trends in Medicaid Nursing Home Reimbursement: 1978-89." Health Care Financing Review, 14:4, pp. 111-132. Thorpe, K. E., Gertler, P. J., and Goldman, P. 1991. "The Resource Utilization Group System: Its Effect on Nursing Home Case Mix and Costs." Inquiry, 28, pp. 357365. U.S. House of Representatives. October 2002. Nursing Home Conditions in Texas: Many Nursing Homes Fail to Meet Federal Standards for Adequate Care. Minority Staff, Special Investigations Divi sion, Committee on Government Reform. Washington, D.C.

PAGE 141

129 Ullmann, S. G. 1981. "Assessment of Facility Quality and Its Relationship to Facility Size in the Long-Term Health Care Industry." The Gerontologist, 21:1, pp. 91-97. ---. 1984. "Cost Analysis and Facility Reim bursement in the Long-Term Health Care Industry." Health Services Research, 19:1, pp. 83-102. ---. 1985. "The Impact of Quality on Cost in the Provision of Long-Term Care." Inquiry, 22, pp. 293-302. United States General Accounting Office (GAO). July 1987. Medicare and Medicaid: Stronger Enforcement of Nursing Home Requirements Needed. GAO/HRD-87113. Washington, D.C. ---. July 1998. California Nursing Homes: Care Problems Persist Despite Federal and State Oversight. GAO/HEHS-98-202. Washington, D.C. ---. March 1999. Nursing Homes: Additional Steps Needed to Strengthen Enforcement of Federal Quality Standards. GAO/HEHS-99-46. Washington, D.C. ---. March 2002. Nursing Homes: More Can Be Done to Protect Residents from Abuse. GAO-02-312. Washington, D.C. ---. July 2003. Nursing Home Quality: Prevalence of Serious Problems, While Declining, Reinforces Importance of Enhanced Oversight. GAO-03-561. Washington, D.C. ---. 2005. Nursing Homes: Despite Increased Oversight, Challenges Remain in Ensuring High-Quality Care and Resident Safety. GAO-06-117. Washington, D.C. Vladeck, B. C. 1980. Unloving Care: The Nursing Home Tragedy. New York: Basic Books, Inc. Waldman, S. 1983. "A Legislative Hi story of Nursing Home Care," in Long Term Care: Perspectives from Research and Demonstrations. Vogel, R. J. and Palmer, H. C. eds. Washington, D.C.: U.S. Department of Health and Human Services, Health Care Financing Admini stration, pp. 507-535.

PAGE 142

130 Weisbrod, B. A. and Schlesinger, M. 1986. "Public, Private, Nonprofit Ownership and the Response to Asymmetric Informatio n: The Case of Nursing Homes," in The Economics of Nonprofit Institutions. Rose-Ackerman, S. ed. New York: Oxford University Press, pp. 133-151. Weissert, W. G. and Musliner, M. C. 1992. "Case Mix Adjusted Nursing-home Reimbursement: A Critical Review of the Evidence." The Milbank Quarterly, 70:3, pp. 455-490. Wiener, J. M. and Stevenson, D. G. 1998. "Repeal of the "Boren Amendment": Implications for Quality of Care in Nursing Homes." The Urban Institute, Series A, No. A-30, pp. 1-7. Winn, S. 1974. "Analysis of Selected Charact eristics of a Matched Sample of Nonprofit and Proprietary Nursing Homes in the State of Washington." Medical Care, 12:3, pp. 221-228. Wooldridge, J. M. 2002. Econometric Analysis of Cr oss Section and Panel Data. Cambridge, MA.: The MIT Press. Zinn, J. S. 1993. "The Influence of Nurse Wage Differentials on Nursing Home Staffing and Resident Care Decisions." The Gerontologist, 33:6, pp. 721-729. ---. 1994. "Market Competition and the Quality of Nursing Home Care." Journal of Health Politics, Policy and Law, 19:3, pp. 555-582. Zwanziger, J., Mukamel, D. B., and Indridason, I. 2002. "Use of Resident-Origin Data to Define Nursing Home Market Boundaries." Inquiry, 39, pp. 56-66.

PAGE 143

131 Bibliography Cowles, C. McKeen. 2002. 2002 Nursing Home Statistical Yearbook. Cowles Research Group. Montgomery Village, MD.

PAGE 144

132 Appendix A: Tables

PAGE 145

133 Table A.1 CMS’ Deficiency Classification System CMS deficiency category Level of severity Compliance status of home cited for this deficiency Immediate jeopardy to resident health or safety Most serious Noncompliant Actual harm that does not put resident in immediate jeopardy Serious Noncompliant No actual harm, with potential for more than minimal harm Less serious Noncompliant No actual harm, with potential for minimal harm Minimal Substantially compliant Adapted from the GAO (1998b) report.

PAGE 146

134 Table A.2 Illustrative Measures of Qu ality of Care in Nursing Structural Measures Staffing levels (nurses, PTs, OTs, etc.) Governance Staffing turnover Age/condition of plant, equipment (include mobility development) Wages/benefits Payer mix (percent mix, etc.) Management/leadership Case mix Facility: size, location, ownership Accreditation Availability of privat e rooms Teaching status Volunteers Process Measures Assists with ADL/IADL (includes bathing, skin care) Bladder training Injury (staff and patient) Delivery of “hotel” services (sanitation) Infection control (includes residents and staff) Assessment (includes care planning), frequency and completeness Resident services: special care to prevent problems Abuse prevention Overuse of restraints Quality assurance (RA and MDS) Use of urinary catheters Acce ss and use of medical care Residents rights Outcome Measures Mortality Urinary incontinence Hospitalization Weight loss Facility-acquired pressure sore, skin breakdown Infectious disease Functional status Patient satisfaction Pain control Family satisfaction Depression Thefts/abuse Injuries Staff injuries/illness Staff satisfaction Notes: ADL = activities of daily living; IADL = instrumental activities of daily living; OT = occupational therapist; PT = physical therapist; RA = resident assessment; MDS = minimum data set

PAGE 147

135Table A.3 Certificate of Need an d Moratorium Policies for the Years 1991-2003 State 1991 1992 1993 1994 1995 1996 1997 1998 1999 2000 2001 2002 2003 AL 1 1 1 3 3 1 1 1 3 1 1 1 3 AK 1 1 1 1 1 3 3 3 1 1 1 1 1 AR* 1 3 3 1 1 1 1 1 1 1 1 1 1 AZ 0 0 0 0 0 0 0 0 0 0 0 0 0 CA 0 0 0 0 0 0 0 0 0 0 0 0 0 CO 2 2 2 2 2 2 2 2 2 2 2 2 0 CT 3 3 3 3 3 3 3 3 3 3 3 3 3 DE 1 1 1 1 1 1 1 1 1 1 1 1 1 FL 1 1 1 1 1 1 1 1 1 1 3 3 3 GA 1 1 1 1 1 1 1 1 1 1 1 1 1 HI 1 1 1 1 1 1 1 1 1 1 1 1 1 IA 1 1 1 1 1 1 1 1 1 1 1 1 1 ID 0 0 0 0 0 0 0 0 0 0 0 0 0 IL 1 1 1 1 1 1 1 1 1 1 1 1 1 IN 1 1 1 1 1 0 1 1 0 0 0 0 0 KS 0 0 0 0 0 0 0 0 0 0 0 0 0 KY 1 3 3 3 1 1 1 1 1 1 3 3 3 LA 1 1 1 1 1 1 3 3 3 3 3 3 3 MA 3 3 3 3 3 3 3 3 3 3 3 3 3 MD 1 1 1 1 1 1 1 1 1 1 1 1 1 ME 3 3 3 3 3 3 3 3 3 3 3 3 3 MI 1 1 1 1 1 3 3 3 3 3 3 3 3 MN 2 2 2 2 2 2 2 2 2 2 2 2 2 MO 3 3 3 3 3 3 3 3 3 3 3 3 1 MS 3 3 3 3 3 3 3 3 3 3 3 3 3 MT 1 1 1 1 1 1 1 1 1 1 1 1 1 NC 1 1 1 1 1 1 1 1 1 1 1 1 1 ND 1 1 1 1 2 2 2 2 2 2 2 2 2 NE 1 1 1 1 1 1 1 1 3 3 3 3 3 NH 1 1 1 1 3 3 3 3 3 3 3 3 3 NJ 3 1 1 1 1 1 1 1 1 1 1 1 1 NM 0 0 0 0 0 0 0 0 0 0 0 0 0 NV1 1 1 1 1 1 0 0 0 0 0 0 0 0 NY 1 1 1 1 1 1 1 1 1 1 3 3 3 OH 1 1 3 3 3 3 3 3 3 3 3 3 3 OK 1 1 1 1 1 1 1 1 1 1 1 1 1 OR 1 1 1 1 1 1 1 1 1 1 1 1 1 PA 1 1 1 1 1 1 0 0 0 0 0 0 0 RI 1 1 1 1 1 3 3 3 3 3 3 3 3 SC 1 1 1 1 1 1 1 1 1 1 1 1 1 SD 2 2 2 2 2 2 2 2 2 2 2 2 2 TN 1 1 1 1 1 1 1 1 1 1 1 1 1 TX 2 2 2 2 2 2 2 2 2 2 2 2 2 UT 2 2 2 2 2 2 2 2 2 2 2 2 2 Legend: 0 = no CON and no moratorium 2 = no CON but moratorium *permit of approval law 1 = CON but not no moratorium 3 = CON and moratorium 1Nevada eliminated certificate of need in its two largest counties begi nning in 1996 and is thus labeled as 1

PAGE 148

136Table A.3 (continued) Certificate of Need an d Moratorium Policies for the Years 1991-2003 State 1991 1992 1993 1994 1995 1996 1997 1998 1999 2000 2001 2002 2003 VT 1 1 1 1 1 1 1 1 1 1 1 1 1 VA 1 1 1 1 1 1 1 1 1 1 1 1 1 WA 1 1 1 1 1 1 1 1 1 1 1 1 1 WV 3 3 3 3 3 3 3 3 3 3 3 3 3 WI 3 3 3 3 3 3 3 3 3 3 3 3 3 WY* 1 1 1 1 1 1 1 1 1 1 1 1 1 Legend: 0 = no CON and no moratorium 2 = no CON but moratorium *intent to construct law 1 = CON but not no moratorium 3 = CON and moratorium

PAGE 149

137Table A.4 Reimbursement Method for the Years 1991-2003 State 1991 1992 1993 1994 1995 1996 1997 1998 1999 2000 2001 2002 2003 AL pfs pfs pfs pfs pfs pfs pfs pfs pfs pfs pfs pfs pfs AK adj adj adj adj adj adj ad j adj adj adj adj adj adj AR1 pcl pcl pcl pcl pcl pcl pcl pcl pcl pcl pfs pfs pfs AZ2 pcl pcl pcl pcl pcl pcl pcl pcl pcl pcl pcl pcl pcl CA pcl pcl pcl pcl pcl pcl pcl pcl pcl pcl pcl pcl pcl CO adj adj adj adj adj adj adj adj adj adj adj adj adj CT adj adj pfs pfs pfs pfs pfs pfs pfs pfs pfs pfs pfs DE pfs pfs pfs pfs pfs pfs pfs pfs pfs pfs pfs pfs pfs FL pfs pfs adj adj adj adj adj adj adj adj adj adj adj GA adj adj adj adj adj adj ad j adj adj adj adj adj adj HI adj adj adj adj adj adj adj adj adj adj adj adj adj IA adj adj adj adj adj adj adj adj adj adj adj adj adj ID adj adj adj adj adj adj adj adj adj adj adj adj adj IL adj adj pfs pfs pfs pfs pfs pfs pfs pfs pfs pfs pfs IN adj adj pfs pfs pfs pfs pfs pfs pfs pfs pfs pfs pfs KS pfs pfs adj adj adj adj adj adj adj adj adj adj adj KY adj adj adj adj adj adj ad j adj adj adj adj adj adj LA pcl pcl pcl pcl pcl pcl pcl pcl pcl pcl pcl pcl pfs MA com com pfs pfs pfs pfs pfs pfs pfs pfs pfs pfs pfs MD adj adj adj adj adj adj adj adj adj adj adj adj adj ME com com adj adj adj adj adj adj adj adj adj adj adj MI adj adj adj adj adj adj adj adj adj adj adj adj adj MN pfs pfs pfs pfs pfs pfs pfs pfs pfs pfs pfs pfs pfs MO adj adj adj adj adj adj adj adj adj adj adj adj adj MS pfs pfs adj adj adj adj adj adj adj adj adj adj adj MT adj adj adj adj adj adj adj adj adj adj adj adj adj NC com com com com com com com com com com com com com ND adj adj adj pfs pfs pfs pfs pfs pfs pfs pfs pfs pfs NE com ret ret ret ret ret ret ret ret ret pfs pfs pfs NH adj adj adj adj adj adj ad j adj adj adj adj adj adj NJ adj adj adj adj adj pfs pfs pfs pfs pfs pfs pfs pfs NM adj adj adj pfs pfs pfs pfs pfs pfs pfs pfs pfs pfs NV adj adj pfs pfs pfs pfs pfs pfs pfs pfs pfs pfs pfs NY adj adj adj adj adj adj ad j adj adj adj adj adj adj OH adj adj adj adj adj adj ad j adj adj adj adj adj adj OK pcl pcl pcl pcl pcl pcl pcl pcl pcl pcl pcl pcl pcl OR com adj adj adj adj adj adj adj adj adj adj adj adj PA ret ret ret ret ret com adj adj adj adj adj adj adj RI pfs pfs pfs pfs pfs pfs pfs pfs pfs pfs pfs pfs pfs SC adj adj adj adj adj adj adj adj adj adj adj adj adj SD pfs pfs pfs pfs pfs pfs pfs pfs pfs pfs pfs pfs pfs TN com com com com com com com pfs pfs pfs pfs pfs pfs TX pfs pfs pfs pfs pfs pfs pfs pfs pfs pcl pcl pcl pcl UT pcl pfs pfs pfs pfs pfs pfs pfs pfs pfs pfs pfs pfs VA com com com com com com com com com com com com pfs VT adj adj adj adj adj adj adj adj adj adj adj adj adj WA adj adj adj pfs pfs pfs pfs pfs pfs pfs pfs pfs pfs Legend: ret = retrospective; pfs = prospective, facility-spec ific; pcl = prospective, class; com = combination; adj = prospective, adjusted 1Harrington, et al. (2000) classifies 1992-1998 as pfs but Medicaid agency c ontact classified method as pcl. 2Harrington, et al. (2000) classifies as pfs but Medi caid agency contact classified method as pcl.

