Citation
A Cost-effectiveness analysis of home health care : implications for public policy and future research

Material Information

Title:
A Cost-effectiveness analysis of home health care : implications for public policy and future research
Creator:
Kurowski, Bettina Tabak
Place of Publication:
Denver, CO
Publisher:
University of Colorado Denver
Publication Date:
Language:
English

Thesis/Dissertation Information

Degree:
Doctorate ( Doctor of public administration)
Degree Grantor:
University of Colorado Denver
Degree Divisions:
School of Public Affairs, CU Denver
Degree Disciplines:
Public administration
Committee Members:
March, Michael S.
Burgess, Philip M.
Shaughnessy, Peter W.

Notes

Abstract:
The subject of this study is the cost-effectiveness of one long-term care policy option: home health care. Funded by the Health Services Administration (DHEW) in 1977, the study highlights long-term care policy issues that need to be addressed through eval uation research. In so doing, it examined the cost per episode of illness of certain home health care services, using 1976 data from a sample of four hospital-based and four free-standing home health care providers in the eastern U.S. The empirical portion of the study determined the effect of patient, provider/community, and out come characteristics on the cost per episode of care. The findings indicated that most of the variation in cost was not explained by factors included in the study. Of the cost variation that was explained, patient characteristics accounted for the greatest portion, followed by outcome and provider characteristics. Important policy implications of the findings are presented; the major research implication of the study is that a comprehensive national research program that is intended to address long-term care issues is essential. Key attributes of the suggested program include comprehensive measurement of costs, effects, patient, and provider character~stics; adequate project duration, and appropriate dissemination of findings.

Record Information

Source Institution:
University of Colorado Denver
Holding Location:
Auraria Library
Rights Management:
Copyright [name of copyright holder or Creator or Publisher as appropriate]. Permission granted to University of Colorado Denver to digitize and display this item for non-profit research and educational purposes. Any reuse of this item in excess of fair use or other copyright exemptions requires permission of the copyright holder.

Downloads

This item has the following downloads:


Full Text
A COST-EFFECTIVENESS ANALYSIS OF HOME HEALTH CARE: IMPLICATIONS FOR PUBLIC POLICY AND FUTURE RESEARCH
by
Bettina Tabak Kurowski
3.S., University of Southern California, 1967 M.P.A., University of Colorado, 1975
A thesis submitted to the Faculty of the Graduate School of Public Affairs of the University of Colorado in partial fulfillment of the requirements for the degree of Doctor of Public Administration
1980


This Thesis for the Doctor
of Public Administration Degree by Bettina Tabak Kurowski has been approved for the Graduate School of Public Affairs by
Michael S. March
Date IlUj 7r / $D


i i i
Kurowski, Bettina Tabak (D.P.A., Public Administration)
A Cost-Effectiveness Analysis of Home Health Care: Implications for
Public Policy and Future Research
Thesis directed by Professor Michael S. March
The subject of this study is the cost-effectiveness of one long-term care policy option: home health care. Funded by the Health Services Administration (DHEW) in 1977, the study highlights long-term care policy issues that need to be addressed through evaluation research. In so doing, it examined the cost per episode of illness of certain home health care services, using 1976 data from a sample of four hospital-based and four free-standing home health care providers in the eastern U.S. The empirical portion of the study determined the effect of patient, provider/community, and outcome characteristies on the cost per episode of care.
The findings indicated that most of the variation in cost was not explained by factors included in the study. Of the cost variation that was explained, patient characteristics accounted for the greatest portion, followed by outcome and provider characteristics.
Important policy implications of the findings are presented; the major research implication of the study is that a comprehensive national research program that is intended to address long-term care issues is essential. Key attributes of the suggested program include comprehensive measurement of costs, effects, patient, and provider characteristics; adequate project duration, and appropriate dissemination of findings.


TV
This abstract is approved as to form and content. I recommend its pub!icat ion.
Signed ____________________________________
Faculty member in charge of thesis


PREFACE
The subject of this study is long-term care policy, specifically in the area of home health care. The vehicle chosen to explore this issue is a study of the cost per episode of illness. In October 1977, the Office of Planning, Evaluation, and Legislation of the Health Services Administration in the Department of Health, Education, and Welfare awarded a contract to the Center for Health Services Research of the University of Colorado Health Sciences Center to conduct research into aspects to home health care relevant to the agency's responsibility for promoting the development and expansion of home health service capacity. One part of that contract was the study described herein.
From the date of the initial award through the end of the eighteen month contract, I served as Project Director for the cost per episode portion of the study. Thus, I was responsible for the development of its design and execution of its analytic methodology, in addition to the preparation of the final report. It is as a result of my role in this contract that I authored this study.
I have been fortunate to have had assistance in the development and improvement of this manuscript. Those who have read and commented on one or another version of this study have contributed greatly to its refinement. I am especially grateful to my colleagues at the Center for Health Services Research, Eileen Tynan, Barbara Harley, Nancy Shanks, and David Landes, for their suggestions and criticisms; Peter Shaughnessy, the Director of the


Center, for concurrently playing the roles of teacher, employer, and friend; Robert Schlenker, the Principal Investigator of the home health evaluation contract, for his guidance and support. Gerri Tricarico and Linda Breed provided invaluable assistance in the data analysis phase of the study. Susan Elgin, Charlotte Gordy, and Phyllis Hunter typed countless drafts of this manuscript and deserve much appreciation for their efforts.
Michael March, my thesis advisor, contributed significantly to this study and to my continuing intellectual growth. His commitment to improving the public policy process through inquiry, analysis, and action should serve as an inspiration to those who wish to improve the quality of government. Philip Burgess, Director of my doctoral program, enlarged my view of the world and helped shape my professional development.
A special thank you goes to my colleague, Ann Jones, who provided needed encouragement and thoughtful suggestions at crucial times throughout the development of tnis manuscript; and to Carol Betson and Laura Dodson, who each helped in their own way. Immense gratitude, which is difficult to put into words, goes to my husband, Jim, and children, Lisa and Thad, who endured my long hours, supported me in my frustration, and continually offered me encouragement. Finally, my mother has been a source of energy and, in a special way, made a valuable contribution to this effort.
Bettina Tabak Kurowski
Denver, Colorado May 1980


TABLE OF CONTENTS
CHAPTER PAGE
I. INTRODUCTION
Overview .................................................. I
The long-term care system ................................. 3
The long-term care population ........................... 3
Rising expenditures for long-term care .................. 7
Institutional care .................................... 9
Home health care...................................... 10
Total costs........................................... 11
Quality and appropriateness of care..................... 12
Home health care........................................ 17
Background............................................ 17
Comparative costs of home health care................. 21
Major barriers to home care........................... 24
Summary of current policy issues ....................... 27
Evaluation research to reduce uncertainty ................ 28
A specific research area ................................. 31
II. STATE OF THE ART
Introduction ............................................. 34
Case mix.................................................. 34
Age..................................................... 35
Living arrangement ..................................... 36
Diagnosis............................................... 37
Functional status ...................................... 38


Vlll
CHAPTER PAGE
Mental health status .................................. 40
Multidimensional health status ........................ 41
Problems............................................... 42
Levels of care......................................... 43
Quality................................................... 46
Structural measures of quality ........................ 46
Process measures of quality ........................... 48
Outcome measures of quality ........................... 50
Specifying appropriate outcomes in general .... 52
Specifying appropriate outcomes for home
health care.......................................... 53
Tenuous relationship between process and outcome . 57
Cost of Care.............................................. 60
Nursing home cost studies.............................. 61
Facility level cost studies ......................... 61
Patient level cost studies .......................... 68
Home health cost studies............................... 69
Cost-effectiveness of long-term care ..................... 71
Cost-effectiveness of home health care versus
nursing home care...................................... 72
Effectiveness of home health care in reducing institutionalization ................................ 74
Cost of home health care versus nursing home care 77
Cost effectiveness of coordinated and expanded
home health care....................................... 81
Effectiveness findings .............................. 82
Cost findings........................................ 84
Cost-effectiveness findings ......................... 84


ix
CHAPTER PAGE
Methodological weaknesses ........................... 85
Conclusion............................................... 91
III. THE STUDY APPROACH AND METHODOLOGY
Introduction ............................................ 92
Study purpose............................................ 92
Conceptual framework .................................... 93
Independent variables ................................. 94
Dependent variable .................................... 96
Overall relationships ................................. 96
Sample selection and data availability .................. 98
Data system selection ................................. 98
Provider selection ................................... 100
Episode selection ..................................... 102
Study variables and data sources.......................106
Reliability of data.....................................106
Specification of the model...............................107
The primary dependent variaole: cost per
episoae of home health care.............................107
Definition of episode ............................... 109
Unit of service.......................................110
Cost per unit of service..............................Ill
Cost per episode......................................113
The independent variables: patient-specific characteristics ....................................... 114
Age...................................................114
Living arrangement................................... 115
Diagnosis and functional status ..................... 116


X
CHAPTER PAGE
Surgical procedure ................................. 118
Goal at admission....................................118
The independent variables: outcome-related characteristics ...................................... 119
Change in functional status ........................ 119
Health status at discharge ......................... 119
Home health care length of use.......................120
Intensity............................................120
The independent variables: provider/health system characteristics ...................................... 121
Primary source of payment............................121
Provider size........................................123
Population density ................................. 123
Acute care admissions................................124
Miscellaneous factors ................................ 125
Income...............................................125
Race.................................................125
Other health care providers..........................125
Regulatory environment ............................. 126
Analytic techniques .................................... 133
Limitations of the study.................................135
Data Limitations.......................................135
Cost Data............................................135
Case Mix Data........................................137
Quality and Effectiveness Data.......................138
Design Issues ........................................ 139
Conclusion...............................................140
I


xi
CHAPTER PAGE
IV. A COMPARISON OF CASE CHARACTERISTICS
Introduction ............................................ 141
Patient-specific characteristics ..................... 141
Age.....................................................142
Living arrangement .................................... 144
Primary diagnosis ..................................... 144
Functional status ..................................... 144
Functional status index ............................... 148
Surgical procedure .................................... 150
Patient's location prior to admission ................. 150
Goal at admission.......................................150
Outcome-related characteristics ..................... 152
Change in functional status index ..................... 152
Health status at discharge ............................ 154
Intensity of care and length of use....................154
Provider/health system characteristics .................. 157
Primary source of paynent ............................. 157
Conclusion................................................159
V. UTILIZATION AND COST FINDINGS
Introduction ............................................ 160
Overall utilization and cost per episode findings . . 161
Utilization findings .................................. 161
Cost findings...........................................163
Cost findings discussion .............................. 165
Charges passed through as costs ................. . 165
Unaudited cost reports .............................. 166


Xll
CHAPTER PAGE
Administrative surcharges ........................... 166
Case mix differences..................................168
Service mix differences ............................. 169
Variations in outcome ............................... 171
Analyses to explain overall variation within
each sample...............................................171
Approach................................................171
Overall findings ...................................... 174
Rehabilitation patients ............................ 175
Terminally ill patients ............................. 175
Analyses to test the relative importance of each independent variable category ........................... 176
Analyses to test the patterns within each
independent variable category ........................... 177
Patient-specific characteristics ...................... 178
Overview of multivariate findings ................... 179
Age...................................................181
Living arrangement .................................. 183
Primary diagnosis .................................. 183
Functional status .................................. 186
Functional status index ............................ 188
Goal at admission....................................190
Outcome-related characteristics ....................... 192
Overview of multivariate findings ................... 192
Change in functional status index ................... 193
Health status at discharge .......................... 195
Utilization-related outcome ......................... 197


xni
CHAPTER PAGE
Provider/health system characteristics ............... 197
Overview of multivariate findings .................. 198
Primary source of payment .......................... 199
Provider size........................................201
Acute care admissions................................203
Comparison of Massachusetts and Philadelphia outcomes 203
Overview...............................................205
Controlling for age..................................205
Controlling for living arrangement ................. 207
Controlling for primary diagnosis .................. 210
Controlling for functional status index ............ 210
Summary of major findings .............................. 212
Overall utilization and cost per episode findings . 214
Overall ability to explain observed variation
within each sample.....................................214
Relative importance of each independent
variable category .................................... 215
Patient-specific characteristics ................... 216
Outcome-related characteristics .................... 217
Provider/health system characteristics ............. 218
Conclusion...............................................219
VI. POLICY IMPLICATIONS
Cost implications........................................220
Inability to explain cost variations ................. 220
Utilization review ................................... 221
Allocation procedures ................................ 221
Cost-effectiveness ................................... 222


xiv
CHAPTER PAGE
Case mix adjustment.................................222
Progran implications..................................223
Continuum of care...................................223
Hospice care........................................223
Home health data....................................224
Housing.............................................225
Tax incentives......................................226
Information and referral .............................. 226
VII. AGENDA FOR FUTURE RESEARCH
Introduction ............................................ 228
Overview of long-term care research problems ............ 228
Lack of comprehensive and systematic approach . . . 229 Inadequate and incomplete goals and objectives . . . 230
Underdeveloped criteria ............................... 231
Inadequate design and data collection ................. 232
Lack of comprehensive cost measurement..............232
Neglect of organizational and management variables . 233
Inadequate patient descriptors ........................ 234
Lack of attention to the longitudinal dimension . . 235
Non-comparable designs ................................. 235
Inadequate utilization and dissemination of findings 236
Systematic analysis for long-term care policy issues . 233
Definition of systems analysis ........................ 240
Overview of the basic elements ........................ 240
Defining the problem..............................241
Identification of goals, objectives and criteria . 242


XV
CHAPTER PAGE
Designing the model .................................243
Development of alternatives ........................ 244
Evaluating alternatives ............................ 244
Predicting consequences ............................ 245
Interpretation ..................................... 245
Communication ...................................... 245
Attributes of national research agenda for
long-term care...........................................246
Key policy questions ................................. 248
Cost.................................................249
Effectiveness ...................................... 251
Cost-effectiveness ................................. 252
Conceptual framework and measurement ................. 253
Comprehensive Patient Descriptors .................. 253
Effectiveness measures ............................. 254
Cost measures........................................255
Duration of the research agenda ...................... 259
Control of the research................................260
Timeliness of the research.............................261
Dissemination of the research findings ............... 262
Centralized data system .............................. 262
Appropriate actions at the federal level ............... 263
Conclusion...............................................266
BIBLIOGRAPHY
267


XVI
CHAPTER PAGE
APPENDICES
A. Description of home health care services....................282
B. Data collection forms.......................................300
C. Data specifications.........................................307
D. Utilization analyses ...................................... 321
E. Multiple regression analyses .............................. 339
I


LIST OF TABLES
TABLE PAGE
I. At Risk Populations In Long-Term Care
Institutions and the Community, 1975-76 4
II. Selected Characteristics of Providers
in the Study Sample......................................103
III. Final Sample of AIT Episodes..............................104
IV. Data Specifications.......................................127
V. A Comparison of the Age of Massachusetts and
Philadelphia Patients .................................. 143
VI. A Comparison of the Living Arrangement of
Massachusetts and Philadelphia Patients ................ 145
VII. A Comparison of the Primary Diagnosis of
Massachusetts and Philadelphia Patients ................ 146
VIII. A Comparison of the Independence in
Activities of Daily Living of Massachusetts
and Philadelphia Patients .............................. 147
IX. A Comparison of the Activities of Daily Living Index of Massachusetts and Philadelphia Patients...............................................149
X. A Comparison of the Preadmission Location of
Massachusetts and Philadelphia Patients ................ 151
XI. A Comparison of the Change in Activities of Daily Living Index of Massachusetts and Philadelphia Patients ........................................ 153
XII. A Comparison of the Health Status at Discharge
of Massachusetts and Philadelphia Patients ............. 155
XIII. A Comparison of the Mean Intensity and the Length of Use of Massachusetts and Philadelphia Patients....................................................156
XIV. A Comparison of the Primary Source of Payment
of Massachusetts and Philadelphia Patients ............. 158


x v i i i
TABLE PAGE
XV. A Comparison of Total Visits Per Episode of Massachusetts and Philadelphia Patients . 162
XVI. A Comparison of the Cost Per Episode for Massachusetts and Philadelphia Patients . 164
XVII. A Comparison of the Average Cost Per Episode by the Age of Massachusetts and Philadelphia Patients . 182
XVIII. A Comparison of the Average Cost Per Episode by the Living Arrangement of Massachusetts and Philadelphia Patients . 184
XIX. A Comparison of the Average Cost Per Episode by the Primary Diagnosis of Massachusetts and Philadelphia patients . 185
XX. A Comparison of the Average Cost Per Episode by the Activities of Daily Living of Massachusetts and Philadelphia Patients . 187
XXI. A Comparison of the Average Cost Per Episode by the Activities of Daily Living Index of Massachusetts and Philadelphia Patients . 189
XXII. A Comparison of the Average Cost Per Episode by the Goal at Admission of Massachusetts Patients . . 191
XXIII. A Comparison of the Average Cost per Episode by the Change in Activities of Daily Living Index of Massachusetts and Philadelphia . 194
XXIV. A Comparison of the Average Cost Per Episode by the Health Status at Discharge of Massachusetts and Philadelphia Patients . 196
XXV. A Comparison of the Average Cost per Episode by the Primary Source of Payment for Massachusetts and Philadelphia Patients . 200
XXVI. A Comparison of the Average Cost per Episode by the Provider Size for Massachusetts and Philadelphia Patients . 202
XXVII. A Comparison of the Average Cost per Episode by the Acute Care Admissions in the Service Area for Massachusetts and Philadelphia Patients . 204
i
I


xix
TABLE PAGE
XXVIII. A Comparison of the Change in Activities of
Daily Living Index By the Age of Massachusetts
and Philadelphia Patients ........................... 206
XXIX. A Comparison of the Change in Activities of Daily Living Index By the Living Arrangement of Massachusetts and Philadelphia Patients ..................... 208
XXX. A Comparison of the Activities of Daily Living
Index By the Living Arrangement of Massachusetts
and Philadelphia Patients ........................... 209
XXXI. A Comparison of the Change in Activities of
Daily Living Index By Selected Primary Diagnoses of Massachusetts and Philadelphia Patients ............. 211
XXXII. A Comparison of the Change in Activities of Daily Living Index By the ADL Index Scores At Time of Admission of Massachusetts and Philadelphia Patients . 213


CHAPTER I
INTRODUCTION
Overview
A disturbing paradox exists in the long-term care delivery system. On the one hand, the system continues to absorb increasing amounts of the public and private dollar. On the other, it is riddled with controversy regarding inappropriate care, Medicare/ Medicaid fraud and abuse, and nursing home scandal.1 Expenditures for services grow, but the system seems to operate contrary to the promotion of the quality of life and provides little incentive for improvement.2 The system meets the mandate of policy makers and the expectation of providers so poorly that both groups advocate major alterations in the provision of and paynent for care.
An overhaul of the system demands the development and implementation of programs designed to assess needs, provide services at reasonable cost, and monitor quality to ensure that needs are
^U.S., Congress, Senate, Special Committee on Aging, Nursing Home Care in the United States: Failure in Public Policy, Nos. 1-4, 94th Cong., 1st Sess., Apri 1 , 1975 (Washington, D.C.: Government Printing Office, 1975).
^Sylvia Sherwood, "Long-Term Care: Issues, Perspectives and Directions," Long-Term Care: A Handbook for Researchers, Planners, and Providers^ ecL Sylvia Sherwood (New York: Spectrum Publications, 1975), p. 26.


2
actually met. The simultaneous optimization of long-term care programs, especially in the areas of cost and quality, is no easy task. Yet, the difficulty of the undertaking does not diminish the necessity for getting the various policy options (with their respective costs) on the table so that more informed decisions can be made.3 It is to this end that this study was undertaken.
In short, this study examines the cost-effectiveness of one of the policy options in long-term care: home health care. However, the study provides more of a vehicle for looking at the utility of evaluation research than making definitive statements about home health care. Further, it highlights the need for a comprehensive research strategy for future policy development in long-term care.
Because the long-term care system provides the policy framework for this study, it is important to review its development. The
balance of this chapter provides such a review, paying particular attention to major issues and problems. Home health care is the focus of the review, as is the potential importance of evaluation research. The second chapter examines recent research and evaluation which bear on this study. Chapters III-VI present the empirical research on home healtn care and include study methodology, findings, and discussion. Finally, Chapter VII suggests an overall research strategy for long-term care, which is based on systems analysis.
^U.S., Congress, Senate, Special Committee on Aging, Home Care Services for Older Americans: Planning for the Future, 96th Cong., TsT Sess., May T, 1979 (Washington, D.C.: Government Printing Office, 1979), p. 2.


3
The Long-Term Care System
Long-term care refers to services for the chronically ill and the physically and mentally handicapped whose conditions are not amenable to brief periods of treatment. It includes health and social services provided in an institutional and/or home-based setting. Examples of institutional long-term care include nursing homes (skilled nursing, intermediate care, and personal care facilities), residential or domiciliary care, adult foster homes, congregate housing, and boarding homes. Home (or community) based longterm care services generally include home health/home health aide services, homemaker/chore services, meal programs, social services, and sometimes transportation services and direct medical care. Because of the wide scope of services, long-term care is often treated as a composite of health and social services, with an emphasis on those service components generally identified as medical or health care services.
The Long-Term Care Population
The population at risk for long-term care services is varied, including the aged, physically and developmentally disabled, mentally retarded, and psychiatricly impaired, etc.^ (Table I presents a summary of recent estimates of the population in need of long-term care.) Yet, it is the functionally dependent elderly, those
^For a more complete discussion of the population at risk, see Judith LaVor, "Long-Term Care: A Challenge to Service Systems" (Washington, D.C.: Office of the Assistant Secretary for Planning and Evaluation, DHEW, 1976), pp. 6-15. (Mimeographed.)


4
Table I
At Risk Populations in Long-Term Care Institutions and the Community 1975-76
Nursing Homes (includes SNF and ICF) - 1,200,000
Chronic Disease Hospitals 25,000
VA Hospitals (Psychotic - long stay)3 25,000
Psychiatric Hospitals 250,000
Institutions for the Mentally Retarded 200,000
Non-institutional Population
Mentally Retarded (substantially handicapped) 670,000
Aged Needing Care at Home 3,400,000
Severely Mentally Disordered 1,000,000
Developmentally Disabled0
Cerebral Palsy 580,000
Epileptic 206,000
Other Neurological Disorders 600,000
aThis represents a minimum rather than a total for all VA patients.
^Does not include other categories of severely physically impaired.
Sources: U.S., Department of Health, Education, and Welfare,
Health, United States: 1975 (Washington, D.C.: Government Printing Office, 1976), p] 255; EThel Shanas, "Measuring the Home Health Needs of the Aged in Five Countries," Journal of Gerontology, XXVI (1971), 38; U.S., Department of Health, Education, and Welfare, Public Health Service, "Changes in the Age, Sex and Diagnostic Composition of the Resident Population of State and County Mental Hospitals. U.S. 1964-1973," Statistical Note 112, Publication No. (ADM) 75-158 ([Washington, D.CTi Government Printing Office, 1975]), pp. 6-7; Murray Ducket, Background and Analysis; Long-term Care for the Mentally Retarded (Washington, D.C.: The Urban Institute, 1975), p. T; U.S., Department of Health, Education, and Welfare, National Institute of Mental Health, "The Severely Mentally Disordered," by Valerie Bradley for use of the White House Conference for Handicapped Individuals (1976), p. 15. All sources cited by Judith LaVor, "Long-Term Care: A Challenge to Service Systems," Reform and Regulation in Long-Term Care, ed. Valerie LaPorte and Jeffrey Rubin (New York: Praeger Pub., 1979), p. 13.


5
individuals over 65 with illnesses or impairments that have become disabling, that are at greatest risk. In 1977 the 65 and over age group was estimated to comprise 11% of the total population, or 23.5 million persons.8 This percentage is expected to rise to 12%, or 31.8 million persons, by the year 2000.® The number of persons age 75 and over, often termed the "frail elderly," is increasing even more rapidly, from 41% of the total aged population in 1977 to an estimated 55% in 2000.7 Because chronic medical conditions, functional dependency, and psychosocial impairments are most prevalent among "frail" elderly persons,8 the increasing age of the population will have a significant impact on the future need for long-term care. Currently, about 2% of the 65-74 age group are residents of nursing homes; this figure increases to 7% for the 75-84 age group and to almost 20% for those over age 85.8 Overall, 87% of all nursing home residents in 1977 were over the age of 65, and 70% were 5 *
5U.S., Department of Commerce, Bureau of the Census, Statistical Abstract of the U.S., 1978 (Washington, D.C.: Government Print-ing Office, 1978), pp. 8-9. (U.S., Department of Commerce is here-
after referred to as Bureau of the Census.)
®Ibid.
^Institute of Medicine, The Elderly and Functional Dependency (Washington, D.C.: National Academy of Sciences, 1977), pp. 1-3.
institute of Medicine, pp. 1-4.
institute of Medicine, p. 3.


6
over the age of 75.^ Given current utilization patterns, the number of people who will at some time reside in nursing homes will increase substantially over the next several decades.
When one considers the nunber of elderly who reside in their own homes, but who are in need of community-based services, the demands placed on the long-term care system become even greater. Of the elderly living in the community in 1974, over three million reported limited mobility, one-third of whom reported the need for assistance by another person or a device for mobility. An additional one-third experienced some difficulty in routine daily functions. Of those individuals with restricted mobility, over one million were reported confined to their homes.^ Of the non-institutionalized
adult population in 1975, over three million elderly persons were
• . . . 1 ? estimated to need some kind of home care on a continuing basis.
It is important to note here that estimates of future nursing home utilization, based on current patterns and the availability of alternatives, are different from estimates based on functional dependency. The former assumes that utilization patterns will
^U.S., Department of Health, Education, and Welfare, Public Health Service, The National Nursing Home Survey; 1977 Simmary for the U.S., Publication No. (PHS) 79-1794 (Washington, D.C.: Govern-ment Printing Office, 1979), p. 29.
^Marie Callender and Judith LaVor, "Home Health Care: Development, Problems and Potential" (Washington, D.C.: Office of the Assistant Secretary for Planning and Evaluation, April, 1975), p. 3ff. (Mimeographed.)
^Saad Nagi, "An Epidemiology of Disability Among Adults in the United States" (Columbus: Ohio State University, Mershon Center, 1976), p. 4-10. (Mimeographed.)


7
continue, while the tatter is not necessarily tied to these patterns. The manner in which a person ultimately receives long-term care is a function of numerous factors including health status, family support, living arrangements, the availability of care alternatives, and funding patterns. Based on estimates of the need for att types of long-term care given current utilization patterns, it is expected that between 7.4 and 12.5 million adults will require long-term care services by 1985, with more than three million util-
1 o
izing institutional care.
As the number of functionally dependent persons increases over the coming decades, expenditures on long-term care can be expected to increase sharply. Concerns about the cost and quality of care become more important in light of the growing demands that will be placed on the health care system.
Rising Expenditures for Long-Term Care
Massive increases in total national expenditures for personal health care have occurred in the fifteen years since the inception of the Medicare and Medicaid programs. Increasing at an 11% annual rate on a per capita basis,^ they reached $142.6 billion in 1977,
^U.S., Congress, Congressional Budget Office, Long-Term Care for the Elderly and Disabled (Washington, D.C.: Government Printing Office, 1977), p. 8. (U.S., Congress, Congressional Budget Office is hereafter referred to as CBO).
^Ann Somers, "The High Cost of Health Care for the Elderly: Diagnosis, and Some Suggestions for Therapy," Journal of Health Politics, Policy and Law, III (Summer, 1978), 163-180.


8
an amount equal to 9% of the gross national product.* *5 The annual rate of increase of those expenditures during the past 10 years was 11% per capita, but most recently reached 13%.I6 Of all health care expenditures, 40% ($57 billion) were covered by public funds, primarily through the Medicare and Medicaid programs ($20.8 and $16.3 billion, respectively) .*7 The annual rate of increase of total Medicare and Medicaid expenditures between 1975 and 1977 was 17% (i.e., total costs rose by one-third in only two years),*8 compared to a 6.1% increase in the Consumer Price Index over the same time period.*9
A disproportionate share of health care expenditures is spent on care of the aged and disabled. Although they comprised only 11% of the total population in 1978, they accounted for 29% of all health care expenditures in that year.^ Nearly half of all pub 1 ic expenditures for personal health care were for the elderly. In addition, 67% of health care costs for the elderly were funded
^Robert Gibson and Charles Fisher, "Age Differences in Health Care Spending, Fiscal Year 1977," Social Security Bulletin, XLII (January, 1979), 3-14.
*&Somers, P* 163.
^Gibson and Fisher, p. 12. l^Gibson and Fisher, p. 10.
*9Sureau of the Census, p. 483.
20|J.S., Department of Health, Education, and Welfare, Public Health Service, Health, United States: 1978, Publication No. (PHS) 78-1232 (Washington, D.C.: Government Printing Office, 1978), pp. 396-400.


9
PI
through public programs, compared to 26-30% for other age groups.
Long-term care expenditures have increased at a greater rate than expenditures for all medical care. This is due in large part to economy-wide inflation and changing demographic patterns, in addition to increased access to care through the government financing described previously.
Institutional Care. In 1970, Medicare expenditures for nursing home care were S236 mi 11 ion.^ By 1977 these payments had reached $349 million^ and they are projected to rise $393 million in 1981.24 Because Medicare benefits are restricted to skilled nursing care, Medicare pays less than Medicaid for nursing home care. The Medicaid program, the dominant payer of nursing home care, expended $1.5 billion^ and $6.4 billion^ in 1970 and 1977, and is expected to spend $9.4 billion11' in 1981 (including the state share of program costs). While the estimated increase in Medicare expenditures for nursing home care from 1977 to 1981 represents an increase of
AGibson and Fisher, pp. 6-12.
^Bureau of the Census, p. 347.
^Gibson and Fisher, p. 12.
^Executive Office of the President, Office of Management and Budget, The Budget of the U.S. Government, Fiscal Year 1981, Appendix (Washington, D.C.: Government Printing Qrfice, 1980), p. 4/4.
^Bureau of the Census, p. 349.
2®U.S., Department of Health, Education, and Welfare (DHEW), Health Care Financing Administration, "HCFA Program Statistics," Health Care Financing Review, I (Fall, 1979), 74.
^Executive Office of the President, p. 470.


10
only 12.6%, Medicaid expenditures are expected to increase by 42.4%
over the same time period.
Home Health Care. Along with nursing home expenditures, Medicare and Medicaid expenditures for home health care have continued to increase over the years, although at a different rate. Medicare expenditures for home health care in 1970 were $70 mill ion,28 but had risen to $358 million by 1977.29 These are expected to rise to $890 million by 1981.Medicaid, which funds a much smaller proportion of home health care, incurred almost no expenses for home health care in 1970,21 spent $179 million in 1977,22 and will reimburse almost $339 million by 1981.22 jn addition to Medicare and Medicaid, most states pay for some type of non-medical in-home services under Title XX of the Social Security Act. In fiscal year 1976 these payments amounted to about $340 million.
While total expenditures for nursing home care are much larger than for home health care, the projected growth for Medicare and Medicaid home health expenditures in the 1977-1981 time period is
^Bureau of the Census, p. 347.
29qhEW, Health Care Financing Administration, pp. 72-73. â– ^Executive Office of the President, p. 474.
2lBureau of the Census, p. 349.
-^DHEW, Health Care Financing Administration, p. 74. 22£xecutive Office of the President, p. 474.
^U.S. Congress, Home Care Services, p. 126.


11
94% and 79%, respectively. This far exceeds the projected growth of nursing home expenditures in the same time period (12.6% and 42.4%, respectively). Similar to institutional long-term care, expenditures for home health care will continue to increase due, in part, to the increasing size of the disabled and elderly population. In addition, the rate of growth in these expenditures is expected to increase with the enactment of pending legislation liberalizing home health benefits. These changes alone will result in an estimated increase of $12 million per year in total Medicare home health expenditures by the year 1982.^
Total Costs. Current long-term care expenditures have reached an all time high and projections of future spending are expected to go even higher. A recent Congressional Budget Office report estimates that total costs (both public and private) of long-term care will run as high as $87 billion in 1985, with federal portions (assuming existing utilization patterns) over $15 billion.
The increases in expenditures for long term care have created a growing burden on the public sector. Pressures to balance the federal budget will likely prompt program cutbacks. Given the major role of government in the support of long-term care services, these cutbacks may have negative consequences in an area which is already 35
35
Executive Office of the President, Office of Management and Budget, The Budget of the U.S. Government, Fiscal Year 1981 (Washington, D.C.: Government Printing Office, 1980), p. 248.
â– ^CBO, p. xv.


12
criticized for suboptimal care. It will, therefore, be increasingly important to adopt program alternatives which provide for care of acceptable quality at reasonable cost.
Quality and Appropriateness of Care
The flow of government funds into long-term care over the last 15 years has not come without concerns over program design or costs. From their inception in 1965, Medicare and Medicaid were accompanied by certification regulations which set forth standards for quality aimed at limiting particpation to those facilities which provided at least minimum levels of quality medical care.37 The regulation of quality focused on external or structural regulations intended to ensure compliance with minimum safety and quality standards. Generally, these standards related to physical facilities, governing and management policies, personnel requirements and qualifications, and the provision of specific services.
Aimed initially at hospitals, Medicare certification was primarily based on the Joint Commission on Accreditation of Hospitals accreditation standards. In January 1967, Medicare certification was expanded to include post-hospital or extended care facilities. Of the more than 13,000 nursing homes which potentially qualified to provide extended care, only 740 were able to comply fully with the
37john Cashman and Beverlee Myers, "Medicare Standards of Service in a New Program - Licensure, Certification, Accreditation," /American Journal of Public Health, LVII (July, 1967), 1107-1117.


13
conditions of participation by July 1967.-^ That number increased to 1,350 by July 1969. Another 3,400 were in substantial compliance with the certification requirements and were conditionally certified in order to increase the availability of services. However, thereafter the nimber of Medicare extended care facilities decreased, and available beds in participating facilities declined.39 At the same time, the use of Medicare extended care services also declined.
The drop in utilization of services reflected changes in the Medicare conditions of participation for extended care facilities in 1969.The federal government was initially willing to pay for more extended care services (subject to lower standards) than planned. Increased costs and the provision of substandard care led to strengthened regulations in 1969 which, as intended, resulted in decreased utilization of services. The 1972 standardization of the regulations for the Medicare extended care facility and the Medicaid skilled nursing home (in which the Medicaid standards were ostensibly upgraded, all facilities becoming skilled nursing facilities) illustrated continued concern for quality.
38u.S., Congress, Senate, Committee on Finance, Medicare and Medicaid: Problems, Issues, and Alternatives, 91st Cong., TsT
Sess., February, 1970 (Washington, O.C.: Government Printing
Office, 1970), p. 94.
39U.S., Department of Health, Education, and Welfare, Social Security Administration, Health Insurance Statistics, Publication No. HI-75 (Washington, D.C.: Government Printing Office, 1977), p. 6.


I
14
Two years after the regulations were standardized, most certified facilities still did not fully comply with the regulatory guidelines. Questions about the adequacy of institutional care continue, and, if anything, are exacerbated by the levels of care legislated under Medicare and Medicaid. The intended step-down system from skilled to intermediate to personal (or maintainance care) has not materialized under the current reimbursement scheme, and home health care has been, until recently, virtually ignored.
In spite of early efforts to assure appropriate placement, a number of studies suggest that many patients in nursing homes are inappropriately placed (i.e., that they do not require the level of care provided) and, in some cases, could maintain a life in the community if the appropriate medical and social supports were provided. Estimates of the rate of inappropriate institutionalization are varied, ranging from a low of 6% in Minnesota^? to 76% in New
^U.S., Congress, Senate, Committee on Government Operations, Subcommittee on Federal Spending Practices, Efficiency and Open Government, Problems Associated with Home Health Agencies and the Medicare Program in the State of Florida (Washington, D.C.: Government Printing Office, August, 1976); UTS., Congress, House, Select Committee on Aging, New York Home Care Abuse, Publication No. 95-145 (Washington, D.C.: Government Printing Office, 1978).
^Robert M. Burton et al., "Nursing Home Cost and Care: An Investigation of Alternatives" (Durham, NC: Center for the Study of Aging and Human Development, Duke University Medical Center, July, 1974). (Mimeographed.)
i


15
York City.4^ A survey of eight Medicaid offices in New Jersey found that, overall, about 35% of the lower intensity patients studied were capable of maintenance in the community if provided appropriate home services^. The Congressional Budget Office conservatively estimates that 10 to 20% of patients in skilled nursing facilities and 20 to 40% of intermediate care facility patients require a lower level of care than they are presently receiving.^ Considering the proportion of the Medicaid long-term care dollar which pays for institutional care, it seems apparent that the money might be more efficiently spent.
Labeling a patient as inappropriately placed or receiving inadequate care raises many questions. First, who makes such a determination? Is it the professional care giver, the patient's family, or the patient himself? Various perspectives yield different judgments. For example, differences in training across professions (such as nursing and physical therapy) are known to affect professional judgments about the appropriateness of care.46 Second, 43 44 45 *
43J. T. Gentry and V. R. Curlin, "The Illinois Long-Term Care Classification Instrument: Use Experience Within the New York City Medicaid Program" (New York: Department of Health, Medical Section, Bureau of Health Care Services, May, 1975). (Mimeographed.)
44Urban Health Institute, "Appropriateness of Long-Term Care Placement: A Study of Long-Term Care Patients in the New Jersey Medicaid Program" (East Orange, NJ: Urban Health Institute, September, 1977). (Mimeographed.)
45CB0, p. 18.
4®A1an Sager, Learning the Home Care Needs of the Elderly: Patient, Family, and Professional Views of an Alternative to Institutionalization (Waltham, MA: Levinson Policy Institute, November, 1979), pp. 3-4.


16
is the decision based on normative or comparative standards? Conclusions do not always take into consideration the limited availability of in-home services, nor do they consider the quality of alternative services. Thus, comparisons may be among no care at all, or care of diminished quality.47 Finally, what cost is society willing to incur in order to increase the quality of care and/or numbers of patients appropriately placed? For example, in a 1974 study in Minnesota, Greenberg found that of the 18% of the skilled nursing facility patients who could be treated at home, less than half could be treated there for less cost than in a nursing home.48 49 In a study of hypothetical home care, the Levinson Policy Institute found that less than half of the nursing home patients reviewed could possibly be treated at home for less cost.^ While these and similar questions are not answered here, they do provide an
indication of the difficulties within the system. Together with the cost concerns mentioned, questions such as these have given impetus to the search for policy options.
47For a more thorough discussion of these issues see, Jay Greenberg, David Doth and Allan Johnson, A Coordinated Approach to the Delivery of Long-Term Care: Urban and Rural Models (Minneapolis: Center for Health Services Research, University of Minnesota, 1980). (Mimeographed.)
48Jay Greenberg, "The Costs of In-Home Services," A Planning Study of Services to Noninstitutionalized Older Persons~in Minnesota, ecT Nancy Anderson (Mi nneapo11s: Schoo ! of Public Affairs,
University of Minnesota, 1974), pp. 35-46.
49
Sager, pp. 240-252.


17
Home Health Care
The costly nature of institutional care, in addition to the inappropriateness of that care for certain types of long-term care patients, has motivated a look at alternative care modalities. Two alternatives, adult day care and home health care, have been subject to intense study in recent years.50 One of these--home health care--is reviewed below.
Background. The concept of providing health care in the home is not a new one in the United States. Its origins date back to the late 1700's when home services were provided on a voluntary basis by visiting community nurses. In the East and Northeast, these services fostered the development of the visiting nurse association (VNA). In the South and West, community nursing became a service provided by public health departments. Interestingly, even though other types of agencies have been organized and expanded, the VNA continues to be located predominantly in the Northeast, and the health department agency in the South. However, in recent years the number of hospital-based providers has increased substantially in the large urban areas.^
^William Weissert, Thomas Wan, and Barbara Livieratos, Effects and Cost of Day Care and Homemaker Services for the Chronically 111" A Randomized Experiment (Washington, O.C.: National Center for Health Services Research, DHEW, 1980), pp. 12-16.
John Hammond, Final Report: Applied Research in Home Health Care Services, Vol. Ill: Community Level Utilization Analysis, Publication No. (OPEL 79-3) (Washinaton, D.C.: Health Services Admin-istration, DHEW, 1979), pp. Ill.10-111.12.


18
Although the nunber of home health agencies and expenditures on home care have increased substantially since the passage of Medicare in 1965, the development of home care as a major component of the health care delivery system has yet to occur. Indeed, in 1977 public programs spent less than 10% of their funds on all home-based care. The result is that certified home health services are still not universally available. Except in the populous Northeast, home health services are only available to some 70% of the Medicare beneficiaries who live in non-metropolitan counties.33
The figures cited above should not be construed to mean that federal programs have had little impact on the development of home health agencies. To the contrary, in 1963 about 250 agencies would have been certified according to the regulations later set forth under Medicare.53 in 1966 approximately 1,275 agencies were certified. By 1975 that figure had more than doubled.
The growth of home health under federal programs has not been consistent, however. With the introduction of benefits for home-based skilled nursing care under Medicare, the nunber of certified agencies increased from 1,275 in 1966 to over 2,400 in 1969.54 As mentioned, the resulting cost of services provided by these and other health care agencies (principally hospitals and nursing homes) increased dramatically. Unplanned expenditures and the alleged * 53 54
52u.S., Congress, Home Care Services, p. 54.
53U.S., Congress, Home Care Services, p. 60.
54
U.S., Congress, Home Care Services, p. 63.


19
provision of substandard care led to the tightening of Medicare regulations. The definition of skilled nursing care provided in both institutions and at home was revised and more narrowly defined. Emphasis was placed on the intent that the Medicare home health care patient have rehabilitative potential. In other words, maintenance care was not covered. The strengthened regulations led to a significant reduction in services by most providers. Retroactive denials imposed by fiscal intermediaries caused some agencies to decrease or stop their participation in the Medicare program, while others were forced to declare bankruptcy. Between 1969 and 1973, the total number of Medicare certified home health agencies actually decreased, from 2,416 to 2,318 providers.^
With increasing attention to the escalating cost of institutional care, federal policy shifted again in 1972 to encourage growth in the home health care field. Amendments to Medicare in that year expanded post-hospital home health benefits to cover individuals with chronic renal failure. Also, the coinsurance payments of the program were eliminated. By 1973, these amendments resulted in significant yearly increases in federal reimbursement for home health care.56 The number of Medicare agencies again grew, and, as
^John Hammond, p. III.29.
56u.S., Department of Health, Education, and Welfare, Health Care Financing Administration, Office of Policy, Planning, and Research, Medicare: Utilization of Home Health Services, 1976,
Research and Statistics Note No. 2 ([Washington, D.C.: Government Printing Office, June, 1978]), p. 3.


20
of June 1978, 2,612 agencies had been certified to provide home
health care.
In summary, a review of the last decade of federal policy points to a certain degree of indecision about the extent to which home health care services should be allowed to grow under federal programs. Current policy shows a similar pattern. On the one hand, the Medicare-Medicaid Anti-Fraud and Abuse /Amendments of 1977 (P.L. 95-142) mandate a review of the delivery of home health and other in-home services provided under Medicare, Medicaid, and Social Services. A report submitted to Congress by the Department of Health, Education, and Welfare (DHEW) pursuant to that legislation (sometimes called the H.R. 3 Report), made no recommendations for the expansion of services primarily on the basis of budgetary constraints. On the other hand, Congress is currently entertaining the idea of expanding benefits.
The policy dilemma related to this area is not confined to the federal government. Providers differ among themselves on a nunber of questions (e.g., whether or not the focus of home health services should be confined to rehabilitative and skilled care; and whether or not the optimum mix of staff should be weighted toward professionals or paraprofessionals). The argument that proprietary (for-profit) agencies should be eligible for Medicare certification has also yet to be settled. At the forefront remain important determinations about what types of providers and for what type of patients home health care is an appropriate substitution for hospital or nursing home care. Although far from conclusive, several studies suggest that home health care may be an appropriate alternative for


21
a portion of the elderly and disabled population.^ Whether or not it substitutes for and is less costly than institutional care is another issue, which is discussed later in this report.
Comparative Costs of Home Health Care. The cost of home health services has been an issue in a nimber of studies. Findings to date indicate that there is some evidence to substantiate the claim that home care is less costly than institutional care for a portion of the long-term care population.
A 1977 General Accounting Office (GAO) study categorized 1,609 elderly people on the basis of their social, economic, mental, physical, and functional status. Other information, such as living arrangements, services received, provider of services, and direct service cost was also gathered. The costs of services furnished by family and friends were estimated at the agency rates. Comparisons were drawn between non-institutionalized and institutionalized people.
The GAO found that a direct relationship existed between the degree of impairment and the nunber and type of services received. At the lower levels of impairment, transportation, periodic screening, and social/recreational services were most important. At the extremely impaired level, social and recreational services were less important, while personal care, full-time monitoring and skilled nursing care were required. Support by family and friends proved to
^Chapter II of this volune contains a detailed review of the major studies in the field.


22
be increasingly important as the level of impairment rose. At all levels, over 50% of the required services were provided by family and friends. For the extremely impaired patients, only 20% of the services received were supplied by agencies. Of the average $845 per month that it cost to maintain these patients, $673 worth of services were furnished by family and friends.
The average cost to Medicaid for a patient in a nursing home in the GAO study was $458 per month. This figure was greater than the average total cost of maintaining all elderly persons in their own homes, except for those in the extremely impaired group. It should be noted, however, that costs for non-institutionalized patients included amounts to cover the services of family and friends, but these were actually donated services. In the most extreme group, the total cost of services provided oy an agency was $172 per month, a figure which represented only about 37% of the cost of nursing home care. Using the GAO calculations, only about 10% of the non-institutional ized elderly fell into the category where total cost (program and living) of home services was more than the cost of institutional care.59 when only costs to the government were considered, it appeared that an even larger number might be maintained at home for less than nursing home costs.
^Comptroller General of the U.S., General Accounting Office, Home Health: The Need for a National Policy to Better Provide for the Elderly, Publication No. HRD 78-19 (Washington, O.C.: Government Printing Office, 1977), pp. 9-22. (Comptroller General of the U.S. is hereafter referred to as GAO.)


23
The GAO findings are supported by a recent study completed by the Visiting Nurse Service of New York and the New York City Health Systems Agency. These agencies found that a disproportionate number of the population accounted for most home care costs (i.e, 10% of the population incurred almost 47% of the costs).60 However, contrary to the GAO report, neither living arrangements nor patient age was a relevant factor. The degree of impairment (in this study, defined as the amount of assistance necessary for carrying out certain activities of daily living) was the factor of primary importance in producing home care which cost more than institutional care. The study concluded that, for the 70% of the patients whose charges were less than the average (of $27 per day, inclusive of service and living expenses), home health care was considerably less expensive than most skilled nursing home care. A study of the Chelsea-Village Program, also in New York City, supported this hypothesis. It estimated that home-based care for semi-ambulatory patients costs only about 50% of nursing home rates.
Although there is some preliminary evidence that home health care may be less costly than institutional care on a per unit basis, whether it will be economically efficient in the long run is yet
^Geraldine Widmer, Roberta Brill, and Adele Schlosser, "Home Health Care Services and Cost," Nursing Outlook (August, 1978), pp. 488-489.
^Philip Brickner et al., "The Homebound Aged: A Medically Unreached Group," Annals of Internal Medicine, LXXXII (1975), 1-6; Linda Scharer, "Analyzing the Cost," Home Health Care for the Aged, ed. Philip Brickner (New York: Appleton-Century-Crofts, 1978),"pp. 229-245.


24
undetermined. Economic efficiency reflects the plausible, technologically efficient method of care given cultural, political and legal constraints and the preferences of patients and their families. It cannot be determined by merely costing out, at existing prices, all technologically efficient methods of care. Significant changes in the long-term care delivery system will likely affect prices by changing the demand and supply of resources and changing the behavior of patients, providers and families. Thus, potential changes in the delivery system through the expansion of home health services are likely to result in a higher total cost of implementing the expanded home program.62 Therefore, although patients and families request additional home care services, policynakers continue to appear reluctant to make large scale commitments, given the continued, although hopeful, uncertainty of its cost saving potential.
Major Barriers to Home Care. Similar to other long-term care services, the provision of home health care is hampered by the lack of a coherent benefit structure, service delivery and reimbursement
system. A series of regional OHEW public hearings in 1976 high-
✓
lighted these deficiencies. Among the problems identified were:
1) The absence of unified federal program definitions and eligibility requirements.
^tewis Freiberg, Jr., "Substitution of Outpatient Care for Inpatient Care: Problems and Experience," Journal of Health Politics, Policy and Law, IV (Winter, 1979), 493-494.



'


25
2) An incomplete knowledge of the system which leaves to the patient and family the task of assessing the appropriateness and quality of available services.
3) The absence of a coordinated range of high quality home care services to provide an essential continuun of care.^
As discussed, federal reimbursement for home services is spread over a nunber of different federal programs, including Medicare, Medicaid, and Title XX (Social Services) of the Social Security Act, and Title III (grants for State and Community Programs on Aging) of the Older Americans Act. Each of these programs has its own definition of the population served, and different requirements for program eligibility and the duration and scope of services. In addition, reimbursement methods differ. However, in spite of program differences, all of them are important to the provision of home care because certain of the services under each are necessary for proper support. No one of these programs provides the full scope of services needed to maintain the functionally impaired patient on a long-term basis in a non-institutional setting. While attempts have been made to bring the elements of these programs together, DHEW officials generally concede that the programs as they stand, given present legislation, defy coordination.^ 63 64
63U.S., Department of Health, Education, and Welfare, Home Health Care: A Report on the Regional Hearings, September 20 -
October 1, 1976, Publication No. 75-133 (Washington, D.C.: Government Printing Office, 1977), pp. 3-5.
64GA0, pp. i-iv.


26
Medicare and Medicaid both provide home health services, although these programs differ in essential aspects such as eligibility and coverage. Medicare hospital insurance (Part A) provides 100 home care visits per benefit period following the start of one spell of illness, only if the patient has been hospitalized for at least three consecutive days for this illness. Medicare supplemental medical insurance (Part B) allows 100 home care visits per year, but without prior hospitalization. The emphasis in Medicare is on the provision of short-term skilled care with little attention given to the social support necessary for long-term in-home maintenance.
Since 1970, home health care has been a required service under Medicaid. Unlike Medicare, it does not require for eligibility a prior hospitalization or the need for skilled nursing care. However, social services are not included. Both Medicare and Medicaid cover a large proportion of the medical and health related services, but tend to ignore necessary social support programs such as transportation and home delivery of meals.
Title XX of the Social Security Act offers a variety of home-based social services, such as home health aides, homemakers, personal care, etc. Title III of the Older Americans Act provides additional health-related services including health education, home repairs and meals either in the home or in a congregate setting.
The fragmentation and lack of coordination of services filters down to the community level where often a patient who could be maintained in the home is institutionalized for lack of a properly organized long-term care system. Thus, while a sufficiently broad


27
range of services is available from the combination of various federal programs to treat a chronically ill patient in the home, it becomes very difficult to do so at the local level. Services are provided by a number of different agencies, and seldom does a local mechanism exist for coordinating them. In addition, services such
as respiratory therapy, respite care, and transportation are not covered.
The 1977 Comptroller General's Report to the Congress on Home
Health Care summarized this issue and recommended a unified policy:
Home Health Care and other related services for the elderly are not being effectively coordinated. Services are available through so many different programs that effective coordination and delivery of home health and other in-home services seems close to impossible. Services provided are not accessible through a single entry point. Inter-agency/intra-agency agreements between federal, state and local agencies have not provided effective coordinated services to beneficiaries.
We recommend that the Secretary of HEW develop a national policy to be considered by the Congress which would consolidate home health activities. HEW should promote the estadishment of a comprehensive single entry system by which individuals are assessed, as to their needs, prior to placement in a program. HEW should consider the services that are currently provided under Titles VIII, XIX, and XX of the Social Security Act and Titles III and VII of the Older Americans Act.
Summary of Current Policy Issues
In summary, long-term care expenditures have risen rapidly in recent years. Increased spending for institutional long-term care, in particular, has been criticized. Dissatisfaction has surfaced
65
GAO, pp. 52-54.


28
with regard to the financial demands of institutional care, as well as the effectiveness of that care. There are major gaps in the knowledge about alternatives to that care. Important questions concern the numbers of people in need of the various kinds of longterm care services, and the quantity, and type of appropriate providers of such services. There is concern about the relative cost and effectivenss of institutional versus non-institutional care. Undoubtedly, uncertainties about cost, utilization, and effectiveness of non-institutional long-term care alternatives will most likely impede diversification of services in the short run. If one issue is clear, it is that insufficient information exists to address adequately all of the questions and concerns posed. It is to this insufficiency that this document now turns.
Evaluation Research to Reduce Uncertainty
The intent of evaluation research is to define, expand, assemble, and make more pertinent the information base upon which decisions are made. Its aim is to assess the effect of specific policy or program on a target population (e.g., individuals, organizations, communities, or systems).56 When used, it has the potential to decrease the uncertainty in decision-making. The assunption is that by providing data on the consequences of policies, the evaluator aids the decision-maker in selecting "wise choices about
^5Carol Weiss, Evaluation Research (Englewood Cliffs: Prentice-Hall, Inc., 1972), pp. 4-7.


29
future courses of action."67 The usefulness of evaluation research is then founded on the premise that decision-makers incur a certain amount of uncertainty when faced with choices. Uncertainty is that "gap" between what is known and what needs to be known to make a
correct choice.68 Indeed, there is much opportunity to reduce uncertainty in the area of long-term care.
Organized efforts to reduce uncertainty with regard to the effects of social programs are a relatively recent phenomenon.
Although program managers and policy-makers have traditionally judged the effects or values of their programs, little systematic
attention was paid to policy analysis/evaluation until the estab-
ft
lishment of the Planning, Programming, and Budgeting System (PPBS) in the Department of Defense in the early I960's.69 Thereafter,
Congress became increasingly interested in evaluation and passed major legislation which required and included supportive funds. For example, amendments to the Economic Opportunity Act of 1967 and to the Social Security Act in 1972 authorized wide-scale evaluation research efforts. Not surprising, the bulk of monies which fund policy research and evaluation comes from the federal government.
66 7Carol Weiss, "Where Politics and Evaluation Research Meet," Evaluation, I (1973), 43.
6®Ruth Mack, Planning on Uncertainty (New York: Wiley-Inter-science, 1971), p. 1.
69oavid Novick, ed., Program Budgeting (2nd ed.; New York: Holt, Rinehart & Winston, Inc., 1969), pp. xix-xxviii.



30
Thus far, the impact of most research in the area of non-insti-tutional long-term care has been minimal.^ There are at least two plausible explanations. First, confidence in policy analysts' projections of cost impacts of non-institutional services is minimal. This lack of confidence is not unwarranted given the low reliability of cost estimates for the Medicare and Medicaid programs in their early years. As noted, the estimated first year cost of the Medicare nursing home program was $25 to $50 million, less than one-fifth the final total cost.* 7^
Second, findings of recent research and evaluation studies in the area of non-institutional long-term care are conflicting. Thus, they often increase rather than reduce uncertainty on the part of policy makers.^ Uncertainty with regard to cost and potential benefits results in minimal efforts to expand services in the direction of any new program.7^ Evidence in support of this argument was the DHEW report to Congress on home health care cited earlier which eventually was stripped of all recommendations to expand service
7^0ne exception is the specific study described herein which did impact the Medicare home health care reimbursement system; but the effects were incremental compared to the enormity of the issues.
71U.S., Congress, Senate, Special Committee on Aging, Developments in Aging, 1969, 91st Cong., 2nd Sess., on S.J. Res. 316 (Wasn-ington: D.C.: Government Printing Office, 1970), p. 87.
7^See a review of representative studies in Chapter II of this volume.
7^Sheila Burke, "Home Health Politics and Policy: Senate Finance Committee" (paper presented at the meeting of the American Hospital Association on Hospital-Based Home and Hospice Care, Arlington, VA, September, 1979).


31
availability, ostensibly on the basis of uncertainty.74 Furthermore, recent experiences with the abuses and disappointments resulting from the deinstitutionalization of the mentally ill reinforces the argument for caution.
The existence of conflicting research and evaluation findings does not mean that these pursuits should be halted. On the contrary, as mentioned, the evaluation of long-term care alternatives has only recently begun. This study, hopefully, is a contribution to these efforts. While it may have a minimal impact on the actual resolution of major long-term care policy issues, it does suggest a systematic framework for a national program of policy-related evaluation research.
A Specific Research Area
Between 1976 and 1979, the Office of Planning, Evaluation and Legislation of the Health Services Administration (DHEW) supported home health care research relevant to that agency's responsibi1ity for promoting the development and expansion of home health service capacity. Part of that effort was a study of the cost per episode of home health care conducted by the Center for Health Services Research, University of Colorado Health Sciences Center. It is this study which is the vehicle for highlighting the long-term care * 7
74U.S., Congress, Home Care Services, pp. 51-68.
7^Ernest Gruenberg and Janet Archer, "Abandonment of Responsibility for the Seriously Mentally 111," Mil bank Memorial Fund
Quarterly, LVII (Fall, 1979), 485-506.


32
policy issues that need to be addressed through research and evaluation.
This study addressed several basic questions:
(1) How were the cost and utilization of home health care influenced by various patient, provider, and community characteristics?
(2) Which of the characteristics studied were the most important in terms of the observed utilization and cost variations?
(3) What are the relevant programmatic, public policy, and research implications of the findings?
These were pursued by analyzing how several groups of characteristics were related to the utilization and cost of care. Two assumptions were implicit: (1) that patient needs varied and were reflected by patient characteristics (e.g., age, living situation and functional status); and (2) that services delivered were in response to differences in need but were affected by provider/health system characteristics (e.g., size of provider and utilization of acute care in the service area). Outcome measures (e.g., change in functional status over time) represented the interface between patient characteristics, provider characteristics and the utiliza-tion/cost of home health care.
The study was thus intended to provide a better understanding of the determinants of the costs of home health care, which, in turn, should improve the reliability of estimates of home health care cost and utilization over the coming decade. In addition, the information pertaining to the types of patients/episodes which were


I
33
lower in cost was particularly helpful in clarifying and shaping public policies regarding the placement of patients in institutional versus home health care settings.
The analyses in the study were carried out using data from two samples. One sample was made up of four agencies (all visiting
nurse associations) participating in a discharge abstract program administered by the Massachusetts Department of Health. The second sample was comprised of four hospital-based home health providers participating in the Blue Cross of Greater Philadelphia Home Care Program. The two samples totaled approximately 2,800 episodes of illness during the study year, 1976.
In order to address the key questions of the study, relationships were examined in varying combinations and levels of detail. Descriptive statistics were used to test utilization and cost differences in terms of single variables. In order to determine the effect of individual variables in conjunction with others, factor analysis, canonical correlation analysis and multiple regression analysis were used to relate the utilization and cost of care to the three categories of variables.
A complete description of the study methodology, findings and implications is contained in the pages which follow. In order that the reader may more fully appreciate the significance of the study, a review of other research in the area of long-term care preceeds its presentation. In view of continuing long-term care policy
i
questions, this document concludes with suggestions for an overall research strategy intended to meet the needs of policy makers.


CHAPTER II
STATE OF THE ART Introduction
As discussed earlier, the primary focus of the study described herein was the cost-effectiveness of long-term care. Thus, the study was concerned with the refinement of institutional and non-institutional long-term care patient descriptors (often referred to as case mix), and measures of the quality and cost of care. The purpose of the following chapter is to provide an overview of the conceptual and analytic state of the art in each of these areas. This overview is presented in order to better understand how this study relates to the efforts of others. Although the topics of case mix, quality, cost, and cost-effectiveness are discussed separately, they are actually closely related to one another; thus a number of the studies cited pertain to multiple issues and are therefore mentioned several times.
Case Mix
Case mix typically refers to certain patient characteristics (e.g., problem, diagnosis, and functional status) which reflect the type and quantity of care needed. In theory, a strong relationship should exist between case mix and the quality and cost of health


35
care. However, while these relationships are conceptually clear, their empirical validation has been less successful in long-term care than for acute care.^ The following discussion of the problems of case mix will measurement points to specific areas of concern.
Age
In long term care, the age of a patient is related to functional incapacities,^ use of health services,* 3 and subsequent cost of care. Scharer and Boehringer found that the most
significant determinant of the cost per visit for home care was the client's age.1^ Another study by Widmer and others of home health care in New York City found that total costs of care actually decreased with advanced age, speculating that those who are elderly
^Larry Shunan and Harvey Wolfe, "The Use of Case Mix and Case Complexity in Prospective Hospital Reimbursement" (Pittsburgh, University of Pittsburgh, August, 1975). (Mimeographed.); Martin Feldstein, "Hospital Cost Variation and Case Mix Differences," Medical Care, III (April-June, 1965), 95-103; Robert Fetter et al., "Case Mix Definition by Diagnosis - Related Groups," Medical Care, XVIII, supplement (February, 1980), 1-53; Judith Lave and Lester Lave, "The Extent of Role Differentiation Among Hospitals," Health Services Research, VI (Spring, 1971), 15-38.
^Thomas Wan, "Age Severity of Disability," Review of Public Data Use, III (1975), 29-32.
3Karen Davis and Roger Reynolds, "Medicare and Utilization of Health Care Services by the Elderly," Journal of Human Resources, X (Summer, 1975), 361-377; Thomas Wan, "Interpreting a General Index of Subjective Well-Being" (paper presented at the meeting of the Gerontological Society, San Francisco, November, 1977). (Mimeographed); William Weissert, "Costs of Adult Day Care: A Comparison to Nursing Homes," Inquiry, XV (March, 1978), 10-19.
^Linda Scharer and John Boehringer, Home Health Care for the Aged: The Program of St. Vincent's Hospital, New York City (New
York: Boehringer Associates, 1976), pp. 9-11. (Mimeographed.)


36
require less costly "basic" care, at less frequent intervals than younger, more acutely ill patients.® It is reasonable to suggest that the relationship of age to utilization or general health services is bimodal, since both the very young and the very old use more health services than those in the middle age range. The direction of the relationship of age to cost of care is inconsistent in long-term care; nevertheless, it is frequently used as one of the many proxy measures of case mix.
Living Arrangement
A number of studies have found that the patient's living arrangement affects the quantity and type of care utilized.® Hence, living arrangement is another proxy measure often used for case mix used in long-term care, and home health care in particular.
The assistance provided by family and friends can signficantly reduce the time spent by professional care givers. With this help, reducing the amount of purchased services should reduce total costs; although not all utilization will be affected in the same manner by the patient's living arrangement. The New York study previously
^Geraldine Widmer, Roberta Brill, and Adele Schlosser, "Home Health Care Services and Cost," Nursing Outlook (August, 1978), pp. 488-489.
^William Weissert, Thomas Wan, and Barbara Livieratos, Effects and Cost of Day Care and Homemaker Services for the Chronical ly 111: A Randomized Experiment (Washington, D.C.: National Center
for Health Services Research, DHEW, 1980), pp. 14-15; Alan Sager, Learning the Home Care Needs of the Elderly: Patient, Family and Professional Views of an Alternative to Institutionalization (Waltham, MA: Levinson Policy Institute, 1979), pp. 241-248.


37
cited suggested that those living alone received essentially the same number of nursing visits, but twice as many social service visits as those living with others.7 Thus, living arrangement is one measure of case mix often found in home health utilization and cost studies; yet, its relationship to either utilization or cost is inconsistent.
Pi agnosis
Although commonly used to characterize the needs of acute care patients, medical diagnoses are not precise or comprehensive enough to measure case mix for chronically ill individuals for several reasons: the occurrence of multiple diagnoses; the difficulty of establishing a primary diagnosis; and the lack of relationship between primary diagnosis and the nature and extent of individual major care needs. Most long-term care studies have found no strong relationship between the primary diagnosis of the patient and utilization of services or personnel time in either the institutional^
7Widmer, Brill, and Schlosser, p. 490.
8peter Shaughnessy et al., Long-Term Care Reimbrusement and Regulation: A Study of Cost, Case Mix and Quality: Working Paper 4, First Year Analysis Report (Denver: Center for Health Services
Research, University of Colorado Health Sciences Center, 1980), pp. VI.4-VI.5; Jay Greenberg, Cost, Case Mix, Quality and Facility Characteristics in Minnesota's Nursing Homes: An Exploratory Analysis, First Year Progress Report (Minneapolis: Center for Health Services Research, University of Minnesota, 1980), pp. IV.31-IV.51. Bernard Ries and Jon Christianson, Nursing Home Costs in Montana: Analysis and Policy Applications (Bozeman: Montana State University, 1977), Iff.; Hirsch Ruchlin and Samuel Levey, "Nursing Home Cost Analysis: A Case Study," Inquiry, IX (September, 1972), 3-15.


38
or home health care setting.^ Thus, alternate measures of need have been developed, leaving primary medical diagnosis as a supplemental measure of case mix.
Functional Status
An accepted method of describing long-term care case mix is by the determination of individual capacities to function in several areas important to daily life (termed "activities of daily living", or ADLs). Early studies by Katz et al.*0 have since been augmented by other works*-'- which have added measures of functional status to the basic set of functions—bathing, dressing, feeding, continence, and mobility. For example, the adaptation by Lawton and Brody*^ which includes such functions as using a telephone, grocery shopping, and maintaining a checkbook, is especially useful for describing the case mix of a home health agency. Others have expanded the * **
Q
^Widmer, Brill, and Schlosser, p. 490; Nancy Anderson, A Comparison of In-Home and Nursing Home Care for Older Persons in Minnesota (Minneapolis: School of Public Affairs, University of
Minnesota, 1977), pp. 32-49.
^Sidney Katz et al., "Studies of Illness in the Aged: The Index of ADL," Journal of American Medical Association, CLXXXV (1963), 914; Sidney Katz and C. Ameci Akpom, “A Measure of Primary Sociobiological Functions," International Journal of Health Services, VI (1976): 493-507.
**Douglas Skinner and Donald Yett, "Debility Index for Long-Term Care Patients," in Health Status Indexes, ed. Robert Berg (Chicago: Hospital Research and Educational Trust, 1973), pp. 69-
82.
*^m. Powell Lawton and Elaine Brody, "Assessment of Older People: Self-Maintaining and Instrumental Activities of Daily
Living," Gerontologist (1969), pp. 179-186.


39
basic list to include a total of eight ADLs plus indicators of impairments in three sensory abilities believed important to the care needs of a long-term care patient.*3 Besides using simple measures of dependency/independency in each ADL category, Shaugh-nessy et al. used gradations of dependency in each.14 Preliminary findings of this and a companion study which used expanded ADL measures, indicated that resource consumption was more strongly related to moderate levels of dependence than .to total dependence. 15 An important effort in functional status measurement has been the attempt to condense the various measures of ADL into indices which maximize the retrievable information while reducing the total number of variables. Katz et al. in early work identified six ADLs which were empirically determined to be hierarchically related to one another, thereby permitting the classification of patients into homogeneous groups that were ranked ordinally.15 Skinner and Yett
l^peter Shaughnessy et al., An Evaluation of Swing-Bed Experiments to Provide Long-Term Care in Rural Hospitals, Volume II: Final Technical Report (Denver: Center for Health Services Research, University of Colorado Health Sciences Center, 1980), pp. V.21-V.59.
l^Shaugnnessy et al., Long-Term Care Reimbursement and Regulation, pp. VI.1-VI.22.
l^Shaughnessy et al., Long-Term Care Reimbursement and Regul a-tion, pp. IX.2-IX.26; Greenberg, Cost, Case Mix and Quality, pp. T3^T2.
^6Katz et al., "Studies of Illness in the Aged," pp. 914-916; Sidney Katz et al., "Program in Development of the Index of ADL," Gerontologist, X (1970), 20; Katz and Akpom, "Primary Sociobiologi-cal Functions," pp. 494-496.


40
derived a similar index through Guttman scaling of analagous d at a.^ Other efforts to define useful case mix groupings have used factor analytic techniques and other classification methodologies. Greenberg used canonical correlation to identify a vector of functional disability which was strongly related to nursing care costs.^ Presently, work in the area of functional status index development is being conducted by Shaughnessy et al. using principal component analysis and canonical correlation to group case mix indi-
. 1Q
cators into homogeneous groups.
Mental Health Status
The description of long-term care case mix has also been expanded through methodologies intended to evaluate the mental health or psychosocial status of long-term care patients. One tool developed by Pfeiffer is a short form for assessment of mental status to rule out organic brain impairment.20 Eisaorfer has argued that specialized, validated instruments are needed to measure depression, withdrawal, and other mental health problems in longterm care.21 Several of the multidimensional assessment systems
^Skinner and Yett, pp. 69-71.
l8Greenberg, Cost, Case Mix, Quality, pp. IV.1-IV.19.
l^Shaughnessy et al., Long-Term Care Reimbursement and Regulation, pp. IV.5-IV.6.
^Eric Pfeiffer, "A Short Portable Mental Status Questionnaire for the Assessment of Organic Brain Deficit in Elderly Patients," Journal of the American Geriatric Society, XXIII (1975), pp. 433-441.


41
which are described in the following paragraphs have recently added
such measures.
Multidimensional Health Status
An important advance in case mix measurement has been the development of multidimensional patient assessment tools that incorporate several of the measures described above, such as activities of daily living, mental health status, impairment indicators (e.g., paralysis and pain) and medical risk factors (e.g., elevated blood pressure).* 22 Additional major assessment methodologies which examine several different patient characteristics include the OARS (Older American Resources and Service) instrument developed at Duke University22 and the PACE II assessment instrument developed by DHEW. The former rates patients in five key areas of economic and social resource, mental and physical health, and activities of daily living. The latter is a complex instrument which uses an algorithmic model to determine the nature and extent of physical, functional, and mental health status. Most current systems of long-term
2^Carl Eisdorfer, "Evaluation of the Quality of Psychiatric Care for the Aged," American Journal of Psychiatry, CXXXIV (March, 1977), 315-317.
22Ellen Jones, Barbara McNitt, and Eleanor McNight, Patient Classification for Long-Term Care: User's Manual, Bureau of Health Services Research and Evaluation, DHEW (Washington, D.C.: Government Printing Office, 1974), pp. Ilff; Paul Densen and Ellen Jones, "The Patient Classification for Long-Term Care Developed by Four Research Groups in the United States," Medical Care, XIV (May, 1975), 126-130.
22Eric Pfeiffer, Multidimensional Functional Assessment: The OARS Methodology (Durham, NC: Duke University, 1975), pp. 3-29.


42
care assessment are closely related to, if not adaptations of, the multidimensional tools described above.2^
Problems
Recently, some researchers have begun to use problem-oriented case mix measures based, in part, on previous work by the Colorado Foundation for Medical Care.25 The system is based upon specification of the most prevalent institutional long-term care problems, including not only medical conditions (e.g., hypertension) and functional and sensory impairment (e.g., immobility and blindness), but also physical, psychological, and social conditions (e.g., pain, depression, and difficult family situations) requiring therapeutic intervention. Twenty-seven problems were found to represent approximately 97% of the patient care needs of long-term care residents.25 Such a problem-oriented approach has the advantage of being directly translatable into service needs, expressed in terms of type and frequency of service. For this reason a major advantage of case mix measured by the prevalence of each problem, is that (at least conceptually) it is directly related to the level of resources
“^Phyllis Giovannetti, Patient Classification Systems in Nursing; A Description and Analysis, Division o7 Nursing, DHEW, Publication No. (HRA) 78-22 (Washington, D.C.: Government Printing Office, July, 1978), pp. 6-10.
â– ^Kenneth Kahn et al., A Multi-Disciplinary Approach to Assessing the Quality of Life and Services in Long-Term Care (Denver; Colorado Foundation for Medical Care, 1975).
25Peter Shaughnessy et al., An Evaluation of Swing-Bed Experi-ments, pp. V.24-V.25.


43
required, and thus to costs incurred. Another advantage of the problem-oriented approach to case mix is that the extent to which normative service needs are actually met can be used as a process measure of the quality of care.
Case mix measurement in home health care has only quite recently expanded to include use of problems as major patient descriptors. The Visiting Nursing Association of Omaha, Nebraska, has been conducting a four year study intended to develop and validate uniform home health care problem descriptors for use in quality assurance programs.27 The problem list developed used factor analytic techniques to cluster 100 problems into four "domains", (i.e., physiological, health behaviors, psychosocial, and environmental). Other home health agencies group problems by prognosis in order to link them to quality assurance activities.23 The relationship of longterm care patient problems to resource consumption and patient outcomes is yet undetermined.
Levels of Care
A final case mix measure relates to current regulatory and reimbursement systems for nursing homes and home health care. As discussed in Chapter I, Congress established two levels of nursing
^Karen Martin et al., "Field Testing of a Problem Classification Scheme and Development of an Expected Outcome Scheme with a Methodology for Use," Draft Exective Summary (Omaha: Visiting Nurse Association of Omaha, September, 1979). pp. 2-6. (Mimeographed.)
28Frances Decker et al., "Using Patient Outcomes to Evaluate Community Health Nursing," Nursinq Outlook (April, 1979), pp. 278-282.



home care (skilled and intermediate) for Medicaid reimbursement
which were intended to be related to the patient's condition and the
intensity of services required for adequate treatment. The
requirement that the higher level of care (skilled) be reimbursed at a higher rate than the intermediate level of care assumed that it was more costly to treat skilled patients. However, these levels of care were not based upon empirical analysis.Uniform definitions of the two levels of care were never made precisely for all states-^0 and as a result, the classification of residents by level of care varies widely by state.^ Preliminary findings of Shaughnessy et al. substantiated the tendency for skilled nursing home residents to need more nursing care, but less psychosocial care than intermediate care residents.^ However, differences between the two were slight, indicating the limited utility of this system of level of care as a valid measure of case mix.
Ironically, as the skilled/intermediate classification system is being increasingly called into question for nursing home
2Q
Sharon Winn, "Assessment of Cost-Related Characteristics and Conditions of Long-Term Patients," Inquiry, XII (December, 1975), 344-353.
'^Thomas Willemain, Christine Bishop, and Alonzo Plough, The Nursing Home "Level of Care" Problem (Waltham, MA: Brandeis
University, 1979), pp. 92-95. (Mimeographed.)
â– ^Douglas Holmes et al., A National Study of Levels of Care in Intermediate Care Facilities, Final Report, Health Services
Administration, DH£w (Washington, D.C.: Government Printing Office, April, 1976), pp. 6-32.
^Shaughnessy et al., Long-Term Care Reimbursement and Regulation, pp. VI.6-VI.7.


45
patients, home health care providers are beginning to advocate a similar system. An ad hoc group of the five national home health care associations issued a position paper in 1979 calling for four levels of home care, similar to those found in institutional care: intensive, intermediate, maintenance and personal.33 The-relation-ship between such a level of care classification system and actual resource consumption thus requires further study.
In summary, although neither a single measurement system, nor a single set of case mix variables exists which adequately and accurately describe the health status of long-term care patients, certain classification systems have been found by researchers to be especially useful. Several of these appear to be related to resource consumption and thus offer potential for use in long-term care planning. By using multiple case mix descriptors, that are designed to measure medical care needs, functional disabilities, psychosocial, and environmental problems, the manner in which these relationships vary by service setting can be determined. This is especially important to determine which modality is best suited to the care of various kinds of patients. Unfortunately, much careful research is still needed to validate the various systems which have been proposed.
33Assembly of Ambulatory and Home Care Services of the American Hospital Association., et al., A Prospectus for a National Home Care Policy ([n.p.: n.n., 1978]), p. 3.


