Open Collections

UBC Theses and Dissertations

UBC Theses Logo

UBC Theses and Dissertations

Equity in health, health care services use and health care financing in British Columbia, 1992 and 2002 McGrail, Kimberlyn Marie 2006

Your browser doesn't seem to have a PDF viewer, please download the PDF to view this item.

Notice for Google Chrome users:
If you are having trouble viewing or searching the PDF with Google Chrome, please download it here instead.

Item Metadata


831-ubc_2006-200452.pdf [ 16.81MB ]
JSON: 831-1.0092824.json
JSON-LD: 831-1.0092824-ld.json
RDF/XML (Pretty): 831-1.0092824-rdf.xml
RDF/JSON: 831-1.0092824-rdf.json
Turtle: 831-1.0092824-turtle.txt
N-Triples: 831-1.0092824-rdf-ntriples.txt
Original Record: 831-1.0092824-source.json
Full Text

Full Text

EQUITY IN H E A L T H , H E A L T H C A R E SERVICES USE A N D H E A L T H C A R E FINANCING IN BRITISH COLUMBIA, 1992 A N D 2002 by K I M B E R L Y N MARIE McGRAIL B.B.A. , The University of Michigan, 1988 M.P.H., The University of Michigan, 1991 A THESIS SUBMITTED IN PARTIAL FULFILMENT OF THE REQUIREMENTS FOR THE DEGREE OF DOCTOR OF PHILOSOPHY in THE F A C U L T Y OF G R A D U A T E STUDIES (Health Care and Epidemiology) THE UNIVERSITY OF BRITISH C O L U M B I A April 2006 © Kimber lyn Marie M c G r a i l , 2006 ABSTRACT The objectives of this thesis are to identify equity objectives in the stated intentions of the health care system in British Columbia, and then to assess the extent to which the system appears to be meeting those objectives. The primary equity principle that emerges through an historical review of legislation and policy documents is the intent to deliver hospital and physician services based on need rather than ability to pay. A principle of equity that is equally important but gets much less attention is a separation between contributions to financing and the use of physician and hospital services. More contentious is how that financing should be shared across income groups. An analysis of health status, health care services use and health care financing in British Columbia in 1992 and 2002 shows that, as expected, health status increases with increasing income. Health care services are largely provided according to need, though the use of surgical day care services became more likely among higher income groups over that decade. In contrast, admission to an acute inpatient facility is more likely for lower-income individuals; lower income groups also use more inpatient acute services, on average, once admitted. Financing for the physician and hospital sectors in B.C. comes from general tax revenues. There were many changes in taxes during this decade; nevertheless, contributions to health care financing were distributed across income groups in about the same way in 1992 and 2002. The analyses here show the tax system is only mildly progressive, meaning that higher income groups contribute only a slightly higher proportion of income to taxes. This finding is at odds with the perception of the general public, which appears to believe that physician and hospital services are (and should be) financed through progressive taxation, more like what is seen with income taxes. Further changes in the physician and hospital sectors and in the health care system overall have the potential to erode equity currently in the system. If there is a commitment to principles of equity, better measures and more monitoring of achievements will be important as these changes continue. TABLE OF CONTENTS Abstract ii Table of contents iii List of tables vi List of figures viii Acknowledgements ix Dedication x Chapter 1: Introduction 1 1.1 Motivation 1 1.2 What is equity? Equity of what in British Columbia? 3 1.3 Measuring and monitoring equity in health, health care services utilization and health care financing 5 1.3.1 Need for health care services 6 1.3.2 Health care services use 7 1.3.3 Health care financing 8 1.4 Achievement of objectives? 10 1.5 A brief introduction to data sources and methods 10 1.6 Contributions of this work 12 Chapter 2: Equity in health and health care 14 2.1 Introduction 14 2.1.1 Methods 15 2.2 Political philosophy and equity 16 2.2.1 Why equity? 16 2.2.2 How does equity apply to health and health care? 26 2.3 What is the place of equity in health care in Canada? 