PAGE 150

138Table A.4 (continued) Reimbursement Method for the Years 1991-2003 State 1991 1992 1993 1994 1995 1996 1997 1998 1999 2000 2001 2002 2003 WI pfs pfs pfs pfs pfs pfs pfs pfs pfs pfs pfs pfs pfs WV adj adj adj adj adj adj adj adj adj adj adj adj adj WY adj adj adj adj adj adj adj adj adj adj adj adj adj Legend: ret = retrospective; pfs = prospective, facility-spec ific; pcl = prospective, class; com = combination; adj = prospective, adjusted

PAGE 151

139Table A.5 Case-mix Reimbursement for the Years 1991-2003 State 1991 1992 1993 1994 1995 1996 1997 1998 1999 2000 2001 2002 2003 AL No No No No No No No No No No No No No AK No No No No No No No No No No No No No AR Yes Yes Yes Yes Yes Yes Yes Yes Ye s Yes Yes Yes Yes AZ Yes Yes Yes Yes Yes Yes Yes Yes Ye s Yes Yes Yes Yes CA No No No No No No No No No No No No No CO No No No No No No No No No No No No No CT No No No No No No No No No No No No No DE Yes Yes Yes Yes Yes Yes Yes Yes Ye s Yes Yes Yes Yes FL No No No No No No No No Yes Ye s Yes Yes Yes GA No No No No No No No No No No No No Yes HI No No No No No No No No No No No No No IA Yes No No No No No No No No Yes Yes Yes Yes ID No No No No No No No No No No Yes Yes Yes IL Yes Yes Yes Yes Yes Yes Yes Yes Ye s Yes Yes Yes Yes IN No No No No No No No Yes Yes Ye s Yes Yes Yes KS No No No Yes Yes Yes Yes Yes Yes Ye s Yes Yes Yes KY Yes Yes Yes Yes Yes Yes Yes Yes Ye s Yes Yes Yes Yes LA No No No No No No No No No No No No Yes MA Yes Yes Yes Yes Yes Yes Yes Yes Yes Yes Yes Yes Yes MD Yes Yes Yes Yes Yes Yes Yes Yes Yes Yes Yes Yes Yes ME No No No Yes Yes Yes Yes Yes Yes Ye s Yes Yes Yes MI No No No No No No No No No No No No No MN Yes Yes Yes Yes Yes Yes Yes Yes Yes Yes Yes Yes Yes MO No No No No No No No No No No No No No MS No No No Yes Yes Yes Yes Yes Yes Ye s Yes Yes Yes MT Yes Yes Yes Yes Yes Yes Yes Yes Yes Yes Yes Yes Yes NC No No No No No No No No No No No No Yes ND Yes Yes Yes Yes Yes Yes Yes Yes Ye s Yes Yes Yes Yes NE No Yes Yes Yes Yes Yes Yes Yes Ye s Yes Yes Yes Yes NH No No No No No No No No Yes Yes Yes Yes Yes NJ Yes Yes Yes Yes Yes Yes Yes Yes Ye s Yes Yes Yes Yes NM No No No No No No No No No No No No No NV Yes Yes Yes Yes Yes Yes Yes Yes Ye s Yes Yes Yes Yes NY Yes Yes Yes Yes Yes Yes Yes Yes Ye s Yes Yes Yes Yes OH Yes No No Yes Yes Yes Ye s Yes Yes Yes Yes Yes Yes OK No No No No No No No No No No No No No OR No No No No No No No No No No No No No PA No No No No No Yes Yes Yes Yes Ye s Yes Yes Yes RI No No No No No No No No No No No No No SC Yes Yes Yes Yes Yes Yes Yes Yes Ye s Yes Yes Yes Yes SD No No No Yes Yes Yes Yes Yes Yes Ye s Yes Yes Yes TN No No No No No No No No No No No No No TX Yes Yes Yes Yes Yes Yes Yes Yes Ye s Yes Yes Yes Yes UT No No No No No No No No No No No No Yes VA Yes Yes Yes Yes Yes Yes Yes Yes Ye s Yes Yes Yes Yes VT No Yes Yes Yes Yes Yes Yes Yes Ye s Yes Yes Yes Yes WA No No No No No No No No Yes Ye s Yes Yes Yes WI1 No No No No No No No No No No No No No WV Yes Yes Yes Yes Yes Yes Yes Yes Ye s Yes Yes Yes Yes WY No No No No No No No No No No No No No 1Harrington, et al. (2000) classifies 1993-1998 as having casemix reimburseme nt but Medicaid agency contact classified as not havin g case-mix.

PAGE 152

140 Table A.6 Average Per Diem Medicaid Reimbur sement Rates for the Years 1991-2003 State 1991 1992 1993 1994 1995 1996 1997 1998 1999 2000 2001 2002 2003 AL 52.58 62.48 68.75 75.37 82.76 85.57 94.73 98.69 103.54 107.18 112.54 119.61 126.98 AK 205.04 217.19 221.27211.21214.77225.38237.47261.67 265.23 275.00 303.71 320.95 329.21 AR 42.22 49.05 54.86 58.02 60.28 61.98 61.98 61.98 64.51 69.01 76.92 94.73 102.51 AZ 69.24 72.30 71.93 77.80 82.36 88.23 88.23 91.88 95.53 100.30 105.83 108.17 117.16 CA 65.32 70.60 72.28 76.27 79.71 79.77 81.54 88.78 91.32 110.19 113.13 113.60 118.05 CO 64.46 72.55 71.28 73.59 75.93 90.31 90.31 101.55 105.88 111.15 114.68 123.37 130.98 CT 112.38 116.57 118.00120.27125.06129.62133.82137.06 147.97 154.37 158.51 161.25 167.95 DE 77.53 81.80 86.16 90.10 93.78 99.14 104.15102.09 105.22 117.66 133.46 157.91 171.62 FL 71.80 76.71 80.48 86.45 87.95 90.62 95.95 97.99 101.74 110.37 119.54 129.36 142.90 GA 61.03 63.49 67.02 68.85 72.99 77.15 72.60 76.30 86.56 88.50 95.10 105.24 117.59 HI 95.03 100.54 105.76119.39124.05129.65132.59127.12 131.06 134.86 140.05 146.85 154.04 IA 50.32 53.10 56.27 57.72 58.87 64.62 68.11 72.78 78.61 85.15 85.90 91.96 95.38 ID 62.28 65.47 74.67 75.45 81.28 88.03 86.29 98.42 104.65 112.65 123.90 130.63 137.66 IL 52.26 62.23 70.08 70.08 70.08 70.22 70.28 77.62 81.44 86.05 90.06 94.86 89.92 IN 60.43 64.02 66.56 71.28 78.48 81.75 80.32 80.32 92.16 94.58 100.39 104.25 101.66 KS 47.72 50.55 51.24 56.14 60.08 63.68 67.11 72.71 76.72 84.12 91.65 97.38 101.04 KY 55.21 59.50 64.03 69.32 73.44 76.00 83.00 80.62 86.34 89.26 100.35 101.76 107.76 LA 50.48 63.82 65.26 60.60 63.78 63.52 61.12 62.48 61.51 71.99 76.67 79.28 80.24 MA 94.77 96.07 96.04 98.84 100.73103.35111.92114.68 119.24 124.47 132.60 140.04 154.26 MD 73.07 77.52 79.33 81.35 89.16 94.19 98.88 106.62 111.93 122.15 134.42 150.64 162.49 ME 84.94 87.25 87.00 101.40105.85114.09113.41115.77 113.83 116.20 130.32 130.37 158.10 MI 58.21 64.88 67.12 71.01 74.25 79.46 84.17 91.49 101.43 107.10 117.44 125.08 134.43 MN 78.32 82.06 81.91 88.21 92.24 95.61 101.79106.47 111.03 117.31 123.04 129.62 138.34 MO 42.65 52.89 57.93 58.15 70.39 73.18 83.35 87.81 89.60 92.32 96.42 96.84 97.01 MS 52.20 55.44 58.07 58.08 68.00 72.89 76.77 83.02 86.40 90.22 97.38 104.46 115.63

PAGE 153

141 Table A.6 (continued) Average Per Diem Medicaid Reimbur sement Rates for the Years 1991-2003 State 1991 1992 1993 1994 1995 1996 1997 1998 1999 2000 2001 2002 2003 MT 56.39 63.22 69.69 74.62 80.15 83.09 85.89 87.55 90.39 93.39 96.97 102.13106.29 NC 78.66 86.75 76.68 79.20 83.53 86.87 92.82 93.12 111.94116.27122.14126.36126.36 ND 66.51 71.98 74.29 75.92 80.86 85.77 90.86 94.13 97.68 104.94115.03127.05129.71 NE 56.82 60.37 60.07 62.03 66.17 70.99 76.70 84.19 90.67 97.97 107.02113.83118.94 NH 92.66 95.58 98.94 105.36105.34104.00108.47109.22 114.58116.94119.24121.99126.42 NJ 88.86 88.86 91.61 100.35102.35102.35112.01115.76 119.39123.93130.03137.57145.29 NM1 68.03 73.84 64.72 75.79 77.69 86.80 111.31129.04 . . NV 65.83 68.58 75.85 76.35 79.33 82.51 85.71 86.17 92.16 97.20 110.04102.88126.59 NY 117.01 122.90 128.10134.48138.94150.15148.91156.83 162.76160.66166.57172.05177.33 OH 79.29 84.63 88.81 93.74 99.54 106.14112.00120.52 126.63134.97145.15153.01152.72 OK 43.90 46.40 48.90 49.70 52.50 54.94 56.77 64.20 66.38 66.57 84.46 93.58 94.61 OR 58.81 67.37 69.55 75.36 76.54 76.54 81.88 88.21 93.52 97.60 108.09111.35111.35 PA 63.46 68.04 78.82 88.07 96.19 102.13109.13114.23 119.54126.14132.86140.65145.57 RI 105.09 110.88 85.76 103.7898.00 101.50101.50109.75 109.25113.44119.43129.97135.71 SC 58.20 65.24 64.99 67.57 71.22 74.69 78.08 84.85 88.38 94.80 96.07 103.11107.57 SD 46.85 54.32 60.00 64.37 68.89 77.91 74.26 75.88 77.43 79.60 83.57 86.62 91.61 TN 55.29 62.32 68.99 56.18 62.75 77.91 87.74 86.02 89.58 95.69 98.80 102.70111.21 TX 51.62 54.51 56.17 60.55 63.34 66.52 71.12 75.15 78.62 83.53 88.50 96.10 95.80 UT 59.68 67.18 67.53 70.38 74.24 76.76 78.53 83.11 85.67 89.11 90.05 94.19 105.55 VA 60.47 63.57 65.50 71.01 72.97 75.07 77.37 79.48 82.60 89.10 99.12 106.33112.06 VT 76.91 76.19 84.90 89.78 94.24 97.20 100.46104.10 105.51113.19118.20132.93139.75 WA 78.74 86.53 85.60 92.74 98.91 104.96109.03112.90 116.89119.55123.64127.64129.23 WI 66.93 71.93 73.41 76.32 80.05 85.85 85.85 95.47 97.68 100.76105.11112.09119.15 WV 68.83 72.28 78.13 80.86 77.27 89.93 101.04106.57 110.02117.86125.34134.94141.50 WY 65.77 70.66 70.06 73.06 75.84 90.09 92.41 93.72 95.15 97.89 105.13117.11119.43 1New Mexico has missing values for 1999-2003.

PAGE 154

142 Table A.7 Nursing Home, County, and State-level Characteristics (1991-2003)a by CON_MORT Variable Description CON_MORT = 1 CON_MORT = 0 Access Measure N Mean Overall SD Within SD N Mean Overall SD Within SD PCTMCAID Percent of Medicaid reside nts 22,682 65.29 19.98 8. 58 128,023 69.59 18.25 8.29 Quality Measures PROPPRESSORE Proportion of residents with decubitus ulcer 22,669 0.07 0.05 0.04 127,902 0.06 0.05 0.04 PROPPARENTERAL Proportion of residents with f eeding tubes 22,669 0.06 0. 07 0.03 127,902 0.06 0.06 0.03 PROPCATHETER Proportion of residents with ca theters 22,669 0.07 0.06 0.04 127,902 0.07 0.05 0.04 PROPMOBLREST Proportion of residents with physical restraints 22,669 0.17 0. 16 0.12 127,902 0.16 0.15 0.12 DEFS1TOT Number of health deficienci es 22,675 9.98 9.46 7. 44 127,995 6.25 6.62 5.43 RNS Registered nurse (RN) hours per resident day 22,682 0.34 0.27 0.17 128,023 0.34 0.32 0.19 LPNS Licensed practical nurse (LPN) hours per resident day 22,682 0.66 0.37 0.26 128,023 0.68 0.42 0.28 AIDES Nurses’ aides hours per resident day 22,682 2.09 0.85 0. 68 128,023 2.15 1.04 0.73 LICNUR RN and LPN hours per resident day 22,682 0.92 0.43 0. 29 128,023 0.94 0.43 0.29 NURSTAFF RN, LPN, and aide hours per reside nt day 22,682 2.86 0.96 0.74 128,023 2.92 1.10 0.77 Facility Characteristics TOTRES Number of residents 22,682 86.88 44.77 9.82 128,023 95.47 50.40 11.51 CENMCAID Number of Medicaid resident s 22,682 57.32 36.42 9. 84 128,023 66.95 41.78 11.24 CENMCARE Number of Medicare resident s 22,682 5.95 6.68 3. 92 128,023 6.58 7.98 4.82 CENOTHER Number of private-pay reside nts 22,682 23.61 19.63 8. 00 128,023 21.94 19.45 8.57 TOTBEDS Total number of beds 22,682 101.88 53.12 9.96 128,023 110.95 55.97 8.94 CONTROL1 =1 if for-profit facility (b ase) 22,682 0.75 0.43 0. 12 128,023 0.74 0.44 0.12 CONTROL2 =1 if nonprofit facility 22,682 0.22 0.42 0.12 128,023 0.21 0.41 0.12 CONTROL3 =1 if government facility 22,682 0.02 0.15 0.05 128,023 0.04 0.21 0.05 MULTI =1 if chain facility 22,682 0.61 0.49 0.23 128,023 0.56 0.50 0.23 ADLINDEX ADL index 22,669 10. 05 1.70 0.88 127,902 9.75 1.48 0.88 OCCUPANCY Number of resident s/total number of beds 22,682 86.27 12.29 7.83 128,023 86.61 14.48 8.35 aThe data are from the Online, Survey, Certification, and Reporting (OSCAR) system unless otherwise noted.