46
Qua!ity
Of the four major areas discussed in this section, the measurement of quality of care is the most controversial. For clarity, this section is arranged according to Donabedian's classic tricotomy (structure, process, and outcome),^4 so that the strengths and weaknesses of each type of measurement as applied to long-term care can be reviewed.
Structural Measures of Quality
Early approaches to quality assessment frequently focused on structural measures (sometimes called input factors) which are precursors to quality (e.g., availability of services, physical facilities, and staffing resources). Structural factors form the basis for institutional liscensure and certification, on the assumption that they represent necessary, albeit minimal, conditions for the delivery of adequate care. Institutional standards are based on structural factors, at least in part, because of ease of measurement. Factors such as ownership, number of beds, safety provisions, and staffing ratios have been used as measures of structural aspects of quality.
Research has failed to show significant associations between many of these measures and process or outcome measures of quality. (No studies to date have looked at this issue in home health care.)
^4Avedis Donabedian, "Evaluating the Quality of Medical Care," Milbank Memorial Fund Quarterly, XLIV (July, 1966), 166-180.


47
Several studies have found little relationship between nursing home ownership and the care provided to residents;-^ one investigator did find that non-profit church affiliated nursing homes provided better care.Holmberg and Anderson^7 and Dennis, Burke, and Garber^8 found that the training and experience of the person in charge were more important than structural measures of quality. Two other studies suggested that the size of the organization was inversely related to quality of care.39 Other studies indicated that the larger the facility, the better the care.40 Still others suggested * 3 4
36r. Hopkins Holmberg and Nancy Anderson, "Implication of Ownership for Nursing Home Care," Medical Care, VI (July-Auaust, 1968), 300-307; Winn, p. 348.
36[_eonard Gottesman, "Nursing Home Performance as Related to Resident Traits, Ownership, Size, and Source of Payment," American Journal of Public Health, LXIV (March, 1974), 269-276.
37Holmberg and Anderson, pp. 300-304.
â– ^|_yman Dennis, Robert Burke, and Kim Garber, "Quality Evaluation System: An Approach for Patient Assessment," Journal of Long-Term Care Administration, V (Summer, 1977), 28-51.
39shayna Greenwald and Margaret Linn, "Intercorrelation of Data on Nursing Homes," Gerontologist (Winter, 1971), pp. 337-340; Peter Townsend, The Last Refuge (London: Routledge and Kegan Paul, 1962).
4°W. Beattie and J. Bullock, "Evaluating Services and Personnel in Facilities for the Aged", Geriatric Institutional Management, ed. M. Leed and H. Shore (New YorlTi Putnan's, 1964), pp. 119-142.


48
that there was no relationship.4l A recent study by Linn, Guerel, and Linn‘S found that a higher professional staff-to-patient ratio, a better medical records system, and the provision of more services were positively related to discharge from institutional care.
Overall there is a weak relationship between various structural characteristics of nursing homes and the quality of care. Results of long-term care research support the general conclusion that the relationship between structural measures of quality and the delivery of care, and between care and its effects on patient health status, is not well established-^ For this reason, recent efforts have focused more intensively on process and outcome measures of quality.
Process Measures of Quality
Process measurement of quality entails the assessment of quality in terms of the resources consumed and procedures (i.e.,
^Timothy J. Curry and Bascom W. Ratliff, "The Effects of Nursing Home Size on Resident Isolation and Life Satisfaction," Gerontologist (Autumn, 1973), pp. 295-298; Beaufort Longest et al., 7rAn Empirical Analysis of the Relationship of Selected Structural Factors to Quality of Patient Care in Nursing Homes," Journal of Long-Term Care Administation, III (Spring, 1975), 16-26.
^Margaret Linn, Lee Guerel, and Bernard Linn, "Patient Outcome as a Measure of the Quality of Nursing Home Care," American Journal of Public Health, LXVII (April, 1977), 337-344.
^Avedis Donabedian, Needed Research in the Assessment and Monitoring of the Quality of Medical Care, NCHSR Research Report Series (Washington, D.C.: Government Printing Office, July, 1978), 2ff.; Robert Brook, Quality of Care Assessment: A Comparison of Five Methods of Peer Review (Rockville, MU: National Center for Health Services Research, DHEW, July, 1973) pp. 16-32; William McAuliffe, "Measuring the Quality of Medical Care: Process Versus Outcome," Milbank Memorial Fund Quarterly, LVII (1979), 118-149.


49
services) performed during the provision of care. Both data collection and the definition of criteria are more difficult for process measurement than in the structural analysis of the quality of care. However, process measures generally have greater validity, since they are more directly related to patient health status. Yet problems with the reliability and generalizability of process measures persist. Since process analyses generally use medical records as their primary source of data, it is difficult to differentiate between procedures that were performed but not recorded and procedures that were never performed. Linn et al.^ concluded in a recent study that both records and staff interviews should be used to collect data, since each was found by itself to be a valid, but incomplete, source of information.
In all areas of health care, the most prevalent technique for evaluating process quality of care is the comparison of process data to explicit criteria (or service standards) which have been generated by panels of multidisciplinary experts.4^ Inclusion of "exception" categories for each service permit the criteria to be applied individually to each long-term care patient. Using such a criteria
^Bernard Linn et al., "Validity of Impairment Ratings Made from Medical Records and from Personal Knowledge," Medical Care, XII (April, 1974), 363-368.
458rook, p. 46; Robert Kane, Prognosing the Course of Nursing Home Patients, proposal accepted for funding by National Center for Health Services Research, DHEW (Los Angeles: Rand, 1978). (Mimeographed.)


50
set, Shaughnessy et al.^6 found that the quality of care in nursing homes was slightly better than the quality of care in hospitals for certain long-term care problems.
Like structural criteria, process criteria represent a useful tool for evaluating the quality of care. However, the use of either structural or process measures is based upon the assumption that there is a causal relationship between them and the outcome of care as expressed in the final condition of the patient. As indicated, this causal relationship has not been empirically demonstrated. As an example, Linn and others'^ reported that patient outcomes were highly correlated with nursing time per patient (a process measure) but showed little relationship to the size of the facility, staff-to-patient ratios, personnel or safety practices (structural measures).
Outcome Measures of Quality
Outcome evaluation in patient care has received a great deal of attention recently as the most appropriate way to approach the concept of quality.^ A number of measures of outcome have been used in long-term care, such as admission to acute care, readmission to nursing homes, or prevalence of particular diseases or undesirable
46Shauqhnessy et al., An Evaluation of Swina-Bed Experiments, pp. V.5-V.30.
^Linn, Guerel, and Linn, pp. 363-368.
^Institute of Medicine, Assessing Quality in Health Care: An Eva!uation (Washington, D.C.: National Academy ot Sciences, Novem-ber, 1976), pp. 6-32.


51
physical occurrences, such as bed sores.Rather than measuring outcomes in terms of indicators of preventable morbidity, changes in patient health status have gained increased recognition. The simplest measure of health status change is, of course, mortality. Beyond this, however, patient status can include virtually any health-related characteristic. Although measures of physical wellbeing have been a primary focus of analysis, it has been increasingly recognized that the scope of outcome quality must be broadened to include such factors as psychosocial well-being, functional status, and, in some instances, satisfaction with care.50
Conceptually the measurement of outcome represents the ultimate validation of the effectiveness of the care process. However, two major problems exist in using outcome measures alone as a basis for effectiveness assessment: the difficulty in specifying appropriate outcomes of care and the difficulty in relating individual outcomes to outputs (or process of care).51 Both are discussed in the following paragraphs.
4Q
Sharon Soroko, "Quality of Care Evaluation in SNFs Using Bedsores as an Output Indicator" (paper presented at the meeting of the American Public Health Association, Washington, D.C., November, 1976). (Mimeographed.)
^Ian McDowell and Carlos Martini, "Problems and New Directions in the Evaluation of Primary Care," International Journal of Epidemiology, V (1976), 247-250.
^Donabedian, "Evaluating the Quality of Medical Care," pp. 166-206; Avedis Donabedian, "Patient Care Evaluation," Hospitals, XLIV (April, 1970), 131-136.


52
Specifying Appropriate Outcomes in General. Specifying appropriate outcomes of long-term care requires prior knowledge in two related aspects of health: the general course of health status expected of individuals at a given level of disability or age, and the expected response of individuals to treatment directed at a specific disease state. Individual outcome in long-term care settings can be measured by major categories of functional status (e.g., physical, psychosocial, and environmental health). These categories are attempts to move away from the earlier acute care criteria of mortality and morbidity rates toward characteristics more relevant to long-term care populations. The Index of Independence in Activities of Daily Living developed by Katz (described earlier) has been used alone and as part of more global assessment instruments to measure and compare changes in patient status over
time.^
Denson and Jones^ demonstrated that significant positive changes in ADLs could be achieved by patients during the several months following admission to a long-term care institution. The condition of nursing home residents is thus neither as static nor are their prospects as bleak as has generally been assumed. However, measurement which depends solely on ADLs is probably not sensitive enough to use in comparing the relative effectiveness of
^Sidney Katz and C. Amechi Akpom, "Index of ADL," Medical Care, XIV (1976), 116-119.
53Densen and Jones, pp. 126-130.


53
alternative care modalities. Mitchell,^ for example, reported that outcomes in various long-term care settings, measured by a behavioral index of health status, were not uniform across settings, but showed differential rates of improvement across sites. She concluded that findings were thus not generalizable beyond the range of the case mix characteristics of the patients examined.
Because of the variation in patient problems, outcome criteria must be specified in relationship to the particular problems which produce dysfunction. For example, degenerative joint disease may cause mobility restrictions which will never show improvement using an ADL scale. Nonetheless, appropriate treatment may produce meas-ureable results in terms of relief of pain, ability to engage in more social activities, and increased well-being. Valid prognostic expectations have been determined for only a limited number of conditions. Katz and Ford5^ and Katz et al.56 looked at medium-range outcomes after stroke and hip fracture, respectively. Further examination of post-stroke outcomes has occurred to evaluate the
^Janet B. Mitchell, "Patient Outcomes in Alternative Long-Term Care Settings," Medical Care, XVI (1978), 439-452.
^Sidney Katz et al., "Prognosis After Strokes, Part II: Long-Term Course of 159 Patients," Medicine, XLV (1966), 236-245.
^“Sidney Katz et al., "Long-Term Course of 147 Patients with Fracture of the Hip," Suraery, Gynecology, and Obstetrics, CXXIV (1967), 1219-1230.
!


54
57
effectiveness of various rehabilitation programs and to determine longer-range expectations.58 However, such studies are very expensive.
Establishing outcomes from a purely empirical base is both difficult and expensive. It requires that large quantities of data be collected and analyzed. Furthermore, analytic methods must be sensitive to identification of the variables which are policy relevant. Careful longitudinal studies appear necessary to validate multidimensional measures of patient status and to measure the effects of alternative methods of long-term patient care.
Such considerations have led to the use of expert opinion as an adjunct to empirical findings in establishing normative outcomes. An example of this is the outcome criteria set developed by the Colorado Professional Standards Review Organization in conjunction
^Carl Granger, Clarence Sherwood and David Greer, "Functional Status Measures in a Comprehensive Stroke Care Program," Archives of Physical Medicine and Rehabilitation, LVIII (December, 1977), 555-
561; J.F. Lehman et al., "Stroke: Does Rehabilitation Affect
Outcome?" Archives of Physical Medicine and Rehabilitation, LVI (August, 1975) j 375-382; Osvaldo Mig 1 ietta, Tae-Soo Chung,and
Vemireddi Rajeswaramma, "Fate of Stroke Patients Transferred to a Long-Term Rehabilitation Hospital," Stroke, VII (January-February, 1976), 76-77.
CO
Eugene Moskowitz, Forrest Lightbody, and Nanci Freitag, "Long-Term Follow-Up of the Poststroke Patient," Archives of Physical Medicine and Rehabilitation, LI 11 (April, 1972), 167-172;
M. Newman, "The Process of Recovery After Hemiplegia," Stroke, I IT (1972), 702-710.


55
with its pilot study of quality in nursing homes.^ Another study which intends to combine projections with empirical evidence is currently underway.Similar outcome criteria, based upon expert judgement, are being developed and tested by Shaughnessy et a!.®*
Specifying Appropriate Outcomes for Home Health Care. Only recently have home health care providers begun to develop outcome criteria for evaluation of the quality of home health care. Based on earlier work which focused on process measures of quality and using expert opinion, Daubert^^ developed outcome criteria for home care. One such criteria set was developed by a voluntary statewide program in Minnesota for 33 different conditions, such as congestive heart failure, depression, and immobility.^3 Unfortunately, the vast majority of patients did not meet their outcomes, primarily because their course was atypical due to the presence of exceptions.
One group that has specified exceptions is the Pennsylvania Assembly of Home Agencies, which began a statewide quality assurance * 50
^Colorado Foundation for Medical Care, The Colorado Experience in PSRO Long-Term Care Review (Denver: Colorado Foundation for Medical Care, October, 1978), pp. A.1-A.25.
50Kane, p. 18.
^Shaughnessy et al., Long-Term Care Reimbursement and Regulation, pp. VII.17-VII.18-
®2Elizabeth A. Daubert, "A System to Evaluate Home Health Care Services," Nursing Outlook, (March, 1977), 158-171.
63
Decker et al., pp. 278-282.


56
project in 1974.64 using expert opinion, participants in the project developed outcome criteria sets for twelve audit topics, including arthritis, cardiovascular disease, hip fracture, and mental illness. Each audit topic included up to thirteen essential outcomes by which to assess the care provided by the home health agencies. The outcome criteria were used to evaluate care given to over 2,500 patients during the following two years in order to compare outcome quality scores across providers. The Pennsylvania Assembly project included in its outcome criteria standards complications which would preclude the meeting of expected outcomes. Thus far, the criteria sets continue in use in western Pennsylvania for the evaluation of care in on-going home care programs, but they have not been used to evaluate the outcome of care as it relates to the process or cost of care.
The Visiting Nursing Association of New Haven has also developed a set of outcome criteria for five types of home health care patients, grouped according to health care problems and rehabilitative potential.65 For example, one group includes patients with acute, non-chronic diseases or disabilities who are expected to return to pre-illness functioning. The ultimate objective (outcome criteria) for this group is the complete elimination of the existing
64Rita Berkoben, "Home Health Care and Quality Assurance: The Experience of the Pennsylvania Assembly Project," Quality Review Bulletin (October, 1977), pp. 25-28.
^Elizabeth A. Daubert, "Patient Classification System and Outcome Critera," Nursing Outlook (July, 1979), pp. 451-453.


57
health problems. Another group of patients for whom outcomes are specified includes those with chronic diseases who, even though a return to pre-illness level of functioning is not possible, have the potential for increasing their level of functioning and may eventually remain in the community without the need for home health services.
The most recent activity in the development of outcome criteria for the evaluation of home health care is the ongoing program of the Visiting Nurse Association (VNA) of Omaha, Nebraska described earlier.66 Through the empirical analysis of problem and outcome data for all patients in four agencies in Nebraska, Iowa, Delaware, and Texas over a two year period, the VNA developed specific outcome criteria for each of the problems commonly found in home health care. Outcomes were then clustered by problem group to identify those outcomes which were met in the majority of cases over the period of time. At this writing, these outcome criteria are being field tested.
Tenuous relationship Between Process and Outcome. Beyond the specification of appropriate outcomes for both institutional and home based care, the second major problem surrounding the use of outcome measures of quality is the relationship between the process and outcome of care. Many factors, in addition to specific treatments and services, have an effect on health status in long-term
65
i
Martin et al., pp. 2-8.


58
care; major influences include the interaction of multiple diagnoses, emotional status, environmental adequacy, intervening life events, and attitude toward, knowledge about, and willingness to comply with, the necessary care regimen. For this reason, although the measurement of outcome has obvious face validity as the ultimate measure of effectiveness, the degree to which the outcome in any particular case is actually attributable to the treatment is typically a matter of conjecture.
These concerns have led to the continuing debate over the relative advantages of using process, as opposed to outcome measures, in assessing the quality of care,67 * and as a basis for corrective actions and policy decisions.^8 The issue has not yet been resolved for any modality of care, least of all for long-term care. Two recent studies found a curvilinear relationship between process and outcome measures of quality, which indicates the presence of a "threshold" level of process quality below which a poor outcome is much more likely and above which the relationship is less predictable.69
67McAuliffe, pp. 118-149.
66Robert H. Brook et al., "Assessing the Quality of Medical Care Using Outcome Measures: An Overview of the Method," Medical Care, XV, Supplement (September, 1977), 6-45.
69lisa Rubenstein, Susan Mates, and Victor Sidel, "Quality of Care Assessment by Process and Outcome Scoring," Annals of Internal Medicine, LXXXVI (1977), 617-625.


59
To confound further the process-outcome association, changes in the health status of patients over time were found to be most significantly related to the initial characteristics of the individuals being treated in several studies. In hypertension'’0 and
congestive heart failure7-*- measures such as age and diastolic blood pressure were stronger determinants of patient status at discharge than the actual care provided to patients. Similar results were noted in the evaluation of adult day care and homemaker programs, where both initial case mix (measured by functional status) and utilization of acute care services were more strongly related to improved health status than was use of the program.70 * 72 Consistent findings emerged from a study of home health services in which younger, less disabled, and medically complicated patients
70Fred T. Nobrega et al., "Quality Assessment in Hypertension: Analysis of Process and Outcome Methods," New England Journal of Medicine, LLXLVI (January 20, 1977), 91-113; S.W. Metcher et al., "Predicting Blood Pressure Control in Hypertensive Patients: An Approach to Quality of Care Assessment," Medical Care, XXVII (March, 1979), 285-292.
71F.J. Romm et al., "Correlates of Outcomes in Patients with Congestive Heart Failure," Medical Care, XIV (1976), 765.
72Ihomas Wan, William Weissert, and Barbara Livieratos, "Geriatric Oay Care and Homemaker Services: An Experimental Study," Journal of Gerontology, XXXV (March, 1980), 256-274.


60
experienced more favorable changes in health status, irrespective of
utilization.
These studies make a strong argument that attempts to prognose the course of long-term care should take into account the initial characteristics and overall individual potential in setting outcome objectives. Changes in physical impairment level are sometimes
meaningfully measured in terms of ambulation. For patients for whom ambulation was not possible, a change in the muscle strength of an affected limb may be considered a desirable outcome. Eliminating depression may be an appropriate index of progress for some patients, but may be irrelevant for those who have been depressed their entire lives. In summary, each patient possesses a unique
combination of disabilities, accompanying medical complications, specific psychosocial and demographic characteristics. Measures used to categorize patients with respect to the effects of treatment should incorporate this uniqueness.
Cost of Care
This section summarizes a number of major studies on the cost of institutional and non-institutional long-term care. Because nursing home cost studies have served as the conceptual basis for
cost studies in home care and are more numerous, the beginning of
7T
Sidney Katz et al., Effects of Continued Care: a Study of Chronic Illness in the Home, Publication No. (HSM) 73-3010 (Washing-National
ton,
1972)
D.C.
, PP-
Center for Health Services Research, DHEW,
23-42.


61
this section primarily discusses nursing home cost studies. The few home health care cost studies are then presented. The final portion of this section discusses studies which compare expanded home health care, traditional home health care, and institutional care in terms of cost and/or effectiveness.
Nursing Home Cost Studies
The majority of existing studies of the cost of nursing home care have concentrated on identifying the various factors influencing the cost of that care. Most use cross-sectional multivariate analytic techniques in order to estimate cost functions. They generally use one of several measures of cost as the dependent variable and a variety of community, facility, and patient characteristics as independent variables.74 Several studies have also tried to include measures, however minimal, of quality of care. The discussion which follows presents the general analytic approach used in these studies and summarizes their findings.
Facility Level Cost Studies. Most nursing home cost studies use the facility as the unit of analysis and the average cost per patient day as the dependent variable. Because there is wide variation in property cost per day across nursing homes, generally due to differences in accounting practices, most studies have focused on operating costs. (Two exceptions were the studies by
74For a review of the use of cost function analyses for nursing home care see, Martin Knapp, "Cost Functions for Care Services for the Elderly," Gerontologist, XVIII (1978), 30-35.


62
Skinner and Yett* 77 78 and McConnel78 which estimated cost functions that used average total cost as the dependent variable.) Although the separate components of operating cost are sometimes analyzed, researchers tend to concentrate on total operating costs. Operating cost components most often used include nursing costs,77 laundry and linen, and dietary.78
Cost functions are the empirical representation of the relationship between the cost of production of services and the level of output, obtained from a multiple regression of total costs (per day) on outputs and other important input influences. The output (of the firm) is assumed to be the patient day of nursing home care and is viewed as varying in relation to the case mix and the quality of care provided. Other factors are then included in the cost function to reflect differences in input prices (e.g., nursing wage costs), efficiency (e.g., occupancy rates), economies of scale (e.g., bed size), and management objectives (e.g., for-profit or non-profit
78Skinner and Yett, p. 79.
^Charles McConnel, "Cost, Size and Quality Structure of
Nursing Home Industry," Journal of Health and Human Resources Administration (November, 1978), pp. 134-149.
77Ruchlin and Levey, pp. 3-15; Moreland Act Commission on Nursing Homes and Residential Facilities, Long-Term Care Regulation: Past Lapses, Future Prospects (New York: Moreland Act Commission on Nursing Homes and Residential Facilities, 1976), pp. 22-27.
78Ries and Christianson, pp. 13-24; Howard Birnbaum et al., Reimbursement Strategies for Nursing Home Care: Developmental Cost Studies: Volume I - Final Report~(Cambridge, MA: Abt Associates,
1979), p. 49.


63
ownership and control). Thus, the factors used in most cost studies can be classified into four general groups: environmental/community factors; facility characteristics; patient characteristics (case mix); and measures of the quality of care.
1) Environmental and Community Factors. Environmental and community factors include measures of location and population density. For example, Ries and Christianson^ usecj location in a town with population greater than 2,000 in their study of Montana nursing homes, while Shaughnessy et al.80 and Birnbaum et al .81 used location in an SMSA. Input prices (e.g., wages) were used in several studies,as were regulatory characteristics, such as the operation of a certificate of need program.83 The study by Birnbaum et al. was especially comprehensive in its approach to environmen-tal/community factors, also using the presence of life safety codes, minimim staffing requirements, and prospective reimbursement systems in its cost functions.84 The significance of factors other than input price in this category varies considerably throughout the studies cited. 79 * * * * 84
79Ries and Christianson, pp. 20-21.
80shaughnessy et al., Long-Term Care Reimbursement and Regulation, p. VIII.10.
81-Birnbaum et al., p. 90.
^Moreland Act Commission, p. 24; Birnbaum et al., p. 90.
83Birnbaun et al., p. 90.
84Birnbaum et al., pp. 39-40, 90-91.
\


64
2) Facility Factors. In terms of facility characteristics, most studies included indicators of facility size, such as the number of beds,85 average daily census®6 and occupancy rate.®'7 Other facility characteristics investigated have included financial status®® and ownership,®9 (e.g., for profit and non-profit). Studies have typically found an association of non-profit ownership with higher cost, controlling for other factors. Results of the other facility characteristics, including size and occupancy rate, are inconsistent across the studies.
3) Case Mix Factors. One of the more difficult groups of
factors to obtain for cost function studies has been adequate measures of patient characteristics. Thus, case mix measures used in cost function studies have included both direct and indirect (proxy) measures. Examples of proxy measures used include the * * * * * * * * * * * *
^5Ries and Christianson, pp. 21-22; Birnbaum et al., p. 93;
Moreland Act Commission, p. 24; Thomas J. Walsh and Michael
Koetting, "Patient Related Reimbursement for Long-Term Care,"
(n.p.: Illinois Department of Public Health, 1978), p. 9.
(Mimeographed.)
^McConnel, pp. 134-149; Birnbaum et al., p. 95.
®^Shaughnessy et al., Long-Term Care Reimbursement and
Regulation, pp. IX.16-IX,18; Birnbaum, p. 97; Ries and Christianson,
p. 21; Ruchlin and Levey, pp. 3-15.
^®Moreland Act Commission, p. 24.
®%ies and Christianson, p. 21; Ruchlin, pp. 3-15; Walsh and
Koetting, p. 78


65
average patient age,90 the level of care classification (skilled or intermediate) ,91 the ratio of skilled nursing care beds to total beds,92 and the percentage of non-ambulatory patients.93 These measures have generally led to inconsistent results.
Several studies have used case mix indicators which directly measure patient care needs. These measures include the patients'
functional abi1ities94 (measured by an ADL Index), medical diagnoses 95 mental status,96 and long-term care problems.97 Findings with respect to functional status were consistent across studies and suggested that it was strongly related to the cost of the nursing home care, with cost generally being highest for those patients in the intermediate range of dependency. Inconsistent results were obtained for other direct measures of case mix, except for medical diagnosis, which was generally not strongly related to cost.
90Moreland Act Commission, p. 23; McConnel, p. 143.
91walsh and Koetting, p. 9; Birnbaum et al., p. 101.
9^Ries and Christianson, p. 21.
93Birnbaun et al., p. 93.
94Skinner and Yett, pp. 69-75; Birnbaum et al., pp 87-88; Shaughnessy et al., Long-Term Care Reimbursement and Regulation, p. III.18; Walsh and Koetting, pp. [16-18J.
^6Birnbaim et al., p. 87.
96Birnbaum et al., p. 88; Shaughnessy et al., Long-Term Care Reimbursement and Regulation, p. 111.18.
97Shaughnessy et al., Long-Term Care Reimbursement and Regulation, p. 111.18.


66
4) Quality of Care Factors. Quality measures are the final group of variables included in cost analyses and have been the most difficult to measure. As might be expected, the most common approach has been to use structural measures, such as the availability of therapeutic services^® or staff-to-patient ratios.99 The proportion of Medicaid patients or patient days was also included as a quality indicator, under the assumption that the greater this proportion, the lower the quality of care.100 The findings of studies using measures of Medicaid utilization substantiate this assumption. However, whether this reflects lower quality or simply greater cost containment pressures is yet unclear.
Few nursing home cost function studies have used process or outcome measures of quality. There are at least two reasons for this omission. First, process measures of quality generally require extensive primary data collection efforts which are usually beyond the scope of (economic) cost function analyses. Second, outcome measures of quality of care require that measures be taken over more than one point in time; this is also beyond the scope of cross-sectional cost function analyses (and their major defect). Consequently, there is only one known study, by Shaughnessy et
QO
Walsh and Koetting, pp. [16-13]; Birnbaum et al., pp. 89-90. "McConnel, pp. 142-145. lOOMcConnel, pp. 143-145.
I


67
al.,^ that used either of these types of quality measures. First year findings of this study indicate a significant and positive relationship between process quality and the cost of nursing home care. In the second year of this study, outcome measures of quality will be added to the analysis.
Cost function studies have thus found that case mix and quality are related to the cost of nursing home care, but methodological and measurement weaknesses of many of these studies suggest that concepts and approaches must be refined before definitive findings will emerge. For example, none of the studies cited above used a comprehensive approach to measurement of case mix and quality of care, while controlling for exogenous factors such as community size or regulatory environment. Two studies included no case mix factors, ^ and several included no qua!ity measures.^ other studies were weak analytically,^ ancj an were cross-sectional, rather than longitudinal. The results were insufficient to address the questions of the relationship between quality of care (effectiveness) and cost, for various types of patients.
l^Shaughnessy et al., Long-Term Care Reimbursement and Regulation, p. III.21.
^Ries and Christianson, p. 21; Ruchlin and Levey, pp. 3-15.
103^oreland Act Commission, pp. 23-25; Skinner and Yett, pp. 69-89; Ruchlin and Levey, pp. 3-15.
104
McConnel, pp. 134-149.


68
Patient Level Cost Studies. Very few nursing home cost studies have been concerned with cost per day at the patient level. Examples of these few are Vannell et al.^5 ancj McCaffree et al.^6 All found significant time (and, by implication, cost) differences across various case mix categories. One reason so few nursing home cost studies have been completed at the patient level is that total cost data are available usually only at the facility level. Thus, patient level analyses require direct observation of staff time required for various patient care activities (e.g., time and motion studies).
Several studies completed as part of the development of the Medicaid reimbursement systems in Illinois, Ohio, and West Virginia have added to estimates of time, weighting factors reflecting intangibles and the various skill levels of providers, calculating cost by multiplying the weighted time figures by a prevailing wage rate.^'7 None of these studies used the resulting cost per patient day in a full scale cost function analysis. Thus, these studies strongly suggest the need for additional analyses at the patient level.
^°^David Vannell et al., Final Report, The Patient Evaluation Review Committee Project 1977-1979, Section I (n.p.: Wisconsin Department of Health and Social Services, August, 1979).
^Kenneth McCaffree et al., Cost Data Reporting System for Nursing Home Care: Final Report, Publication No7 (hPa) 77-3169
(Rockville, MD: National Center for Health Services Research, DHEW, 1977).
^Donald St. John, "The Illinois Automated Long-Term Care System - Three Years of Experience," Medical Care, XIV, Supplement (May, 1976), 192-197.


69
Home Health Cost Studies
No home health cost studies to date have analyzed cost at the provider level. Contrary to the situation for nursing home costs, the few studies of home health cost that do exist are at the patient 1 evel.
Two studies in particular have dealt with the cost of home health care independent of comparisons to nursing home care costs. Both focused primarily on case mix, facility/provider, and environmental characteristics in relationship to cost, omitting any measures of quality of care. The first looked at unit costs as part of an evaluation of the federal Home Health Grant Program. Part of the study included the analysis of factors affecting the cost per skilled nursing care visit in a small sample of grantee and control group agencies. In that study the major factor found to be related to cost per visit was input price, as measured by nursing salary. However, other factors, such as case mix, were not taken into account in the analysis.^8
The second project that studied the cost of home health care was the recent experiment completed in New York City (described earlier in this report).In that study the relationship of the total cost of home care to the disability level of the patient was analyzed. Cost included expenditures for a wide array of in-home
^Robert Schlenker et al., Applied Research in Home Health Services Volume I: Grant Program Evaluation (Denver: Center for Health Services Research, University oT Co lor ado Health Sciences Center, 1979), pp. VII.23-VII.27.


70
services, including homemaker and chore services, in addition to an estimate of living expenses (i.e., room, board and transportation). Findings, based solely on univariate statistical analyses, indicated that the most highly disabled group of patients, which comprised 10% of the total study population, incurred almost 47% of the costs of the entire project. There was generally a significant difference in the cost of care by functional status; per diem cost for the least disabled patient was $11 while the per diem cost for the most disabled was $20.1^9 whether the relationship between cost and functional status is linear is still uncertain since other studies found that costs for the most disabled patients were lower than costs for those slightly more functional
One unusual study was recently completed at the Levinson Policy Institute. In that study, the estimated cost of home health care (including hypothetical direct program expenditures and costs of living costs) for 50 patients was analyzed. Factors which significantly increased cost were the functional dependency and the age of the patient at admission to home care. On the other hand, the greater the number of individuals residing at home with the patient, the lower the cost. No other case mix or environmental factors were significant in the cost functions. No quality of care measures were
^Widmer, Brill, and Schlosser, p. 490.
^A.L. Creese and R. Fielden, "Hospital or Home Care for the Severely Disabled: A Cost Comparison," British Journal of
Preventive and Social Medicine, XXXI (1977), 116.
i


71
|
included in this study because patients did not actually receive
home health services.m
i
No doubt, there remains significant room for methodological improvement of home health cost studies. None of these studies reaches the level of methodological sophistication of the nursing home cost studies cited previously. Researchers interested in home health care cost analysis would do well to build upon the strengths of these earlier studies (and take account of added suggestions in the last chapter of this volume).
Cost-Effectiveness of Long-Term Care
Cost-effectiveness studies are an attempt to relate the cost of a program or service to the outcomes, or effectiveness, of that program or service. One major component of any cost-effectiveness study is the measurement of effectiveness. Another is the determination of the relationship between effectiveness, however measured, and the services whose impact is being assessed. (That is, it must be determined that the observed outcomes are the result of the services provided and the costs incurred for them). The final component of significance is the appropriate definition of cost.
The adequate completion of a cost-effectiveness analysis depends upon the ability of the analyst and program administrators to determine measureable program objectives (e.g., a positive change in health status or a reduction in length of stay), alternative ways
Sager, pp. 241-248.


72
of attaining those objectives (comparison of alternative programs), and economic constraints on the program (the value of resources available). These requirements suggest that alternatives compared should be appropriate, that the dimensions of costs and outcomes be specified, the period of observation be meaningful, and that the appropriate analytic technique be employed. None of these areas has been comprehensively addressed in any single study thus far. Therefore a variety of work is discussed in this section.
Several cost-effectiveness approaches have been used in comparing various forms of long-term care.H2 studies in this area have focused on comparisons at two levels: the cost-effectiveness of home health care compared to nursing home care; and the costs and benefits of enriched home service programs as opposed to more conventional and limited home care.These are discussed in the following paragraphs.
Cost-Effectiveness of Home
Health Care Versus Nursing Home Care
Most of the available studies of home health care versus
HZpor a general discussion of cost-effectiveness in the longterm care field see, Neville Doherty, Joan Segal, and Barbara Hicks, "Alternatives to Institutionalization for the Aged: Viability and Cost-Effectiveness," Aged Care and Services Review, I (January/ February, 1978), 8-14; Jonn Hammond, "Home Health Care Cost Effectiveness: An Overview of the Literature," Public Health Reports, XLIV
(July/August, 1979), 305-312; Neville Doherty an3 Barbara Hicks, "Cost-Effectiveness Analysis and Alternative Health Care Programs for the Elderly," Health Services Research, XII (1977) 190-203.
l^No known studies have compared among different types of nursing home care.


73
nursing home care, compare the actual program cost of home health care to the hypothetical cost of nursing home care. All assume that the quality of care (effectiveness) is equal among the settings. Few include the total cost of living in the community (e.g., room, board and transportation) in the cost of home health care. Only a few of the more recent studies compare total public costs across the care modalities, including expenditures for Old Age Pension (O.A.S.D.I.), Supplemental Security Income, etc. No study to date has compared the actual cost of care for cohorts of home health and nursing home patients over (a significant amount of) time to the outcome of that care. Thus, weaknesses in existing methodology make it difficult to draw definitive conclusions about the cost-effectiveness of any form of long-term care.
The findings of early studies addressing the first question (of traditional home care versus nursing home care) were conflicting due to major methodological deficiencies. A review of twenty-five projects begun before 1975^ identified numerous problems with previous studies: noncomparability of experimental and control groups; lack of significant differences between experimental and control treatment variables; wide variation in (and lack of control for) the needs of individual members of the study populations; variation in the scope and extent of treatments from one study to another;
l^Sonia Conly, Critical Review of Research on Long-Term Care Alternatives ([Washington, D.C.]: Office of the Assistant Secretary for Planning and Evaluation, DHEW, June, 1977), pp. 4-16. (Mimeographed.)


74
differences in the methods used to collect cost information and the scope and kinds of costs included; and noncomparability of costs considered for the experimental group compared to the control group. The review confirmed problems cited in other evaluations of longterm care^ ancj concluded that evidence at that time was insufficient to either refute or support arguments that either enriched or traditional home health care alternatives were more cost-effective than institutional long-term care.
Effectiveness of Home Health Care In Reducing Institutionalization . Several small studies in recent years have focused on the question of the effectiveness of home health care in preventing institutionalization. However, results of these efforts are inconclusive.
Some studies concluded that home care prevents institutionalization. Using the number of home health care admissions per 1,000 Medicare beneficiaries as his dependent variable, Dunlop found that increases in home health care utilization were associated with decreases in the utilization of nursing homes (on an area-wide
^Marie Callender and Judith LaVor, Home Health Care: Development, Problems and Potential (Washington” D.C.: Office of Soc i a 1 Services a?T3 Human Development, Office of the Assistant Secretary for Planning and Evaluation, DHEW, April, 1975). (Mimeographed.); Applied Management Sciences, Evaluation of Personal Care Organizations and Other In-Home Alternatives to Nursing Home Care for the Elderly and Long-Term Disabled-! final Report ana Executive Summary (Revised) (Silver Spring, MD: Applied Management Sciences, May, 1976).


75
basis).In another study of 245 patients in a New York City program for the homebound aged, Brickner et al.^ reported that
after 24 months, 23 patients improved to the extent that they were
no longer homebound, 116 remained stabilized under the program's continuing care, and 40 patients were institutionalized (either in hospitals or nursing homes). Relying solely on clinical judgment, Brickner et al. estimated that 85 of the patients would have
required institutional care and 25 would have died without the program. A recent Canadian study concluded on a similar basis, that
a significant number of study patients avoided nursing home utilization.^
Other studies have found inconsistent relationships between use of home care and institutionalization. A one-year study by the Benjamin Rose Institute in Cleveland of 50 elderly patients receiving home health aide services after hospitalization indicated that the group receiving services had significantly fewer days in and fewer admissions to long-term care institutions than a control
group. The service group patients also appeared to be significantly
^Burton Dunlop, Determinants of Long-Term Care Facility Utilization by the Elderly: An Empirical Anaysis, Working Paper 963-35 (Washington, O.C.: THe Urban Institute, revised March 1, 1976).
^Philip W. Brickner et al., "The Home Bound Aged: A
Medically Unreached Group," Annals of Internal Medicine, LXXXII (January, 1975), 1-3.
1 1 o
A.S. Kraus and M.I. Armstrong, "Effect of Chronic Home Care on Admission to Institutions Providing Long-Term Care," Canadian Medical Association Journal, CXVII (October, 1977), 747-749.


76
1iq ion
more contented. On the other hand, Bryant et al. reported on a small study of home care provided to stroke victims who had been discharged from the hospital. This ex post study compared a group that had received home care and a mixed group which received either only physical therapy or no care. After a nine-month follow-up, two home care and eight control group patients were living at home. However, close examination of the study indicates that the experimental and control groups were probably not well matched according to severity of illness.
The underlying assumption of these studies is that if home health care can prevent (or postpone) institutionalization, it will, by definition, be cost-effective when compared to nursing home care. Unfortunately none of these studies investigated the validity of this assumption. One study which did try to test the assumption was the recent experimental study of homemaker services by Weissert, Wan, and Livieratos.^ They concluded that the study group did not have significantly lower utilization of hospital or nursing home services than the control group (which received the usual array of services). Not surprisingly, total costs of the study group were actually higher than for the controls.
^Margaret Neilson et al., "A Controlled Study of Home Health Aid Services," American Journal of Public Health, CLXII (1972), 1094-1101.
^Nancy Bryant, Louise Candland, and Regina Loewenstein, "Comparison of Care and Cost Outcomes for Stroke Patients, With and Without Home Care," Stroke, V (1974), 54-59.
12^Weissert, Wan, and Livieratos, pp. 21-26.


77
Cost of Home Health Care Versus Nursing Home Care. Few studies are available which empirically support the general proposition that home care is less costly than nursing home care. Indeed, the most widely cited home health care studies concerning cost savings are of short-term acutely ill patients.^2 most studies of home care which are purported to demonstrate cost savings from home care utilization suffer from severe methodological weaknesses. They either merely estimate savings from home care compared to hypothetical nursing home care, or they fail to measure costs in a comprehensive manner and assume that the effect of long-term care is the unit of output (e.g., cost per visit, patient day, month).
Many studies merely estimate the cost savings from community-based experimental programs. Brickner and Scharerl23 -jn their study of a progran for the homebound in New York City, claimed considerable cost savings, but their claims were based solely on physician estimates of "probable" institutionalization and "probable" program
^Joseph R. Stone, Elizabeth Patterson, and Leon Felson, "Effectiveness of Home Care for General Hospital Patients," Journal of the American Medical Association, CCV (July 15, 1968), 145-148; Lowell Gerson and Owen Hughes, ‘‘A Comparative Study of the Economics of Home Care," International Journal of Health Services, VI (1976), 543-555; Creese and Fielden, pp. 116-121.
^3ph-n-jp Brickner and Linda Keen Scharer, "Hospital Provides Home Care for Elderly at One-Half Nursing Home Cost," Forum (Novem-ber/December, 1977), no page numbers.


78
costs. There are at least four other major studies which estimate cost savings in the same manner.124
In a more sophisticated cost study, Greenberg disaggregated a Minnesota target population into four disability levels and two living arrangements (living alone and living with others) on the assumption that home care costs would vary on these dimensions. Costs were taken as total costs, including room and board. For only the worst disability level was home care as expensive or more expensive than nursing home care. All other levels of disability and living arrangements were cheaper with home care. From this, Greenberg estimated that 9% of the 1974 Minnesota skilled nursing facility patient population could be cared for at home, resulting in substantial cost savings.^5
In studies that have looked at the population already in nursing homes, cost savings from hypothetical home health care have not been found. One study in Durham, North Carolina by Burton et al-126 estimated that for approximately 87% of the patients in nursing homes, the only suitable alternatives were not economically
^4Avery Colt et al., "Home Health Care is Good Economics," Nursing Outlook (October, 1977), pp. 632-636; Anthony Amado, Beatrice Cox, and Rich Mileo, "Cost of Terminal Care: Home Hospice vs. Hospital," Nursing Outlook (August, 1979), pp. 522-526; Kraus and Armstrong, pp. 747-749; Gerald M. Edgert and Joyce E. Bowlyow, "Preliminary Findings: Monroe County's Access Project to Prevent
Unneeded Nursing Home Admissions," Perspectives on Medicaid and Medicare Management (September, 1979), pp. 5-13.
^Greenberg, "jhe Costs of In-Home Services," p. 45.
*26Burton et al., pp. 3-12.


79
feasible, costing approximately four times more than nursing home care. For the other 13%, alternatives outside the nursing home were feasible but with no great reduction in costs. The Levinson study cited earlier compared the actual costs of nursing home care to the costs of hypothetical home health care plans for a group of 50 patients awaiting hospital discharge. This study found that institutional care was less costly than home care for 80% of the
patients, with an estimated cost of room and board of $60 per week included in the cost of home care.-^7
Only recently have actual costs of home care and institutional care been compared. One General Accounting Office (GAO) study found that home care services were less costly only if the patients were not severely disabled, and were more costly in cases of disability where extensive services were needed. In that study, the GAO
compared the cost of providing home health care to the elderly,
including the value of services provided by family and friends, to the cost of nursing home care. When comparing the functional and health status of elderly persons residing in the community with those institutionalized, the study found that 87% of the institutionalized older persons were greatly or extremely impaired, compared to 13% of those at home. Overall, half of the services
received by the elderly were provided by family and friends; over 70% of the services received by the greatly or extremely impaired came from that same source. At all levels of impairment, the value
127$ager, p. 247A.


80
of services (computed at the same rate as purchased services) provided by the informal social support systems greatly exceeded the cost of services provided at public expense. (Needless to say, the pricing of services rendered by family or friends is subject to a wide range of uncertainty.) At the greatly impaired level, where the break-even point between the cost of home care and institutional care was reached, family and 'friends provided about $287 per month in services for every $120 spent by the public agencies.^8 Unfortunately, no comparison of the quality of care (effectiveness) was made across the care modalities.
Another study that compared the actual cost of home care to institutional care was completed by the Stanford Research Insti' tute. Although similar to the GAO study in that it did not investigate the effectiveness of care, it suffered from more severe methodological weaknesses because of its limited conceptualization of cost. Like many other studies, it compared the average direct program cost per month of home health to nursing home care, omitting the costs of maintaining the patient in his home as part of the cost of in-home services. Thus, it was not surprising that the study
^comptroller General of the U.S., General Accounting Office, Home Health: The Need for a National Policy to Better Provide for the Elderly. Publication No. HRD-78-19. (Washington, D.C.: Government Printing Office, 1977), pp. 9-22. (Comptroller General of the U.S. is hereafter referred to as GAO.)


81
concluded that home health care was cost-effective compared to nursing home care.129
One important study of the relative cost-effectiveness of nursing home versus home care is currently being conducted in Minnesota. In that study the total costs of services and living needs for a group of 367 home health patients and 350 nursing home patients will be compared in terms of outcomes (measured by changes in functional status, satisfaction, and social contacts). In addition, the study will compare total costs using a cost-finding method which tries to develop consistent allocation procedures at the level of the cost center for both institutional and non-institu-tional long-term care. Although a significant improvement over previous efforts, this study treats each modality as if it were a generic product (e.g., no consideration is given to organizational and management issues that may affect the cost and outcome of care).120 Thus, no truly comprehensive long-run studies of longterm care cost-effectiveness have been conducted to date.
Cost-Effectiveness of Coordinated and Expanded Home Health Care
A recent key federal policy initiative to address the cost-effectiveness of an expanded home health service system has been the
l^Neill Pi land, Feasibility and Cost-Effectiveness of Alternative Long-Term Care Settings: Executive Summary, (Menlo
Parx, CXi Stanford Research Institute Internationa!, Ray, 1978), pp..3-6.
130Anderson, A Comparison of In-Home Care, pp. 2ff.
i


Full Text

PAGE 1

A COST-EFFECTIVENESS ANALYSIS OF HOME HEALTH CARE: IMPLICATIONS FOR PUBLIC POLICY AND FUTURE RESEARCH by Bettina Tabak Kurowski B.S., University of Southern California, 1967 M.P.A., University of Colorado, 1975 A thesis submitted to the Faculty of the Graduate School of Public Affairs of the University of Colorado in partial fulfillment of the requirements for the degree of Doctor of P ublic Administration 1980

PAGE 2

This Thesis for the Doctor of Public Administration Degree by Bettina Tabak Kurowski has been approved for the Graduate School of Public Affairs by

PAGE 3

i i i Kurowski, Bettina Tabak (D.P.A., Public Administration) A Cost-Effectiveness Analysis of Home Health Care: Implications for Public Policy and Future Research Thesis directed by Professor Michael S. March The subject of this study is the cost-effectiveness of one long-term care policy option: home health care. Funded by the Health Services Administration (DHEW) in 1977, the study highlights long-term care policy issues that need to be addressed through evaluation research. In so doing, it examined the cost per episode of illness of certain home health care services, using 1976 data from a sample of four hospital -based and four free-standing home health care providers in the eastern U.S. The empirical portion of the study determined the effect of patient, provider/ community, and outcome characteristics on the cost per episode of care. The findings indicated that most of the variation in cost was not explained by factors included i n the study. Of the cost variation that was explained, patient characteristics accounted for the greatest portion, followed by outcome and provider characteristics. Important policy implications of the findings are presented; the major research implication of the study is that a comprehensive national research program that is intended to address long-term care issues is essential. Key attributes of the suggested program include comprehensive measurement of costs, effects, patient, and provider adequate project duration, and appropriate dissemination of findings.

PAGE 4

iv This abstract is approved as to form and content. I recommend its pub 1 i cation.

PAGE 5

PREFACE The subject of this study is long-term care policy, specifically in the area of home health care. The vehicle chosen to explore this issue is a study of the cost per episode of illness. In Octo ber 1977, the Office of Planning, Evaluation, and Legislation of the Health Services Admi ni strati on in the Department of Health, Education, and Welfare awarded a contract to the Center for Health Ser vices Research of the University of Colorado Health Sciences Center to conduct research into aspects to home health care relevant to the agency• s responsibility for promoting the deve 1 opment and expansion of home health service capacity. One part of that contract was the study described herein. From the date of the initial award through the end of the eighteen month contract, I served as Project Director for the cost per episode portion of the study. Thus, I was res pons i b 1 e for the development of its design and execution of its analytic methodology, in addition to the preparation of the final report. It is as a result of my role in this contract that I authored this study. I have been fortunate to have had assistance in the development and improvement of this manuscript. Those who have read and commented on one or another version of this study have contributed greatly to its refinement. I an especially grateful to my colleagues at the Center for Health Services Research, Eileen Tynan, Barbara Harley, Nancy Shanks, and David Landes, for their suggest ions and criticisms; Peter Shaughnessy, the Director of the

PAGE 6

Center, for concurrently playing the roles of teacher, employer, and friend; Robert Schlenker, the Principal Investigator of the home health evaluation contract, for his guidance and support. Gerri Tricarico and Linda Breed provided inva l uable assistance in the data analysis phase of the study. Susan Elgin, Charlotte Gordy, and Phyllis Hunter typed countless drafts ofthis manuscript and deserve much appreciation for their efforts. Michael March, my thesis advisor, contributed significantl y to this study and to my continuing intellectual growth. His commitment to improving the public policy process through inquiry, ana lysis, and action should serve as an inspiration to those who wish to improve the ctuality of gover n ment. Philip Burgess, Director of my doctoral program, enlarged my view of the world and helped shape my professional development. A special thank you goes to my colleague, Ann Jones, who pro vided needed encouragement and thoughtful suggest i ons at crucial times throughout the development of this manuscript; and to Carol Betson and Laura Dodson, who each helped in their own way. Immense gratitude, which is difficult to put into words, goes to my husband, J im, and children, Lisa and Thad, 1vho endured my long hours, supported me in my frustration, and continually offered me encouragement. Finally, my mother has been a source of energy and, i n a special way, made a valuable contribution to this effort. Denver, Colorado May 1980 Bettina Tabak Kurowski

PAGE 7

TABLE OF CONTENTS CHAPTER I . INTRODUCTION Overview . The long-term care system The long-term care population Rising expenditures for long-term care Institutional care Home health care Total c osts Quality and appropr i ateness of care Home health care Background .. Comparative costs of home heal th care M ajor barriers to home care Summary o f current policy issues E val uation research to reduce uncerta i nty A specific research area . . . . ..... II. STATE OF THE ART I ntroduction Case mix Age Living arrangement Diagnosis Functional status PAGE 1 3 3 7 9 10 11 12 17 17 21 2 4 27 2 8 31 34 34 35 36 37 38

PAGE 8

CHAPTER Mental health status ..... Multidimensional health status Prob 1 ems . . . Levels of care Quality Structural measures of quality . Process measures of quality Outcome measures of quality Specifying appropriate outcomes i n general Specifying appropriate outcomes for home health care . . . . . . . . viii PAGE 40 41 42 43 46 46 48 50 52 53 Tenuous relationship between process and outcome 57 Cost of Care . 60 Nursing home cost studies 61 Facility level cost studies 61 Patient level cost stu dies 68 Home health cost studies . . 69 Cost-effectiveness of long-term care 71 Cost-effectiveness of home health care versus nursing home care . . . . . . . . . . . . 72 Effectiveness of home health care in reducing institutionalization . . . . . . . . . . 74 Cost of home health care versus nursing home care 77 Cost effectiveness of coordinated and expanded home health care . . . . 81 Effectiveness findings 82 Cost findings 84 Cost-effectiveness findings 84

PAGE 9

CHAPTER Methodological weaknesses Conclusion ... III. THE STUDY APPROACH AND METHODOLOGY Introduction . Study purpose Conceptua 1 fr arne work Independent variables Dependent variable ... Overall relationships Sample selection and data avai lability Data system selection Provider selection Episode selection Study variables and data sources Reliability of data Specification of the model The primary dependent variab l e: cost per . episode of home health care Definition of episode Unit of service Cost per unit of service Cost per episode The independent variables: patient-specific character i s t i c s Age Living arrangement Diagnosis and functional status ix PAGE 85 91 92 92 93 94 96 96 98 98 100 102 106 106 107 107 109 llO 111 113 114 114 115 116

PAGE 10

CHAPTER Surgical procedure Goal at admission The independent variables: outcome-related characteri sties Change in functional status Health status at discharge . Home health care length of use Intensity The independent variables: provider/health system X PAGE 118 118 119 119 119 120 120 characteristics . . . . . 121 Primary source of payment Provider size Population density Acute care admissions Miscellaneous factors Income Race . Other health care providers Regulatory environment Analytic techniques Limitations of the study Data Limitations Cost Data Case Mix Data Quality and Effectiveness Data Design Issues Conclusion ... 121 123 123 124 125 125 125 125 126 133 135 135 135 137 138 139 140

PAGE 11

xi CHAPTER PAGE IV. A COMPARISON OF CASE CHARACTERISTICS Introduction .. Patient-specific characteristics Age .... Living arrangement Primary diagnosis Functional status Functional status index Surgical procedure Patient's location prior to admission Goal at admission Outcome-related characteristics Change in functional status index Health status at discharge . Intensity of care and length of use Provider/health system characteristics Primary source of payment Conclusion . . . . ..... V. UTILIZATION AND COST FINDINGS Introduction ....... . 141 141 142 144 144 144 148 150 150 150 152 152 154 154 157 157 159 160 Overall utilization and cost per episode findings 161 Utilization findings 161 Cost findings 163 Cost findings discussion 165 Charges passed through as costs 165 Unaudited cost reports 166

PAGE 12

Xi i CHAPTER PAGE Admi ni strati ve surcharges 166 Case mix differences . . 168 Service mix differences 169 Variations in outcome 171 Analyses to explain overall variation within each sample 171 Approach . . . . 171 Overall findings 174 Rehabilitation patients 175 Terminally ill patients 175 Analyses to test the relative importance of each independent variab l e category 176 Analyses to test the patterns within each i ndependent variable category 177 Patient-specific characteristics 178 Overview of multivar i ate findings 179 Age Living arrangement Primary diagnosis Functional status Functional status index Goal at admission Outcome-related characteristics Overview of multivariate findings Change in functional status index Health status at discharge . Utilization-related outcome 181 183 183 186 188 190 192 192 193 195 197

PAGE 13

CHAPTER Provider/health system characteristics Overview of multivariate findings Primary source of payment Provider size Acute care admissions Comparison of Massachusetts and Philade l phia Overview . . . Controlling for age . Contra 11 i ng for 1 i vi ng arrangement Controlling for primary diagnosis Controll ing for functional status index Summary of m ajor findings outcomes xiii PAGE 197 198 199 201 203 203 205 205 207 210 210 212 Overall utili zation and cost per episode findings 214 Overall ability to explain observed var i ation within each sample . . . . . . . . . . . 214 Relative importance of each independent variable category . . . . . . . . 215 Patient-specific characteristics 216 Outcome-re 1 a ted characteristics 217 Provider/health system characteristics 218 Conclusion .... 219 VI. POLICY IMPLICATIONS Cost implications Inability to explain cost variations Utilization review . Allocation procedures Cost-effectiveness . . 220 220 221 221 222

PAGE 14

CHAPTER Case mix adjustment Progran implications Continuum of care Hospice care Home health data Housing xiv Tax incentives Information and referral VII. AGENDA FOR RESEARCH PAGE 222 223 223 223 224 225 226 226 Introduction . . . . . . Overview of long-term care research problems Lack of comprehensive and systematic approach Inadequate and incomplete goals and objectives Underdeveloped criteria ..... . Inadequate design and data collection Lack of comprehensive cost measurement Neglect of organizational and management variables Inadequate patient descriptors ..... 228 228 229 230 231 232 232 233 234 Lack of attention to the longitudinal dimension 235 Non-comparable designs 235 Inadequate utilization and dissemination of findings 236 Systematic analysis for long-term care policy issues 238 Definition of systems analysis 240 Overview of the basic elements 240 Defining the problem . . 241 Identification of goals, objectives and criteria 242

PAGE 15

CHAPTER Designing the model Development of alternatives Evaluating alternatives Predicting consequences Interpretation Communication Attributes of national research agenda for long-term care .... Key policy questions Cost .. Effectiveness Cost-effectiveness Conceptual framework and measurement Comprehensive Patient Descriptors Effectiveness measures Cost measures Duration of the research agenda Control of the research Timeliness of the research Dissemination of the research findings Centralized data system Appropriate actions at the federal level Conclusion BIBLIOGRAPHY XV PAGE 243 244 244 245 245 245 246 248 249 251 252 253 253 254 255 259 260 261 262 262 263 266 267

PAGE 16

CHAPTER APPENDICES A. Description of home health care services B. Data collection forms C. Data specifications D. Utilization analyses E. Multiple regression analyses xvi PAGE 282 300 307 321 339

PAGE 17

TABLE I. LIST OF TABLES At Risk Populations In Long-Term Care Institutions and the Community, 1975-76 II. Selected Characteristics of Providers in the Study Sample . . . III. Final Sample of All Episodes IV. Data Specifications v. VI. VII. A Comparison of the Age of Massachusetts and Philadelphia Patients . . . A Comparison of the Living Arrangement of Massachusetts and Philadel phia Patients A Comparison of the Primary Diagnosis of Massachusetts and Philadelphia Patients VIII. A Comparison of the Independence in IX. Activities of Daily Living of Massachusetts and Philadelphia Patients . . A Comparison of the Activities of Daily Living Index of Massachusetts and Philadelphia Patients . . . X. A Comparison of the Preadmission Location of Massachusetts and Phil adelphia Patients XI. A Comparison of the Change in Activities of Daily Living Index of Massachusetts and XI I. XIII. XIV. Philadelphia Patients . . A Comparison of the Health Status at Discharge of Massachusetts and Philadelphia Patients . A Comparison of the Mean Intensity and the Length of Use of Massachusetts and Philadel-phia Patients . . . A Comparison of the Primary Source of Payment of Massachusetts and Philadelphia Patients PAGE 4 103 104 127 143 145 146 147 149 151 153 155 156 158

PAGE 18

TABLE XV. XVI. XVI I. XVI I I. XIX. A Comparison of Total Visits Per Episode of Massachusetts and Philadelphia Patients A Comparison of the Cost Per Episode for Massachusetts and Philadelphia Patients A Comparison of the Average Cost Per Episode by the Age of Massachusetts and Philadelphia Patients ................. . A Comparison of the Average Cost Per Episode by the Living Arrangement of Massachusetts and Philadelphia Patients ........... . A Comparison of the Average Cost Per Episode by the Primary Diagnosis of Massachusetts and Philadelphia patients ........... . XX. A Comparison of the Average Cost Per Episode by the Activities of Daily Living of Massachu-xviii PAGE 162 164 182 184 185 setts and Philadelphia Patients ......... 187 XXI. A Comparison of the Average Cost Per Episode by the Activities of Daily Living Index of Massachusetts and Philadelphia Patients XXII. A Comparison of the Average Cost Per Episode 189 by the Goal at Admission of Massachusetts Patients 191 XXI I I. A Comparison of the Average Cost per Episode by the Change in Activities of Daily Living Index of Massachusetts and Phil adelphia XXIV. A Comparison of the Average Cost Per Episode by the Health Status at Discharge of Massachusetts 194 and Philadelphia Patients . . . . . . . . . . . 196 XXV. A Comparison of the Average Cost per Episode by the Primary Source of Payment for Massachusetts and Philade l phia Patients . . . . . . . . . . . 200 XXVI. A Comparison of the Average Cost per Episode by the Provider Size for Massachusetts and Philadelphia Patients . . . . . . . . . . . . . . . . . . . . . 202 XXVII. A Comparison of the Average Cost per Episode by the Acute Care Admissions in the Service Area for Massachusetts and Philadelphia Patients . . . . . 204

PAGE 19

TABLE XXVIII. A Comparison of the Change in Activities of Daily Living Index By the Age of Massachusetts xix PAGE and Philadelphia Patients . . . . . . . . . . 206 XXIX. A Comparison of the Change in Activities of Daily Living Index By the Living Arrangement of Massa-chusetts and Philadelphia Patients . . . . . . . 208 XXX. A Comparison of the Activities of Daily Living Index By the Living Arrangement of Massachusetts and Philadelphia Patients . . . . . . . . . . . 209 XXXI. A Comparison of the Change in Activities of Daily Living Index By Selected Primary Diagnoses of Massachusetts and Philadelphia Patients . . . 211 XXXII. A Comparison of the Change in Activities of Daily Living Index By the ADL Index Scores At Time of Admission of Massachusetts and Philadelphia Patients 213

PAGE 20

CHAPTER I I NTROOUCTI ON Overview A disturbing paradox exists in the long-term care delivery system. On the one hand, the system continues to absorb increasing amounts of the public and private dollar. On the other, it is riddled with controversy regarding inappropriate care, Medicare/ Medicaid fraud and abuse, and nursing home scandal.l Expenditures for services grow, but the system seems to operate contrary to the promotion of the quality of life and provides l ittle incentive for improvement.2 The system meets the mandate of policy makers and the expectation of providers so poorly that both groups advocate major alterations in the provision of and payment for care. An overhaul of the system demands the development and implemen-tation of programs designed to assess needs, provide services at reasonable cost, and monitor quality to ensure that needs are 1u.s., Congress, Senate, Special Committee on Aging, Nursing Home Care in the United States: Failure in Public Polic , Nos. 1-4, 94t ong., st ess., Apr1 , 975 Wash1ngton, D.C.: Government Printing Office, 1975). 2sylvia Sherwood, "Long-Term Care: Issues, Perspectives and Directions," Lon -Term Care: A Handbook for Researchers, Planners, and Providers, ed. Sy via erwood New York: pectrum Pub 1cations, 1975), p. 26.

PAGE 21

2 act u a 11 y met. The simultaneous optimization of long-term care programs, especially in the areas of cost and quality, is no easy task. Yet, the difficulty of the undertaking does not diminish the necessity for getting the various policy options (with their respective costs) on the table so that more informed decisions can be made.3 It is to this end that this study was undertaken. In short, this study examines the cost-effectiveness of one of the policy options in long-term care: home health care. However, the study provides more of a vehicle for looking at the utility of evaluation research than making definitive statements about home health care. Further, it h ighlights the need for a comprehensive research strategy for future policy development in long-term care. Because the long-term care system provides the policy framework for this study, it is important to review its development. The balance of this chapter provides such a rev iew, paying particular attention to major issues and problems. Home health care is the focus of the review, as is the potential importance of evaluation research. The second chapter examines recent research and ev al ua-tion which bear on this study. Chapters III-VI present the empiri-cal research on home health care and include study methodology, findings, and d1scussion. Finally, Chapter VII suggests an overall research strategy for long-term care, which is based on systems analysis. 3u.s., Congress, Senate, Special Committee on Aging, Home Care Services for Older Americans: Planning for the Future, 96th Cong., 1st Sess., May 7, 1979 (Wash1ngton, D.C.: Government Printing Office, 1979), p. 2.

PAGE 22

3 The Long-Term Care System Long-term care refers to services for the chronically ill and the physically and mentally handicapped whose conditions are not amenab 1 e to brief periods of treatment. It inclu des health and social services provided in an institutional and/or home-based setting. Examples of institutional long-term care include nursing homes (skilled nursing, intermediate care, and personal care facilities), residential or domicillary care, adult foster homes, congre-gate housing, and boarding homes. Home (or community) based long term care services generally include home health/home health aide s .ervices, homemaker/chore services, meal programs, social services, and sometimes transportation services and direct medical care. Because of the wide scope of services, long-term care is often treated as a composite of health and social services, with an empha-sis on those service components generally identified as medical or health care services. The Long-Term Care Population The population at risk for long-term care services is varied, including the aged, physically and developmentally disabled, mental l y retarded, and psychiatricly impaired, etc.4 ( Table I presents a summary of recent estimates of the population in need of 1 ong-term care.) Yet, it is the functionally dependent elderly, those 4 For a more complete discussion of the population at risk, see Judith LaVor, "Long-Term Care: A Challenge to Service Systems" ( Washington, D.C.: Office of the Assistant Secretary for Planning and Evaluation, DHEW , 1976), pp. 6-15. (Mimeographed. )

PAGE 23

Tab 1 e I At Risk Populations in Long-Term Care and the Community 1975-76 Institutions Nursing Homes (includes SNF and ICF) Chronic Disease Hospitals VA Hospitals (Psychotic-long stay)a Psychiatric Hospitals Institutions for the Mentally Retarded Non-institutional Population Mentally Retarded (substantially handicapped) Aged Needing Care at Home Severely Mentally Digordered Developmentally Disabled Cerebral Palsy Epileptic Other Neurological Disorders 1,200,000 25,000 25,000 250,000 200,000 670,000 3,400,000 1,000,000 580,000 206,000 600,000 4 aThis represents a minimum rather than a total for all VA patients. bDoes not include other categories of severely physically impaired. Sources: U.S., Department of Health, Education, and Welfare, Health, Un1ted States: 1975 (Washington, D.C.: Government Printing Off1ce, 1976), p. 255; Ethel Shanas, .. Measuring the Home Health Needs of the Aged in Five Countries,.. Journal of Gerontology, XXVI (1971), 38; U.S., Department of Health, Education, and Welfare, Public Health Service, 11Changes in the Age, Sex and Diagnostic Composi tion of the Resident Population of State and County Mental Hospitals. U.S. 1964-1973,11 Statistical Note 112, Publication No. (ADM) 75-158 ([Washington, D.C.: Government Pr1nting Office, 1975]), pp. 6-7; Murray Ducket, Background and Long-term Care for the Retarded (Wash1ngton, D.C.: re Urban Institute, 1975), p. 4; u .. , Department of Health, Education, and Welfare, National Institute of Mental Health, 11The Severely Mentally Disordered, .. by Valerie Brq.dley for use of the White House Conference for Handi capped Individuals (1976), p. 15. All sources cited by Judith LaVor, 11Long-Term Care: A Challenge to Service Systems, .. Reform and Regulation in Long-Term Care, ed. Valerie LaPorte and Jeffrey Rubin (New York: Praeger Pub., 1979), p. 18.

PAGE 24

5 individuals over 65 with illnesses or impairments that have become disabling, that are at greatest risk. In 1977 the 65 and over age group was estimated to comprise 11% of the total population, or 23.5 mill ion persons. 5 This percentage is expected to rise to 12%, or 31.8 million persons, by the year 2000.6 The number of persons age 75 and over, often termed the 11frai1 elderly, .. is increasing even more rapidly, from 41% of the total aged population in 1977 to an estimated 55% in 2000.? Because chronic medical conditions, func-tional dependency, and psychosocial impairments are most prevalent among 11frail11 elderly persons,8 the increasing age of the population will have a significant impact on the future need for long-term care. Currently, about 2% of the 65-74 age group are residents of nursing homes; this figure increases to 7% for the 75-84 age group and to almost 20% for those over age 85.9 Overall, 87% of all nursing home residents in 1977 were over the age of 65, and 70% were 5u.s., Department of Commerce, Bureau of the cal Abstract of the U.S., 1978 (Washington, D.C.: ing Office, 1978), pp. 8-9. (U.S., Department of after referred to as Bureau of the Census.) 6Ibid. Census, StatistiGovernment Pr1ntCommerce is here-7Institute of Medicine, The Elderly and Functional Dependency (Washington, D.C.: Nat1onal Academy ot Sc1ences, 197/J, pp. 1-3. 8Institute of Medicine, pp. 1-4. 9Institute of Medicine, p. 3.

PAGE 25

6 over the age of 75.10 Given current utilization patterns, the number of people who will at some time reside in nursing homes will increase substantially over the next several decades. When one considers the nLJTlber of elderly who reside in their own homes, but who are in need of community-based services, the demands placed on the long-term care system become even greater. Of the elderly living in the community in 1974, over three million reported limited mobility, one-third of whom reported the need for assistance by another person or a device for mobility. An additional one-third experienced some difficulty in routine daily functions. Of those individuals with restricted mobility, over one million were reported confined to their homes.ll Of the non-institutionalized adult population in 1975, over three million elderly persons were estimated to need some kind of home care on a continuing basis.12 It is important to note here that estimates of future nursing home utilization, based on current patterns and the availability of alternatives, are different from estimates based on functional dependency. The former assumes that utilization patterns will 10u.s., Department of Health, Education, and Welfare, Public Health Service, The National Nursin Home Surve : 1977 SLJTlmary for the U.S., Publicat1on No. PH 9-94 Was 1ngton, D .. : Government Pr1nting Office, 1979), p. 29. 11Marie Callender and Judith LaVor, 11Home Health Care: Devel opment, Problems and Potential" (Washington, D.C.: Office of the Assistant Secretary for Planning and Evaluation, April, 1975), p. 3ff. (Mimeographed.) 12saad Nagi, 11An Epidemiology of Disability Jlmong Adults in the United States11 (Columbus: Ohio State University, Mershon Center, 1976), p. 4-10. (Mimeographed.)

PAGE 26

7 continue, while the latter is not necessarily tied to these patterns. The manner in which a person ultimately receives long-term care is a function of numerous factors including health status, support, living arrangements, the availability of care alternatives, and funding patterns. Based on estimates of the need for all types of long-term care given current utilization patterns, it is expected that between 7.4 and 12.5 million adults will require long-term care services by 1985, with more than three million utilizing institutional care.13 As the number of functionally dependent persons increases over the coming decades, expenditures on long-term care can be expected to increase sharply. Concerns about the cost and quality of care become more important in light of the growing demands that will be placed on the health care system. Rising Expenditures for Long-Term Care Massive increases in total national expenditures for personal health care have occurred in the fifteen years since the inception of the Medicare and Medicaid programs. Increasing at an 11% annual rate on a per capita basis,14 they reached $142.6 billion in 1977, 13u.?., Congress, Congressional Budget Office, Long-Term Care for the Elderly and Disabled (Washington, D.C.: Government Printing Office, 1977), p. 8. (U.S., Congress, Congressional Budget Office is hereafter referred to as CBO). 14Ann Somers, "The High Cost of Health Care for the Elderly: Diagnosis, and Some Suggestions for Therapy," Journal of Health Politics, Policy and Law, III (Summer, 1978), 163-180.

PAGE 27

8 an amount equal to 9% of the gross national product.l5 The annual rate of increase of those expenditures during the past 10 years was 11% per capita, but most recently reached 13%.16 Of all health care expenditures, 40% ($57 billion) were covered by public funds, primarily through the Medicare and Medicaid programs ($20.8 and $16.3 billion, respectively).!? The annual rate of increase of total Med-icare and Medicaid expenditures between 1975 and 1977 was 17% (i.e., tot a 1 costs rose by one-third in only t\ltO years), 18 compared to a 6.1% increase in the Consumer Price Index over the same time period.19 A disproportionate share of health care expenditures is spent on care of the aged and disabled. Although they comprised only 11% of the tot a 1 popu 1 at ion in 1978, they accounted for 29% of a 11 health care expenditures in that year.20 Nearly half of all public expenditures for personal health care were for the elderly. In addition, 67% of health care costs for the elderly were funded 15Robert Gibson and Charles Fisher, 11Age Differences in Health Care Spending, Fiscal Year 1977,11 Social Security Bulletin, XLII (January, 1979), 3-14. 16s omers, p. 163. 17Gibson and Fisher, p. 12. 18Gibson and Fisher, p. 10. 1 9sureau of the Census, p. 483. 20u.s., Department of Health, Education, and Welfare, Public Health Service, Health, United States: 1978, Publication No. (PHS) 78-1232 (Washington, D.C.: Government Printing Office, 1978) , pp. 396-400.

PAGE 28

9 through public programs, compared to 26-30% for other age groups.21 Long-term care expenditures have increased at a greater rate than expenditures for all medical care. This is due in large part to economy-wide inflation and changing demographic patterns, in addition to increased access to care through the government financ-ing described previously. Institutional Care. In 1970, Medicare expenditures for nursing home care were $236 million.2 2 By 1977 these payments had reached $349 mi 11 i on23 and they are projected to rise $393 mi 11 ion in 1981. 2 4 Because Medicare benefits are restricted to skilled nursing care, Medicare pays less than Medicaid for nursing home care. The Medicaid program, the dominant payer of nursing home care, expended $1.5 billion25 and $6.4 . billion26 in 1970 and 1977, and is expected to spend $9.4 billion27 in 1981 (including the state share of program costs). While the estimated increase in Medicare expenditures for nursing home care from 1977 to 1981 represents an increase of 21Gibson and Fisher, pp. 6-12. 22Bureau of the Census, p. 347. 23Gibson and Fisher, p. 12. 25Bureau of the Census, p. 349. 26u.s., Department of Health, Education, and Welfare (DHEW}, Health Care Financing Administration, 11HCFA Progran Statistics, .. Health Care Financing Review, I (Fall, 1979), 74. 27Executive Office of the President, p. 470.

PAGE 29

10 only 12.6%, Medicaid expenditures are expected to increase by 42.4% over the same time period. Home Health Care. Along with nursing home expenditures, Medicare and Medicaid expenditures for home health care have continued to increase over the years, although at a different rate. Medicare expenditures for home health care in 1970 were $70 million,28 but had risen to $358 million by 1977.29 These are expected to rise to $890 million by 1981. 3 0 Medicaid, which funds a much smaller proportion of home health care, incurred almost no expenses for home health care in 1970,31 spent $179 million in 1977,32 and will reim burse almost $339 million by 1981.33 In addition to Medicare and Medicaid, most states pay for some type of non-medical in-home services under Title XX of the Social Security Act. In fiscal year 1976 these payments amounted to about $340 million.34 While total expenditures for nursing home care are much larger than for home health care, the projected growth for Medicare and Medicaid home health expenditures in the 1977-1981 time period is 28sureau of the Census, p. 347. 29oHEW, Health Care Financing Administration, pp. 72-73. 30Executive Office of the President, p. 474. 31sureau of the Census, p. 349. 32oHEW, Health Care Financing Administration, p. 74. 33Executive Office of the President, p. 474. 34u.s. Congress, Home Care Services, p. 126.

PAGE 30

11 94% and 79%, respectively. This far exceeds the projected growth of nursing home expenditures in the same time period (12.6% and 42.4%, respectively). Similar to institutional long-term care, expenditures for home health care will continue to increase due, in part, to the increasing size of the disabled and elderly population. In addition, the rate of growth in these expenditures is expected to increase with the enactment of pending legislation liberalizing home health benefits. These changes a 1 one wi 11 result in an estimated increase of $12 million per year in total Medicare home health expenditures by the year 1982.35 Total Costs. Current long-term care expenditures have reached an all time high and projections of future spending are expected to go even higher. A recent Congressional Budget Office report estimates that total costs (both public and private) of long-term care will run as high as $87 billion in 1985, with federal portions (assuming existing utilization patterns) over $15 bil1ion.36 The increases in expenditures for long term care have created a growing burden on the public sector. Pressures to balance the federal budget will likely prompt progran cutbacks. Given the major role of government in the support of long-term care services, these cutbacks may have negative consequences in an area which is already 35Executive Office of the President, Office of Management and Budget, The Bud et of the U.S. Government, Fiscal Year 1981 (Washington, D.C.: Government Printing Office, 1980 , p. 248. 36cBo, p. xv.

PAGE 31

12 criticized for suboptimal care. It will, therefore, be increasingly important to adopt program alternatives which provide for care of acceptable quality at reasonable cost. Quality and Appropriateness of Care The flow of government funds into long-term care over the last 15 years has not come without concerns over program design or costs. From their inception in 1965, Medicare and Medicaid were accompanied by certification regulations which set forth standards for quality aimed at limiting particpation to those facilities which provided at least minimLm levels of quality medical care.37 The regulation of quality focused on external or structural regulations intended to ensure compliance with minimum safety and qua 1 tty standards. Generally, these standards related to physical facilities, governing and management policies, personnel requirements and qualifications, and the provision of specific services. Aimed initially at hospitals, Medicare certification was pri-marily based on the Joint Commission on Accreditation of Hospitals accreditation standards. In January 1967, Medicare certification was expanded to include post-hospital or extended care facilities. Of the more than 13,000 nursing homes which potentially qualified to provide extended care, only 740 were able to comply fully with the 37John Cashman and Beverlee Myers, 11Medicare Standards of Service in a New Program-Licensure, Certification, Accreditation, .. American Journal of Public Health, LVII (July, 1967), 1107-1117.

PAGE 32

13 conditions of participation by July 1967.38 That number increased to 1,350 by July 1969. Another 3,400 were in substantial compliance with the certification requirements and were conditionally certified in order to increase the av ai 1 ability of services. However, there-after the number of Medicare extended care facilities decreased, and available beds in participating facilities declined.39 At the sane time, the use of Medicare extended care services also declined. The drop in utilization of services reflected changes in the Medicare conditions of participation for extended care facilities in 1969.40 The federal government was initially willing to pay for more extended care services (subject to lower standards) than planned. Increased costs and the provision of substandard care led to strengthened regulations in 1969 which, as intended, resulted in decreased utilization of services. The 1972 standardization of the regulations for the Medicare extended care facility and the Medicaid skilled nursing home (in which the Medicaid standards were ostensi bly upgraded, all facilities becoming skilled nursing facilities) illustrated continued concern for quality. 38u.s., Congress, Senate, Corrmittee on Finance, Medicare and Medicaid: Problems, Issues, and Alternatives, 91st tong., 1st Sess., February, 1970 (Washington, D.C.: Government Printing Office, 1970), p. 94. 39u. S. , Department of He a 1 th, Education, and We 1 fare, Socia 1 Security Administration, Health Insurance Statistics, Publication No. HI-75 (Washington, D.C.: Government Printing Off1ce, 1977), p. 6. 40Ibid.

PAGE 33

14 T'ftQ years after the regulations were standardized, most cert i-fied facilities still did not fully comply with the regulatory guidelines. Questions about the adequacy of institutional care continue, and, if anything, are exacerbated by the levels of care legislated under Medicare and Medicaid. The intended step-down system from skilled to intermediate to personal (or maintainance care) has not materialized under the current reimbursement scheme, and home health care has been, until recently, virtually ignored.41 In spite of early efforts to assure appropriate placement, a number of studies suggest that many patients in nursing homes are inappropriately placed (i.e., that they do not require the level of care provided) and, in some cases, could maintain a life in the community if the appropriate medical and social supports were pro vided. Estimates of the rate of inappropriate institutionalization are varied, ranging from a low of 6% in Minnesota42 to 76% in New 41u.s., Congress, Senate, Committee on Govenment Operations, Subcommittee on Federal Spending Practices, Efficiency and Open Government, Prob 1 ems Associ a ted with Home Health A enc i es and the Medicare Program 1n the State o onda as 1ngton, D .. : overnment Printing Office, August, l976); U.S., Congress, House, Select Committee on Aging, New York Home Care Abuse, Publication No. 95-145 (Washington, D.C.: Government Printing Office, 1978). 42Robert M. Burton et al., 11Nursing Heine Cost and Care: An Investigation of Alternatives11 (Ourhcm, NC: Center for the Study of Aging and Human Develoj:XIlent, Duke University Medical Center, July, 1974). (Mimeographed.)

PAGE 34

rs York City.43 A survey of eight Medicaid offices in New Jersey found that, overall, about 35% of the lower intensity patients studied were capable of maintenance in the community if provided appropriate home services44. The Congressional Budget Office conservatively estimates that 10 to 20% of patients in skilled nursing facilities and 20 to 40% of intermediate care facility patients require a lower level of care than they are presently receiving.45 Considering the proportion of the Medicaid long-term care dollar which pays for institutional care, it seems apparent that the money might be more efficiently spent. Labeling a patient as inappropriately placed or receiving inadequate care raises many questions. First, who makes such a determination? Is it the professional care giver, the patient's family, or the patient himself? Various perspectives yield differ-ent judgments. For example, differences in training across professions (such as nursing and physical therapy) are known to affect professional judgments about the appropriateness of care.46 Second, 43J. T. Gentry and v. R. Curlin, 11The Illinois Long-Term Care Classification Instrument: Use Experience Within the New York City Medicaid Program11 (New York: Department of Health, Medical Sec tion, Bureau of Health Care Services, May, 1975). (Mimeographed.) 4 4urban Health Institute, .. Appropriateness of Long-Term Care Placement: A Study of Long-Term Care Patients in the New Jersey Medicaid Program.. (East Orange, NJ: Urban Health Institute, September, 1977). (Mimeographed.) 45 CBO, p. 18. 46Alan Sager, Learning the Home Care Needs of the Elderl y : Patient, Famil , and Profess1onal Views of an Alternat1Ve to .tnstltutionalization Waltham, MA: Levinson Po 1cy Institute, November, 1979)' pp. 3-4 ' .

PAGE 35

16 is the decision based on normative or comparative standards? Con-elusions do not always take into consideration the 1 imited availability of in-home services, nor do they consider the quality of alternative services. Thus, comparisons may be among no care at all, or care of diminished quality.47 Finally, what cost is society willing to incur in order to increase the quality of care and/or numbers of patients appropriately placed? For example, in a 1974 study in Minnesota, Greenberg found that of the 18% of the skilled nursing facility patients who could be treated at home, less than half could be treated there for less cost than in a nursing home.48 In a study of hypothetical home care, the Levinson Policy Institute found that less than half of the nursing home patients reviewed could possibly be treated at home for less cost.49 While these and similar questions are not answered here, they do provide an indication of the difficulties within the system. Together with the cost concerns mentioned, questions such as these have given i mpetus to the search for policy options. 47For a more thorough discussion of these issues see, Jay Greenberg, David Doth and Allan Johnson, A Coordinated Approach to the Delivery of Long-Term Care: Urban and Rural POdels (Mlnneapolis: Center for Health Services Research, University of Minnesota, 1980). (Mimeographed.) 48Jay Greenberg, 11The Costs of In-Home Services, .. A Planning Study of Services to Noninstitutional i zed Older Persons 1n Mlnnesota, ed. Nancy Anderson (Mlnneapolls: School of Publ1c Affa1rs, University of Minnesota, 1974), pp. 35-46. 49sager, pp. 240-252.

PAGE 36

17 Home Health Care The costly nature of institutional care, in addition to the inappropriateness of that care for certain types of long-term care patients, has motivated a look at alternative care modalities. Two alternatives, adult day care and home health care, have been subject to intense study in recent years.SO One of these--home health care--is reviewed below. Background. The concept of providing health care in the home is not a new one in the United States. Its origins date back to the late 1700's when home services were provided on a voluntary basis by vis it i ng community nurses. In the East and Northeast, these services fostered the development of the visiting nurse association (VNA). In the South and West, community nursing became a service provided by public health departments. Interestingly, even though other types of agencies have been organized expanded, the VNA continues to be located predominantly in the Northeast, and the health department . agency in the South. However, in recent years the number of hospital-based providers has increased substantially in the large urban areas.Sl Thomas Wan, and Barbara Livieratos, Effects Homemaker Services for the Chronicall Ill: 51John Hanmond, Final Report: Applied Research in Home Health Care Services, Vol. III: Commun1ty Level Utilization Analys1s, Pub licat}on No. (OPEL 79-3) (Washington, D.C.: Health Services Administration, DHEW, 1979), pp. III.l0-III.l2.

PAGE 37

18 Although the number of home health agencies and expenditures on home care have increased substantially since the passage of Medicare in 1965, the development of home care as a major component of the health care delivery system has yet to occur. Indeed, in 1977 pub-1 ic programs spent less than 10% of their funds on all home-based care. The result is that certified home health services are still not universally available. Except in the populous Northeast, home health services are only available to some 70% of the Medicare beneficiaries who live in non-metropolitan counties.52 The figures cited above should not be construed to mean that federal programs have had little impact on the development of home health agencies. To the contrary, in 1963 about 250 agencies Y.Quld have been certified according to the regulations later set forth under Medicare.53 In 1966 approximately 1,275 agencies were certified. By 1975 that figure had more than doubled. The growth of home health urider federal programs has not been consistent, however. With the introduction of benefits for homebased skilled nursing care under Medicare, the number of certified agencies increased from 1 ,275 in 1966 to over 2,400 in 1969.54 . As mentioned, the resulting cost of services provided by these and other health care agencies (principally hospitals and nursing homes) increased dramatically. Unplanned expenditures and the alleged 52u.s., Congress, Home Care Services, p. 54. 53u.s., Congress, Home Care Services, p. 60. 54u.s., Congress, Home Care Services, p. 63.

PAGE 38

19 provision of substandard care led to the tightening of Medicare regulations. The definition of skilled nursing care provided in both institutions and at home was revised and more narrowly defined. Emphasis was placed on the intent that the Medicare home health care patient have rehabilitative potential. In other \I!Ords, maintenance care was not covered. The strengthened regulations led to a significant reduction in services by most providers. Retroactive denials imposed by fiscal intermediaries caused some agencies to decrease or stop their participation i n the Medicare program, while others were forced to declare bankruptcy. Between 1969 and 1973, the total number of Medicare certified home health agencies actually decreased, from 2,416 to 2,318 providers.55 With increasing attention to the escalating cost of institu-tional care, federal pol icy shifted again in 1972 to encourage growth in the home health care f ield. Amendments to Medicare in that year expanded post-hospital home health benefits to cover indi-viduals with chronic renal failure. Also, the coinsurance pa.)111ents of the progran were eliminated. By 1973, these amendments resulted i n significant year 1 y increases in fed era 1 reimbursement for home health care.56 The ni.ITlber of Medicare agencies again grew, and, as 5 5John Hammond, p. III.29. 56u.s., Department of Health, Education, and Welfare, Health Care Financing Administration, Office of Policy, Planning, and Research, Medicare: Utilization of Home Health Services,. 1976, Research and Statistics Note No. 2 ([Washington, D.C.: Government Printing Office, June, 1978]), p. 3.

PAGE 39

20 of June 1978, 2,612 agencies had been certified to provide home health care. In summary, a review of the last decade of federal policy points to a certain degree of indecision about the extent to which home health care services should be allowed to grow under federal programs. Current policy shows a similar pattern. On the one hand, the Medicare-Medicaid Anti-Fraud and Abuse /lmendments of 1977 (P.L. 95-142) mandate a review of the de 1 ivery of home health and other in-home services provided under Medicare, Medicaid, and Social Services. A report submitted to Congress by the Department of Health, Education, and Welfare (DHEW) pursuant to that legislation (some times called the H.R. 3 Report), made no recommendations for the expansion of services primarily on the basis of budgetary constraints. On the other hand, Congress is currently entertaining the idea of expanding benefits. The policy d ilemma related to this area is not conf ined to the federal government. Providers differ among themselves on a nlJllber of questions (e.g., whether or not the focus of home health services should be confined to rehabilitative and skilled care; and whether or not the optimum mix of staff should be wei ghted toward professionals or paraprofessionals). The argument that proprietary (forprofit) agencies should be eligible for Medicare certification has also yet to be settled. At the forefront remain important determinations about what types of providers and for what type of patients home health care is an appropriate substitution for hospital or nursing home care. Although far from conclusive, several studies suggest that home health care may be an appropriate alternative for

PAGE 40

21 a portion of the elderly and disabled population.57 Whether or not it substitutes for and is less costly than institutional care is another issue, which is discussed later in this report. Comparative Costs of Home Health Care. The cost of home health services has been an issue in a number of studies. Findings to date indicate that there is some evidence to substantiate the claim that home care is less costly than institutional care for a portion of the long-term care population. A 1977 Accounting Office (GAO) study categorized 1,609 elderly people on the basis of their social, economic, mental, phys-ical, and functional status. Other information, such as living arrangements, services received, provider of services, and direct service cost was also gathered. The costs of services furnished by family and friends were estimated at the agency rates. Comparisons were drawn between non-institutionalized and institutionalized people. The GAO found that a direct relationship existed between the degree of impairment and the number and type of services received. At the lower levels of impairment, transportation, periodic screen ing, and social/recreational services were most i mportant. At the extremely impaired level, social and recreational services were less important, while personal care, full-time monitoring and skilled nursing care were required. Support by family and friends proved to 57chapter II of this volume contains a detailed review of the major studies in the field.

PAGE 41

22 be increasingly important as the level of impairment rose. At all levels, over 50% of the required services were provided by family and friends. For the extremely impaired patients, only 20% of the services received were supplied by agencies. Of the average $845 per month that it cost to maintain these patients, $673 ....orth of services were furnished by family and friends.58 The average cost to for a patient in a nursing home in the GAO study was $458 per month. This figure was greater than the average total cost of maintaining all elderly persons in their own homes, except for those in the extremely impaired group. It should be noted, however, that costs for non-institutionalized patients included amounts to cover the services of family and friends, but these were actually donated services. In the most extreme group, the total cost of services provided Dy an agency was $172 per month, a figure which represented only about 37% of the cost of nursing home care. Using the GAO calculations, only about 10% of the non-institutionalized elderly fell into the category where total cost {program and living) of home services was more than the cost of institutional care.59 When only costs to the government were con-sidered, it appeared that an even larger number might be maintained at home for less than nursing home costs. 58comptroller General of the U.S., General Accounting Office, Home Health: The Need for a National Polic to Better Provide for the Elderly, Publication No. HRD 78-19 Washington, O.C.: Government Pr1nt1ng Office, 1977), pp. 9-22. {Comptroller General of the U.S. is hereafter referred to as GAO.) 59Ibid.

PAGE 42

23 The GAO findings are supported by a recent study completed by the Visiting Nurse Service of New York and the New York City Health Systems Agency. These agencies found that a disproportionate number of the population accounted for most home care costs (i.e, 10% of the population incurred almost 47% of the costs). 60 However, contrary to the GAO report, neither living arrangements nor patient age was a relevant factor. The degree of impairment (in this study, defined as the amount of assistance necessary for carrying out cer-tain activities of daily living) was the factor of primary impor-tance in producing home care which cost more than institutional care. The study concluded that, for the 70% of the patients whose charges were less than the average {of $27 per day, inclusive of service and living expenses), home health care was considerably less expensive than most skilled nursing home care. A study of the Chelsea-Village Program, also in New York City, supported this hypo-thesis. It estimated that home-based care for semi-ambulatory patients costs only about 50% of nursing home rates.61 Although there is some preliminary evidence that home health care may be less costly than institutional care on a per unit basis, whether it will be economically efficient in the long run is yet 6 0Geraldine Widmer, Roberta Brill, and Adele Schlosser, "Home Health Care Services and Cost, .. Nursing Outlook (August, 1978), pp. 488-489. 61Philip Brickner et al., "The Homebound Aged: A Medically Unreached Group," Annals of Internal Medicine, LXXXII (1975), 1-6; Linda Scharer, "Analyzing the Cost,11 Home Health Care for the Aged, ed. Philip Brickner (New York: Appleton-Century-Crofts, 1978), pp. 229-245.

PAGE 43

24 undetermined. Economic efficiency reflects the plausible, techno-logically efficient method of care given cultural, political and legal constraints and the preferences of patients and their fami-lies. It cannot be determined by merely costing out, at existing prices, all techno 1 ogi cally efficient methods of care. Significant changes in the long-term care delivery system will likely affect prices by changing the demand and supply of resources and changing the behavior of patients, providers and families. Thus, potential changes in the delivery system through the expansion of home health services are likely to result in a higher total cost of implementing the expanded home program.62 Therefore, although patients and families request additional home care services, policymakers continue to appear reluctant to make large scale commitments, given the contin-ued, although hopeful, uncertainty of its cost saving potential. Major Barriers to Home Care. Similar to other long-term care services, the provision of home health care is hampered by the lack of a coherent benefit structure, service delivery and reimbursement system. A series of regional DHEW public hearings in 1976 high-lighted these deficiencies. Among the problems identified were: 1) The absence of unified federal progran definitions and eligibility requirements. 62Lewis Freiberg, Jr., .. Substitution of Outpatient Care for Inpatient Care: Problems and Experience, .. Journal of Health Politics, Policy and Law, IV (Winter, 1979), 493-494.

PAGE 44

25 2) An incomplete knowledge of the system which leaves to the patient and family the task of assessing the appropriateness and quality of available services. 3) The absence of a coordinated range of high qual .ity home care services to provide an essential continuum of care.63 As discussed, federal reimbursement for home services is spread over a number of different federal programs, including Medicare, Medicaid, and Title XX (Social Services) of the Social Security Act, and Title III (grants for State and Community Programs on Aging) of the Older .Americans Act. Each of these programs has its own defini-tion of the population served, and different requirements for pro gram eligibility and the duration and scope of services. In addi-tion, reimbursement methods differ. However, in spite of program differences, all of them are important to the provision of home care because certain of the services under each are necessary for proper support. No one of these programs provides the fu 11 scope of ser-vices needed to maintain the functionally impaired patient on a long-term basis in a non-institutional setting. While attempts have been made to bring the e 1 ements of these programs together, OHEW officials generally concede that the programs as they stand, given present legislation, defy coordination.64 63u.s., Department of Health, Education, and Welfare, Home Health Care: A Report on the Re ional Hearin s, September zu-:October , 976, Pub 1cat1on No. 76-33 Wash1ngton, 0 .. : Government Print1ng Office, 1977), pp. 3-5. 64GAO, pp. i-iv.

PAGE 45

26 Medicare and Medicaid both provide home health services, although these programs differ in essential aspects such as eligibility and coverage. Medicare hospital insurance (Part A) provides 100 home care visits per benefit period following the start of one spell of illness, only if the patient has been hospitalized for at least three consecutive days for this illness. Medicare supplemental medical insurance (Part B) allows 100 home care visits per year, but without prior hospitalization. The emphasis in Medicare is on the provision of short-term skilled care with little attention given to the social support necessary for long-term in-home maintenance. Since 1970, home health care has been a required service under Medicaid. Unlike Medicare, it does not require for eligibility a prior hospitalization or the need for skilled nursing care. However, social services are not included. Both Medicare and Medicaid cover a large proportion of the medical and health related services, but tend to ignore necessary social support programs such as transportation and home delivery of meals. Title XX of the Social Security Act offers a variety of home based social services, such as home health aides, homemakers, personal care, etc. Title III of the Older .Americans Act provides additional health-related services including health education, heme repairs and meals either in the home or in a congregate setting. The fragmentation and lack of coordination of services filters down to the community level where often a patient who could be maintained in the home is institutionalized for lack of a properly organized long-term care system. Thus, while a sufficiently broad

PAGE 46

27 range of services is available from the combination of various federal programs to treat a chronically ill patient in the home, it becomes very difficult to do so at the local level. Services are provided by a number of different agencies, and seldom does a local mechanism exist for coordinating them. In addition, services such as respiratory therapy, respite care, and transportation are not covered. The 1977 Comptroller General's Report to the Congress on Home Health Care summarized this issue and recommended a unified policy: Home Health Care and other related services for the elderly are not being effectively coordinated. Services are avail able through so many different programs that effective coordination and delivery of home health and other in-home services seems close to impossible. Ser vices provided are not accessible through a single entry point. Inter-agency/intra-agency agreements between federal, state and local agencies have not provided effective coordinated services to beneficiaries. We recommend that the Secretary of HEW develop a national policy to be considered by the Congress which would consolidate home health activities. HEW should promote the establishment of a comprehensive single entry system by which individuals are assessed, as to their needs, prior to placement in a program. HEW should consider the services that are currently provided under Titles VIII, XIX, and XX of the Social Act and Titles III and VII of the Older Americans Act. Summary of Current Policy Issues In summary, long-term care expenditures have risen rapidly in recent years. Increased spending for institutional long-term care, in particular, has been criticized. Dissatisfaction has surfaced 65 GAO, pp. 52-54.

PAGE 47

... 28 with regard to the financial demands of institutional care, as well as the effectiveness of that care. There are major gaps in the knowledge about alternatives to that care. Import ant questions concern the numbers of people in need of the various kinds of long-term care services, and the quantity, and type of appropriate providers of such services. There is concern about the relative cost and effectivenss of institutional versus non-institutional care. Undoubtedly, uncertainties about cost, utilization, and effectiveness of non-institutional long-term care alternatives will most likely impede diversification of services in the short run. If one issue is clear, it is that insufficient information exists to address adequately all of the questions and concerns posed. It is to this insufficiency that this document now turns. Evaluation Research to Reduce Uncertainty The intent of evaluation research is to define, expand, assemble, and make more pertinent the information base upon which decisions are made. Its aim is to assess the effect of specific policy or program on a target population (e.g., individuals, organizations, communities, or systems).66 When used, it has the potential to decrease the uncertainty in decision-making. The assumption is that by providing data on the consequences of policies, the eval-uator aids the decision-maker in selecting 11wise choices about 66carol Weiss, Evaluation Research Prentice-Hall, Inc., 1972), pp. 4-7. (Englewood Cliffs:

PAGE 48

29 future courses of action ... 67 The usefulness of evaluation research is then founded on the premise that decision-makers incur a certain amount of uncertainty when faced with choices. Uncertainty is that 11gap11 between what is known and what needs to be known to make a correct choice.68 Indeed, there is much opportunity to reduce uncertainty in the area of long-term care. Organized efforts to reduce uncertainty with regard to the effects of social programs are a relatively recent phenomenon. Although progrcm managers and policy-makers have traditionally judged the effects or values of their programs, little systematic attention was paid to policy analysis/evaluation until the estab• lishment of the Planning, Programming, and Budgeting System (PPBS) in the Department of Defense in the early 1960's.69 Thereafter, Congress became increasingly interested in evaluation and passed major legislation which required and included supportive funds. For example, amendments to the Economic Opportunity Act of 1967 and to the Social Security Act in 1972 authorized wide-scale evaluation research efforts. Not surprising, the bulk of monies which fund policy research and evaluation comes from the federal government. 67carol Weiss, "Where Politics and Evaluation Research Meet, .. Evaluation, I (1973), 43. 68Ruth Mack, Planning on Uncertainty (New York: Wiley-Inter-science, 1971), p. 1. 69David Novick, ed., Program Budgeting (2nd ed.; New York: Holt, Rinehart & Winston, Inc., 1969), pp. xix-xxviii.

PAGE 49

30 Thus far, the impact of most research in the area of non-institutional long-term care has been minimal JO There are at least two plausible explanations. First, confidence in policy analysts' projections of cost impacts of non-institutional services is minimal. This lack of confidence is not unwarranted given the low reliability of cost estimates for the Medicare and Medicaid programs in their early years. As noted, the estimated first year cost of the Medicare nursing home progran was $25 to $50 million, less than one-fifth the final total cost.71 Second, findings of recent research and evaluation studies in the area of non-institutional long-term care are conflicting. Thus, they often increase rather than reduce uncertainty on the part of po 1 icy makers J2 Uncertainty with regard to cost and potentia 1 benefits results in minimal efforts to expand services in the direction of any new progran.73 Evidence in support of this argument was the DHEW report to Congress on home health care cited earlier which eventually was stripped of all recommendations to expand service 70one exception is the specific study described herein which did impact the Medicare home health care reimbursement system; but the effects were incremental compared to the enormity of the issues. 71u.s., Congress, Senate, Special Committee on Aging, Develop ments in Aging, 1969, 91st Cong., 2nd Sess., on S.J. Res. 316 (Washington: D.C.: Government Printing Office, 1970), p. 87. 72see a review of representative studies in Chapter II of this vo 1 ume. 73sheila Burke, "Home Health Politics and Policy: Senate Finance Committee" (paper presented at the meeting of the American Hospital Association on Hospital-Based Home and Hospice Care, Arlington, VA, September, 1979).

PAGE 50

31 availability, ostensibly on the basis of uncertaintyJ4 Further-more, recent experiences with the abuses and disappointments result-ing from the deinstitutionalization of the mentally ill reinforces the argument for caution.75 The existence of conflicting research and evaluation findings does not mean that these pursuits should be halted. On the con-trary, as mentioned, the evaluation of long-term care alternatives has only recently begun. This study, hopefully, is a contribution to these efforts. While it may have a minimal impact on the actual resolution of major long-term care policy issues, it does suggest a systematic framework for a nati onal program of policy-related evalu-ation research. A Specific Research Area Between 1976 and 1979, the Office of Planning, Evaluation and Legislation of the Health Services Administration (DHEW) supported home health care research relevant to that agency• s responsibility for promoting the development and expansion of home health service capacfty. Part of that effort was a study of the cost per episode of home health care conducted by the Center for Health Services Research, University of Colorado Health Sciences Center. It is this study which is the vehicle for highlighting the long-term care 74u.s., Congress, Home Care Services, pp. 51-68. 7 5Ernest Gruenberg and Janet Archer, .. Abandonment of Respon-sibility for the Seriously Mentally Ill," Milbank Memorial Fund Quarterly, LVII (Fall, 1979), 485-506.

PAGE 51

32 policy issues that need to be addressed through research and evaluation. This study addressed several basic questions: (1) How were the cost and uti1ization of home health care influ-enced by various patient, provider, and corrmunity characteristics? (2) Which of the characteristics studied were the most important in terms of the observed utilization and cost variations? (3) What are the relevant programmatic, public policy, and research implications of the findings? These were pursued by analyzing how several groups of characteristics were related to the utilization and cost of care. Two assumptions were implicit: (1) that patient needs varied and were reflected by patient characteristics (e.g., age, living situation and functional status); and (2) that services delivered were in response to differences in need but were affected by provider/health system characteristics (e.g., size of provider and utilization of acute care in the service area). Outcome measures (e.g., change in functional status over time) represented the interface between patient characteristics, provider characteristics and the utilization/cost of home health care. The study was thus intended to provide a better understanding of the determinants of the costs of home health care, which, in turn, should improve the reliability of estimates of home health care cost and utilization over the coming decade. In addition, the information pertaining to the types of patients/episodes which were

PAGE 52

33 lower in cost was particularly helpful in clarifying and shaping public policies regarding the placement of patients in institutional versus home health care settings. The analyses in the study were carried out using data from two samples. One sample was made up of four agencies (all visiting nurse associations) participating in a discharge abstract program administered by the Massachusetts Oepar . tment of Health. The second sample was comprised of four hospital-based home health providers participating in the Blue Cross of Greater Philadelphia Home Care Program. The two samples totaled approximately 2,800 episodes of illness during the study year, 1976. In order to address the key questions of the study, relationships were examined in varying combinations and levels of detail. Descriptive statistics were used to test utilization and cost dif-ferences in terms of single variables. In order to determine the effect of individual variables in conjunction with others, factor analysis, canonical correlation analysis and multiple regression analysis were used to relate the utilization and cost of care to the three categories of variables. A complete description of the study methodology, findings and implications is contained in the pages which follow. In order that the reader may more fully appreciate the significance of the study, a review of other research in the area of long-term care preceeds its presentation. In view of continuing long-term care policy questions, this document concludes with suggestions for an overall research strategy intended to meet the needs of policy makers . .

PAGE 53

CHAPTER II STATE OF THE ART Introduction As discussed earlier, the primary focus of the study described herein was the cost-effectiveness of long-term care. Thus, the study was concerned with the refinement of institutional and noninstitutional long-term care patient descriptors (often referred to as case mix), and measures of the quality and cost of care. The purpose of the following chapter is to provide an overview of the conceptual and analytic state of the art in each of these areas. This overview is presented in order to better understand how this study relates to the efforts of others. Although the topics of case mix, quality, cost, and cost-effectiveness are discussed separately, they are actually closely related to one another; thus a number of the studies cited pertain to multiple issues and are therefore mentioned several times. Case Mix Case mix typically refers to certain patient characteristics (e.g., problem, diagnosis, and functional status) which reflect the type and quantity of care needed. In theory; a strong relationship should exist between case mix and the quality and cost of health

PAGE 54

35 care. However, while these relationships are conceptually clear, their empirical validation has been less successful in long-term care than for acute care.1 The following discussion of the problems of case mix will measurement points to specific areas of concern. In long term care, the age of a patient is related to functional incapacities,2 use of health services,3 and subsequent cost of care. Scharer and Boehringer found that the most significant determinant of the cost per visit for home care was the client•s age.4 Another study by Widmer and others of home health care in New York City found that total costs of care actually decreased with advanced age, speculating that those who are elderly 1 Larry Shtman and Harvey Wolfe, 11The Use of Case Mix and Case Comp 1 ex i ty in Prospective Hospi t a 1 Re imbursemen t11 (Pittsburgh, University of Pittsburgh, August, 1975). (Mimeographed.); Martin Feldstein, 11Hospital Cost Variation and Case Mix Differences, .. Medical Care, III (April-June, 1965), 95-103; Robert Fetter et al . , 11Case M1x Definition by Diagnosis-Related Groups, .. Medical Care, XVI I I, supplement (February, 1980) , 1-53; Judith Lave and Lester Lave, 11The Extent of Role Differentiation Among Hospitals, .. Health Services Research, VI (Spring, 1971), 15-38. 2Thomas Wan, 11 Age Severity of Dis ab il i ty, ,. Review of Pub 1 i c Data Use, III (1975), 29-32. 3Karen Davis and Roger Reynolds, 11Medicare and Utilization of Health Care Services by the Elderly, .. Journal of Human Resources, X (Summer, 1975), 361-377; Thomas Wan, .. Interpreting a General Index of Subjective Well-Being .. (paper presented at the meeting of the Gerontological Society, San Francisco, November, 1977). (Mimeographed); William Weissert, 11Costs of Adult Day Care: A Comparison to Nursing Homes, .. Inquiry, XV (March, 1978), 10-19. 4Linda Scharer and John Boehringer, Home Health Care for the Aged: The Program of St. Vincent•s Hosp1tal, New York c1ty (New York: Boehringer Associates, 1976), pp. 9-11. (Mimeographed.)

PAGE 55

36 require less costly 11basic11 care, at less frequent intervals than younger, more acutely ill patients.5 It is reasonable to suggest that the relationship of age to utilization or general health ser-vices is bimodal, since both the very young and the very old use more health services than those in the middle age range. The direc-tion of the relationship of age to cost of care is inconsistent in long-term care; nevertheless, it is frequently used as one of the many proxy measures of case mix. Living Arrangement A number of studies have found that the patient's living arrangement affects the quantity and type of care utilized.6 Hence, living arrangement is another proxy measure often used for case mix used in long-term care, and home health care in particular. The assistance provided by family and friends can signficantly reduce the time spent by professional care givers. With this help, reducing the amount of purchased services should reduce total costs; although not all utilization will be affected in the same manner by the patient • s 1 i vi ng arrangement. The New York study previously 5Geraldine Widmer, Roberta Brill, and Adele Schlosser, 11Home Health Care Services and Cost,11 Nursing Outlook (August, 1978), pp. 488-489. 6 william Weissert, Thomas Wan, and Barbara Livieratos, Effects and Cost of Oa Care and Homemaker Services for the Chron1ca lly I : A Random1zed Expenment Wash1ngton, D.C.: Nat1ona enter for Health Services Research, OHEW, 1980), pp. 14-15; Alan Sager, Learnin2 the Home Care Needs of the Elderly: Patient, Family and Profess1onal Views of an Alternative to Institutionali-zation (Waltham, MA: Levinson Policy Institute, 1979), pp. 241-248.

PAGE 56

37 cited suggested that those living alone received essentially the same number of nursing visits, but twice as many social service visits as those living with others.? Thus, living arrangement is one measure of case mix often found in home health utilization and cost studies; yet, its relationship to either utilization or cost is inconsistent. Diagnosis Although commonly used to characterize the needs of acute care patients, medical diagnoses are not precise or comprehens ive enough to measure case mix for chronically ill individuals for several rea-sons: the occurrence of multiple diagnoses; the difficulty of establishing a primary diagnosis; and the lack of relationship between primary diagnosis and the nature and extent of individual major care needs. Most long-term care studies have found no strong relationship between the primary diagnosis of the patient and utilization of services or personnel time in either the institutional8 7Widmer, Brill, and Schlosser, p. 490. 8peter Shaughnessy et a 1., Long-Term Care Reimbrusement and Re ulation: A Study of Cost, Case MlX and Quallty: Work1ng Paper 4, First Year Ana ysis Report Denver: Center for Health Services Research, University of Colorado Health Sciences Center, 1980), pp. VI.4-VI.5; Jay Greenberg, Cost, Case Mix, Quality and Facility Characteristics in Minnesota•s Nursin Homes: An Explorator Analysis, First Year Progress Report Minneapolis: Center for Health Services Research, University of Minnesota, 1980), pp. IV.31-IV.51. Bernard Ries and Jon Christianson, Nursing Home Costs in Montana: Analysis and Policy Applications (Bozeman: MOntana State Un1Vers1ty, 1977), lff.; Hirsch Ruchlin and Samuel Levey, 11Nursing Home Cost Analysis: A Case Study, .. Inquiry, IX (September, 1972), 3-15.

PAGE 57

38 or home health care setting.9 Thus, alternate measures of need have been developed, leaving primary medical diagnosis as a supplemental measure of mix. Functional Status An accepted method of describing long-term care case mix is by the determination of individual capacities to function in several areas important to daily life (termed of daily living", or ADLs). Early studies by Katz et a1.10 have since been augmented by other worksll which have added measures of functional status to the basic set of functions--bathing, dressing, feeding, continence, and mobility. For example, the adaptation by Lawton and Brody12 which includes such functions as using a telephone, grocery shopping, and maintaining a checkbook, is especially useful for describ-ing the case mix of a home health agency. Others have expanded the 9 widmer, Brill, and Schlosser, p. 490; Nancy Anderson, A Com-parison of In-Home and Nursin Home Care for Older Persons 1n M1nnesota 1nneapo 1 s: Schoo Pub 1c A a1rs, Unwers1ty o M1nnesota, 1977}, pp. 32-49. 10sidney Katz et al., "Studies of Illness in the Aged: The Index of AOL," Journal of American Medical Association, CLXXXV (1963), 914; Sidney katz and C. kiec1 Akpom, 11A Measure of Primary Sociobiological International Journal of Health Ser vices, VI 493-507. 11Douglas Skinner and Donald Yett, Index for LongTerm Care Patients," in Health Status Indexes, ed. Robert Berg (Chicago: Hospital Research and Educational Trust, 1973), pp. 6982. 12M. Powell Lawton and Elaine Brody, "Assessment of Older People: Self-Maintaining and Instrumental Activities of Daily Living," Gerontologist (1969), pp. 179-186.

PAGE 58

39 basic list to include a total of eight ADLs plus indicators of impairments in three sensory abilities believed important to the care needs of a long-term care patient.l3 Besides using simple measures of dependency/independency in each ADL category, Shaugh nessy et al. used gradations of dependency in each.14 Preliminary findings of this and a companion study which used expanded ADL measures, indicated that resource consumption was more strongly related to moderate levels of dependence than total dependence.lS An important effort in functional status measurement has been the attempt to c . ondense the various measures of ADL into indices which maximize the retrievable information while reducing the total number of variables. Katz et al. in early work identified six ADLs which were empirically determined to be hierarchically related to one another, thereby permitting the classification of patients into homogeneous groups that were ranked ordinally.l6 Skinner 'and Yett 13Peter Shaughnessy et al., An Evaluation of Swing-Bed Experi ments to Provide Lon -Term Care in Rural Hosp1tals, Volume II: F1nal Techn1ca Report Denver: Center for Hea th Serv1ces Research, University of Colorado Health Sciences Center, 1980), pp. V.21-V.S9. 14shaughnessy et al., Long-Term Care Reimbursement and Regula t ion, pp. VI.1-VI.22. 15sh aughnessy et tion, pp. IX.2-IX.26; 18-!2. 16Katz et al., 11Studies of Illness in the Aged,11 pp. 914-916; Sidney Katz et al., 11Program in Development of the Index of AOL, .. Gerontologist, X (1970), 20; Katz and Akpom, 11Primary Sociobiological Functions,11 pp. 494-496.

PAGE 59

40 derived a similar index through Guttman scaling of analagous data.l7 Other efforts to define useful case mix groupings have used factor analytic techniques and other class ifi cation methodologies. Greenberg used canonical correlation to identify a vector of func-tional disability which was strongly related to nursing care costs.l8 Presently, work in the area of functional status index development is being conducted by Shaughnessy et al. using principal component analysis and canonical correlation to group case mix indicators into homogeneous groups.l9 Mental Health Status The description of long-term care case mix has also been expanded through methodologies intended to evaluate the mental health or psychosocial status of long-term care patients. One tool developed by Pfeiffer is a short form for assessment of mental status to rule out organic brain impairment.20 Eisdorfer has argued that specialized, validated instruments are needed to measure depression, withdrawal, and other mental health problems in longtenn care.21 Several of the multidimensional ass . essment systems 1 7 Skinner and Yett, pp. 69-71. 18Greenberg, Cost, Case Mix, Quality, pp. IV.l-IV.l9. 19shaughnessy et al., Long-Term Care Reimbursement and Regula-tion, pp. IV.S-IV.6. 20Eric Pfeiffer, "A Short Portab 1 e Menta 1 Status Questionnaire for the Assessment of Organic Brain Deficit in Elderly Patients," Journal of the American Geriatric Society, XXIII (1975), pp. 433-441.

PAGE 60

41 which are described in the following paragraphs have recently added such measures. Multidimensional Health Status An important advance in case mix measurement has been the development of multidimensional patient assessment tools that incorporate several of the measures described above, such as activities of daily living, mental health status, impairment indicators (e.g., paralysis and pain) and medical risk factors (e.g., elevated blood pressure) .22 Additional major assessment methodologies which examine several different patient characteristics include the OARS (Older American Resources and Service) instrument developed at Duke University23 and the PACE II assessment instrument developed by DHEW. The former rates patients in five key areas of economic and social resource, mental and physical health, and activities of daily living. The latter is a complex instrument which uses an algorithmic model to determine the nature and extent of physical, func-tional, and mental health status. Most current systems of long-term 21carl Eisdorfer, 11Evaluation of the Quality of Psychiatric Care for the Aged,11 American Journal of Psychiatry, CXXXIV (March, 1977)' 315-317. 22E11 en Jones, Barbara McNitt, and Eleanor McNight, Patient Classification for Lon -Term Care: User's Manual, Bureau of Health erv1ces esearch and va uat1on, DH W Was 1ngton, D.C.: Govern-ment Printing Office, 1974), pp. llff; Paul Densen and Ellen Jones, 11The Patient Classification for Long-Term Care Developed by Four Research Groups in the United States,.. Medical Care, XIV (May, 1976), 126-130. 23Eric Pfeiffer, Multidimensional Functional Assessment: The OARS Methodology (Durham, NC: Duke University, 1975), pp. 3-29.

PAGE 61

42 care assessment are closely related to, if not adaptations of, the multidimensional tools described above.24 Problems Recently, some researchers have begun to use problem-oriented case mix measures based, in part, on previous 'r'tOrk by the Colorado Foundation for Medical Care.25 The system is based upon specifica-tion of the most prevalent institutional long-term care problems, including not only medical conditions (e.g., hypertension) and functional and sensory impairment (e.g., immobility and blindness), but also physical, psychological, and social conditions (e.g., pain, depression, and difficult family situations) requiring therapeutic intervention. Twenty-seven problems were found to represent approximately 97% of the patient care needs of long-term care residents.26 Such a problem-oriented approach has the advantage of being directly translatable into service needs, expressed in terms of type and fre-quency of service. For this reason a major advantage of case mix measured by the prevalence of each problem, is that (at least conceptually) it is directly related to the level of resources 24Phyllis Giovannetti, Patient Classification in Nursing: A Description and Analysis, Divis1on of Nurs1ng, DHEW, Publication No. (HRA) 78-22 (Wash1ngton, D.C.: Government Printing Office, July, 1978), pp. 6-10. 26peter Shaughnessy et al., An Evaluation of Swing-Bed Experi ments, pp. V.24-V.25.

PAGE 62

43 required, and thus to costs incurred. Another advantage of the problem-oriented approach to. case . mix is that the extent to which normative service needs are actually met can be used as a process measure of the quality of care. Case mix measurement in home health care has only quite recent-ly expanded to include use of problems as major patient descriptors. The Visiting Nursing Association of Omaha, Nebraska, has been con ducting a four year study intended to develop and validate uniform home health care problem descriptors for use in quality assurance programs.27 The problem list developed used factor analytic tech-niques to cluster 100 problems into four "domains", (i.e., physic-logical, health behaviors, psychosocial, and environmental). Other home health agencies group problems by prognosis in order to link them to quality assurance activities.28 The relationship of longterm care patient problems to resource consumption and patient out-comes is yet undetermined. Levels of Care A final case mix measure relates to current regulatory and reimbursement systems for nurs ing homes and home health care. A s discussed in Chapter I, Congress established two levels of nurs ing 27Karen Martin et al., 11Field Testing of a Problem Cl assification Scheme and Development of an Expected Outcome Scheme with a Methodology for Use," Draft Exective Summary (Omaha: Visiting Nurse Association of Omaha, September, 1979). pp. 2-6. (Mimeographed. ) 28Frances Decker et al., " U sing Patient O utcomes to Eval uate Community Health N urs i ng," Nursi n g Outlook ( Apri 1, 1979) , pp. 278 282.

PAGE 63

44 home care (skilled and for Medicaid reimbursement ' i which were intended to be related to the patient's condition and the intensity of services required for adequate treatment. The requirement that the higher level of care (skilled) be reimbursed at a higher rate than the intermediate level of care assumed that it was more costly to treat skilled patients. However, these levels of care were not based upon empirical analysis.29 Uniform definitions of the two levels of .care were never made precisely for all states30 and as a result, the classification of residents by level of care varies widely by state.31 Preliminary findings of Shaughnessy et al. substantiated the tendency for skilled nursing home residents to need more nursing care, but less psychosocial care than intermediate care residents.3 2 However, differences between the two were slight, indicating the limited utility of this system of level of care as a valid measure of case mix. Ironically, as the skilled/intermediate classification system is being increasingly called into question for nursing home 29sharon Winn, .. Assessment of Cost-Related Characteristics and Conditions of Long-Term Patients, .. Inquiry, XII (December, 1975), 344-353. 30Thomas Willemain, Christine Bishop, and Alonzo Plough, The Nursing .Home 11Level of Care11 Problem (Waltham, MA: Brandeis Un1versity, 1979), pp. 92-95. (Mimeographed.) 31oouglas Holmes et al., A National Study of Levels of Care in Intermediate Care Facilities, Final Report, Health Services Adm1n1strat1on, DREw (Wash1ngton, D.C.: Government Printing Office, April, 1976), pp. 6-32. 32sh aughnessy et a 1., Long-Term Care Reimbursement and Regulation, pp. VI.6-VI.7.

PAGE 64

45 patients, home health care providers are beginning to advocate a similar system. An ad hoc group of the five national home health care associations issued a position paper in 1979 calling for four levels of home care, similar to those found in institutional care: intensive, intermediate, maintenance and personal .33 The relation-ship between such a level of care classification system and actual resource consumption thus requires further study. In summary, although neither a single measurement system, nor a single set of case mix variables exists which adequately and accurately describe the health status of long-term care patients, certain classification systems have been found by researchers to be especially useful. Several of these appear to be related to resource consumption and thus offer potentia 1 for use in 1 ong-term . care planning. By using multiple case mix descriptors, that are designed to measure medical care needs, functional disabilities, psychosocial, and environmental problens, the manner in which these relationships vary by service setting can be determined. This is especially important to determine which modality is best suited to the care of various kinds of patients. Urifortunately, much careful research is still needed to validate the various systems which have been proposed. 33Assembly of Ambulatory and Home Care Services of the American Hospital Association., et al., A Prospectus for a National Home Care Policy ([n.p.: n.n., 1978]), p. 3.

PAGE 65

46 Quality Of the four major areas discussed in this section, the measure-ment of quality of care is the most controversial. For clarity, this section is arranged according to Oonabedian•s classic tricotomy (structure, process, and outcome),34 so that the strengths and weaknesses of each type of measurement as applied to long-term care can be reviewed. Structural Measures of Quality Early approaches to quality assessment frequently focused on structural measures (sometimes called input factors) which are pre-cursors to quality (e.g., availability of services, physical facilities, and staffing resources). Structural factors fonn the basis for institutional liscensure and certification, on the assunption that they represent necessary, albeit minimal, conditions for the delivery of adequate care. Institutional standards are based on structural factors, at least in part, because of ease of measure-ment. Factors such as ownership, number of beds, safety provisions, and staffing ratios have been used as measures of structural aspects of quality. Research has failed to show significant associ at ions between many of these measures and process or outcome measures of quality. (No studies to date have looked at this issue in home health care.) 3 4 Avedis Donabedian, .. Evaluating the Quality of Medical Care, .. Milbank Memorial Fund Quarterly, XLIV (July, 1966}, 166-180.

PAGE 66

47 Several studies have found little relationship between nursing home ownership and the care provided to residents;35 one investigator did find that non-profit church affiliated nursing homes provided better care. 36 Holmberg and Anderson37 and Dennis, Burke, and Garber38 found that the training and experience of the person in charge were more important than structural measures of quality. Two other studies suggested that the size of the organization was inversely related to quality of care.39 Other studies indicated that the larger the facility, the better the care.40 Still others suggested 35R. Hopkins Holmberg and Nancy Anderson, "Implication of Ownership for Nursing Home Care," Medical Care, VI (July-August, 1968), 300-307; Winn, p. 348. 36Leonard Gottesman, "Nursing Home Performance as Related to Resident Traits, Ownership, Size, and Source of Pa.J111ent," American Journal of Public Health, LXIV (March, 1974), 269-276. 37Holmberg and Anderson, pp. 300-304. 38LJ111an Dennis, Robert Burke, and Kim Garber, "Quality Evaluation System: An Approach for Patient Assessment," Journal of LongTerm Care Administration, V (Summer, 1977), 28-51. 39shayna Greenwald and Margaret Linn, "Intercorrelation of Data on Nursing Homes," Gerontologist (Winter, 1971), pp. 337-340; Peter Townsend, The Last Refuge (London: Routledge and Kegan Paul, 1962). 40w. Beattie and J. Bullock, "Evaluating Services and Personnel in Facilities for the Aged", Geriatric Institutional Management, ed. M. Leed and H. Shore (New York: Putnan' s, 1964), pp. 119-142.

PAGE 67

48 that there was no relationship.4 1 A recent study by Linn, Guerel, and Linn42 found that a higher professional staff-to-patient ratio, a better medical records system, and the provision of more services were positively related to discharge from institutional care. Overall there is a weak relationship between various structural characteristics of nursing homes and the quality of care. Results of long-term care research support the general conclusion that the relationship between structural measures of quality and the delivery of care, and between care and its effects on patient health status, is not well established.43 For this reason, recent efforts have focused more intensively on process and outcome measures of quality. Process Measures of Quality Process measurement of quality entails the assessment of quality in terms of the resources consumed and procedures (i.e., 4 1Timothy J. Curry and Bascom W. Ratliff, 11The Effects of Nursing Home Size on Resident Isolation and Life Satisfaction,11 Gerontologist (Autumn, 1973), pp. 295-298; Beaufort Longest et al., 11An Empirical Analysis of the Relationship of Selected Structural Factors to Quality of Patient Care in Nursing Homes,11 Journal of Long-Term Care Administation, III (Spring, 1975), 16-26. 4 2Margaret Linn, Lee Guerel, and Bernard Linn, 11Patient Outcome as a Measure of the Quality of Nursing Home Care, 11 American Journal of Public Health, LXVII (April, 1977), 337-344. 43Avedis Oonabedian, Needed Research in the Assessment and Monitorin of the Qualit of Med1cal Care, NCHSR Research Report Series Wash1ngton, D.C.: Government Pnnting Office, July, 1978), 2ff.; Robert Brook, Quality of Care Assessment: A Comparison of Five Methods of Peer Review (Rockville, MD: National Center for Health Services Research, OHEW, July, 1973) pp. 16-32; William McAuliffe, 11Measuring the Quality of Medical Care: Process Versus Outcome,11 Milbank Memorial Fund Quarterly, LVII (1979), 118-149.

PAGE 68

49 services) performed during the provision of care. Both data collection and the definition of criteria are more difficult for process measurement than in the structural analysis of the quality of care. However, process measures generally have greater validity, since they are more directly related to patient health status. Yet problems with the reliability and generalizability of process measures persist. Since process analyses generally use medical records as their primary source of data, it is difficult to differentiate between procedures that were performed but not recorded and proce dures that were never performed. Linn et a 1 . 4 4 cone 1 uded in a recent study that both records and staff interviews should be used to collect data, since each was found by itself to be a valid, but incomplete, source of information. In all areas of health care, the most prevalent technique for evaluating process quality of care is the comparison of process data to explicit criteria (or service standards) which have been generated by panels of multidisciplinary experts.45 Inclusion of 11excep tion11 categories for each service permit the criteria to be applied individually to each long-term care patient. Using such a criteria 44sernard Linn et al., 11Validity of Impairment Ratings Made from Medical Records and from Personal Knowledge, .. Medical Care, XI I (April, 1974), 363-368. 45Brook, p. 46; Robert Kane, Prognosing the Course of Nursing Home Patients, proposal accepted for funding by National Center for Health Services Research, OHEW (Los Angeles: Rand, 1978). (Mimeographed.)

PAGE 69

50 set, Shaughnessy et a1.46 found that the quality of care in nursing homes was slightly better than the quality of care in hospitals for certain long-term care problems. Like structural criteria, process criteria represent a useful tool for evaluating the quality of care. However, the use of either structural or process measures is based upon the assumption that there is a causal relationship between them and the outcome of care as expressed in the final condition of the patient. As indicated, this causal relationship has not been empirically demonstrated. As an example, Linn and others47 reported that pati"ent outcomes were highly correlated with nursing time per patient (a process measure) but showed little relationship to the size of the facility, staff-to-patient ratios, personnel or safety practices (structural measures). Outcome Measures of Quality Outcome evaluation in patient care has received a great deal of attention recently as the most appropriate way to approach the concept of qua 1 ity. 48 A number of measures of outcome have been used in long-term care, such as admission to acute care, readmission to nursing homes, or prevalence of particular diseases or undesirable 46shaughnessy et al., An Evaluation of Swina-Bed Experiments1 pp. V.5-V.30. 4 7Linn, Guerel, and Linn, pp. 363-368. 48rnstitute of Medicine, Assessin1 Quality in Health Car. e: An Evaluation (Washington, D.C.: Nat1ona Academy of Sc1ences, Novem-ber, 1976), pp. 6-32.

PAGE 70

51 physical occurrences, such a . s bed sores.4 9 Rather than measuring outcomes in terms of indicators of preventable morbidity, changes in patient health status have gained increased recognition. The simplest measure of health status change is, of course, mortality. Beyond this, however, patient status can include virtually any health-related characteristic. Although measures of physical well being have been a primary focus of analysis, it has been increasing-ly recognized that the scope of outcome quality must be broadened to include such factors as psychosocial well-being, functional status, and, in some instances, satisfaction with care.50 Conceptually the measurement of outcome represents the ultimate validation of the effectiveness of the care process. However, two major problems exist in using outcome measures alone as a basis for effectiveness assessment: the difficulty in specifying appropriate outcomes of care and the difficulty in relating individual outcomes to outputs (or process of care) .51 Both are discussed in the fol-lowing paragraphs. 49Sharon Soroka, "Quality of Care Evaluation in SNFs Using Bedsores as an Output Indicator" (paper presented at the meeting of the Prnerican Public Health Association, Washington, D.C., November, 1976). (Mimeographed.) 50ran McDowell and Carlos Martini, "Problems and New Directions in the Evaluation of Primary Care," International Journal of Epidemiology, V (1976), 247-250. 51Donabedian, "Evaluating the Quality of Medical Care," pp. 166-206; Avedis Donabedian, "Patient Care Evaluation," Hospitals, XLIV (April, 1970), 131-136.

PAGE 71

52 Specifying Appropriate Outcomes in General. Specifying appro-priate outcomes of long-term care requires prior knowledge in two related aspects of health: the genera 1 course of he a 1 th status expected of individuals at a given level of disability or age, and the expected response of individuals to treatment directed at a specific disease state. Individual outcome in long-term care set-tings can be measured by major categories of functional status (e.g., physical, psychosocial, and environmental health). These categories are attempts to move away from the earlier acute care criteria of mortality and morbidity rates toward characteristics more relevant to long-term care populations. The Index of Indepen dence in Activities of Daily Living developed by Katz (described earlier) has been used alone and as part of more global assessment instruments to measure and compare changes in patient status over time.52 Denson and Jones53 demonstrated that significant positive changes in AOLs could be achieved by patients during the several months following admission to a long-term care institution. The condition of nursing home residents is thus neither as static nor are their prospects as b 1 eak as has genera 11 y been asst.med. How-ever, measurement which depends solely on ADLs is probably not sensitive enough to use in comparing the relative effectiveness of 52sidney Katz and C. AA1echi Akpom, "Index of ADL," Medical Care, XIV (1976), 116-119. 53oensen and Jones, pp. 126-130.

PAGE 72

53 alternative care Mitchen,54 for example, reported that outcomes in various long-term care settings, measured by a behav-ioral index of health status, were not uniform across settings, but showed differential rates of improvement across sites. She conclud-ed that findings were thus not generalizable beyond the range of the case mix_characteristics of the patients examined. Because of the variation in patient problems, outcome criteria must be specified in relationship to the particular problems which produce dysfunction. For example, degenerative joint disease may cause mobility restrictions which will never show improvement using an ADL scale. Nonetheless, appropriate treatment may produce measureable results in terms of relief of pain, ability to engage in more social activities, and increased well-being. Valid prognostic expectations have been determined for only a limited number of conditions. Katz and Ford55 and Katz et a1.56 looked at medium-range outcomes after stroke and hip fracture, respectively. Further examination of post-stroke outcomes has occurred to evaluate the 5 4Janet B. Mitchell, "Patient Outcomes in Alternative Long-Term Care Settings, .. Medical Care, XVI (1978), 439-452. 55sidney Katz et al., 11Prognosis After Strokes, Part II: LongTerm Course of 159 Patients, .. Medicine, XLV (1966), 236-245. 56sidney Katz et al., "Long-Term Course of 147 Patients with Fracture of the Hip,.. Surqer y , Gynecology, and Obstetrics, CXXIV (1967)' 1219-1230.

PAGE 73

54 effectiveness of various rehabilitation programs57 and to determine longer-range expectations.58 However, such studies are very expensive. Establishing outcomes from a purely empirical base is both difficult and expensive. It requires that 1 arge quantities of data be collected and analyzed. Furthermore, analytic methods must be sensitive to identification of the variables which are policy relevant. Careful longitudinal studies appear necessary to validate multidimensional measures of patient status and to measure the effects of alternative methods of long-term patient care. Such considerations have led to the use of expert opinion as an adjunct to empirical findings in establishing normative outcomes. An example of this is the outcome criteria set deve 1 oped by the Colorado Professional Standards Review Organization in conjunction 57car1 Granger, Clarence Sherwood and David Greer, "Functional Status Measures in a Comprehensive Stroke Care Progran," Archives of Physical Medicine and Rehabilitation, LVIII (December, 1977), 555-561; J.F. Lehman et al., 11Stroke: Does Rehabilitation Affect Outcome?" Archives of Physical Medicine and Rehabilitation, LVI (August, 1975) , 375-382; Osvaldo M1glietta, Tae-Soo Chung,and Vemireddi Rajeswaramma, "Fate of Stroke Patients Transferred to a Long-Term Rehabilitation Hospital," Stroke, VII (January-February, 1976)' 76-77. 58Eugene Moskowitz, Forrest Lightbody, and Nanci Freitag, "Long-Term Follow-Up of the Poststroke Patient," Archives of Physical Medicine and Rehabilitation, LIII (April, 1972), 167-172; M. Ne\'ITlan' "The Process of Recovery After Hemi pl egi a, II Stroke, In ( 1972)' 702-710.

PAGE 74

55 with its pilot study of quality in nursing homes.59 Another study which intends to combine projections with empirical evidence is currently underway.60 Similar outcome criteria, based upon expert judgement, are being developed and tested by Shaughnessy et a1.61 Specifying Appropriate Outcomes for Home Health Care. Only recently have home health care providers begun to develop outcome criteria for evaluation of the quality of home health care. Based on earlier work which focused on process measures of quality and using expert opinion, Daubert62 developed outcome criteria for home care. One such criteria set was developed by a voluntary statewide program in Minnesot"a for 33 different conditions, such as congestive heart failure, depression, and immobility.63 Unfortunately, the vast majority of patients did not meet their outcomes, primarily because their course was atypical due to the presence of exceptions. One group that has specified except i ons is the Pennsylvania Assembly of Home Agencies, which began a statewide quality assurance 59calorado Foundation for Medical Care, The Colorado Experience in PSRO Lon6-Term Care Review (Denver: Colorado Foundat1on for Medical Care, ctober, 1978), pp. A.1-A.25. 60K ane, p. 18. 61shaughnessy et al., Long-Term Care Reimbursement and Regula t ion, pp. VII.17-VII.18. 62Elizabeth A. Daubert, 11A System to Evaluate Home Health Care Services, 11 Nursing Out 1 oak, (March, 1977), 168-171. 63oecker et al., pp. 278-282.

PAGE 75

56 project in 1974.64 Using expert opinion, participants in the project deve 1 oped outcome criteria sets for twe 1 ve audit topics, including arthritis, cardiovascular disease, hip fracture, and mental illness. Each audit topic included up to thirteen essential outcomes by which to assess the care provided by the home health agencies. The outcome criteria used to evaluate care given to over 2,500 patients during the following t'NO years in order to compare outcome quality scores across providers. The Pennsylvania Assembly project included in its outcome criteria standards compli-cations which would preclude the meeting of expected outcomes. Thus far, the criteria sets continue in use in western Pennsylvania for the evaluation of care in on-going home care programs, but they have not been used to evaluate the outcome of care as it relates to the process or cost of care. The Visiting Nursing Association of New Haven has also developed a set of outcome criteria for five types of home health care patients, grouped according to health care problems and rehabilitative potentia1.65 For example, one group includes patients with acute, non-chronic diseases or disabilities who are expected to return to pre-illness functioning. The ultimate objective (outcome criteria) for this group is the complete elimination of the existing 64Rita Berkoben, 11Home Health Care and Quality Assurance: The Experience of the Pennsylvania Assembly Project, .. Quality Review Bulletin 1977), pp. 25-28. 65Elizabeth A. Daubert, 11Patient Classification System and Outcome Critera,11 Nursing Outlook (July, 1979), pp. 451-453.

PAGE 76

57 health problems. Another group of patients for whom outcomes are specified includes those with chronic diseases who, even though a return to pre-illness level of functioning is not possible, have the potential for increasing their level of functioning and may eventually remain in the community without the need for home health services. The most recent activity in the development of outcome criteria for the evaluation of home health care is the ongoing program of the Visiting Nurse Association (VNA) of Omaha, Nebraska described earlier.66 Through the empirical analysis of problan and outcome data for all patients in four agencies in Nebraska, Iowa, Del aware, and Texas over a two year period, the VNA developed specific outcome criteria for each of the problems commonly found in home health care. Outcomes were then c 1 ustered by prob 1 en group to identify those outcomes which were met in the majority of cases over the period of time. At this writing, these outcome criteria are being field tested. Tenuous relationship Between Process and Outcome. Beyond the specification of appropriate outcomes for both institutional and home based care, the second major problem surrounding the use of outcome measures of quality is the relationship between the process and outcome of care. Many factors, in addition to specific treatments and services, have an effect on health status in long-term 66Martin et al., pp. 2-8.

PAGE 77

58 care; major influences include the interaction of multiple diagno-ses, emotional status, environmental adequacy, intervening life events, and attitude toward, knowledge about, and wi 11 i ngness to comply with, the necessary care regimen. For this reason, although the measurement of outcome has obvious face validity as the ultimate measure of effectiveness, the degree to which the outcome in any particular case is actually attributable to the treatment is typic-ally a matter of conjecture. These concerns have led to the continuing debate over the relative advantages of using process, as opposed to outcome measures, in assessing the quality of care,67 and as a basis for corrective actions and policy decisions.68 The issue has not yet been resolved for any modality of care, least of all for long-term care. T'hO recent studies found a curvilinear relationship between process and outcome measures of quality, which indicates the pres-ence of a "threshold" level of process quality below which a poor outcome is much more likely and above which the relationship is less predictab 1 e. 69 67McAuliffe, pp. 118-149. 68Robert H. Brook et al., "Assessing the Quality of Medical Care Using Outcome Measures: An Overview of the Method," Medical Care, XV, Supplement (September, 1977), 6-45. 69Lisa Rubenstein, Susan Mates, and Victor Sidel, "Quality of Care Assessment by Process and Outcome Scoring," Annals of Internal Medicine, LXXXVI (1977), 617-625.

PAGE 78

I . I 59 To confound further the process-outcome association, changes in the health status of patients over time were found to be most significantly related to the initial characteristics of the individ-uals being treated in several studies. In hypertension70 and congestive heart failurell measures such as age and diastolic blood pressure were stronger determinants of patient status at discharge than the actual care provided to patients. Similar results were noted in the evaluation of adult day care and homemaker programs, where both initial case mix (measured by functional status) and utilization of acute care services were more strongly related to improved health status than was use of the Consistent findings emerged from a study of home health services in which younger, less. disabled, medically complicated patients 7 Fred T. Nobrega et al., 11Quality Assessment in Hypertension: Analysis of Process and Outcome Methods, .. New England Journal of Medicine, LLXLVI (January 20, 1977), 91-113; S.W. Fletcher et al., 11Pred1cting Blood Pressure Control in Hypertensive Patients: An Approach to Quality of Care Assessment, .. Medical Care, XXVII (March, 1979), 285-292. 71F.J. Rorrm et al., 11Correlates of Outcomes in Patients with Congestive Heart Failure, .. Medical Care, XIV (1976), 765. 72Thomas Wan, William Weissert, and Barbara Livieratos, 11Geriatric Day Care and Homemaker Services: An Experimental Study, .. Journal of Gerontolooy, XXXV (March, 1980), 256-274.

PAGE 79

60 experienced more favorable changes in health status, irrespective of utilization )3 These studies make a strong argument that attempts to prognose the course of long-term care should take into account the initial characteristics and overall individual potential in setting outcome objectives. Changes in physical impairment level are sometimes meaningfully measured in terms of ambulation. For patients for whom ambulation was not possible, a change in the muscle strength of an affected limb may be considered a desirable outcome. Eliminating depression may be an appropriate index of progress for some patients, but may be irrelevant for those who have been depressed their entire 1 ives. In SllTlmary, each patient possesses a unique combination of disabilities, accompanying medical complications, specific psychosocial and demographic characteristics. Measures used to categori ze patients with respect to the effects of treatment should incorporate this uniqueness. Cost of Care This section Sllllmarizes a number of major studies on the cost of institutional and non-institutional long-term care. Because nursing home cost studies have served as the conceptual basis for cost studies in home care and are more nLmerous, the beginning of 73Sidney Katz et al., Effects of Continued Care: a Study of Chronic Illness in the Home, Publicat1on No. (HSM) 73-3010 (Washington, D.C.: National Center for Health Services Research, DHEW, 1972)' pp. 23-42. -

PAGE 80

61 this section primarily discusses nursing home cost studies. The few home health care cost studies are then presented. The final portion of this section discusses studies which compare expanded home health care, traditional home health care, and institutional care in terms of cost and/or effectiveness. Nursing Home Cost Studies The , majority of existing studies of the cost of nursing home care have concentrated on identifying the various factors influenc ing the cost of that care. use cross-sectional multivariate analytic techniques in order to estimate cost functions. They generally use one of several measures of cost as the dependent variable and a variety of community, facility, and pat i ent characteristics as independent variables.74 Several studies have also tried to include measures, however minimal, of quality of care. The discussion which follows presents the general analytic approach used in these studies and summarizes their findings. Facility Level Cost Studies. nurs ing home cost studies use the facility as the unit of analysis and the average cost per patient day as the dependent variable. Because there is wide variation in property cost per day across nursing homes, generally due to differences in accounting practices, most studies have focused on operating costs. (Two exceptions were the studies by 74For a review of the use of cost function analyses for nursing home care see, Martin Knapp, "Cost Functions for Care Services for the Elderly," Gerontologist, XVIII (1978), 30-35.

PAGE 81

62 Skinner and Yett75 and McConne 176 which estimated cost functions that used average total cost as the dependent variable.) Although the s-eparate components of operating cost are sometimes analyzed, researchers tend to concentrate on total operating costs. Operating cost components most often used include nursing costs,77 laundry and linen, and dietary.78 Cost functions are the enpirical representation of the rela-tionship between the cost of production of services and the level of output, obtained from a multiple regression of total costs (per day) on outputs and other important input influences. The output (of the firm) is assumed to be the patient day of nursing home care and is viewed as varying in relation to the case mix and the quality of care provided. Other factors are then included in the cost function to reflect differences in input prices (e.g., nursing wage costs), efficiency (e.g., occupancy rates), economies of scale (e.g., bed size), and management objectives (e.g., for-profit or non-profit 75skinner and Yett, p. 79. 76charles ftt:Connel, "Cost, Size and Quality Structure of Nursing Home Industry," Journal of Health and Human Resources Administration (November, 1978), pp. 134-149. 77Ruchlin and Levey, pp. 3-15; Moreland Act Commission on Nursing Homes and Residential Facilities, Long-Term Care Regulation: Past Lapses, Future Prospects (New York: Morel and Act Comm1ss1on on Nursing Homes and Residential Facilities, 1976), pp. 22-27. 13-24; Howard Birnbaum et al., Home Care: Developmental Cost Canbr1dge, MA: Abt Assoc1ates,

PAGE 82

63 ownership and control). Thus, the factors used in most cost studies can be classified into four general groups: environmental/community factors; facility characteristics; patient characteristics (case mix); and measures of the quality of care. 1) Environmental and Community Factors. Environmental and community factors include measures of location and population density. For example, Ries and Christianson79 used location in a town with population greater than 2,000 in their study of Montana nursing homes, while Shaughnessy et a1.BO and Birnbaum et a1.81 used location in an SMSA. Input prices (e.g., wages) were used in several studies,B2 as were regulatory characteristics, such as the operation of a certificate of need program.83 The study by Birnbaum et al. was especially comprehensive in its approach to environmen tal/community factors, also using the presence of life safety codes, minimum staffing requirements, and prospective reimbursement systems in its cost functions.84 The significance of factors other than input price in this category varies considerably throughout the studies cited. 79Ries and Christianson, pp. 20-21. BOshaughnessy et a 1., Long-Term Care Reimbursement and Regulation, p. VIII.10. 81Birnbaum et al ., p. 90. 82Moreland Act Commission, p. 24; Birnbaum et al., p. 90. 83Birnbaum et al., p. 90. 84Birnbaum et al., pp. 39-40, 90-91. \

PAGE 83

64 2) Facility Factors. In terms of facility characteristics, most studies included indicators of facility size, such as the number of beds,85 average daily census86 and occupancy rate.87 Other facility characteristics investigated have included financial status88 and ownership,89 (e.g., for profit and non-profit). Studies have typically found an association of non-profit ownership with higher cost, controlling for other factors. Results of the other facility characteristics, including size and occupancy rate, are inconsistent across the studies. 3) Case Mix Factors. One of the more difficult groups of factors to obtain for cost function studies has been adequate measures of patient characteristics. Thus, case mix measures used in cost function studies have included both direct and indirect (proxy) measures. Examples of proxy measures used include the 85Ries and Christianson, pp. 21-22; BirnbaLrn et al., More 1 and Act Commission, p. 24; Thomas J. Wa 1 sh and Koetting, "Patient Related Reimbursement for Long-Term (n.p.: Illinois Department of Public Health, 1978), ( Mimeographed . ) 86McConnel, pp. 134-149; Birnbaum et al., p. 95. p. 93; Mi chae 1 Care," p. 9. 87 Shaughnessy et a 1., Long-Term Care Reimbursement and Regulation, _ pp. IX.16-IX,18; Birnbai.JTl, p. 97; Ries and Christianson, p. 21; Ruchlin and Levey, pp. 3-15. 88Moreland Act Commission, p. 24. 89Ries and Christianson, p. 21; Ruchlin, pp. 3-15; Walsh and Koetting, p. 78

PAGE 84

65 average patient age,90 the level of care classification (skilled or intermediate) ,91 the ratio of skilled nursing care beds to total beds,92 and the percentage of non-anbulatory patients.93 These measures have generally led to inconsistent results. Several studies have used case mix indicators which directly measure patient care needs. These measures include the patients' functional abilities94 (measured by an AOL Index), medical diagnoses,95 mental status,96 and long-term care problems.97 Findings with respect to functional status were consistent across studies and suggested that it was strongly related to the cost of the nursing home care, with cost generally being highest for those patients in the intermediate range of dependency. Inconsistent results were obtained for other direct measures of case mix, except for medical diagnosis, which was generally not strongly related to cost. 90Moreland Act Commission, p. 23; McConnel, p. 143. 9lwalsh and Koetting, p. 9; Birnbaum et al., p. 101. 92Ries and Christianson, p. 21. 93Birnbaum et al., p. 93. 94skinner and Yett, pp. 69-75; Birnbaum et al., pp 87-88; Shaughnessy et al., Long-Term Care Reimbursement and Regulation, p. III.l8; Walsh and Koett1ng, pp. [16-18]. 95Birnbaum et al., p. 87. 9 6sirnbaum et al., p. 88; Shaughnessy et al., Long-Term Care Reimbursement and Regulation, p. III.18. 97shaughnessy et al., Long-Term Care Reimbursement and Regulation, p. III.18.

PAGE 85

66 4) Quality of Care Factors. Quality measures are the final group of variables included in cost analyses and have been the most difficult to measure. As might be the most common approach has been to use structural such as the availability of therapeutic services98 or staff-to-patient ratios.99 The proportion of Medicaid patients or patient days was also included as a quality under the assumption that the greater this pro the lower the quality of care.100 The findings of studies using measures of Medicaid utilization substantiate this assumption. However, whether this reflects lower quality or simply greater cost containment pressures is yet unclear. Few nursing home cost function studies have used process or outcome measures of quality. There are at least two reasons for this amiss ion. First, process measures of qua 1 i ty generally require extensive primary data collection efforts which are usually beyond the scope of (economic) cost function analyses. outcome measures of quality of care require that measures be taken over more than one point in time; this is also beyond the scope of crosssectional cost function analyses (and their major defect). there is only one known study, by Shaughnessy et 98wa1sh and pp. [16-18]; Birnbaum et al., pp. 89-90. 99McConnel, pp. 142-145. 100McConnel, pp. 143-145.

PAGE 86

67 al. ,101 that used either of these types of quality measures. First year findings of this study indicate a significant and positive relationship between process quality and the cost of nursing home care. In the second year of this study, outcome measures of quality will be added to the analysis. Cost function studies have thus found that case mix and quality are related to the cost of nursing home care, but methodological and measurement weaknesses of many of these studies suggest that concepts and approaches must be refined before definitive findings will emerge. For example, none of the studies cited above used a comprehensive approach to measurement of case mix and quality of care, while contra 11 i ng for exogenous factors such as community size or regulatory environment. Two studies included no case mix factors,102 and several included no quality measures.l03 Other studies were weak analytically,l04 and all were cross-sectional, rather than longitudinal. The results were insufficient to address the questions of the relationship between quality of care (effectiveness) and cost, for various types of patients. 101shaughnessy et al., Long-Term Care Reimbursement and Regulation, p. II 1.21. 102Ries and Christianson, p. 21; Ruchlin and Levey, pp. 3-15. 103Moreland Act Commission, pp. 23-25; Skinner and Yett, pp. 69-89; Ruchlin and Levey, pp. 3-15. 104McConnel, pp. 134-149.

PAGE 87

68 Patient Level Cost Studies. Very few nursing home cost studies have been concerned with cost per day at the patient level. Examples of these few are Vannell et al.105 and McCaffree et al.106 All found significant time (and, by implication, cost) differences across various case mix categories. One reason so few nursing home cost studies have been completed at the patient level is that total cost data are available usually only at the facility level. Thus, patient level analyses require direct observation of staff time required for various patient care activities (e.g., time and motion studies). Several studies completed as part of the development of the Medicaid reimbursement systems in Illinois, Ohio, and West Virginia have added to estimates of time, weighting factors reflecting intangibles and the various skill levels of providers, calculating cost by multiplying the weighted time figures by a prevailing wage rate.107 None of these studies used the resulting cost per patient day in a full scale cost function analysis. Thus, these studies strongly suggest the need for additional analyses at the patient 1 eve 1 . 105oavid Vannell et al., Final Report, The Patient Evaluation Review Corrmittee Project 1977-1979, Section I (n.p.: Wisconsin Department of Health and Soc1al Services, August, 1979). 106Kenneth McCaffree et al., Cost Data Reporting System for Home Care: Final Report, Publ1cat1on No. (HRA) 77-3169 (Rockvllle, MD: National Center for Health Services Research, DHEW, 1977). 107oonald St. John, 11The Illinois Automated Long-Term. Care System-Three Years of Experience,11 Medical Care, XIV, Supplement (May, 1976), 192-197.

PAGE 88

69 Home Health Cost Studies No home health cost studies to date have analyzed cost at the provider level. Contrary to the situation for nursing home costs, the few studies of home health cost that do exist are at the patient level . Two studies in particular have dealt with the cost of home health care independent of comparisons to nursing home care costs. Both focused primarily on case mix, facility/provider, and environmental characteristics in relationship to cost, omitting any meas ures of quality of care. The first looked at unit costs as part of an evaluation of the federal Home Health Grant Progran. Part of the study included the analysis of factors affecting the cost per skill-ed nursing care visit in a small sample of grantee and control group agencies. In that study the major factor found to be related to cost per visit was input price, as measured by nursing salary. How-ever, other factors, such as case were not taken into account in the analysis.l08 The second project that studied the cost of home health care was the recent experiment completed in New York City (described earlier in this report).In that study the relatianship of the total cost of home care to the disability level of the patient was analyzed. Cost included expenditures for a wide array of in-home 108Robert Sch 1 enker et a 1 . , App 1 i ed Research in Home Health Services Volume I: Grant Program Evaluation (Denver: Center for Health Serv1ces Research, Onwers1ty of Colorado Health Sc.iences 1979), pp. VII.23-VII.27.

PAGE 89

70 services, including homemaker and chore services, in addition to an estimate of living expenses (i.e., room, board and transporta-tion). Findings, based solely on univariate statistical analyses, indicated that the most highly disabled group of patients, which comprised 10% of the total study population, incurred almost 47% of the costs of the entire project. There was generally a significant difference in the cost of care by functional status; per diem cost for the least disabled patient was $11 while the per diem cost for the most disabled was $20.109 Whether the relationship between cost and functional status is linear is still uncertain since other studies found that costs for the most disabled patients were lower than costs for those slightly more functiona1.110 One unusual study was recently completed at the Levinson Policy Institute. In that study, the estimated cost of home health care (including hypothetical direct progrilll expenditures and costs of living costs) for 50 patients was analyzed. Factors which signifi-cantly increased cost were the functional dependency and the age of the patient at admission to home care. On the other hand, the greater the number of individuals residing at home with the patient, the lower the cost. No other case mix or environmental factors were significant in the cost functions. No quality of care measures were 109widmer, Brill, and Schlosser, p. 490. llOA.L. Creese and R. Fielden, "Hospital or Home Care for the Severely Disabled: A Cost Comparison," British Journal of Preventive and Social Medicine, XXXI (1977), 116.