PAGE 155

143 Table A.7 (continued) Nursing Home, County, and State-level Characteristics (1991-2003)a by CON_MORT Variable Description CON_MORT = 1 CON_MORT = 0 Market (County) Characteristics POP65UP Elderly population (aged 65+)b 22,682 215,562 329,335 13, 286 128,023 53,346 94,543 3,022 EMPTYELDERLY Number of empty beds per 1000 noninstitutionalized elderlyd 22,682 7.17 8.85 4.23 128,023 8.61 10.17 4.51 INCOME Per capita personal income (2003 $)c 22,682 28,284 6,518 1,774 128,023 27,102 7,200 2,220 HHI Herfindahl-Hirschman index 22, 682 0.18 0.23 0.06 128,023 0.26 0.27 0.08 State-level Characteristics RATE Average, annual Me dicaid per diem reimbursement rate (2003 $)e 22,682 101.68 21.58 12.73 128,023 104.41 26.34 10.55 RATEMISSING =1 when New Mexico’s rate is missing 22,682 0.01 0.10 0.08 128,023 0.00 0.00 0.00 METHOD0 =1 if state uses retrospective reimbursemente 22,682 0.00 0.00 0.00 128,023 0.32 0.18 0.08 METHOD1 =1 if state uses prospective, facilityspecific reimbursemente 22,682 0.17 0.37 0.17 128,023 0.37 0.48 0.25 METHOD2 =1 if state uses prospective, class reimbursemente 22,682 0.50 0.50 0.00 128,023 0.08 0.27 0.14 METHOD3 =1 if state uses prospective, adjusted reimbursemente (base) 22,682 0.34 0.47 0.17 128,023 0.44 0.50 0.17 METHOD4 =1 if states combines retrospective and prospective reimbursemente 22,682 0.00 0.00 0.00 128,023 0.08 0.26 0.13 CASEMIX =1 if state uses case-mix Medicaid reimbursemente 22,682 0.42 0.49 0.23 128,023 0.55 0.50 0.22 aThe data are from the Online, Survey, Certification, and Reporting (OSCAR) system unless otherwise noted. bThis variable is from the Area Resource File (ARF). cThis variable is from the Bureau of Economic Analysis’ (B EA) Regional Economic Info rmation System (REIS). dThis variable is constructe d using OSCAR and ARF files. eThese variables are from the 1998 State Data Book on Long-Term Care Program and Market Characteristics (Harrington, et al. 2000b) and the author’s state surveys.

PAGE 156

144Table A.8 Regression Results for Process-Based Quality Models State Fixed Effects Dependent Variable PROPCATHETER PROPPARENTERAL PROPMOBLREST Explanatory Variables Coefficient Standard Errora Coefficient Standard Errora Coefficient Standard Errora CON_MORT -0.00957 0.00132*** -0.00096 0.00111 -0.05028 0.00321*** RATE -0.00002 0.00003 -0.00004 0.00003 -0.00061 0.00008*** RATEMISSING 0.00336 0.00512 -0.00902 0.00461** -0.07038 0.01653*** CONTROL2 -0.01008 0.00072*** -0.01040 0.00093*** 0.00309 0.00151* CONTROL3 -0.00772 0.00141*** -0.00241 0.00252 0.01414 0.00339*** TOTBEDS 0.00007 5.74e-06*** 0.00008 8.03e-06*** 0.00001 0.00001 EMPTYELDERLY 0.00005 0.00003 -0.00019 0.00004*** 0.00018 0.00007*** MULTI 0.00423 0.00056*** 0.00306 0.00067*** -0.01193 0.00118*** ADLINDEX 0.01005 0.00027*** 0.01606 0.00068*** 0.02203 0.00050*** HHI -0.00442 0.00119*** -0.02634 0.00160*** 0.01163 0.00258*** INCOME -2.47e-07 4.98e-08*** 9.48e-08 7.54e-08 -8.80e-08 1.06e-07 METHOD0 -0.00122 0.00172 -0.00119 0.00149 0.01015 0.00451** METHOD1 0.00154 0.00080* -0.00275 0.00067*** 0.01518 0.00183*** METHOD2 -0.00370 0.00158** 0.00236 0.00135* 0.04824 0.00357*** METHOD4 -0.00177 0.00132 0.00114 0.00120 0.00679 0.00337** CASEMIX 0.00195 0.00064*** -0.00380 0.00059*** 0.01002 0.00166*** CONSTANT -0.00209 0.00386 -0.09614 0.00677*** 0.06852 0.00866* N 150,571 150,571 150,571 R2 0.1449 0.2941 0.2351 Prob > F 0.0000 0.0000 0.0000 aThe standard errors represent the Huber-White robust standard errors corrected for intra-home correlation using the cluster option in Stata 9 S/E. Each regression contains 15892 clusters. *p < .10 **p < .05 ***p < .01 All models include state and year fixed effects.

PAGE 157

145Table A.9 Regression Results for Process-Based Quality Models Facility Fixed Effects Dependent Variable PROPCATHETER PROPPARENTERAL PROPMOBLREST Explanatory Variables Coefficient Standard Errora Coefficient Standard Errora Coefficient Standard Errora CON_MORT -0.01037 0.00133*** -0.00194 0.00099** -0.04827 0.00325*** RATE -0.00002 0.00003 -0.00004 0.00002* -0.00063 0.00008*** RATEMISSING 0.00589 0.00493 -0.01060 0.00414*** -0.08244 0.01552*** CONTROL2 -0.00278 0.00111** -0.00091 0.00093 0.01006 0.00324*** CONTROL3 -0.00069 0.00282 -0.00098 0.00165 0.01138 0.00791 TOTBEDS 0.00002 0.00002 -5.90e-06 0.00001 0.00005 0.00004 EMPTYELDERLY 0.00016 0.00004*** 0.00003 0.00002 0.00047 0.00008*** MULTI -0.00121 0.00060** 0.00056 0.00051 -0.00460 0.00163*** ADLINDEX 0.00564 0.00016*** 0.00668 0.00016*** 0.01409 0.00046*** HHI 0.00570 0.00160*** 0.00153 0.00130 0.02149 0.00444*** INCOME 5.17e-07 1.31e-07*** -5.82e-07 1.66e-07*** -3.08e-06 5.44e-07*** METHOD0 -0.00133 0.00171 -0.00034 0.00127 0.01356 0.00457*** METHOD1 0.00094 0.00077 -0.00274 0.00057*** 0.01701 0.00183*** METHOD2 -0.00567 0.00154*** 0.00121 0.00126 0.04961 0.00360*** METHOD4 -0.00214 0.00129* 0.00079 0.00105 0.00772 0.00339** CASEMIX 0.00252 0.00061*** -0.00181 0.00049*** 0.01198 0.00168*** CONSTANT 0.01455 0.00484*** -0.00014 0.00525 0.19108 0.01719*** N 150,571 150,571 150,571 R2 0.0535b 0.0785b 0.1571b Prob > F 0.0000 0.0000 0.0000 aThe standard errors represent the Huber-White robust standard errors corrected for intra-home correlation using the cluster option in Stata 9 S/E. Each regression contains 15892 clusters. *p < .10 **p < .05 ***p < .01 bThis value is the R2 within. All models include year fixed effects.

PAGE 158

146Table A.10 Regression Results for Outcome and Composite-Based Quality Models State Fixed Effects Dependent Variable Outcome-Based PROPPRESSORE Composite-Based DEFS1TOT Explanatory Variables Coefficient Standard Errora Coefficient Standard Errora CON_MORT 0.00081 0.00109 0.11049 0.15330 RATE -0.00018 0.00002*** -0.00801 0.00370** RATEMISSING -0.02305 0.00464*** -0.66741 0.62357 CONTROL2 -0.00752 0.00048*** -1.33765 0.06161*** CONTROL3 -0.00701 0.00102*** -1.46497 0.11455*** TOTBEDS 0.00005 3.76e-06*** 0.01388 0.00054*** EMPTYELDERLY -0.00007 0.00002*** -0.00015 0.00285 MULTI 0.00512 0.00039*** 0.13362 0.05319** ADLINDEX 0.00735 0.00017*** 0.03289 0.01819* HHI -0.00910 0.00082*** -0.71961 0.11509*** INC03 1.29e-07 3.47e-08*** 7.14e-06 4.47e-06 METHOD0 -0.00213 0.00144 -0.34340 0.19223* METHOD1 9.75e-06 0.00073 1.04717 0.11105*** METHOD2 -0.00112 0.00123 1.11955 0.17030*** METHOD4 -0.00085 0.00110 -0.01786 0.16576 CASEMIX 0.00008 0.00058 0.46302 0.08352*** CONSTANT 0.00340 0.00279 7.20068 0.38482*** N 150,571 150,537 R2 0.1209 0.1830 Prob > F 0.0000 0.0000 aThe standard errors represent the Huber-White robust standard errors corrected for intra-home correlation using the cluster option in Stata 9 S/E. Each regression contains 15892 clusters. *p < .10 **p < .05 ***p < .01 Both models include state and year fixed effects.

PAGE 159

147Table A.11 Regression Results for Outcome and Composite-Based Quality Models Facility Fixed Effects Dependent Variable Outcome-Based PROPPRESSORE Composite-Based DEFS1TOT Explanatory Variables Coefficient Standard Errora Coefficient Standard Errora CON_MORT 0.00032 0.00104 0.34217 0.14785** RATE -0.00018 0.00002*** -0.11655 0.00366* RATEMISSING -0.02165 0.00465*** -0.57068 0.63383 CONTROL2 0.00036 0.00106 0.10461 0.16863 CONTROL3 0.00244 0.00201 -0.74093 0.44590* TOTBEDS 0.00004 0.00001*** 0.02200 0.00198*** EMPTYELDERLY 0.00008 0.00003*** -0.00677 0.00362* MULTI 0.00137 0.00057** -0.20407 0.08834** ADLINDEX 0.00448 0.00015*** 0.10078 0.02063*** HHI 0.00203 0.00157 0.55473 0.21274*** INCOME 4.98e-08 9.26e-08 0.00010 0.00002*** METHOD0 -0.00191 0.00143 -0.09740 0.19073 METHOD1 0.00015 0.00072 1.18294 0.11003*** METHOD2 -0.00165 0.00123 1.32699 0.17023*** METHOD4 -0.00094 0.00111 0.01779 0.16702 CASEMIX 0.00090 0.00057 0.49594 0.08272*** CONSTANT 0.03044 0.00387*** 2.98307 0.77156*** N 150,571 150,537 R2 0.0142b 0.0441b Prob > F 0.0000 0.0000 aThe standard errors represent the Huber-White robust standard errors corrected for intrahome correlation using the cluster option in Stat a 9 S/E. Each regression contains 15892 clusters. *p < .10 **p < .05 ***p < .01 bThis value is the R2 within. Both models include year fixed effects.

PAGE 160

148Table A.12 Regression Results for Structure-Based Quality Models State Fixed Effects Dependent Variable RNS LPNS AIDES Explanatory Variables Coefficient Standard Errora Coefficient Standard Errora Coefficient Standard Errora CON_MORT 0.00638 0.00548 0.02957 0.00844*** -0.05294 0.01833*** RATE 0.00028 0.00015* 0.00081 0.00020*** -0.00203 0.00051*** RATEMISSING 0.07977 0.02639*** 0.02723 0.03697 -0.23163 0.09461** CONTROL2 0.05902 0.00472*** 0.02486 0.00587*** 0.24987 0.01392*** CONTROL3 0.03751 0.00909*** 0.02317 0.01017** 0.25495 0.02299*** TOTBEDS -0.00029 0.00003*** -0.00003 0.00004 -0.00022 0.00010** EMPTYELDERLY 0.00063 0.00012*** 0.00198 0.00021*** 0.00209 0.00046*** MULTI 0.01110 0.00311*** 0.02901 0.00408*** -0.02409 0.00974** ADLINDEX 0.02223 0.00162*** 0.02508 0.00216*** 0.10279 0.00429*** HHI -0.02315 0.00627*** -0.12060 0.00879*** -0.13185 0.02094*** INCOME 6.40e-06 3.72e-07*** -7.57e-08 4.41e-07 5.66e-06 1.03e-06*** METHOD0 0.00055 0.00915 0.00711 0.01065 -0.21510 0.02744*** METHOD1 0.01893 0.00407*** -0.00109 0.00506 -0.03343 0.01333** METHOD2 0.02780 0.00548*** 0.03998 0.01221*** -0.05716 0.02830** METHOD4 0.00961 0.00664 -0.00171 0.00865 -0.04944 0.02158*** CASEMIX 0.00019 0.00391 -0.01649 0.00453*** -0.08810 0.01114*** CONSTANT -0.24675 0. 02062*** 0.38955 0.02663*** 1.06145 0.06128*** N 150,571 150,571 150,571 R2 0.1954 0.1266 0.0944 Prob > F 0.0000 0.0000 0.0000 aThe standard errors represent the Huber-White robust standard errors corrected for intra-home correlation using the cluster option in Stata 9 S/E. Each regression contains 15892 clusters. *p < .10 **p < .05 ***p < .01 All models include state and year fixed effects.