PAGE 90

71 included in this study because patients did not actually receive home health services.111 No doubt, there remains significant room for methodological improvement of home health cost studies. None of these studies reaches the level of methodological sophistication of the nursing home cost studies cited previously. Researchers interested in home health care cost analysis would do well to build upon the strengths of these earlier studies (and take account of added suggestions in the last chapter of this volume). Cost-Effectiveness of Long-Term Care Cost-effectiveness studies are an attempt to relate the cost of a program or service to the outcomes, or effectiveness, of that program or service. One major component of any cost-effectiveness study is the measurement of effectiveness. Another is the determination of the relationship between effectiveness, however measured, and the services whose impact is being assessed. (That is, it must be determined that the observed outcomes are the result of the services provided and the costs incurred for them) . The fi na 1 component of significance is the appropriate definition of cost. The adequate completion of a cost-effectiveness analysis depends upon the ability of the analyst and program administrators to determine measureable program objectives (e.g., a positive change in health status or a reduction in length of stay), alternative ways 111sager, pp. 241-248.

PAGE 91

72 of attaining those objectives (comparison of alternative programs), and economic constraints on the program (the value of resources available). These requirements suggest that alternatives compared should be appropriate, that the dimensions of costs and outcomes be specified, the period of observation be meaningful, and that the appropriate analytic technique be employed. None of these areas has been comprehensively addressed in any single study thus far. There-fore a variety of work is discussed in this section. Several cost-effectiveness approaches have been used in compar ing various forms of long-term care.ll2 Studies in this area have focused on comparisons at two levels: the cost-effectiveness of home health care compared to nursing home care; and the costs and benefits of enriched home service programs as opposed to more conventional and limited home care.ll3 These are discussed in the following paragraphs. Cost-Effectiveness of Home Health Care Versus Nursing Home Care Most of the available studies of home health care versus 112For a general discussion of cost-effectiveness in the long term care field see, Neville Doherty, Joan Segal, and Barbara Hicks, 11Alternatives to Institutionalization for the Aged: Viability and Cost-Effectiveness, .. Aged Care and Services Review, I (January/ February, 1978), 8-14; John Hammond, "Rome Health Care Cost Effectiveness: An Overview of the Literature, .. Public Health Reports, XLIV (July/ August, 1979), 305-312; Nevi 11 e Doherty and Barbara Hicks, 11Cost-Effectiveness Analysis and Alternative Health Care Programs for the Elderly, .. Health Services Research, XII (1977) 190-203. 113No known studies have compared among different types of nursing home care.

PAGE 92

73 nursing home care, compare the actual cost of home health care to the hypothetical cost of nursing home care. All assume that the quality of care (effectiveness) is equal among the settings. Few include the total cost of living in the community (e.g., room, board transportation) in the cost of home health care. Only a few of the more recent studies compare total public costs across the care modalities, including expenditures for Old Age Pension (O.A.S.D.I.), Supplemental Security Income, etc. No study to date has compared the actual cost of care for cohorts of home health and nursing home patients over (a significant of) time to the outcome of that care. Thus, weaknesses in existing methodology make it difficult to draw definitive conclusidns about the cost-effec-tiveness of any form of long-term care. The findings of early studies addressing the first question (of traditional home care versus nursing home care) were conflicting due to major methodological deficiencies. A review of twenty-five projects begun before 1975114 identified numerous problems with previous studies: noncomparability of experimental and control groups; lack of significant differences between experimental and control treatment variables; wide variation in (and lack of control for) the needs of individual members of the study populations; variation in the scope and extent of treatments from one study to another; 114sonia Conly, Critical Review of Research on Lon -Term Care Alternatives ([Washington, D.C. : Office of the Assistant Secretary for Planning and Evaluation, DHEW, June, 1977), pp. 4-16. (Mimeographed.)

PAGE 93

74 differences in the methods used to collect cost information and the scope and kinds of costs included; and noncomparability of costs considered for the experimental group compared to the control group. The review confirmed problems cited in other evaluations of long term care115 and concluded that evidence at that time was insuffi-cient to either refute or support arguments that either enriched or traditional home health care alternatives were more cost-effective than institutional long-term care. Effectiveness of Home Health Care In Reducing Institutionaliza-tion. Several small studies in recent years have focused on the question of the effectiveness of home health care in preventing institutionalization. However, results of these efforts are incon-elusive. Some studies concluded that home care prevents institutional-ization. Using the nllllber of home health care admissions per 1,000 Medicare beneficiaries as his dependent variable, Dunlop found that increases in home health care utilization were associated with decreases in the utilization of nursing homes (on an area-wide 115Marie Callender and Judith LaVor, Home Health Care: Development, Problems and Potential (Washington, D.C.: Office of Soc1al Serv1ces and Human Development, Office of the Assistant Secretary for Planning and Evaluation, DHEW, April, 1975). (Mimeographed.); Applied Management Sciences, Evaluation of Personal Care Organizations and Otl1er In-Home Alternatives to Nursing Home Care for the Elderly and Long-Term 01sabled, F1na! Report and Execubve Summary (Revised) (Silver Spring, MD: Applied Management Sci .ences, May, 1976).

PAGE 94

75 basis) .116 In another study of 245 patients in a New York City program for the homebound aged, Brickner et al.ll7 reported that after 24 months, 23 patients improved to the extent that they were no longer homebound, 116 remained stabilized under the program's continuing care, and 40 patients were institutionalized (either in hospitals or nursing homes). Relying solely on clinical judgment, Brickner et al. estimated that 85 of the patients would have required institutional care and 25 would have died without the program. A recent Canadian study concluded on a similar basis. that a significant number of study patients avoided nursing home utilization.118 Other studies have found inconsistent relationships between use of home care and institutionalization. A one-year study by the Benjamin Rose Institute in Cleveland of 50 elderly patients receiv-ing home health aide services after hospitalization indicated that the group receiving services had significantly fewer days in and fewer admissions to long-term care institutions than a control group. The service group patients also appeared to be significantly 117Philip w. Brickner et al., 11The Home Bound Aged: A Medically Unreached Group,11 Annals of Internal Medicine, LXXXI I (January, 1975) , 1-3. 118A.S. Kraus and M.I. Armstrong, 11Effect of Chronic Home Care on Admission to Institutions Providing Long-Term Care,11 Cal")adian Medical Association Journal, CXVII (October, 1977), 747-749.

PAGE 95

76 more contented.119 On the other hand, Bryant et a1.120 reported on a small study of home care provided to stroke victims who had been discharged from the hospita 1. This post study compared a group that had received home care and a mixed group which received either only physical therapy or no care. After a nine-month follow-up, two home care and eight control group patients were living at home. However, close examination of the study indicates that the experi-menta 1 and contra 1 groups were prob ab 1 y not we 11 matched according to severity of illness. The underlying assllTlption of these studies is that if home health care can prevent (or postpone) institutionalization, it will, by definition, be cost-effective when compared to nursing home care. Unfortunately none of these studies investigated the validity of this assumption. One study which did try to test the asst.mption was the recent experimental study of homemaker services by Weissert, Wan, and Livieratos.121 They concluded that the study group did not have significantly lower utilization of hospital or nursing home services than the control group (which received the usual array of services). Not surprisingly, total costs of the study group were actually higher than for the controls. 119Margaret Neilson et al., 11A Controlled Study of Home Health Aid Services,11 ,American Journal of Public Health, CLXII (1972), 1094-1101. 120Nancy Bryant, Louise Candland, and Regina Loewenstein, 11Com pari son of Care and Cost Outcomes for Stroke Patients, With and Without Home Care,11 Stroke, V (1974), 54-59. 121weissert, Wan, and Livieratos, pp. 21-26.

PAGE 96

77 Cost of Home Health Care Versus Nursing Home Care. Few studies are available which empirically support the general proposition that home care is less costly than nursing home care. Indeed, the most widely cited home health care studies concerning cost savings are of short-term acutely ill patients.122 Most studies of home care which are purported to demonstrate cost savings from home care utilization suffer from severe methodological weaknesses. They either merely estimate savings from home care compared to hypothetical nursing home care, or they fail to measure costs in a comprehensive manner and assume that the effect of long-term care is the unit of output (e.g., cost per visit, patient day, month). Many studies merely estimate the cost savings from community based experimental programs. Brickner and Scharer123 in their study of a progran for the homebound in New York City, claimed considerable cost savings, but their claims were based solely on physician estimates of 11probable11 institutionalization and 11probable11 program 122Joseph R. Stone, Elizabeth Patterson, and Leon Felson, 11Effectiveness of Home Care for General Hospital Patients, .. Journal of the M!erican Medical Association, CCV (July 15, 1968), 145-148; Lowell Gerson and Owen Hughes, "A Comparative Study of the Economics of Home Care, .. International Journal of Health Services, VI (1976), 543-555; Creese and Fielden, pp. 116-121. 123Philip Brickner and Linda Keen Scharer, 11Hospital Provides Home Care for Elderly at One-Half Nursing Home Cost, .. ForLITl (November/December, 1977), no page numbers.

PAGE 97

78 costs. There are at least four other major studies which estimate cost savings in the same manner.124 In a more sophisticated cost study, Greenberg disaggregated a Minnesota target population into four disability levels and two living arrangements (living alone and living with others) on the assumption that home care costs \'Kluld vary on these dimensions. Costs were taken as total costs, including room and board. For only the worst disability level was home care as expensive or more expen-sive than nursing home care. All other levels of disability and living arrangements were cheaper with home care. From this, Greenberg estimated that 9% of the 1974 Minnesota skilled nursing facil-ity patient population could be cared for at home, resulting in substantial cost savings.125 In studies that have looked at the population already in nursing homes, cost savings from hypothetical home health care have not been found. One study in Durham, North Carolina by Burton et a1.126 estimated that for approximately 87% of the patients in nursing homes, the only suitable alternatives were not economically 124Avery Colt et al., 11Home Health Care is Good Economics, .. Nursing Outlook (October, 1977), pp. 632-636; Anthony Amado, Beatr1ce Cox, and Rich Mileo, 11Cost of Terminal Care: Home Hospice vs. Hospital, .. Nursin' Outlook (August, 1979), pp. 522-526; Kraus and Armstrong, pp. 74 -749; Gerald M. Edgert and Joyce E. Bowlyow, 11Preliminary Findings: Monroe County's Access Project to Prevent Unneeded Nursing Home Admissions, .. Perspectives on Medicaid and Medicare Management (Septenber, 1979), pp. 5-13. 125Greenberg, 11The Costs of In-Home Services, .. p. 45. 126surton et al., pp. 3-12.

PAGE 98

79 feasible, costing approximately four times more than nursing home care. For other 13%, alternatives outside the nursing home were feasible but with no great reduction in costs. The Levinson study cited earlier compared the actual costs of nursing home care to the costs of hypothetical home health care plans for a group of 50 patients awaiting hospital discharge. This study found that institutional care was less costly than home care for 80% of the patients, with an estimated cost of room and board of $60 per week included in the cost of home care.l27 Only recently have actual costs of home care and institutional care been compared. One General Accounting Office (GAO) study found that home care services were less costly only if the patients were not severely disabled, and were more costly in cases of disability where extensive services were needed. In that study, the GAO compared the cost of providing home health care to the elderly, including the value of services provided by fami1y and friends, to the cost of nursing home care. When comparing the functional and health status of elderly persons residing in the community with those institutionalized, the study found that 87% of the institutionalized older persons were greatly or extremely impaired, com-pared to 13% of those at home. Overc .ll, half of the services received by the elderly were provided by family and friends; over 70% of the services received by the greatly or extremely impaired came from that same source. At all levels of impairment, the value 127sager, p. 247A.

PAGE 99

80 of services (computed at the same rate as purchased services) pro vided by the i nforma 1 soc i a 1 support sys terns great 1 y exceeded the cost of services provided at pub.lic expense. (Needless to say, the pricing of services rendered by family or friends is subject to a wide range of uncertainty.) At the greatly impaired 1 eve 1, where the break-even point between the cost of home care and institutional care was reached, family and friends provided about $287 per month in services for every $120 spent by the public agencies.128 Unfortunately, no comparison of the quality of care (effectiveness) was made across the care modalities. Another study that compared the actual cost of home care to institutional care was completed by the Stanford Research Insti-tute. Although simi 1 ar to the GAO study in that it did not i nves-tigate the effectiveness of care, it suffered from more severe methodological weaknesses because of its limited conceptualization of cost. Like many other studies, it compared the average direct program cost per moryth of home health to nursing home care, omitting the costs of maintaining the patient in his home as part of the cost of in-home services. Thus, it was not surprising that the study 128comptroller General of the U.S., General Accounting Office, Polic to Better Provide for ash1ngton, . . . Govern ( Comptroller General the

PAGE 100

. I 81 concluded that home health care was cost-effective compared to nursing home care.129 One important study of the relative cost-effectiveness of nursing home versus home care is currently being conducted in Minnesota. In that study the total costs of services and 1 iving needs for a group of 367 home health patients and 350 nursing home patients will be compared in terms of outcomes (measured by changes in functional status, satisfaction, and social contacts). In addition, the study will compare total costs using a cost-finding method which tries to develop consistent allocation procedures at the level of the cost center for both institutional and non-institu-tional long-term care. Although a significant improvement over previous efforts, this study treats each modality as if it were a generic product (e.g., no consideration is given to organizational and management issues that may affect the cost and outcome of care) .130 Thus, no truly comprehensive long-run studies of long term care cost-effectiveness have been conducted to date. Cost-Effectiveness of Coordinated and Expanded Home Health Care A recent key federal policy initiative to address the cost-effectiveness of an expanded home health service system has been the 129Neill Piland, Feasibility and Cost-Effectiveness of Alternative Long-Term Care Executwe Summary, (Menlo Park, CA: Stanford Research Inst1tute Internabonal, May, 1978), pp .. 3-6. 130Anderson, A Compariso-n of In-Home Care, pp. 2ff.

PAGE 101

82 funding of sever a 1 dernonstrat ion projects .131 Each of these projects implements both an intervention strategy and a methodology by which to evaluate the effect of that intervention. Three character-istics common to these demonstrations include centralized case man-agement; expansion of benefits and services; and development of appropriate service plans for individual patients. The overall conclusion of these projects is that most people needing long-term care need maintenance rather than medical care.l32 Beyond this, no other consistent conclusions have been drawn from these demonstrations. In part, this was because the pro-jects did not use comparable measures of cost or patient outcomes, nor did they operate for common periods of time. Effectiveness Findings. On the issue of effectiveness, some projects reported that their respective interventions were more effective than the present system in maintaining the functional status and satisfaction of their clients, but other projects found 131These include the Colorado Community Care Organization for the Aged and Disabled; Home Care: An Alternative to Institutionalization (Massachusetts); Monroe County (New York) Long-Term Care Program; A Model Service Delivery System for the Aging (Pennsylvania); Day Care for the Elderly (Montefiore Hospital, New York); Cost-Effective Alternatives to Nursing Home Utilization (Georgia); Wisconsin Community Care Organization; Effects and Costs of Day Care and Homemaker Services for the Chronically Ill (National Center for Health Services Research); Triage: Coordinated Delivery of Services for the Elderly (Connecticut).

PAGE 102

83 the opposite to be true. One day care project found no difference between the experimental and control groups in terms of their functional status or satisfaction at discharge. The participants in another day care experiment exhibited dissatisfaction, with over 20% of the experimental group leaving the project.133 However, one project site reported that its clients• quality of life increased.l34 The National Center for Health Services Research {NCHSR) study showed that both homemaker and day care programs he 1 ped to improve the functioning, or slowed the rate of deterioration of the elderly. Both the day care and homemaker programs resulted in lower death rates. In addition, participants in both programs had higher levels of contentment, mental functioning, and social activity than did controls.135 Another study that is testing the impact of community-based services for the elderly appears to corraborate the NCHSR findings of lower mortality in the group receiving services.136 In this con-tinuing four year demonstration project administered by Georgia•s Medicaid program, several types of alternative services are provided 133weissert, Wan, and Livieratos, p . 6. 134Fredrick Seidl, Kevin Mahoney, and Carol D. Austin, 11Providing and Evaluating Home Care: Issues of Targetting11 (paper presented at the meeting of the Gerontological Society, Dallas, November 20, 1978). (Mimeographed.) 135weissert, Wan, and Livieratos, pp. 6-12. 136F. Albert Skellie et al., 11Community-Based Long-Term Care and Mort a 1 ity: Impact One Year After Enro 11 ment,.. (paper presented at the meeting of the Prnerican Public Health Association, New York, Novenber, 1979). (Mimeographed.)

PAGE 103

84 to Medicaid beneficiaries certified eligible for nursing home placement. Differing from previous studies, early reports indicate that utilization of experimental services appears to have had the greatest impact on mortality rates of all variables studied. In particular, home health services were especially significant since the experimental group (which received these services) exhibited a death rate of six percent, compared to the control group of 27%. Cost Findings. Some projects demonstrated a cost saving from experimental programs, while others did not. Using direct program expenditures only, the r.tlnroe County Long-Term Care Program found that their clients could be kept at home for one-half the cost of placing them in an institution.l37 When looking at total (program and living) costs, the Wisconsin Community Care Organization found that it cost over $2.22 more per day to maintain a client at home than to place them in a comparable level of institutional care.l38 Cost-Effectiveness Findings. Findings of the most . recent analysis of the NCHSR projects comparing homemaker to day care services do not support the argument that coordinated communitybased programs are cost-effective)39 Although day care services did appear to reduce the rate of institutionalization in the study population, the cost savings were more than offset by the cost of 137Edgert and Bowlyow, pp. 5-13. 138seidl, Mahoney, and Austin, pp. 6-12. 139weissert, Wan, and Livieratos, p. 3ff.

PAGE 104

85 the program. In the population using homemaker services, rates of institutionalizaiton were, if anything, higher than in the control population. In any case, it does not appear that the addition of new outcome measures to the studies wi 11 improve the consistency of their findings. The demonstration projects yield no consensus as to cost-effectiveness. Methodological Weaknesses. Greenberg, Doth, and Johnson identified major methodological problems which reoccur in each of these demonstration projects with varying degrees of intensity and frequency. THey incJuded the following: 1) Lack of conceptual framework; 2) Unsound sampling design; 3) Inadequate data collection; and 4) Inappropriate project 1 ife.140 In addition, the demonstrations lack comprehensive cost measurements and analyses of the organization and management of the service delivery system. Not all of the problems can be attibuted to all of the projects, but the presence of one of two of then is serious enough to question the reliability and validity of the findings and conclusions. Each of these weaknesses is discussed in this section. 1) Lack of Conceptual Framework. An adequate conceptual dis-cussion of the long-term care system was missing in most of these projects. The experimental services were seldom justified on the 140Greenberg, Doth, and Johnson, pp. C.24-C.39.

PAGE 105

86 basis of clear relationships with some desired outcome; it. was implicitly assLmed that the services would be cost-effective. As a result, there was no way to know whether the outcome was produced by the experimental services or by the interaction of program components. Normally, the conceptual model should generate the specific research questions of the project. Without such a model to identify the important questions and provide a framework from which to evaluate findings, evaluation of the intervention was difficult. 2) Unsound Sampling Design. f1bst of the projects had small sample sizes. Requirements for minimun sample sizes necessary for statistically significant findings do not appeaer to have been considered in designing the projects. Consequently, the validity of their conclusions is questionable on sampling grounds. In addition, the selection of the study population was often not systematic. Many projects included anyone who walkea in the door until the nt.mber of desired clients was reached. Thus, considerable self-selection occurred; thereby threatening the validity of the findings. When a non-random sample is added to a small sample size, there is no way of knowing whether the intervention actually made the difference or whether it was due to the particular characteristics of a few individuals. These problems also affect the validity of the control group. Thus meaningful analysis of the intervention becomes very difficult.

PAGE 106

87 3) Inadequate Data Collection. The data collection process for most of the projects lacked consistency and comparability. There was a lack of documentation regarding what data was collected, how it was collected and analyzed, and whether it was valid and reliable. Few projects had systematic procedures for the collection and storage of data, or for its periodic analysis. Only one project planned for data retrieval so that time series analysis could be carried out.l41 Without good data and an adequate means for storing and using it, the evaluations were severely constrained. 4) Inappropriate Project Life. The actual project life (or amount of time in which the project was fully operable) was often short. With the majority of project grants being for three years, by the time they got started, admitted sufficient numbers of patients and corrected imp 1 ementat ion problems, the project often had one year or less to run . Therefore many of the sites had only one year of data with which to evaluate the project, which itself may have been working well only its final few months. Furthermore, it is reasonable to assi.ITle that the effects of long-term care intervention will not surface until several years after the actual intervention. With only one year of data to evaluate the treatment effect, the results were, at best, specu-1 at ive. 14lstatement by Michael March, Director of the evaluation ofthe Colorado Community Care Organization, Denver, November, 1979.

PAGE 107

88 5) Lack of Comprehensive Cost Data. Most projects only compared direct program expenditures (e.g., Medicare and/or Medicaid), in spite of previous studies indicate that substantial contributions to living expenses were made by cash benefit programs, patients, families and friends.l42 One project assumed that all living expenses and use of medical services \'Kluld be the same for the experimental and control groups. Another tried to collect information on out-of-pocket expenditures but gave up when clients would not complete their diaries. With comparisons limited to direct programs expenditures, it is difficult to discern the real cost-effectiveness of the demonstration projects or the long-term care system. In part, this is due to the composition of the system (i.e., nutrition transportation, soc ial services, cash benefits, and multiple medical care programs) . Limited comparisons run the risk of substantially under-or over-esti mating the cost implications of program and benefit expansions. 6) Lack of Organizational Analyses. Only one of the projects included an in-depth analysis of the service delivery system as part of its research design. This shortcoming of the other projects is unfortunate since the impact of the availability of a multiservice, well-organized provider system on the outcome of care seems potentially significant. Aggregate national cost and utilization data suggest that all home-health care is not the same. For example, analysis of the availability of manpower, the degree of control over 142GAO, p. 6

PAGE 108

89 case monitoring, or the willingness of physicians to prescribe treatment are all important to the efficiency of a project. The treatment of long-term care demonstration projects as generic pro reduced the probability of discovering the unique nature of any one project that was linked to its success. Therefore, it becomes impossible to duplicate successful programs and avoid failures. 7) One Exception. One exception cited earlier which deserves further recognition is the Georgia Alternative Health Services Project. Under a Medic.aid waiver, the state of Georgia began offering three types of community-base services (adult day rehabilitation; home health care, including personal care services and meals; and alternative living services, such as adult foster care, boarding homes and congregate living) to all Medicaid clients over the age of 50 who would otherwise have been eligible for nursing home placement. The project design called for one-quarter of all clients judged appropriate for community-based services to be randomly assigned to a control group (who are eligible for traditional Medicaid/Medicare services). Determination of eligibility is based on a multi dimensional assessment of patients physical, mental, and social status, including a mental status questionnaire, a morale index, activities of daily living score, and the instrLITlental activities of daily living index. Each client is assessed every six months using the complete array of assessment tools in order to develop outcome measures. Additional outcome information recorded as it becomes

PAGE 109

90 available are dates of institutionalization and death.14 3 Preliminary analysis of the impact of the project on 267 clients indicates that, although the experimental and control groups were not significantly different on key base 1 i ne measures, project services appeared to have decreased mortality rates by two-thirds for clients using the experimental services compared to those having access to traditional Medicaid services. In the first six months of the project, six percent of the service group patients died, while 19% of the control group died. Mortality rates in the second sixmonth period were not significantly different between the groups. Overall, during the one year period after the start of the program, 15% of the experimental group died compared to 29% of the control group.l4 4 These findings suggest there may be a stronger influence on mortality rates in the early period of a community-based service progran, tapering off as length of participation in the program increases. Whether the impact on mortality (and life expectancy) will remain significant over time remains to be seen. In any case, the necessary evaluation components have been included in the project design so that the differential impact on mortality and functional status over, at least three years, can be tested. The cost comparisons in the study focus on average monthly cost to public health and social services payers, (Medicare, Medicaid and Title XX) by obtaining data from claims files. Only actual costs to 143skellie et al., p. 8. 144skel1ie et al., pp. 11-12.

PAGE 110

91 the programs are compared (i.e., cash benefits, patient's contribu-tions and deductibles are not included). In the case of nursing home costs, this results in a net cost to the state for Medicaid recipients of only $500 per month for intermediate care, compar ed to the average monthly service cost for the experimental group of $169.145 Although a step in the right direction, expenditures for income maintenance, food or nutrition supplement programs and subsidized housing are not included in the cost comparisons. There-fore, costs are not totally comparable. In any case, this design is a substantia 1 improvement over most others deve 1 oped by the demonstration projects and warrants closer inspection over time. Even so, the project does not meet all the criteria for a comprehensive, systematic review. Conclusion As discussed at the outset of this chapter, the primary focus of the study which follows was the cost-effectiveness of home health care. It thus was necessary to review a wide range of research covering the case mix, qua 1 i ty, cost, and cost-effectiveness of long-term care. The intention of this review was to place the study in its appropriate context and to provide the foundation for the concluding chapter of this study. 145Ruth Coan and Judy Hagebak, "The Alternative Health Services Project: A Prudent Approach to Adult Day Health Care and Other Corrmunity-Based Long-Term Care Services," (paper presented at the meeting of the American Public Health Association, New York, November, 1979), pp. 5-6. (Mimeographed.)

PAGE 111

CHAPTER III THE STUDY APPROACH AND METHODOLOGY Introduction This chapter first describes the purpose of the specific study cited herein and presents an overview of the conceptual framework used to address the study's goals. The next portion of the chapter describes the study sample and avail able data. The model that was used to operationalize the conceptual framework of the study follows. The chapter closes with a summary of the analytic approach and empirical techniques employed. Study Purpose This study addressed several basic questions: 1) How were the cost and -utilization of home health care influenced by various patient, provider, and c001munity char acteri st i cs? 2) Which of the characteristics studied were the most important in terms of the observed utilization and cost variations? 3} What were the relevant progranmatic, public policy, and research implications of the findings? The study was thus intended to provide a better understanding of the determinants of the cost of home health care for the purposes

PAGE 112

93 of improving the reliability of estimates of home health care cost and. utilization over the coming decade. In addition, the information derived on the types of patients which were lower in cost (and treated with equal effect) was intended to assist in the clarifica-tion and specification of public policies regarding the placement of patients in institutional versus home health care settings. Conceptual Framework The overall conceptual framework developed for the study was built upon earlier work in the field of long-term care which identified major groups of factors potentially impacting utilization and cost of care.l The subsequent model described herein was specific-ally designed to analyze characteristics expected to have the greatest impact on 'these issues. Certain characteristics were selected because they were potentially related to factors that could be controlled by, or serve as the basis for, public policy.2 Hence, the main criterion for inclusion of the explanatory variables was amen-ability to policy control. As in most applied studies, a secondary criterion was the availability of data. Figure 1 depicts the hypo-thesized relationship between the cost and utilization of care and these variables. lsee Chapter .II for an overview of these works. 2rhese are termed policy variables because they are amenable to manipulation by external forces.

PAGE 113

94 Independent Variables Variables hypothesized to affect utilization and cost were included in the overall analytic model and related to the dependent variables of the study. Major categories of variables identified were conceptually and analytically clustered into three groups: patient-specific, outcome-related and provider/health system variables (referred to in Chapter II as case mix, quality of care, and community/facility factors, respectively). Patient-specific variables, which included those case mix factors generally categorized as demographic and health status indicators (e.g., age, diagnosis, and functional status) were hypothesized to directly affect the cost of care. This group of variables was, in turn, hypothesized to interact with provider/health system variables to influence both the unit cost and the total utilization of home health care, which together comprised the cost per episode. Examples of provider/health system variables used in this study included population density and hospital admiss i ons per thousand population. The third major category of variables included in the study was comprised of those factors directly related to patient outcomes (i.e., quality of care). It included measures such as the change in the patients• functional status from t ime of admission to discharge and a global measure of change in health status. Utilization-related outcome factors, which were the result 'of interaction between the first two variable groups and utilization of services comprised the second part of the variable grouping labelled outcomerelated variables. Examples of these variables included the av-erage

PAGE 114

95 Cost per Episode ..... r I I I I Patient Provider/Hea lth I I Char acteri st i cs .... System I I Characteristics I I I I I I ... Outcome Ch ar act er i s t i c s Figure 1 Conceptual Framework of the Study

PAGE 115

96 niJllber of visits per day and the average length of use per episode. These three categories of variables together formed the independent variables in this study. Dependent Variable Because the primary purpose of the study was to explain home health care costs and utilization, the main dependent variable was the total program cost of selected home health care services per episode of illness. For the sake of comparability, it was calculated on the basis of all Medicare allowable costs. A secondary concern was home care utilization, measured by the total number of visits per episode. This included all skilled nursing, physical therapy, speech therapy, occupational therapy, medical social service, and home health aide visits provided during each episode of i 11 ness. Overall Relationships In Figure 1, the cost per episode of home health care is depicted as a function of the patient (case) mix, as measured by patient-specific characteristics, and the patient's outcome. It was hypothesized that outcome-related characteristics, although not necessarily solely determined by patient mix, would vary with it. For example, the service capacity of a more intensive hospital-based program might be better su ited to caring for a patient with medically intensive needs than a free-standing provider, which was oriented toward chronic disease care. Consequently, Figure 1 depicts patient characteristics as both directly influencing the cost per episode,

PAGE 116

97 as well as acting through outcomes to influence cost. Further, outcomes themse -lves were expected to exert a direct influence on the cost per episode since home health services require certain input resources which can influence the expense structure of the provider. Finally, the figure depicts the relationship whereby home health patient mix interacts with provider/health system characteristics to effect the outcome and cost of care. Although the variab 1 es re 1 ated to patient outcomes might account for some of the variation in cost, the direction of causation may also go from cost to outcomes as shown by the broken line (i.e., more services at a higher cost may result in a greater degree of patient improvement). Because the approach taken was primarily data analytic, the objective was to identify those factors associated with differences in cost per episode; hence, cost was used as the dependent variable and outcome was one of the categories of independent variables. The relationships depicted were based on the assumptions that patient needs varied and were reflected by patient characteristics, while services delivered were in response to the different needs of each patient. Services were also affected by provider/health system characteristics, which, in turn, affected the outcome and cost of care. Hence, the interrelationships between cost,. outcome and patient characteristics were studied in the context of, and where possible controlling for, additional factors that could influence, inter act with, and act independent 1 y of, patient mix and outcome in determining the cost per episode (CPE) of home health care. These relationships were described as,

PAGE 117

where, CPE = f(E_, .Q, .!!) , P = Patient-specific variables; 0 = Outcome-specific variables; H = Provider/Health system variables3. Sample Selection and Data Availability 98 This section begins with a discussion of data systems avail able for the study, then moves to a description of the selection of pro-viders from within each system and the final episodes ultimately used in the study. The section closes with a discussion of data availability which constrained the operationalization of the study's conceptual framework. Data System Selection Resource constraints (e.g., time and money) in addition to the expressed preference of the author for "real-world" situations, prompted the search for on-going home health programs that regularly collected data necessary to address the major questions of the study in accord with the conceptual framework presented above . Thus, criteria were developed regarding the availability, reliability, and generalizability of potential data sources. 3underscoring, such as P, is used in accord with accepted mathematical convention of denoting a group (vector) of several components of variables rather than a single variable.

PAGE 118

99 Several large-scale home health care data bases which contained patient care and utilization data were identified. Two of them were ultimately selected because they met most of the established criteria. Both data sets exhibited the following characteristics: (1) Similar and reliable data were available on a majority of the variables identified which were expected to affect home health care utilization and cost. (2) The data systems linked home health care utilization and cost at the patient level. (3) Data were availabl e for the same year in both systems, which resulted in the use of 1976 data. (4) Data were abstracted in a consistent manner in both data systems. (5) Data were availabl e for providers of varying size and location within a 1 imited geographic area. While allowing for diversity in provider characteristics, this fea-ture partially controlled for the introduction of exogen ous variables which might otherwise confound the analysis (e.g., region, economic and cultural variations). Similarly, s i nce one of the data sets represented hospitalbased providers and the other represented visiting nurse

PAGE 119

100 associations, the impact of provider control could be addressed. 4 (6) The use of large data bases assured a sufficient number of episodes in the study sample. (7) Individuals in control of each data system were willing to provide the data at minimal cost. (8) The data represented utilization and costs for ongoing home health care programs, rather than experimental or demonstration programs which are often under much scrut-iny, potentially jeopardizing the generalizability of the study findings. In summary,. two data systems were selected for use in this study because of their consistency, reliability, and representative nature. They were the Massachusetts Home Health Care Discharge Summary Systen and the Blue Cross of Greater Phi 1 ade 1 phi a Home Care Program. Provider Selection Si nee each of the data bases inc 1 uded a 1 arge number of home health care providers, it was necessary to select a sample of programs representing various provider characteristics which would allow the testing of hypothesized relationships. process is described in the following paragraphs. Th i s s e 1 ec t i o n 4To the extent that provider control measured other characteristics of the (e.g., levels of care or location), the interpretation of f1ndings regarding control was difficult.

PAGE 120

101 First, it was important to determine the general nature of the providers in each system. The Home Health Care Discharge Summary Systan was administered on a demonstration basis by the Massachusetts Department of Health over the three-year period, 1974-1976, with 46 agencies ultimately participating in the 1976 data collection process. Most of the agencies using the system were located outs ide the major metropolitan areas of Massachusetts and were visiting nurse associations of varying size and service capacity. All providers participating in the Greater Philadelphia Blue Cross Home Care Program were affiliated with major hospitals in the Philadelphia metropolitan area, the result of a home care progran which had been initiated by Blue Cross 15 years earlier. Although the providers in the Philadelphia area were similar to each other in some respects (e.g., they were all part of full-service non-federal hospitals), they represented home care departments of varying service capacity and case mix. In order to provide enough variation in the provider/health systan variables, while minimizing the cost of data verification and editing, it was decided that four providers should be selected from each data base. Criteria were then developed in order to select the four providers in each system. All providers selected used a Medi-care cost finding method which identified costs by type of service; this information was necessary in order to compute the cost per

PAGE 121

102 episode of home health care.5 To the extent possible, the providers selected represented an array of provider/health system variables which were hypothesized to affect home health care utilization and cost (e.g., size, location, level of service). Table II summarizes the principal characteristics of each of the providers selected for the study sample. Episode Selection Episodes were excluded from consideration for several reasons: incomplete treatment, incomplete or missing data, and costs per episode beyond ten standard deviations from the mean. The condi-tions which led to the final selection of episodes from both data sets are summarized in Table III. To assure consistency, the patient records used in this analy-sis were for episodes of care which began and ended in 1976. Limiting patient records to completed episodes meant that a number of cases were omitted frcxn the final sample.6 All cases open at the beginning or end of the study year were omitted from the final sample in order to assure that the utilization measures pertained to 5only one of the four cost finding methods avail able to home agencies separates the cost of therapies from other services. Thus, the use of the combined Public Health Service-National League of Nursing method was an important factor in provider selection. 6The extent to which this decision unduly limited the scope of the study to shorter 1 engths of use was investigated in the ear 1 y steps of the analyses. Because a sufficient number of episodes over 200 days were in the final study sample, it was assumed that this decision would not unnecessarily restrict the generalizability of the findings. Information regarding specific patterns of length of use and cost can be found in Chapter V.

PAGE 122

---------------TAIIL II SElECtED CIIAAACTERI SIICS Of l 'R0VIIlRS ' " 111 S IIIDY SNII'U : llas sa chuselts Philade lph i a ----r-------------------P rovi d e r l'rov ldcr Provid e r Pr o vid e r Prov ldt!r Provider Provider Prov i der I 2 3 4 I 2 3 4 ---------------Tota l Popuhtlon1 I thousands ) 136 1 ,397 304 6 1 9 633 I ,017 586 197 Populltlon Per Square Hlle 257 1 ,693 587 1 ,511 1 ,277 14,006 3,187 259 Type s of Care Prov l dedb SIIC SIIC SNC SIIC SHC SHC SHC SNC PT PT PT PT P I P T PT PI ST sr SJ ST S T SJ S T OJ OT o r or O J O J OT I* lA IUIA In lA IIIlA IUIA IN IA I MIA lilA ss ss s s ss s s Jotal llo. of Episodes 2 1 5 612 498 844 215 213 239 126 Tota l No. of Hunlng VIsits 3,810 0,557 6,247 11,000 2,533 2,025 2,61 9 1 ,22 0 Total No. of T herapy VIs Its 666 2,445 344 1,52 5 146 295 9 4 6 325 Total No. of A i de VIsits 892 6,63 2 4,765 4,48 1 40 31 178 6 3 2 Ave rage length of Use ( days ) 74.13 104.53 70.91 61.54 30.71 30.43 35.77 38.26 Pilyo r llh (S) llcdlcare 65.6 12. 2 51. 1 63. 0 67. 7 5 2. I 79. 8 H e d l ca l d 9 . 6 11. 8 14. 9 9 . 3 0 8 . 2 8 . 8 1.8 I nsuran c e 15. 3 1 0 . 5 1 1.4 11. 7 43. 7 24. 1 37. 7 1 6 . 5 Private Pay 2 . 9 1.9 1 8 . 0 11. 5 0 0 . 9 1.8 Cost per Nursing VIs It $10.29 $13.76 $15.77 $12. 5 8 $17. 5 7 $19 .63 $18 . 6 1 $17 . 3 1 Source : Hassa chuselts Ols charCJe S u n • nar y S y s l P o M , .lcdl c are C o s t R eports , Olue Cross llotne Care P r ograow O ala S y s tcto, B l u e Cross of Phila d elphia lloalP. Care r • , ogr a n • Direc tor, and _ _!!!)!!!_. 1ror county I n whi c h toafn offic e w a s l o cated. bSIIC refers t o skille d nursing care, PT to thenpy, Sl lo s p eec h the raJiy, O T to o ccupationa l therapy, IIIlA to h011e heallh 1 lde, and SS to s o c ia l servi c es . 1-' 0 w

PAGE 123

TABLE I II FINAL SAMPLE OF ALL EPISODES Massachusetts Philadelphia Beginning Total Cases in Sample 2,229 791 Cases Deleted Outliers a 39 25 Discharged after Study Year 56 Missing Data 29 No Home Care Used 2 Final Sample b 2,190 679 Source: Massachusetts Discharge Summary System and Blue Cross Home Care Program Data System. 104 aAll cases with costs per episode 10 standard deviations beyond the mean were eligible for deletion from the final sample. Each such case was reviewed and those judged to be sufficiently unique were omitted from the sample. bThe final Massachusetts sample represented 1 ,971 patients, while the Philadelphia sample represented 679.

PAGE 124

105 a complete episode of care. In addition, a11 cases transferred to community-based home health agencies, outside the control of the hospital-based agencies, were omitted because those cases did not represent an entire episode of illness. Several Massachusetts cases had missing patient identification numbers. Because of potential duplication within the sample, these cases were excluded from the analysis. The result was a final sample of 2,190 episodes in Massachusetts and 679 in Philadelphia. One final issue is noteworthy with regard to the episode as the unit of analysis. In Massachusetts, the sample of 2,190 episodes represented 1,971 different patients; no patient had more than two episodes in any year. Therefore, there were 219 patients with two episodes each during the study year. In Philadelphia, there were no patients with duplicate episodes. Sensitivity analysis was conduc ted to determine the effect of the duplicate epi sodes on the distri-bution of key patient descriptors (e.g., functional status, age, diagnosis, and cost). Analysis of variance indicated that no significant differences existed between those patients with more than one episode and a11 others. Therefore, patients with duplicate episodes were included in the study sample. This finding was contrary to that of the New York study described earlier, hence it may merit further attention in the future.? ?Geraldine Widmer, Roberta Brill, and Adele Schlosser, 11Home Health Care Services and Cost,11 Nursing Outlook (August, 1978), pp. 488-493.

PAGE 125

106 Study Variables and Data Sources Once the samples had been selected, both data sets were reviewed for information appropriate to test the hypothesized relationships. Both sets already contained patient level data such as age, living arrangement, diagnosis, and source of referral. In addition, utilization data of specific service types and frequency of use were available. Most of the data were already computerized. Additional variables were abstracted from patient records, while others were obtained from secondary data sources as necessary. The specific variables used in this study are discussed in the next section of this chapter and Table IV, which sunmarizes the study variables, concludes the chapter. Appendix B contains copies of the data collection instruments used by each of the home care data systems and a copy of the instrument used to abstract additional patient data from Philadelphia records; Appendix C contains a detailed description of each variable used in the study. Reliability of Data Relatively few difficulties were encountered with regard to internal consistency and reliabilty of the data used in this study. Both data sets were edited carefully prior to use in this study; consequently, only minimal receding was necessary to assure consistency. Data abstracted from Philadelphia case records were reviewed by at least two additional persons to assure their completeness, accuracy, and reliability. Massachusetts data were verified at the time of collection by health department personnel.

PAGE 126

107 Specification of the Model This section discusses the components of the model presented above in greater det ai 1 . First, the primary dependent variable, cost per episode, and later, each of the independent variables is described. The Primary Dependent Variable: Cost Per Episode of Home Health Care Attempts to address the cost of home health care in the past have seldom gone beyond a determination of cost per visit. Yet experience indicates that in order to make appropriate comparisons between institutional and non-institutional care settings, accurate cost data must be presented not only on the basis of a unit of ser-vice but on the basis of cost per case, per diem, diagnosis, and per episode of i11ness The problem with comparing cost per visit is illustrated by the fact that in 1976 visiting nurse association agencies provided care to over half of the Medicare enrollees using home health benefits nationally, while ranking below the national average in all utilization measures and charges per person. On the other hand, private non-profit agencies exhibited the highest aver-age number of visits per case and average cost per visit, well above the national average.9 Consequently, a comparison of the cost per 8Judith LaVor and Marie Callendar, "Home Health Cost-Effective-ness: What Are We Measuring," Medical Care, XIV (October, 1976), 866-872. 9u.s., Department of Health, Education, and Welfare, Health Care Financing Administration, Medicare: Utilization of Home Health Services, 1976, Research and Stabst1cs Note No. 2, (\1Jash1ngton, o.C.: Government Printing Office, June, 1978), pp. 10-18.

PAGE 127

108 visit or the average nllllber of visits per episode is inadequate to address the basic policy issue of the total programmatic costs of alternative care approaches. Thus, a more accurate indication of the cost of home health care is provided by the measure of cost of home health care per episode of illness (CPE) .10 As noted earlier, the primary dependent variable in this study was the cost of home health care per episode, which equalled the sum of the products of the cost per unit of service times the number of units of each type of service provided to a patient from the day of home care admission to day of discharge.11 To examine the extent to which observed variations in the total progrclTl costs were not principally a function of variations in unit costs, the total number of visits per episode was also examined as a dependent variable.l2 Because utilization and cost were highly correlated (R = .97), the findings discussed i n this report focus on cost. When the relationships between the two dependent variables and the independent vari-ables differed, utilization is discussed separately. Otherwise, a lOAs argued earlier, in the case where the cost comparison is made across home care programs (as in this study), the importance of including total living costs is diminished. For this reason, in addition to limitations in data availability, only direct program costs were included in the calculation of CPE. llTo obtain the total CPE for Philadelphia cases, an administrative surcharge was added to this sum. This is discussed in more detail in a later section of this chapter. 12As part of the preliminary analysis, another dependent variable, total nursing visits per episode, was examined. S i nce this variable was found to exhibit relatively low and insignificant ordinary and multiple correlation coefficients with each of the independent variables, it was not included in the final analyses as a separate dependent variable.

PAGE 128

109 discussion of the specific relationships between utilization and the independent variables can be found in Appendix D. Since the major dependent was comprised of several measures, it was essential to define its composition carefully. The remainder of this section provides such information. Definition of Episode. A major area of concern in this study was the definition of episode, which was defined as the length of use of home health care for each admission, from day of admission to day of discharge. This definition proved somewhat problematic due to inconsistencies in the definitions of an episode used by existing providers. Often a client may be discharged by a home health agency, later admitted to an acute or skilled nursing care facility, and readmitted to the home care program. Upon readmission, the patient might be classified as a new episode even though his admission was the result of a "flare-up" of the original problem. Unfortunately, the issue could not be resolved in this study. Thus conflicting concepts of episode were avoided, to the extent possible, when selecting the study providers. The result was that all providers in this study based their working definition of an episode on the Medicare concept of "spell of illness." This does not imply that all providers imposed Medicare eligibility requirements on their patients, but rather that definitions were used to a large extent in the classification of patients and administration of programs. The term "spell of illness" pertains to a period of consecutive days that begins with the first day on which a patient receives

PAGE 129

110 acute or skilled nursing facility services provided by a Medicarecertified provider. It ends with the close of a period of 60 con-secutive days during which the patient is neither a hospital inpatient nor a patient of a skilled nursing facility. Although technically an individual may be discharged from and readmitted to a hospital or skilled nursing facility during one spell of illness and still be in the same spell, all home health episodes included in this study were considered concluded upon admission to an inpatient facility. Unit of Serv ice. In this study, a vi sit was defined as a per-sonal contact in the place of residence of the patient, made for the purpose of prov i d ing a health service. Although some providers record home health aide visits on an hour l y basis, all services pro-vided to patients in this study were recorded on a per visi t basis. In order to improve the comparability of service patterns, only ( ---the six Medicare reimburseable services (skill ed nursing care, phys-ical therapy, speech therapy, occupational therapy, medical social services, and home health aides) were used in calculating the total number of visits per episode.l3 In addition, neither administrative function visits nor evaluation visits were included, since neither 13Almost 18% of the total visits provided by the Philadelphia programs was for ancillary services. Since the cost of professional coordination of these services was included in the administrative surcharge allocated to the cost of the Medicare reimburseable services, the total cost for these cases may be slightly overestimated.

PAGE 130

111 type of visit provided direct patient care, nor was it an allowable Medicare cost.l4 Cost Per Unit of Service. The basic unit cost data used to compute cost per episod e in this study was the average Medicare allowable cost per visit for each of the six services analyzed, as derived from intermediary reports and Medicare Statements of Reim-burseab 1 e Cost, verified and adjusted for comparability. The following characteristics of these cost data should be noted. First, the data were based upon Medicare allowable costs, which may be different from total agency costs and may not reflect actual reimbursement, depending on the payor mix of the agency. Second, the data did not take into consideration the intensity of resource utilization per visit. Although such information ....ould be optimal in order to develop a sensitive cost analysis, few agencies collect service data by time spent per visit in patient care. (None of the agencies providing data for this study did so.) Third, the cost data were representative of the general situation in that there was wide variation in unit costs by type of service across the eight providers. Finally, these cost data did not include the value of services contributed by other agencies or by volunteers. Generally, neither the frequency nor the cost of those services was recorded by the providers. Therefore, the extent to which additional support 14To the extent that the costs of these visits were allocated to the cost of agency administration, a portion of the costs were included in the cost per episode computat i on.

PAGE 131

112 was provided to the patient by other agencies and/or fani ly and friends could not be included in the cost computation for this study. Nor were cash benefits such as Social Security or public aid considered. Hence, the cost data for this study reflect direct progran costs standardized on the basis of Medicare reimbursement principles.l5 Cost per unit of service for the Massachusetts agencies in this study was taken directly from audited Medicare Statements of Reimburseable Cost obtained from the Massachusetts Rate Setting Commis sion. Cost data for the hospital-based Philadelphia programs were obtained from unaudited cost reports provided by Blue Cross of Greater Phil adelphi a.l6 These programs contracted for a majority of their patient care services with community-based agencies. Thus, the charge per v isit to the hospitals became the cost of care at the hospital level (to third party payers). These charges were then passed on to the payer as the cost of care, with the addition of an administative surcharge to cover the hospital's cost of management and overhead .17 15The specific costs per unit of service used to compute cost per episode are contained in Appendix C. 16sas.ed on past experience, the extent to which unaudited cost reports overestimated Philadelphia costs was minimal. 17 The administrative surcharge included the costs associated with the identification and direct referral of inpatients to community-based agencies when patients did not require the intensive level of care provided by the hospital's home care department. The implication of such allocation procedures is d iscussed more fully i n Chapter V of this report.

PAGE 132

113 Cost Per Episode. The dependent vari ab 1 e components described above were combined for the six home care services to produce cost per episode as follows: C = Cost per unit of service by type of service (i = nursing -1 service (N), physical therapy (PT), etc.) Q = Units of service by type of service (i) per episode of _, illness. CPE = Cost Per Episodel8, in vector notation: [ CN, CpT, . . . CK] varies by provider X QK -varies by patient The vector of costs per unit was comprised of the average cost for each provider of each type of service; these unit costs were, of course, the same for all cases treated by each provider, but varied across providers. The vector of units of serv i ce was measured as the number of units of each service per episode. 18To this computation was added the administrative surcharge in the case of Philadelphia episodes.

PAGE 133

114 The Independent Variables: Patient-Specific Characteristics Patient factors have probably received the most attention in the health care literature in relationship to utilization and cost. As discussed in Chapter II, several studies concluded that various patient characteristics affect the cost of care, yet few attempted to control for more than a few at a time. The following justification for the inclusion of each variable is based on earlier studies. Age. The age of a patient is clearly related to his functional incapacities,19 use of health services20, femaleness,21 and subsequent cost of care. Specific findings of several studies cited in Chapter II are relevant to this issue. In one study, the most significant determinant of the cost per unit of service was the 19Thomas Wan, "Age Severity of Disability," Review of Public Data Use, III (1975), 29-32. 20Karen Davis and Roger Reynolds, "Medicare and Utilization of He a 1 th Care Services by the Elder 1 y," Journ a 1 of Human Resources, X (Summer, 1975) , 361-377; Thomas Wan, "Interpreting a Genera 1 Index of Subjective Well-Being," (paper presented at the meeting of the Gerontological Society, San Francisco, November, 1977); William Weissert, "Costs of Adult Day Care: A Comparison to Nursing Homes," Inquiry, XV (March, 1978), 10-19. 21In part, because the patient•s sex was highly correlated with his age (i.e., the vast majority of elderly people are women),. analysis was completed using only the age of the patient.

PAGE 134

115 c 1 i ent • s age. 22 The New York study of home care found that the total cost of home care may actually decrease with advanced age, speculating that those who are elderly require less costly 11basic11 care at less frequent intervals than younger, more acutely ill patients.23 Many have argued that the relationship of age to utili-zation of general health services is bimodal, since both the very young and the old use more health services than those in the middle age range. That is, the young have a higher prevalence of acute conditions while the elderly have a higher incidence of chronic disease. Thus, it was clear that the re 1 ationshi p between age and cost warranted investigation. In fact, age was a policy-relevant variable of particular interest since it serves as the basis for Medicare eligibility.24 Living Arrangement. Many studies have found that the patient•s living arrangement affects the quantity and type of care utilized.25 22Linda Scharer and John Boehringer, 11Home Health Care for the Aged: The ProgrcJll of St. Vincent•s Hospital, New York City .. (New York: Boehringer Associates, 1976), pp. 9-11. (Mimeographed.) 23widmer, Brill, and Schlosser, pp. 488-489. 24The age of patient was measured in years at time of admission to home care. The square of this variable was also used in the multivariate analyses. Regression coefficients for its exponential form were not statistically significant, consequently the exponential age variable was omitted from subsequent analyses.

PAGE 135

116 If another adult resides in the household and can assist in the pro-vision of services, then the time spent by professional care givers may be reduced. A reduction in the time spent in performing each service should subsequently reduce the cost of that service, although not all utilization will be affected in the same manner by the patient's 1 i vi ng arrangement. In the New York study, those living alone received essentially the same number of nursing visits, but twice as many social service visits as those living with others.26 Thus, living arrangement was expected to impact total cost of services utilized. Diagnosis and Functional Status. A major area of concern in the study was the appropriateness of descriptors for the patient's illness. As discussed earlier, concern over the appropriate cate-gorization of patients is not limited to long-term care, home health care, or the elderly. Classification systems developed as alter-natives to the International Classification of Diseases (ICDA) can be found in discussions of ambulatory medical care, nursing home care, and hospital care. One reason for the original development of the problem-oriented record was the lack of consistency and the nonrepresentative nature of the ICDA coding and classification system. The Long-Term Care Patient Classification System27 and the Older 26widmer, Brill, and Schlosser, p. 490. 27Ellen Jones, Barbara McNitt, and Eleanor McKnight, Patient Classification for Lon -Term Care: Users Manual, Bureau of Health Services Research and Evaluation Washington, D.C.: Government Printing Office, 1974).

PAGE 136

117 American Resources and Services methodology (OARS) developed by Pfeiffer28 are both multidisciplinary ways of assessing patients. The Instrumental Activities of Daily Living (IADL) scale developed by Lawton and Brody29 is another. Other methods of measuring func-tional status were reviewed in Chapter II. The proliferation of assessment instruments reflects the necessity for multiple patient descriptors or stratifiers and the need for their standardization. Both primary diagnosis and a measure of functional ability at the time of admission were used in this study because each offered distinct advantages over the other.30 Diagno-sis defines the patient's condition in terms of his medical or surgical condition and provides information which suggests the general type of care needed. The diagnostic condition can provide an estimate of the patient' s need for care in the future. One problen in using diagnosis alone is its lack of specificity in describing the extent and kind of functional disability affecting the patient in his current state. Linn and others argue that the same diagnosis can result in various functional levels and behavior 28Eric Pfeiffer, ed., Multidimensional Functional Assessment: The OARS (Durham, NC: Duke Un1versity, 1975), pp. 6-28. 29M. Powell Lawton and Elaine Brody, "Assessment of Older People: Self-Maintaining and Instrumental Activities of Daily Living," Gerontologist (1969), pp. 179-186. 30Another factor which can affect the intensity and subsequent cost of service is the stage of disease progression of the client. Although it was recognized that the cost of home care may vary throughout the stages of a patient's convalescence, documentation was not available on the variation in costs over the entire length of use of home care.

PAGE 137

118 patterns depending on other patient characteristics.31 As a patient descriptor, the functional status index offers the opposite advantage of diagnosis, since it focuses on current need (especially those related to physical disabilities), and lacks the future perspective associated with diagnosis. Hence, both measures were used in this study. Primary diagnosis was recorded, as well as a six-item AOL index. Each of its components was recorded separately in order to facilitate secondary analyses. Surgical Procedure. Tl:le extent to which the presence of a surgical procedure prior to admission to home care affected the cost of home care was expected to be an important issue. For each episode in the analysis, this variabl e ind icated the presence or absence of any surgical procedure immediately prior to admission to the home care program. Goal at Admission. Because utilization and cost patterns may be different for patients with various goals (e.g., rehabilitation versus maintenance care), as determined by providers, this issue was included. Differing patterns of utilization and cost were expected to be the result of different expectations on the part of the care providers (based on some notion of the potential of each patient), or the genera 11 y agreed upon protoco 1 s for the care of different types of patients (e.g., all rehabilitation patients with a diag-31Margaret Linn, Lee Guerel, and Bernard Linn, 11Patient Outcome as a Measure of Quality of Nursing Home Care,11 American Journal of Public Health, LXVII (April, 1977), 337-343.

PAGE 138

119 nosis of stroke should receive speech and physical therapy twice a week). The extent to which this variable was significant in explaining observed variations was the subject of inquiry. The Independent Variables: Outcome-Related Characteristics As discussed earlier, this set of variables was separated from the preceding because, conceptually, it was intended to measure the impact of care (or quality) on both patient status and outcomerelated utilization over time. Although, in an absolute sense, it was impossible to attribute any change in patient status totally to the receipt of home health care, in a relative sense, it was possible to use the factors listed on the following pages as, at least, proxy indicators of impact or outcome of care. Change in Functional Status. A variable was created for this study which computed the change in functional status (ADL) index score (assigned by the primary care giver) from the time of admi ssion to the time of discharge for each patient. Although there were difficulties in attributing substantial significance to the score itself, in part due to problems of reliability, it appeared to be a useful estimate of change in patient functional status over time, especially for comparative purposes. Consequently, the six-item ADL index at the time of admission was subtracted from the comparable score at time of discharge to find the numerical value associated with the change over time. Health Status at Discharge. A simple estimate of change in general health status over time made by the provider (e.g.,

PAGE 139

120 improved, same, worsened, or dead) offered additional information, since it was an estimate of outcome which included diagnostic or medical intensity, as well as psychosocial needs. Because this estimate of change was broader and thus less precise than the change in ADL measure, it was included in order to develop as broad a based outcome measure as possible. At the same time, it allowed for the simultaneous evaluation of the reliability and consistency of both measures. Home Health Care Length of Use. The number of days from date of admission to date of discharge was used as a measure of the interaction of patient characteristics, alternative available health care resources, and the patterns of practice within an agency and corrununity. A 1 though 1 ength of use was high 1 y co 11 i near with the dependent variable, cost per episode, to the extent that it captured the above interaction, it was treated as a separate factor. Intensity. Levels of care are of major concern to home health care providers since they serve to define type of care provided and patient served. Whether all agencies do or should provide each of the four levels of care described in Chapter II, is a matter of debate. In addition, little is known about the relative intensity of service provision to various patient types across providers. Consequently, a proxy measure (of levels of care) was calculated as a measure of intensity (i.e., the total number of visits per episode was divided by the total number of days per episode).

PAGE 140

121 The Independent Variables: Provider/Health System Characteristics Provider/health system variables were included in this study in order to aid in the interpretation of the variation in cost per episode. It was intended that the inclusion of these variables, while controlling for certain patient characteristics, help explain the effect of the organization and delivery of services on the cost of care. Since differences among providers probably affected utilization and cost patterns, these were taken into account using several variables. . Also, because the main focus of this study was on case mix and quality factors, it was necessary to control for major possible confounding influences due to provider differences such as size, and health system differences such as payer and general acute care util ization in the community. Consequently, the following factors were included. Primary Source of Payment. In an attempt to capture the impact of variations in reimbursement policies and benefit packages by type of payer, the primary payment source for each episode (e.g., Medicare, Medicaid, private pay) was recorded. This variable was expected to be of importance for several reasons. As noted earlier, there are widespread variations in the benefit programs. For example, some states limit Medicaid benefits to 50 visits per calendar year, while other states impose far different limitations. In Pennysylvania, hospital-based home health providers are reimbursed at a rate of $5 per day for Medicaid patients, for any type of service, while free-standing agencfes are

PAGE 141

122 reimbursed twice as much. It is not surprising that the percentage of Medicaid patients in hospital-based programs in Pennsylvania has decreased in recent years. In other areas of th e country, it is alleged that some agencies discontinue the provision of services to Medicaid beneficiaries whose benefits are no longer available, ir-respective of need. 32 The extent to which these types of varia-tions affect the utilization of services was therefore investigated. While differences exist across benefit programs, they also exist within them. The extent to which such differences were the result of variations in fiscal intermediary policies could not be addressed because the amount of variation in this variable was extremely limited. Although both primary and secondary payers were included in this study, emphasis was placed on the principal payer because few episodes of home care have more than one. 33 Thus it was hypothesi zed that the benefits covered by the principal payer would have the major impact on utilization, with little, if any, effect 32u.s., Congress, Senate, Special Committee on Aging, Home Care Services for Older .Americans: Pl ann in for the Future, 96th Cong., st ess., May , 2 , 979 Wash1ngton, D.C.: Government Printing Office, 1979), pp. 86-90. 33sensitivity analysis of the findings using a sample that excluded episodes with secondary payment sources was not s i gnificantly different from the finding with all episodes included. This was, in part, due to the small nl.lllber of cases in the study sample with more than one payer (none in Philadelphia and 20% in Massachusetts). Ten percent of the Massachusetts episodes had Medicaid as their second payer, four percent were private pay (out-ofpocket), t"'-0 percent had Medicare, and four percent were unknown sources of payment.

PAGE 142

123 exhibited by the secondary payer (since it reimbursed for a relatively small proportion of the total cost). Provider Size. In order to test the impact of economies of scale, a variable measuring provider size was included. Partly because all Medicare-certified agencies must provide skilled nursing care, the majority of visits provided are for the provision of such care. Hence, size was measured as the total nlJTlber of nursing visits provided during the study year. Because many home care pro-viders use part-time employees to provide care, the nlJTlber of nurs-ing visits provided either directly or through contractual arrange ments was used as a measure of provider capacity.34 Population Density. The problem of recruitment of professional health personnel by agencies located in rural or semi-rural environ ments is well known. The extent to which this problem aids in the understanding of the variation in utilization (which affect the cost per episode) was partially addressed by the inclusion of population density as an independent variable. There may also be differences in utilization due to other factors measured by population density 3 4 This measure may have underestimated the size of the Phila delphia programs, since their activity level was greater than such a measure might indicate. Each of the Philadelphia programs referred more than twice as many patients directly to free-standing home health agencies for intermediate (level of) care not requiring close monitoring for post acute and unstable conditions, than it actually treated through its own program. The extent of actual resource utilization for such activities was unknown; hence, total nursing visits remained the most reliable and comparable available measure, although it may have been consistently low for Philadelphia providers.

PAGE 143

124 (e.g., low demand due to size of potential market or high costs due to increased travel time). Thus, the population per square mile within each service area was used as a measure of urban/rural differences. Acute Care Admissions. The extent to which alternative health care resources in a community affected home health utilization was unclear. As discussed in Chapter II, although there seems to be agreement that home health care can reduce acute care lengths of stay, the impact of home care on nursing home utilization is less clear. There are several explanations for the association between home health and acute care utilization. It could be argued that a low number of acute care admissions per thousand population might reduce the number of referrals to home health agencies and subsequent utilization. The result is increased home health care unit costs. On the other hand, low acute care utilization may result in the reduced availability of patients for home care referral; perhaps due to increased pressures on hospitals to retain patients for longer stays when the patient census is low. Also, low numbers of acute care admissions may be an indication that the area is served by an effective and efficient home health agency which is able to divert utilization from an inpatient to an outpatient setting. High acute care admissions per thousand population should produce diversion effects of the reverse order, increasing home health care utilization in the aggregate because of an increased number of discharges from hospi-

PAGE 144

ta 1 s. 125 The resulting effect is potentially more utilization per agency and lower unit costs. In either instance, the utilization of acute care services is an important variable in relation to the utilization and costs of home health care. Consequently, it was included in this analysis. Because the ability of home health care to substitute for nurs ing home care has not been consistently demonstrated empirically, it was not included in this study. To the extent that nursing homes serve as an alternative to home health care, they were considered in the interpretation of the study findings. Their omission was not expected to substantially.alter the study findings. Miscellaneous Factors Several other characteristics or factors not included in this analysis, but potentially relevant, were omitted due to data constraints. The following are illustrative of such characteristics. Income. A patient's income may potentially affect his ability to withstand out-of-pocket expenses for the actual provision of care and/or drugs and medical supplies necessary for adequate compliance with medical regimens. To the extent that Medicaid, as primary payer, was a proxy measure for low income, it was addressed in this study. Race. Although race may be a factor contributing to variations in utilization and cost, data were not available on the race of the patient for this study. To the extent that the racial mix of the patients in the tv.o provider groups was different, the patterns of

PAGE 145

126 utilization and cost may be at least partially explained by the differences. The impact of such characteristics was unknown in this study. Other Health Care Providers. The impact of the availability of alternative home health care providers within the community is important since 11competition11 from other agencies may impact the patterns of practice within the community. In general, there were few other home health care providers in the communities served by the four agencies in Massachusetts; Philadelphia was served by many other home health agencies besides those included in this study. The extent to which this impacted the study findings was considered in the interpretation of the findings. Regulatory Environment. The regulatory environment in the states from which study agencies were selected influenced utilization and cost. No doubt, the existence of the Massachusetts Rate Setting Commission, with its statutory review of all costs of home care, affected patterns of utilization and cost in that state. Other issues such as certification and licensure may also potentially affect the findings of this study. Although it was not possible to systematically collect data and thoroughly control for all possible factors potentially influencing home health costs, the existence of mitigating factors was considered in the interpretation of the findings of this study to the extent possible. Each of the variables is summarized in Table IV. A de .tailed description of their measurement is contained in Appendix C of this

PAGE 146

Data E1 ements Dependent Variables Total Visits Per Episode Cost Per Episode Massachusetts Philadelphia a Acronyms TOTVTS CPE TABLE IV DATA SPECIFICATIONS Measurement Specifications Total number of visits for each type of service: Skilled Nursing Care Physical Therapy Speech Therapy Occupational Therapy Social Services Home Health Aide Services Total number of visits times cost per visit, for each type of visit. Total chargesb for each type of visit plus a of the administrative surchal"lge. Data Sources l. Massachusetts Discharge Summary System (MOSS) 2. Blue Cross Home Care Program Data System (BCDS) l. MOSS 2. Medicare Statements of Reimbursable Cost l. BCDS 2. Blue Cross of Greater Philadelphia Home Care Program Director ...... N .........

PAGE 147

Data Elements Acronyms Independent Variables Age AGE Living Arrangement LIVARNG Primary Diagnosis INFECT NEOP EN DOC BLOOD MENTAL NERVS CIRCU RESP DIGEST GENITO SKIN MUSCO ACCIO Other TABLE IV (continued) DATA SPECIFICATIONS Measurement Specifications Age recorded at time of admission, in years Alone 0/1 With Re 1 ati ves 0/1 Others Infection (1-137) 0/l Neoplasm (140-239) 0/1 Endocrine (250-279) 0/l Blood 0/l Mental (794,299-315) 0/l Nervous (320-389) 0/l Circulatory 0/l Respiratory (490-520) 0/l Digestive (530-575) 0/1 Genitourinary (599-620) 0/1 Skin (690-710) 0/1 Musculoskeletal (715-739) 0/l Accidents (829-980) 0/l Data Sources 1. MOSS 2. BCOS 1. MOSS 2. Philadelphia Abstracted Patient Records 1. MOSS 2. BCOS -N co

PAGE 148

TABLE IV ( continued ) DATA SPECIFICATIONS Data Elements Acronyms Measurement Specifications Data Sources Goal at Admission Recovery l. MOSS Self-Care REHAB Rehabilitation 0/l Maintenance TERM Tenninal 0/l None Change in Functional AOLCHG AOL Index at Discharge minus AOL l. MOSS Status--Change in AOl Index at Admission: 2. Philadelphia Abstracted Index Patient Records Improved (-6) to Deteriorated (+6) Change in Health Status PATSTAT Improved l. MOSS Same 2. Philadelphia Abstracted -Worsened Patient Records OEAO Expired 0/l Change in Health Status BETWORSE Improved or same 0/l l. MOSS at Discharge 2. Philadelphia Abstracted Patient Records

PAGE 149

: Data Elements Acronyms Functional Status at BATH Admission DRESS AMBO TOILET TRANS EAT Functional Status--Admitting ADL Index ADADL Surgical Procedure YNSURG TABLE IV (continued) DATA SPECIFICATIONS Measurement Specifications Bathing 0/l Dressing 0/l Ambulation 0/1 Toileting 0/l Transferring 0/l Eating 0/l Total score of functional status rating, when independent is given value of zero Total Scores range from zero to six Bathing 0/l Dressing 0/l Ambulation 0/l Toileting 0/1 Transferring 0/l Eating 0/1 Presence of any surgical diagnosis at admission 0/l Data Sources l. MOSS 2. Philadelphia Abstracted Patient Records 1. MOSS 2. Philadelphia Abstracted Patient Records 1. MOSS 2. BCDS t-' w 0

PAGE 150

Data Elements Acronyms Home Health Length HHLOU of Use Intensity INTENS Provider/Health System Variables Primary Payor MEDICAR MEDICAID INSUR PPAY Provider Size SIZE Population Density DENS TABLE IV (continued) DATA SPECIFICATIONS Measurement Specifications Number of days case open Total number of visits per episode divided by the total number of days (length of use) per episode Medicare 0/l Medicaid 0/l Insurance or Blue Cross 0/l Private Pay 0/l Other 0/l Total number of nursing visits, 1976 Population per square mile of county in which main office is located, 1976. Data Sources l. MOSS 2. BCDS l. MOSS 2. BCDS l. MOSS 2. BCDS l. MOSS 2. BCOS l. and Data Book --

PAGE 151

Data Elements Acronyms Acute Care Admissions ADMISS a TABLE IV (continued) .DATA SPECIFICATIONS Measurement Specifications Non-federal general acute care admissions per thousand po8ulation in County of main office, 1 76. Acronyms are the variable names used in the multivariate analyses. bRatio of costs to charges for all visits was 100%. Data Sources 1. AHA Guide 2. and Data Book cRatio of costs to charges varied for Philadelphia providers from 77% to 100%. City dlndicates the category was used as a dichotomous variable in selected regression analyses in one study. eEach diagnosis was coded with its specific ICDA code and then grouped into the appropriate categories. Numbers in parentheses are the ICDA-8 codes. w N

PAGE 152

133 report with specific reference to the data sources used in this study. Each of the variables was used in various ways to test the hypothesized model depicting the variation in utilization and cost of home health care. The following section briefly describes the analytic techniques used to test these relationships. Analytic Techniques In a non-experimental design such as that used in this study, there are several approaches to controlling for bias.35 One of the most commonly used involves retrospectively matching the study subjects on the basis of key factors (e.g., age, primary diagnosis, functional status) that are expected to highly influence the depen dent variable, CPE. Another method for controlling bias is basically a statistical one, using stratification by key factors and multivariate sta-tistical techniques to control for any number of factors. Because it was believed that a large number of variables simultaneously influenced CPE, it was decided that the matching of study episodes would be highly problematic (i.e., it would be very difficult to obtain good matches across the two provider groups), and even when good matches were made, many factors waul d st i 11 remain uncon-trolled. Therefore, the approach taken in this study was a statistical one. 35Fred Kerlinger, Foundations of Behavioral Research (2nd ed.; New York: Holt, Rinehart and Winston, Inc., 1973), pp. 315-348.

PAGE 153

134 A combination of analytic techniques was used to test the basic model and to assess the consistency of the findings. First, descriptive one-and two-way tables were developed which Sllllmarized the relationships anong the patient-specific, outcome-related, and provider/health system characteristics (the independent variables) and total utilization and cost per episode of home health care (the dependent variables). Selected three-way tables were also constructed to assess the relationships anong some of the _independent variables. Second, statistical tests, such as chi-square, analysis of variance and t-tests were applied to obtain preliminary information regarding the main factors influencing the dependent variables. The results were then used to help specify the subsequent multivariate analyses. Third, canonical correlation coefficients were computed for all pairs of the three groups of the independent variables; zero-order correlation coefficients were calculated for all possible pairs of the independent and dependent v ari ab 1 es. These were used to examine the linear relationships between the groups and pairs of variables and provided a further basis for the specification of the regression equations used as the final empirical analytic technique. In an attempt to better understand the underlying relationships, both fixed and stepwise least-squares regression equations were estimated for the relationships between the dependent variables and the three major sets of independent variables. All equations were estimated separately using pooled data from the two groups of providers. In addition, selected equations were estimated separately on a subset of patients (of particular interest, for their policy relevance) in each of the pooled data sets.

PAGE 154

135 Limitations of the Study In light of the shortcomings identified in the studies described in the prior chapter, it is essential to also point out the limitations of the in-depth statistical study presented herein. Constrained by the limitations of the available data and study resources, it had some of the same shortcomings as other studies, while avoiding the problems of others. Only the limitations of the study are presented here. Its strengths should be obvious in subsequent chapters. Data Limitations As is the case with most studies that re 1 y on the use of data from ongoing programs, the available information was less precise and not always collected in a manner conducive to the analysis of desired variables, was sometimes inaccurate, and not available at all on some key points. Making do with operating data from ongoing programs increased the study's generalizability to the 11real11 world, but at the same time limited its analytic possibilities. This study suffered from this generic problem, and was affected i n several areas because of it. Cost Data. The measurement and analysis of the cost of home health care was particularly influenced by 1 imitations of data avai 1 ability. 1) Only the six Medicare reimburseable services were included in the calculation of cost per episode. Although this was done to

PAGE 155

136 assure the availability of standardized cost measures and the comparability of the components of the utilization vector, it eliminated from consideration several important home health care services (e.g., nutrition counseling, homemaker/chore services, friendly visitors, etc). Whether the provision of such services (not included in the cost calculations) affected the cost structure of the relevant providers was unknown. To the extent that the cost per unit of service for other services (included in this study) was affected by the total mix of services offered, average costs per visit were underor over-stated. The other major implication of the elimination of selected home care services from the study was that the dependent variable, cost per episode, did not truly represent the total cost of home health care. Other services may have been provided to the patient during the course of treatment. How this omission affected the cost and effectiveness measures was unknown. For example, it may have been the provision of inhalation therapy services to the patient with chronic lung disease which actually improved his health status during his course of home health care use; because inhalation therapy services were not studied, any observed improvement in the patient's conditions would be inappropriately attributed to the six home health services that were studied. 2) Because cost data were available only on home health service utilization during each episode of illness, the cost per episode calculation did not include the cost of other medical and health care services received by the patient during the same_ time

PAGE 156

137 period. Thus, it was not possible to discern the nature and extent of any substitution effects between home health services and ambulatory or institutional care. 3) Another important shortcoming of the use of 1 imi ted cost data, was the inability of the study to include other public, nonmarket, and out-of-program costs. Whether the study patients received informal care from family and friends or cash benefits, of what type and duration, was unknown. Because of these omissions, this study, along with many others, does not permit conclusions to be drawn regarding the comparative costs of home health care and nursing home care. 4) Cost per episode was calculated using the average allowable Medicare cost per visit. This did not allow the computation of actual costs at the patient level, since some visits may have taken more time than others. Although some agencies collect information on the actual time spent with the patient and in travel, these data were not avail ab 1 e for this study. Because the providers were reimbursed on the basis of the average cost per visit, not on the actual cost of the visit in question, the analysis of cost per episode was less sensitive to variations than was desirable for analysis of cost-effectiveness. Thus, the ability of the study to explain variations in cost per episode was probably reduced. Case Mix Data. Because the needs of the patient appear to most highly influence the cost per episode, it is important to recognize the limitations of this study with respect to these measures.

PAGE 157

138 1) Only limited information about the physical, functional, and psychosocial status of the patients in this study was available. No information about medical risk factors (which might complicate a case) was available. Only a general estimate of health condition, made by the discharge nurse, was used. The reliability and validity of the measure was unknown. Consequently, greater emphasis was placed in the final analysis on the AOL index as a measure of health status. However, the ADL measures used in this study were also quite simple; six items were scored on the basis of independence or dependence. No gradations in status were recorded. In addition, only the primary medical diagnosis at admission was used to describe patient -status. Other studies suggest that primary diagnos i s is not a highly useful measure of case mix in long-term care; the findings of this study substantiated this contention. 2) No measure of level of care required by each patient was available, nor were any measures of patient prob lems. Other categories of case mix descriptors found useful in predicting resource consumption were not available. Had such information been available, the ability of the study to explain cost variation might have been improved. Quality and Effectiveness Data. This category of variables suffered most from the constraints of data avai lability. The only reliable measure of program effectiveness (and i mplied quality) was the change in functional status from admission to discharge, meas ured by the six-item ADL score. This ADL index score was recorded twice for each case; no truly longitudinal measure was available in

PAGE 158

the study. 139 Consequently, it was impossible to closely trace the changes in patient status over time. In addition, no process measures of quality (or other outcome measures) were available. The strength of the study findings relating to outcome suggest that this is an important issue which should be studied in greater depth, since even the simple measures used appeared to significantly affect the analyses performed. Design Issues Several limitations of the study design precluded its ability to generalize regarding the entire home health care patient population. 1) In Philadelphia, only cases with one episode of illness during the study year were included. In Massachusetts, there were slightly over 200 cases with two episodes of illness which fell in one year. It seems logical that individuals with more than one episode of illness in a year may be more acutely ill than patients with a single episode. Therefore analysis of variance tests were completed on key patient and outcome variables to test for significant differences between cases with one episode and all others. No significant differences were found. Whether this was due to idiosyncracies in case management procedures within the providers can only be speculated. Further analysis of the cases with more than one episode of illness may be warranted on intuitive grounds. Fortunately, because their number was few, they were not expected to confound the study findings.