PAGE 161

149Table A.13 Regression Results for Structure-Based Quality Models Facility Fixed Effects Dependent Variable RNS LPNS AIDES Explanatory Variables Coefficient Standard Errora Coefficient Standard Errora Coefficient Standard Errora CON_MORT 0.00998 0.00489* 0.03114 0.00705*** -0.04119 0.01646** RATE 0.00050 0.00013*** 0.00093 0.00018*** -0.00200 0.00049*** RATEMISSING 0.10221 0.02429*** 0.05815 0.03099* -0.18694 0.08930** CONTROL2 -0.00428 0.00481 0.00214 0.00703 -0.00376 0.01855 CONTROL3 0.00289 0.00805 -0.00374 0.01696 0.07022 0.04590 TOTBEDS -0.00062 0.00010*** -0.00037 0.00013*** -0.00095 0.00037*** EMPTYELDERLY 0.00163 0.00014*** 0.00307 0.00023*** 0.00462 0.00057*** MULTI 0.00068 0.00274 0.00492 0.00397 -0.00912 0.01062 ADLINDEX 0.00571 0.00090*** 0.01196 0.00126*** 0.04015 0.00330*** HHI 0.04683 0.00677*** 0.06716 0.01164*** 0.10397 0.02990*** INCOME -3.20e-06 7.40e-07*** -1.45e-06 7.04e-07** -6.55e-06 2.23e-06*** METHOD0 0.00417 0.00790 0.00382 0.00952 -0.20872 0.02595*** METHOD1 0.01398 0.00348*** 0.00132 0.00457 -0.03751 0.01214*** METHOD2 0.02154 0.00503*** 0.01061 0.00933 -0.12369 0.02245*** METHOD4 0.00950 0.00572* -0.00374 0.00777 -0.05187 0.01879*** CASEMIX -0.00016 0.00357 -0.01119 0.00427*** -0.07188 0.01076*** CONSTANT 0.35169 0.02612*** 0.44031 0.02968*** 2.20549 0.08781*** N 150,571 150,571 150,571 R2 0.0130b 0.0273b 0.0105b Prob > F 0.0000 0.0000 0.0000 aThe standard errors represent the Huber-White robust standard errors corrected for intra-home correlation using the cluster option in Stata 9 S/E. Each regression contains 15892 clusters. *p < .10 **p < .05 ***p < .01 bThis value is the R2 within. All models include year fixed effects.

PAGE 162

150Table A.14 Regression Results for Structure-Based Quality Models State Fixed Effects Dependent Variable LICNUR NURSTAFF Explanatory Variables Coefficient Standard Errora Coefficient Standard Errora CON_MORT 0.03481 0.00933*** -0.00769 0.02114 RATE 0.00111 0.00021*** -0.00066 0.00055 RATEMISSING 0.12539 0.04245*** -0.05558 0.10582 CONTROL2 0.08555 0.00639*** 0.34693 0.01591*** CONTROL3 0.08903 0.01234*** 0.39290 0.02981*** TOTBEDS -0.00038 0.00005*** -0.00075 0.00011*** EMPTYELDERLY 0.00241 0.00021*** 0.00429 0.00052*** MULTI 0.00660 0.00426 -0.07697 0.01060*** ADLINDEX 0.03611 0.00288*** 0.11776 0.00571*** HHI -0.10736 0.00907*** -0.18517 0.02210*** INCOME 4.64e-06 4.89e-07*** 7.79e-06 1.10e-06*** METHOD0 -0.00234 0.01300 -0.22118 0.03186*** METHOD1 0.00896 0.00554 -0.03159 0.01429** METHOD2 0.04853 0.01068*** -0.02749 0.02894 METHOD4 0.00476 0.00962 -0.03836 0.02250* CASEMIX -0.01297 0.00288*** -0.08851 0.01241*** CONSTANT 0.28647 0.03166*** 1.57380 0.07170*** N 150,571 150,571 R2 0.1127 0.1193 Prob > F 0.0000 0.0000 aThe standard errors represent the Huber-White robust standard errors corrected for intrahome correlation using the cluster option in Stat a 9 S/E. Each regression contains 15892 clusters. *p < .10 **p < .05 ***p < .01 Both models include state and year fixed effects.

PAGE 163

151Table A.15 Regression Results for Structure-Based Quality Models Facility Fixed Effects Dependent Variable LICNUR NURSTAFF Explanatory Variables Coefficient Standard Errora Coefficient Standard Errora CON_MORT 0.03915 0.00788*** 0.00704 0.01845 RATE 0.00134 0.00019*** -0.00052 0.00052 RATEMISSING 0.15351 0.03558*** -0.01243 0.10158 CONTROL2 0.00402 0.00748 0.00929 0.02027 CONTROL3 0.01347 0.01682 0.10864 0.05227** TOTBEDS -0.00103 0.00015*** -0.00213 0.00041*** EMPTYELDERLY 0.00430 0.00026*** 0.00833 0.00064*** MULTI -0.00309 0.00427 -0.02712 0.01138** ADLINDEX 0.01542 0.00149*** 0.05049 0.00379*** HHI 0.10038 0.01351*** 0.18436 0.03242*** INCOME -4.01e-06 9.45e-07*** -9.07e-06 2.52e-06*** METHOD0 0.00311 0.01167 -0.20223 0.02967*** METHOD1 0.00934 0.00506* -0.03033 0.01311** METHOD2 0.02981 0.00626*** -0.08105 0.02468*** METHOD4 0.00582 0.00880 -0.03369 0.02013* CASEMIX -0.00810 0.00149*** -0.06977 0.01193*** CONSTANT 0.76865 0.03599*** 2.89170 0.09794*** N 150,571 150,571 R2 0.0174b 0.0109b Prob > F 0.0000 0.0000 aThe standard errors represent the Huber-White robust standard errors corrected for intra-home correlation using the cluster option in Stata 9 S/E. Each regression contains 15892 clusters. *p < .10 **p < .05 ***p < .01 bThis value is the R2 within. Both models include year fixed effects.

PAGE 164

152Table A.16 Regression Results for Access Model Dependent Variable PCTMCAID State Fixed Effects Facility Fixed Effects Explanatory Variables Coefficient Standard Errora Coefficient Standard Errora CON_MORT 0.95788 0.34786*** 1.55143 0.29731*** RATE -0.00507 0.00777 0.00302 0.00616 RATEMISSING -0.42287 1.64902 1.78212 1.27866 CONTROL2 -9.38626 0.34208*** -0.91782 0.27432*** CONTROL3 0.79328 0.60256 0.92467 0.65133 TOTBEDS 0.01952 0.00235*** -0.01413 0.00460*** EMPTYELDERLY -0.04217 0.01136*** 0.02472 0.00735*** MULTI -0.96125 0.22794*** -0.17373 0.14230 ADLINDEX -1.10705 0.09050*** 0.15339 0.03652*** HHI 0.70194 0.46441 0.39247 0.35295 INCOME -0.00038 0.00003*** 0.00012 0.00003*** METHOD0 -1.91870 0.48414*** -1.97860 0.40190*** METHOD1 -0.22563 0.23044 -0.40020 0.18285** METHOD2 -1.40048 0.37974*** -0.78225 0.31524** METHOD4 0.21310 0.36760 0.62293 0.29102** CASEMIX -0.08600 0.17625 -0.09821 0.14293 CONSTANT 96.13814 1.30675*** 64.50688 1.19857*** N 150,571 150,571 R2 0.2016 0.0060b Prob > F 0.0000 0.0000 aThe standard errors represent the Huber-White robust standard errors corrected for intra-home correlation using the cluster option in Stata 9 S/E. Each regression contains 15892 clusters. *p < .10 **p < .05 ***p < .01 bThis value is the R2 within. The state fixed effects model include s state and year fixed effects. The facility fixed effects model includes year fixed effects.

PAGE 165

153Table A.17 Regression Results fo r Heavy-Care Access Model Dependent Variable ADLINDEX State Fixed Effects Facility Fixed Effects Explanatory Variables Coefficient Standard Errora Coefficient Standard Errora CON_MORT -0.12390 0.02821*** -0.10086 0.02636*** RATE 0.00109 0.00069 0.00081 0.00064 RATEMISSING -0.06513 0.13310 -0.06249 0.13122 CONTROL2 0.06312 0.02116*** 0.02638 0.02532 CONTROL3 0.32173 0.04918*** 0.08470 0.06216 TOTBEDS 0.00210 0.00017*** -0.00069 0.00036* EMPTYELDERLY -0.00769 0.00085*** 0.00070 0.00070 MULTI 0.06928 0.01599*** -0.03014 0.01354** HHI -0.11673 0.03379*** 0.05923 0.03650 INCOME -1.24e-07 1.70e-06 -1.12e-06 2.52e-06 METHOD0 0.04236 0.04044 0.02111 0.03785 METHOD1 0.06795 0.01859*** 0.03508 0.01670** METHOD2 -0.04475 0.03208 -0.05763 0.03013* METHOD4 -0.11659 0.03047*** -0.09409 0.02806*** CASEMIX 0.20506 0.01521*** 0.20339 0.01400*** CONSTANT 9.24604 0.07997*** 9.34348 0.09638*** N 150,571 150,571 R2 0.2424 0.0556b Prob > F 0.0000 0.0000 aThe standard errors represent the Huber-White robust standard errors corrected for intra-home correlation using the cluster option in Stata 9 S/E. Each regression contains 15892 clusters. *p < .10 **p < .05 ***p < .01 bThis value is the R2 within. The state fixed effects model include s state and year fixed effects. The facility fixed effects model includes year fixed effects.

PAGE 166

154Table A.18 Regression Results for Process-Based Quality Models without Method State-Fixed Effects Dependent Variable PROPCATHETER PROPPARENTERAL PROPMOBLREST Explanatory Variables Coefficient Standard Errora Coefficie nt Standard Errora Coefficient Standard Errora CON_MORT -0.00869 0.00119*** -0.00116 0.00113 -0.05192 0.00314*** RATE -0.00002 0.00003 -0.00003 0.00003 -0.00072 0.00007*** RATEMISSING 0.00451 0.00508 -0.00905 0.00455** -0.08074 0.01637*** CONTROL2 -0.01008 0.00072*** -0.01040 0.00093*** 0.00309 0.00151** CONTROL3 -0.00773 0.00141*** -0.00241 0.00252 0.01420 0.00339*** TOTBEDS 0.00007 5.74e-06*** 0.00008 8.03e-06*** 0.00001 0.00001 EMPTYELDERLY 0.00005 0.00003 -0.00019 0.00004*** 0.00015 0.00007** MULTI 0.00425 0.00056*** 0.00304 0.00067*** -0.01205 0.00118*** ADLINDEX 0.01006 0.00027*** 0.01605 0.00068*** 0.02202 0.00050*** HHI -0.00443 0.00119*** -0.02633 0.00160*** 0.01191 0.00258*** INCOME -2.45e-07 4.98e-08*** 9.25e-08 7.53e-08 -8.94e-08 1.06e-07 CASEMIX 0.00203 0.00063*** -0.00361 0.00058*** 0.00571 0.00164*** CONSTANT -0.00267 0.00365 -0.09779 0.00680*** 0.10298 0.00798** N 150,571 150,571 150,571 R2 0.1446 0.2939 0.2339 Prob > F 0.0000 0.0000 0.0000 aThe standard errors represent the Huber-White robust standard errors corrected for intra-home correlation using the cluster option in Stata 9 S/E. Each regression contains 15892 clusters. *p < .10 **p < .05 ***p < .01 All models include state and year fixed effects.

PAGE 167

155Table A.19 Regression Results for Process-Based Quality Models without Method Facility-Fixed Effects Dependent Variable PROPCATHETER PROPPARENTERAL PROPMOBLREST Explanatory Variables Coefficient Standard Errora Coefficient Standard Errora Coefficient Standard Errora CON_MORT -0.00948 0.00118*** -0.00239 0.00092*** -0.05112 0.00316*** RATE -0.00001 0.00003 -0.00003 0.00002 -0.00076 0.00008*** RATEMISSING 0.00737 0.00490 -0.01109 0.00409*** -0.09399 0.01537*** CONTROL2 -0.00287 0.00112** -0.00083 0.00093 0.01021 0.00325*** CONTROL3 -0.00088 0.00282 -0.00088 0.00163 0.01249 0.00793 TOTBEDS 0.00002 0.00002 54.51e-06 0.00001 0.00005 0.00004 EMPTYELDERLY 0.00017 0.00004*** 0.00003 0.00002 0.00040 0.00008*** MULTI -0.00113 0.00061* 0.00051 0.00051 -0.00500 0.00164*** ADLINDEX 0.00566 0.00016*** 0.00666 0.00016*** 0.01406 0.00046*** HHI 0.00551 0.00160*** 0.00162 0.00130 0.02269 0.00446*** INCOME 5.55e-07 1.36e-07*** -6.20e-07 1.70e-07*** -3.10e-06 5.41e-07*** CASEMIX 0.00274 0.00060*** -0.00161 0.00049*** 0.00742 0.00167*** CONSTANT 0.01175 0.00490** -0.00020 0.00534 0.21883 0.01670*** N 150,571 150,571 150,571 R2 0.0529b 0.0779b 0.1553b Prob > F 0.0000 0.0000 0.0000 aThe standard errors represent the Huber-White robust standard errors corrected for intra-home correlation using the cluster option in Stata 9 S/E. Each regression contains 15892 clusters. *p < .10 **p < .05 ***p < .01 bThis value is the R2 within. All models include year fixed effects.

PAGE 168

156Table A.20 Regression Results for Outcome and Composite-Based Quality Models without Method State-Fixed Effects Dependent Variable Outcome-Based PROPPRESSORE Composite-Based DEFS1TOT Explanatory Variables Coefficient Standard Errora Coefficient Standard Errora CON_MORT 0.00162 0.00101 0.40607 0.13719*** RATE -0.00017 0.00002*** -0.00955 0.00360*** RATEMISSING -0.02197 0.00459*** -0.53315 0.61513 CONTROL2 -0.00751 0.00048*** -1.33853 0.06162*** CONTROL3 -0.00702 0.00102*** -1.46728 0.11458*** TOTBEDS 0.00005 3.76e-06*** 0.01389 0.00054*** EMPTYELDERLY -0.00007 0.00002*** -0.00037 0.00285 MULTI 0.00512 0.00039*** 0.13460 0.05320** ADLINDEX 0.00735 0.00017*** 0.03510 0.01817* HHI -0.00910 0.00082*** -0.71115 0.11508*** INC03 1.30e-07 3.47e-08*** 7.59e-06 4.47e-06* CASEMIX 0.00021 0.00056 0.33073 0.08227*** CONSTANT 0.00292 0.00259 8.36010 0.35627*** N 150,571 150,537 R2 0.1209 0.1821 Prob > F 0.0000 0.0000 aThe standard errors represent the Huber-White robust standard errors corrected for intra-home correlation using the cluster option in Stata 9 S/E. Each regression contains 15892 clusters. *p < .10 **p < .05 ***p < .01 Both models include state and year fixed effects.

PAGE 169

157Table A.21 Regression Results for Outcome and Composite-Based Quality Models without Method Facility-Fixed Effects Dependent Variable Outcome-Based PROPPRESSORE Composite-Based DEFS1TOT Explanatory Variables Coefficient Standard Errora Coefficient Standard Errora CON_MORT 0.00111 0.00098 0.56605 0.13340*** RATE -0.00018 0.00002*** -0.00925 0.00356*** RATEMISSING -0.02056 0.00461*** -0.53416 0.62653 CONTROL2 0.00034 0.00106 0.09161 0.16870 CONTROL3 0.00238 0.00201 -0.73398 0.45020 TOTBEDS 0.00004 0.00001*** 0.02153 0.00197*** EMPTYELDERLY 0.00008 0.00003*** -0.00757 0.00362** MULTI 0.00139 0.00057** -0.20468 0.08854** ADLINDEX 0.00449 0.00015*** 0.10394 0.02064*** HHI 0.00199 0.00156 0.57849 0.21298*** INCOME 6.00e-08 9.29e-08 0.00011 0.00003*** CASEMIX 0.00104 0.00056* 0.33201 0.08151*** CONSTANT 0.02885 0.00379*** 3.56265 0.79317*** N 150,571 150,537 R2 0.0142b 0.0425b Prob > F 0.0000 0.0000 aThe standard errors represent the Huber-White robust standard errors corrected for intrahome correlation using the cluster option in Stat a 9 S/E. Each regression contains 15892 clusters. *p < .10 **p < .05 ***p < .01 bThis value is the R2 within. Both models include year fixed effects.

PAGE 170

158Table A.22 Regression Results for Structure-Based Quality Models without Method State-Fixed Effects Dependent Variable RNS LPNS AIDES Explanatory Variables Coefficient Standard Errora Coefficient Standard Errora Coefficient Standard Errora CON_MORT 0.00832 0.00567 0.02713 0.00750*** 0.01921 0.01565 RATE 0.00023 0.00015 0.00073 0.00019*** -0.00109 0.00050** RATEMISSING 0.08015 0.02618*** 0.01457 0.03639 -0.12940 0.092981 CONTROL2 0.05881 0.00472*** 0.02487 0.00587*** 0.24992 0.01392*** CONTROL3 0.03749 0.00808*** 0.02326 0.01017** 0.25434 0.02300*** TOTBEDS -0.00029 0.00003*** -0.00003 0.00004 -0.00022 0.00010** EMPTYELDERLY 0.00063 0.00012*** 0.00195 0.00021*** 0.00215 0.00046*** MULTI 0.01107 0.00311*** 0.02886 0.00409*** -0.02387 0.00974** ADLINDEX 0.02225 0.00162*** 0.02503 0.00216*** 0.10275 0.00422*** HHI -0.02297 0.00627*** -0.12038 0.00878*** -0.13214 0.02093*** INCOME 6.41e-06 3.72e-07*** -8.50e-08 4.41e-07 5.67e-06 1.03e-06*** CASEMIX -0.00246 0.00382 -0.01938 0.00440*** -0.07356 0.01064*** CONSTANT -0.22148 0. 01966*** 0.40798 0.02593*** 1.06145 0.06128*** N 150,571 150,571 150,571 R2 0.1953 0.1264 0.0940 Prob > F 0.0000 0.0000 0.0000 aThe standard errors represent the Huber-White robust standard errors corrected for intra-home correlation using the cluster option in Stata 9 S/E. Each regression contains 15892 clusters. *p < .10 **p < .05 ***p < .01 All models include state and year fixed effects.

PAGE 171

159Table A.23 Regression Results for Structure-Based Quality Models without Method Facility-Fixed Effects Dependent Variable RNS LPNS AIDES Explanatory Variables Coefficient Standard Errora Coefficient Standard Errora Coefficient Standard Errora CON_MORT 0.00971 0.00465** 0.03047 0.00636*** 0.02952 0.01448** RATE 0.00046 0.00013*** 0.00089 0.00017*** -0.00104 0.00047** RATEMISSING 0.10148 0.02409*** 0.05301 0.03048* -0.07747 0.08755 CONTROL2 -0.00434 0.00481 0.00220 0.00703 -0.00406 0.01859 CONTROL3 0.00310 0.00805 -0.00337 0.01698 0.06594 0.04602 TOTBEDS -0.00063 0.00010*** -0.00037 0.00013*** -0.00090 0.00037** EMPTYELDERLY 0.00161 0.00014*** 0.00306 0.00023*** 0.00489 0.00058*** MULTI 0.00057 0.00275 0.00482 0.00397 -0.00766 0.01062 ADLINDEX 0.00572 0.00090*** 0.01196 0.00126*** 0.04024 0.00330*** HHI 0.04725 0.00676*** 0.06745 0.01164*** 0.10258 0.02989*** INCOME -3.15e-06 7.33e-07*** -1.47e-06 7.03e-07** -6.39e-06 2.21e-06*** CASEMIX -0.00221 0.00338 -0.01234 0.00412*** -0.05326 0.01065*** CONSTANT 0.36358 0.02548*** 0.44642 0.02882*** 2.04959 0.08531*** N 150,571 150,571 150,571 R2 0.0128b 0.0272b 0.0097b Prob > F 0.0000 0.0000 0.0000 aThe standard errors represent the Huber-White robust standard errors corrected for intra-home correlation using the cluster option in Stata 9 S/E. Each regression contains 15892 clusters. *p < .10 **p < .05 ***p < .01 bThis value is the R2 within. All models include year fixed effects.

PAGE 172

160Table A.24 Regression Results for Structure-Based Quality Models without Method State-Fixed Effects Dependent Variable LICNUR NURSTAFF Explanatory Variables Coefficient Standard Errora Coefficient Standard Errora CON_MORT 0.03653 0.00850*** 0.06552 0.01778*** RATE 0.00106 0.00021*** 0.00029 0.00053 RATEMISSING 0.11969 0.04206*** 0.04740 0.10447 CONTROL2 0.08555 0.00639*** 0.34698 0.01591*** CONTROL3 0.08906 0.01234*** 0.39228 0.02982*** TOTBEDS -0.00038 0.00005*** -0.00075 0.00011*** EMPTYELDERLY 0.00238 0.00021*** 0.00433 0.00052*** MULTI 0.00646 0.00426 -0.07686 0.01060*** ADLINDEX 0.03608 0.00287*** 0.11767 0.00571*** HHI -0.10708 0.00907*** -0.18531 0.02209*** INCOME 4.64e-06 4.89e-07*** 7.79e-06 1.10e-06*** CASEMIX -0.01629 0.00514*** -0.07500 0.01182*** CONSTANT 0.31231 0.03110*** 1.46827 0.06787*** N 150,571 150,571 R2 0.1126 0.1190 Prob > F 0.0000 0.0000 aThe standard errors represent the Huber-White robust standard errors corrected for intrahome correlation using the cluster option in Stat a 9 S/E. Each regression contains 15892 clusters. *p < .10 **p < .05 ***p < .01 Both models include state and year fixed effects.

PAGE 173

161Table A.25 Regression Results for Structure-Based Quality Models without Method Facility-Fixed Effects Dependent Variable LICNUR NURSTAFF Explanatory Variables Coefficient Standard Errora Coefficient Standard Errora CON_MORT 0.03885 0.00689*** 0.07442 0.01590*** RATE 0.00129 0.00019*** 0.00040 0.00050 RATEMISSING 0.14931 0.03518*** 0.09160 0.10011 CONTROL2 0.00414 0.00748 0.00934 0.02032 CONTROL3 0.01409 0.01682 0.10533 0.05243** TOTBEDS -0.00103 0.00015*** -0.00209 0.00041*** EMPTYELDERLY 0.00425 0.00026*** 0.00853 0.00064*** MULTI -0.00333 0.00427 -0.02611 0.01139** ADLINDEX 0.01539 0.00149*** 0.05048 0.00379*** HHI 0.10117 0.01351*** 0.18417 0.03240*** INCOME -4.03e-06 9.45e-07*** -9.03e-06 2.52e-06*** CASEMIX -0.01043 0.00476** -0.05356 0.01128*** CONSTANT 0.78275 0.03519*** 2.75653 0.09567*** N 150,571 150,571 R2 0.0173b 0.0103b Prob > F 0.0000 0.0000 aThe standard errors represent the Huber-White robust standard errors corrected for intra-home correlation using the cluster option in Stata 9 S/E. Each regression contains 15892 clusters. *p < .10 **p < .05 ***p < .01 bThis value is the R2 within. Both models include year fixed effects.

PAGE 174

162Table A.26 Regression Results for Access Model without Method Dependent Variable PCTMCAID State-Fixed Effects Facility-Fixed Effects Explanatory Variables Coefficient Standard Errora Coefficient Standard Errora CON_MORT 1.54133 0.33592*** 2.07968 0.28453*** RATE 0.00561 0.00768 0.01424 0.00612** RATEMISSING 0.88529 1.64061 3.06620 1.27895** CONTROL2 -9.38668 0.34207*** -0.91092 0.27417*** CONTROL3 0.78491 0.60253 0.88465 0.65033 TOTBEDS 0.01949 0.00235*** -0.01368 0.00459*** EMPTYELDERLY -0.04113 0.01135*** 0.02621 0.00735*** MULTI -0.95677 0.22788*** -0.16680 0.14229 ADLINDEX -1.10707 0.09045*** 0.15052 0.03656*** HHI 0.69350 0.46433 0.39209 0.35261 INCOME -0.00038 0.00003*** 0.00012 0.00003*** CASEMIX 0.13168 0.17320 0.12142 0.14065 CONSTANT 94.7137 1.27952*** 63.19044 1.18258*** N 150,571 150,571 R2 0.2015 0.0053b Prob > F 0.0000 0.0000 aThe standard errors represent the Huber-White robust standard errors corrected for intra-home correlation using the cluster option in Stata 9 S/E. Each regression contains 15892 clusters. *p < .10 **p < .05 ***p < .01 bThis value is the R2 within. The state fixed effects model include s state and year fixed effects. The facility fixed effects model includes year fixed effects.

PAGE 175

163Table A.27 Regression Results for Heavy-Ca re Access Model without Method Dependent Variable ADLINDEX State-Fixed Effects Facility-Fixed Effects Explanatory Variables Coefficient Standard Errora Coefficient Standard Errora CON_MORT -0.11325 0.02709*** -0.09062 0.02483*** RATE 0.00060 0.00068 0.00053 0.00063 RATEMISSING -0.10095 0.13164 -0.08152 0.13003 CONTROL2 0.06315 0.02117*** 0.02452 0.02536 CONTROL3 0.32201 0.04919*** 0.08328 0.06255 TOTBEDS 0.00210 0.00017*** -0.00071 0.00036** EMPTYELDERLY -0.00762 0.00085*** 0.00086 0.00070 MULTI 0.06996 0.01599*** -0.02900 0.01356** HHI -0.11670 0.03380*** 0.05660 0.03657 INCOME -6.10e-08 1.70e-06 -2.80e-07 2.50e-06 CASEMIX 0.19390 0.01489*** 0.19775 0.01369*** CONSTANT 9.31124 0.07808*** 9.33862 0.09448*** N 150,571 150,571 R2 0.2423 0.0551b Prob > F 0.0000 0.0000 aThe standard errors represent the Huber-White robust standard errors corrected for intra-home correlation using the cluster option in Stata 9 S/E. Each regression contains 15892 clusters. *p < .10 **p < .05 ***p < .01 bThis value is the R2 within. The state fixed effects model include s state and year fixed effects. The facility fixed effects model includes year fixed effects.