PAGE 159

140 2) Only cases open and closed within the study year were included in the sample. Thus those cases with extremely long lengths of use were not included. Whether the cost and utilization patterns of those cases were significantly different from those studied was not known. 3) The study design did not include a detailed analysis of the organization and management of the study providers. Since there were clear differences in the patterns of practice, cost structure, etc., across the sites, such a study component would have been most helpful in understanding the findings of the study. 4) Because most patients used several different types of home health care services, with varying intensity, it was impossible to detect the impact of any one type, or intensity, of service on the study population. Sample size was not large enough to allow analy ses to be stratified at the service level, controlling for case mix factors. Findings suggested that this would be a useful analytic approach in the future. Conclusion In spite of the methodological limitations of the study presented, it has important public policy and research implications for the development of a comprehensive, systematic 1 ong-term care system. The study findings, upon which these implications are based, are presented in the following two chapters. The final two chapters of this work discuss the implications themselves.

PAGE 160

CHAPTER IV A COMPARISON OF CASE CHARACTERISTICS Introduction The following discussion sunmarizes the relative distribution of episodes in the Phil adelphi a and Massachusetts samples in terms of key independent variables. Findings are displayed in a tabular format. Information contained therein includes frequency distribu-tions and statistical tests for proportional and mean differences.l To the extent that widespread differences existed between the two samples, some of the variation in cost per episode was explain-ed. A discussion of the relationship between key independent vari-ables and cost per episode is contained in Chapter V of this report. Patient-Specific Characteristics On the who 1 e, there were generally consistent patterns, both within and across the two S
PAGE 161

142 specific factors: (1) the most prevalent age group was the range 70-79; (2) most patients did not live alone; (3) similar types of primary diagnoses were prevalent; and (4) a consistent functional dependency pattern existed. On the other hand, there were a few differences. Philadelphia patients were generally more debilitated than Massachusetts patients, even though their average age was younger, and more of them lived with another person.2 Although the age group 70-79 was most prevalent in both samples, there were distinct differences in the age distribution of patient episodes between Massachusetts and Philadelphia, as shown in Table V. Philadelphia patients were generally younger than Massachusetts patients, with 29.0% of the episodes in Philadelphia attributed to patients in the 20-59 age category, while only 17.1% of the Massachusetts patients were in the same age range. A similar varia-tion was found at the older end of the age distribution; 15.0% of the Philadelphia patients were over the age of 80, while 24.9% of the Massachusetts patients were in the same category. The result of this difference in the age distributions was a significantly different average age for the two provider groups. Massachusetts patients were, generally, four years older than Philadelphia patients. 2one explanation for this observation is that Philadelphia patients were more acutely i 11 than their Massachusetts counterparts. In light of differences in functional ability, age, living situation, etc., this was consistent with the general finding that the average cost per episode for Philadelphia was higher than Massachusetts.

PAGE 162

Age 0-19 20-59a 60-69c 70-79b 80+a Total Mean Agea S.D. TABLE V A COMPARISON OF THE AGE OF MASSACHUSETTS AND PHILADELPHIA PATIENTS 143 Massachusetts Philadelphia N Percent N Percent 68 3.1 14 2.1 374 17.1 197 29.0 444 20.3 168 24.7 757 34.6 198 29.2 544 24.9 102 15.0 2187 100.0 679 100.0 68.74 64.55 17.23 15.92 Source: Massachusetts Discharge Summary System, Blue Cross Home Care Program Data System, and Philadelphia Abstracted Patien t Records. 01.

PAGE 163

144 Living Arrangement Another difference between Massachusetts and Philadelphia patients is illustrated in Table VI, which indicates the frequency and percentage of patients 1 iving alone or with others. In part because Philadelphia patients were significantly younger than those in Massachusetts, 20% fewer lived alone. Chapter V contains further discussion about this difference; its implications are discussed in Chapter VI and VII of this report. Primary Diagnosis There were consistent patterns, in terms of prevalence of primary diagnosis at admission, for all episodes in this study (see Table VII). Circulatory system disorders and neoplasms were the most often occurring primary diagnoses in both samples. Several other diagnoses were also prevalent in both samples, particularly digestive system and endocrine disorders. There were far fewer nervous system disorders, musculoskeletal disorders and accidents in Philadelphia than in Massachusetts. The differences in the distributions were primarily accounted for by the significantly greater percentages of patients with primary diagnoses of neoplasm and circulatory system disorders in Philadelphia than in Massachusetts. This clustering of diagnoses may be an indication of a more highly specialized program in Philadelphia than in Massachusetts. Functional Status Philadelphi. a patients tended to be, on the average, more debilitated than Massachusetts patients at admission. As shown in Table VIII in terms of functional status (ADL) scores, a significantly

PAGE 164

TABLE VI A COMPARISON OF THE LIVING ARRANGEMENT OF MASSACHUSETTS AND PHILADELPHIA PATIENTS Living Massachusetts Philadelphia Arrangement N Percent N Percent A1onea 617 28.3 56 8.2 With re1ativesa 1472 67.3 596 87.9 Other 95 4.4 26 3.9 Total 2186 100.0 678 100.0 145 Source: Massachusetts Discharge Summary System, Blue Cross Home Care Program Data System, and Philadelphia Abstracted Patient Records.

PAGE 165

TABLE VII A COMPARISON OF THE PRIMARY DIAGNOSISa OF'MASSACHUSETTS AND PHILADELPHIA PATIENTS Primary Massachusetts Philadelphia 146 Diagnosis N Percent N Percent Infectionc 45 2.1 4 0.6 Nepolasm b 295 13.5 150 22.1 Endocrine 160 7.3 47 6.9 Bloodb 53 2.4 4 .6 Mental 49 2.2 8 1.2 Nervousb 209 9.6 12 1.8 Circulatoryb 516 23.6 237 34.9 Respiratory 87 4.0 36 5.3 Digestiveb 233 10.7 39 5.7 Genitourinaryb 56 2.6 1 . 1 Skinb b 61 2.8 5 .7 Musculoskeletal 233 10.7 16 2.4 166 7.6 17 2.5 Othe 21 1. 0 102 15.0 Total 2184 100.0 678 100.0 Source: Massachusetts Discharge Summary System and Blue Cross Home Care Data System. aA11 primary diagnoses were grouped according to the standard International Classification of Diseases, (8th Edition). c .

PAGE 166

ADL TABLE VII I A COMPARISON OF THE INDEPENDENCE IN ACTIVITIES OF DAILY LIVING OF MASSACHUSETIS AND PHILADELPHIA PATIENTS Massachusetts Philadelphia 147 Category N Percent N Percent Bathing a 1198 55.0 290 42.7 Dressing a 1395 64.0 280 41.2 Ambu1ation a 1562 71.8 395 56.7 Toileting a 1651 75.7 365 53.8 Transferring a 1651 75.7 391 57.6 Eatingb 1809 82.9 594 87.5 Total 2190 100.0 679 100.0 Source: Massachusetts Discharge Summary System, Blue Cross Home Care Program Data System, and Philadelphia Abstracted Patient Records.

PAGE 167

148 1 ower percent age of patients was independent in their ADL score at admission in Philadelphia than in Massachusetts, for all ADL categories except eating. In fact, the highest percentage of patients in both samples was independent in the eating category. Indepen dence in transferring (in and out of a bed or chair) was second in prevalence, while the most difficult areas of functioning for all patients were bathing and dressing. The similarity between the two patient groups in eating was not surprising, since it is generally the 1 ast function to be affected by ill ness. These patterns are consistent with findings of other studies of functional status and wjth Katz• hypotheses concerning the relationships of the various components of the ADL index.3 Functional Status Index with the results for the individual ADL categories, Philadelphia patients were more functionally dependent than Massa-chusetts patients when compared on the basis of the overall functional status (ADL) index score at the time of admission. Table IX compares the ADL index scores of both samples. When the first two and the last two groups were collapsed, the difference in the distributions between the two data sets was easily discernible. Over 60% of the Massachusetts patients were independent in at least five of the ADL components, while only 40% of the Philadelphia patients 3sidney Katz et al., 11Studies of Illness in the Aged: The Index of ADL,11 Journal of the Jlmerican Medical Association, CLXXXV (1963), 914; S1dney katz and C. AITiechl Akpom, 11Index of AOL,11 Medical Care, XIV (1976), 116-119.

PAGE 168

TABLE IX A COMPARISON OF THE ACTIVITIES OF DAILY LIVING INDEXa OF MASSACHUSETTS AND PHILADELPHIA PATIENTS Massachusetts Phi.ladelphia ADL Index I N Percent N Percent Independent in 6 ADLsb 1086 50.0 259 38.1 Independent in 5 ADLsb 227 10.5 17 2.5 Independent in 4 ADLsc 214 9.9 89 13. 1 Independent in 3 ADLsc 106 4.9 23 3.4 Independent in 2 ADLs 102 4.7 32 4.7 Independent in 1 ADL b 179 8.2 177 26.1 Independent in None 257 11.8 82 12. 1 Total 2171 100.0 679 100.0 b Mean ADL' . 1. 76 2.61 S.D. 2.22 2.39 149 Source. Massachusetts Discharge Summary System, B lu e Cross Home Care Program Data System, and Philadelphia Abstracted Patient Records. aScore computed in such a way that a score of zero represents total independence while a score of six represents total dependence.

PAGE 169

150 were as independent. On the other hand, almost 40% of the Phi 1 adelphia patients were totally independent in, at most, one of the ADLs categories, while only 20% of the Massachusetts patients were so debilitated. Surgical Procedure The difference in functional status between the two samples may be accounted for partially by the presence of a surgical procedure prior to admission. The frequency of such an occurrence was not significantly different in the t\\Q samples {p = .088); 35.6% of Massachusetts and 39.2% of Phil adelphi a episodes were admitted to home care with surgical diagnoses. Patient's Location Prior to Admission The location of each patient prior to home care admission may be reflective, in some way, of his health care needs. Although admission directly from an acute care facility was the most prevalent in both samples, the distributions were significantly different. Table X illustrates this point. Almost all of the cases in Philadeiphia were admitted directly from a hospital to home care, while only 50.8% of Massachusetts patients were admitted directly from an acute care facility to a home health program. This finding was consistent with conditions discussed earlier regarding the pos sible specialized nature of the Philadelphia program. Goal at Admission A final patient-specific characteristic, available only for the Massachusetts cases inthis study, was the goal at admission as

PAGE 170

TABLE X A COMPARISON OF THE PREADMISSION LOCATION OF MASSACHUSETTS AND PHILADELPHIA PATIENTS Preadmission Massachusetts Philadelphia 151 Location N Percent N Percent Acute Carea 1110 50.8 652 96.2 Home a 865 39.6 27 3.8 Long-Term Care 172 7.9 0 0 Other 40 1.8 0 0 Total 2187 100.0 679 100.0 Source: Massachusetts Discharge Summary System and Blue Cross Home Care Program Data System.

PAGE 171

152 determined by the care provider. Most episodes had goals which were oriented toward the care of continuing problems (e.g., self-care (28.2%) or maintenance (27.8%)). Goals based on expectations of improvement comprised 36.0% of the episodes; recovery was the goal of 23.3% and rehabilitation was the goal of 12.7% of the episodes. Of all episodes, 4.0% were classified as terminally ill at time of admission. These last t'ftO goal classifications (rehabilitation and terminally ill) were singled out for further investigation in this study. A discussion of the findings of these analyses is contained in Appendix C. Outcome-Related Characteristics Although genera 1 patterns of patient outcomes were similar in the Massachusetts and Philadelphia samples (e.g., more patients improved or stayed the same than not), a significantly higher proportion of cases improved in Philadelphia than in Massachusetts. Change in Functional Status Index As illustrated in Table XI, most cases in Philadelphia (59.1%} and in Massachusetts (73.6%) went unchanged in their ADL index scores from time of admission to discharge. The next most prevalent cagetory of change in both samples was improvement, and the least prevalent category was deterioration. There was no significant difference between the two populations with respect to deterioration in functional status over time, but the difference in the other two categories was significant (p .001). Philadelphia patients reduced their dependency by .922 units (where each of the six

PAGE 172

TARLE Xf a A COMPARISON OF THE CHANGE IN ACTIVITIES OF DAILY LIVING INDEX MASSACHUSETTS AND PHILADELPHIA PATIENTS Change in Massachusetts Philadelphia ADL Index N Percent N Percent No Changeb 1555 73.6 401 59.1 Improvedb 386 18.3 230 33.9 Deteriorated 171 8.1 47 6.9 Total 2112 100.0 678 100.0 Mean Changeb -.189 -.922 S.D. 1. 534 2.105 153 Source: Massachusetts Discharge Summary System, Blue Cross Home Care Program Data System, and Philadelphia Abstracted Patient Records. aChange score computed by subtracting admission ADL Index Score from discharge score (with a zero score representing total independence and a score of six representing total dependence). Thus, the lower the mean ADL change score the more improvement from admission to discharge, e .g., the greater the reduction i n functional dependency.

PAGE 173

154 component i terns was assigned one unit v a 1 ue for dependency) , as compared to an improvement of .189 for Massachusetts patients. These findings indicate, in general, significantly greater improvementon the part of Philadelphia patients. Health Status at Discharge The health status at discharge distributions (Table XII) showed similar patterns to those found in the change in ADL index variable. For example, a significantly higher proportion of patients in Phila delphia than in Massachusetts were improved upon discharge, while the percentage of patients worsening or dying was similar. Thus, the general outcomes of Philadelphia episodes were better than Massachusetts' .4 This finding was not unexpected since greater improvement could be generally expected in an acutely ill population as compared to one which is predominantly chronically ill. Intensity of Care and Length of Use Table XIII presents the utilization-related outcome character-istics examined for the two samples which were the average number of visits per day (intensity) and the average length of use of home care. These characteristics were important since they assisted in the interpretation of patient-specific and cost analyses findings. 4 since this issue warranted further analysis, a discussion of outcomes, while controlling for other patient-specific characteristics, was included in Chapter V of this report. That discussion indicates that the general patterns outlined here were consistent, even when other factors were taken into consideration.

PAGE 174

Health Status Improved a Same a Worsened Expired Total TABLE XI I A COMPARISON OF THE HEALTH STATUS AT DISCHARGE OF MASSACHUSETTS AND PHILADELPHIA PATIENTS Massachusetts Philadelphia N Percent N Percent 977 45.6 460 67.7 641 30.0 62 9.1 437 20.4 121 17.9 85 4.0 36 5.3 2140 100.0 679 100.0 155 Source: Massachusetts Discharge Summary System, Blue Cross Home Care Program Data System, and Philadelphia Abstracted Patient Records.

PAGE 175

TABLE XIII A COMPARISON OF THE MEAN INTENSITYa AND THE LENGTH OF USEb OF MASSACHUSETIS AND PHILADELPHIA PATIENTS Massachusetts Philadelphia Length of Use (days) Mean Mean 156 N Percent Intensity N Percent Intensity 1-9 623 28.5 .84 79 11.6 .74 10-29 513 23.5 .48 276 40.6 .55 30-59 366 16.7 .38 246 36.2 .45 60-89c 185 8.5 .33 60 8.8 .53 90+ c 500 22.9 .22 18 2.7 .47 Total 2187 100.0 • 50 679 1 00.0 .53 Mean LoUc 77.94 33.48 S.D. 143. so 22.84 Source: Massachusetts Discharge Summary System and Blue Cross Home Care Program Data System. alntensity is defined as the total number of v i s its per d ay. blength of use is the number of days from admission to discharge.

PAGE 176

157 In both instances, the patterns were significantly different in Philadelphia and Massachusetts. The average number of visits per day for Philadelphia patients was .53, while the average number of visits per day for Massachusetts patients was .SO (p .05). These values were similar, although statistically different, because of the bimodal distribution of length of use in the Massachusetts sample which caused the average length of use to be significantly longer for Massachusetts than Philadelphia episodes. Subsequent computation of the intensity variable produced results which indicated greater similarity than would otherwise be expected in the two samples. These differences in intensity and length of use were partially accounted for by differences in case mix between the two samples. This issue is explored more fully in Chapter V. Provider/Health System Characteristics The major factor of interest in this section is the primary source of payment for each episode of care, si nee each of the other provider/health systen variables (i.e., size, density and hospital admissions) were not patient-or episode-specific (i.e., they had only eight values, corresponding to the eight providers in the study samples. Primary Source of Payment Table XIV illustrates the difference in primary source of payment across the two samples. Medicare was the domi :1ant payer in both samp 1 es. The primary differences between the two samp 1 es were a lower prevalence of Medicaid and private pay episodes and a higher

PAGE 177

Source of Payment Medicare Medica ida Insurance a Private Pay Other a Total TABLE XIV A COMPARISON OF THE PRIMARY SOURCE OF PAYMENT OF MASSACHUSETTS AND PHILADELPHIA PATIENTS Massachusetts Philadelphia N Percent N Percent 1385 63.4 417 61.4 247 11.3 34 5.0 253 11.6 223 32.8 a 201 9.2 4 .6 98 4 . 5 1 • 1 2184 100.0 679 100.0 Source: Massachusetts Discharge Summary System and Blue Cross Home Care Program Data System. 158

PAGE 178

159 prevalence of third-party insurance (including Blue Cross) in Philadelphia than in Massachusetts. Because Philadelphia patients were, on the average, younger than Massachusetts patients, with a greater probability of subscribing to the Blue Cross home care benefit p1an, those episodes reimbursed by private insurers (rather than Medicaid) were greater in nlJTiber. In addition, the low reimbursement rates of the Pennsylvania Medicaid program may have partially accounted for the significant differences in the payer mix across the two samples. Conclusion Generally, there were significant differences between the Massachusetts and Philade l phia samples with respect to all three categories of independent variables. How these differences influenced the cost of care is the subject of the next chapter.

PAGE 179

CHAPTER V UTILIZATION AND COST FINDINGS Introduction The purpose of this chapter is to address two key questions: How were the cost and utilization of home health care influenced by patient, provider, and community characteristics? of the characteristics studied were the most important in terms of the observed utilization and cost variations? To address these questions, this chapter is organized in the fol-lowing manner. A general description of the average m.mber of visits and cost per episode for the two samples is presented first. A discussion follows which describes the analyses used to test the overall relationship between the cost per episode and all three categories of independent variables (i.e., patient-specific, outcome-related, and provider/health system characteristics).l Then information about the relative importance of each of the individual lThe findings in this chapter focus on the principal dependent variable, cost episode; because it was, as expected, highly correlated (R = .97), with the secondary dependent variable, total number of visits per episode, given the use of average costs per visit as input prices. To the extent that findings differed for the two dependent variables, they are discussed in the body of the text. Appendix D contains a general discussion of the relationships observed between the total niJllber of visits per episode and the three categories of independent variables.

PAGE 180

161 categories of independent variables with respect to its ability to explain variation observed in the dependent variable is presented. The next section identifies those variables within each of the inde pendent variable categories, most significant in explaining differences in cost and utilization. Findings of the regression analyses are also included. The next discussion presents the findings related to a comparison of the outcomes of care between the two samples. Finally, the overall cost and utilization findings of the study are summarized. Overall Uti lization and Cost Per Episode Findings There was a greater degree of similarity between the two samples, in terms of total visits per episode than cost per episode. On the average, patients in Massachusetts and Philadelphia received the same total number of visits per episode; the subsequent cost of home health care was not, on the average, similar. These findings are presented and discussed in this section. Utilization Findings Table XV sl.lllmarizes the utilization findings of this study. The average number of visits per episode in both samples was not significantly different. On the average, patients received 18.25 visits per episode in Massachusetts and 16.35 visits per episode in Phi 1 ade 1 phi a. , 1\lthough there was no significant difference between the means of the two samples, the standard deviation for the Massachusetts episodes was twice as large as Philadelphia's. The chi-square test

PAGE 181

Visits per Episode 1-10 a ll-20a 21-4rf 41-60 61+ a Total Mean Visits S.D. TABLE XV A COMPARISON OF TOTAL VISITS PER EPISODE OF MASSACHUSETTS AND PHILADELPHIA PATIENTS Massachusetts Philadelphia N Percent N Percent 1250 57.2 293 43.2 378 17.3 215 31.7 310 14.2 123 18.1 115 5.3 32 4.7 l31 6.0 16 2.4 2184 100.0 679 100.0 18.25 16.35 29.49 14.39 Sources: Massachusetts Discharge Summary System and. Blue Cross Home Care Program Data System. 162

PAGE 182

163 results, identified in the footnotes on the table, indicate whether or not the proportions of episodes in any visit per episode category (e.g., the 1-10 visit category) were significantly different across the Massachusetts and Philadelphia samples. The majority of patients in Massachusetts (74.5%), and a similar proportion of patients in Philadelphia (74.9%) received less than 21 visits per episode. Far fewer patients in Philadelphia (2.4%) than in Massachusetts (6.0%) received more than 60 visits per episode. This finding does not take into account service mix or intensity of care, both of which are discussed later in this chapter. Cost Findings Subsequent analyses revealed significant differences in the average cost per episode between the t....o samples. These findings are presented in Table XVI. The average cost per episode in Massachusetts was significantly lower than the average cost per episode in Philadelphia (p .001). In Massachusetts the mean was $257.39, with a standard deviation twice as large; in Philadelphia the mean was double that of Massachusetts ($522.11), but the standard deviation was less than the mean. Although the costs in Phil adelphi a averaged twice those in Massachusetts, the lower value for the standard deviation in Philadelphia indicates that costs were less widely dispersed about the average there than in Massachusetts. This finding, coupled with the preceedi ng one, suggests greater uni formity of case mix and service patterns in the Philadelphia sample than in Massachusetts.

PAGE 183

TABLE XVI A COMPARISON OF THE COST PER EPISODE FOR MASSACHUSETTS AND PHILADELPHIA PATIENTS Cost Per Massachusetts Philadelphia 164 Episode ($) N Percent N Percent J 0-100 a 1048 48.0 31 4.6 101-300 618 28.3 223 32.8 301-500 235 10.8 169 24.9 501-700 102 4.7 99 14.6 701-900 58 2.7 59 8.7 900+ 123 5.6 98 14.4 Total 2184 100.0 679 100.0 Mean Cost a $257.39 $522.11 S.D. 488.30 441.96 Source: Massachusetts Discharge Summary System, Medicare Cost Reports, Blue Cross Home Care Program Data System, and Blue Cross of Greater Philadelphia Home Care Program Director. a

PAGE 184

165 Cost Findings Discussion Since there was little variation across the two samples in terms of the total nlJllber of visits per episode, the variation observed in the cost per episode across the samples warrants further discussion. It is important to point out that it is not the intention of this discussion to justify the differential in cost between the Massachusetts and Philadelphia samples, but rather to provide information upon which polic)111akers can subsequently evaluate the relative merits of such differences. The differences may be cate-gorized into six areas: 1) charges passed through as costs, 2) unaudited cost reports, 3) administrative surcharges, 4) case mix differences, 5) serv1ce mix differences, and 6) variations in outcome. Charges Passed Through as Costs. The vast majority of nursing and therapeutic visits in Philadelphia were provided by community based agencies under contract to the hospita 1-b ased programs. The intent of such arrangements was to combine a rehabilitative approach, generally found in community health nursing care, with the close, professional monitoring and coordination of services often provided by hospital-based programs. One criticism of hospital-based programs often levied by providers affiliated with freestanding agencies is that hospital-based programs focus on acute, medically-oriented nursing care. The contractual arrangements

PAGE 185

166 negotiated in Philadelphia many years ago with the aid of Blue Cross of Greater Philadelphia were intended to address this criticism. Unfortunately, the result of the arrangement was that the charges of the community-based agencies became costs to the hospital-based pro grams, which were then passed on to third party payers who would not have otherwise paid full charges. The actual effect of this arrangement on total costs could not be determined in this study because of data constraints. However, estimates ranged from a 15% to 20% increase in total cost per unit of service. Unaudited Cost Reports. The cost per unit of service data were obtained from the Blue Cross of Greater Philadelphia Data System which abstracted cost data from unaudited Medicare cost reports. To the extent that una 11 owab 1 e costs were reported, the Phi 1 ade 1 phi a costs per unit of service (and cost per episode) were overstated when compared to Massachusetts costs. In recent years the amount of unallowable costs averaged around 5%. Administrative Surcharges. A major portion of the difference in cost per episode between the two samples was accounted for by the administrative surcharge which was added to the cost of home health care visits by Philadelphia providers. For some episodes, the amount of the surcharge was almost as much as the total cost of nursing and therapeutic visits. The administrative surcharge was computed on the basis of total nlJTiber of home health visits per episode (at a fixed rate) and included the direct and indirect costs of the hospital-based program. These overhead costs for the

PAGE 186

167 Philadelphia programs included the items listed below which were not typically incurred by community-based agencies. 1) Hospital Administrative Expenses. The expenses of the bospital' s non-revenue producing cost centers were allocated to the home care department and a 11 other revenue producing cost centers. This included a portion of such costs allocated to the other hospital departments from which the home care department obtained direct services (e.g., central supply). Thus, the home care department was directly allocated a portion of the hospital's general administrative expenses because it was a revenue producing cost center, and was also indirectly allocated an additional portion of the hospital's overhead through its portion of the costs of the nonrevenue producing departments. 2) Patient Care Planning. The hospital-based programs pro-vided a patient care planning function for the hospitals with which they were affiliated, the cost of which was allocated to the home care department. The administrator of each hospital-based program received a copy of the patient record for every admission to the hospital. Daily rounds were made to visit patients and review charts in an effort to identify candidates with potential for early discharge to a home care progrCITl. Because the hospital-based programs tended to admit the more medically intensive, acutely ill patient who needed close monitoring, they referred twice as many patients directly to community-based agencies than they referred to themselves. The administrative overhead for this function was a direct cost to the home care department and thus allocated to the

PAGE 187

168 visits provided to patients of the hospital home care Hence, the administrative costs which were added to the cost of visits purchased by the hospital-based programs included the cost of reviewing each hospital admission and referring appropriate cases to community-based agencies outside of the hospital's jurisdiction. This situation resulted in higher costs for the Philadelphia hospital-based programs.2 3) Professional Coordination. The costs of professional coordination of all services, including ancillary services not reim-bursed by Medicare or Medicaid, were included in the administrative surcharges. Case managers in the hospital-based programs maintained daily contact with care-givers and retained copies of all service records for review in the programs' offices. Since the Philadelphia programs also provided transportation to bring patients to the hasPit a 1 for needed services and de 1 i ver other spec i a 1 services in the patients's home, the cost of coordination of these services was also included in the administrative surcharge. The result was a higher cost per episode for Philadelphia providers. Case Mix Differences. Another reason that the aver age cost per episode in Philadelphia was significantly greater than in Massachu-setts was the variation in case mix between the t'ftQ groups. The data presented in Chapter IV show that the Philadelphia case mix was substantially different from that of Massachusetts. \.Jhen com-pari sons were made in patient-specific characteri sties across the 2 For example, one hospital in Philadelphia referred over 300 patients directly to community-based agencies while admitting less than 150 patients to its own home care department in the study year.

PAGE 188

169 two samples, Philadelphia consistently exhibited a more debilitated, acutely ill case mix. Although its patients were younger, their functional status was more impaired, and their diagnoses were concentrated in a few categories (which potentially required more intensive services). A1most the entire Philadelphia patient population was admitted to the home care program directly from the hospital; consequently, it was not surprising that a substantially larger portion of the Philadelphia sample was admitted with a surgical diagnosis which may, in turn, have led to higher costs. When controlling for diagnostic factors, the average number of visits per day was significantly higher in Philadelphia than in Massachusetts. Conversely, the average length of use in Philadelphia was significantly shorter than in Massachusetts. Findings of the overall regression analyses of this study suggested that intensity of service utilization, general resource consumption, and the necessity to closely monitor a patient's condition may have accounted for some of the increased costs of home care. Service Mix Differences. In addition to, and partially because of, a more acutely ill patient population in the Philadelphia sample, the service mix of the two provider groups was significantly different. Several of the Massachusetts agencies provided only a limited number of therapeutic visits during the study year; in addition, they provided a substantially greater nlJTlber of home health aide visits than did the Philadelphia providers. The average number of skilled nursing visits per episode for Massachusetts cases (11.37 visits) was not significantly different from that of Philadelphia

PAGE 189

170 cases (12.38 visits), but the average number of therapy visits was different. Approximately twice as many physical and speech therapy visits were provided, on the average, to Philadelphia patients, and more than nine times as many occupational therapy visits. Although Massachusetts cases received fewer therapy visits, they received almost four times as many home health aide visits, on the average, as Philadelphia cases (5.10 versus 1.30 visits). Hence, there was a significantly different service mix in the two samples. When the cost of services was taken into account, the variation in cost per episode was partially explained. In Philadelphia, the unit cost of therapies was higher than the unit cost of nursing or home health aide care. Findings from supplemental analysis indicated that since Philadelphia provided therapeutic services to 12% more patients than did Massachusetts, the provision of a greater number of therapeutic services resulted in a higher average cost per episode than for Massachusetts agenc ies. Massachusetts agencies provided a similar number of visits, but with an emphasis on nursing and home health aide care. When the preceding two factors were taken into consideration, it was apparent that Philadelphia providers treated patients who were, as a group, more. acutely ill and received more highly intensive therapeutic services than did Massachusetts patients. This finding suggested that, in the case of Philadelphia providers, home care was used as an alternative care modality, not only for chronically ill long-term care patients, but also for the treatment of post-acute short-term hospital patients. Home health care in Philadelphia could thus be viewed as an alternative to acute inpatient

PAGE 190

171 care, thereby emphasizing the need to control for case mix in any future comparisons of costs. Variations in Outcome. If the variation in service mix and other resource uti 1 i zat ion contributed to better outcomes for the Philadelphia patients as noted earlier, another possible explanation for Philadelphia's higher costs is provided. A later section of this chapter addresses this topic in more detail. Analyses to Explain Overall Variation Within Each Sample For the majority of home health patients, factors other than those which were included in this study affected the cost of care. The low multiple correlation coefficients resulting from the overall regression analyses provided evidence for. this finding. When analyses were performed on two subsets of patients, the ability of the equations to account for variati ons in the cost per episode increased in one case and decreased in another. For patients with rehabilitation as the goal at admission, the factors available in this study explained a greater proportion of the variation than for the other patients. In contrast, when analyses were performed on the subset of patients categorized as terminally ill at admission, the ability of the equations to account for variation in the dependent variable was less than for the remainder of the patient population. Approach In order to test the ability of the three categories of inde pendent variables to explain variation in the dependent variable,

PAGE 191

172 several regression analyses were conducted as part of this study. Although the specific variables in each of the equations and the samples varied, each of the regression equations could be depicted in the following functional form: CPE = f (.f.,]_,.!!) ' where, CPE = Cost Per Episode p = Patient-Specific Characteristics 0 = Outcome-Related Char acteri st i cs H = Provider /Health System Characteristics The three groups of characteristics represent various combinations of the independent variables included in each category and previous-ly described in Chapter III. The approach used was basically statistical, allowing an empir-ical estimation of the general form of the equations previously described. Ordinary 1 east squares (forced and stepwise) regression were used for all the analyses.3 Because early descriptive analyses indicated significant variation in case mix and provider character-istics for the t\'tO samples, all equations were estimated for each 3 simultaneous equation method s for estimating regression equations were considered due to the presence of joint dependent, or endogenous, variables. Consultations with experienced statisticians indicated that simultaneous equation methods would add little to the policy implications of this study. Hence, a decision was made to use ordinary least squares.

PAGE 192

173 sampJe (Philadelphia and Massachusetts) separately. Consequently, the reliability of patterns observed in cost per episode and in the re 1 at i ve impact of the independent vari ab 1 e categories was increased, to the extent that results of the estimated equations were consistent across both data sets. As outlined previously, the conceptual and analytic framework which served as the basis for these analyses required the estimation of regression equations in a staged process. Consequently, the regressions were estimated first, using each of the three sets of independent variables separately. Then, (in part, based on the results of the canonical correlation analysis) the independent variables were entered in the fo11owing sequence: patient-specific variables, provider/health system variables, and outcome-related variables. The resulting analyses allowed for the estimation of the relative impact of the three categories of variables on the dependent variable. When taken together, the overall ability of the independent variables to explain variation in the dependent variable could be assessed. All equations were estimated for the sample of 2,101 Massachusetts episodes and for the 697 Philadelphia episodes. In addition, to test the ability of the independent variables to explain variation in cost for particular patients, equations were estimated for two subsets of Massachusetts patients: those who were categorized at admission with a goal of rehabilitation, and those categorized as

PAGE 193

174 having a terminal i11ness.4 Additional analyses were completed on other subsets of the sample population and are also described. Overall Findings The analyses to determine the impact of the independent vari-ables on the cost per episode indicated that data available on characteristics typically believed to account for variation in home health care costs at the patient level were inadequate to explain variations observed. In Massachusetts, the overall ability to explain variation in the final equations was slightly lower (R2 = .099) than in Philadelphia (R2 = .129) .5 Although all equations were significant at the .001 level, the small R2 for both samples . indicated the minimal explanatory power of the equations. Thus, when using all three categories of variables together, the vast majority of variation in the dependent variable remained unexplained.6 There were several possible explanations for this finding. They are discussed in a later part of this chapter. Its implication for policymakers and analysts is significant and is discussed in Chapter VI. Appendix E provides a detailed discussion of the multivariate analyses. 4since these data were not available at the time of the study for the Philadelphia sample, analyses on these subsets of cases were carried out for the Massachusetts sample only. 5The value of the R2 represents the proportion of variation in the dependent variable which could be explained in terms of all the independent variables contained in any equation. 6when selected equations were estimated using data from each provider separately, the R2 of several of the equations increased slightly (e.g., R2 = .32 and R2 = .34), indicating there were variations in patterns of practice between providers in each The relative significance of the independent variables was unchanged.

PAGE 194

175 Rehabilitation Patients. When equat i ens were estimated for rehabilitation patients in the Massachusetts sample, the ability to explain the variation in the utilization and cost of home care improved. A significantly larger portion of the variation was accounted for by the three categories of vari ab 1 es for this sarnpl e than for the total sample (R2 = .265). The higher R2 value suggests that factors such as functional status at admission, age, and di agnes is were better predictors of cost for rehabilitation patients than for others. This is, in part, because patterns of practice are generally less discretionary, more uniform, and more precisely defined for rehabilitation patients than for other types of patients receiving home care. This finding substantiates the contention that cost projections for this population may be more reliable than most . Terminally I ll. When only those patients categorized as terminally ill at admission were selected from the Massachusetts sample, the result was 87 cases, 90% of which were categorized with the diagnosis of neoplasm. The average cost per episode for this sample of patients was slightly lower than the average cost for all other Massachusetts patients, partly because not all terminally ill patients died at home. Most patients in Massachusetts were admitted to an acute care facility before death, which resulted in the shift of service costs from home care to the inpatient facility. The result was a lower average cost per episode. When a subset of patient-specific and pro vi der/hea 1 th systems vairables was included in regression analyses for the terminally ill, the overall ability to explain variation in cost per episode

PAGE 195

176 was low (R2 = .089) and not significant. None of the independent variables included in the equation were significant in explaining variations observed. One reason for this may be the lack of uniform and/or consistent approaches to caring for the terminally ill patient. In addition, since there was almost no potential for rehabilitation, the number of nursing visits provided was generally more dependent on the care-giver and the providers pattern of practice than on any patient-specific condition. Clearly, traditional patient descriptors were inadequate to explain much variation in this subsample of patients. The implications of this finding for the expansion of home health care services to the terminally ill are considerable since current estimates of the costs of hospice-type care are based on these factors. This finding suggests that such estimates of costs warrant additional analyses before they become the basis o.f policymaking. Analyses to Test the Relative Importance of Each Independent Variable Category The major purpose of the final regression analyses was to examine variations in cost per when all three categories of independent variables were included in the equations. This approach first estimated equations using ' each of the independent variable categories alone, then combined patient-specific and provider/health system variables to estimate their impact, and finally, used all three groups together. In so doing, the relative impact of each of the variable categories and its components could be determined, taking into account other characteristics.

PAGE 196

177 The findings of the separate regression analyses previously described indicated that for the Massachusetts sample, patientspecific variables accounted for slightly more of the variation than did the outcome-related variables. The other category of indepen dent variables, provider/health system characteristics, account7d for the smallest portion of the variation in cost per episode. For the Philadelphia sample, the patient-specific variables again accounted for the greatest anount of variation in cost per episode, followed by the outcome variables. These findings were consistent throughout the staged analyses when all sets of variables were included. Those independent variables which were highly significant and contributed most to the overall R2 tended to retain their rel a-tive importance in all equations. Analyses to Test the Patterns Withi n Each Independent Variable Category The relationship between home health care patient, outcome, and provider characteristics, and the cost of care is examined in this section. The main focus is on the variation in the average cost per episode in terms of the categories of the independent variables . It should be noted that consistency of patterns across the two samples was of prime importance in this study, rather than the absolute differences between the sanples. However, when particularly important, the narrative focuses on distributional differences within and across the samples.

PAGE 197

178 Patient-Specific Characteristics A clear understanding of the nature of the relationship between patient-specific characteristics and the cost of home health care is essential to the develollllent of a cost-effective health care deli-very system. A basic assumption which underlies current contentions that home health care is less expensive than institutional care for all but the most impaired patients is that the costs of home health care are uniformly low for patients of varying ages, diagnoses, and living arrangements . That a strong relationship exists between patient characteristics and costs is the basis of recent long-range projections of total home health care expenditures.? This study was, in part, intended to test these assumptions . A list of the key variables discussed in this section follows, as does a discussion of regression findings which highl i ght impor-tant patient-specific variables. Then, the descriptive analyses identify those key patient-specific character istics which affect cost. The regression findings specific to each of the variables are also included. The variables discussed include: Age Living Arrangement Primary Diagnosis Functional Status Functional Status Index Goal at Admission 7u.s., Congress, Congressional Budget Office, Long-Term Care for the Elderly and Disabled (Washington, D.C.: Government Pri nting Off1ce, 1977), pp. 7-14.

PAGE 198

179 Overview of Multivariate Findings. As noted above, patient-specific variables accounted for the greatest portion of the variation in cost per episode in both samples, although the size of the overall R2 in both cases was small. The significance of a few of the variables was noteworthy. For the Massachusetts sample, several of the patient-specific variables had significant impact on cost per episode. Age was significant when patient-specific variables were the only variables included in the regressions. However, age was not significant when all variable groups were included in the equations because it was highly correlated with the dichotomous variable which signified Medicare as a source of paj111ent for the episode . Medicare was highly significant in the final equations. Whether or not a patient lived alone was also positively related to cost, as was the presence of a preadmission surgical diagnosis. Another highly significant patient-specific variable was the functional status at admission, which was positively related to the cost per episode ( i . e. , the more dependent the patient, the greater the cost). T....o diagnostic categories, neoplasms and blood system disorders, were also significant in their ability to explain variation, and were positively related to cost per episode. For the Philadelphia sample, the effect of patient living arrangement was not significant in explaining variations observed. The only diagnostic variable which was significant was accidents. Accidents exhibited a positive effect on the cost per episode, probably due to the generally traumatic nature of the resulting

PAGE 199

180 injuries.8 Functional status of the patient at admission and the presence of a surgical procedure were highly significant and posi-tively associated with cost per episode for the Philadelphia sam-ple. As measured by the size of the regression coefficients, the presence of a surgical procedure had greater impact for Phi 1 ade 1-phi a cases than for Massachussets, whereas the impact of functional status was similar in both samples. Analyses on subs amp 1 es of Massachusetts cases revea 1 ed contra-dietary findings. For the equations which used the subsample cate-gorized as having a goal of rehabilitation, patient-specific vari-ab les accounted for the majority of overall explanatory power ( R2 = .221) when inc 1 uded in equations by thenselves. When all three categories of variables were included, the R2 was 1 arger ( . 265) and remained significant ( p = . 001) . Seven of the 11 patient-specific variables were significant for this subset of patients. These inc 1 uded: patient's living arrangement; funct i ana 1 status of the patient at admission; and, diagnoses of endocrine dis-orders, circulatory system disorders, respiratory system disorders, musculoskeletal system disorders, and accidents. For the equations using the termi n a 11 y ill as the subs
PAGE 200

181 The remaining portion of this section presents the results of the descriptive and multivariate analyses for each of the key patient-specific variables. The sample sizes presented in the tables vary because the number of episodes with missing data was different for many of the variables. The F-test (ANOVA) results shown in the tables represent the significance of the relationship across average cost per episode and the independent variable categories for each sample. The t-tests in the tables compare the mean value for each category of average cost per episode across samples, while the chi-square tests compare the differences in each sample between the independent variable categories in the distribution of episodes by cost range categories. Table XVII illustrates consistency across both samples in the re 1 at ion ship bet ween the cost per episode and the age of the patient. The general pattern was for cost to increase with age. In both the Massachusetts and Philadelphia samples, significant coefficients were exhibited for the correlation of age with cost per episode ( .074 and .077 respectively). In Massachusetts, the youngest age group was significantly lower than the older age groups in cost per episode. In Philadelphia, the youngest age group was significantly different from and lower than all other age groups. In general, the younger ages were less costly to treat than the older. This relationship was especially noticeable in Massachusetts. Age was a highly significant variable in the regression findings when patient-specific variables were used alone with cost per episode. When provider/health system variables were added to the

PAGE 201

( TABLE XVII A COMPARISON OF THE AVERAGE COST PER EPISODE BY THE AGE OF MASSACHUSETTS AND PHILADELPHIA PATIENTS Massachusetts Philadelphia Age N Mean Cost S.D. N M ean Cost 0-19b 68 $162.90 257.08 14 $ 349.68 20-59a 373 208.25 344.01 198 492.43 60-69a 443 245.43 387.49 168 488.61 70-79a 755 269.54 444.54 197 598.46 80+a 542 296.77 564.34 102 511 . 04 Total a 2181 $ 257.60 448.57 679 $ 522.10 182 S.D. 183 .'39 420.05 402.81 503.09 429.00 411.95 Source: Massachusetts Discharge Summary System, Medicare Cost Reports, Blue Cross Home Care Program Data System, and Blue Cross of Greater P hiladelphia Home Care Program Director.

PAGE 202

183 analyses, age dropped out as a significant variable in favor of Medicare as the primary source of paj111ent due to the high collinearity between the two variables noted above. However, irrespective of the relative significance of the age variable, older patients tended to cost more than younger ones. Living Arrangement. In a pattern which proved relatively con-sistent across the samples, the difference between the average cost per episode for patients who lived alone compared to those who did not was significant. Table XVIII illustrates this finding. A strong relationship existed in Massachusetts between living arrangement and the average cost per episode. Those individuals who lived alone cost substantially more than those who did not. In Philadelphia, the cost per episode for patients who lived alone was greater than the overall cost per episode, but not significantly different from those who l ived with others.9 In the regression analyses, the positive and significant coefficient (at the . 001 level) of the dichotomous variable representing whether the patient lived alone was not surprising. Primary Diagnosis. Table XIX indicates that, in general, the variation in average cost per episode across diagnostic groupings , was significant. However, few of the categories were significant 9This finding took on additional importance when functional status at admission was compared to 1 iving arrangement across the samples. The average AOL index score for patients who lived alone in both samples was twice that (indicating greater disability) of patients who did not live alone.

PAGE 203

TABLE XVIII A COMPARISON OF THE AVERAGE COST PER EPISODE BY LIVING ARRANGEMENT OF MASSACHUSETTS AND PHILADELPHIA PATIENTS Massachusetts Philadelphia Living Arrangement Mean Mean N Cost a S.D. N Cost S.D. Alone a 617 $313.52 527.10 56 $564.89 591.87 With Othersa 1564 235.25 411.59 623 518.25 426.32 Total 2181 $257.39 448.57 679 $522.10 441.95 184 Source: Massachusetts Discharge Summary System, Medicare Cost Reports, Blue Cross Home Care Program Data System, Blue Cross of Greater Philadelphia Home Care Program Director, and Phila delphia Abstracted Patient Records.

PAGE 204

I I I TABLE XIX A COMPARISON OF THE AVERAGE COST PER EPISODE BY THE PRIMARY DIAGNOSIS OF MASSACHUSETTS AND PHILADELPHIA PATIENTS Massachusetts Philadelphia Primary Diagnosis N Mean Cost . S.D. N Mean Cost Infectiog 45 $254.05 454.78 4 $497.87 Neoplasm a 295 222.31 356.78 150 496.08 Endocrine 160 231.71 486.06 47 580.07 Blood 53 454.08 675.61 4 474.28 ' Menta 1 49 251.11 525.00 8 601.35 Nervous a 209 302.58 466.89 12 469.24 Circulatory a 516 309.52 553.00 237 484.62 87 207.89 252.99 36 542.40 Digestive 233 183.11 421.70 39 510.81 Genitourinarya 56 206.93 429.32 1 2424.38 Skin 61 205.93 287.95 5 430.03 233 283.99 380.89 16 685.72 Accidant 166 197. 13 261.65 17 776.93 Other 21 206.48 231.47 102 534.79 Total a 2184 $257.38 448.30 678 $522.10 185 S.D. 337.33 434.74 472.37 133.16 362.85 487.71 446.56 317.99 338.99 -295.96 693.15 438.49 436.73 441.95 Source: Massachusetts Discharge Summary System, Medicare Cost Reports, Blue Cross Home Care Program Data System, and Blue Cross of Greater Philadelphia Home Care Program Director.

PAGE 205

186 when included in the regression analyses. Of the 14 categories, in Massachusetts 10 exhibited an average cost per episode which was lower than the overall grand mean; in Philadelphia, there were seven categories which had lower cost per episode across the data sets for all but a few diagnostic categories. As indicated earlier, regression analyses indicated that musculoskeletal system disorders and ace idents strongly affected average cost per episode. Although there was a pattern between cost and primary diagnosis, the nature of the pattern was difficult to interpret. The extent to which variations occurred was a complicated issue which could not be adequately addressed in this study. Implications for further research are summarized later in this discussion. Functional Status Clear and consistent effects of functional status on cost per episode could be seen in both samples. Table XX illustrates the difference in cost per episode by each of the components of the ADL index. The only component which was not consistently significant in Hs relation to cost was eating. The prevalence of dependency for this category was one of the lowest in both samples which may, in part, explain this finding. The average cost per episode for patients dependent in each of the components of the ADL index was higher than the cost per episode for those who were independent. The strong relationship between functional status, services pro vided, and the cost of those services suggested a relative efficiency in the provision of services which was somewhat encouraging.

PAGE 206

TABLE XX A COMPARISON OF THE AVERAGE COST PER EPISODE BY COMPONENTS OF THE ACTIVITIES OF DAILY LIVING OF MASSACHUSETTS AND PHILADELPHIA PATIENTS ADL a Massachusetts Phil a del phi a N Mean Cost S.D. N Mean Cost Bathing Independent 1198 $188.51 341 .26 290 $ 437.00 Dependent 982 342.23 540.31 389 585.54 Dressing Independent 1395 205.00 347.91 280 423.27 Dependent 784 351 .88 575.07 399 591 .46 Ambu1ation Independent 1562 219.22 357.86 385 455.67 Dependent 614 355.57 613.01 294 609. 1 0 Toileting Independent 1651 227.27 381 . 76 365 443.83 Dependent 529 352.90 602.63 314 613.09 Transferring Independent 1614 221 .66 370.79 391 351 . 98 Dependent 565 361 . 11 608.16 288 617.30 Eating Independent 1809 236.79 391 .08 594 51 5. 35 Dependent 372 359.04 651 . 49 85 569.25 187 S.D. 361 .34 484.37 346.59 486.65 399.44 479.03 366.84 501 .14 387.80 491 .19 429.52 521 . 1 0 Source: Massachusetts Discharge Summary System, Medicare Cost Reports, Blue Cross Home Care Program Data System, Blue Cross of Greater Philadelphia Home Care Program Director, and Phil adelphia Abstracted Patient Records. aAverage cost per episode was significantly different across and within the two samples for all categories of functional status (ps.OOl), except in the category of " Eating" within the Phil adelphia sample.

PAGE 207

188 Functional Status Index The ADL index score at admission was consistently significant and positively related to cost per episode in all of the analyses conducted. It appeared the single most important factor in explain-ing variation in cost. Table XXI illustrates the variation in cost per episode in relationship to the total functional status (ADL) index score at time of admission. Patterns were consistent across both samples, since costs were lower for those individuals either totally dependent or totally independent; and the average cost per episode was significantly different across the levels of independence on the ADL categories. These differences caul d be accounted for, at 1 east conceptually, by the fact that individuals who were relatively independent used fewer services than individuals who were more dependent. Those totally dependent actually used fewer services, ' since there was little potential for rehabilitation. In addition, they were often readmitted to the hospital for more intensive care and these costs were not included in the calculation of CPE. A 1 though the highest cost per episode in Phi 1 ade 1 phi a was for those who were almost independent, this was probably the result of the small nllllber of cases in the category (17). I n the Phil adelphi a sample, if this category was ignored, then the relationship between the ADL index score and cost became clearer. The three most depend-ent ADL index categories ( i . e. , independent in zero, one, and two ADLs) had higher average costs per episode than the remaining cate-gories which indicated greater independence (i.e., independent in

PAGE 208

189 TABLE XXI A COMPARISON OF THE AVERAGE COST PER EPISODE BY THE ACTIVITIES OF DAILY LIVING INDEX OF MASSACHUSETTS AND PHILADELPHIA PATIENTS Massachusetts Philadelph i a ADL Index N Mean Cost S.D. N Mean Cost: S.D. Independent in 6 ADLsa 1086 $179.76 289.70 259 $395.39 295.36 Independent in 5 ADLsa 227 283.73 446.22 17 833.69 724.74 Independent in 4 AOLsa 214 320.18 525.05 89 542.22 478.52 Independent in 3 ADLs 106 346.39 531.98 23 538.14 433.26 Independent in 2 ADLsb 102 316.44 509.49 32 611.15 502.26 Independent in 1 ADLa I 179 402.24 597.51 177 626.83 469.32 Independent in Noneb 257 355.48 652.50 82 570.59 526.02 Total a 2171 $258.17 449.39 679 $522.10 4 41.95 Source: Massachusetts Discharge Summary System, Medicare Cost Reports, Blue Cross Home Care Program Data System, Blue Cross of Greater Philadelphia Home Care Program Director, and Philadelphia Abstracted Patient Records. a < p-. 001

PAGE 209

190 three through six AOLs). In Massachusetts a similar pattern existed. Otherwise, the episodes classified as independent in five and six ADLs exhibited lower average cost per episode than those classified as independent in four and five AOLs, which in turn had 1 ower average costs per episode than the t\>Kl most dependent cate-gories. The only exception was the category 11Independent in two AOLS,11 which indicated lower costs per episode than all but the two most independent categories. Hence, overall the patterns appeared consistent across the two samples. As indicated earlier, the ADL index score at time of admission was a key independent variable in the regression analyses. I t was high.ly significant in all equations in which it was included and indicated a strong positive relationship between dependence and cost per episode. In addition, it was the first variable entered in the stepwise equations for both samples, thus substantiating its importance in explaining observed var iation. Goal at Admission. Table XXII presents the relationships between goal at admission and the cost per episode for the Massachusetts sample only, s i nce data were not available on this variable for the Philadelphia episodes. The most costly episodes were those with the goal of rehabilitation at admission.lO The dichotomous variable representing the goal of rehabilitation was lORehabilitation patients had greater functional dependency at admission than all patients other than the terminally ill, • .vith an average AOL index score of 2.51 (indicating dependenc! on 2 . 5 out of six items in the scale).

PAGE 210

TABLE XXII A COMPARISON OF THE AVERAGE COST PER EPISODE BY THE GOAL AT ADMISSION-oF MASSACHUSETTS PATIENTS Massachusetts Goal N a Mean Cost Recovery 507 $ 188.38 Self-Care 613 259.94 Rehabilitation 276 379.22 Maintenance 605 295.45 Tenni na 1 87 251.72 None 87 19.54 Total 2175 $ 258.33 191 S.D. 229.64 444.77 532.14 561.44 339.27 11.89 448.98 Source: Massachusetts Discharge Summary System and Medicare Cost Reports.

PAGE 211

192 positive and significant in all of the regression equations in which it was included, thus emphasizing its importance. Patients whose care plans predicted recovery, cost the least of all patients in the Massachusetts sample. The average cost per episode of terminally ill patients was slightly lower than the average for all patients. However, when the terminally ill were compared to all other patients in the regression analyses, the F-test for overall significance of the R2 was insignificant. This suggested that home care costs for terminally ill patients were not explained by the same factors which applied to the remainder of the patient population. Outcome-Related Characteristics A critical issue in the delivery of health care is the relationship between cost and the outcome of care. This relationship is, at best, indirect in most cases. This portion of the chapter describes the sample populations• cost experiences in terms of two of the most import ant patient outcome measures. are: Change in Functional Status Index Score Health Status at Discharge Overview of Multivariate Findings These variables Overall, although both the Massachusetts and Philadelphia equations using only outcome variables as independent variables were significant, the multiple regression coefficients in both cases were extremely small. Thus, by themselves, the outcome-related variables did not have a strong impact on CPE. In both samples, the variable

PAGE 212

193 measuring change in ADL index score from admission to discharge was significant 1 y negative 1 y associ a ted with cost per episode, i ndi eating that improvement in functional status was positively associated with CPE. The findings regarding the vari ab 1 e me as uri ng general change in health status were inconsistent. The coefficients of the variable in the two samples were different in magnitude, direction,. and significance. Change in Functional Status Index In Table XXIII the change in functional status (ADL) index was measured in terms of the change in the ADL score from admission to discharge, and categorized as no change, improved, or deteriorated. The index was composed of six items (toileting, eating, dres .sing, bathing, transferring, and anbulation) and scored according to the patients' independence or dependence. The measure was more sophi sticated and quantitative than the one which follows this discussion, yet narrower in scope. It was more fccused, since it measured only the individual's ADL status, and excluded measures of psychosocial function or medica l intensity. Consequently, it was generally a more consistent measure of change and more strongly correlated with the utilization and cost of home care. In both the Massachusetts and Philadelphia samples, the average cost per episode for patients with no change in functional status was lower than for patients in either of the other two groups. In Massachusetts, those cases with no change in functional status were

PAGE 213

TABLE XXIII A COMPARISON OF THE AVERAGE COST PER EPISODE BY T H E CHANGE IN ACTIVITIES OF DAILY LIVING INDEX OF MASSACHUSETTS AND PHILADELPHIA PATIENTS Change in Massachusetts Philadelphia ADL Index N Mean Cost S.D. N Mean Cost S.D. No Change a 1553 $ 204.01 380.93 401 $ 447.88 393.41 Improved a 383 387.04 484.58 230 649.58 475.95 Deteriorated 170 469.42 724.52 47 $ 535.33 524.52 Total 2106 $ 258.72 447.81 678 $ 522.37 4 42.22 194 Source: Massachusetts Discharge Summary System, Medicare Cost Reports, Blue Cross Home Care Program Data System, Blue Cross of Greater Philadelphia Home Care Program Director, and Phila delphia Abstracted Patient Records.

PAGE 214

195 less costly than those that improved, which in turn were less costly than those that deteriorated. In contrast, the average cost per episode in Philadelphia was highest for patients who improved. In the regression equations, the change in functional status variable was generally significant and negative (indicating that decreased dependency was associated with higher cost per episode}. However, the general patterns of the regression and the descriptive findings suggested that both improvement and deterioration were more costly than no change. These results were reasonable, in that the care provided to assist a patient toward independence was likely to be more costly than maintenance care, as was the care required by a patient whose condition was deteriorating. Health Status at Discharge. As indicated, the general health status at discharge was an estimate of both medical and psychosocial status of the patient. Its inconsistent relationship to cost per episode is depicted in Table XXIV. In both samples the highest cost per episode was for those patients who expired during the course of treatment. It should be noted that this finding was consistent with the finding relating goal at admission to cost per episode for terminally ill patients, since many terminally ill patients did not die at home, but were readmitted to an inpatient facility when death was near. In Philadelphia, the average cost per episode for those who stayed the same was about average; in Massachusetts it was below. The average cost for patients who worsened in Philadelphia was below the average, but in Massachusetts it was above.

PAGE 215

/ 196 TABLE XXIV A COMPARISON OF THE AVERAGE COST PER EPISODE BY THE HEALTH STATUS AT DISCHARGE OF MASSACHUSETTS AND PHILADELPHIA PATIENTS Health Status Massachusetts Philadelphia at Discharge N Mean Cost S.D. N Mean Cost S.D. Improved a 977 $249.48 341.94 460 $537.29 425.09 Same a 641 148.74 305.00 62 567.11 516. 19 Worsened 437 394.44 637.42 121 424.01 417.14 Expired 85 493.47 827.33 36 580.17 555.55 Total a 2140 $258.59 447.77 679 $ 522.10 441.95 Source: Massachusetts Discharge Summary System, Medicare Cost Reports, Blue Cross Home Care Program Data System, Blue Cross of Greater Philadelphia Home Care Program Director, and Phila delphia Abstracted Patient Records.

PAGE 216

197 The significance of the health status variable in regression equations was inconsistent across the samp 1 es. When outcome vari-ables were included alone in the regression equations, this variable was generally significant, but when it was included with the other variable categories, its significance diminished. In light of these somewhat anbiguous findings, it appeared appropriate to place most reliance on the results for the other major outcome measure, namely, change in functional status. Utilization-Related Outcome. Both of the additional outcome-related variables, length of use of home health care and the average mmber of visits per day, were highly positively correlated with cost per episode in both samples. As indicated in the discussion of the conceptual framework, this relationship was to be expected since both variables were, in part, measures of utilization (and therefore highly related to total cost). Because their measurement was similar to that of the dependent variables, they were omitted from the regression analyses but used as criterion variables in selected secondary ana lyses. Provider/Health System Characteristics Recently published Medicare data cited earlier indicate that the cost per unit of service and the total cost per beneficiary vary by region and provider type.ll In addition, differences in benefit llwayne Callahan, Medicare: Utilization of Home Health Services, 1978 (Baltimore: Health Care Financing Administration, DHEW, l980), pp. 2-6.

PAGE 217

198 packages and reimbursement rates of the major home health payers may impact service utilization and cost. Hence, the primary source of payment for each episode was included as a key independent variable in this study. In an effort to determine the impact of such factors on cost per episode, several other provider/health system variables were used in this study. The key variables discussed in this section include: Primary Source of Payment Provider Size Acute Care Admissions Overview of Multivariate Findings. In the regression analyses, the effect of thi. s category of variables was different in the two samples. In Massachusetts, when other factors were taken into consideration, three or four provider/ health system characteristics were significant. Both Medicare and Medicaid, as primary payment sources, exhibited positive and significant regression coefficients, indicating that cost per episode for Medicare and Medicaid patients was higher than for other patients. In Massachusetts, where all hospitals served as major sources of referrals to home care, the variables measuring hospital admis sions per thousand population had a positive and significant effect on cost per episode. When all cases were included in the Massachusetts equations, the variable measuring agency size was not signi fie ant. However, when equations were estimated using a sub sample of those cases within five standard deviations of the mean cost per episode, size of provider was negatively associated with

PAGE 218

199 cost per episode and significant. These findings were consistent with descriptive analyses described in this section. In Phil adelphi a, different regression results emerged. When all variables were included in the equations, none of the provider/ health systen characteristics were significant in the Philadelphia regressions. Consequently, for the total Philadelphia sample, variation in cost per episode was accounted for entirely by patient-related variables (i.e., patient-specific and outcome-related characteristics). When all variables were included in the equa-tions, for the subsample of Philadelphia rehabilitation episodes, the only provider/health systen characteristic of significance was the variable which measured acute care admissions per thousand population. Primary Source of Payment. The descriptive analyses depicted in Table XXV illustrate the variation in cost per episode by primary source of payment for each episode of care. In both samples, episodes financed through public programs (e.g., Medicare and Medicaid) had the highest cost per episode. Given the low per diem rate paid by Medicaid in Pennsylvania, it was somewhat surprising to find consistency across both samp 1 es. Suppl ementa 1 an a lyses for both samples indicated that patients whose episodes were reimbursed out of pocket or by commercial insurers were significantly younger and

PAGE 219

TABLE XXV A COMPARISON OF THE AVERAGE COST PER EPISODE BY THE PRIMARY SOURCE OF PAYMENT OF MASSACHUSETTS AND PHILADELPHIA PATIENTS Massachusetts Philadelphia Source of Payment N Mean Cost S.D. N Mean Cost S.D. Medica rea 1385 $288.46 479.70 417 $547.01 466.32 Medicaida 247 274.98 471.22 34 595.75 458.92 Insurancea 253 172.49 211.98 223 470.86 388.33 Private Pay 201 179.29 455.07 4 167.59 50.15 Other 98 153. 21 275.39 1 475.00 Tota1a 2184 $ 257.38 448.30 679 $522.10 441.95 200 Source: Massachusetts Discharge Summary System, M edicare Cost Reports, Blue Cross Home Care Program Data System, and Blue Cross of Greater Philadelphia Home Care Program Director.

PAGE 220

201 less functionally disabled than all others.l2 Consequently, the differences in cost between the payer categories were no longer surprising. Provider Size. When comparing provider size to the dependent variable, cost per episode, there was weak evidence that economies of scale exist in the provision of home health services. Table XXVI illustrates this point. In both samples, as the size of the provider increased (measured by the total number of nursing visits per year), the cost per episode decreased. The larger dispersion about the mean cost per episode for the Massachusetts sample tended to weaken this relationship.l3 The sign of the correlation coeffi-cient of cost per episode and provider size was negative and signi-ficant (p .001) in both samples. Yet, the regression analyses indicated that the relationship between size and cost per episode was not always statistically significant. This was partially due to the high correlation coefficients between size and the other health system variables, which made the interpretation of its effect on the cost per episode difficult. 12In the Massachusetts sample, the average age of private pay patients was 66 years, while patients whose episodes of care were reimbursed by commercial insurers averaged 48 years of age. Their corresponding average ADL index scores were 1.05 and 1.68 respectively (indicating less debilitation than the average). In the Philadelphia sample, the average age of private pay patients was 42 years, while patients whose episodes of care were reimbursed by commercial insurers averaged 51 years of age; their corresponding aver age ADL index scores were 2.00 and 2.90 respectively (also significantly 1 ower than the aver age). 13As indicated earlier, when only cases within five standard deviations of the mean were used in the sample, size became significant in the regression equation.

PAGE 221

TABLE XXVI A COMPARISON OF THE AVERAGE COST PER EPISODE BY THE PROVIDER SIZE FOR MASSACHUSETTS AND PHILADELPHIA PATIENTS M assachusetts Philadelphia Provider Size Mean Mean N Cost S.D. N Cost 0-1500 -$ -109 $ 610.55 1501-5000b 209 237.82 275.21 570 505.19 5001-10000 1147 277.57 456.26 -10000 + 828 234.35 470.94 --Total 2184 $ 257.38 448.30 679 $ 522.10 202 S.D. 583.04 407.96 --441.95 Source: Massachusetts Discharge Summary System, Medicare Cost Reports, and Blue Cross Home Care Program D ata System. a P rovider Size was m easured by total number of nursing v isits in 1976.

PAGE 222

203 Acute Care Admissions. Table XXVII illustrates the relationship between acute care admissions per thousand population and cost per episode. Although difficult to interpret, in part because of the limited number of data points, there was a significant positive relationship between the two variables in both samples. This relationship was also evident (for the Massachusetts sample only) in the regression findings. The direction of this relationship and its significance suggested that there was a substitution effect between home care and inpatient care which prompted home care providers to increase service provision and hence costs, all other things being equal. However, because this variable took on only eight possible values, the interpretation of these findings remains tentative. Comparison of Massachusetts and Philadelphia Outcomes The average cost per episode was significantly greater in the Philadelphia sample than in Massachusetts. As indicated earlier, there were severa1 reasons for this, including differences in case mix, cost allocation methods, and the general management of the pro grams. One important potential explanation which warrants additional discussion i s the difference in outcomes of care i n the samples. In other words, did patients treated by Philadelphia providers improve more often than patients treated by Massachusetts providers, when controlling for key patient characteristics? This sections addresses this question.

PAGE 223

TABLE XXVII A COMPARISON OF THE AVERAGE COST PER EPISODE BY THE ACUTE CARE ADMISSIONS IN THE SERVICE AREA FOR MASSACHUSETTS AND PHILADELPHIA PATIENTS Acute Care Massachusetts Philadelphia Admissionsa Mean Mean N Cost S.D. N Cost S.D. 0-10.0 1520 $238.83 427.40 --10.1-15.0 b 664 299.89 490.52 412 $494.33 417.95 15.1 + ---267 564.97 474.26 Total 2184 $257.39 448.30 679 $522.11 441.96 Source: AHA Guide. 204 aAcute care admissions was measured as acute care admissions per thousand population in county of location of main office.

PAGE 224

205 Overview In an effort to systematically address the issue, the mean change in the functional status (ADL) index scores and the proport ion of episodes with improvement in these scores were compared across the t\\Q samples, while controlling for case mix variables.l4 These variables included age, living arrangement, diagnosis, and the admitting AOL index. The specific findings are presented below for each variable. In general, they indicated that there was a consis-tent pattern across the samples; Phi 1 adelphi a patients tended to improve more often than Massachusetts patients. Whether or not this gain in functional status offsets the increased cost of care is a value judgment. Clearly, further investigation of this matter is warranted. Controlling for Age When controlling for age, the proportion of the Philadelphia sample with improved outcomes was almost twice that of the Massachu-setts sanple. Table XXVIII indicates the relationship between the age of a patient and aver age change in ADL index score from admi s-sian to discharge. Of the episodes in Philadelphia, 33.9% reported improved function a 1 status, whi 1 e 18.3% of the Massachusetts epi-sodes reported similarly. The difference between the proportions in the t\\Q samples was most pronounced in both the youngest and oldest 14rhe comparison was also made using the outcome variable, patient health status at discharge, with similar results.

PAGE 225

206 TABLE XXVIII A COMPARISON OF THE CHANGE IN ACTIVITIES OF DAILY LIVING INDEX BY THE AGE OF MASSACHUSETTS AND PHILADELPHIA PATIENTS Massachusetts Philadelphia Age Categories Mean Change a % b Mean % N Improved N Change Improved 0-19 65 -.21 10.8 14 -1.14 42.8 20-59 363 -.04 12. 1 198 -.79 26.8 60-69 427 -.20 19.6 167 -.87 34.1 70-79 736 -.22 18.9 197 .. :95 38.1 80 + 518 -.23 9.8 102 -1.16 38.2 Total 2109 -.19 18.3 678 -.92 33.9 Source: M assachusetts Discharge Summary System, B lue Cross Home Care Program Data System, and Philadelphia Abstracted Patient Records. aMean changes in ADL index scores for all age categories were significantly different across the two samples (p .01) . bThe percentage of patients with improved ADL index scores for all age categories were significantly different across the two samp 1 es ( p s. 021 ) .

PAGE 226

207 age categories. The average change in functional status when controlling for age, was also consistently different in the two sam-ples. Improvement in the Massachusetts sample was not significantly different across Philadelphia age categories. Similar to the proportional findings, the greatest average improvement in the functional status index score was exhibited by patients in the youngest and oldest age categories.l5 Differences in proportions and means across the two samples in each age category were significant.16 Controlling for Living Arrangement Since the living arrangement of the patient was an important factor in the determination of utilization and cost, it was also contro1led in these comparisons (shown in Table XXIX). There was not a significant difference in the outcome of care in the two samples of patients living alone, yet there was a significant dif-ference in the outcomes for patients who did not live alone. This finding gained additional significance when the admitting functional status index of patients living alone was compared across the two samples. Findings presented in Table XXX indicate that a significantly larger proportion of patients in the Philadelphia sample (45.7%) than in the Massachusetts sample (31.1%) were independent in two or fewer of the items in the AOL index score, thus 15The negative signs indicated reduction in dependency. 16rhe percentage of patients deteriorating in both samples was not significantly different, but more episodes in Massachusetts recorded no change in health status at discharge than in Phi 1 adelphi a.

PAGE 227

TABLE XXIX A COMPARISON OF THE CHANGE IN ACTIVITIES OF DAILY LIVING INDEX BY THE LIVING ARRANGEMENT OF AND PHILADELPHIA PATIENTS Massachusetts Philadelphia Living Mean % Mean % 208 Arrangement N Change a Improvedb N Change Improved Alone 595 -.19 20.0 56 -.92 With Others 1515 -.20 17.6 622 -.97 34.9 Total 2110 -.19 18.3 678 -.92 33.9 . Source: Massachusetts Discharge Summary System, Blue Cross Home Care Program Data System, and Philadelphia Abstracted Patient Records. aMean changes in ADL index scores for those living 11Alone11 and 11With Others11 were significantly different across the two samples bThe percentage of patients with improved ADL index scores for those patients living 11With Othersu was significantly different across the two samples

PAGE 228

TABLE XXX A COMPARISON OF THE ACTIVITIES OF DAILY L IVING INDEX BY THE LIVING ARRANGEMENT OF AND PHILADELPHIA PATIENTS 209 Massachusetts Philadelphia ADL Index % % With % % With N Alone Others N Alone Others Independent in 6 ADLs 1084 61.2 45.4 . 258. 60.7 36.0 Independent in 5 ADLs 228 14.6 8. 9 . 17 1.8 2.6 Independent in 4 ADLs 215 11 . 5 9.2 8 9 23.2 12.2 Independent in 3 ADLs 108 3.9 5.4 23 1. 8 3.5 Independent in 2 ADLs 103 3.4 5.3 32 1.8 5.0 Independent in 1 ADL 179 2.4 10.5 177 8.9 27.7 Independent in None 257 2 . 9 15.3 82 1.8 13.0 Total 2174 28.3 71.7 678 8.3 91.7 Mean ADL 1. 76 .93 2.09 2. 61 1.16 2.73 ' S.D. 2.22 1. 52 2.36 2.39 l. 72 2.40 Source: Massachusetts Discharge Summary System, Blue Cross Home Care Program Data System, and Philadelphia Abstracted Patient Records.

PAGE 229

210 indicating significantly greater disability on the part of the Philadelphia patients. Consequently, the difference in the improvement of patients who did not 1 ive alone was of greater importance than would otherwise be the case. The increased proportion of patients who lived with others with greater functional dependency (e.g., independent in one or zero AOLs) indicated that patients who did not live alone could remain at home while exhibiting greater dependency than other patients. This finding, considered in view of regression findings for living arrangement presented earlier, suggests that the living arrangement of the patient is more important than previously recognized in its impact on home health case mix. Controlling for Primary Diagnosis Table XXX.I presents the comparison of change in functional status across the t\'tO samples by the primary diagnosis of the patient at admission for a selected m.mber of diagnostic categories. As illustrated, a substantially larger proportion of patients in the Philadelphia sample than in Massachusetts was found, in almost all categories with improved ADL index scores. In addition, the aver age improvement in those diagnostic categories with higher than average costs per episode was significantly larger in Philadelphia. Controlling for Functional Status Index The outcome comparison also controlled for functional status at admission because the maximum possible improvement for any patient depended, in part, on their initial score. For example, those

PAGE 230

211 TABLE XXXI A COMPARISON OF THE CHANGE IN ACTIVITIES OF DAILY LIVING INDEX BY SELECTED-PRIMARY DIAGNOSES OF MASSACHUSETTS AND PHILADELPHIA PATIENTS Massachusetts Philadelphia Primary Diagnosis Mean % Mean % N Changea Improvedb N Change Improved Neoplasms 290 .30 13.1 150 -.49 24.7 Endocrine 155 .... 07 9.7 47 -. 91 34.0 Nervous 206 -.17 10.2 12 -1.17 25.0 Circulatory 497 -.26 21.9 236 -.75 28.4 Digestive 223 -.12 10.8 39 -1.26 48.7 Muscu1 oske 1 eta 1 225 -.52 30.2 16 -1.38 50.0 Accidents 164 -.93 40.9 17 -1.94 58.8 Total 2112 . . 18 18.3 679 -.92 33.9 Source: Massachusetts Discharge Summary System, Blue Cross Home Care Program Data System, and Philadelphia Abstracted Patient Records. a Mean c .hanges in ADL index scores for all diagnoses were sig-nificantly different across the two samples for all but the Musculoskeletal system category. bThe percentage of patients with improved ADL index scores for all primary diagnoses were significantly different across the two samples for all but the Circulatory system category.

PAGE 231

212 patients who were totally independent at admission could not (by definition) improve their scores. Table XXXII illustrates the com-parison between the two samples when controlling for functional status at admission. It shows that a significantly higher proper-tion of Philadelphia episodes than Massachusetts episodes had improved functional status index scores (except for those most independent at admission). The average change in functional status index score, when controlling for ADL scores at admission, was sig-nificantly higher in Philadelphia than in Massachusetts. Again, this indicates greater overall improvement by Philadelphia patients. In conclusion, Phil adelphi a patients generally improved more often and to a greater degree than Massachusetts patients. This improvement was spread across the patient characteristics examined.!? Whether or not the increased cost of such care was warrant-ed by the differences in outcome between the Massachusetts and Philadelphia samples is a question still to be answered. SlJllmary of Major Findings This section summarizes the major findings of the study. A general description of the average number of visits and cost per 17This finding was, to some extent, not unexpected since the Philadelphia case mix was generally more acutely ill than the Massachusetts case mix; consequently, there was a greater potential for improvement in Philadelphia than in the more chronically ill long term care case mix of Massachusetts. Other explanations which may account for the variation in outcomes across the two samples include service mix differences, availability of ancillary services, and intensive professional coordination of services.

PAGE 232

213 TABLE XXXII A COMPARISON OF THE CHANGE IN ACTIVITIES OF DAILY LIVING INDEX BY THE ADL INDEX SCORES AT TIME OF ADMISSION OF MASSACHUSETTS AND PHILADELPHIA PATIENTS Massachusetts Philadelphia ADL Index Mean % Mean % N Change Improved a N Change Improved Independent in 6 ADLs 1057 .24 o _ 259 . 21 Independent in 5 ADLs 221 -.31 52.0 17 .94 Independent in 4 ADLs 207 -.43 44.9 89 -.72 Independent in 3 ADLs 105 -.42 40.0 23 -1.35 Independent in 2 ADLs 102 -.67 32.4 31 -2.65 Independent in 1 ADL 177 -1.16 35.6 177 -2.25 Independent in None 243 -.74 16.5 82 -1.07 Total 2112 -.18 18.3 678 -.92 Source: Massachusetts Discharge Summary System and Philadelphia Abstracted Patient Records. 0 94. l 62.9 60.9 74.2 54.8 29.3 33.9 aThe percentage of patients with improved ADL index scores for all admitting ADL index scores were significantly different across the two samples for all but the 11Independent in 3 ADLs11 category.