PAGE 176

164Table A.28 Regression Results for Process-Based Quality Models Most Restrictive Sample State-Fixed Effects Dependent Variable PROPCATHETER PROPPARENTERAL PROPMOBLREST Explanatory Variables Coefficient Standard Errora Coefficient Standard Errora Coefficient Standard Errora CON_MORT 0.00460 0.00219** 0.00157 0.00258 -0.05470 0.00724*** RATE -0.00013 0.00004*** 0.00010 0.00004*** -0.00052 0.00011*** RATEMISSING -0.00940 0. 00636 0.00403 0.00614 -0.05932 0.01840*** CONTROL2 -0.00813 0.00097*** -0.01004 0.00135*** 0.00050 0.00219 CONTROL3 -0.00253 0. 00214 0.00092 0.00421 0.01412 0.00511*** TOTBEDS 0.00007 7.47e06*** 0.00007 0.00001*** 0.00001 0.00002 EMPTYELDERLY 0.00058 0.00012*** 0.00051 0.00015*** 0.00023 0.00029 MULTI 0.00442 0.00076*** 0.00166 0.00106 -0.01235 0.00178*** ADLINDEX 0.00958 0.00038*** 0.01805 0.00096*** 0.02253 0.00071*** HHI -0.00436 0.00162*** -0.02604 0.00247*** 0.01167 0.00367*** INCOME -2.11e-07 5.99e-08*** -1.93e08 9.83e-08 2.29e-07 1.33e-07* METHOD0 0.01183 0.00249*** 0.00455 0.00258* 0.01899 0.00810** METHOD1 0.00779 0.00116*** -0.00313 0.00116*** 0.02467 0.00289*** METHOD2 0.00241 0.00447 0.01799 0.00602*** 0.01887 0.01505 METHOD4 0.00761 0.00181*** 0.00306 0.00178* -0.00120 0.00447 CASEMIX 0.00280 0.00097*** -0.00431 0.00101*** 0.02536 0.00248*** CONSTANT 0.00028 0.00836 -0.07501 0.01266*** 0.10057 0.01931*** N 75,904 75,904 75,904 R2 0.1500 0.3093 0.2488 Prob > F 0.0000 0.0000 0.0000 aThe standard errors represent the Huber-White robust standard errors corrected for intra-home correlation using the cluster optio n in Stata 9 S/E. Each regression contains 8089 clusters. *p < .10 **p < .05 ***p < .01 All models include state and year fixed effects.

PAGE 177

165Table A.29 Regression Results for Process-Based Quality Models Most Restrictive Sample Facility-Fixed Effects Dependent Variable PROPCATHETER PROPPARENTERAL PROPMOBLREST Explanatory Variables Coefficient Standard Errora Coefficient Standard Errora Coefficient Standard Errora CON_MORT 0.00153 0.00206 -0.00203 0.00224 -0.05707 0.00715*** RATE -0.00009 0.00004** 0.00008 0.00003** -0.00566 0.00011*** RATEMISSING -0.00226 0. 00630 0.00077 0.00544 -0.06747 0.01859*** CONTROL2 -0.00250 0.00147* -0.00231 0.00135* 0.00222 0.00469 CONTROL3 0.00096 0.00443 -0 .00008 0.00304 -0.00959 0.01371 TOTBEDS 8.66e-06 0.00003 2.20e-06 0.00003 0.00011 0.00007 EMPTYELDERLY 0.00045 0.00010*** 0.00006 0.00009 2.09e-06 0.00031 MULTI -0.00074 0.00088 0.00097 0.00080 -0.00510 0.00249** ADLINDEX 0.00447 0.00022*** 0.00705 0.00024*** 0.01400 0.00065*** HHI 0.00591 0.00210*** 0.00030 0.00184 0.00263 0.00647 INCOME 6.32e-07 1.98e-07*** -6.86e-07 2.51e-07*** -2.89e-06 6.93e-07*** METHOD0 0.01147 0.00240*** 0.00528 0.00233** 0.01965 0.00805** METHOD1 0.00701 0.00113*** -0.00395 0.00094*** 0.02512 0.00288*** METHOD2 0.00421 0.00439 0.01335 0.00582** 0.02095 0.01515 METHOD4 0.00621 0.00178*** 0.00138 0.00156 -0.00115 0.00447 CASEMIX 0.00425 0.00092*** -0.00709 0.00083 0.02815 0.00251*** CONSTANT 0.02018 0.00775*** -0.01031 0.00871 0.20482 0.02576*** N 75,904 75,904 75,904 R2 0.0508b 0.0910b 0.1881b Prob > F 0.0000 0.0000 0.0000 aThe standard errors represent the Huber-White robust standard errors corrected for intra-home correlation using the cluster optio n in Stata 9 S/E. Each regression contains 8089 clusters. *p < .10 **p < .05 ***p < .01 bThis value is the R2 within. All models include year fixed effects.

PAGE 178

166Table A.30 Regression Results for Outcome and Composite-Based Quality Models Most Restrictive Sample State-Fixed Effects Dependent Variable Outcome-Based PROPPRESSORE Composite-Based DEFS1TOT Explanatory Variables Coefficient Standard Errora Coefficient Standard Errora CON_MORT 0.00424 0.00210** -0.94130 0.34912*** RATE -0.00015 0.00003*** -0.04641 0.00477*** RATEMISSING -0.01783 0.00593*** -5.2114 0.73940*** CONTROL2 -0.00746 0.00069*** -1.28071 0.08418*** CONTROL3 -0.00773 0.00156*** -1.33318 0.17198*** TOTBEDS 0.00005 5.28e-06*** 0.01491 0.00073*** EMPTYELDERLY 0.00036 0.00010*** 0.01668 0.01348 MULTI 0.00520 0.00058*** 0.05069 0.07705 ADLINDEX 0.00756 0.00024*** 0.07397 0.02582*** HHI -0.00815 0.00115*** -0.51059 0.15554*** INCOME 2.54e-08 4.25e-08 0.00002 5.31e-06*** METHOD0 0.00167 0.00241 -1.66448 0.37454*** METHOD1 0.00171 0.00113 -0.19344 0.17132 METHOD2 0.00660 0.00387* 1.35087 0.67555** METHOD4 0.00044 0.00157 -0.87429 0.24996*** CASEMIX 0.00148 0.00091 0.42792 0.13086*** CONSTANT 0.00184 0.00554 9.29651 0.66411*** N 75,904 75,890 R2 0.1258 0.2310 Prob > F 0.0000 0.0000 aThe standard errors represent the Huber-White robust standard errors corrected for intrahome correlation using the cluster option in Stat a 9 S/E. Each regression contains 8089 clusters. *p < .10 **p < .05 ***p < .01 Both models include state and year fixed effects.

PAGE 179

167Table A.31 Regression Results for Outcome and Composite-Based Quality Models Most Restrictive Sample Facility-Fixed Effects Dependent Variable Outcome-Based PROPPRESSORE Composite-Based DEFS1TOT Explanatory Variables Coefficient Standard Errora Coefficient Standard Errora CON_MORT 0.00152 0.00200 -0.62845 0.35597* RATE -0.00015 0.00003*** -0.04333 0.00475*** RATEMISSING -0.01595 0.00601*** -4.86745 0.75864*** CONTROL2 0.00056 0.00158 0.29073 0.26288 CONTROL3 0.00245 0.00308 0.07535 0.81683 TOTBEDS 0.00005 0.00002** 0.02368 0.00333*** EMPTYELDERLY 0.00322 0.00010*** 0.00384 0.01354 MULTI 0.00144 0.00085* -0.37322 0.13841*** ADLINDEX 0.00405 0.00021*** 0.06487 0.03035** HHI 0.00275 0.00204 0.54486 0.29289* INCOME 7.46e-08 1.15e-07 0.00009 0.00003*** METHOD0 0.00064 0.00232 -1.33310 0.38046*** METHOD1 0.00144 0.00111 -0.00531 0.17215 METHOD2 0.00605 0.00107 1.68600 0.67179** METHOD4 -0.00038 0.00158 -0.82442 0.25577*** CASEMIX 0.00249 0.00087*** 0.44276 0.13181*** CONSTANT 0.03010 0.00581*** 7.00379 1.11310*** N 75,904 75,890 R2 0.0128b 0.0384b Prob > F 0.0000 0.0000 aThe standard errors represent the Huber-White robust standard errors corrected for intrahome correlation using the cluster option in Stat a 9 S/E. Each regression contains 8089 clusters. *p < .10 **p < .05 ***p < .01 bThis value is the R2 within. Both models include year fixed effects.

PAGE 180

168Table A.32 Regression Results for Structure-Based Quality Models Most Restrictive Sample State-Fixed Effects Dependent Variable RNS LPNS AIDES Explanatory Variables Coefficient Standard Errora Coefficient Standard Errora Coefficient Standard Errora CON_MORT 0.02150 0.01254* -0.00421 0.01615 -0.00521 0.03494 RATE 0.00038 0.00020* 0.00126 0.00025*** 0.00015 0.00063 RATEMISSING 0.07942 0. 03237** 0.09096 0.04279** 0.02671 0.12215 CONTROL2 0.04987 0.00621*** 0.01504 0.00711** 0.24503 0.01510*** CONTROL3 0.03924 0.01052*** 0.03386 0.01349** 0.30788 0.02922*** TOTBEDS -0.00042 0.00004*** -0.00015 0.00005*** -0.00052 0.00010*** EMPTYELDERLY 0.00546 0.00077*** 0.00868 0.00111*** 0.01348 0.00271*** MULTI 0.01280 0.00436*** 0.02266 0.00517*** -0.03736 0.01107*** ADLINDEX 0.01839 0.00204*** 0.01958 0.00293*** 0.08297 0.00475*** HHI -0.01733 0.00797* -0.10584 0.01233*** -0.05532 0.02404** INCOME 5.05e-06 4.07e-07*** -1.10e-06 4.77e-07** 3.33e-06 9.78e-07*** METHOD0 -0.01118 0.01567 -0.02707 0.01798 -0.15958 0.04261*** METHOD1 0.00803 0.00702 -0.00684 0.00795 -0.05563 0.02065*** METHOD2 -0.00216 0.01415 0.00336 0.03250 -0.09861 0.09063 METHOD4 0.00185 0.00955 -0.00537 0.01127 -0.04464 0.02801 CASEMIX -0.00652 0.00761 -0.03475 0.00819*** -0.09060 0.01702*** CONSTANT -0.04645 0. 03496 0.33916 0.07850*** 1.33449 0.17449*** N 75,904 75,904 75,904 R2 0.1675 0.1390 0.0922 Prob > F 0.0000 0.0000 0.0000 aThe standard errors represent the Huber-White robust standard errors corrected for intra-home correlation using the cluster optio n in Stata 9 S/E. Each regression contains 8089 clusters. *p < .10 **p < .05 ***p < .01 All models include state and year fixed effects.

PAGE 181

169Table A.33 Regression Results for Structure-Based Quality Models Most Restrictive Sample Facility-Fixed Effects Dependent Variable RNS LPNS AIDES Explanatory Variables Coefficient Standard Errora Coefficient Standard Errora Coefficient Standard Errora CON_MORT 0.00468 0.00986 -0.02867 0.01518* -0.02507 0.03303 RATE 0.00059 0.00017*** 0.00161 0.00023*** 0.00095 0.00061 RATEMISSING 0.10579 0. 03030*** 0.14160 0.04085*** 0.15987 0.11567 CONTROL2 -0.00755 0. 00727 0.00388 0.01000 0.01446 0.02342 CONTROL3 -0.01929 0.01188 -0.00514 0.02073 -0.04004 0.05818 TOTBEDS -0.00093 0.00020*** -0.00032 0.00022 -0.00169 0.00068** EMPTYELDERLY 0.00592 0.00066*** 0.00789 0.00094*** 0.01594 0.00259*** MULTI -0.00055 0.00398 0.00668 0.00584 -0.01841 0.01456 ADLINDEX 0.00378 0.00119*** 0.00861 0.00152*** 0.02963 0.00340*** HHI 0.05282 0.01053*** 0.06519 0.01819*** 0.12331 0.04984** INCOME -3.86e-06 1.13e-06*** 2.76e-07 9.11e-07 -3.68e-06 2.40e-06 METHOD0 -0.02156 0.01294* -0.05077 0.01729*** -0.16466 0.04107*** METHOD1 0.00528 0.00637 -0.00892 0.00763 -0.05429 0.02066*** METHOD2 -0.01374 0.01445 0.00056 0.02705 -0.10632 0.08343 METHOD4 0.00845 0.00873 -0.01852 0.01086* -0.04671 0.02744* CASEMIX -0.01382 0.00708* -0.03321 0.00801*** -0.08781 0.01711*** CONSTANT 0.47167 0.04296*** 0.39456 0.04326*** 2.13435 0.11889*** N 75,904 75,904 75,904 R2 0.0146b 0.0318b 0.0124b Prob > F 0.0000 0.0000 0.0000 aThe standard errors represent the Huber-White robust standard errors corrected for intra-home correlation using the cluster optio n in Stata 9 S/E. Each regression contains 8089 clusters. *p < .10 **p < .05 ***p < .01 bThis value is the R2 within. All models include year fixed effects.

PAGE 182

170Table A.34 Regression Results for Structure-Based Quality Models Most Restrictive Sample State-Fixed Effects Dependent Variable LICNUR NURSTAFF Explanatory Variables Coefficient Standard Errora Coefficient Standard Errora CON_MORT 0.01867 0.01847 0.02890 0.03945 RATE 0.00161 0.00027*** 0.00170 0.00675** RATEMISSING 0.19055 0.04897*** 0.24058 0.13770* CONTROL2 0.07804 0.00814*** 0.33726 0.01819*** CONTROL3 0.10005 0.01446*** 0.45467 0.03623*** TOTBEDS -0.00058 0.00006*** -0.00121 0.00012*** EMPTYELDERLY 0.01186 0.00119*** 0.02148 0.00268*** MULTI 0.00162 0.00579 -0.09267 0.01289*** ADLINDEX 0.03048 0.00406*** 0.09855 0.00728*** HHI -0.09248 0.01268*** -0.10792 0.02783*** INCOME 2.83e-06 5.47e-07***4.55e-06 1.14e-06*** METHOD0 -0.04651 0.02134** -0.20704 0.04935*** METHOD1 0.00044 0.00936 -0.04879 0.02310** METHOD2 -0.00148 0.02642 -0.10236 0.09013 METHOD4 -0.00453 0.01351 -0.03657 0.03118 CASEMIX -0.03662 0.00996*** -0.11244 0.01958*** CONSTANT 0.35313 0.07087*** 1.75655 0.17854*** N 75,904 75,904 R2 0.0854 0.1063 Prob > F 0.0000 0.0000 aThe standard errors represent the Huber-White robust standard errors corrected for intrahome correlation using the cluster option in Stat a 9 S/E. Each regression contains 8089 clusters. *p < .10 **p < .05 ***p < .01 Both models include state and year fixed effects.

PAGE 183

171 Table A.35 Regression Results for Structure-Based Quality Models Most Restrictive Sample Facility-Fixed Effects Dependent Variable LICNUR NURSTAFF Explanatory Variables Coefficient Standard Errora Coefficient Standard Errora CON_MORT -0.01321 0.01605 -0.01135 0.03594 RATE 0.00210 0.00025*** 0.00291 0.00065*** RATEMISSING 0.24970 0.04667*** 0.40824 0.13043*** CONTROL2 0.00375 0.01064 0.02791 0.02593 CONTROL3 0.00047 0.02033 0.00779 0.05970 TOTBEDS -0.00119 0.00027*** -0.00290 0.00074*** EMPTYELDERLY 0.01107 0.00105*** 0.02226 0.00240*** MULTI -0.00538 0.00640 -0.03971 0.01600** ADLINDEX 0.01204 0.00181*** 0.03988 0.00446*** HHI 0.09287 0.02183*** 0.17693 0.05187*** INCOME -3.09e-06 1.06e-06*** -5.83e-06 2.71e-06** METHOD0 -0.06496 0.01991*** -0.21151 0.04662*** METHOD1 -0.00370 0.00910 -0.05162 0.02313** METHOD2 -0.01073 0.02333 -0.10881 0.08425 METHOD4 -0.01380 0.01280 -0.04150 0.02983 CASEMIX -0.03627 0.00954*** -0.10217 0.01942*** CONSTANT 0.81163 0.05067*** 2.84249 0.13258*** N 75,904 75,904 R2 0.0157b 0.0121b Prob > F 0.0000 0.0000 aThe standard errors represent the Huber-White robust standard errors corrected for intrahome correlation using the cluster option in Stat a 9 S/E. Each regression contains 8089 clusters. *p < .10 **p < .05 ***p < .01 bThis value is the R2 within. Both models include year fixed effects.

PAGE 184

172Table A.36 Regression Results for Access Model Most Restrictive Sample Dependent Variable PCTMCAID State-Fixed Effects Facility-Fixed Effects Explanatory Variables Coefficient Standard Errora Coefficient Standard Errora CON_MORT 2.10167 0.71449*** 3.68244 0.56805*** RATE 0.01873 0.00995* 0.01122 0.00795 RATEMISSING 3.35836 1.95300* 3.64824 1.51695** CONTROL2 -8.50022 0.48757*** -1.11712 0.40088*** CONTROL3 4.35162 0.77855*** 1.91969 1.08148* TOTBEDS 0.02720 0.00312*** -0.02642 0.00707*** EMPTYELDERLY 0.09504 0.04670** 0.00670 0.02408 MULTI -1.76661 0.33569*** -0.59560 0.21049*** ADLINDEX -0.74316 0.12425*** 0.16310 0.05107*** HHI 3.26867 0.65618*** -0.41052 0.49224 INCOME -0.00041 0.00003*** 0.00008 0.00004** METHOD0 -1.01304 0.77661 -0.75264 0.61478 METHOD1 0.18235 0.38161 -0.07545 0.29549 METHOD2 -1.95926 1.21465 0.31312 1.12446 METHOD4 1.11780 0.53190** 1.45567 0.41672*** CASEMIX -0.30968 0.28800 -0.22946 0.23233 CONSTANT 88.31466 2.67226*** 66.41340 1.63747*** N 75,904 75,904 R2 0.1823 0.0099b Prob > F 0.0000 0.0000 aThe standard errors represent the Huber-White robust standard errors corrected for intra-home correlation using the cluster optio n in Stata 9 S/E. Each regression contains 8089 clusters. *p < .10 **p < .05 ***p < .01 bThis value is the R2 within. The state fixed effects model include s state and year fixed effects. The facility fixed effects model includes year fixed effects.

PAGE 185

173Table A.37 Regression Results fo r Heavy-Care Access Model Most Restrictive Sample Dependent Variable ADLINDEX State-Fixed Effects Facility-Fixed Effects Explanatory Variables Coefficient Standard Errora Coefficient Standard Errora CON_MORT -0.26803 0.06390*** -0.27698 0.05058*** RATE 0.00284 0.00093*** 0.00298 0.00085*** RATEMISSING 0.27901 0.16209* 0.38902 0.15658** CONTROL2 -0.03202 0.03121 0.05377 0.03643 CONTROL3 0.37891 0.07373*** 0.19462 0.10332* TOTBEDS 0.00194 0.00024*** -0.00095 0.00060 EMPTYELDERLY -0.02158 0.00347*** 0.00127 0.00254 MULTI 0.05436 0.02401** -0.03056 0.02029 HHI -0.13302 0.04773*** 0.02480 0.05278 INCOME -7.02e-07 2.12e-06 2.26e-06 3.13e-06 METHOD0 0.06635 0.07163 0.04578 0.06057 METHOD1 0.01923 0.03098 -0.03994 0.02674 METHOD2 -0.07406 0.10565 -0.05278 0.09838 METHOD4 -0.21592 0.04542*** -0.18824 0.03970*** CASEMIX 0.22287 0.02457*** 0.22514 0.02285*** CONSTANT 10.06110 0.18597*** 9.45304 0.14257*** N 75,904 75,904 R2 0.2087 0.0606b Prob > F 0.0000 0.0000 aThe standard errors represent the Huber-White robust standard errors corrected for intra-home correlation using the cluster optio n in Stata 9 S/E. Each regression contains 8089 clusters. *p < .10 **p < .05 ***p < .01 bThis value is the R2 within. The state fixed effects model include s state and year fixed effects. The facility fixed effects model includes year fixed effects.

PAGE 186

174Table A.38 Regression Results for Process-Based Quality Models Least Restrictive Sample State-Fixed Effects Dependent Variable PROPCATHETER PROPPARENTERAL PROPMOBLREST Explanatory Variables Coefficient Standard Errora Coefficient Standard Errora Coefficient Standard Errora CON_MORT -0.01149 0.00163*** -0.00202 0.00123* -0.04263 0.00386*** RATE 0.00009 0.00004** -0.00018 0.00004*** -0.00072 0.00011*** RATEMISSING 0.01318 0.00883 -0.01973 0.00714*** -0.07095 0.03727* CONTROL2 -0.12296 0.00106*** -0.01023 0.00126*** 0.00616 0.00206*** CONTROL3 -0.01268 0.00183*** -0.00522 0.00276* 0.01455 0.00449*** TOTBEDS 0.00007 8.83e06*** 0.00008 0.00001*** 0.00002 0.00002 EMPTYELDERLY 0.00002 0.00004 -0.00015 0.00004*** 0.00007 0.00007 MULTI 0.00398 0.00082*** 0.00441 0.00079*** -0.01153 0.00155*** ADLINDEX 0.01075 0.00040*** 0.01366 0.00095*** 0.02130 0.00070*** HHI -0.00310 0.00187* -0.02137 0.00185*** 0.00882 0.00384** INCOME -3.42e-07 9.73e-08*** 7.09e-07 1.15e-07*** -9.00e-07 1.85e-07*** METHOD0 -0.00649 0.00247*** -0.00307 0.00162* -0.01969 0.00633*** METHOD1 -0.00324 0.00113*** -0.00087 0.00077 -0.00174 0.00263 METHOD2 -0.00804 0.00182*** 0.00477 0.00142*** 0.02468 0.00414*** METHOD4 -0.00942 0.00232*** 0.00142 0.00172 0.01808 0.00668*** CASEMIX 0.00156 0.00090* -0.00180 0.00076** -0.00319 0.00238 CONSTANT -0.00917 0.00566 -0.08346 0.01001*** 0.12087 0.01181*** N 74,667 74,667 74,667 R2 0.1447 0.2648 0.2271 Prob > F 0.0000 0.0000 0.0000 aThe standard errors represent the Huber-White robust standard errors corrected for intra-home correlation using the cluster optio n in Stata 9 S/E. Each regression contains 7821 clusters. *p < .10 **p < .05 ***p < .01 All models include state and year fixed effects.

PAGE 187

175Table A.39 Regression Results for Process-Based Quality Models Least Restrictive Sample Facility-Fixed Effects Dependent Variable PROPCATHETER PROPPARENTERAL PROPMOBLREST Explanatory Variables Coefficient Standard Errora Coefficient Standard Errora Coefficient Standard Errora CON_MORT -0.01161 0.00166*** -0.00116 0.00114 -0.03956 0.00388*** RATE 0.00006 0.00004 -0.00015 0.00004*** -0.00071 0.00012*** RATEMISSING 0.011187 0.00789 -0.02018 0.00670*** -0.09721 0.03043*** CONTROL2 -0.00286 0. 00166* 0.00028 0.00127 0.01808 0.00437*** CONTROL3 -0.00165 0.00371 -0.00131 0.00190 0.02390 0.00929*** TOTBEDS 0.00003 0.00002* -0.00003 0.00002** 0.00008 0.00005 EMPTYELDERLY 0.00011 0.00004*** 0.00009 0.00003*** 0.00033 0.00009*** MULTI -0.00169 0.00083** 0.00026 0.00064 -0.00396 0.00213* ADLINDEX 0.00686 0.00023*** 0.00623 0.00021*** 0.01436 0.00064*** HHI 0.00568 0.00239** 0.00483 0.00184*** 0.03207 0.00607*** INC03 1.94e-07 2.13e-07 -5.28e-07 2.12e-07*** -2.72e-06 7.34e-07*** METHOD0 -0.00793 0.00253*** -0.00256 0.00150* -0.01614 0.00639** METHOD1 -0.00421 0.00108*** -0.00046 0.00071 0.00125 0.00262 METHOD2 -0.01100 0.00176*** 0.00477 0.00133*** 0.02808 0.00416*** METHOD4 -0.00942 0.002270*** 0.00194 0.00164 0.02057 0.00656*** CASEMIX 0.00152 0.00084* -0.00062 0.00062 -0.00106 0.00241 CONSTANT 0.00987 0.00680 0.00698 0.00633 0.16933 0.02151*** N 74,667 74,667 74,667 R2 0.0598b 0.0686b 0.1282b Prob > F 0.0000 0.0000 0.0000 aThe standard errors represent the Huber-White robust standard errors corrected for intra-home correlation using the cluster optio n in Stata 9 S/E. Each regression contains 7821 clusters. *p < .10 **p < .05 ***p < .01 bThis value is the R2 within. All models have year fixed effects.

PAGE 188

176Table A.40 Regression Results for Outcome and Composite-Based Quality Models Least Restrictive Sample State-Fixed Effects Dependent Variable Outcome-Based PROPPRESSORE Composite-Based DEFS1TOT Explanatory Variables Coefficient Standard Errora Coefficient Standard Errora CON_MORT 0.00030 0.00138 0.39374 0.18673** RATE -0.00023 0.00004*** 0.04098 0.00579*** RATEMISSING -0.03164 0.00743*** 5.45617 1.18333*** CONTROL2 -0.00742 0.00066*** -1.37505 0.09018*** CONTROL3 -0.00579 0.00133*** -1.50386 0.15082*** TOTBEDS 0.00005 5.36e-06*** 0.01253 0.00078*** EMPTYELDERLY -0.00008 0.00002*** -0.00303 0.00299 MULTI 0.00511 0.00051*** 0.19029 0.07275*** ADLINDEX 0.00716 0.00023*** -0.00824 0.02499 HHI -0.00652 0.00122*** -0.98078 0.17867*** INCOME 4.48e-07 6.65e-08*** -8.47e-06 9.21e-06 METHOD0 -0.00328 0.00212 0.66507 0.27790** METHOD1 -0.00777 0.00103* 2.36500 0.16135*** METHOD2 -0.00412 0.00148*** 2.64283 0.21232*** METHOD4 -0.00157 0.00203 0.78846 0.37235** CASEMIX -0.00136 0.00079* 0.40608 0.11215*** CONSTANT 0.00410 0.00398 4.49264 0.55528*** N 74,667 74,647 R2 0.1123 0.1322 Prob > F 0.0000 0.0000 aThe standard errors represent the Huber-White robust standard errors corrected for intra-home correlation using the cluster optio n in Stata 9 S/E. Each regression contains 7821 clusters. *p < .10 **p < .05 ***p < .01 Both models include state and year fixed effects.