PAGE 233

214 episode for the t\'tO samples is presented. Also, a general discussion is presented which describes the analyses used to test the overall relationship between the cost per episode (and total number of visits) and all three categories of independent variables. Finally, within each set of independent variables, the most impor-tant findings with respect to individual variables are highlighted. Policy implications of the findings are discussed in Chapter VI of this volume, and research implications in Chapter VII. Overall Utilization and Cost Per Episode Findings The average number of total visits per episode was similar in the two samples, 16.35 visits per episode for the Philadelph i a sample and 18.25 for the Massachusetts. However, the average cost per episode was considerably higher in Philadelphia ($522) than in Massachusetts {$257). Several explanations for this cost difference were explored in the study. Overall Ability to Explain Observed Variation Within Each Sample Regression results suggested that for the study patients many factors other than those inc 1 uded affected the cost of home he a 1 th care. The uniformly low multiple correlation coefficients which resulted from the overall regression analyses (e.g., R2 = .13) sug gested that the majority of the variation in cost per episode was unexplained. Secondary analyses performed on a subsample of Massachusetts patients who were categorized as terminally i 11 at admission to home care indicated that even less variation in cost

PAGE 234

215 per episode was explained by the independent variables for this pop-ulation than for all others: Similar analyses for rehabilitation patients resulted in larger multiple correlation coefficients, although still relatively low, than for the remainder of the population. This suggested a somewhat greater ability to explain the variation in cost for rehabilitation patients than for other categories of patients.l8 Relative Importance of Each Independent Variable Category The patient-specific variable category was the most significant of the three in terms of its ability to explain variation in cost per episode. The proportions of variation explained by the other two categories of independent variables, outcome-related and provider/health system characteristics, exhibited varying explanatory l8The rehabilitation patient is more likely to be served by a few health care providers, even possibly a single one. Typically, the therapies used by such patients (physical therapy, occupational therapy, and speech therapy) are provided by a single 11Specialist11 agency. The result is less duplication and overlapping of service. In addition, the service protocol for many rehabilitation patients is more definitive than for the majority of patients (e.g., stroke patients generally receive occupational, physical, and speech therapy in some combination of services). Since protocols for most home health care patients are less well established (or more discretionary) than for rehabilitation patients, it is not surprising that the utilization and cost of care for the majority of episodes of illness are not well explained by the independent variables. This explanation relates to the broader issue of increased inconsistency in patterns of medical practice dealing with less acutely ill patients. Health services research literature (some of which was summarized in Chapter II) suggests that the less acutely ill the patient, the less consistency is found among physicians in prescriptions for care. Thus, it is not surprising that in the case of home health care, where patients are usually not acutely ill, there is little consistency in utilization practices.

PAGE 235

216 power. Individual variables from within each of the three categories exhibited different relationships and significance levels with the dependent variable. These findings are summarized below. Patient-Specific Characteristics. 1) There was a consistent pattern in the two samples for the average cost per episode to increase with age up to, but not including, the most elderly patient groups. The age of a patient was a key factor in explaining costs in the regression analyses when only patient-specific variables were included. 2) Living Arrangement. The average cost per episode was significantly higher in both samples for patients living alone than for those living with others. Whether or not a patient lived alone was also a key variable in the regression analyses in explaining variation in cost. 3) Primary Diagnosis. There was little consistency in cost patterns across the primary diagnostic categories. Only a few of the diagnostic categories were significant when included in the regression analyses (e.g., blood disorders and accidents). Although there were general patterns the various diagnostic groupings, it was difficult to discern any overall impact on cost per episode in terms of diagnosis. 4) Functional Status Index. The functional status index of the patient at admission was an important variable in explaining the cost of home care. Costs were lower for those individuals either totally dependent or totally independent. Patterns were consistent for both samples and in the regression analyses.

PAGE 236

217 5) Goal at Admission. The goal specified for patients by the home health provider at the time of admission was found to be related to cost in the Massachusetts sample. Patients for whom the goal was rehabilitation had the highest average cost per episode. Patients for whom the goal was recovery had the least cost. Outcome-Related Characteri sties. The two main outcome vari-ables used in this study were: 1) the change in the functional status (ADL) index score from admission to discharge, and 2) a global e stimate of the change in health status from admission to discharge. Because outcome variables are subject to methodological weaknesses, the general pattern of findings was emphasized than the specific findings in terms of each outcome variable. Three general findings emerged. The first was that, although there were similarities in the two samples with regard to patient outcome, there were significant differences. The majority of patients in both samples either improved or maintained their functional status from time of admission to discharge (as measured by changes in AOL index scores). However, a significantly higher proportion of patients improved in Philadelphia than in Massachusetts. Even after controlling for key patient characteristics such as age, diagnosis, and admitting functional status, significantly more Phil adelphi a patients improved than did

PAGE 237

218 Massachusetts patients.l9 Thus, part of the difference in cost per episode between the two s amp 1 es was apparent 1 y re 1 a ted to differ-ences in outcomes. The second outcome-related finding was that, in both samples, patients who died at home while receiving care had higher costs than other patients. As indicated earlier, regression findings indicated that the independent variables explained observed cost variations for termi na 11 y i 11 patients (some of whom died at home) 1 ess than for all other patients. The implications of this finding for the expansion of hospice-type care are discussed later in this work. The third finding was that for the remaining groups of patients in both samples, when patient health status improved from admission to discharge, the cost per episode was greater than otherwise. The number of visits (utilization) did not show a similar pattern, suggesting that the mix of services varied, and that more costly ser-vices were provided to those who improved. Provider/Health System Characteristics. 1) Primary Source of Pa.]1Tlent. Episodes reimbursed by Medicare and Medicaid resulted in higher costs per episode than those reim bursed by other payers. This difference may have reflected differences in patient characteristics (which were included in the regres-19This finding was, to some extent, not unexpected since the Philadelphia case mix was generally more acutely ill than the Massachusetts case mix; consequently there was greater potential for improvement than in the more chronically i11 long-term care case m i x of Massachusetts.

PAGE 238

219 sian analyses), as well as other progrcl'll factors (which could not be tested), such as the types of services which were reimbursed, the ease of obtaining reimbursement, and the degree to which reimbursement covered cost. 2) Provider Size. Because only four providers from each sample were included in this study, the effects of variation in provider-specific characteristics could only be examined for four different situations in each sample. Of the three variables (size, density of population, and acute care admissions), the most significant was provider size (measured as the total nunber of nursing visits provided in 1976). In general, as the size of the provider increased, the cost per episode decreased. The regression findings suggested that this relationship, although tenuous, was not due to differences in case mix or outcomes. Therefore, service efficiencies may exist for larger agencies in the delivery of home health care services. Conclusion There are several clear and consistent patterns across the two samples with respect to the influence of certain patient characteristics on cost per episode. Patterns are less consistent for outcome and provider/health system variables. The implications of these findings are discussed in the following chapter.

PAGE 239

CHAPTER VI POLICY IMPLICATIONS The findings SLITlmarized in the previous chapter have several implications for the development of long-tenn care public policy. The discussion which follows outlines the program and policy implications of the study with respect to t\\0 main areas. The first implications deal with the general area of the cost (and reimburse ment) of home health care. The second group of implications deals with specific program issues at the federal level which warrant further consideration in view of the findings of this study. The final chapter of this vohme, Chapter VII, discusses the implications of this study for further research. Cost Implications Inability to Explain Cost Variations The consistently low multiple correlation coefficients suggest that 11typical11 patient descriptors (which often serve as the basis for national projections) may be inadequate for the development of highly reliable utilization and expenditure estimates because patient-level cost (and utilization) does not vary sig.nificantly with patient or provider characteristics. Thus, it may be difficult to predict, with much confidence, national estimates of home health care expenditures without further knowledge of the relationship of

PAGE 240

221 home health care utilization and cost to factors other than those included in this study. Utilization Review Another implication of the overall regression findings is that the impact of discretionary judgment on the part of the provider is significant in its effect on the utilization and cost of home care. This conclusion is suggested by the regression finding that more of the variation in cost is explained for rehabilitation patients than for other patients. This may be due to greater uniformity in treatment patterns or protocols for rehabilitation patients compared to those for the average home care patient. If the explanation of cost patterns is a necessary condition for contra 1 of total expenditures, then the discretionary judgment of the care giver may need to be restricted through increased util. " ization re. view efforts and the development of more uniform standards of care. Utilization review and quality assurance programs are well developed in the hospital sector where they restrict variation in patterns of practice by reducing the discretion of the care giver. Perhaps federal policy makers should require such practices in home care programs. Allocation Procedures The finding with regard to the overall difference in cost per episode across the two samples can be explained by several factors. Some of the difference is probably a function of the (Medicare) cost allocation procedures imposed on hospital-:-based

PAGE 241

programs. 222 One way to reduce the amount of the difference is to alter this allocation procedure so that the cost of utilization review and discharge planning for inpatients is not totally allocated to the home care department of the hospital. Consideration should be given to review of allocation procedures for home care programs in order to make them more consistent, not only within, but across provider types (i.e., free-standing and hospita 1-based pro grams). Cost-Effectiveness The finding of the study which relates cost per episode to changes in functional status indicates that, generally, greater improvement in outcomes was obtained when higher costs were incurred, especially for the more acute Ty ill patient. The break-even point at which the improvement in outcome is no longer "worth .. the additional cost is a public policy question which must be addressed openly. Thus, to some extent, efforts to contain home health care costs may conflict with simultaneous efforts to provide high-quality care. Case Mix Adjustment Since the study suggests there may be differences in the case mix intensity of hospital-based and free-standin' g home health care programs, cost reimbursement 1 imitations developed by third party payers which do not account for this variation in intensity may unduly restrict the patient mix for whom home care is an economically viable alternative. Thus, Medicare cost limitations should include

PAGE 242

223 an adjustment factor for case mix (similar to those in use in Medicaid programs in Illinois and West Virginia) .1 Program Implications Continuum of Care !he average cost per episode varied significantly across categories of patients, thus indicating that home health care was not uniformly low in cost. For example, rehabilitation care for stroke patients at home is generally more costly than other types of home care. Although the cost of home care was not directly compared to that of institutional care in this study, this finding suggests there may be types of patients for whom alternative care settings, such as adult day care, are more cost-effective. Thus, serious con-sideration should be given to the development of a full 11Continuum of care11 which matches the efficient provision of services to the needs of the patient, rather than fitting the needs of the patient to the available benefit package or program. Hospice Care One area of home health care currently under consideration for expansion is care for the terminally ill by hospice programs. This 1 Medicaid reimbursement for nursing home care in these states is based, in part, on the complexity of the needs of each patient. In these states, empirical analysis was completed at the facility level using cost per patient day as the dependent variable. Many of the typical long-term care case mix descriptors were used as the independent variables. In the estimated cost functions, the resulting regression coefficients are then used as weights in determining the appropriate level of reimbursement for each facility.

PAGE 243

224 study found that less of the variation in the utilization and cost of care for terminally ill patients treated at home was explained than for the remainder of the home health care population. Yet, the average cost of care for patients who died at home was the highest of any patient group. This suggests that recent moves to expand the availability of home health care services to terminally ill patients should be approached cautiously in view of current cost containment efforts. This is not meant to imply that home hospice care is more cost-ly than institutional care for the terminally ill, nor that total health care expenditures will necessarily increase if the availabil-ity of hospice care is expanded. Substitution for higher cost alternatives may take place. To the contrary, the findings indicate that home care patterns of practice for the terminally ill are difficult to explain; consequently, costs may be difficult to predict and contra 1 . Home Health Data I The overall regression findings substantiate the weaknesses of patient-specific data collected by ongoing home health programs. Without knowledge of other providers serving a particular patient, it may be impossible to explain much of the variation in home care program costs at the patient level. Consequently, it is important that home health care uniform data collection be implemented, not only for program-specific utilization and cost data, like that required by the Uniform Systen for Home Health Agency Reporting recently proposed by DHEW, but also for case mix descriptors,

PAGE 244

225 outcome measures, and the frequency of utilization of alternative services. With specific reference to case mix indicators, the inability of primary diagnosis to explain variation in cost and utilization reinforces the contention that additional case mix measures must be used in 1 eng-term care. Consequent 1 y, a prob 1 em-oriented record system which incorporates multidimensional case mix measures should be encouraged for primary data collection by home health agencies. In addition, three diagnoses, at a minimum, should be routinely noted by providers in order to aid in the determination of interactive effects of multiple diagnoses. Housing Because the patients• living arrangement was a major determi nant of the utilization and cost of home health care, housing warrants special attention. In this study, those individuals who lived alone used significantly more services, at a higher cost, than did individuals who lived with others. Since the number of elderly living alone can be expected to increase significantly over the next 20 years due to changing demographic patterns and the decline of the nuclear family, this study suggests that an important issue in the health care delivery system is the availablity of supportive housing programs for the elderly and disabled. Thus, expansion of the sup ply of sheltered or congregate housing which provides basic health and social services should be considered part of the health care system and included in the long-range planning for home health

PAGE 245

226 care. Area, regional, and state health planning agencies should be encouraged to include the supply of such housing in their plans.2 Tax Incentives Another implication of the effect of living arrangement on cost per episode relates to tax incentives intended to encourage families and friends to care for the elderly at home. Because patients living with others were generally less costly than those living alone, in .spite of their being more functionally disabled, federal tax policy should encourage the disabled and elderly to remain in the homes of others. Tax incentives in this regard may be, in the long run, both more "humane" and less costly. Information and Referral One final implication pertains to the ability of the data to account for only a small portion of the variation in utilization and cost. This is probably due, in part, to duplication and overlap within the health care delivery systen. Lack of coordinated services generally results in the home care provider having little access to information regarding the nature of other providers treat-ing a particular patient at any time. Therefore, when various combinations of health and social service agencies provide services to the same patients, variation in provider-specific utilization and 2Ann Jones et al., "Projecting the Demand for Public Housing for the Elderly: A Model for Health Planning," (paper presented at the meeting of the Gerontological Society, Dallas, November, .1979). (Mimeographed.)

PAGE 246

227 costs is difficult to explain without information on the frequency and cost of services provided to the patient by each provider. In an attempt to address the problem of a 1 ack of coordination of services, several alternatives should be considered. One suggestion is to develop health and social service information systems which can provide necessary information about the identity and frequency of each care giver serving any particular client. Another suggestion has been the of single entry 11umbre11a11 programs.(currently known as channeling agencies) where patients• utilization of services are continually assessed and monitored. Neither of these suggestions represents a new concept; yet each deserves renewed attention. as the home health care delivery system expands service availability and subsequent costs. The reader is encouraged to add other program and po 1 icy imp 1 ications to this list. Those more experienced in the clinical aspects of long-term care may find additional interesting options within the findings of this study.

PAGE 247

CHAPTER VII AGENDA FOR FUTURE RESEARCH Introduction The United States lacks adequate information for making effective long-term care policies. The purpose of this final chapter is to suggest an improved, systematic research program for developing the information necessary to improve this policy-making effort. It begins with a review of the key problems effecting current research in the ffeld (as reviewed in Chapters II through V of this volume). The next section describes the application of systems analysis to 1 ong-term care research in an effort to improve the unsatisfactory current incremental approach. The specific application of systems analysis to a national research agenda for long-term care policy is then presented. Characteristics and attributes of such a program are outlined. Finally, the chapter closes with recommendations for important actions necessary at the federal level to promote the development of a systematic approach to long-term care policy research. Overview of Long-Term Research Problems I 1 Many of the past and on-going long-term care research studies I l reviewed in Chapter II have limitations which restrict their utility I

PAGE 248

229 for policy-making. The problems and limitations of this research hampers its ability to reduce the uncertainty surrounding long-term care issues. These weaknesses stan from inadequate approaches to systematic and comprehensive research, incomplete research efforts, and inadequate utilization of relevant research, among other things. Each of these is discussed in this section. Lack of a Comprehensive and Systematic Approach There has been no comprehensive or systematic approach to the resolution of long-term care policy questions through research. Few of the studies reviewed in Chapter II were built upo. n conceptual and methodological approaches that were comprehensive and systematic. Although meeting the rigorous requirements of the particular disci-pline of the principal investigator, most dealt with only a small part of long-term care policy problems, often paying little atten-tion to the interfaces within the system and their impact on the outcome of care .1 Few of the projects were the result of a coor dinated and systematic research effort. One common pitfall of applied research in a public pol icy area is the inclination of investigators to avoid difficult and time consuming analytic problems (i.e., they do what is most easily !Geraldine Widmer, Roberta Brill, and Adele Schlosser, "Home Health Care Services and Cost," Nursing Outlook (August, 1978), pp. 488-491; Neill Piland, and Cost-Effectiveness of Alternative Lon -Term Care Setbn s: xecut1Ve Summary (Menlo Park, CA: Stanford Research Inst1tute International, 1978 , pp. 3-6.

PAGE 249

230 done) . 2 In long-term care, this has resulted in projects with limited lives,3 inadequate data collection,4 and conceptually shallow ana lyt i ca 1 constructs. 5 Thus, l eng-term care research must inc 1 ude a comprehensive and systematic approach to the problem. Multidisciplinary efforts are essential in the health care field. Long-term care is a complicated issue which is not easily addressed through fragmented and disjointed efforts. It thus requires an approach which draws upon the knowledge base of many disciplines in an order-ly, comprehensive and systematic manner. Inadequate and Incomplete Goals and Objectives The goals and objectives of many of the studies have been unclear. Consequently, they sometimes provide answers, but to the 2 Alice Rivlin, sxstematic for Social Action (Washing ton, D.C.: The Brook1ngs Inst1tute, 19 1), p. 142. 3 sharon Soroka, "Quality of Care Evaluation in SNFs Using Bedsores as an Output Indicator" (paper presented at the meeting of the .American Public Health Association, Washington, D.C., November, 1976). (Mimeographed.); Margaret Neilson et al., "A Controlled Study of Home Health Aid Services," American Journal of Public Health, CLXII (1972), 1094-1101. 4 Edgar Bernier and Joan Quinn, '"Project Triage: Toward a Comprehensive National Policy for the Delivery of Home Health Ser vices to the Elderly" (paper presented at the meeting of the American Public Health Association, New York, November, 1979). (Mimeographed.); Linda Scharer and John Boehrenger, Home Health Care for the A ed: The Pro rem of St. Vincent • s Hos ital, New York C1t New 1ty: Boehr1nger Assoc1ates, 976 , pp. 9-5Anthony .Amado, Beatrice Cox, and Rich Mileo, "Cost of Terminal Care: Home Hospice vs. Hospital," Outlook (August, 1979), pp. 522-526; A.L. Greese and R. Fielden, 11 osp1ta! or Home Care for the Severly Disabled: A Cost Comparison," British Journal of Pre ventive and Social Medicine, XXXI (1977), 116-121.

PAGE 250

231 wrong questions.6 For example, several of the comprehensive demon-stration projects which intended to test the viability of substituting home health care services for institutional care, studied only home health services. The result was mere speculation about the potential for substitution of the two services. Had the goals and objectives been clear from the outset, the design and evaluation would have probably acquired more pertinent specificity; hence, the objectives might have been met. Finally, there have been no sys tematic specifications of goals and objectives for the overall long term care research effort. The major unresolved issues that could be addressed through research have not been delineated, and no plan ned, coordinated research program has been developed. Plan ned, systematic research is an important step in any process intended to bring about change in a complex field such as long-term care. Underdeveloped Criteria Many projects suffer from underdeveloped criteria for success. Lack of operational criteria can mean the inability to adequately evaluate the impact of a project. For example, when is day care a vi ab 1 e substitute for nursing home care? If a home care program succeeds in diverting five percent of the patients in a community who would, under normal circumstances, have been placed in a nursing home, has it been a successful program? Clear specifications of criteria for evaluation are imperative at the onset of any study. 6 E.S. Quade, Analysis for Public Decisions (New York City: Elsevier, 1975), pp. 34-44.

PAGE 251

232 Inadequate Design and Data Collection In recent l eng-term care evaluation research projects, design and data collection efforts were inadequate. Many researchers tried to "make do" with data from operational records which were often inadequate for drawing valid policy research conclusions. Sometimes designs did not address the multivariate nature of long-term care programs in a manner that permitted the investigator to determine the impact of any one treatment variable separately from all others. A simple before-after design to test the impact of a coordinated service delivery system which provides care and information referral services, will be inadequate to determine the impact of either the services or the information and referral system. Lack of Comprehensive Cost Measurement Costs are commonly not measured comprehensively. This is a major problem in long-term care evaluation where services are multidimensional in nature. To compare the costs of the 24-hour service in a nursing home to home health services provided intermittently is inappropriate, since the costs of home health care are underestimated. Direct health care program expenditures are commonly the focus of cost analyses. It is noteworthy that many long-term care patients receive support through several publicly funded programs (e.g., Supplemental Security Income, Social Security, Food Stamps, etc.) in addition to the benefits provided by Medicare and Medicaid. Since benefits provided to i nst i tuti ana 1 i zed patients from 'fnccxne

PAGE 252

233 support programs are substantially 1 ower than benefits provided to those at home, it is important to include all public expenditures. In spite of this, many past studies have compared only direct pro-gram costs. Neglect of Organizational and Management Variables Organization of the program and characteristics of the providers are often overlooked in evaluation studies. Home health care is not a generic product nor is nursing home care. They are both affected by the organization and characteristics of the service providers. Home care in most communities is highly fragmented. Literature on organization and management highlights the necessity of viewing the organization and its environment as a system; thus, prompting one to focus on interorganizational arrangements, financial management, legal structure, etc.? For example, few studies have differentiated between hospital-based and free-standing provid-ers, in spite of well-documented differences in unit cost and utilization rates between the two. Many of the demonstration projects described earlier were intended to test the impact of a single entry community-based long-term care system; but f ai 1 ed to differentiate programs with a high degree of control over case management from those with less control. To indict as ineffectual one type of care over another because of the inefficiencies of a 7 oaniel Katz and Organizations (New York: 171. Robert Kahn, The Social Psychology of John Wiley & Sons, Inc., 1966), pp. 149-

PAGE 253

234 particular organizational arrangement does disservice to long-term care policy in the long-run. Research on the organization of long term care and, especially, home care is needed. Inadequate Patient Descriptors Another problem which appears i n current long-term care research efforts is the collection of incomplete and inadequate patient descriptors. Many studies utilize very limited measures of functional status (e.g., dichotomous measures of a six-iten ADL scale), rather than multidimensional patient assessment i nstruments such as The Older Americans Resources and Services (OARS) .8 Because long-term care is a complicated issue, with the medical status of the patient, his economic situation and available social support system affecting his ability to function independently in the com-munity, it is inappropriate to compare the cost-effecti veness of various care modalities on the basis of one or two patient charac-teristics (e.g., age, functional status, or pr imary d i agnosis) . . Although it may be easier for the researcher to use simple measures, these measures may be relatively insensitive to variation in patient needs over time. Composite indices of health status may be desir-able. At a miniml.l11, assessment instruments need to be standardized across research projects. 8 Eric Pfeiffer, Multidimensional Functional Assessment: .. The OARs Methodology (Durham, N.C.: Duke University, 1975).

PAGE 254

235 Lack of Attention to the Longitudinal Dimension Another related problan is the limited life of previous longterm care projects. Most studies are short-term and often only cross-sectional data are used in the analysis. In a chronic disease population, it is especially important to measure the impact of various services on long-term patients over time. For these individuals, it is unrealistic to expect an intervention to cause a measurable change in health status within six months (or a year) from the onset of the intervention. Thus, it becomes the job of the researcher, admittedly difficult, to detect minute changes in patient status over time. Comparison of cohorts of patients over five, ten, fifteen, or more years is necessary to measure the effects of different types of long-term care. Improved techniques are also needed to measure patient outcomes. To repeat insensitive measures does little good. Non-Comparable Desions A major problem across evaluation projects has been inconsistent and non-comparable research designs which prevent valid comparisons across program sites. Some studies follow patients longitudinally, with functional status measures taken at various points in time, while others use control (or comparison) groups to evaluate the impact of care. Most look only at partial program expenditures, while a few try to estimate total costs, usually on the basis of substantially different assumptions about resource consumption. Many studies compare multiservice coordinated home service programs to the normal array of available services. Few I

PAGE 255

236 compare home health patients to similar patients in nursing homes. It becomes a matter of conjecture to discover the relative impact of the care modalities on the different patient populations. Clearly, design consistency and implementation must be improved. In 1969, Schulberg and his colleagues concluded, about the status of evaluation research in health care, that: There is a striking difference between the high quality and sophistication of the theoretical papers in this field and the lower quality research emerging from actual studies of the progr
PAGE 256

237 on-going programs which have not been adequately analyzed. Their availability might improve the probability of obtaining answers to the many long-term care policy questions. With regard to dissemination of information, long-term care is no different than many other areas of health services research. Findings and policy recommendations often sit on the shelves of 1 ibraries and researchers offices, never to find their way to the policy makers or progran administrators where they can be used to aid decision-making. Unfortunately, long-term care researchers are often unaware of others in the same fiel d who d o similar work and are involved in addressing s i milar questions. The progran evaluation staff of several DHEW demonstrat ion projects were brought together after several months of discussion only when it was discovered that the evaluation designs of the projec t s were sometimes inappropriate, inconsistent, and noncomparable. Had a uniform, consistent design been developed jointly by evaluators and federal project officers at the onset of the demonstration programs, the utility of their efforts might have been substantially improved. In view of the several limitations of past research efforts, i t appears likely that a systematic, planned and comprehensive approach to long-term care evaluation research will substantially improve the level of understanding by policy makers in the field.

PAGE 257

238 Systematic Analysis for Long-Term Care Policy Issues With regard to long-term care policy issues, we have seen the "muddling through" of incrementalism for too long.lO Attempted remedial solutions to the problems do not seem to have improved the efficiency or effectiveness of the system. To exclude arbitrarily certain consequences or considerations because of political infeasibility, reduces rational problem solving to no more than a device to help decision makers by contributing to their bargaining power, rather than through the provision of sufficient information intended to improve the appropriateness and accuracy of their decisions.!! The current approach to long-term care policy issues has been conspicuously devoid of any systematic effort. When the existing, incremental approach is compared to systems analysis by Archibald, its importance to long-term care policy becomes obvious. While both approaches recognize that certain alternatives and consequences have to be omitted from an analysis, systems analysts attempt to make those omissions rationally or strategically, while (the current system) is content to have them made quite arbitrarily. The incrementalist feels he can afford to make only minor changes and mistakes because policy-making is serial and fragmented. Problems are never solved; instead some analysis is done, a decision is made, unanticipated adverse consequences show up, more analysis is done, and more decisions are made to remedy the adverse consequences, etc., ad infinitum. The incrementalist feels he can ignore consequences at will because if those ignored should prove damaging to certain groups such groups wi 11 press for new analyses and new decisions. He arbitrarily excludes some lOcharles Lindbloom, "The Science of Muddling Through," Public Administration Review, XVIV (1959), 79-88. llQuade, p. 29.

PAGE 258

consequences from his analysis because he depends on "politics" in a pluralist democracy to provide, eventually, whatever degree of closure is feasible in an obstreperous world. 2 239 Long-term care services are too costly to leave to the vagaries of politics. Analytically spund research can improve decisions which are now left completely to political bargaining. Other important differences between the current, disjointed approach and systems analysis are noteworthy. First, the current system assumes that the cost of analysis and delay (the cost of delaying action) are greater than the costs of error. This is a questionable assumption. In light of the current trends in longterm care expenditures, it seems hard to imagine that more orderly research that is intended to strengthen the decision-making process could be any more costly than the current situation. Second, sys-terns analysis is more 1 ikely to act on the notion of "integrity of design11 than is the current system. In other words, if researchers approached long-term care from a systems analytic perspective, they might say, the consequences of a particular compromise are likely to defeat major intents of the progr therefore if the compromise is a condition of the program'l.:fcceptability, it may be better to have no program at a 11 . 12K. A. Archibald, "Three Views of the Expert's Making: Systems Analysis, Incrementalism and Approach," Policy Sciences, I (1970), 76. 13Archibald, pp. 77-78. Role in Po 1 icythe Clinical

PAGE 259

240 The risk is one of 11throwi ng the baby out with the bath water" compared to the current situation where the bath water is saved, but the baby is forfeited. Definition of Systems Analysis Systems analysis is a comprehensive and systematic approach to policy research which evaluates alternatives on the basis of specific criteria. For the purpose of this discussion, the definition of Quade and Boucher may be helpful. Systems analysis is, ... a systematic approach to helping a decision-maker choose his course of action by investigating his full problem, searching out objectives and alternatives, and comparing them in the light of their consequences, using an appropriate framework-insofar as possible analytic-to bring expert judgement and intuition to bear on the prob lem.I4 Overview of the Basic Elements Successful systems analysis depends on a continuous cycle of identifing program goals and objectives; formulating research problems; selecting objectives and specified criteria; designing alter-natives; examining assumptions, uncertainties, and consequences; creating and applying analytic models; and, applying them to alternatives. It is an iterative process which results in policy adop-tion and implementation, ultimate evaluation, and subsequent feedback into the policy-making system.l5 14E.S. Quade and W.I. Boucher, eds., Systems Analysis and Policy P l anning: Applications in Defense (New York: American Elsevier, 1968), p. 2. 15 Quade, pp. 34-44.

PAGE 260

241 The basic elements of systems analysis include 1) an assessment of the problem for which relief is sought and the environment in which it exists; 2) the identification of progran goals and objectives; 3) the development of models which comprehensively reflect the relevent factors of the issue under consideration; 4} analysis in terms of objectives and criteria; 5) generation of alternatives which are various options available to the policy-maker by which to obtain objectives; and 6} evaluation of the consequences associated with each alternative.The process might then begin over again with feedback from the evaluation into the policy-making cycle, which provides input into policy formulation. The systems analytic process has much to offer the investigator concerned with long-term care policy issues. It is holistic, future-oriented, and orderly. Thus, its application highlights areas of concern previously over 1 ooked. Consequently, it pro vi des insights into those issues which demand further investigation. The components of systems analysis applied to long-term care are discussed in the following paragraphs. Defining the Problem. The first step in the process is defining the program or problem. Systems analysis is initiated when someone perceives a problem exists, or might exist. Initial problem definition (rarely the definition that is finally used) can come from a variety of sources, including the decision-maker, the analyst, the legislator, or the environment (press, interest groups, citizens, etc.). Problems can be categorized by the type of relationship, the degree of specificity, or the mode of analysis approp-

PAGE 261

242 riate to this solution. There are generally three types of problems (operational, program and strategic) and each may require a different analytical approach.l6 Identification of Goals, Objectives and Criteria A key step in systems analysis is the identification of program goals, objectives and evaluation criteria. For analysis to be useful to a decision-maker, it is necessary to determine what results are to be achieved. This is called the goal of the progran or the policy. For home health care, one goal might be to develop a service delivery system which is less costly than institutional care. Goals should be specified in terms of objectives (and sub-objectives) which are operational (measurable) specifications of desired outcomes. Objectives, ideally, state what must be done, the conditions under which it must be done, and standards of successful per-formance. For example, an operationalized objective for a home health care progran might be to care for certain patients as well as or better than a nursing home, but at 25% lower cost. Objectives should be described in terms that are clearly measurable, otherwise it is difficult to know when they have been achieved. They list the actions that have to be taken before a goal can be realized. In long-term care, the specification of appropriate objectives is often difficult because of the global nature of the care modality (i.e., social service, nutrition, medical, etc.). 16Quade, pp. 14-19.

PAGE 262

243 The development of comprehensive and measurable criteria or standards by which the objectives can be evaluated and the program alternatives compared or ranked is the next step in the process. The decision-maker is faced with the problem of how to choose different actions intended to achieve objectives. There are several criteria from which he may choose (e.g., benefit/cost ratio, netbenefit, cost-effectiveness).!? Other useful criteria include per-formance indicators, timing requirements, and risk and uncertainty. The policy literature classifies criteria according to effectiveness measures (e.g., benefit/cost, cost-effectiveness), efficiency (i.e., achievement of objectives at the lowest cost with fewest resources and least undesirable consequences), equity (i.e., fairness), and feasibility (i.e., realistic expectations).l8 In practice, long-term care benefits and costs cannot be easily measured or quanti-fied, but this does not deny the utility of trying to do so. Designing the Model The analyst must next design an analytic model which can be used to eva 1 uate inputs and outcomes to assess the consequences of different alternatives. Unfortunately, most long-term research thus far has devoted few resources to this step which plays several 17Leonard Merewitz and Stephen Sosnick, The Budget's New Clothes (Chicago: Markhan, 1971), pp. 85-92; Peter Sassone and William Schaffer, Cost-Benefit Analysis: A Handbook (New York: Academic Press, 1978), pp. 14-29. 18Theodore Poister, Public Program Analysis (Baltimore: University Park Press, 1978), pp. 8-15.

PAGE 263

244 roles. The model is used to identify particular elements or aspects of the problem under consideration. It explicitly defines the significant relationships among the and facilitates the formulation of hypothesis about causal relationships and interactions. It also helps to identify the type of data needed before a decision can be made. One of the main tasks in long-term care research is to develop 4ppropriate systems analytic models that can be applied on a comparable basis to the evaluation of demonstrations and research projects. Development of Alternatives The development of policy alternatives, which are the options or means by which objectives achieved, is the next step in the process. Because most decision-makers have neither time nor the training to conduct an exhaustive search of the various ways to achieve an objective, designing alternatives is one of the essential tasks of the researcher. Alternatives should be discovered (or invented) by the analyst in consultation with the policy-makers. Ideally, the analyst is supposed to create new alternatives, but genuinely new and workable alternatives are not easy to develop. Often the application of alternatives to new situations is the best that can be done. Evaluating Alternatives Alternatives must be evaluated on the basis of the criteria that are built into the analytic model generated earlier, only after the appropriate data has been collected. There are various sources

PAGE 264

245 of data for the analysis of policy alternatives,. including primary and secondary documents, and major actors in the policy system. There are several data collection strategies available to the analyst, some of which were described earlier. Predicting Consequences After the problems and alternatives have been defined, the analyst must predict the consequences of the alternatives. The consequences, in terms of outputs and outcomes, are generated by the model and interpreted by the analyst. The models are the chief mechanism for predicting consequences. They can be effective problem-solving tools to the extent that they simulate the key relationships in the problem and are appropriately used. Interpretation Which ever model is used, the analyst must interpret its results. This involves the assessment of the significance of the information generated by the model. It requires comparing the impacts of alternatives, ranking alternatives by the criteria developed, drawing conclusions, and, if appropriate, recommending the preferred alternative or the need for more analysis. Ideally, the decision-maker should be actively involved in this phase of the analysis, since ultimately she must exercise her own judgment. Conmunication The final formal step in the analytic process is the communication of findings. This is the primary link between the analyst and the policy world and includes strategically presenting inform .ation,

PAGE 265

246 anticipating clients' reaction, and planning responses to reactions. Knowing how to communicate effectively is a key attribute of a sue-cessful analyst. Too often, long-term care research has been ignored by decision-makers because it was presented in a way that highlighted its limitations, rather than its strengths. Attributes of A National Research Agenda For Long-Term Care Relatively little has been learned during the last fifteen years about the organization and delivery of long-term care. Much work remains to be done if the early gains are to lead to answers to current policy questions. A national, 10 to 15 year, research program in long-term care, costing between $100 to $150 million is essential if necessary policy-relevant information is to be obtained. As is true in much research, first approximations of knowledge are relatively easy to come by; further advances require more elaborate, time-consuming, and expensive efforts. Considering that public expenditures for long-term care are likely to exceed $350 bi 11 ion between 1980 and the year 2000, the proposed research program is modest. The potential for savings is great. Long-term care policy can now benefit significantly from increased emphasis on a long-range and comprehensive planned program of research that encourages more effective use of existing knowledge in policy planning and program development. A systems analysis approach to problem solving can be especially helpful in the development of this needed program.

PAGE 266

247 The systems approach as an organizing tool for long-term care research is basically a planning and reasoning process (based on previous research and familiarity with long-term care programs) which is aimed at identifying the critical variables in the pro gram(s) and environment; thus indicating presumed relationships among them. The resulting.model represents the underlying logic of the system, including outputs of programs (e.g., days of care or number home health visits) which are expected to produce desired outcomes (e.g., change in health status); and environmental factors which are presLmed to influence the ability of the program to meet its objectives.19 A systems approach to long-term care problems views the organ ; zation and behavior of the long-term care systen in terms of the dynamic interactions among the interdependent elements of the system within a contextual environment.20 Thus, the parts of the system (e._g., the availability of alternative health care providers, training of staff, characteristics of the patients and their living conditions, etc.) are viewed as part of larger systems rather than in isolation from one another. Consequently, because systems analysis focuses on the interface between the various elements of the long term care system, it encourages analysis of the design, administration and management of the system. Thus, a systems approach encourages a comprehensive, mul t idimens i ana 1 analytic process. 19Poister, p. 42. 20Poister, p. 33 In the

PAGE 267

248 following section, the key elements of a systematic long-term care research agenda are discussed. Key Policy Questions The first part of the design of any systematic research agenda should be the development of key policy, program, and research ques tions to be addressed, based on a complete understanding of the sys tem and its environment. Care must be taken to avoid the tendency of the existi-ng system to focus on programs, policies, outcomes, and costs which are the most easily measured and for which data are available, irrespective of their policy significance. Rather, it is crucial that the key questions have direct policy relevance (i.e., they are highly related to the key policy issues) so that the answers to these questions will result in information that can be used to reduce the uncertainty surrounding these issues. The resulting list is likely to be composed of highly complicated health and social policy questions. Earlier sections of this volume suggest that the principal pol icy concerns surrounding long-term care on the national level are those of the cost and effectiveness of home care versus institu-tional care. Thus, these are the general topics that should be sys tematically addressed in a planned long-term care research program. The translation of these policy concerns into a research

PAGE 268

249 framework might result in the following specific research questions which the proposed research adgenda should be designed to address.21 Cost. Since cost is probably the most crucial current policy concern, these questions are of a major importance. 1) Which type of care delivers comparable service (and out-come) at the lowest cost? This question is one of determining the difference between the cost of existing services (i.e., nursing home and home health care) and/or alternative service delivery systems (e.g., channeling agencies). Because alternatives may be more costly initially, but in the long run actually save money, it is important to investigate how the costs of any intervention change over time. 2) Are there populations for whom one type of care is 1 ess costly than another? Because the total cost of care for a given population is the weighted sum of caring for the different 11types11 (case mix) of patients who comprise that population, one type of care may be more costly in general than another, but less costly for a particular subpopulation. This issue is of prime importance in view of current cost containment efforts. 21Jay Greenberg, David Doth and Allan Johnson, A Coordinated Approach to the Deliver of Lon -Term Care: Urban and Rural Models M1nneapo 1s: Center for Heath Services Research, Un1Versity of Minnesota, 1980) . (Mimeographed.)

PAGE 269

250 3) Does the use of home health care postpone, avoid entry into, or change the patient mix of nursing homes? Popular belief is that a system of expanded and coordinated home health services will delay entry into and reduce the number of patients entering institutions, while simultaneously reducing expenditures. This question is designed . to test whether this is myth or reality. 4) Does the use of long-term care affect the use of other health and social services? Even if home health care expend itures rose dramatically and increased the total cost of long-term care in the short run, the use of additional resources which serve as a substitute for other, more expensive services (e.g., social services in lieu of high cost health services) may eventually reduce total expenditures for long term health care. On the other hand, a reduction in the use of nursing home care may lead to increased use of acute care services. Thus, to distinguish real savings (from increased efficiency or effectiveness) from illusionary savings (from the transfer of costs from one payer to another), it is important to study the utilization and cost of other health and social services. 5) Does the increased use of professional long-term care services reduce the use of informal (and less costly) services? As the recent General Accounting Office study suggested, a significant part of the long-term care system is the informal care provided by spouse, family and friends, which ranges from an occasional

PAGE 270

251 shopping trip to extensive personal care.2 2 Such care may be sufficient to keep a person in his or her own home and out of an institution. Certainly, the intent of additional public support of long-term care is not to diminish these efforts. Rather, it is expected that formal care will aid the i nformal care. Thus, it is important to examine the. affect of long-term care on the type and quantity of informal service that patients receive. Any substitution of formal, publicly funded care for informal care, will surely impact total public expenditures. Effectiveness. 1) Does one type of care or another favorab l y impact mortality rates over time? Although it is uncertain whether there will be a significant difference in mortality rates between care modalities, the prelim-inary findings of the Georgia Medicaid project suggest this is a possibility.23 Thus, an important public policy question arises. 2) Does one type of care or another favorably impact psycho-social and physical functioning over time? 22camptroller General of the U.S., General Accounting Office, Home Health: The Need for a National Polic to Better Provide for t e der y, Pub 1cat1on No. H D 8-9 Wash1ngton, overnment Printing Office, 1977), pp. 9-22. 23F. Albert Skellie et al., 11Community Based Long-Term Care and Marta 1 i ty: Impact One Year After Enro llment11 (paper presented at the meeting of the American Public Health Association, New . . York, November, 1979) . ( Mimeographed.)

PAGE 271

252 Changes in psychosocial and physical functioning are the key patient outcome measures with which a national long term research agenda should be concerned. Whi 1 e these are known to sometimes deteriorate over time, it is believed by some that with . appropriate support, deterioration can be retarded. If this were true, then a key policy question would become how much money is society willing to pay for improved functioning on the part of the elderly and disab 1 ed? 3) Is one type of care or another preferred by patients and their families? The clamour of organized groups which represent the elderly (e.g., the Grey Panthers), in addition to the demands of an aging Congress suggest that home care is usually preferred to institu-tiona 1 care. Whether this preference persists under all circum-stances is an important public policy question (e.g., are patients willing to accept a reduced l evel of functioning in order to remain in their own homes?) . Cost-effectiveness. What is the relative cost-effectiveness of home health care versus institutional care? Once the costs and effects of the various care modalities are det ermined, it is essential to simultaneously compare their inputs and outputs. The cost-effectiveness comparison must include not only for which types of patients is one type of care more effective and efficient than another, but also for which service mixes and

PAGE 272

253 organizational arrangements. For example, do home health aide vis its, of a particular quantity and frequency, assist in the improvement of patient status?; do channeling agencies coordinate care in a manner that improves patient status more than traditional service de 1 i very mode 1 s? Whether the re 1 at i ve cost-effectiveness of one type of care retains its superiority over another in all instances is an important issue which needs further investigation. Conceptua 1 Framework and Measurement In order to address the key pol icy questions presented, a 1 ong-term care research progr an must adopt a systems approach to measuring costs and effects. In other words, both program outputs and outcomes should be studied in relation to total progran costs over time. The organization of the service delivery system should also be considered because it affects cost and effectiveness. Comprehensive Patient Descriptors. The starting point, or baseline, for any health progran evaluation is adequate and standardized data on the socio-economic, family, medical, health, and functional status of the patient. Without such measures it is difficult to control for variations in case mix when analyzing program effectiveness or costs. In addition, it is difficult, if not impossible, to actually determine the outcomes of care for particular patients without valid and reliable measures of their health status, etc., at entry into the program of interest. These measures should be comprehensive in nature and standardized across programs, to the extent possible, in order to develop the potential for comparisons nationwide.

PAGE 273

254 Effectiveness Measures. Effectiveness measures of the 1 ongterm care system are used to measure program objectives which are the important short and long range goals of the systan. Primary outcomes are those direct effects which the programs intend to produce. These include the improvement (or maintenance) of the patient's functional and medical health status, improvement (or maintenance) of mental status, increased satisfaction, reduction in social isolation, etc. These outcomes can be measured using changes in activities of daily living indexes, changes in the number of days a patient is restricted to his bed or home, changes in number of hospital admissions or lengths of stay, etc. Each of these is an important and necessary direct measure of the system's quality using an outcome approach. Some method of aggregating these diverse measures into a composite, overall measure of effectiveness is also highly desirable. Measures of program outcomes should be standardized across pro-jects in order to facilitate comparisons. Indeed, all research efforts should be required to collect a standardized minimum data set which describes the physical, functional and mental health status of each patient at three or six month intervals. Such a data set, which could be used to measure progrclTl outcome, has already been developed, but not implemented, by the National Center for Health Statistics.24 24u.s., Department of Health, Education, and Welfare, Public Health Service, National Center for Health Statistics, Long-Term Health Care: Minimum Data Set (Washington, D.C.: Government Printing Office, Septanber, 1978), pp. 12-37.

PAGE 274

255 Additional important measures of program effect are those unanticipated outcomes which sometimes are the opposite of those intended. For example, one argument in support of recent regulations which allow nursing homes that violate Medicare quality stand ards to remain open is that the effect of rigid enforcement of those standards in the past has been to require the transfer of patients (who are at greater risk of dying upon relocation) and to reduce the supply of long-term care services in rural areas (where specialized personnel needed to meet the standards are not readily available). Thus, effects on the availability of services to various populations should be an important outcome measure in all long-term care research. Both direct and indirect effects should be studied in each project which is part of the national long-term care research agenda. Using a systems approach highlights the insufficiency of comparing the outputs (e.g., day, visit, etc.). of programs and then inferring changes in outcomes. Outputs are important to evaluate progran efficiency, but outcome is necessary to measure program effect. Each must be appropriate to the particular program and patient under consideration. Cost Measures. First, regardless of the type of analysis, cost comparisons should be based on standardized cost accounting proced ures which break total costs down by the various types of services. Costs should be comprehensively measured and should inc 1 ude, at least, all public costs (including cash and "in-kind" benefits), and ideally private costs.

PAGE 275

256 Similar definitions of costs (e.g., depreciation, capital expenditures, etc.) should be used consistently in the calculations of costs per unit of output. Unfortunately, nursing homes currently use a different method of cost allocation than do free-standing home health agencies; and hospital-based home health programs use another. Thus, recent actions on the part of the Health Care Financing Administration (DHEW) to impose uniform accounting pro cedures on all long-term care providers receiving public funds are especially important to the proposed research efforts. Second, it is important when evaluating home health care and institutional care to compare the total costs of maintaining the individual, rather than direct program costs. An appropriate evaluation can be made only by comparing total costs across modalities (including food, housing, transportation, etc. provided in the institutional setting), since these costs are not included in the cost per uni t of service for home health care. On the other hand, the cost of therapeutic services (e.g., physical or speech therapy) are often excluded from the calcu lation of the average cost per day in nursing homes, but may be included in the cost per day (or per case, episode, etc.) calculation for home care. Consequently, if indirect costs are not added to the cost per unit of service of home health care before comparison with institutional care, home health care costs will be substantially underestimated; and if the cost of therapy is not added to the cost of nurs ing home care, it may also be underestimated. These costs need not be considered when comparing alternative in-home programs, s i nce

PAGE 276

257 they will probably either have no, or a randomly distributed, effect on living expenses. It is not possible to ignore the value of ther-apeutic services when comparing across nursing homes since patients in one home may receive therapy services not included in the average cost per day, while patients in another institution may not receive any, or those therapies they do receive will be included in the average cost per day. Thus, it is especi a 11y important to compare costs at several levels. Third, if possible, especially in comparison across care modal-ities, it is important to consider tertiary (or opportunity) costs. These are those costs incurred because other goods and services might have otherwise been produced with the resources used to provide long-term care. Typical tertiary costs are the value of earnings forgone by a fantily member or volunteer who could have earned income had they not been providing 11free" services. In cer-tain instances, time inputs of patients might also be considered. Although opportunity costs should theoretically be based on the marginal costs of earnings forgone, market wage rates have been used as shadow prices to estimate the value of corresponding opportunity costs. 25 Opportunity costs are important because of the different impacts of programs on such costs (i.e., the amount of volunteer and family services necessary to maintain a patient in the home is greater than if the patient were in a nursing home). 25comptroller General of the U.S., General Accounting Office, Home Health, pp. 18-24.

PAGE 277

258 Fourth, in all cases, it is important to distinguish between cost reduction and cost redistribution. As the modality of the care changes, from nursing home to home health care, the proportion of total costs reimbursed by any particular level (or program) of gov ernment may change. Out-of-pocket expenses may a 1 so change s i nee Medicare Part A currently has no deductible while Medicare Part B includes deductible and co-insurance. Thus, it is important to distinguish between decreases in costs due to increased efficiency and decreases in cost because one level of government (or program) . is shifting its cost burden to another or to private individuals. The importance of differentiating cost saving from cost redistribution makes it necessary to compare total costs. This issue is more important than previously realized when comparing within care modalities. For example, the cost incidence of official agencies providing home health care is probably different than the cost incidence of care provided by proprietary agencies. Thus far, no studies have been completed which investigate this theoretical contention, in spite of its conceptually inherent appeal. A national progrc3'Tl of long-term care research which required the collection of detailed and uniform cost data would permit such analysis. Finally, it is essential to compare costs and effects over a long period of time (at least three to five, and for some purposes ten to fifteen, years). As discussed ear 1 i er, cost per day, per visit, or per month are the units generally used in health services research. In long-term care, these units may significantly distort

PAGE 278

259 findings bee ause they do not detect the shifting of service frequency or intensity with cost. Use of cost per unit of output (case or episode) controlling for outcome of care has been advocated for use in the evaluation of long-term care modalities. The justification is its ability to account for trade-offs between high intensity resource consl.ITlption and low intensity resource consi.ITlption provided with greater frequency. The study described herein docu ments the importance of comparing costs in a manner that takes the mix of services into account over time. This is an especially important issue to long term care policy-makers who are currently faced with (in H.R. 3990) the decision to expand benefits for homemaker/chore services or for occupational therapy. Such questions will continue to dominate the concerns of Congress until analysis which compares the relative cost and effectiveness of alternative care modalities (and service mixes) over time (and outcome) is completed. Duration of the Research Aaenda Determining the appropriate duration of any social research is difficult. Where probable range and timing of outcomes across populations is known, an educated guess can be made. In long-term care research, where most patients are suffering from chronic illnesses, a considerable amount of time must elapse before the complete effects of care become apparent. The duration of any study also depends on the nature of the questions addressed. If they encompass short-term expectations (e.g., death within six months in a hospice program) , then a shorter duration may be appropriate. However, in

PAGE 279

260 order: to address each of the major long-term care policy questions listed previously, the proposed research agenda should be conducted over a period of, at least, five and probably ten to fifteen years. Even though changes in case mix, cost and effectiveness may occur in a shorter period oftime, because the research agenda requires attention to system-wide changes, it is critical to follow a sufficiently large cohort people over time. It is expected that this time frame will be adequate to obtain longitudinal data on a cohort of patients and thereby avoid some of the difficulties of earlier efforts. Of course, in instances where adequate system impacts can be discerned in shorter periods of time (as i n the case of hospice care) selected projects should be conducted with one year durations. In any case, each project should incorporate the necessary time frame appropriate to a systems analytic approach to longterm care research. Control of the Research Currently, management of long-term care research and evaluation efforts are dispersed throughout DHEW, from the Office of the Assistant Secretary for Planning and Evaluation to divisions within the Health Service Administration and the Health Care Financ ing Administration. As discussed earlier, there is little consistency in evaluation design. In addition, management styles of project officers vary widely from laissez-faire to extensive involvement in day-today research activities; monies are so dispersed that the potential impact of any one evaluation is reduced.

PAGE 280

261 Because long-term care research and development is most appropriately fostered at the federal level due to its intensive resource requirements, it is appropriate that efforts be centralized at this level, where expertise can be accumulated efficiently. The establishment of a high-ranking evaluation research office, whose major role is to assure the development and completion of high quality appropriate evaluation research projects would be a major improvement over the current system. Timeliness of the Research In spite of the continued efforts of Congress to encourage timely evaluation and policy analysis, requests for proposals (RFPs) and grant solicitations for the evaluation of long-term care demonstrati on and research projects cant i nue to be issued months, sometimes years after the implementation of the programs. Recently the Health Care Financing Administration (HCFA) awarded 26 contracts to i nst itut i ana 1 and home-based hospice programs throughout the country. These programs should begin providing services in June of 1980. As of April, 1980, the only uniform design or data collection requirements were related to cost information. No base-line data regarding patient status, family situation, etc. have been required by HCFA. Undoubtedly, an evaluation contract will be awarded, prob ably more than a year after the initial provision of services. In addition, the evaluation contractor will be asked to perform the difficult task of comparing non-comparable programs, using inconsistent and incomplete data. Another example of the need to improve the timeliness of long-term care research and policy

PAGE 281

262 analysis efforts is the recently rejected H.R. 3 report that was submitted to Congress by DHEW 18 months 1 ate. A primary concern of a centralized, high quality, federal evaluation office should be the correction of this problem. Dissemination of the Research Findings Current efforts to disseminate research and evaluation findings are, at best, haphazard. States and local governments seldom have available to them relevant findings of the many long-term evaluation projects conducted across the country. Results should be automatically disseminated by the funding agency, in a timely and appropriate manner. Evidence of the need to disseminate findings to state and local policy makers is the recent award of a contract to the Western Interstate Co11111i ssion for Higher Education to prepare syntheses of relevant research in two areas of long-term care: quality assurance in nursing homes and home health care, and reimbursement systems for nursing homes. Similar efforts must be continued and expanded throughout DHEW since long-term care research findings that are not made available to progrcltl administrators and policy makers will, in all liklihood, have little impact on the system. Knowledge for the sheer sake of enlightenment is an expensive commodity that the longterm care system can do without. Centralized Data System Another important responsibility of a centralized long-term care evaluation unit should be the establishment and maintenance of a centralized data bank and library. The computerized data bank

PAGE 282

263 should be the repository for all long-term care information obtained through evaluation and research efforts funded with federal monies. Consistent documentation across data systems should be required of all contractors and grantees, thus enab 1 i ng the reanalysis or completion of new analyses of the data at future points in time. Verification of findings from one study to another could be assisted through such a data bank. The notion of a centralized data bank might eventually be expanded to include a national longitudinal data system covering all elderly people in the U.S. Not an unforeseen event, a recent federal report recommended that such a system be developed.26 Based on the success of their Cleve 1 and study, the GAO suggested that information on the well-being of elderly people (including health and psychosocial indicators) could be collected over time in a manner that permitted the assessment of changes in status, and the measurement of direct program and indicate social costs. Appropriate Actions at the Federal Level This section suggests appropriate actions which might be taken in the near future at the federal level to develop a systematic and comprehensive long-term care national research program. The specific research questions discussed previously, the strengths and

PAGE 283

264 weaknesses of current efforts, and the components of such an agenda were used to develop this list of actions. First, the Department of Health, Education, and Welfare (renamed the Department of Health and Human Services as of May, 1980) might establish a departmental-wide task force to develop a plan for a national research program which is comprehensive and systematic, taking into consideration the available knowledge of and gaps in current long-term care policy information. When a program plan is developed, the Department should seek and secure congres sional legislative authorization for a program housed in the Health Care Financing Administration. Once the program is authorized, annual or multi-year appropriations should be sought. The progrclTl authorization should include a specific dollar amount (e .g., $100 to $150 million) over the next 10 years. Authority should also be provided for waiver of individual program requirements in the authorized research projects, so that broader mixes of services might be evaluated and pooled funding of serv i ce costs might be possible under one set of accounting and auditing rules. Second, the Department should identify the types and nature of research needed to address l ong•term care policy questions through a comprehensive research strategy. The identification of these needs should be approached through a systematic analytic frame\\()rk in order to minimize fragmented or disjointed efforts. Once the relevant types of research are identified, it would be essential for the Department to make its preferences known to researchers and po 1 icy

PAGE 284

265 analysts so that they can direct their efforts towards the needs of the Department. Third, a major responsibility of DHEW should be the development of the tools necessary for a coordinated research effort. The centralized research office, proposed earlier, could be responsible for the development of a comprehensive, standard 1 ong-term care data set, including detailed identification of relevant services, that would be used by all current and proposed research projects so that comparative analysis 'M:>uld be facilitated. Such a data set should include standardized patient descriptors, assessment instruments and comprehensive and uniform cost accounting methods across the various types of long-term care programs. Cost elements should also be consistently specified across programs (e.g., FTE measures, visits by unit of time by type) so t hat comparisons cou l d be made using the same unit of analysis. A methodology for analysis of organizational issues should be developed as a part of the foregoing tasks. In addition to the specification of data elements, DHEW should suggest a set of standardized, basic analyses to be completed for all research projects. In the cases where such analyses require spec i a 1 computer software, the Department should deve 1 op the software and make it available to all relevant projects and researchers participating in the national research program. Finally, DHEW should assllTle a major role in the development of training and instructional materials for evaluators and project managers. Written materials should include information about systems analysis and research design, particularly adapted for long-term

PAGE 285

266 care programs. Seminars should be conducted periodically in convenient locations (using these materials). This notion has ample precedents because similar types of orientation and instruction is currently provided by the National Center for Health Statistics (DHEW) Applied Statistical Training Institute (ASTI) for health planners. The Department clearly has the responsibi 1 ity to assure the best possible research effort for its dollar. More importantly, the long-term care system can no longer afford evaluation and research that does not address major po 1 icy quest ions in a methodologically sound manner. OHEW can take steps to assure that this is done. Conclusion Given the increased size of prospective public and private costs for long-term care of the elderly and disabled in the next several decades, it would be in the national interest to determine which methods of such care are cost-effective. I t i s believed that a well-designed and proper l y coordinated research progrcm along the lines suggested above could, in the end, save taxpayers and citizens many billions of dollars and simultaneously result in better care.

PAGE 286

267 SELECTED BIBLIOGRAPHY BOOKS Berg, Robert, ed. Health Status Indexes. Research and Educational Trust, 1973. Chicago: Hospital Brickner, Philip, ed. Home Health Care for the Aged. New York: Appleton-Century-Crofts, 1978. Katz, Daniel, and Robert Kahn. The Social Psychology of Organiza-tions. New York: John Wiley & Sons, Inc., l966. Kerlinger, Fred. Foundations of Behavioral Research. 2d. ed. New York: Holt, Rinehart & Winston, Inc., 1973. LaPorte, Valerie, and Jeffrey Rubin. Reform and Regulation in LongTerm Care. New York: Praeger, 1979. Leed, M., and H. Shore. Geriatric Institutional Management. New York: Putnam•s, 1964. Merewitz, Leonard, and Stephen Sosnick. The Budget•s New Clothes. Chicago: Markham, 1971. Mack, Ruth. Planning on Uncertainty. New York: WileyInterscience, 1971. Novick, David, ed. Program Budgeting. 2d. ed. New York: Holt, Rinehart & Winston, Inc., 1969. Poi ster, Theodore. Pub 1 i c Program An a 1 ys is. Ba 1 t imore: University Park Press, 1978. Quade, E.S. Analysis for Public Decisions. New York: Elsevier, 1975. Quade, E.S., and W.I. Boucher, eds. Systems Analysis and Policy Planning: t.plications in Defense. New York: Aiiiencan Elsevier, 196 . Rivlin, Alice. Systematic Thinking for Social Action. Washington, D.C.: The Brookings Institution, 1971. Sassone, Peter, and William Schaffer. Cost-Benefit Analysis: A Handbook. New York: Academic Press, 1978.

PAGE 287

268 Schulberg, Herbert, Alan Sheldon, and Frank Baker. Program Evalua-tion in the Health Fields. New York: Behavioral Publ1cat1ons, 1969. Sherwood, Sylvia, ed. Long-Term Care: A Handbook for Researchers, Planners, and Providers. New York: SpectrLm Publications, 1975. Townsend, Peter. The Last Refuge. London: Routledge and Kegan Paul, 1962. Weiss, Carol. Evaluation Research. Englewood Cliffs: Prentice-Hall, Inc., 1972. PERIODICALS .Amado, Anthony, Beatrice Crok, and Rich Mileo. "Cost of Terminal Care: Home Hospice vs. Hospital. i l Nursing Outlook (August, 1979)' pp. 522-526. Archibald, K.A. "Three Views of the Expert's Role in Policy-Making: Systems Analysis, Incrementalism and the Clinical Approach," Policy Sciences, I (1970), 73-86. Berkoben, Rita. "Home Health Care and Quality Assurance: The Experience of the Pennsylvania Assembly Project," Quality Review Bulletin (October, 1977), pp. 25-28. Brickner, Philip W., et al. "The Home Bound Aged: A Medically Unreached Group," Anna 1 s of Internal Medicine, LXXXII (January, 1975)' 1-6. Brickner, Philip W., and Linda Keen Scharer. "Hospital Provides Home Care for Elderly at One-Half Nursing Home Cost," Forum, (November/December, 1977), reprint. Brook, Robert H., et al. "Assessing the Quality of Medical Care Using Outcome Measures: An Overview of the Method," Medical Care, XV, Supplement (September, 1977). Bryant, Nancy, Louise Candland, and Regina Loewenstein. "Comparison of Care and Cost Outcomes for Stroke Patients, With and Without Home Care," Stroke, V (1974), 54-59. Cashman, John, and Beverlee Myers. "Medicare Standards of Service in a New Program -Licensure, Certification, Accreditation," American Journal of Public Health, LVII (July, 1967), 1107Colt, Avery, et al. "Home Health Care is Good Economics," Nursing Outlook (October, 1977), pp. 632-636.

PAGE 288

269 Creese, A.L., and R. Fielden. "Hospitalization or Home Care for The Severely Disabled: A Cost Comparison," British Journal of Pre ventive and Social Medicine, XXXI (1977}, 116-121. Curry, Timothy J., and Bascom W. Ratliff. "The Effects of Nursing Home Size on Resident Isolation and Life Satisfaction," Gerontologist (Autumn, 1973), 295-298. Daubert, Elizabeth A. "A System to Evaluate Home Health Care Services," Nursing Outlook (March, 1977), pp. 168-171. . "Patient Classification System and Outcome Cri n-g Out l oak (July, 1979), pp. 450-454. Davis, Karen, and Roger Reynolds. "Medicare and Utilization of Health Care Services by the Elderly," Journal of Human Resources, X (SllTlmer, 1975), 361-377. Decker, Frances, et al. "Using Patient Outcomes to Evaluate Community Health Nursing," Nursing Outlook (April, 1979}, pp. 278282. Dennis, Lyman, Robert Burke, and Kim Garber. "Quality Evaluation System: An Approach for Patient Assessment," Journa l of Long-Term Care Administration, V (Summer, 1977), 28-51. Densen, Paul M., and Ellen W. Jones. "The Patient Classification for Long-Term Care Developed by Four Research Groups in the United States," Medical Care, XIV (May, 1976), 126-133. Doherty, Nevil1e, and Barbara Hicks. "Cost Effectiveness Analysis and Alternative Health Care Programs for the Elderly," Health Services Research, XII (1977), 190-203. Doherty, Neville, Joan Segal, and Barbara Hicks. "Alternatives to Institutionalization for the Aged: Viability and Cost Effectiveness," Aged Care and Services Review, I (January-February, 1978}' 8-14. Donabedian, Avedis. "Evaluating the Quality of Medical Care," Milbank Memorial Fund Quarterly, XLIV (July, 1966), 166-206. . "Patient Care Evaluation," Hospitals, XLIV -11":1:9"1111":7 olf""!')-, -..131-136 . Edgert, Gerald M., and Joyce E. Bowlyow. "Preliminary Findings: Monroe County's Access Project to Prevent Unneeded Nursing Home Admissions," Perspectives on Medicaid and Medicare Management, (September, 1979), pp. 5-13.

PAGE 289

270 Eisdorfer, Carl. "Evaluation of the Quality of Psychiatric Care for the Aged," AA1erican Journal of Psychiatry, CXXXIV (March, 1977)' 315-317. Feldstein, Martin. "Hospital Cost Variation and Case Mix Differences," Medical Care, III (April-June, 1965), 95-103. Fetter, Robert, et al. "Case Mix Definition by Diagnosis Related Groups," Medical Care, XVIII, Supplement (February, 1980), 1-53. Fletcher, S.W., et al. "Predicting Blood Pressure Control in Hyper-tensive Patients: An Approach to Quality of Care Assessment," Medical Care, XXVII (March, 1979), 285-292. Freiberg, Jr., Lewis. "Substitution of Outpatient Care for Inpa-tient Care: Problems and Experience," Journal of Health Politics, Policy and Law, IV (Winter, 1979), 492-498. Gerson, Lowell, and Owen Hughes. "A Comparative Study of the Econ omics of Home Care." International Journal of Health Services, VI (1976), 543-555. Gibson, Robert, and Charles Fisher. "Age Differences in Health Care Spending, Fiscal Year 1977," Social Security Bulletin, XLII (January, 1979), 3-14. Gottesman, Leonard. "Nursing Home Performance as Related to Resident Traits, Ownership, Size, and Source of Pajment," American Journal of Public Health, LXIV (March, 1974), 269-276. Granger, Carl V., Clarence Sher\'ttlod, and David Greer. "Functional Status Measures in a Comprehensive Stroke Care Program," Archives of Physical Medicine and Rehabilitation, LVIII (1977), 555-561. Greenwald, Shayna, and Margaret Linn. "Intercorrelation of Data on Nursing Homes," Gerontologist (Winter, 1971), pp. 337-340. Gruenberg, Ernest, and Janet Archer. "Abandonment of Responsibility for the Seriously Mentally Ill," Milbank Memorial Fund Quarterly, LVIII (Fall, 1979), 485-506. Hammond, John. "Home Health Care Cost Effectiveness: An Overview of the Literature," Public Health Reports, XCIV (July-August, 1979)' 305-311. Holmberg, R. Hopkins, and Nancy Anderson. "Implication of Ownership for Nursing Home Care," Medical Care, VI (July-August, 1968), 300-307. -

PAGE 290

271 Katz, Sidney, and C. Amechi Akpom. 11Index of AOL,11 Medical Care, XIV (1976), 116-119. • 11A Measure of Primary Sociobiological Func--"":"t..,.io_n_s-,"TT"n--...I-nt:-e-r-n-at:-ional Journal of Health Services, VI (1976), 493-507. Katz, Sidney, et al. 11Studies of Illness in the Aged: The Index of AOL,11 Journal of Americal Medical Association, CLXXXV (1963), 914-919. Katz, Sidney, et al. 11Prognosis After Strokes, Part II: Long-Term Course of 159 Patients,11 Medicine, XLV (1966), 236-245. Katz, Sidney, et al. 11Program in Development of the Index of AOL,11 Gerontologist, X (1970), 20-22. Knapp, Martin. 11Cost Functions for Care Services for the Elderly,11 Gerontologist, XVIII (1978), 30-35. Kosberg, J.I. 110ifferences in Proprietary Institutions Caring for Affluent and Non-Affluent Elderly,11 Gerontologist (Autumn, 1973), pp. 299-304. Kraus, A.S., and M.I. Armstrong. 11Effect of Chronic Home Care on Admission to Institutions Providing Long-Term Care,11 Canadian Medical Association Journal, CXVII (October, 1977), 747-749. Lave, Judith, and Lester Lave. 11The Extent of Role Differentiation Among Hospitals,11 Health Servcices Research, VI (Spring, 1971), 15-38. LaVer, Judith, and Marie Callendar. 11Home Health Cost-Effective-ness: What are We Measuring?11 Medical Care, XIV (October, 1975)' 866-872. Lawton, M. Powell, and Elaine Brody. 11Assessment of Older People: Self-Maintaining and Instrumental Activities of Daily Living,11 Gerontologist, (1969), pp. 179-186. Lehman, J.F., et al. 11Stroke: Does Rehabilitation Affect Outcome,11 Archives of Physical Medicine and Rehabilitation, LVI (1975), 375-382. Lindbloom, Charles. 11The Science of Muddling Through,11 Public Administration Review, (1959), 79-88. Linn, Bernard, et al. 11Validity of Impairment Ratings Made from Medical Records and from Personal Knowledge,11 Medical Care, XII (April, 1974), 363-368.

PAGE 291

272 Linn, Margaret, Lee Guerel, and Bernard Linn. "Patient Outcome as a Measure of Quality of Nursing Home Care," American Journal of Public Health, LXVII (April, 1977), 337-344. Longest, Beaufort, et al. "An Empirical Analysis of the Relation-ship of Selected Structural Factors to Quality of Patient Care in Nursing Homes," Journal of Long-Term Care Administration, III (Spring, 1975), 16-26. . McAuliffe, William E. "Measuring the Quality of Medical Care: Pro cess Versus Outcome," Mil bank Memorial Fund Quarterly, LVI I (1979)' 118-149. McConnel, Charles. "Cost, Size and Quality Structure of Nursing Home Industry," Journal of Health and Human Resources Administration (November, 1978), pp. 134-149. McDowell, Ian, and Carlos Martini. "Problems and New Directions in the Evaluation of Primary Care," International Journal of Epidemiology, V (1976), 247-250. Miglietta, Osvaldo, Tae-Soo Chung, and Vemireddi Rajeswaramma. "Fate of Stroke Patients Transferred to a Long-Term Rehabilitation Hospital," Stroke, VII (1976), 76-77. Mitchell, Janet B. "Patient Outcomes in Alternative Long-Term Care Settings," Medical Care, XVI (1978), 439-452. Moskowitz, Eugene, Forrest Lightbody, and Nanci Freitag. "Long-Term Follow-Up of the Poststroke Patient," Archives of Physical Medicine and Rehabilitation, LIII (1972), 167-172. Neilson, Margaret, et al. "A Controlled Study of Home Health Aid Services," American Journal of Public Health, CLXII (1972), 1094-1101. Newnan, 1\1. "The Process of Recovery After Hemi pl egi a," Stroke, I II (1972)' 702-710. Nobrega, Fred T., et al. "Quality in Hypertension: Analysis of Process and Outcome Methods," New England Journal of Medicine, CCXCVI (January 20, 1977), 91-113. Pfeiffer, Eric. "A Short Portable Mental Status Questionnaire for the Assessment of Organic Brain Deficit in Elderly Patients," Journal of the American Geriatric Society, XXIII (1975), 443441. RolliTl, F .J., et al. "Carrel ates of Outcomes in Patients with Con-gestive Heart Failure," Medical Care, XIV (1976), 765-l67.

PAGE 292

273 Rubenstein, Lis a, Susan Mates, and Victor Side 1. 11Qua 1 ity of Care Assessment by Process and Outcome Scoring, .. Annals of Internal Medicine, LCCCVI (1977), 617-625. Ruchlin, Hirsch, and Samuel Levey. 11Nursing Home Cost Analysis: A Case Study, .. Inquiry, IX (Septenber, 1972), 3-15. Somers, Ann. 11The High Cost of Health Care for the Elderly: Diag-nosis, and Some Suggestions for Therapy, .. Journal of Health Politics, Policy and Law, III (Summer, 1978), 163-180. St. John, Donald B. 11The Illinois Automated Long-Term Care SystemThree Years of Experience, .. Medical Care, XIV, Supplement (May, 1976), 192-197. Stone, J., E. Patterson, and L. Felson. 11Effectiveness of Home Care for General Hospital Patients, .. Journal of American Medical Association, CCV (July, 1968), 145-148. U.S., Department of Health, Education, and Welfare, Health Care Financing Administration. 11HCFA Program Statistics, .. Health Care Financing Review, I (Fall, 1979), 74-80. Wan, Thomas. 11Age Severity of Disability, .. Review of Public Data Use, III (1975), 29-32. Wan, Thomas, William Weissert, and Barbara Livieratos. 11Geriatric Day Care and Homemaker Services: An Experimental Study,n Journal of Gerontology, XXXV (March, 1980), 256-274. Weiss, Carol. 11Where Politics and Evaluation Research Meet, .. Eval-uation, I (1973), 43-49. Weissert, William. 11Costs of Adult Day Care: A Comparison to Nursing Homes, .. Inquiry, XV (March, 1978), 10-19. Widmer, Geraldine, Roberta Brill, and Adele Schlosser. 11Home Health Care Services and Cost, .. Nursing Outlook (August, 1978), pp. 488-493. Winn, Sharon. 11Assessment of Cost-Related Characteristics and Conditions of Long-Term Patients, .. Inquiry, XII (Decenber, 1975), 344-353. Zimmer, James. 11Characteristics of Patients and Cases Provided i n Health-Related and Skilled Nursing Facilities, ... Medical Care, XI I I ( 1975) , 992-1010.

PAGE 293

GOVERNMENT DOCUMENTS Brook, Robert. Quality of Care Assessment: Methods of Peer Rev 1 ew. Rockv i 11 e, MD: Health Services Research, July, 1973. 274 A Comparison of Five Nat1onal Center for Callahan, Wayne. Medicare: Utilization of Home Health Services, 1978. Baltimore: Health Care Financing Administration, DHEW, -mrr. Comptroller General of the U.S., General Accounting Office. Home Health: The Need for a National Policy to Better Providet"'"r the Elderly. Publication No. HRD 78-19. Washington, D.C.: Government Printing Office, 1977. Comptroller General of the Health Care Services: cat1on No. HRD 79-17. Office, 1979. U.S., General Accounting Office. Home Tighter Fiscal Controls Needed. Publi-Washlngton, D.C.: Government Printing Comptroller General of the U.S., General Accounting Office. Condi-tions of Older People: National Information System Needed. Publication No. HRD 79-95. Washington, D.C.: Government Printing Office, 1979. Donabedian, Avedis. Needed Research in the Assessment and rt>nitoring of the Quality of Med1cal Care. Washington, D.C.: National Center for Health Services Research, 1978. Executive Office of the President, Office of Management and Budget. The Budget of the U.S. Government, Fiscal Year 1981. Washing ton, D.C.: Government Printing Office, 1980. Executive Office of the President, Office of Management and Budget. The Bud et of the U.S. Government, Fiscal Year 1981, Appendix. Was 1ngton, D .. : Government Pnnt1ng 0 1ce, 9 0. Giovannetti, Phyllis. Patient Classification Systems in Nursing: A Description and Analysis. Dwis1on of Nursing. Public Health Serv1ce, Publ1cabon No. (HRA) 78-22. Washington, D.C.: Government Printing Office, July, 1978. John. Final Report: Applied Research in Home Health Care Services, Vol. III: Commumt Level Db J1Zabon Anal s1s, Publication No. OPEL 79-3 . Washington, D.C.: Health Services Administration, DHEW, 1979.

PAGE 294

275 Jones, Ellen, Barbara McNitt, and Eleanor McKnight. Patient Class ification for Long-Term Care: User•s Manual. Bureau of Health Services Research and Eva luat1on, DHEW. Washington, D.C.: Government Printing Office, 1974. Katz, Sidney, et al. Effects of Continued Care: A Study of Chronic Illness in the Home. Publlcabon No. (HSM) 73-3010. Washing-ton, D.C.: National Center for Health Services Research, 1972. McCaffree, Kenneth, et al. Cost Data Reporting System for Nursing Home Care: Final Report. Battelle Human Affairs Research Cen ter, Publ1cat10n No. (HRA) 77-3169. Rockville, MD: National Center for Health Services Research, 1977. Moreland Act Commission on Nursing Homes and Residential Facilities. Long-Term Care Regulation: Past Lapses, Future Pros New York: Moreland Act Commission on Nursing Homes and ReSfCfential Facilities, 1976. U.S. Congress. Congressional Budget Office. Long-Term Care for the Elderly and Disabled. Washington, D.C.: Government Printing Office, 1977. U.S. Congress. House. Select Committee on Aging. New York Home Care Abuse. Publication No. 95-145. Wash1ngton, D.C.: Government Printing Office, 1978. u.s. Congress. Senate. Special Committee on Aging. Developments in Aging, 1969. S. Res. 316, 9lst Cong., 2nd Sess. Washlng-ton, D.C.: Government Printing Office, 1970. U.S. Congress. Senate. Committee on Finance. Medicare and Medi-caid: Problems, Issues, and Alternatives. 91st Cong., 1st Sess., Washington, D.C.: Government Printing Office, 1970. U.S. Congress. Senate. Special Committee on Aging. Nursing Home Care in the United States: Failure in Public Policy, Nos. 1-4, 94th Cong., 1st Sess. Washington, D.C.: Government Printing Office, April, 1975. U.S. Congress. Senate. Committee on Government Operations. Subcommittee on Federal Spending Practices, Efficiency and Open Government. Prob 1 ems Associ a ted with Home Health Agencies and the Medicare Program in the State of Florida. Washington, D.C.: Government Printing Office, August, 1976. U.S. Congress. Senate. Special Committee on Aging. Home Care Services for Older Americans: Planning for the Future. 96th Cong., 1st Ses s., May 7, 1979. Washington, D.C.: Government Printing Office, 1979.