PAGE 189

177Table A.41 Regression Results for Outcome and Composite-Based Quality Models Least Restrictive Sample Facility-Fixed Effects Dependent Variable Outcome-Based PROPPRESSORE Composite-Based DEFS1TOT Explanatory Variables Coefficient Standard Errora Coefficient Standard Errora CON_MORT 0.00053 0.00131 0.57946 0.17650*** RATE -0.00023 0.00004*** 0.04103 0.00579*** RATEMISSING -0.02894 0.00697*** 5.27053 1.24988*** CONTROL2 0.00010 0.00141 0.00142 0.21261 CONTROL3 0.00179 0.00262 -1.2186 0.50713** TOTBEDS 0.00004 0.00002** 0.01941 0.00241*** EMPTYELDERLY 0.00004 0.00003 -0.00192 0.00383 MULTI 0.00133 0.00075* -0.07906 0.11193 ADLINDEX 0.00495 0.00021*** 0.12874 0.02773*** HHI 0.00096 0.00239 0.78495 0.31125** INCOME 1.07e-07 1.69e-07 0.00011 0.00003*** METHOD0 -0.00376 0.00214* 0.73696 0.27617*** METHOD1 -0.00168 0.00103 2.42352 0.15862*** METHOD2 -0.00463 0.00149*** 2.71901 0.20974*** METHOD4 -0.00205 0.00206 0.62906 0.37838* CASEMIX -0.00071 0.00080 0.44168 0.11078*** CONSTANT 0.02692 0.00576*** -0.97425 0.95677 N 74,667 74,647 R2 0.0167b 0.0594b Prob > F 0.0000 0.0000 aThe standard errors represent the Huber-White robust standard errors corrected for intrahome correlation using the cluster option in Stat a 9 S/E. Each regression contains 7821 clusters. *p < .10 **p < .05 ***p < .01 bThis value is the R2 within. Both models include year fixed effects.

PAGE 190

178Table A.42 Regression Results for Structure-Based Quality Models Least Restrictive Sample State-Fixed Effects Dependent Variable RNS LPNS AIDES Explanatory Variables Coefficient Standard Errora Coefficient Standard Errora Coefficient Standard Errora CON_MORT -0.01613 0.00622*** 0.03371 0.01030*** -0.10149 0.02246*** RATE -0.00002 0.00023 0.00011 0.00033 -0.00463 0.00084*** RATEMISSING 0.09933 0.04898** -0.04867 0.06762 -0.49284 0.14188*** CONTROL2 0.05067 0.00701*** 0.03390 0.00944*** 0.25115 0.02382*** CONTROL3 0.03758 0.01212*** 0.01614 0.01531 0.21321 0.03490*** TOTBEDS -0.00021 0.00006*** 0.00003 0.00007 -0.00004 0.00017 EMPTYELDERLY 0.00069 0.00013*** 0.00186 0.00022*** 0.00239 0.00049*** MULTI 0.01052 0.00436** 0.03598 0.00625*** -0.00926 0.01580 ADLINDEX 0.02704 0.00255*** 0.03211 0.00319*** 0.12633 0.00734*** HHI 0.00292 0.00982 -0.10605 0.01289*** -0.14898 0.03551*** INCOME 9.92e-06 7.79e-07*** 2.54e06 1.04e-06** 0.00001 2.75e-06*** METHOD0 0.03198 0.01521** -0.00364 0.01650 -0.09587 0.05340* METHOD1 0.03232 0.00512*** -0.00100 0.00701 0.01901 0.01928 METHOD2 0.03896 0.00655*** 0.04920 0.01350*** 0.01122 0.03252 METHOD4 0.01640 0.01490 -0.01700 0.02268 -0.02571 0.05974 CASEMIX 0.00279 0.00430 -0.00201 0.00571 -0.06514 0.01613*** CONSTANT -0.39987 0. 03457*** 0.27847 0.04265*** 0.78037 0.10672*** N 74,667 74,667 74,667 R2 0.2152 0.1239 0.0868 Prob > F 0.0000 0.0000 0.0000 aThe standard errors represent the Huber-White robust standard errors corrected for intra-home correlation using the cluster optio n in Stata 9 S/E. Each regression contains 7821 clusters. *p < .10 **p < .05 ***p < .01 All models include state and year fixed effects.

PAGE 191

179Table A.43 Regression Results for Structure-Based Quality Models Least Restrictive Sample Facility-Fixed Effects Dependent Variable RNS LPNS AIDES Explanatory Variables Coefficient Standard Errora Coefficient Standard Errora Coefficient Standard Errora CON_MORT -0.00163 0.00567 0.04317 0.00826*** -0.07180 0.02016*** RATE 0.00260 0.00020 0.00011 0.00028 -0.00526 0.00078*** RATEMISSING 0.10920 0.04122*** -0.03028 0.04021 -0.55575 0.12313*** CONTROL2 -0.00175 0.00636 -0.00092 0.00981 -0.02562 0.02829 CONTROL3 0.01877 0.01057* -0.000264 0.02461 0.15239 0.06428** TOTBEDS -0.00043 0.00011*** -0.00046 0.00016*** -0.00082 0.00041* EMPTYELDERLY 0.00126 0.00014*** 0.00286 0.00024*** 0.00461 0.00060*** MULTI 0.00205 0.00378 0.00406 0.00540 0.00122 0.01529 ADLINDEX 0.00793 0.00135*** 0.01519 0.00201*** 0.04990 0.00523*** HHI 0.05089 0.00900*** 0.08529 0.01550*** 0.13731 0.03593*** INCOME -1.56e-06 8.62e-07* -4.71e06 1.62e-06*** -0.00001 3.68e-06*** METHOD0 0.02305 0.01378* -0.00318 0.01558 -0.08624 0.04947* METHOD1 0.02040 0.00415*** -0.00030 0.00600 0.00722 0.01605 METHOD2 0.03040 0.00584*** 0.01798 0.01053* -0.05596 0.02578** METHOD4 0.02335 0.01105** -0.00775 0.01928 0.00952 0.04192 CASEMIX 0.00836 0.00376** 0.00501 0.00513 -0.03480 0.01470** CONSTANT 0.23228 0.03220*** 0.52687 0.05031*** 2.3422 0.12998*** N 74,667 74,667 74,667 R2 0.0147b 0.0262b 0.0142b Prob > F 0.0000 0.0000 0.0000 aThe standard errors represent the Huber-White robust standard errors corrected for intra-home correlation using the cluster optio n in Stata 9 S/E. Each regression contains 7821 clusters. *p < .10 **p < .05 ***p < .01 bThis value is the R2 within. All models include year fixed effects.

PAGE 192

180Table A.44 Regression Results for Structure-Based Quality Models Least Restrictive Sample State-Fixed Effects Dependent Variable LICNUR NURSTAFF Explanatory Variables Coefficient Standard Errora Coefficient Standard Errora CON_MORT 0.01996 0.01120* -0.06266 0.02578** RATE 0.00022 0.00034 -0.00372 0.00089*** RATEMISSING 0.04489 0.08459 -0.39278 0.15122*** CONTROL2 0.09157 0.00980*** 0.35219 0.02631*** CONTROL3 0.08222 0.01979*** 0.34392 0.04659*** TOTBEDS -0.00024 0.00008*** -0.00043 0.00019** EMPTYELDERLY 0.00234 0.00022*** 0.00450 0.00055*** MULTI 0.01314 0.00617** -0.05823 0.01663*** ADLINDEX 0.04333 0.00405*** 0.14083 0.00905*** HHI -0.07646 0.01296*** -0.18579 0.03471*** INCOME 9.13e-06 1.04e-06*** 0.00002 2.67e-06*** METHOD0 0.01423 0.02306 -0.09974 0.06211 METHOD1 0.01211 0.00715* 0.00638 0.01983 METHOD2 0.05533 0.01207*** 0.02602 0.03324 METHOD4 -0.00092 0.02199 -0.02582 0.05568 CASEMIX 0.00155 0.00640 -0.05651 0.01754*** CONSTANT 0.11218 0.04982*** 1.25096 0.11896*** N 74,667 74,667 R2 0.1359 0.1180 Prob > F 0.0000 0.0000 aThe standard errors represent the Huber-White robust standard errors corrected for intrahome correlation using the cluster option in Stat a 9 S/E. Each regression contains 7821 clusters. *p < .10 **p < .05 ***p < .01 Both models have state and year fixed effects.

PAGE 193

181Table A.45 Regression Results for Structure-Based Quality Models Least Restrictive Sample Facility-Fixed Effects Dependent Variable LICNUR NURSTAFF Explanatory Variables Coefficient Standard Errora Coefficient Standard Errora CON_MORT 0.03729 0.00927*** -0.02571 0.02242 RATE 0.00029 0.00031 -0.00447 0.00084*** RATEMISSING 0.03999 0.04586 -0.49537 0.14552*** CONTROL2 0.00254 0.01049 -0.01420 0.03066 CONTROL3 0.02189 0.02452 0.18073 0.07651** TOTBEDS -0.00096 0.00018*** -0.00196 0.00047*** EMPTYELDERLY 0.00383 0.00027*** 0.00806 0.00070*** MULTI -0.00020 0.00570 -0.01345 0.01608 ADLINDEX 0.01890 0.00236*** 0.06067 0.00612*** HHI 0.12451 0.01697*** 0.24296 0.04002*** INCOME -5.11e-06 1.69e-06*** -0.00002 4.26e-06*** METHOD0 0.01518 0.02154 -0.07700 0.05775 METHOD1 0.00965 0.00620 0.00536 0.01675 METHOD2 0.03673 0.01044*** -0.02269 0.02817 METHOD4 0.01071 0.02070 0.01221 0.04577 CASEMIX 0.01012 0.00581* -0.02672 0.01621* constant 0.75621 0.05415*** 3.01955 0.14603*** N 74,667 74,667 R2 0.0234b 0.0150b Prob > F 0.0000 0.0000 aThe standard errors represent the Huber-White robust standard errors corrected for intrahome correlation using the cluster option in Stat a 9 S/E. Each regression contains 7821 clusters. *p < .10 **p < .05 ***p < .01 bThis value is the R2 within. Both models include year fixed effects.

PAGE 194

182Table A.46 Regression Results for Access Model Least Restrictive Markets Dependent Variable PCTMCAID State-Fixed Effects Facility-Fixed Effects Explanatory Variables Coefficient Standard Errora Coefficient Standard Errora CON_MORT -0.62873 0.41804 0.66737 0.36207* RATE -0.03963 0.01215***-0.01374 0.00971 RATEMISSING -6.97784 2.6512*** -3.31099 2.36212 CONTROL2 -10.13519 0.47517***-0.73226 0.37400** CONTROL3 -2.57698 0.86979***0.12175 0.77340 TOTBEDS 0.00728 0.00346** -0.00706 -0.00706 EMPTYELDERLY -0.03419 0.01192***0.02668 0.00798*** MULTI -0.20703 0.30485 0.20340 0.19248 ADLINDEX -1.50868 0.13121***0.14916 0.05219*** HHI -1.75475 0.64923***1.03556 0.50782** INCOME -0.00028 0.00004***0.00019 0.00005*** METHOD0 -1.74078 0.71196** -0.81800 0.59639 METHOD1 -0.36103 0.29609 -0.27111 0.24335 METHOD2 -1.35941 0.43857***-0.76737 0.36560** METHOD4 -0.34194 0.71408 0.17570 0.58264 CASEMIX 0.20832 0.23109 0.01360 0.19147 CONSTANT 101.29370 1.89618***62.32554 1.41084*** N 74,667 74,667 R2 0.2364 0.0047b Prob > F 0.0000 0.0000 aThe standard errors represent the Huber-White robust standard errors corrected for intrahome correlation using the cluster option in Stat a 9 S/E. Each regression contains 7821 clusters. *p < .10 **p < .05 ***p < .01 bThis value is the R2 within. The state fixed effects model include s state and year fixed effects. The facility fixed effects model includes year fixed effects.

PAGE 195

183Table A.47 Regression Results fo r Heavy-Care Access Model Least Restrictive Market Dependent Variable ADLINDEX State-Fixed Effects Facility-Fixed Effects Explanatory Variables Coefficient Standard Errora Coefficient Standard Errora CON_MORT -0.04861 0.03376 -0.00141 0.03263 RATE -0.00005 0.00104 -0.00061 0.00098 RATEMISSING -0.56527 0.23961** -0.69092 0.22297*** CONTROL2 0.16315 0.02835*** -0.00477 0.03503 CONTROL3 0.26759 0.06476*** -0.00948 0.07494 TOTBEDS 0.00223 0.00025*** -0.00057 0.00045 EMPTYELDERLY -0.00539 0.00090*** 0.00160 0.00076** MULTI 0.08487 0.02107*** -0.03106 0.01809* HHI -0.11583 0.04907** 0.10844 0.05091** INCOME 1.36e-06 2.96e-06 -5.07e-06 4.32e-06 METHOD0 -0.16837 0.06291*** -0.16757 0.06172*** METHOD1 0.07910 0.02506*** 0.07337 0.02270*** METHOD2 -0.03807 0.03722 -0.02008 0.03482 METHOD4 -0.03295 0.05968 -0.04687 0.05667 CASEMIX 0.21721 0.02042*** 0.20908 0.01879*** CONSTANT 9.17985 0.11400*** 9.15576 0.13787*** N 74,667 74,667 R2 0.1949 0.0537b Prob > F 0.0000 0.0000 aThe standard errors represent the Huber-White robust standard errors corrected for intra-home correlation using the cluster optio n in Stata 9 S/E. Each regression contains 7821 clusters. *p < .10 **p < .05 ***p < .01 bThis value is the R2 within. The state fixed effects model include s state and year fixed effects. The facility fixed effects model includes year fixed effects.

PAGE 196

About the Author Barbara Caldwell obtained her undergraduate degree in Industrial Engineering from Georgia Institute of Technology in 1981. After holding several positions in the manufacturing industry she earne d her MBA in 1999 at the Univer sity of South Florida. While studying at USF she was inducted into the Phi Kappa Phi and Beta Gamma Sigma Honor Societies. Upon completion of her MBA, Ms. Cald well entered the PhD program in the Department of Economics with her main area of research in Health Economics. She received the Howard S. Dye Memorial Scho larship in Economics in the Fall of 2004 and was a 2005 recipient of the National Science Foundation Doctoral Di ssertation Research Grant.