PAGE 295

276 U.S. Department of Commerce. Bureau of the Census. Statistical Government u.s. u.s. Abstract of the U.S., 1978. Washington, D.C.: Pr1nt1ng Off1ce, 1978. Department of Health, Education, and Welfare. Administration. Health Insurance Statistics. HI-75. Washington, D.C.: Government Print1ng 1977. Social Security Pub 1 icat ion No. Office, February U.S. Department of Health, Education, and Welfare. Public Health Service. Health, United States: 1978. Publication No. (PHS) 78-1232. Wash1ngton, D.C.: Government Printing Office, 1978. U.S. Department of Health, Education, and Welfare. Public Health Service. Long-Term Health Care: Minimum Data Set. Washington, D.C.: Government Pr1nting Office, September, 1978. U.S. Department of Health, Education, and Welfare. Public Health Service. The National Nursin Home Survey: 1977 Summary for the U.S. Publication No. PHS 79-1794. Washington, D.C.: Government Printing Office, 1979. U.S. Department of Health, Education, and Welfare. Health Care Financing Administration. Office of Policy, Planning, and Research. Medicare: Utilization of Home Health Services, 1976. Research and Statistics Note No. 2. Washington, D.C.: Government Printing Office, June, 1978. Vanell, David, et al. Final Report, the Patient Evaluation Review Co111T1ittee Project 1977-1979. Wiscons1n: Department of Health and Social Services, August, 1979. Walsh, Thomas J., and Michael Koetting. Patient Related Reimbursement for Long-Term Care. Illinois Department of Public Health, 1978. (Mlmeographed.) PUBLISHED REPORTS Effects and Chronical l .: Nat1ona Anderson, Nancy N. A ComMarison of In-Home and Nursing Home Care for Older Persons 1n 1nnesota. M1nneapol1s: School of . Publ1c Affairs, Un1versity of Minnesota, 1977.

PAGE 296

277 Applied Management Sciences. Evaluation of Personal Care Organiza-tions and Other In-Home Alternatives to Nursing Home Care for the Elderly and Long-Term 01sabled. F1nal Report and Execut1ve Summary (Revised). Silver Spring, MD: ADS, May 1, 1976. Assembly of Ambulatory and Home Care Services of the American Hospital Association, et al. A Prospective for a National Home Care Policy. n.p. :n.n., 1978. Birnbaum, Howard, et al., Reimbursement Strategies for Nursing Home Care: Developmental Cost Studies. Volume I Final Report. Cilnbridge, MA: Abt Associates, l979 . Colorado Foundation for Medical Care. The Colorado Experience in PSRO Long-Term Care Review. Denver: Colorado Foundation for Medical Care, October, 1978. Dunlop, Burton. Determinants of Long-Term Care Facility Utilization by the Elderly: An Emgirical Analysis. Working Paper 963-35. Washington, D.C.: Therban !nst1tute, revised March 1, 1976. Greenberg, Jay. Cost, Case Mix, uality and Facility Characteris-tics in Minnesota s Nurs1ng Homes: An xp oratory Ana ys1s. First Year Progress Report. nneapo lis: Center for He a 1 th Services Research, University of Minnesota, 1980. Holmes, Douglas, et al. A National study of Levels of Care in Intermediate Care Facilities. Fina l Report. New Y ork: Community Research Applications, April, l976. Institute of Medicine. Assessing Quality i n Health Care: An Evaluation. Washington, D.C.: Nat1onal Academy at Sc1ences, November, 1977. Kurowski, Bettina. Co lorado Nagi, Saad. An Epidemiology of Disability Adults in the United States. Columbus: Ohio State Universlty, Mershon Cen-ter, 1976. Pfeiffer, Eric. Multidimensional Functional Assessment: The OARS Methodology. Durham, NC: Duke Un1vers1ty, 1975. Pi 1 and, Nei 11. Feasibility and Cost-Effectiveness of Alternative Long-Term Care Setbngs: Execut1Ve Summary. Menlo Park, CA: Stanford Research Institute International, May, 1978.

PAGE 297

278 Ries, Bernard, and Jon Christianson. Nursing Home Costs in Montana: Analtsis and Policy Applications. Bozeman: Montana State Universlty, 1977. Sager, Alan. Learning the Home Care Needs of the Elderly: Patient, Family and Professional Views of an Alternative to Instltutlonalization. Waltham, MA: Levinson Policy Institute, November, 1979. Schlenker, Robert, et al. Applied Research in Home Health Services Volume I: Grant Program Evaluabon. Denver: Center for Health Serv1ces Research, OnlVersHy of Colorado Health Sciences Center, 1979. Evaluation of an Experiment to Hospitals in Utah: Shaughnessy, Peter, et al. An Evaluation of Swing-Bed Experiments to Provide Long-Term Care 1n Rural Hosp1tals, Volume II: Final Technical Report. Denver: Center for Health Services Research, On1versity of Colorado Health Sciences Center, 1980. Willemain, Thomas, Christine Bishop, and Alonzo Plough. The Nursing Home 11Level of Care11 Program. Walthc111, MA: Brandeis Un1Vers1-ty, 1979. (Mimeographed.) UNPUBLISHED SOURCES Bernier, Edgar, and Joan Quinn. 11Project Triage: Toward a Compre-hensive National Policy for the Delivery of Home Health Services to the Elderly... Paper presented at the meeting of the American Public Health Association, New York, November, 1979. (Mimeographed.) Burke, Sheila. 11Home Health Politics and Policy: Senate Finance Committee... Paper presented at the meeting of the American Hospital Association on Hospital-Based Home and Hospice Care, Arlington, VA, September, 1979.

PAGE 298

279 Burton, Robert, et al. "Nursing Home Cost and Care: An Investiga-tion of Alternatives." Durham, NC: Center for the Study of Aging and Human Development, Duke University Medical Center, July, 1974. (Mimeographed.) Callender, Marie, and Judith LaVer. "Home Health Care: Development, Problems, and Potential." Washington, D.C.: Office of Social Services and Human Development, Office of the Assistant Secretary for Planning and Evaluation, DHEW, April, 1975. (Mimeographed.) Coan, Ruth, and Judy Hagebak. "The Alternative Health Services Project: A Prudent Approach to Adult Day Health Care and Other Community-Based Long-Term Care Services." Paper presented at the meeting of the American Public Health Association, New York, November, 1979. (Mimeographed.) Conly, Sonia. "Critical Review of Research on Long-Term Care Alter-natives." Washington, D.C.: Office of the Assistant Secretary for Planning and Evaluation, DHEW, June, 1977. (Mimeographed.) Gentry, J.T. and V.R. Curlin. 11The Illinois Long-Term Care Classi-fication Instrument: Use Experience Within the New York City Medicaid Program." New York: Department of Health, Medical Section, Bureau of Health Care Services, May, 1975. (Mimeographed.) Greenberg, Jay, David Doth, and Allan Johnson. A Coordinated Approach to the Delivery of Long-Term Care: Urban and Rural Models. Minneapolis: Center for Health Services Research, Un1versity of Minnesota, 1980. (Mimeographed.) Jones, Ann, et al. 11Projecting Demand for Public Housing for the Elderly: A Model for Health Planning." Paper presented at the meeting of the Gerontological Society, Dallas, Texas, November, 1978. (Mimeographed.) Kane, Robert. 11Prognosing the Course of Nursing Home Patients ... Los Angeles: Rand, 1978. (Mimeographed.) LaVor, Judith. "Long-Term Care: A Challenge to Service Systems," Washington, D.C.: Office of the Assistant Secretary for Planning and Eva 1 uat ion, DHEW, 1976. (Mimeographed.) Martin, Karen, et al. "Field Testing of a Problem Classification Scheme and Development of an Expected Outcome Scheme with a Methodology for Use," Draft Executive SLmmary. Omaha: Visit-ing Nurse Association of Qnaha, September, 1979. (Mimeographed.)

PAGE 299

280 Seidl, Frederick, Kevin Mahoney, and Carol D. Austin. "Providing and Evaluating Home Care: Issues of Targetting." Paper presented at the meeting of the Gerontological Society, Dallas, TX, November, 1978. (Mimeographed.) Shuman, Larry, and Harvey Wolfe. "The Use of Case Mix and Case Com-plexity in Prospective Hospital Reimbursement." Pittsburgh: University of Pittsburgh, August, 1975: (Mimeographed.) Skellie, F. Albert, et al. "Community-Based Long-Term Care and Mortality: Impact One Year After Enrollment." Paper presented at the meeting of the American Public Health Association, New York, November, 1979. (Mimeographed.) Soroka, Sharon. "Quality of Care Evaluation in SNFs Using Bedsores as an Output Indicator." Paper presented at the meeting of the American Public Health Association, Washington, D.C., November, 1976. (Mimeographed.) Urban Health Institute. "Appropriateness of Long-Term Care Placement; A Study of Long-Term Care Patients in the New Jersey Medicaid Program." East Orange, NJ: Urban Health Institute, September, 1977. (Mimeographed.) Wan, Thomas. "Interpreting a General Index of Subjective WellBeing." Paper presented at the meeting of the Gerontological Society, San Francisco, November, 1977. (Mimeographed.)

PAGE 300

APPENDICES

PAGE 301

282 APPENDIX A Description of Home Health Care Services1 Home health and other in-home services benefits are provided through Title XVIII (Medicare), Title XIX (Medicaid) and Title XX (Social Services) of the Social Security Act. In fiscal year 1977, these programs spent a combined tot a 1 of $1 b i 11 ion ( $425 mi 11 ion for Medicare, $179 million in federal and state funds for Medicaid, and $445 million for Title XX) for in-home care. In fiscal year 1977, in-home services were used by 530,000 Medicare beneficiaries, 208,000 Medicaid beneficiaries, and 1,634,000 Title XX beneficiaries. In addition, a large but undetermined anount of home care is paid for privately by individuals, private insurers, and philanthropic programs. Many persons in need of services who wish to stay at home or in other non-institutional settings have few, if any, benefits covering in-home services available to assist them. Each of the three Social Security Act programs described below has different restrictions on the availability and utilization of in-home services.

PAGE 302

283 Medicare (Title XVIII of the Social Security Act) Medicare is a nationwide health insurance plan for people aged 65 and over, for persons eligible for social security disability payments for over two years, and for certain workers and their dependents who need kidney transplantation or dialysis. Health insurance protection is available to insured persons without regard to income. The progran was enacted July 30, 1965, as Title XVIII-Health Insurance for the Aged--of the Social Security Act and became effective on July 1, 1966. The Medicare consists of two separate, but coordtnated parts: hospital insurance (Part A) and supplementary medical insurance (Part B). Part A pays, after various cost sharing requirements are met, for hospital and skilled nursing facility care and services by home health agencies following a period of hospitalization. Part B covers physician services, home health care (up to 100 visits), medical and other health services, outpatient hospital services, and laboratory, pathology and radiologic services. Participation in Part B of Medicare is voluntary, and any individual over 65 may elect to enroll. About 95% of those eligible for Part A elect to enroll in Part B. Eligibility for Medicare Home Health Services In order to receive home health care under Medicare, a Medicare beneficiary must be confined to his or her residence (homebound), have the services prescribed by a physician and be under the care of a physician, and need part-time or intermittent skilled nursing service and/or physical or speech therapy. Unless these

PAGE 303

284 requirements are met, the Medicare beneficiary cannot receive covered home health services under either Parts A or B of Medicare. In addition, eligibility for Part A home health benefits requires that the beneficiary must have been in a hospital for at least three consecutive days prior to entry into home care. The care to be provided must be for an illness for which the patient received services as a patient in the hospital, and a plan of care must be established within 14 days after discharge from the hospital. Under Part A, a person's coverage is 1 imited to 100 home care vists a year after the start of one spell of illness and before the beginning of a new spell of illness in the year following the last discharge from hospitalization. Under Part B, the Medicare beneficiary must be homebound and require skilled nursing services, but there is no prior hospitalization requirement. For Part B, a beneficiary is limited to 100 home care visits in any one calendar year. Home Health Benefits Under Medicare The Medicare home health care benefits are, by law, oriented toward the need for skilled care. They were not designed to provide coverage for care related to helping with activities of daily living unless the patient requires skilled nursing care or physical or speech therapy. Home health services, as defined by Title XVIII of the Social Security Act, include:

PAGE 304

285 o Part-time or intermittent nursing care provided by or under the supervision of a registered professional nurse; o Physical, occupational, or speech therapy; o Medical social services under the direction of a physician; o Part-time or intermittent services of a home health aide to the extent permitted in regulations; o Medical supplies (other than drugs and medications including serums and vaccines) and the use of durable medical equipment; and o Medical services provided by an intern or resident-intraining under the teaching progran of a hospital which is affiliated or under common control with a home health agency. The statute specifies that these services can be covered if furnished by a home health agency to individuals under the care of a physician, or by others under arrangements with then made by such agency under a plan established and periodically reviewed by a physician. These services are to be provided generally on a visiting basis in the individual's home. Under certain circumstances these services can be provided also on an outpatient

PAGE 305

286 basis at a hospital, skilled nursing facility, or a rehabilitaion center. Medicare Home Health Providers Medicare limits the provisions of home health services to organizations certified as Home Health Agencies (HHAs). Participating HHAs must provide skilled nursing and at least one other home health service. Home Health Agencies must meet all federal, state and local licensure and certification requirements. Proprietary agencies may participate only if they are licensed by the state. At present, twenty states have licensure laws for Home Health Agencies, 16 of which allow proprietary agencies. Only 126 of 2,612 HHA's participating in Medicare are proprietary agencies; the majority are visiting nurse associations or public health departments. However, the limitation on the participation of proprietary agencies has sometimes been circumvented through the formation of private notfor-profit corporations and through subcontracting arrangements. Medicare pays for services provided by an HHA on the basis of the 1 esser of its reasonab 1 e costs or charges. Reasonable cost is defined as, the cost actually incurred, excluding therefore any cost found to be unnecessary in the efficient delivery of needed health services. Utilization of Home Health Services Under Medicare In fiscal year 1977, 530,000 Medicare beneficiaries used inhome services resulting in expenditures of $525 million. Home health expenditures under Medicare have been consistently

PAGE 306

287 increasing. In fiscal year 1974, $100 million was spent on home health compared to $298 million in fiscal year 1976 and $724 million estimated for fiscal year 1979. In 1974, only 393,000 Medicare benefi ci aries used home he a 1 th benefits compared to over 530,000 today. Of the beneficiaries utilizing home health benefits in 1975, 10.4% received visits under both parts A and B while 61.7% used Part A benefits only and 27.9% used Part B visits only. Beneficiaries using both Part A and B benefits used an average of 55.5 visits per year compared to 17.8 visits annually for Part A only beneficiaries and 17.2 visits annually for Part B only beneficiaries. These data suggest that those persons using both Parts A and B benefits are the most in need of such services because although this group represents only 10% of the beneficiaries, they receive about 25% of the total number of visits. Use of home health services under Medicare related to age shows a fairly even distribution among home health beneficiaries. In fiscal year 1975, the 65-70 age group had 101,700 home health beneficiaries receiving 2,137 visits compared to 109,700 beneficiaries in the 70-79 age group using 2,356 visits and 97,200 beneficiaries in the 80-84 age group using 2,076 visits. All age groups averaged about 21 visits per home health beneficiary. Utilization of home health services varies geographically. Over one third of all beneficiaries using home health services reside in the northeast. However, the beneficiaries in the South received the most visits annually and had the highest total charges per person.

PAGE 307

288 Medicaid (Title XIX of the Social Security Act) Medicaid, Title XIX of the Social Security Act, is the major vehic 1 e for financing health care services for 1 ow-income people. It was enacted in 1965 for the purpose of enabling states to furnish the aged, blind, and disabled and families with dependent children, whose income and resources were insufficient to meet the costs of necessary medical services, with medical assistance and rehab i 1 it at ion. Medicaid programs have been implemented in 49 states, the District of CollJJlbia, Guam, Puerto Rico, the Virgin Islands, and the Mariana Islands. Only the State of Arizona has not imp 1 emented a program. Medicaid is state administered under federal regulations. Progran costs are shared by the states and federal government with the federal share ranging from 50% in states with high per capita incomes to 79% in Mississippi, the state with the lowest per capita income. Subject to federal legislation and regulations, states have broad discretion in establishing eligibility criteria, benefit packages, and reimbursement rates. States must provide Medicaid coverage to all people receiving AFDC and, with certain exceptions, to beneficiaries of Supplemental Security Income (SSI), the federalized blind, disabled, and aged welfare program. Income-related eligibility criteria are determined by the states. States may, at their option, extend coverage to the .. medically needy... These are persons or families who meet the SSI or AFDC eligibility criteria (e.g., aged, disabled, etc.) but whose incomes are slightly above welfare levels. States establish the

PAGE 308

289 income eligibility standards for the medically needy, which may not exceed 133-1/3% of the state AFDC paj111ent standard. States also have the option of covering other categories, including: families headed by an unemployed male; children who are financially eligible, but not in a federal welfare category; spouses who are 11essential11 to the well-being of an SSI recipient; and persons eligible for, but who voluntarily elect to decline, AFDC or SSI cash payments. Medicaid plays a significant role in assisting its sister progran, Medicare, in providing health insurance for the aged poor. Approximately 3.9 million aged, 16.9% of Medicare beneficiaries, are also covered by Medicaid. For these people, in most cases, Medicaid both pays the Medicare Part B premiums, coinsurance and deductibles and provides more extensive benefits than are available under Medicare. Most notably Medicaid provides the aged poor with drugs and long-term care services, especially institutional care. In their Medicaid benefit packages, states must cover hospital, physician, skilled nursing facility, family planning, home health, laboratory, and x-ray services. They must also cover early and periodic screening, diagnosis, and treatment (EPSDT) of children under 21, and rural clinic services. They have the option of covering other services such as outpatient prescription drugs, dental services, eyeglasses, intermediate care facilities, prosthetic devices and care for patients over 65 in tuberculosis or mental institutions. If a state's program includes the medically needy, it must provide that group with either the basic required

PAGE 309

290 services or seven of the seventeen optional services authorized for matching funds under Medicaid. Eligibility for Medicaid Home Health Services States are required to provide home health coverage to any Medicaid beneficiary who is covered for institutional skilled nursing care under Medicaid. By statute, states must provide skilled nursing facility (SNF) benefits to adult Medicaid beneficiaries (any individual over 21 years of age). Coverage of skilled nursing facility benefits for individuals under 21 is at state option. Since eligibility for home health services is tied to eligibility for SNF services, Medicaid beneficiaries under 21 are . covered for home health benefits only if their state has opted to cover them for SNF care. All Medicaid beneficiaries over 21 are covered for both home health and SNF benefits. Unlike Medicare, Medicaid does not require a patient to be home bound or in need of skilled care to be eligible 'for home health services. However, a physician must certify that the patient needs home health services. Home Health Benefits Under Medicaid Medicaid, from its enactment in 1965 to 1970, specified 11home health services .. in its list of services to be provided at state option. However, definitions, criteria, and requirements were not included. The 1967 amendments to the Social Security Act mandated home health services effective July 1, 1970. New regulations clarifying the Medicaid benefits and eligibility were published on

PAGE 310

291 Novenber 18, 1976 to clear up the confusion over eligibility and benefits by requiring the states to meet certain basic criteria. Under these regulations, states must: o Provide coverage of nursing, medical supplies, equipment and applfances, and home health aide services to Medicaid home health beneficiaries. o Allow an RN to provide covered services where no organized home health agency exists (LPNs are now excluded). o Permit medical rehabilitation centers to provide therapy services (if they meet the standards as written in the regulations). o Require all agencies to meet Medicare standards of certification or be eligible to meet them. o Define nursing according to each state•s Nurse Practice Act. o Provide home health services for: All categorically needy individuals over 21 years of age if the state plan provides for SNF services; for those under 21 if they are eligible for SNF services; and

PAGE 311

to a 11 corresponding individuals to whom Eligibility shall not 292 groups of medically needy SNF services are available. depend upon need for or discharge from institutional care. o Permit coverage of home health services in an ICF if the ICF is not required to provide such services (such as RN services during a short, acute illness to avoid the need to transfer patients). In addition to the required nursing, medical supply, equipment and home health services, a state has the option of providing coverage for physical, occupational, and speech therapies, medical social services, and personal care services. All services must be authorized by a physician and supervised by a professional nurse. However, although home health benefits are mandatory, states have the discretion to place limits on the amount, duration, and scope of home pealth benefits. Thus, several states place limits on the number of covered home health visits. Personal care services Under Medicaid Personal care services are an optional benefit under Medicare. Nine states cover personal care services: District of Coltmbia, Massachusetts, Minnesota, 1'-bntana, Nebraska, Nevada, New York, Oklahoma, and Wisconsin. Personal care services include health related supportive services, such as assistance with

PAGE 312

293 household maintenance and activities of daily living. Personal care services are provided in a beneficiary's home by an individual who is qualified to receive such but is not a member of the family. The services are to be prescribed by a physician in accordance with a plan of treatment and super vi sed by a registered nurse. Many of the disabled receive attendant care services under the personal care benefit in the nine states that have elected such coverage. Medicaid Home Health Providers Medicaid requirements for participating HHAs are the same as for Medicare. Medicaid also permits states to provide personal care services by individuals not employed by an HHA. Payment rates for home health services under Medicaid in some states are inadequate to attract sufficient provider part ici pat ion. Reimbursement methods and rates for home health as for physician are left completely at the state• s discretion. In contrast, SNFs and ICFs must be reimbursed on a cost-related basis which HEW must first approve. Some states attempt to contain program costs by keeping these rates and in an undetermined number of states the rates are lower than the cost of providing services. Medicaid Utilization of In-home Services In the last ten years, total Medicaid and total Medicaid payments for long-term care services (primarily consisting of nursing home and home health care) have risen roughly five

PAGE 313

294 fold. During this same time span, Medicaid expenditures for home health have increased 25 fold. Even with this huge increase, however, Medic aid in-home benefits st i 11 only anount to about one percent of total Medicaid expenditures. In 1977, Medicaid spent almost $180 million on home health for its 261,331 beneficiaries. Approximately 80% of all expenditures and 70% of all recipients are accounted for by the aged and disabled beneficiary groups. On a state-by-state basis, Medicaid home health benefits constitute about 0.1% to 0.5% of total state Medicaid expenditures for most states. The greatest deviation is New York, which spends 4.4% of its total on home health. New York is also responsible for 63% of all home health recipients and 80% of all national Medicaid home health payments. Between 1975 and 1976, both New York•s beneficiary population and expenditures more than doubled. Hence, New York alone is mainly responsible for the huge increases in Medicaid in-home benefits during the past three years. Federal-State Social Service Programs (Title XX of the Social Security Act) In January, 1974, the U.S. Congress passed Title XX, 11Grants to States for Services, .. with implementation scheduled for October 1, 1975. The legislative goal of Title XX was to enable states to make available services for: o Self support; o Self-sufficiency;

PAGE 314

295 o Protection of children and vulnerable adults from abuse, neglect, or exploitation, and strengthening family life; o Prevention or reduction of inappropriate institutional care by providing for community-based care, home-based care, or other forms of less intensive care; or o Appropriate institutional placement and services when in a person's best interest. Title XX is a grant-in-aid program that allows the states a large degree of discretion in providing a range of social services to their populations. A ceiling of $2.5 billion annually is currently imposed; these funds are distributed to the states on the basis of their populations. The states are required to provide 25% matching and required to publish, in an annual plan, a description of the services they will provide, to whom, and by what methods. Eligibility for Title XX Home Health Services Individuals are eligible for Title XX services if they are eligible for cash assistance, have a low income, need particular social services, or are members of certain designated groups. Fees for Title XX services are mandatory for families whose monthly gross income exceeds 80% of the state's median income for a family of four adjusted for family size. Fees for other individuals are at state

PAGE 315

option. 296 At least 50% of the state's aggregate federal allotment must be used for cash assistance recipients. In-home Benefits Under Title XX Services vary widely from state to state with eligibility and emphasis dependent primarily on decisions made within the state under an open planning process. This needs assessment and planning process gives concerned individuals and organizations a chance to help identify needs, establish priorities, suggest service providers, and assist to coordinate resources to build a systematic services delivery network that responds to the social services needs of local communities. Local government representatives, interested organizations and concerned citizens can help to decide what services should be offered. At least three services must be made available for SSI recipients and at least one must be directed toward each of the five Title XX goals. Information and referral, family planning, and services directed toward the goal of protection may be offered without regard to income. A variety of home-based services--including homemaker, home health aide, choreworker, home management, personal care, consumer education, and financial counselling services--can be provided under a state's Title XX progrcl'Tl. Covered services vary from state to state. It is difficult to present a concise of in-home services delivered under Title XX. However, certain generalities may be noted and patterns observed from one state to another.

PAGE 316

297 The following four services are particularly relevant to helping maintain individuals in their own homes: o Home Health Aide Services are medically related home care activities similar to those provided by nursing aides in hospitals. Such activities include maintaining an individual's health by assisting him or her in carrying out physicians' instructions. These services may be provided under Title XX only if they are an essential but subordinate service provided as part of a social service. o Homemaker Services are described as general household activities (meal preparation, child care, and routine household care) provided by a trained homemaker when the individual who usually performs activities is temporarily absent or unable to adequately manage the home and care for the personal needs of others. o Chore Services are most often described as home maintenance activities (repairs, yard work, shopping, house cleaning) performed by an untrained person for individuals unabl e to do such chores themse 1 ves. often inc 1 uded. Personal care activities are o Home Management Services are described as formal or informal instruction and training i n home maintenance, meal

PAGE 317

298 preparation, budget management, child care, and consLmer education. Home Health Providers Under Title XX There are no federal standards for participation as home health providers under Title XX and, in fact, states many contract with private individuals to provide in-home services. States provide in-home services in the following ways: o Direct provision by individuals enployed by state or local Title XX agency. o Purchase-of-service through contractual arrangement with public . or private (voluntary, non-profit, or proprietary) agencies. (States vary between state-administered and state-supervised programs. In some cases the local Title XX agency contracts directly with the provider agency.) o Independent provider offers service provided by individual who is not affiliated with an agency--may be self-employed or considered under employment to the service recipient. Since states have wide latitude within federal regulations in defining services and establishing regulations for the program, those regulations vary substantially from state to state. Only 12 states responding to a recent HEW survey reported having a licensing requirement for providers under Title XX.

PAGE 318

299 In-home Services Utilization Under Title XX In 1976, over one million individuals received in-home services under Title XX programs. Over 90% of all Title XX beneficiaries are adults who gain access to these services as an SSI or AFDC recipient or by being income eligible. Chore, homemaker, and protective services represent 70% of both total recipients and expenditures. Approximately 56% of these Title XX services are made available through direct provision; 11% are purchased from public sources and 33% from private providers. Most states spend a substantial portion of their total Title XX budgets on these services. For example, California spends 62% of its Title XX monies for this purpose. Many other states spend between 40% and 60%, and most all states spend at least 10% of their total. Thus, Title XX now plays an important role in the provision of in-home services.

PAGE 319

APPENDIX B DATA COLLECTION FORMS This appendix contains the primary data collection forms used by both provider groups included in this study. The Home Health Discharge Abstract Summary Form of the Massachusetts Department of Health is presented on page 301. This form was completed at the agency level and then forwarded to the health department. It was the primary source of patient-and outcome-related data for the Massachusetts sample. Then, samples . of the patient records are presented (pp. 302-304) from which the information relevant to Philadelphia episodes was computerized by staff of the Blue Cross of Greater Philadelphia Home Care Program. Finally, on pages 305-306, the forms used to abstract additional patient-specific information from Philadelphia case records are included .

PAGE 320

---• ....,elllt'\ ' .......... ,.. r z i -.. .10 01. u •• ID I LIJ .._..",.... ' [IJ I [I] ........... 0 I •• n R [ I I 1 • tr rr rt I ! tfl __,...,..,.. c u ! I . i 1 ..., ....... r rr l! It I I I J IU'\ ' • i I t B I I 1 .,.,..,. • SWIQIACMN a lO ............ -....n I .. ,,., . _ .. ,_ ...... c n n -'""' ........... .,. .c 0 D . c ... CIY ''"'M ._,,u..,IMQ s .. , .. .. ,... ' ,._., c 0 LJ_.....,_ ..... r . . , ... L 1.-1 CIY K 0 It It [0 -..rl •...-s -o ................... .., --et---: _ ... -..-.;) ...... , . ... CST. I Cl _..., --...s _..., L I:I .. YN:ISICl SfU.Y c --n -ao ,..,, .. _,_., ._, .. .,.--., .._, . . _ ... ........... .._ --1.1.1$ a .__.__. -Nlll-..... _J_, ......... ,._... ...., . ...._ NOISSIWOY I .... ---s ,._,.., ....ao _...., .... __ .... ) ...... , ... _ .. ........ Cl " " " ft It et IT It f! I I I I I ) • .. I ! ! I I ] 0 0 " -lllltl ....., ' -ao • rt cr rr rr n q I I I I i I 1 " 4 • AO OW . 0 D '':r=: 1 ....,,_'II r _., .. _._.. r ' ...,...., ... ..,.,, ' NO atOll J:>AIQ CIY I i nljll]" R11TJR JffOI AC:'C'WWnS ,, tt u u I I I I I ... 'l ., u I I II e1 I t ll1 I ! , , ., Z' tJ " I I I I '

PAGE 321

! I ! 302 i In-•• .._:e .. 1o. ...... lly tt.. ,.,. ... ,.; •• ,. ... ;l• •u• .. •••' : .. 1 .. ._ ... : .. ,.,. ,.,;..,,•, ;n-•• ......... .,,,;_, ; .. 'lt-lel'1 , , , .................. ,_ .. , ............. ,; ... , ... -c... o.,.,._ .... I I I MOS ... ItO. I HOME CARE DEPARTMENT . ...c:. R!PURAL. fi'ORM c:..sc NO. I ..... CI.AIM NO. I I '"t:--. -Oft IIC,.C .. IUI. I A Oil. OISC)I. AOC css :__ MOS ... Olt S.lf.,.: 04TC OAfC CIT'" TCt.. I'A,.TY ltCSI'ONStel.( ItO"'o. C:lf,.UU:(S TO ll4 TICIIT " .. 0A1 01' a.c. 011 '"' !,Ill. II"T ie. •GC "X C:CitT. NO. o. ""' .. , ... sex • 1'CI 0 aT A us I ,. 11' SCJit. 01'1. ltAC:C •• ...... . . , .... ,. IS tt.L.HIU Cliii'I.OTMCitT ltCt..ATCOf "" c ltCJ!. ONSrlt.l Jlt.ATICN,..S JtCit oie ,,. .. ouc _,.,_,.I.CYCII ' us AOO,.CSS ""' , ... I PLAN OF TREATMEUT-TO BE COMPLETED 6Y PHYSICIAN OtAC NOSI5 (JitiiUIAIIY AIIO SCC:OHO.AitY-IN OIIQ&Itt SURC IC:AL. PRCC:CUR! DATI IIA TICN T IN "Ciii"IO: 014C:MOSIS YCSC NOC 'l'!SC: JtRO HOSIS ,.4Wit.'r IH,O.II,..CO: OIA:,.OSIS 1'UQ i'tOC ltltO:NOSIS '!se ,.o= THI' CQ.a.:,.< M!C c:J.I. SUP!il"IISIOH IH HOM! IT TIL. HO. .A.OO !SS C:tTY %1,.. ' H !AI. TH JII11YSIC:A L. M! CIIC:AI. SP!!C:K oc:c:. HOM! SIR te:!S 1'4URSIHG0. THU.APYCJ SOC:. S!R. C TH!RAI'T 0 TH!RAI'T 0 HEAL. TH .A.IQ 0 MID TIONS : I.Ail RATORT T!STS I 0111 AC:T YITIIS AL.I.C"i1tl0 na. TMIHT .A.HC SP!CI.A.'-!QUIPM!HT I I I SPEC JAI. UCSTitUC:TI:NS, TQ IE IUPQRT!O TQ PHTSIC:IAM ! I I .MT TO II IT 'MTSICIAMt OATI HQMC 0 O"ICI CJ 01"00 P&1!1 I ftcltt S!"YIC:IS TO II i"liCVtQ!O .UC: HUDID TQ Tl-41! C:QHQI'TICNISl 1'0. WHICH Tl!ll ,4t'lll'fT n.e •• VICIS. CURUtG THI :tiLA Tli2 lTAT ll'f A HwS"TAI. a• SXII.I..IQ NURSI NG I'AC:U.ITT. TIS 0 HO 0 I I Ct:l .,. ,. .. , r.c ••r•c"r ,,_ '" .., .. ct.,••• •z• r-c H•••cts ••:t:.rco •• .,, =• •• ••''"""''"'•' h"S. :ooe .. roc.r ..... &• f.C :1 a •twO .• ,._. •C••C• :_,1 #\.•'t Q, "l,.!OOICA ... I.f • j I :ur& ..... ,,_ .. ., ..... ,., .... ,,c, ... , •••C • --..tate., •• .,, I HmPtTAL COPY ' I

PAGE 322

I ! 303 . I i PLEASE .WRITE LECIBL 'f WITH BALL-POINT PEN 1 -i n... Is •• Le .. u .. Lr If. ,,. C:e•• Oc,_,,_,.,, -••• ..... IN••-IIIcel ,_..,._, se..,icec •• tecenl ;,.( .. I• •-•••"'"" ,,.., ""';,,,.,, !.,.,., •isit '"" ;ate lie rep.,tell. Twe cepies el tl•is 1-.,. te lie -ilec •• th., HC:..e o., .,_,., ..... '"• co;.r •• t"'o ett .. ..Ci"t physicle" withewt 4efey. I I I ' AOCitC:'I' I'IIIOVI Qatt.! SCAVtC:C I KOMI C:.AR! I SCAVICC I' A fi!HT PROCRISS UPORT IIIAT'l tltY',._ .. o. ---=--= -I j DATI scavacr . . . ftA.1 C:OtiOITIQN ".ANO PROCR!SS R!JIORTiiMC:&.IIUC ,.,JII.., A!'IIO '"CCTUIG IIIATICIIT"S II.I.NUS1 .. I ' su irJC:I AHO IHSTRUc:TIOH PROVIO!C '" liNT c:.t.R! P\..AH• IHITI.A\. AHO lltOOIJilc:.t. TIONS IIMCI.I.ICC IIICMA.II.ITATION .. I I I QUI STIOHS, R!QU!STS, AHO OTHER C:OMMUHIC:.o\ TIOHS I I I I I I I I I YCIQ YCSQ '''" Tl io• RIPCRT, TO I'IUSICl.AII NOQ MOMC C:A"C IC .. ,U 11' • S•Clta"'IIC lr ....... VJSmNO NURSE C:t PARAMtl)IC:Al SiRVlC:E C:CPY I . . I I I

PAGE 323

I 304 I ' . HOIAE CARE DEPARTMEIIT -- •ae ll1Jif!RR4&. ,OR).& ..... lll&n HT -----------------------------CAst i . TO I! COMPL!T!O BY HO.t"! !'fURS! 0, PA Tll!tiT•s ILI.:I!SS 4!'40 C:URP.!NT C:OliOITIOH CIHC:. "UVIOUS 11tfJfT ..t.L STATUS. PUl'lc:TtQM.U.. t.I.,.IT S.I.1!TY . .. !ASUR!S TO Ill 08SU'I!O, I. . U 11NOIHCS .4li0 s::W11C.:.UT SCO.I.I. IM101llo4ATiOSI -----=-I I --+------------------------------------------------------------------------------1 I ... SIJ l'tiCIS CU.' I OJif "RST VtSITa 'Rt OU!HC:T" CF VIS:TS1 S.T. O.T. . M.S.S. '!.1 TU•I.OfT IH,ORMATlOH RICAROIHC CARl TO 5 CIV!H-------------------------I i I Ml btG\L. SUiliii.I!S AHO 1 I _..._,_____ ------------------------------I o• • nna I HOST-IT.\L c:c;.y I -+-----------------------------------------------------------------------------

PAGE 324

I .I I ! PATIENT DATA HOME CARE PROGRAM 305 Date ------------------Initial -------------------BLUE CROSS OF GREATER PHILADELPHIA I 1. 0. of Hospital Program : I 2. H :nne Care Case No. 3. Age (in years) 4. Pa.tients Sex ! 5. Arrangements of Patient I 6. General Health Status at Discharge "TS I • D Bryn Mawr Hospital G -Mercy Catholic I Crozer .Chester J Paoli 1 -Male 2 -Female 1 -Alone 2 -With Relative 3 Other 4 Unknown 1 -Improved 2 -Same 3 \'lorsened r 4 -Worsened due condition to terminal 5 -Expired 9 -Unknown

PAGE 325

(sponae, .;ha"'er or tub) ! ! ' ! I ! er encc of Urine Conti rtenc• of Feces. Assistance only in bathina a sinale part (e.a •• back) or b4thcs self co8pletely. AdllissiOft -1 Dischara• -1 Ctts clothes closets & drawers; J'Uts on clothes, outer aarments, braces; .. naaes fasteners. (Tleina shoes excluded.) Adaission -1 Discharae -1 Gets to toilet; aets on and off toilet; arranaes clothes, cleans oraans of excretion; aanaae own bedpan used at ni&ht only and uy or uy not be usina •chanica! supports. Ad•ission -1 Discharae -1 Moves in and out of bed indepenand moves in and out of independently (may or may not b• usin1 mechanical supports). Adaission -1 DischarJe • 1 Urination entirely self controlled. Adai ssion -1 Discharae -1 Defecation entirely self controlled. Adaission -1 Disc:harae -1 Gets food fro• plate or its into IIIOUth. Pnc:uttjng of men & preparation of food, as butterin& bread, are excluded fro• evaluation. Adaission -1 Discharae -1 Mobile no human assistance (can use walker, cane, crutches, and/or wheelchair). Ad111ission -1 Discharte • 1 If p3ticnt has at discharre o•it discnarae ADL. Ocpe!!dent in bathinc .are than one of body, assistance in ,cttina in or out of tub; or does not bathe self. Adai ssion 2 Discharae -2 Docs not dress self or r..ains partly undressed. Ad•hsion 2 Oischarae 2 Uses bedpan or co..ode or receives assistance in tina to and usina toilet. Adaission -2 Discharae 2 in .avina in and out of bed chair; does not perfora transfers. Adahsion 2 Discharae -2 or total incontinence in urination, partial or contTOl by or usc of urinals and/or bedpans. Adaission • 2 Di schara• • % rartial or total incontinence in defecation; partial or total control by en...s, lax atives, etc. Adaission -2 Discharae -2 in act of feedin1; does not eat at all or rarent•ral feedinJ. Adllission • 2 Dischar1e % Human assistance needed for patient or patient is bedfast. Admission % Oischarae 2

PAGE 326

I I I . APPENDIX C Data Specifications Each of the variables used in this study is specifically defined below. In addition, relevant data sources are listed accordingly. The reader is referred to Chapter III for a more analytic discussion of these variables. Episode A major area of concern in this study was the definition of an episode of illness. Inconsistencies were found in the recording of this information. Some agencies recorded the admission/acceptance date as the day the patient returned from the hospital or nursing home; others recorded it as the day the client was formal l y accepted into the home health program; still others recorded it as the day the first service was rendered. As one might expect, the same vari-ety occurred in the recording of discharge date. Thus, patient and financial records had to be carefully scrutinized and appropriate adjustments made when necessary. Measurement The number of days from date of first vis it through date of last visit (with no further visits for 60 days) or admission to an inpatient facility.

PAGE 327

308 Data Sources Massachusetts -Massachusetts Discharge Summary System; Phi 1 adelphia-Blue Cross Home Care Program Data System. Unit of Service In this study a visit was defined as a personal contact in the place of residence of the patient, made for the purpose of providing a health service. If a visit was made simultaneously by t'WO or more people from one agency to provide one type of service, for which one person supervised another, it was recorded as one visit. If a provider vi sited the home more than once during a day to provide the same services, each was recorded as a separate visit. If a visit was made by two or more persons for the purpose of providing different services at the same time, each was recorded as as visit. Although some providers record home health aide visits on an hourly basis, all services provided to patients in this study were recorded on a per-visit basis. Although both data sets contained utilization information pertaining to miscellaneous service types, such as nutrition counseling, inhalation therapy, and laboratory services, the frequency of such visits and the subsequent cost of care was relatively minimal. Consequently, a common set of visit types was developed for this analysis which was based on the six Medicare reimbursable services. The direct cost of administrative function and evaluation visits was excluded by both provider groups, since neither type of visit provided direct patient care, nor was it an allowable Medicare cost.

PAGE 328

Measurement The number of visits of each of the following service types: Oat a Sources Skilled Nursing Care Physic a 1 Therapy Speech Therapy Occupational Therapy Social Services Home Health Aide Services 309 Massachusetts Massachusetts Discharge Summary System; Phil adelphia-Blue Cross Home Care Program Data System. Cost Per Unit of Service Cost per unit of service for the Massachusetts agencies was taken directly from audited Medicare Statements of Reimbursable Cost obtained from the Massachusetts Rate Setting Commission. Cost data for the hospital-based Philadelphia agencies were obtained from Blue Cross of Greater Philadelphia. Philadelphia programs contracted for most of their patient care services with community-based agencies, thus the price/charge per visit to the hospital became the cost of care at the hospital level. These charges/costs were then passed on to the payer as the cost of care, with the addition of an administrative surcharge to cover the hospital's cost of case management and overhead. The administrative surcharge included the costs associated with

PAGE 329

' ' . I I 310 identification and direct referral of inpatients to community-based agencies when patients did not require the intensive level of care provided by the hospital-based program. Although these differences in cost measurement affected the comparison between Phi 1 adelphi a and Massachusetts, the main purpose of this study was an assessment of the relative differences in cost per episode within each of the t\tl data sets due to patient, pro-vider, and outcome characteristics. The main purpose of using the two data sets was thus to determine if the pattern of relative dif-ferences was similar to the two settings. Measurement The average cost per unit of service by type of service, cal-culated according to Medicare allowable cost criteria. Table C.1 lists the specific cost per visit values by type of service. Data Source Massachusetts Medicate Statement of Reimbursable Cost for 1976; Philadelphia-Blue Cross of Greater Philadelphia. Cost Per Episode For Massachusetts providers, the sLrn of the products of the average cost per unit of service for each service type times the number of vi sits per episode of each type was the cost per episode for a specific patient. For Philadelphia providers, the SliTl of the product of the average cost times visits did not yield the total cost per episode since the administrative surcharge was not yet included. Surcharges were calculated by each hospital on a patient-

PAGE 330

TABLE C.1 COST PER VISIT BY TYPE OF VISIT . . . .. . . . . . . Massachusetts Philadelphia Provider Provider Provider Provider Provider Provider Provider Provider l 2 3 4 l 2 3 4 Skilled Nursing Care $10.29 $13.76 $15.77 $12.58 $17.57 $19.63 $18.61 $17.31 Physical Therapy 14.07 15.39 12.54 13.56 18.39 17.44 15.34 l7 .45 Speech Therapy a 14.39 11.00 20.41 17.56 19.93 --Occupational Therapy 30.00 12.67 20.29 --18.81 16.50 Social Services -30.11 16.65 19.50 20.52 24.24 Home Health Aide 16.14 17.27 8.62 22.16 6.04 12.53 11.25 11.04 Source: Medicare Cost Reports and Blue Cross Home Care Program Data System. aThe dash(-) indicates that the service was not provided.

PAGE 331

312 specific basis. After addition of the the calculation of total cost per episode for Philadelphia providers was complete. Because the surcharges were idiosyncratic to each space restrictions preclude their presentation herein. Age Measurement The age of the patient was measured in years at the time of admission. Data Sources Massachusetts Massachusetts Discharge Summary System; Phi 1 adelphia-Blue Cross Home Care Program Data Systen. Living Arrangement Measurement The living arrangement of each patient at time of admission to home care was classified as 1 iving alone, 1 iving with relatives, or other living situations. Data Sources Massachusetts Massachusetts Discharge Summary System; Phil adelphia-Abstracted Patient Records.

PAGE 332

313 Primary Diagnosis Measurement All primary diagnoses were recorded at the time of admission and classified according to the International Classification of Diseases, 8th edition into the following groups: Data Sources Infections Neoplasms Endocrine Disorders Blood Disorders Mental Disorders Nervous System Disorders Circulatory Disorders Respiratory Disorders Digestive Disorders Genitourinary System Disorders Skin Disorders Musculoskeletal System Disorders Accidents Other Massachusetts Massachusetts Oi scharge Summary System; Phi 1 adelphia-Blue Cross Home Care Program Data System.

PAGE 333

314 Functional Status-Admitting ADL Index Measurement The total cumulative score on a six-item functional status assessment at time of admission was comprised of the values asso-ci ated with six components of the index, when dependence was given the value of one and independence a value of zero.l (See Appendix B for a copy of the Philadelphia data collection form which provided the specific criteria used to classify patients.) The six items included: Bathing Dressing .Ambulation Toileting Transferring Eating The six scores were summed for each patient, so the aggregate values ranged from zero to six (i.e., totally independent (zero) to dependent Data Sources Massachusetts Massachusetts Discharge Summary System; Phi 1 adelphia-Abstracted Patient Records. loata were collected on continence of urine and bowel for all Philadelphia patients but not included in the current analysis to assure comparability with Massachusetts data.

PAGE 334

315 Surgical Procedure Measurement Presence or absence of any surgical procedure immediately prior to admission to home care program. Data Sources Massachusetts Massachusetts Discharge Summary System; Phil adelphia-Blue Cross Home Care Program Data System. Goal at Admission Me as ur emen t The goal for each patient classified by home care provider at the time of admission: Recovery (Post Acute) Self-Care (Oriented Toward Health Education) Rehabilitation (Active Efforts at Restoration) Maintenance (Little Change in Condition Expected) Terminal Care (Deterioration Expected) None Data Sources Massachusetts Massachusetts Discharge Summary System; Phil adelphia-not available.

PAGE 335

316 Change in Functional Status Index Measurement Total ADL index score at discharge minus ADL index score at admission. Values ranged from minus six (indicating greatest improvement) to plus six (indicating greatest deterioration). Data Sources Massachusetts Massachusetts Discharge Summary System; Phi 1 adelphia-Abstracted Patient Records. Change in Health Status Measurement An estimate by home health care providers of the health status of the patient at discharge relative to status at admission, using the categories: Data Sources Improved Same Worsened Expired Massachusetts Massachusetts Oi scharge Summary System; Phi 1 adelphia Abstracted Patient Records. Length of Use Measurement The nunber of days from date of admission to date of di s . charge for each episode.

PAGE 336

317 Data Sources Massachusetts Massachusetts Discharge Summary System; Phi 1 adelphia-Blue Cross Home Care Program Data System . Intensity Measurement Total number of visits per episode divided by the length of use of home care. Data Sources Massachusetts Massachusetts Discharge Summary System; Phil adelphia-Blue Cross Home Care Program Data System. Primary Source of Payment Measurement The primary source of payment for each episode was categorized according to the following: Data Sources Medicare Medicaid Commercial Insurance Private Pay Other Massachusetts Massachusetts Discharge Summary System; Phil adelphia-Blue Cross Home Care Program Data System.

PAGE 337

318 Size Although the number of professional nursing FTEs was originally intended as the measure of provider size, because most nursing visits were provided under contractual arrangement in Philadelphia, the use of such a measure would underestimate the capacity of the Phil adelphia agencies. Consequently, the ntmber of nursing visits pro vided either directly or through contractual arrangements was used as the measure of size. Measurement The total number of nursing visits, either provided directly or under contract in 1976, was calculated for each agency. The specific values were as follows: Data Massachusetts Phil adelphi a Sources Provider 1 = 3,810 Provider 2 = 8,557 Provi der 3 = 8,247 Prov ider 4 = 11,000 Provider 1 = 2,533 Provider 2 = 2,025 Provider 3 = 2,819 Provider 4 = 1,220 Massachusetts Massachusetts Discharge Summary delphia-Blue Cross Home Care Program Data Systan. System; Phil a-

PAGE 338

319 Acute Care Admissions Measurement The total number of full-service, non-federa 1 hospita 1 admi ssions per thousand total population in the county in which the main office was located in 1976 was used as a measure in this study. The specific values were as follows: Massachusetts Phi 1 ade 1 phi a Data Sources Provider 1 = 9.6 Provider 2 = 12.8 Provider 3 = 7.7 Provider 4 = 9.6 Provider 1 = 13.6 Provider 2 = 17.1 Provider 3 = 10.2 Provider 4 = 15.9 AHA Guide , .1977 and County and City Data Book, 1977. Density The population per square mile within the county in which the main office of each provider was located was used as a measure of urban/rural or metropolitan/non-metropolitan differences. In this study, the population per square mile was highly correlated with the U.S. Bureau of Census definition of SMSA; when the providers were grouped into three density categories, all those located within SMSAs fell into the upper two density groups while those outs . ide of SMSAs fell into the lower group.

PAGE 339

320 Measurement Total population per square mile of the county in which the main office of the provider was located in 1976. The specific values were as follows: Massachusetts Phi 1 adelphi a Data Source Provider 1 = 257 Provider 2 = 1,693 Provider 3 = 587 Provider 4 = 1,571 Provider 1 = 1,277 Provider 2 = 14,087 Provider 3 = 3,187 Provider 4 = 259 County and City Data Book, 1977. Table IV in Chapter III contains a summary of this information.

PAGE 340

APPENDIX 0 UTILIZATION ANALYSES Overview This appendix presents a brief overview of the findings of analyses investigating the relationship between the three categories of independent variables and the secondary dependent variable, total visits per episode of illness. It begins with a sl.lTlmary of the findings of the multi.variate analyses which were intended to discern the overall impact of all three categories of variables on the dependent variable and continues with a discussion of specific variables which were related to utilization of home health care. Since the relative importance of the independent variables in the formula tions of the regression equations developed for total nl.lTlber of visits per episode were highly similar to those found in the regres sion analyses for cost per episode, they are discussed only briefly. The reader is referred to the detailed discussions of the regression analyses in Appendix E of this report for a complete discussion of the relative importance of each of the independent variables in explaining variation in the dependent variable. In general, the overall ability of the independent variables to explain the observed variation in total nl.lTlber of visits per episode was small. The anount of variation explained for the Massachusetts

PAGE 341

322 sample was 11% (R2 = .106), while the variation explained in Philadelphia was 15% (R2 = .147). These values were less than those obtained for the regressions on cost per episode. Patient-specific variables tended to explain the largest proportion of variation, with outcome-related characteristics explaining the next largest portion of variation, and provider/health system variables explaining the smallest portion of variation observed. The following discussion focuses on those variables within each of the three main categories which exhibited a highly significant, either statistically or programatically, impact on utilization of home health care. Findings are displayed in a tabular format. Information contained therein includes frequency distributions and statistical tests for association and .mean differences.1 Patient-Specific Variables The general pattern observed was for vis its per episode to increase with age in both Massachusetts and Philadelphia samples, for all but the oldest of patients. In both samples (Table 0.1), 1To test associations the chi-square statistic was computed. It tests the significance of the differences between the independent variable categories in the distribution total visits per episode categories. To test mean differences analyses of variance was used and the F statistic was computed. The ANOVA (F-test) tests the significance of differences among the mean values of total visits per episode in each independent variabl'e'Category (e.g., the mean number of visits per episode for the various age categories). If only two categories are involved, this test is analagous to a t-test.

PAGE 342

Age 0-19 20-59 60-69 70-79 80+ Total TABLE 0.1 A COMPARISON OF THE MEAN TOTAL VISITS PER EPISODE BY AGE OF MASSACHUSETTS AND PHILADELPHIA PATIENTS I Massachusetts . Philadelphia Mean Mean N Visitsa S.D. N Visitsa S.D. 68 11.27 17.25 14 9.85 4.92 ' . 373 14.78 22.50 198 15.60 13.73 443 17.45 25.90 168 15.20 12.85 755 19.52 31.87 197 18.76 16.73 542 20.44 33.73 102 15.96 13.44 2181 18.26 29.51 679 16.35 14.38 Source: Massachusetts Discharge Summary System and the Blue Cross Home Care Program Data System. 323

PAGE 343

324 there was a positive significant relationship between the average number of visits per episode and age. For the Massachusetts sample, the younger patients tended to use fewer visits than the older ones. The oldest patients used fewer visits than those in the middle age range. The standard deviation for the Massachusetts sample was larger than the mean to which it pertained in all cases, while this was not true for Philadelphia, indicating a wider dispersion in the Massachusetts sample than in Philadelphia about the mean total visits by age group. Similar patterns of utilization occurred for the Philadelphia sample. Younger patients generally used fewer visits than the average, while older patients used more. The eldest of the patients actually used fewer visits than the average. In general, the tendency was for the number of visits per episode to increase with age. Living Arrangement Table 0.2 depicts the variation in utilization accounted for by the living arrangement of the patient at the time of admission. In terms of the overall relationship in both samples, the average number of visits per episode was higher for patients who lived alone than for patients with other types of living arrangements. In supplementary analyses comparing the t'.'tO samples, utilization was not significantly different across the samples when controlling for patient living situations.

PAGE 344

Living TABLE D.2 A COMPARISON OF MEAN TOTAL VISITS PER EPISODE BY LIVING ARRANGEMENT OF MASSACHUSETTS AND PHILADELPHIA PATIENTS Massachusetts Philadelphia Arrangement Mean Mean N Visitsa S.D. N Visits Alone 617 21.. 74 33.66 56 18.12 All Others 1564 16.83 27.58 623 16.19 Total 2181 18.24 29.51 679 16.35 325 S.D. 19.18 13.88 14.38 Source: Blue Cross Home Care Program Data System and Massachusetts Discha.rge Surrmary System. a . p$.001.

PAGE 345

326 Primary Diagnosis Although there were patterns in the utilization of home health services among the various diagnostic groupings, it was difficult to attach any significance to these patterns. Table 0.3 illustrates this point. There was little consistency across the data sets with regard to the impact of primary diagnosis on utilization of home health services. In Massachusetts, patients with blood disorders generally used more visits than the average patient, while those with digestive disorders used the lowest niJTlber of visits per episode. Patients with muscul oske leta 1 and circulatory disorders used more services than the average patient. This may be of particular importance because both diagnoses tended to include patients classified as having a goal of rehabilitation. The impact of the diagnosis on utilization in Philadelphia, although significant, was not consistent in direction and magnitude with that of the Massachusetts sample. A final note regarding diagnostic categories concerns the supplementary analyses comparing utilization across the two samples. This comparison, although not shown in the table, indicated that service patterns, measured by the total niJTlber of visits per episode, were significantly different between the two samples only for two categories of illness (circulatory system disorders and accidents). The Philadelphia hospital-based programs provided more intensive services to these patient types since they often included individuals who were suffering from more acute episodes of illness. The category of circulatory system disorders included those patients

PAGE 346

TABLE D.3 A COMPARISON OF MEAN TOTAL VISITS PER EPISODE BY PRIMARY DIAGNOSIS OF MASSACHUSETTS AND PHILADELPHIA PATIENTS Massachusetts Philadelphia Primary a Mean Mean Diagnosis N Visitsb S.D. N Visi .tsc Infection 45 19.08 34.29 4 12.75 Neoplasm 295 15.58 22.82 150 15.46 Endocrine 160 15.70 28.58 47 19.63 Blood 53 32.47 48.38 4 15.25 Mental 49 16.26 31.92 8 20.50 Nervous 209 22.11 30.95 12 15.83 Circulatory 516 21.78 35.35 237 15.08 Respiratory 87 14.96 16.93 36 16.91 Digestive 233 12.73 25.39 -39 15.12 Genitourinary 56 16.16 38.39 1 77.00 Skin 61 15.57 21.96 5 12.80 Musculoskeletal 233 20.05 25.31 16 21.25 Accident 166 14.46 21.53 17 23.64 Other 21 14.80 16.02 102 16.66 . T otal 2184 18.25 29.49 678 16.35 327 S.D. 8.77 14.25 16.32 5.73 13.91 17.84 14.00 11.04 10.06 0 8.58 21.23 15.94 13.57 14.38 Source: Massachusetts Discharge Summary System and Blue Cross Home Care Program Data System. aA11 primary diagnoses were grouped according to standard International Classtffcation o f (8th Edition).

PAGE 347

328 with the primary diagnosis of stroke, and the accident category included patients often suffering from To the extent that these categories represented patients who required an increased intensity of service, especially in a post-acute situation, this finding was not surprising. Functional Status Table 0.4 presents the patterns of utilization for patients categorized as independent or dependent on each of the components of the ADL index. Consistently for both the Massachusetts and Phila delphia samples, the findings indicated significantly greater utilization by patients classified as dependent for all of the cate-gories. Further, when comparing the Massachusetts s amp 1 e to the Philadelphia sample in the supplementary analyses, there was a significant difference in the total number of visits provided to dependent patients in all categories other than dressing and eating (p.s_ .OS). Functional Status Index When controlling for the functional status index score, the average n1.1Tlber of visits per episode was generally greater in the Massachusetts sample than in the Philadelphia sample for all levels of disability, as shown in Table 0.5. Partially because of the dis' parity in sample size, these differences were not significant when the t\'tQ samples were compared in supplementary analyses using a t-test across al 1 categories, except for the one indicating 11!ndependence in Five ADLs11 (p .s_ .01). Thus, there was a generally

PAGE 348

ADL N Bathing 1198 Dressing 1395 Ambulation 1562 Toileting 1651 Transferring 1614 Eating 1809 TABLE D.4 A COMPARISON OF THE MEAN TOTAL VISITS PER EPISODE BY ADLs OF MASSACHUSETTS AND PHILADELPHIA CASES Massachusetts Philadelphia Independent Dependent Independent Mean Mean Mean Visits S.D. N Vi sits S.D. N Vi sits S.D. N 13.87 24.13 982 23.65 34.22 290 13.23 10.81 389 15.00 24.76 784 24.11 35.75 280 12.86 10.50 399 15.85 24.55 614 24.42 38.82 385 13.89 12.27 294 16.31 25.80 529 24.38 38.26 365 13.52 11 . 52 314 15.89 24.84 565 25.09 39.20 391 13.86 12.10 288 17.26 27.74 372 23.13 36.55 594a 15.99 13.76 85 Dependent Mean Visits S.D. 18.68 16.18 18.80 16.14 19.58 16.21 19.64 16.54 19.73 16.43 18.85 18.03 Source: Massachusetts Discharge Summary System, Blue Cross Home Care Program Data System and Philadelphia Abstracted Patient Records. :aThe only pair of mean values not significant (p=.086). w N 1.0

PAGE 349

TABLE D.5 A COMPARISON OF THE MEAN TOTAL VISITS PER EPISODE BY ADLa INDEX OF MASSACHUSETTS-AND PHILADELPHIA PATIENTS Massachusetts Philadelphia ADL Index Mean Mean N Visitsb S.D. N Visitsb Independent in 6 ADLs 1086 13.32 21.41 259 11.94 Independent in 5 ADLs 227 20.43 31.97 17 25.52 Independent in 4 ADLs 214 22.10 31.58 89 16.79 -Independent in 3 ADLs 106 23.71 30.96 23 17.21 Independent in 2 ADLs 102 20.85 30.72 32 19.25 Independent in 1 ADL 179 29.56 43.83 177 19.87 Independent in None 257 23.21 37.14 82 18.93 Total c 2171 18.30 29.56 679 16.35 330 S.D. 8.86 21.65 15.32 -14.33 15.81 15.65 18.24 14.38 Source: Massachusetts Discharge Summary System, Blue Cross Home Care Program Data System and Philadelphia Abstracted Patient Records. aAbbreviation for Activities of Daily Living Index.

PAGE 350

331 consistent pattern in both samples that utilization increased as dependency increased, up to but not including the most dependent patients. Goal at Admission Table 0.6 illustrates the variation in utilization by patient goal at the time of admission for the Massachusetts sample only since this information was not available at the time of the study for Philadelphia patients. Patients with rehabilitation as their goals received the highest average number of visits and those with recovery as their goals received the fewest. Those classified as self-care, which may have entailed visits for educational/training purposes, received slightly more visits than those classified as recovery and almost the same average number of visits as those classified as terminally ill. Outcome-Related Variables Shown in Table 0.7, the majority of Massachusetts episodes with no change in status received fewer than 21 visits. This pattern was similar to that in the Philadelphia sample. Patients whose conditions deteriorated received, on the average, the greatest number of visits. In comparing the tYtO samples in supplementary analyses, there was no significant difference in the average number of visits between Massachusetts and Philadelphia for those patients who improved, but there were significant differences between all other groups (p .01).

PAGE 351

TABLE D.6 A COMPARISON OF THE MEAN TOTAL VISITS PER EPISODE BY GOAL AT ADMISSION OF MASSACHUSETTS PATIENTS Mean Goal at Admission N I Visitsa S.D. Recovery 507 13.98 16.87 Self-care 613 17.55 26.70 Rehabilitation 276 26.97 36.73 Maintenance 605 21.29 37.18 Tenninal 87 17.70 23.82 None 87 1.35 00.80 Total 2175 18.31 29.53 Source: Massachusetts Discharge Summary System 332

PAGE 352

Change in ADL Index No Change Improved Deteriorated Total TABLE D.7 A COMPARISON OF THE MEAN TOTAL VISITS PER EPISODE BY CHANGE IN ADL INDEXa OF MASSACHUSETTS AND PHILADELPHIA PATIENTS Massachusetts Philadelphia Mean Mean N Visitsb S.D. N Visitsb 1553 14.56 24.70 401 13.89 383 26.95 31.66 230 20.54 170 33.39 48.51 47 17.00 2106 18.33 29.37 678 16.36 333 S.D. 12.45 15.86 17.54 14.40 Source: Massachusetts Discharge Summary System, Blue Cross Home Care Program Data System and Philadelphia Abstracted Patient Records. aRepresents change in Activities of Daily Living Index Score from time of admission to discharge.

PAGE 353

334 Patients in Phil adelphi a whose status remained the sane received the highest average nl.ITiber of visits (18.34), while those deteriorating received the fewest (Table 0.8). Patterns of practice were not the same in Philadelphia as in Massachusetts, nor was the overall outcome of care. Variation in the overall outcome of care was discussed in more detail in Chapters IV and V of this report. As these findings indicate, the patterns for the t'ltO key outcome-related measures in the Philadelphia sample were quite different from those for the Massachusetts sample. This difference was partially due to the differences in programs and patterns of practice. Many patients in the Philadelphia sample whose health status worsened were readmitted to the hospi ta 1 at time of discharge from home care, thereby reducing their lengths of stay and potential uti 1 i zation of the program. In most cases, this readmission to acute care was planned at the time of the original discharge from the hospital. Provider/Health System Variables Primary Source of Payment Table 0.9 presents the findings for the variable measuring the primary source of payment for home care. The average total visits per episode varied significantly by primary payor in the Massachu-setts sample (p .001) and to a lesser extent in the Philadelphia sample {p .10). Episodes reimbursed by Medicare and Medicaid both exhibited higher utilization than those in other categories.

PAGE 354

Health Status Improved Same Worsened Expired Total TABLE D.8 A COMPARISON OF THE MEAN TOTAL VISITS PER EPISODE BY HEALTH STATUS OF MASSACHUSETTS AND PHILADELPHIA PATIENTS Massachusetts Philadelphia Mean Mean N Visitsa S.D. N Visits 977 17.74 22.59 460 16.75 641 10.92 21.14 62 18.34 437 27.36 40.43 121 13.50 85 33.83 54.40 36 17.58 2140 18.30 29.30 679 16.36 335 S.D. 13.96 16.24 13.10 19.15 14.39 Source: Massachusetts Discharge Summary System, Blue Cross Home Care Program Data System and Philadelphia Abstracted Patient Records.

PAGE 355

TABLE D.9 A COMPARISON OF THE MEAN TOTAL VISITS PER EPISODE BY SOURCE OP PAYMENT OF MASSACHUSETTS AND PHILADELPHIA PATIENTS Massachusetts Philadelphia Source of I Payment Mean Mean N Visitsa S.D. N Visits Medicare 1385 20.24 30.35 417 17.08 Medicaid 247 19.68 31.93 34 19.61 Insurance 253 12.54 15.04 223 14.66 Private Pay 201 13.34 34.92 4 6.00 Other 98 11.30 22.31 1 19.00 Total 2184 18.25 29.49 679 16.35 S.D. 15.06 16.53 12.59 2.16 14.38 Source: Massachusetts Discharge Summary Systemand Blue Cross Home Care Program Data System. 336

PAGE 356

337 Size The of provider size to utilization was shown in Table 0.10. There appeared to be a generally downward trend in both sanples. That is, as size increased the average mmber of visits per episode decreased. This implied that larger providers possibly serve their patients more efficiently than smaller. Because there were only four possible values for this measure in each sample, the relationship was, at best, tenuous. Regression findings served to reinforce the presence of the relationship, albiet weak. This pat-tern is worthy of further investigation. Summary Although subtle differences are evident which indicate the need for future investigation (.e.g., service mix variations), the relationships observed between the total minber of visits per episode and the independent variables were similar to those for the primary dependent variable, cost per episode. In an effort to avoid redundancy, those relationships have been only briefly summarized in this appendix. In those instances where patterns were different, a more I careful review of the utilization distributions may be warranted in the future.

PAGE 357

Provider Size 0-1500 1501-5000 5001-10000 10000+ Total TABLE 0.10 A COMPARISON OF THE MEAN TOTAL VISITS PER EPISODE BvSIZE OF .PRQVIDERa FOR MASSACHUSETTS .. AND PHILADELPHIA PATIENTS Massachusetts Philadelphia Mean Mean b N Visits c S.D. N Visits 109 19.82 209 20.40 23.36 570 15.69 1147 19.45 31.54 -828 16.04 27.78 --2184 18.25 29.49 679 16.35 S.D. 19.58 13.07 -14.38 Source: Mass achusetts Discharge Summary System and 81 ue Cross Home Care Program Data System. aProvider size was measured on. the basis Of total number of nursing visits in 1976. 338

PAGE 358

APPENDIX E MULTIPLE REGRESSION ANALYSES Introduction This section presents the results of several of the regression ana lyses conducted as part of this study. Although the specific variables in each of the equations and the samples varied, there was enough similarity in the forms of the equations that it was felt most appropriate to discuss all of the regression analyses together. Each of the regression equations discussed in this section can be written in the following functional form: CPE = f (f., .Q, 1:!) where CPE = Cost Per Episode P = Patient-Specific Characteristics 0 = Outcome-Related Characteri-stics H = Provider/Health System Characteristics Each of these groups of characteristics represented a vector of a set of variables which were included in different combinations in the various regression equations. The general categories of variables were discussed in Chapter III, as were the variables used in each regression analyses. Their specific measurement was detailed in Appendix C. The reader is referred to Table IV in Chapter III

PAGE 359

340 for definitions of the acronyms used in this appendix. Each variable is discussed as it specifically relates to each estimated regression equation. Although each regression equation was also estimated using the secondary dependent variable (total visits per episode), the major dependent variable discussed in this section is cost per episode. To the extent that the relationships identified in the regression analyses on cost per episode differed substantially from findings when using total visits per episode, differences are identified. Since the major focus of this study was an analysis of the impact of the independent variables on cost per episode, equations estimating these relationships serve as the focus of this section. In each section, the independent variables and samples are noted. The regression analysis summaries are presented in the following order. Independent Variables Patient-Specific Variables Provider/Health System Variables Outcome-Related Variables Patient-Specific and Provider/Health System Vari ab 1 es Sample Massachusetts, All Cases Phi 1 ade 1 phi a, All Cases Massachusetts, All Cases Phi 1 adelphi a, All Cases Massachusetts, All Cases Phil adelphi a, All Cases Massachusetts, All Cases Phi 1 ade 1 phi a, All Cases

PAGE 360

All Independent Variables All Independent Variables All Independent Variables Massachusetts, All Cases Phil ade 1 phi a, All Cases Massachusetts, Rehabilitation Cases Massachusetts, Terminally I 11 Cases 341 In terms of empirical techniques, ordinary least squares and forced stepwise regression were used for all equations. A listwise deletion technique was used to exclude missing data. The approach was basically statistical and was used because it allowed an empiri-cal estimation of the equations of the type presented in the general form at the outset of this section. All regression equations were estimated separately for both samples since the generalizability of a pooled sample was believed to be limited. To the extent that the functional forms of the esti-mated equations were consistent across the data sets, the reliabil-ity of observed patterns was increased. The analytic framework which served as the basis for the .regression analyses, as well as the results of the canonical carrelation analysis, prompted the estimation of regression equations in a staged process. First, all equations were estimated utilizing only the patient-specific variables. Then, provider/health system variables were added to the equations. Finally, outcome-related variables were added. In this way, the relative impact of each of the three vectors of independent variables on the overall R 2 could

PAGE 361

342 be evaluated. Findings are presented in this same order. Selected final equations for subsamples of the populations are presented at the conclusion of this Appendix. The discussions of each of the analyses that follow are organ-ized in the same manner. First, when appropriate, preliminary statistics are presented on the means, standard deviations and zero-order correlation coefficients for all the included variables in each major section. Second, the approach and purpose of the analysis are noted and the sample is described. Third, the independent variables are specified and discussed.l Fourth, selected regression results are presented and discussed. The regression tables presented throughout this section include the following items. The dependent variable is l isted, and its mean and standard deviations are presented. The sample size is indica ted. For each estimated regression equation, the coefficient of each variable and its significance are presented. Significance is presented using the t-test, and the results are generally presented for the two-tailed test.2 The two tailed t-test is used to test the null hypothesis that the coefficient is zero against the alternative hypothesis that it is not equal to zero. A one-tailed t-test is used to test the null hypothesis against the alternative hypothesis 1 As noted ear 1 i er, the dependent v ar i ab 1 e in a 11 regression equations discussed in this section is the cost per episode of home health care. A detailed discussion of its conceptual and intuitive meaning can be found in Chapter III of this report. 2 The one-tailed t-test was used for the following variables: AGE, ALONE, ADAOL, and YNSURG.

PAGE 362

343 that the coefficient is actually either positive or negative. The level of the significance of the test is the probability that a value as large (either positive or negative) as the estimated value of the coefficient \'tOu,ld result if the true value were actually zero. The final items presented in the regression tables are the R2 and the statistical significance of each overall equation. The equation's significance is tested using an F-test, which tests the null hypothesis that the estimated regression coefficient is actually zero. Major Regression Analyses Preliminary Statistics The means and standard deviations of the dependent and independent variables included in the primary regression analyses are presented in Tables E.1 and E.2 for the samples of 2,101 Massachusetts episodes and the 677 Philadelphia episodes, respectively. Tables E.3 and E.4 present the zero order correlation matrices for all major variables included in this study for the t\'tO samples. Because listwise deletion was used in the development of the matrices, the sample sizes of the correlation matrices were slightly smaller than those of the regression analyses. Since the extent of the variation was minimal, its impact on the correlation coeffi-cients was negligible. Coefficients of .04 or better in the Massachusetts matrix and .06 or better. in the Phi 1 ade 1 phi a matrix were significant at the .OS level.

PAGE 363

TABLE E.l MEANS AND STANDARD DEVIATIONS OF SELECTED VARIABLES FOR ALL MASSACHUSETTS CASES (N=2101) Variables Mean S.D. CPE $259.04 448.28 AGE 68.90 17.39 ALONE .28 .45 NEOP . 14 .34 EN DOC .07 .26 BLOOD .02 . 15 MENTAL .02 . 14 NERVS . 10 .30 CIRCU .24 .42 RESP .04 . 19 DIGEST .11 .31 GENITO .03 . 16 MUSCO .11 .31 ACCID .08 .27 ADADL 1. 75 2.21 YNSURG .36 .48 ADLCHG -.19 1.53 BETWORSE .76 .43 DEAD .04 .20 MEDICARE .64 .48 MEDICAID .11 .32 INSUR . 12 .32 SIZE 8913.78 2054.75 ADM ISS 10. 17 1. 93 344 Source: Massachusetts Discharge Summary System, Medicare Cost Reports and AHA Guide.

PAGE 364

TABLE E.2 MEANS AND STANDARD DEVIATIONS OF SELECTED VARIABLES FOR ALL PHILADELPHIA CASES (N = 677) Variable Mean CPE $521.82 AGE 64.56 ALONE .08 NEOP .22 EN DOC .07 BLOOD . 01 MENTAL . 01 NERVS .02 CIRCU .35 RESP .05 DIGEST .06 GENITO .00 MUSCO .02 ACCID .02 ADADL 2.60 YNSURG .40 ADLCHG -.92 BETWORSE .77 DEAD .05 MEDICARE .61 MEDICAID .05 INSUR .33 SIZE 2294.20 ADM ISS 13.71 345 S.D. 442.32 15.94 .28 .42 .25 .08 .11 . 13 .48 .22 .23 .04 . 15 . 1 6 2.39 .49 2.11 .42 .22 .49 . 21 .47 553.80 2. 71 Source: Blue Cross Home Care Program Data System, Philade l p hia A bstracted Patient Records, and AHA Guide.

PAGE 365

Ho•trh (or the Coet Per frltock'lr&rt'•tlon S.all!'h for "-•••duucttt C8tcl (N•l06S)8 'IOM1 l.H. CPI . 81 1.00 Act .01 . 07 1,00 ALOHI . 01 . 01 • U I :oo INFECT HEOI' !HOOC .oo •• oo • . ao • . oJ 1.oo .04 • • OJ . 04 . 01 .... 1.00 •• OJ .01 -.OS -.01 . • 04 -.11 1.00 11.000 .85 .05 .04 . OJ -.IZ -.06 -.04 1 .00 • HI!NTAI. .00 .01 . 04 .01 •• OJ • . 06 ,.04 •.OJ 1,00 IIElVS .OS .OC . 01 .01 .tl •.IJ .ot -.05 -.05 1 . 00 CIIICU . 01 . 07 .U .00 -.06 . U -.16 •.01 -.01 -.11 1 .00 IESP . OJ .OJ -.OJ .01 •• OJ -.01 •.06 -.OJ .OJ -.07 -.11 1 .00 DICtsT -.07 . 06 .00 -.04 .OS -.14 • . JO -.OS -.OS •.ll -.It . 07 1 .00 C . 01 -.OJ .OJ -.00 -.01 -.06 -.05 -.OJ -.01 -.05 -,01 .OJ .06 1.00 • I I \IJ '-'-• '• >..1 SliM -.01 -.01 -.04 . OJ -.IJ -.07 •.OS -.OJ -.OJ •.06 •.10 .OJ -.'06 .OJ 1 .00 IIJSCO .01 .01 .OJ . 04 -.05 -.Jc -.1o .os -.05 •• IJ -.It -.o7 -.u -.06 . ... a .oo ACCJD -.04 . 04 .01 .01 .14 •.U -.01 .04 -.04 -.10 -.16 . 06 -.10 -.05 . 05 -.10 1 .00 .. Y14SlllC . 17 .11 . 01 -.JJ -.00 . U -.01 -.07 -.OJ • . 05 . Ot .IJ • • IJ .OJ -.07 .00 .OS 1 .00 .01 -.oo -.Jo . 04 .01 .J6 .11 • • o7 -.1o . u . 16 •• u . o7 .os .01 .u .oo -.n a . oo .06 . 07 -.ot ,OJ -.01 .01 -.00 •.01 .OJ . 04 .OS .05 .04 .OJ -.11 -.01 .00 -.04 . U 1 .00 TUH -.01 -.00 -.05 -.Ot -.OJ .46 •.05 -.OJ -.OJ • . 07 -.ot •.OJ -.06 -.00 . 04 •.07 -.06 .10 •.00 . 04 1 . 00 REIIAI .U .JO • . 01 -.OJ • . OJ . 10 -.OJ .OJ .00 -.07 . 05 .04 . 11 • • OJ -.04 . 11 .ll .IJ . 05 . 05 -.01 1.00 ADLOIG -.OJ • • OJ • • OJ .OJ -.OS .U .01 .04 ,OJ .00 •.01 .U .OJ .OJ .01 -.01 . 14 •.ll • . 06 . 00 .II -.IJ 1.00 llSin .11 .15 .06 .14 • • 01 . 11 .OJ . 06 .OJ •• Of .06 .OJ -.ot .OJ •• OJ -.OS • • OJ .JJ -.U -.OJ .II .00 .JI I . M PATSTAT ,IJ ,IJ .05 -,10 •• 04 .U .01 .OS .01 . U , 01 .OJ .07 .01 . 04 -.01 -.11 .Jl . Zl -.07 .SC .OJ . 45 . 67 1.00 JElVOI$1 • • JO •• Jt •. 01 . 07 ,OJ -.U . 01 -.OJ .04 . 10 .09 t.OS .ll .OJ .01 . 01 .01 •.U .IJ -.It •,JJ .OJ • .JI •.70 •.IS 1.00 llEAD ,II .11 .OJ • • 01 -.OJ .10 -.05 .OJ •• OJ . 04 .01 .01 -.04 .OJ . 01 . 04 -.04 .zz -.01 •.00 .JS • . OJ .11 . 16 .$1 .J7 1 .00 IIEDICAII .Ot .ot . U .14 .,01 .01 •• oz -.06 •• JO .OJ . 07 .II' •.04 .01 •.01 . 01 .05 . 06 . 05 . 10 • . OJ .01 • . 07 .It .04 .OS .01 1.00 HDICAJD .OJ .OJ •. JO .OJ .00 •• 01 . 04 .05 .OJ .01 • • oo .01 . 01 .OJ .OJ . 00 -.OJ . 01 •.07 -.04 : oo . 01 . 05 .OJ . 01 . 04 .01 -.47 1 .00 IKSUA -.07 .07 .CS •.17 ,OS . 01 .OJ .OS -.01 •. OJ •.01 •. OJ .01 .01 .01 .01 .01 .01 . U . 01 . 01 . OJ . 01 -.OJ . 04 .OJ .01 • . ct . U 1 .00 fPAY -.06 •.06 •• 05 .OS •• 01 -.01 .04 ,IJ . 01 -.05 -.00 -.01 .U .OJ . 01 •• 01 • . 04 • , 01 • . U •.It -.05 . 05 . 04 -.05 .01 .ot -.OJ . 40 . 11 -.11 1.00 ltfLOS .60 .11 .OS . 04 •• oc .07 . 05 .10 .OS . 00 .Ol -.10 .OJ -.OJ .OJ . 04 ."10 -.11 .OJ . 06 . 10 .OJ . U .U -.14 .17 -.04 ,IJ . 06 . 01 1.00 lUI -.05 •• 01 • . 01 .04 •.II .OJ .01 .00 -.04 .OJ •.06 .01 .06 . 01 . 05 . 00 -.06 ,JI4 .05 .OJ .OJ -.04 -.OJ •.01 -.06 .07 .01 .01 •,01 -.OJ .04 -.06 1.00 DeliS -.OJ .OJ .01 .01 . , n .OJ .01 -.OJ •• oo . 11 . 07 .01 • • OJ .01 . 05 .01 • .06 -.00 .ot .14 .01 •.01 .OJ -.14 .U .06 .OJ , U • . oc .OJ -.U .OJ ,71 1 .00 ADIIISS .OJ .06 ,OJ .01 •.06 . . 01 -.00 -.OJ .04 .IS .OJ •.00 . 10 .01 .01 .OJ • .OJ -.04 . 07 . IS -.01 •.01 .OS •.U .II .OJ .01 ,IJ •.OJ •• oz • • zo .10 • • 01 .61 1.00 INnNS •.01 .01 .01 .02 .04 .01 . 04 •.II -.OJ . 11 -.IS -.ot .17 .01 .01 -.06 •,04 .01 .. U . 05 . 10 .01 .04 .OJ . 01 -.01 .OJ . • • oJ •. 01 .OJ .07 -.17 .07 .OJ .10 1.00 I:! 1!1 1 -II i

PAGE 366

TAILEG.41 TABLE E.4 Corrolotlon N.Hrh for CaoJ hr Eploodo lcarrulan S•orh for Phlladol('lllo Cun (NU)• 1.00 CP! " a .oo AGe .01 .01 1.00 . AI.ONI .04 .OJ .OS 1.00 Sex .01 .OJ .OS .10 1.00 INFECT .01 . 01 .01 -.ol .oz 1.00 IIP -.OJ .OJ -.01 -.01 •• OJ -.04 1 .00 ENDOC ,06 .o4 .01 .ot .01 -.u . u 1.00 .. .000 -.01 -.01 .04 -.01 .04 -.01 -.04 -.01 1.00 .OJ .01 .04 .01 .01 -.01 -.06 -.OJ -.01 1 .00 NEIYS .00 -.01 -.U -.04 .06 .01 . 07 -.04 -.01 -.01 1.00 CIICU -.06 -.06 .11 .OJ -.el -.OS •. Jt -.20 -.06 . 01 . 10 1.00 USP .01 .01 .OJ .01 • .II -.01 -.U . 07 .-01 -.OJ -.OJ -.11 1 . 00 . 01 •.01 .06 ,OJ • .IJ -.01 -.U -.07 -.01 -.OJ -.OJ . 11 '-. 06 1.00 .16 .IJ .OJ •• 01 -.04 -.00 . 02 . 01 -.00 .,to . 01 • .OJ . 01 •• 01 1.00 SliM -.01 .OJ .01 .04 .tS -.01 -.OS .OJ . 01 -.01 -.01 -.06 .OJ . 01 .00 1.00 NJSCO .os .06 -.01 -.o1 -.os •• 01 -.01 -.04 .01 . o2 -.01 -.11 . 04 -.04 -.01 -.o1 1 .00 ACCIII .00 .ot , 07 -.01 .M .01 • • Of -.04 .01 . 02 -.OJ . U -.04 . 04 . 01 -.01 •. OJ 1.00 'AOAOL .u .11 .u -.u .01 -.06 .u •. o1 -.o4 .n .01 . u -.o1 •• o7 .os •• os .02 .o1 1.00 YNSUIIG .IJ .14 -.IS .00 -.04 .01 .It -.01 -.01 . 06 . 01 -.U -.01 . U .OS .07 .01 . 06 .OJ 1 . 00 AllLOIG -.11 -.11 . 04 .01 . 06 .01 .11 .00 .01 • • 01 . 02 .06 .OJ . 04 .01 -.01 -.OJ . 01 -.U -.IJ 1 .00 IJSITI •.07 .01 , 04 -.04 -.06 .OJ .14 -.07 .04 ,02 -.01 . 01 .OS -.04 .OJ -.01 -.OJ . 06 .JO .OS . H 1.00 PATSTAT -.OS .OS .Ot -.07 .OS .OS .Z7 -.07 -.01 .OS .OJ -.07 .OS -.10 .06 -.01 -.OS . 01 .J4 -.10 .47 .U 1 . 00 tWOilSI .01 .01 -.01 .01 .04 .01 .16 .OS . 00 -.04 .OS .OS . 04 .11 -.07 .01 . 04 .ot .JI . 09 •.4J -.11 .IS 1.00 PEAO .01 . 03 . 01 .00 -.OS .01 .16 -.07 -.01 -.OJ -.OJ -.06 .OS . 06 -.01 .OJ .01 . 04 . 10 .OJ .16 . 56 .51 • . U 1.00 MEDICAII .07 .07 .11 . 01 .01 .01 . 14 .00 .06 .OJ -.01 .IJ .OJ .05 .OJ .OJ . 10 -.01 .IJ -.11 •.OS . OJ . 01 -.07 -.01 1.00 NEDICAID . 04 . OJ -.15 .01 .OS -.01 "01 . 07 -.01 • • OJ .IJ .01. .01 -.06 .01 .06 .OS . 01 -.01 ,06 .04 -.11 -.11 . 01 .OS .It 1 . 00 INSUR -.01 •• 01 -.60 • . OJ -.06 .. 00 .16 -.OJ .OS -.01 .00 -.10 .OJ -.01 .OJ -.06 .01 . 01 -.11 .09 .OS •.01 -.04 .OJ . 04 . 11 -.16 1.00 PrAY . 06 -.06 -.11 •.01 .07 -.01 -.04 -.01 .01 . 01 .14 -.01 . 01 .01 •.00 -.01 -.01 •.01 -.Ol . 01 .00 . 04 -.OS .04 .01 .11 . 01 -.OS 1 .00 ltiLOS • 74 .71 .01 -.OS • • OJ •• OJ -.06 .OJ .OS .OJ -.OJ -.00 .01 .DS ,14 .01 .06 . 07 -.IS .01 . U -.11 . 11 .IS -.07 .01 .01 . 01 .07 1.00 IIZB . 04 -.ot -.16 .01 -.14 -.01 .09 .01 • . OJ .01 .07 .01 -.01 .01 . 07 .01 -.or . 07 . 01 .01 .OS . 00 .01 -.Ol .01 . 10 .OS .It . 04 . 04 1 .00 DENS -.01 .00 .01 -.00 .01 . 06 .01 .04 . 00 -.OS -.OJ -.U . U . 06 .OJ .OJ. .OJ .04 .01 . U -.OJ -.04 . 04 .OJ • .00 . 04 .II -.01 -.01 -.07 •.10 1.00 AIJMISS IIITI!NS -.06 .OS .11 .01 . 11 . 05 • . 10 -.04 .01 -.06 •.09 .OJ .09 .OJ .OJ .02 .II .OS -.14 .01 .OJ . 06 • , 10 . Ot .01 .16 .04 -.14 •.01 . 04 -.71 . U 1.00 , J7 .st . 05 .06 . 01 -.00 . 01 .05 . 06 .OJ . 01 -.07 .01 .OJ .01 •.01 .01 . 05 .It . 09 .ot .JS . a6 .25 .II .04 .04 -.07 .01 -.It . 01 .01 -.07 t.OO w 0 .. ii •eorr10latl01t cooffldo•t• of .06 or Muor are alpl,.caftl at tt.e .OS lev" l for • saapl• she of 614 ohserntlon•. E IS tl I

PAGE 367

348 Massachusetts For the Massachusetts samp 1 e, the primary dependent vari ab 1 e, CPE, and the secondary dependent variable, TOTVTS, were positive1y corre1ated ( .97). This served to reinforce the appropriateness of focusing the discussion on the primary dependent variab1e, CPE; therefore, CPE is the principal dependent variable discussed in this section. Several correlations within the patient-specific variables are noteworthy. Patient's living situation, ALONE, was positively correlated with age of the patient ( .22) . This suggests that older patients tend to 1 ive alone more than other patients. Few of the diagnostic categories exhibited strong correlations with the other independent variab1es. Those worth noting included the fo11owing. First, the diagnosis of circulatory system disorders, CIRCU, was negatively associated with the presence of a surgica1 procedure prior to admission to home care, YNSURG (-.26). Second, longer 1ength of use, HHLOS, was associated with the diagnosis of blood disorders, BLOOD ( .20). Third, the diagnosis of neop1asms, NEOP, was positively associated with the discharge site vari ab 1 e, DSITE ( . 17), the patient's health status at discharge, PATSTAT (.22), and the variable indicating whether the patient died while receiving home care, DEAD ( .20). NEOP was positive1y associated with the variable indicating terminal illness, TERM (.46), and was negative1y corre1ated with the dichotomous variable measuring health status at discharge, BETWORSE (-.23). These findings indicate that the diagnosis of neoplasm was high1y corre1ated with deterioration and/or death during the patient's episode of i11ness.

PAGE 368

349 The patient's functional status index, ADADL, was positively associated with the patient's health status at discharge, PATSTAT (.31), but negatively correlated with the dichotomous health status variable, BETWORSE (-.29), and positively correlated with DEAD ( .22); the signs of the correlation coefficients indicated that the less dependent the patient was at admission to the home care pro gran, the better his health status at discharge. As was expected, those patients diagnosed as terminally ill, TERM, were likely to die during the episode, DEAD (.35). Other outcome variables strongly correlated included the change in ADL index, ADLCHG, with PATSTAT (.45) and BETWORSE (-.39); DSITE (discharge site) was strongly correlated with BETWORSE (-.70) and PATSTAT ( .67). In addition, as expected, PATSTAT and BETWORSE were highly correlated (-.85). Of the provider/health system characteristics, MEDICARE was positively associated with the patient's age at admission, AGE, (.53). Those episodes with commercial insurers or Blue Cross as the primary payor, INSUR, exhibited a negative correlation with patient's AGE (-.43). The population per square mile variable, DENS, was highly correlated with both agency size, SIZE, and acute care admissions per thousand population, ADMISS, indicating that a strong association existed between the density of an area's population, the size of the agency which that population could support, and the associated acute care admissions for the same population. Canonical correlation coefficients for all possible pairs of the three groups of variables suggest that patient-specific variables were more strongly correlated with outcome variables (R =

PAGE 369

350 .383) than health system variables (R = .383); health system variables exhibited the weakest relationship of all with the outcome variables (R = .116). All coefficients were significant at the .001 level. An encouraging finding in itself, this suggests that outcomes of care were more a function of patient than provider needs. Because of the wide dispersion of CPE in the Massachusetts sample, preliminary statistics and equations were estimated using both the entire sample and a subsample of cases within five standard deviations of the mean CPE. This was done to test the impact of this dispersion on overall R2, the regression coefficients, and the mean values for all variables. As indicated in Table E .5, the average cost per episode was substantially reduced when all cases five standard deviations from the mean CPE were eliminated from the sanple. The skewness about the mean CPE was also reduced, but the standard deviation was still greater than the mean CPE. In general, there was little impact on the mean values of the independent vari-ab 1 es when the samp 1 e was reduced: ( 1) the aver age age. of the patients in the sample was reduced by .05 years; (2) the percentage of patients with digestive disorders as primary diagnosis was reduced by .01%; and (3) the admitting AOL Index, ADADL, was reduced by .02 units. In all other cases, the patient-specific variables retained the same mean values they had when all cases were included in the sample. The only outcome-related variable whose mean was affected by the reduction in samp 1 e size was BETWORSE, whose mean was increased by .01% in the smaller sample. The only provider/ health system variable whose mean was affected by the sample reduction was SIZE, which was reduced, on the average, by 5.88 units.

PAGE 370

I I I I 351 TABLE E.5 MEANS AND STANDARD DEVIATIONS OF SELECTED. VARIABLES FOR THE SAMPLE OF MASSACHUSETTS CASES_ WITHIN FIVE OF THE MEAN CPE ( N=2078) Variable Mean S.D. CPE $226.39 310.58 AGE 68.85 17.44 ALONE .28 .45 NEOP . 14 .34 EN DOC .07 .26 BLOOD .02 . 14 MENTAL .02 . 1 4 NERVS . l 0 .30 CIRCU .24 .42 RESP .04 . 19 DIGEST . l 0 . 31 GENITO .03 . 16 MUSCO .11 .31 ACCID .08 .27 ADADL l. 73 2.20 YNSURG .36 .48 ADLCHG -.19 l. 51 BETWORSE .77 .42 DEAD .04 . 19 MEDICARE .64 .48 MEDICAID .11 .32 INSUR . 12 .32 SIZE 8907.90 2061 . 13 ADM ISS l 0.17 l. 93 Source: Massachusetts Discharge Summary System, Medicare Cost Reports, and AHA Guide.

PAGE 371

352 In general, this analysis suggested that the mean values of the independent variables were not significantly affected by the elimination of episodes whose CPE exceeded plus or minus five standard deviations of the dependent variable, CPE. This suggested that the original, larger sample was appropriate for the regression analyses and that the mean values of the independent variables did not change substantially between the two Massachusetts samples; therefore, the observed relationships were not expected to be significantly different in the two samples. Preliminary regression analyses were completed utilizing the smaller sample of Massachusetts cases in order to test the impact of this reduction in sample size on the estimated R2 and regression coefficients. The estimated regression equations which were the result of this analysis were not substantially different from equations estimated using the entire Massachusetts sample. In light of these findings, all final regression equations estimated for the Massachusetts sample utilized the larger sample of all cases. The findings of the correlation analyses provided i nformation upon which to base preliminary formulations and subsequent refinements which produced the final regression equations. When variables exhibited strong collinearity, different formulations of the equations were estimated using other combinations of variables. Only the final equations were reported in this section. Philadelphia The correlation matrix included in Table E.4 for the Phila delphia sample exhibited both similarities and differences when

PAGE 372

353 compared to the Massachusetts matrix. As in the Massachusetts sample, both the primary and secondary dependent variables were highly correlated ( .97). The major patient-specific variable with strong correlation coefficients, in relationship to the dependent variables, was ADADL, indicating a strong correlation between the patient's functional status at admission and the dependent variable. This was similar to the finding for Massachusetts. ADADL was also highly correlated with several of the outcome-related variables. It was positively related to PATSTAT ( .34) and DSITE ( .30) and negatively correlated with . BETWORSE (-.31). The direction of these findings was consistent with the conclusion that the better the functional status of the patient at admission, the better the patient status at discharge. A similar relationship existed between the change in functional status, ADLCHG, and the admitting ADL Index, ADADL (-.43), serving to reinforce this contention. The only two diagnostic variables which exhibited strong correlation with other independent variables were circulatory system disorders, CIRCU, and neoplasms, NEOP. Consistent with findings in the Massachusetts sample, the variable NEOP was positively correlated with DEAD (.26), DSITE (.24), and PATSTAT ( .27), but negatively correlated with BETWORSE (-.26). These relationships indicated that, similar to the Massachusetts sample, patients with the diagnosis of neoplasms generally deteriorated during their use of home care services. Consistent with the Massachusetts correlation findings, but more pronounced, MEDICARE was highly correlated with the age of the

PAGE 373

354 patient, AGE {.72), while MEDICAID was negatively associated with AGE (-.25). The variable INSUR was strongly, negatively correlated with the variable MEDICARE (-.88), indicating that almost all cases in the Philadelphia sample could be characterized by either of the two payor sources as the primary payor. INSUR was negatively correlated with AGE (-.60), which was also similar to findings in the Massachusetts matrix. In general, the length of use, HHLOS, was negatively correlated with the outcome variables which were measured in such a way as to indi .cate that the longer the length of use, the better the outcome for Philadelphia episodes. This was not the relationship found among similar variables in the Massachusetts matrix. Similar to the Massachusetts findings, several outcome-related variables were strongly correlated with one another, e.g., ADLCHG with PATSTAT ( .47), ADLCHG with BETWORSE {-.42), and PATSTAT with DSITE ( .82). These correlation analyses substantiated findings of the descriptive analyses which indicated a stronger positive relationship between cost and outcome of care in the Philadelphia than in the Massachusetts sample. Consistent with the Massachusetts canonical carrel ation analysis, but stronger, was the relationship in Philadelphia of patientspecific variables to the outcome variables (R = .603), and to the health systen variables (R = .543). Health systen variables exhibited the weakest correlation with outcome variables (R = .111). All coefficients were significant at the .001 level.

PAGE 374

355 Regressions on Cost Using Patient-Specific Variables Approach and Sample The purpose of the regression analysis described in this section was to test the impact of patient characteristics on the cost per episode of care. Si nee much of the health care 1 iterature substantiates the contention that patient characteristics are the major determinants of cost and utilization, this regression equation was estimated in-order to determine the nature and extent of this relationship. The sample for the regression. analysis included all cases with complete data in both the Massachusetts and Philadelphia samples. The source of data for each of the variables was discussed in Chap ter III and detailed in Appendix C. Independent Variables The independent variables used in this analyses were those patient characteristics at admission which were hypothesized to account for differences in the cost per episode in care. Patientspecific variables were included in various formulations of the regression equations for the two different primary samples, Massachusetts and Philadelphia. The equations presentedin this section include the group Qf patient-specific independent variables which were common to both data sets. The independent variables used in this regression analysis were all included in the patient-specific category of variables. The specific factors chosen were the following:

PAGE 375

356 AGE. This was the age at admission, measured in years, of each patient whose episode of care was the unit of analysis. This vari-able was included as a potential indicator of demand, under the hypothesis that the older the patient, the greater the potential need and demand for home health services. ALONE. This was a 0/1 categorical variable used to indicate whether or not the patient 1 ived alone or with others prior to and during treatment by the home care program (1 = alone; 0 = not alone). The hypothesized effect of this variable on the cost per episode was positive. PRIMARY DIAGNOSIS. The major diagnostic categories pertaining to primary diagnosis were included as 0/1 categorical variables and used to test the impact of medical diagnoses on the cost of care. No hypotheses regarding the re 1 at ive impacts of various diagnoses were made; the only hypothesis was that cost per episode was expec-ted to differ by diagnostic category. The variables used were:3 NEOP: Neoplasms 0/1 ENDOC: Endocrine Disorders 0/1 BLOOD: Blood Disorders 0/1 MENTAL: Mental Disorders 0/1 NERVS: Nervous System Disorders 0/1 CI RCU: Circulatory System Disorders 0/1 3Preliminary equations were estimated diagnostic variable. Because of the small and two other categories, final equations variables INFECT (infections) and SKIN omitted. using a 11 but the OTHER n lJTlber of cases in this were estimated with the (skin disorders) also

PAGE 376

357 RESP: Respiratory System Disorders 0/1 DIGEST: Digestive System Disorders 0/1 GENITO: Genitourinary System Disorders 0/1 MUSCO: Muscul oske 1 eta 1 System Disorders 0/1 ACCID Accidents 0/1 ADADL. This was the value of the ADL index score at admission for each patient whose episode of care was the unit of analyses. This variable was included as another potential indicator of demand, under the hypothesis that the greater the functional disability, the greater the potential need and demand for home health services. YNSURG. This was a 0/1 categorical variable used to indicate whether or not a surgical procedure was noted in the case records at time of admission (1 = surgical procedure; 0 = no surgical proced ure). It was hypothesized that the effect of this variable on cost per episode was positive. The source of referral and the goal at admission variables were omitted from the final regression equations presented here for two reasons: (1) preliminary analyses which included these variables for Massachusetts cases in an R2 which was minimally higher than that in final equations estimated with the variables omitted (less than a difference of .001); and (2) these variables were not available for Philadelphia cases. Hence, their inclusion in the final analyses would result in regression functions which were different from those of the Philadelphia sample.

PAGE 377

358 Regression Findings Table E.6 presents regression results for the final estimated equations for both the Massachusetts and Philadelphia samples. When including only the patient-specific variables, the overall regression equations were both significant (at the .001 level), although the R2 was quite low for both For Massachusetts cases the R2 equalled .06, indicating that only 6% of the variation in CPE was accounted for by the patient-specific variables. This value was slightly higher in Philadelphia with an R2 of .11. Consistent with the findings of the descriptive analyses, several variables included in the estimated equations were significant. The AGE regression coefficient was positive and significant in both the Massachusetts and Philadelphia equations. The relative impact of AGE on CPE was also consistent with earlier findings since the regression coefficients were relatively small in both samples (1.17 and 2.10 for Massachusetts and Philadelphia, respectively). In the Massachusetts sample, ALONE was highly significant (p = .001), and its regression coefficient was large (112.16) and positively related to the dependent variable. This was consistent with the descriptive findings which indicated that the patient's living condition was strongly related to cost per episode. Again, consistent with the descriptive findings, this variable was not significant in the regression equations estimated for the Philadelphia sample. Step wise analyses supported these findings in both cases. The only diagnostic category with a significant coefficient in the Massachusetts sample was BLOOD, while for the Philadelphia sample coefficients in three diagnosti c categories (GEN ITO, MUSCO

PAGE 378

Independent Variables Regression Coefficient AGE 1.17 ALONE 112.16 NEOP -56.13 EN DOC 31.81 BLOOD 183.11 MENTAL 62.76 NERVS 46.37 CIRCU 49.73 RESP DIGEST -26.33 GENITO -54.28 MUSCO 12.54 ACCID -79.44 ADADL 44.57 YNSURG 52.52 Constant 43.11 TABLE E.6 FIXED REGRESSION RESULTS ON CPE AND THE PATIENT-SPECIFIC VARIABLES Massachusetts Significance .045 .000 .242 .553 .017 .412 .363 .280 .757 .599 . 451 .802 .134 .000 .017 R 2 = .062 N = 2101 Correl. w/ Dep. Var. .071 .075 . 031 -.013 .044 .004 .035 .070 -.024 -.053 -.017 .017 -.038 .177 -.001 F = 9.22 (ps.OOl) Regression Coefficient 2.10 92.18 -51.41 94.75 28.51 35.16 -.4684 26.70 39.54 -23.46 1703.58 189.71 221.96 34.44 148.54 212.60 Philadelphia Significance .052 .129 .334 .204 .895 .882 .997 .604 .629 .769 .000 .095 .045 .000 .000 R 2 = .1 05 N = 677 Cerrel. w/ Oep. Var. .078 .029 -.031 .036 -.008 .020 -.016 .061 .011 -.006 . 166 .058 .093 .192 .146 F = 5.17 Source: Massachusetts Discharge Summary System, Medicare Cost Reports, Blue Cross Home Care Program Data System, and Philadelphia Abstracted Patient Records. w U'1 1.0

PAGE 379

360 and ACCID) were significant. The significance of the GENITO coefficient in the Philadelphia sample was partially the result of the extremely small prevalence of this diagnosis. The significance of the other two diagnostic categories was consistent with earlier descriptive findings. Those patients with MUSCO and ACCID diagnoses in the Philadelphia sample were generally those patients for whom rehabilitation was the goal at admission. Consequently, they generally utilized a highly intense therapeutic service mix. ADADL had a significant and positive regression coefficient for both samples. The pattern was consistent when the equations were estimated using a stepwise procedure, since this variable was the first to enter for both siJTlp 1 es. When these regressions were run in preliminary analyses using the components of ADL, rather than the ADL index (ADADL), the only significant regression coefficient was for the variable which meas ured independence at admission on bathing. The R2 of the equations were affected very little. The R2 for the Philadelphia sample increased from .105 to .110, while in Massachusetts it increased from .062 to .065. All equations were significant at the .000 level. The variables with significant regression coefficients remained essentially unchanged in the new formulations, except that in the Massachusetts sample, ACCID became significant, as did the variable which measured independence at admission for bathing. In the Philadelphia equations, when ADADL was omitted, none of the components of the ADLs beciJTle significant. It appears that when each of the components was included in the analyses separately, their

PAGE 380

361 impact was insignificant on the dependent variable, but when aggre gated to their index level of measurement, their overall impact was significant. The final patient-specific variable with a significant coefficient was YNSURG in both samples. The size of this coefficient was significantly higher in Philadelphia than in Massachusetts, as was its correlation coefficient with the dependent variable. These findings were consistent with descriptive analyses contained in Chapter VI. In conclusion, when the impact of the patient-specific variables on cost per episode was estimated, a pattern consistent with the earlier findings emerged: age, living arrangement, admitting AOL, and the presence of a surgical procedure were important variables in explaining variation in cost per episode. The impact of the specific diagnostic categories was also consistent with earlier findings, since several key diagnoses identtfied in descriptive findings appeared with significant regression coefficients (BLOOD, MUSCO and ACCIO). Regressions on Cost Using Provider/Health System Variables Approach and Sample The approach and sample used in these regression equations was similar to those described in the preceding section. The purpose of these equations was to estimate the relative impact of the provider/ health system characteristics on CPE when no other variables were taken into account. It was expected that within each sample, these

PAGE 381

362 variables did not account for a substantial portion of the observed variation. Dependent and Independent Variables . major dependent variable in t l 1ese equations was, as previously indicated, CPE. The independent variables used in these analyses were selected to include those factors which were hypothe sized to account for differences in CPE. Various combinations of each of the provider/health system variables were included in preliminary formulations of the regression equations. Based upon the findings of the correlation and preliminary regression analyses, final equations were estimated utilizing five provider/health system v ari ab 1 es. Several of these variables were included in order to reflect potential variation in both the nunber of referrals to home health agencies and in possible competition for patients among agencies and other providers. In addition, other variables reflect the hypothe sized relationships between the source of payment for home care and CPE. The following are the specific variables selected. MEDICARE. This was a 0/1 dichotomous variable indicating whether or not the payment source for a particular episode was Medicare. Since Medicare was the primary payor for much home health care, it was hypothesized that this variable would have a direct impact on the variation in CPE. Since only Medicare reimbursable types of home care were included in the calculation of CPE, it was hypothesized that this variable was positively related to the depen dent variable.

PAGE 382

363 MEDICA! 0. A similar vari ab 1 e to the preceding one, this was intended to determine the impact of Medicaid as the primary payor source on CPE. It was hypothesized to have a positive relationship with CPE. INSUR. The final variable measuring payment source was included in order to estimate the impact of private insurers and Blue Cross on CPE. This was also a 0/1 categorical variable. Prelimi nary analysis suggested a negative directional effect of this variable on CPE. SIZE. This variable was measured as the number of skilled nursing care visits provided to patients either directly or through contract during 1976. No specific hypothesis regarding the directional impact of this variable was tested. ADMISS. This was the nunber of short-term hospital admissions per thousand population in 1976 for the county in which the provid er• s main office was located. Because most patients were referred to home care after an acute episode, this measure was included as a primary measure of potential referral. A negative relationship was hypothesized. Regression Findings In estimating the regression equations using only provider/ health system characteristics, the findings (shown in Tables E.7 and E.8), were again consistent with descriptive analyses reported earlier. The overall R2 for the Massachusetts sample was significant at the .000 level, although its magnitude was small (R2 = .015).

PAGE 383

Independent Variables MEDICARE MEDICAID SIZE ADMISS Constant FIXED AND STEPWISE REGRESSION RESULTS ON CPE VARIABLES AND THE PROVIDER/HEALTH SYSTEM FOR MASSACHUSETTS SAMPLE Fixed Equation Regression Coefficient 114.10 105.93 -.0031 8.49 115.50 R2 = .015 N = 2101 Signifi-cance .000 .003 .507 .097 F = 7 • 7 8 ( • 001 } Regression Coefficient 114.00 106.29 8.56 86.88 Stepwise Equation Signifi-cance .000 .002 .094 R 2 = . 014 N = 2101 Entering Step (1) (2} (3} F = 1 0. 23 ( ps: • 001 } Source: Massachusetts Discharge Summary System, Medicare Cost Reports, and AHA Guide. Correl. w/ Dep. Var. .092 .016 -.015 .054

PAGE 384

Independent Variables MEDICARE MEDICAID SIZE ADM ISS Constant FIXED AND STEPWISE REGRESSION RESULTS ON CPE AND THE PROVIDER/HEALTH SYSTEM VARIABLES FOR PHILADELPHIA SAMPLE Fixed Equation Regression Signifi-Coefficient cance 67.69 119:45 -.0912 -7.50 728.26 R 2 = . 015 N = 677 .068 .146 .069 .460 F = 2.59 (p= .035) Regression Coefficient 67.46 119.31 -.0623 617.48 Stepwise Equation Significance .069 . 146 .046 R 2 = .014 N = 677 Entering Step (2) (3) ( 1 ) F = 3.28 (p=.021) Carrel w/ Dep. Var. .073 .033 -.090 .054 Sourc e: Blue Cross Home Care Program Data System, Philadelphia Abstracted Patient Records, and AHA Guide. w 0\ Ol

PAGE 385

I . I 366 The size of the R2 increased slightly when the sample was reduced to only those cases within five standard deviations of the mean (R2 = .021). The overall R2 of the Philadelphia sample was identical to the main Massachusetts equation (R2 = .015) and significant at the .035 level. These findings indicated that little of the variation in CPE (less than 2%) was explained by the provider/health system characteristics when no other factors were taken into consideration. A key variable in both samples was MEDICARE. It was also sig-nificantly different from the other payor sources, and its coefficient was positive in preliminary equations for both samples. Although its sign and the magnitude of the regression coefficients were essentially the same in both samples, MEDICAID was significant-ly different from other payor sources ..!!.!1. in the preliminary Massachusetts equations. Other differences between the equations for the two samples occured in the case of the SIZE and ADMISS variables. The SIZE variable was significantly related to CPE in Philadelphia, but was not significant for the Massachusetts sample. When the Massachusetts equations were estimated using the sample within five standard deviations of the mean, SIZE became significant. In the Massachusetts equations, the ADMISS variable was significantly related to CPE, while in the Philadelphia equations the opposite was true (p = .460). This indicated that acute care admissions served as a source of referral in Massachusetts, but not so in Phil adelphi a. This was not surprising since acute care admissions, which provided the primary source of home care admissions in Philadelphia, cane

PAGE 386

367 from each hospital to its home care program. Admissions to other hospitals were not relevant to the Philadelphia programs. These regression findings were consistent with earlier descriptive analyses which indicated that MEDICARE and MEDICAID were significantly different from other payors in their impact on CPE in Philadelphia. This could be partially explained by: (1) the low per dien actually paid to hospital-based programs in Philadelphia which .may have prompted them to reduce the average length of use of Medicaid patients to something more similar to the average case and hence eliminate the impact of Medicaid on CPE as observed in the Massachusetts sample; and (2) the small mrnber of Medicaid episodes in Philadelphia which may have impacted its statistical significance in the equations. In terms of public policy, the first explanation is of more "real-world" significance. Regressions on Cost Using Outcome-Related Variables Approach and Samp 1 e The purpose of these regression analyses was to determine whether outcome characteristics, when considered alone, were systematically related to the dependent variable, CPE. The sample for these regressions was the same as that in the preceding sections. The approach used was also similar to those preceding, since all equations were estimated using fixed and stepwise procedures. Dependent and Independent Variables As noted previously, CPE was the primary dependent variable used for these regressions. Three of the potential

PAGE 387

368 variables were used in the final regression equations reported here. These variables were selected after completion of the correlation analyses and preliminary regression formulations utilizing various combinations of outcome-related variables. The specific variables used in this analysis include the following. ADLCHG. This key outcome variable was measured by subtracting the AOL index score at admission frcxn the AOL index score at discharge for each patient's episode of care. Hence a negative value indicated greater independence at discharge. It was hypothesi zed that the greater the ' improvement in functional status, the higher the cost per episode; hence, the coefficient was hypothesized to be negative. BETWORSE. This was a 0/1 categorical variable used to indicate whether or not a patient's general health status at discharge was improved or ( 1 = improved or sane; 0 = worsened or dead) . This measure was the other key outcome variable used in the regression analyses and was hypothesi zed to have a positive effect on the dependent variable. DEAD: This was also a 0/1 categorical variable used to indi-cate those patients who died at home during the course of their treatment. This measure was included to test the impact of the treatment of patients who died at home, versus those who did not, on the dependent variable. After preliminary analysis using descriptive statistics, it was hypothesized that there would be a positive relationship between this variable and CPE.

PAGE 388

369 Regression Findings The purpose of estimating regression equations using only the outcome-re 1 a ted v ar i ab 1 es was to test their impact upon CPE i rrespective of the other variable categories. The findings illustrated in Table E.9 suggest that, although the equations for both Massachusetts and Philadelphia were significant at the .001 level, the multiple regression coefficients in both cases were extremely small (i.e., the Massachusetts R2 equalled .056, while for Philadelphia the R2 was .038). This indicated that, by themselves, the outcomerelated variables explained little of the variation in cost per epi sode. Each of the variables included in these equations was significant for both the . Massachusetts and Philadelphia samples. ADLCHG was highly significant (p = .001) with regression coefficients of relatively the same magnitude in both samples. The negative sign on the ADLCHG variable indicated that improvement in functional status was positively associated with the cost per episode in both samples. This finding served to reinforce earlier descriptive findings for Philadelphia, and strengthened the argument that, indeed, a similar, although weaker, relationship existed in Massachusetts. The coefficients of the BETWORSE variable in the two samples were different in magnitude, direction, and significance. BETWORSE was significant at the .001 level for Massachusetts cases, but insignificant in Phila delphia. In Massachusetts, BETWORSE was negatively associated with CPE (-236.31); in Philadelphia, it was positively associated with

PAGE 389

Independent Variables Regression Coefficient ADLCHG -42. 17 BET WORSE -236.31 DEAD 151.82 Constant 425.00 FIXED REGRESSION RESULTS ON CPE AND THE OUTCOME-RELATED VARIABLES Significance .000 .000 .004 R 2 = .056 . N = 2101 F = 41.43 Carrel. w/ Regression Dep. Var. Coefficient -.038 -38.20 -.193 42.74 . 107 190.71 443.59 Philadelphia Significance .000 .364 .022 R 2 = .038 N = 677 Carrel. w/ Dep. Var. -.174 .077 . 031 F = 8.82 Source: Massachusetts Discharge System, Medicare Cost Reports, Blue Cross Home Care Program Data System, and Philadelphia Abstract Patient Records. w ""-.,! 0

PAGE 390

371 CPE and of much smaller magnitude (42.74) .4 The size of regression coefficients of the DEAD variable in both samples suggested that patients who died at home had a strong positive and significant impact on CPE. Stepwise formulations of the same equations resulted in consis-tent findings. ADLCHG entered on the first step in both samples and DEAD on the second. BETWORSE was exc 1 uded in the Phi ade 1 phi a equa-tions, but was included and significant in Massachusetts. This finding was consistent with earlier descriptive analyses which inves-tigated the relationship of the health status at discharge variable to CPE. In conclusion, change in functional status, ADLCHG, and the death of the patient at home, DEAD, were key variables in both sam-ples, although outcome-related characteristics taken together seemed to explain little of the variation in CPE when no other factors were considered. Regressions on Cost Using Patient-Specific and Provider/Health System Variables Approach and Sample The purpose of these regression an a 1 yses was to determine the incremental impact on CPE of the t\'tO categories of independent 4The reliability of this variable seemed questionable, since various formulations of measurement using the PATSTAT variable were tested and none appeared completely satisfactory. Therefore, interpretation of findings regarding BETWORSE were tentative at best.

PAGE 391

372 variables, patient-specific and provider/health system character-istics, when considered together. Since each of th . e equations had been estimated using each of the categories of variables alone, the step, when followed by regressions using all three categories of the variables together, provided intermediate information about the relative impact of the variable groups. The sanples were the same as those used for the preceding analyses. Dependent and Independent Variables The dependent v ari ab 1 e of interest was CPE. The independent variables were a combination of all variables used in equations described earlier in this Appendix. The independent variables included the following, all of which have been defined earlier: Patient-Specific Variables: Provider/Health System Variables: AGE ALONE NEOP EN DOC BLOOD MENTAL NERVS CIRCU RESP DIGEST GENITO MUSCO ACCID ADADL YNSURG MEDICARE MEDICAID SIZE ADMISS

PAGE 392

373 Regression Findings In order to simp 1 ify this discussion, Massachusetts findings are presented first, followed by those for Phil adelphi a. Differences between the two samplse are described in the last portion of this section. Table E.lO presents the findings of the final Massachusetts equations. The overall R2 of the equation { .069} was higher than those for either of the equations estimated with only one category of independent variable, but not as l .arge as it would have been if the R2 for each equation were simply added together. For example, the R2 for the patient-specific equation was .062 and the R2 for the provider/health system equation was .015 which, if added together, equalled .077. This was significantly greater than the R2 of tne combined equation which was .069. This was due partially to the correlation of variables across the two groups. When both patient and provider variables were combined in these equations, two previously significant variables, AGE and ADMISS, dropped out. The coefficient of the AGE variable was no longer significant because of its collinearity with the variable MEDICARE, which retained its significant coefficient in the new formulation. ADMISS was significant at the .094 level in the separate equations. Partly due to correlations between this variable and others in the added category, in addition to its relatively low significance level, the significance of its coefficient was decreased to .113 when it was included with the patient-specific variables. All other variables retained the relative size and significance of their coefficients in these formulations for Massachusetts.

PAGE 393

Independent Variable AGE AlONE NEOP EN DOC BlOOD MEIITAl NERVS CIRCU RESP DIGEST GENITO MUSCO ACCID ADADL YNSURG MEDICARE MEDICAID SIZE Constant TABLE E.IO FIXED AND STEPWISE REGRESSION RESULTS ON CPE AND TilE PATIEIIT-SPECIFIC AND PROVIDER/IIEALTII SYSTEM VARIABLES FOR MASSACHUSETTS SAMPLE Fixed Equation Regression Coefficient .3148 105.43 -53.73 36.78 197.60 68.54 41.45 52.07 -21.28 -II. 23 -50.07 15.30 -80.89 43.70 48.56 74.20 77.61 .0057 8.07 15.11 -R 2 = .069 N = 2101 Significance .619 .000 .264 .492 .010 .372 .416 .258 .736 .823 .486 .760 .127 .000 .029 .006 .025 .219 .113 F = 8.08 (ps.001) -Stepwise Equation Regression SignHiEntering Coefficient cance Step ---107.72 .000 58.14 .053 (6 193.43 .003 ( 4) 36.65 .288 13) 47.59 .058 10) -86.44 .021 43.47 .000 43.58 .040 78.49 .001 3) 77.01 .025 (lq -.0061 .192 8 .46 .094 (9 -41.71 ------------------n2 = .067 N = 2101 F = 12.53 (ps.OOl) ------------------------------------Correl. w/ Dcp. Var. .071 .075 .03 1 -.013 .044 .004 .035 .070 .024 -.053 -.017 .017 -.038 .177 -.001 .092 .016 -.015 .054 Source: Blue Cross Home Care Program Data System, Philadelphia Abstracted Patient Records and AliA Guide.

PAGE 394

I I I I I I I I I 375 When the two categories of independent variables were combined for the Philadelphia sample, the impact of the interaction on CPE of the two categories was similar to Massachusetts, as shown in Table E.ll. The overall R2 of the equation was greater than either of those estimated separately, but not as large as it would have been if the mu 1 tip 1 e regression coefficients were simp 1 y added together (R2 = .111); yet the effect on the significance of particular coefficients was more dramatic in Philadelphia than in Massachu-setts. Four variables which were significant in earlier formulations no longer had significant coefficients. Similar to Massachusetts, AGE dropped out of this equation because of its collinearity with MEDICARE. In contrast to the Massachusetts equation, MEDICARE also dropped out of the combined equation in the Phi 1 adelphi a analyses. MUSCO had been significant at the .095 level, but due to its correlation, albeit low, with the provider variables, it no longer exhibited a significant coefficient. SIZE was no longer significant when included in these combined formulations. In summary, when patient and provider characteristics were com-bined in the equations regressed on CPE for the Philadelphia sample, none of the provider/health system characteristics retained their significant correlation coefficients. Consequently, the main difference between the two samples was that for Massachusetts episodes, provider/health system variables (e.g., the sources of payment, MEDICARE and MEDICAID) retained their significance when included with patient-specific variables, while in Philadelphia they did not.

PAGE 395

Independent Variable AGE ALONE NEOP ENOOC BLOOD MENTAL NERVS CIRCU RESP DIGEST GENITO KISCO ACCIO ADADL YNSURG MEDICARE MEDICAID SIZE ADM ISS Constant TABLE E.11 FIXED AND STEPWISE REGRESSION RESULTS ON CPE AND THE PATIENT-SPECIFIC ANO PROVJOER/IIEALTII SYSTEM VARIABLES FOR PHILAOELPIIIA SAI-IPLE Fixed Equation Regress ton Coefficient 1.88 94.38 -39.21 98.30 30.97 43.22 3.09 33.11 32.85 -11.96 1660.39 186.91 212.25 35.48 145.91 6.86 87.39 .0398 3.39 255.37 R 2 " . 111 N " 677 Slgn1f1cance .216 .120 .469 .190 .886 .783 .981 .521 .688 .882 .000 . 101 .055 .000 .000 .89 1 .275 .418 .736 F = 4.32 (ps.OOl) --Regression Coefficient 2.09 92.99 -58.73 76.71 1641.75 166.44 195.28 34.72 137.39 85.46 -.0532 248.91 Stepwise Equation Sign 1ftEntering cance .056 .122 .153 .239 .00 0 .122 .064 .000 .000 .273 .077 R 2 = .110 N = 677 F = 7.46 (ps.001) Step 1:! (5 (10 g H (11l (4 Source: Blue Cross llome Care Prog.ram Data System, Ph1lade1phla Abstracted Patient Records and AliA Guide. Correl. w/ Oep. Var. .078 .029 -.031 .036 -.008 .020 -.016 -.061 .011 -.006 .166 .058 .093 .192 .146 .073 . -33 -.09 0 .054

PAGE 396

I I 377 Regressions on Cost Using All Variables Approach and Sample The purpose of the final regression analyses for the entire was to examine variations in CPE when all three categories of variables were included. In so doing, the relative impact of each of the variable categories and its components could be determined, taking into account other characteristics. This approach first estimated equations utilizing each of the independent variable categories alone, then combined patient-specific and provider/health system variables together to estimate their impact on CPE, and finally utilized all three groups of independent variables together. All equations were estimated using both fixed and stepwise analyses. The relative impact of each of the independent variable categories on CPE could then be evaluated using this approach. Dependent and Independent Variables The dependent variable used in the analyses, as in earlier ones, was the cost per episode of illness of home health care, CPE. The independent variables were selected from those used in the preceding analyses. Zero-order and canonical correlation analysis and preliminary regression findings resulted in the final equations discussed in this section. The independent variables included the following: Patient-Specific Variables: AGE ALONE NEOP EN DOC BLOOD

PAGE 397

I . I I I I Provider/Health System Variables: Outcome-Related Variables: Regression Findings MENTAL NERVS CIRCU RESP DIGEST GENITO MUSCO ACCID ADADL YNSURG MEDICARE MEDICAID SIZE ADMISS ADLCHG BETWORSE DEAD 378 To ease the following discussion which details the relative impact of the independent variables on CPE, the Massachusetts findings are presented first; those of Philadelphia then follow. Differences between the two samples are discussed in the last portion of this section. When estimating regression equations for CPE using all three groups of variables, the equations (shown in Tables E.l2 _and E.l3) were significant at the .000 level for both Massachusetts and Phila delphia samples. In Massachusetts, the overall ability of the equation to explain variation in CPE was slightly lower (R2 = .099) than in Philadelphia (R2 = .129). For the Massachusetts sample, both the stepwise equations using the entire sample and the equations whose sample was limited to those cases within five standard deviations of the mean, resulted in a slightly higher R2 (.102).

PAGE 398

Independent Variable AGE ALONE HEOP EN DOC BLOOD MENTAL HERVS CIRCU RESP DIGEST GENITO 14JSCO ACCID ADADL YNSURG ADLCIIG BETWORSE DEAD M EDICARE M EDICAID SIZE ADM I S S Constant --TABLE E. 12 FIXED AND STEPWISE REGRESSION RESULTS ON CPE AND All GROUPS Of VARIABLES FOR HASSACIIUSETTS SAMPLE fixed Equation Regression Slgnlfl-Coefficient cance .0205 103.88 -104.20 30.75 171.90 89.30 55.24 40.41 53.07 .7095 65.05 20.14 -66.89 26.13 55.29 25.70 -206.19 140.14 66.04 65.95 -.0028 9 .66 187.14 R2= .099 H "' 2101 .976 .000 .030 .559 .023 .238 .271 .372 .394 .989 .359 .682 .201 ,000 .012 .001 .000 .009 .013 .053 .548 .055 Stepwise Equation R e gression SlgnHI-Coefficient cance 103.78 .000 140.15 .000 135. 19 .037 -B9.96 .069 -3 7.02 .244 -100.93 .091 -100.93 .005 2 6 .03 .000 52.72 .010 -25.24 .001 -206.75 .000 138.88 .009 65.87 . 004 65.68 .053 10.26 .040 192.58 R 2 " .098 N = 2101 Entering SJ*ep (3l (4 ( 10) (8) j!l (1:1 (12) (1) f = 10.44 f " 15.15 ( flS .001) Sour ce : Massac husetts D is c h arge S u u wuary S y stem , Medi care Cost R eports a n d AHA Guide . Carrel. w/ Oep. Var. .071 .077 .031 -.013 .044 .004 .035 .070 .024 -.053 -.017 .017 -.038 .177 -.001 -.038 -.193 .107 .092 .016 -.015 .054 w "'-.1 1.0

PAGE 399

Independent Variable AGE ALONE NEOP ENOOC BLOOD MENTAL NERVS CIRCU RESP DIGEST GENITO CO ACCIO ADADL YNSURG AOCIIG BETWORSE DEAD MEDICARE •EDICAID SIZE ADM ISS Constant TABLE E. 13 FIXED AND STEPWISE REGRESSION RESULTS ON CPE AND ALL GROUPS OF VARIABLES FOR PHILADELPHIA SAMPLE Ftxed Equation Regression Signifl-Coefficient cance 1.68 83.38 -9.24 106.86 50.43 67.81 6.16 41.20 61.55 -15.77 1795.08 179.96 189.82 38.91 132.82 -3.0 6 158.61 152.54 20.40 80.98 -.0456 .8330 162.14 Rl = . 1 ?9 N = 677 .265 .167 .868 .151 .815 .665 .962 .421 .451 .843 .000 .112 .085 .000 .000 .792 .004 .067 .683 .308 .350 .934 F "4.41 Regression Coefft ctent 1.95 83.03 92.25 1781.90 163.91 174. 16 40.17 117.20 169.47 131.54 -.0538 2D7.45 Stepwise Equation Stgntftcance .065 .165 .148 .000 .123 .096 .000 .000 .000 .100 .070 R 2 = . 125 II = 677 Entering Step llJ ( 11) (2 ( 10 n (5) F = 8.62 Source: Blue Cross llome Care Program Data System, Philadelphi a Abstracted Patient Records and IIIlA Guide. Carrel w./ Dep. Var. .078 .029 .031 .036 .008 .020 .016 .061 .011 .006 .166 .058 .093 . 192 .146 .114 .077 .031 .073 .033 .090 .054 w (X) 0

PAGE 400

381 Massachusetts. Several of the patient-specific variables had significant regression coefficients in Massachusetts. Whether or not a patient lived ALONE was positively related to CPE and significant at the .000 level. The other significant patient-specific variable was ADADL (p = .001). These findings were consistent with earlier descriptive analyses and indicated that for Massachusetts providers, the admitting functional status of the patient and his living situation were important determinants of CPE. The impact of YNSURG was also significant (p = .012), indicating that the presence of a surgical diagnosis at admission accounted for some of the variation in CPE. Two diagnostic categories, NEOP and BLOOD, had significant regression coefficients (p = .030 and .023 respectively). The coefficient of NEOP was negatively related, while that of BLOOD is positively related to CPE. The magnitude of the regression coefficients for both variables was similar. These findings were similar to those of the regression equations estimating the impact of patient-specific variables alone and with provider/health system variables with two exceptions: (1) AGE was insignificant because of its strong correlation with the MEDICARE variable; and (2) NEOP was significant but with a change in the sign of its coefficient. When other factors were taken into consideration, three of the four provider/health systen characteristics included in the equations were significant. Both MEDICARE and MEDICAID exhibited positive and significant regression coefficients, indicating that the impact of these two payors was significantly different from all

PAGE 401

382 others on CPE. In Massachusetts, where it appeared that hospitals served as sources of referral for home care, ADMISS had a positive regression coefficient (p = .055). The renaining variable in the provider/health system category, SIZE, was insignificant when all cases were included in the analysis; when the equations were estimated on the basis of the subsample of cases within five standard deviations of the mean, SIZE changed to become negatively associated and significant in its relationship to CPE. These findings were genera 11 y consistent with those discussed earlier in Chapter V and with other equations reported earlier in this Appendix. All of the outcome-related patient variables included in the analyses were significant in the Massachusetts sample. ADLCHG was negatively related to CPE at the .001 level, indicating that an improvement in functional status was positively associated with CPE. The coefficient of BETWORSE, although significant at the .001 level, was negatively related to CPE, indicating that patients whose health status at discharge was 'ftOrse than at admission were more costly than others. This may be partially explained by the fact that patients who eventually died at home were coded zero for this variable, and those with no change were coded one. Based on findings of this and the ealier regression analyses, patient-specific variables accounted for slightly more of the variation in CPE than did the outcome-related variables. The other category of independent variables, provider/health systen characteristics, accounted for the smallest portion of variation in CPE . . For

PAGE 402

383 the Massachusetts sample, the impact of pati ent-speci fi c vari ab 1 es on CPE was concentrated in the variables categorizing diagnoses, surgical procedures and functional status at admission. Phi 1 ade 1 phi a. The equations estimated for the Philadelphia sample also exhibited patterns consistent with earlier findings. The impact on CPE of the patient•s living arrangement, ALONE, was not significant. The only diagnostic variables which were significant were ACCID (at the. 085 level) and GENITO (at better than the .001 level). Although the GENITO variable was significant, this was probably an artifact of the extremely small nt.mber of cases within this diagnostic category. Similar to the Massachusetts equations, ADADL was highly significant (p = .001) and positively associated with CPE, as was the variable YNSURG. Consistent with earlier discussions, the presence of a surgical procedure prior to admission had greater impact for Phi 1 adelphi a cases than for Massachusetts, whereas the impact of ADADL was relatively consistent in both samples. These findings were similar to early regressions estimating the impact of patient-specific variables. With regard to the outcome-related variables, ADLCHG was not significant in the Philadelphia sample; contrary to the Massachusetts findings, BETWORSE was positively related to CPE and significant at the .004 level. This relationship indicated that those patients who improved their health status at discharge were more costly than others. The patterns in the two samp 1 es were generally consistent. Those patients who improved in general health status or in functional status were usually more costly than those .

PAGE 403

384 who did not. This was consistent with the regressions using only outcome-related characteristics as independent variables, with the exception of the Philadelphia coefficients for BETWORSE and ADLCHG. This inconsistency may be partially due to the correlation between the t'.\() variables and other independent variables. Again, consistent across samples,. and with earlier descriptive analyses, was the positive, significant relationship of the DEAD variable to CPE. This finding indicated that patients who died at home were substantially more costly than those who did not. As expected, in 1 ight of the findings of equations reported earlier, when all the other factors were taken into consideration, none of the provider/health system characteri sties were significant in the Philadelphia sample. This finding suggested that for Phila delphia patients, the variation in CPE was accounted for almost entirely by patient-related variables, either directly related to medical/health status or to the outcome of care. In conclusion, patterns within the two samples appeared consistent throughout the staged regression analyses which were presented in this section. Those independent variables which were h ighly significant and contributed the most to the overall R2 of the equations tended to retain their relative importance throughout the analyses. Patient-specific variables accounted for the greatest portion of the variation in CPE in both samples, although the size ofthe overall R2 of the equations tended to retain their relative importance throughout. Patient-specific variables accounted for the greatest portion of the variation in CPE in both samples, although the size

PAGE 404

385 of the overall R2 in both cases was relatively small. In general, the small R2 for both samples, although highly significant, indicated that data available on characteristics typically thought to account for home health care utilization and subsequent cost (at the patient level), were inadequate to explain the variations observed. One important implication is that policy analysts who have traditionally relied on such information for the purposes of developing long-range estimates of home health care costs might reconsider the basis of these estimates in an effort to improve their utility to policymakers. Regression on Cost Using all Variables Rehabilitation Cases Only Approach and Sample The purpose of the regression equations presented here, and based on earlier .formulations, was to determine the relative impact on CPE of the three categories of independent variables for a subset of episodes which were described at time of admission to home care as having a goal of rehabilitation. The purpose of this analysis was to determine (1) if the independent variables of interest explained any more of the variation observed in CPE for the subset of rehabilitation cases than for the total sample, and (2) if pat-terns of significance were similar in the two sanples. Since data on this characteristic were not available for Philadelphia cases at the time of this study, the sample for this analysis included 267 episodes classified with the goal of rehabilitation from within the total Massachusetts sample only.

PAGE 405

386 The process used to estimate the final equation presented here was similar to that described in the preceding sections. In an effort to be parsimonious, only the final equation is discussed. When appropriate, comments regardi'ng preliminary formulations were included. Preliminary Statistics Means and standard deviations of the dependent and independent variables for all cases with the goal of rehabilitation at admission are listed in Table E.l4. It should be noted that the average cost per episode of this sample of patients was substantially higher than the average cost per episode of all cases in the Massachusetts sample. The age of the patients whose episodes were included in these analyses was slightly lower than the total overall average age of Massachusetts patients, and the percentage of patients living alone was also slightly lower. The diagnostic case mix shifted dramatically in this subsample of cases, clustering about the diagnoses circulatory system disorders, CIRCU, musculoskeletal system disorders, MUSCO, and accidents, ACCID. The average ADL index score at time of admission, ADADL, was significantly higher for these patients, indicating greater debilitation than for the average patient. Other important differences should be noted. A higher per-centage of patients in this subsample were admitted to home care with a surgical diagnosis. Average improvement in AOL index scores from time of admission to time of discharge, ADLCHG, was substantially higher for these patients (-.72 versus -.19 for the total

PAGE 406

387 TABLE E.l4 MEANS AND STANDARD DEVIATIONS OF SELECTED VARIABLES FOR ALL CASES WITH THE GOAL OF REHABILITATION FOR MASSACHUSETTS SAMPLE (N=267) Variable Mean S.D. CPE $381.41 522.71 AGE 68.54 17.88 ALONE .26 .44 NEOP .04 .21 EN DOC .05 .22 NERVS .04 .21 CIRCU .29 .46 RESP .06 .24 MUSCO .25 .44 ACCID . 15 .36 ADADL 2.47 2.32 YNSURG .42 .49 REFSOR2 .87 .33 ADLCHG -.72 1. 75 BETWORSE .79 .41 MEDICARE .67 .47 MEDICAID . 10 .30 INSUR . 14 .35 SIZE 8693.12 2532.44 ADM ISS 9.74 1.63 Source: Massachusetts Discharge Summary System, Medicare Cost Reports and AHA Guide.

PAGE 407

388 Massachusetts sample). The proportion of episodes which improved, according to the general health status at discharge variable, BETWORSE, was essentially the same as in the total Massachusetts population. Of the provider/health system variables, the average value associated with each in this category was approximately the same as in the total sample of Massachusetts cases. Table E.lS presents the zero order correlation matrix for the Massachusetts rehabi 1 it at ion cases. The coefficient of ALONE was positive, highly related to AGE ( .22) and negatively related to INSUR (-.22). The AGE of the patient at admission was negatively related to tY.O of the payor variables, INSUR (-.47) and MEDICAID(.38), but positively related to MEDICARE (.63). Few of the diagnostic variables were strongly correlated with any of the other inde pendent variables, except for the following: endocrine disorders, ENDOC, was positively related to the payor source, PPAY ( .30); mental disorders, MENTAL, was positively related to MEDICAID ( .20); nervous system disorders, NERVS, was negatively related to MEDICARE (-.21), but positively related to MEDICAID (.20); finally, circulatory disorders, CIRCU, was negatively related to the presence of a surgical procedure, YNSURG (-. 40) . The on 1 y other diagnostic category with a strong corre 1 ati on coefficient was muscul oske 1 eta 1 disorders, MUSCO, which was positively associated with surgical procedures, YNSURG (.32). Consistent with earlier findings in the overall sample, population density, DENS, was highly correlated with both provider size, SIZE, and admissions per thousand population, ADMISS.

PAGE 408

TAII.E C . TABLE E.l5 Cotrelatla. Natrla of Patients• Eplaodes wtth Coal at Adalssion of RehabllJtatia. for the Saaple (N-276}8 t:I'E INFf.CT NEOP ESOOC llOOil II:STAL NERVS CIAI:U RESP DICEST GE.'HTO SUN ACCIIJ AIJAIJL \'NSURG CARE INSIJR PPAY SIZI: DESS AlCISS 1.00 .06 1.00 .12 .22 1.00 .07 .02 .02 1.00 .00 -.02 -.01 .02 1 .00 . 09 .OS -.01 -.OJ -.OS 1.00 .20 .114 .02 -.01 -.02 -.OJ 1.00 .01 -.17 -.OJ -.02 -.OJ -.04 -.02 1.00 .II .38 -.02 -.02 -.OS -.06 -.02 -.03 1.00 .OJ .17 -.04 -.07 -.13 -.16 -.07 -.09 -.14 1 . 00 -.03 .04 .03 -.Ol -.OS -.06 -.OJ .04 -.06 -.16 1.00 .03 .OS -.00 -.01 -.OJ -.OJ -.01 . 02 -.Ol -.01 -.OJ 1.00 -.OJ .00 -.07 .Ill -.OJ -.OJ -.01 -.02 . OJ -.08 .03 -.01 1.00 -.06 -.OJ -.06 -.01 -.02 -.OJ -.01 .02 .02 -.07 -.Ol -.01 .01 1.00 . 12 -. 06 -. ll -. 15 -. 06 . 09 -. ll . 37 -. IS . 07 -. 07 -. 06 l. 00 -.09 .01 -.OJ -.OS -.09 -.11 -.OS . 06 -.10 -.27 .11 -.OS .US .OS . 26 1.00 .32 • OJ -. 21 .OS -.04 -.07 .02 -.10 .17 .IS -.OS -.OS -.OJ -.07 .io .01 1.00 .02 -.10 -.OJ .OS .14 -.OS .OS -.ll .OS -.40 -.11 .01 .08 .02 .32 .10 -.IJ 1.00 . 00 . 06 .OS .04 .OJ -.00 .04 . 06 -.2J -.01 .OS .OS -.OS .04 -.00 .07 .15 .OS 1 . 00 .01 .6J .17 •.00 -.00 -.02 .07 -.16 -.21 .12 .01 .09 .02 . 01 -.00 .07 .12 -.03 .11 1.00 • 04 • l8 • 00 . OJ • 07 . 08 . OJ . 20 .21 .10 -.OJ -.04 -.0 • .20 .OS .11 .02 .02 -.10 -.47 1.00 -.0 7 -.47 -.22 .04 -.07 -.06 -.04 .01 -.06 -.02 .114 -.OS .04 .04 -.Ol -.00 -.07 -.01 -.01 -.51 -.13 1 . 00 . 08 -.02 . 10 -.02 -.OS .JO -.02 -.OJ -.OS .00 -.06 -.OJ -.Ol .02 -.02 -.01 -.II -.OJ -.06 -.JJ -.08 -.09 1.00 . 00 .04 .IS -.08 .OS .10 .06 -.01 -.12 .03 . 09 .oa .01 .OJ -.06 . 01 .06 -.14 .ll .01 .05 -.lJ .06 1.00 .II .01 .U -.OS .OJ .23 -.OS .01 .02 -.01 .09 .Ol z ... .oa .06 .02 -.01 -.01 .06 .15 -.07 ... :z .09 .04 -.01 .09 .OS -.12 .05 -.11 .02 -.OS -.02 .ll -.02 -.09 .02 .01 8Correlatlon of .10 or better are slJnlflcant et the .OS level for a saaple size of 276 observations. .IS .04 .07 -.10 •.00 .07 -.04 .08 .00 -.Ill .n a.oo .OJ .51 l.llO "' :z :!: "' v. j v 0 lJ

PAGE 409

390 Dependent and Independent Variables The final regression equations estimated for the sample of rehabilitation patients in Massachusetts inc 1 uded a subset of vari-ables selected after review of zero-order and cano. nical cor-relation and descriptive findings. The variables included in the analysis were: Patient-Specific Variables: Provider/Health System Variables: Outcome-Related Variables: Regression Findings AGE ALONE NEOP NERVS CIRCU RESP MUSCO ACCID ADADL YNSURG REFSOR25 MEDICARE MEDICAID INSUR SIZE ADMISS ADLCHG BETWORSE The regression findings presented in Table E.l6 illustrate the difference in the overall R2 when this subsample of episodes in Massachusetts was compared to the total sample of episodes. The SThis was a dichotomous variable indicating the source of referral to home care (0 =self; 1 =other).

PAGE 410

391 TABLE E. 16 FIXED AND STEPWISE REGRESSION RESULTS ON CPE AND ALL GROUPS OF VARIABLES FOR MASSACHUSETTS SAMPLE WITH Independent Variables AGE ALONE NEOP EN DOC NERVS CIRCU RESP MUSCO ACCID ADADL YNSURG REFSOR2 ADLCHG BETWORSE MEDICARE MEDICAID INSUR SIZE ADM ISS Constant GOAL OF REHABILITATION Regression Coefficient Significance 2. 72 202.26 -253.66 -331.72 -116.53 -252.55 -328.38 -246.72 -379.68 78.59 63.52 96.66 9.59 -235.08 69.76 145.87 143.88 -1.62 63.28 -293.94 R2 = .265 N = 267 .264 .005 .134 .309 .514 .029 .037 .030 .002 .000 .358 .203 .637 .003 . 515 .301 .278 .179 .001 F = 4.68 (p=.OOO) Carrel. w/ Dep. Var.a .048 .096 .001 -.075 . 121 .033 -.029 -.060 -.087 .327 -.002 .007 -.080 -.214 .076 .052 -.065 -.012 .238 Source: Massachusetts Discharge Summary System, Medicare Cost Reports, and AHA Guide. aThe correlation coefficients presented here differ slightly from those shown in the correlation matrix because the regression equations were run on a smaller sample.

PAGE 411

392 findings indicated that a significantly larger portion of the variation in CPE was accounted for by the three categories of variables for this sample than for the total sample. The patient-specific variables accounted for the majority of the overall explanatory power of the equation (R2 = .221), when included in the estimated equations by themselves. When all three categories of variables were included, the R2 was larger (.265) and remained significant at the .001 level. When all the variable categories were included, six of the patient-specific variables had significant regression coefficients. ALONE was significant at the .005 level, indicating the relative importance of 1 iving with others, and its effect on the cost per episode of patients receiving rehabilitative care. Several of the diagnostic categories were significant, and all exhibited negative regression coefficients. They included ENDOC, CIRCU, RESP, MUSCO and ACCID. The final significant variable in the patient-specific category was AOADL (at the .001 level). This was consistent with earlier findings of descriptive analyses and regression equations and indicated the importance of admitting functional status as an explanatory factor. The only outcome-related variable with a significant regression coefficient in the final equation was BETWORSE. It also exhibited one of the highest correlation coefficients with the dependent variable of any of the independent variables included in the equation. Si nee it was carrel ated with the other outcome-re 1 ated vari ab 1 e included in the equation, it probably represented the impact of

PAGE 412

393 improvement in health or functional status combined on the cost per episode of illness. The negative sign of the regession coefficient indicated that for rehabilitation patients, the greater the improve ment the lower the average cost per episode. This was contrary to findings for the general patient population. Only one of the provider/health system characteristics was significant when included with all other variables in the regression equation. It was the variable AOMISS which was significant and positively related to CPE at the .001 level. This finding was con-sistent with earlier equations estimating the variation in CPE for the total sample of Massachusetts cases. In SliTlmary, patient-specific variables accounted for a substan-ti ally 1 arger portion of the variation in CPE for the subset of Massachusetts cases with the goal of rehabilitation at admission than for the total sample of patients in Massachusetts. One implication of this finding is that because typical patient descriptors explained more variation in cost for this subsample of patients, policymakers may be faced with a more stable basis for cost projections for rehabilitation patients than for others. Regression on Cost Using All Variables, Terminally Ill Cases Only Approach and Sample The purpose of the final set of regression equations presented here was to determine the re 1 at ive impact on CPE of the two categories of independent variables, patient-specific and provider/

PAGE 413

394 health systan characteristics, for a subset of cases described at the time of admission as terminally ill. Outcome-related variables were omitted since all patients in the sample deteriorated over time. The purpose of the analysis was to determine (1) if the inde pendent variables explained any more of the variation observed in CPE for this subset of cases than for the total sample, and (2) if patterns of significance were similar in the two samples. Data on this variable were not available for Philadelphia cases at the time of the study; consequently, the sample was comprised of the 87 episodes classified as terminally ill at time of admission to home care in Massachusetts. The process used to estimate the final equation presented here was similar to that described in the preceding sections. Pre 1 imi nary Statistics Table E.17 presents the zero order correlation matrix for the Massachusetts terminally ill cases. Patterns within the correlation matrix were similar to those observed for the entire sample of Massachusetts patients. MEDICARE, MEDICAID and INSUR were highly correlated with the variable AGE. The other payor source variable, PPAY, was positively correlated with the patient's living arrange ment, ALONE ( .24). As in earlier analyses, the provider/health system variable measuring density, DENS, was correlated with provider size and admissions per thousand population.

PAGE 414

TABLE 6.11: TABLE E. 17 Correlation of Patients' Episodes Claaalfled aa Ter.lnally .Ill at Adalaslon for the Nassecbusetta Se!ple (N•I7)a CPE 1.00 AGE .09 1.00 ALONE -.OJ -.OS 1.00 NEOP -.OS -.04 .10 1.00 ENOOC .01 -.17 .OJ -.J2 1.00 CIRCU .12 .09 -.06 -.6S -.02 1.00 RESP -.04 -.07 -.03 -.32 -.01 -.02 1.00 DIGEST .03 .06 -.OJ -.32 -.01 -.02 .01 1.00 GENJlO -.01 .01 -.OS -.4S -.02 -.OJ -.02 -.02 1.00 ADAPL .ll .09 -. 16 -.ll .09 .OS -.OS .09 .10 1.00 YNSURG .OS -.11 • 04 .25 -.08 -.16 -.08 -.08 -.11 -.19 1.00 Rr:FSOR2 .06 -.01 .11 -.12 .04 .08 .04 .04 .06 .04 -.03 1.00 .04 .7S -.10 -.13 .09 .18 -.13 .09 -,Ol .12 -.06 -.01 1.00 .16 -.23 .OJ .00 .04 -.01 .30 -.04 -.06 -.03 -.04 .13 -.43 l.OO INSUR -.17 -.54 -.07 .19 -.06 -.12 -.06 -.06 -.09 -.OS .14 -.05 -.67 -.20 1.00 PPA -.08 -.05 • 24 -.20 -.02 -.03 -.02 -.02 .49 .07 -.11 .06 . 11 -.06 -.09 1.00 SIZE -.02 -.03 -.12 .OS .08 -.06 .08 .08 -.20 .17 -.01 .02 -.02 .05 .OS -.20 1.00 PENS -.04 -.08 -.u .IS . 06 -.12 .06 . 06 -.25 .u .01 .01 -.03 -.03 • 10 -.25 .80 1.00 ADHISS -.06 -.08 -.05 .19 -.03 -.13 -.03 -.03 -.13 .01 -.OJ -.02 -.OJ -.12 . 09 -.u -.05 .56 1.00 N g c ... Iii ..., "' ::l ... !i !i .... .... "" ... i ,.. "" 0 ..., z ... c !2 : (J "" ... c rJ ::;! i ... <( z u ... .... en •correlation coefficient• of .17 or better are slenlflcent at the ,05 level for • se8ple size of 17 observettona. UJ \0 c.n

PAGE 415

396 Means and standard deviations of the dependent and independent variable for all cases categorized as terminally ill at admission are 1 i sted in Tab 1 e E .18. The aver age cost per episode for this sample of patients was slightly lower than the average cost per episode for all Massachusetts cases ($251.72 versus $259.05), but significantly lower than the average cost per episode for patients with a goal of rehabilitation. It should be noted that the average age of patients whose episodes were included in this analysis was significantly lower than the overall average age of Massachusetts patients, and the proportion of patients living alone was also significantly lower. Primary diagnoses shifted dramatically in this with 90% of the cases having the diagnosis of neoplasm, NEOP. The aver age admitting ADL index score at the time of admi ssion, ADADL, was significantly higher than for the majority of patients. This indicated significantly greater debilitation on the part of the terminally ill, since the average patient was dependent on four items out of a six-item scale. The majority of patients in this sample were referred to home care by a health care provider/ facility, as indicated by the variable REFSOR2. Of the provider/ health system variables, the average value associated with each of the variables in this category was approximately the same as in the total sample of Massachusetts cases. Dependent and Independent Variables The final regression equation estimated for the terminally ill cases in Massachusetts included a subset of variables selected after

PAGE 416

Table E.l8 MEANS AND STANDARD DEVIATIONS OF SELECTED VARIABLES FOR ALL TERMINALLY ILL CASES FOR THE MASSACHUSETTS SAMPLE (N=87) V ariable Mean 397 S.D. CPE $251.72 339.27 AGE 64.72 12.47 ALONE .08 .27 NEOP .90 .30 CIRCU .OS . 21 ADADL 3.97 2.32 YNSURG .36 .48 REFSOR2 .89 .32 MEDICARE .59 .50 MEDICAID .11 .32 SIZE 9149.58 2365.15 ADM ISS 10.07 1.63 Source: Massachusetts Discharge Summary System, Medicare Cost Reports, and AHA Guide.

PAGE 417

•398 review of the zero order correlation matrix and descriptive findings. These variables included the following: Patient-Specific Variables: Provider/Health System Variables: Regression Findings AGE ALONE NEOP CIRCU ADADL YNSURG REFSOR2 MEDICARE MEDICAID SIZE ADMISS Table E.l9 indicates that for all terminally ill cases in the Massachusetts sample, regression equations estimating the impact on CPE of the two categories of variables produced significantly different findings than for the preceding ana lyses. The R2 was s i gnificantly reduced (to .089) and was no longer significant (p = .765). None of the variables had significant regression coefficients. One implication of this finding is that traditional patient descriptors, in addition to characteristics of the health care system, are inadequate to explain variation in CPE for this subsample of patients. This has significant implications for the expansion of home health care services to the terminally ill, since ability to explain costs is limited. This implication was discussed further in Chapter VI.

PAGE 418

TABLE E. 19 FIXED REGRESSION RESULTS ON THE CPE AND ALL GROUPS OF VARIABLES FOR TERMINALLY ILL PATIENTS FOR MASSACHUSETTS SAMPLE Independent Variables AGE ALONE NEOP CIRCU ADADL YNSURG REFSOR2 MEDICARE MEDICAID SIZE ADM ISS Constant Coefficients Significance ' 3.41 -65.50 44.15 260.23 -19.64 49.41 41.28 17.42 217.81 -.66 1.89 -1 .69 R 2 = .089 N = 87 . 471 .645 .794 .273 .251 . 551 .731 .893 . 107 .968 .937 F = .67 (p=.765) Carrel. w/ Dep. Var. .093 -.033 -.048 .125 -.128 .052 .059 . 041 . 158 -.022 -.057 399 Source: Massachusetts Discharge Summary System, Medicare Cost Reports, and AHA Gu. i de.