30 2.3.1 The first decades of the 20th century 31 2.3.2 British Columbia Royal Commissions on Health Insurance 33 2.3.3 A short digression to federal-provincial fiscal relations 36 2.3.4 The war years 38 2.3.5 The Hospital Insurance and Diagnostic Services Act 41 2.3.6 The Hall Commission 44 2.3.7 The Medical Care Act 46 2.3.8 The Established Programs Financing Act 48 2.3.9 The Canada Health Act 49 2.3.10 The B.C. Royal Commission on Health Care and Costs and the Medicare Protection Act 52 2.3.11 The Canada Health and Social Transfer 54 2.3.12 Three recent federal reports 55 2.4 Identifying and monitoring equity 58 Introduction to Chapters 3, 4 and 5 62 Chapter 3: The distribution of income and need for health care services 63 3.1 Introduction 63 3.2 Methods I - measuring inequality 63 3.2.1 A summary measure for inequality 64 in 3.2.2 Why Gini coefficients? 67 3.2.3 Preference weightings in the Gini coefficient 68 3.2.4 Concentration indices 74 3.3 Methods II - quantifying income distribution in British Columbia 75 3.3.1 Income statistics from the Canada Revenue Agency 75 3.3.2 Choosing an ecological measure of income 76 3.3.3 Income statistics from Statistics Canada tax-filer data 77 3.4 Methods III - Measuring the need for health care services 86 3.5 Results I - income distribution 92 3.6 Results II - need for health care services 96 3.7 Discussion and conclusion 108 Chapter 4: Equity in health care services use 113 4.1 Introduction 113 4.2 Methods and data sources 114 4.2.1 Statistics Canada 114 4.2.2 The BC Linked Health Database 115 4.2.3 Analysis 125 4.3 Results 127 4.3.1 A brief comparison of administrative- and survey-based measures of inequality 143 4.4 Discussion and conclusions 146 Chapter 5: The redistributive effect of health care financing 150 5.1 Introduction 150 5.2 Data sources and variables 153 5.2.1 Overall distribution of health care finance 154 5.2.2 Estimating consumption taxes 156 5.2.3 Calculating the payment of the B.C. premium 158 5.2.4 Adding a family identifier 159 5.3 Analysis 161 5.4 Results 164 5.4.1 Distribution of income and progressivity of payments 164 5.4.2 The progressivity of health care financing and redistribution of health care expenditures 170 5.5 Discussion and conclusions 174 5.5.1 British Columbia in an international context 175 5.5.2 Equity and health care financing in British Columbia 177 Chapter 6: The equity agenda 179 6.1 Introduction 179 6.2 Health care systems and equity of what? 180 6.3 Is there equity in British Columbia's health care system? 181 6.3.1 Health care services use 181 6.3.2 Health care financing 183 6.3.3 What are the prospects for the future? 184 6.4 Future areas of inquiry 187 6.4.1 Limitations of the current work 187 6.4.2 The research agenda for equity 189 iv References 193 Appendix 1 - Sample Stata code for data analysis 207 Appendix 2 - Population distribution by A C G group 222 Appendix 3 - Sample SAS code for data preparation 226 Appendix 4 - Standard errors from analyses of health care services utilization 238 v LIST OF TABLES Table 1: Data requirements and data sources for empirical analyses 11 Table 2: Theories of social organization and application to health and health care 27 Table 3: Initial and modified income distributions in a 10 person society 72 Table 4: Income, government transfers and tax information requested ! 78 Table 5: Total income and adjusted income by household, year and equivalization method 81 Table 6: Calculations required for the construction of Lorenz curves 82 Table 7: Gini coefficients for market income, total income and disposable income, British Columbia, 1992 and 2002 , 93 Table 8: Population characteristics for records that were dropped from analysis 97 Table 9: Characteristics of the study population, 1992 and 2002 98 Table 10: Need characteristics of selected Adjusted Clinical Groups (ACGs) in British Columbia, 1992 and 2002 100 Table 11: Study population characteristics for individuals in high frequency and high need ACGs 102 Table 12: Unstandardized and standardized Gini coefficients for need for health care services and concentration indices for income-related need for health care services 108 Table 13: Life expectancy for Canadian males, 1996, by neighbourhood income quintile 110 Table 14: Resource intensity weight by level of care, 2002 119 Table 15: Population characteristics for users and non-users of health care services 128 Table 16: Number and percentage of population that use services by health care service type, 1992 (excluding non-users) 130 Table 17: Number and percentage of population that use services by health care service type, 1992 (excluding non-users) 131 Table 18: Average use of services (in $$) 1992 133 Table 19: Average use of services (in $$) 2002 134 Table 20: Average use of services (in $$) by income decile for select high frequency and high cost ACGs, 1992 137 Table 21: Average use of services (in $$) for select high frequency and high cost ACGs, 2002 138 Table 22: Inequality and inequity in health care services utilization, 1992 139 Table 23: Inequality and inequity in health care services utilization, 2002 140 Table 24: Comparison of administrative data-based and survey data-based results for British Columbia, 2002 144 Table 25: Consolidated provincial government revenues for fiscal year ending March 31 s t, British Columbia, 1989 - 2003 155 Table 26: Estimates of household-level consumption taxes, as derived from the Social Policy Simulation Database and Model 158 Table 27: B.C. premium rates and income-based subsidy cut-points, 1992 159 Table 28: B.C. premium rates and income-based subsidy cut-points, 2002 159 Table 29: Gini coefficients for market income, total income and disposable income, British Columbia, 1992 and 2002 164 Table 30: Income, taxes, transfers and health care services use by decile, Gini coefficients, concentration indexes and Kakwani indexes, 1992 and 2002 ($ in 000,000) 166 Table 31: Summary of Kakwani indexes and "g", British Columbia, 1992 and 2002 169 Table 32: Overall progressivity of financing of physician and hospital services in British Columbia, 1992 and 2002 170 vi Table 33: A comparison of ratios of income of highest income and lowest income deciles in Manitoba and British Columbia 171 Table 34: International progressivity indices from previous analyses, progressivity indices for British Columbia from present analyses 176 Table 35: Summary table for future areas of inquiry related to equity 189 vn LIST OF FIGURES Figure 1: A policy and legislative time-line for Canada and British Columbia 31 Figure 2: Hypothetical Lorenz curve of income inequality 64 Figure 3: A method for calculating the Gini coefficient 66 Figure 4: "Weights" when the parameter v is varied from 1.5 to 2 to 4 70 Figure 5: Lorenz curves for the baseline distribution and transfers between the tails 73 Figure 6: Lorenz curves for the baseline distribution and transfers around the middle 73 Figure 7: Lorenz curves based on different household weighting concepts 83 Figure 8: Lorenz curves based on Canada Revenue Agency data, 1992 and 2002 92 Figure 9: Range (95 t h-5 t h percentile) of income, by income band 95 Figure 10: X Y scatter plot of 5 t h percentile against 95 t h percentile of income 96 Figure 11: Range (95 t h percentile - 5 t h percentile) of health care services use within each A C G 104 Figure 12: Unadjusted and adjusted Lorenz curves for need for health care services, 1992 105 Figure 13: Adjusted Lorenz curves for need for health care services, 1992 and 2002 106 Figure 14: Concentration curves (unadjusted and adjusted) for need according to ranking by income, 1992 107 Figure 15: What do we know about postal codes in the BC Linked Health Database? 122 Figure 16: Hypothetical Lorenz curve of pre-tax income (red) and concentration curve for tax payments (pink) 162 Figure 17: Average tax rates in British Columbia by income group based on Canada Revenue Agency data, 1992, 1997 and 2002 165 Figure 18: Concentration curves for income tax payments, Statistics Canada and Canada Revenue Agency data, 1992 and 2002 167 Figure 19: Tax payments and hospital and physician expenditures by income decile, British Columbia 2002 172 Figure 20: Taxes paid and physician and hospital expenditures by need deciles, British Columbia 1992 and 2002 174 Figure 21: Hospital and physician expenditures as a share of total health care expenditures, Canada and British Columbia, 1975 to 2003 185 Figure 22: Proportion of public financing of health care, Canada and British Columbia, 1975 to 2003 186 viii ACKNOWLEDGEMENTS This thesis represents my work, but I am grateful for the support, encouragement and guidance of many individuals, many who may not even recognize the extent of their contributions. With apologies to anyone I neglect to mention, special thanks go to the following: Members of Health Care and Epidemiology thesis screening panel offered encouragement and useful comments at the start of the process. Thanks especially to Jeannie Shoveller, who provided time and assistance with earlier thoughts on this project and consequently helped me steer clear of potential pitfalls. Charlyn Black and staff and faculty of CHSPR understood and were supportive when I needed to back off and dedicate myself to this work. My colleagues at CHSPR were also always willing to talk about data and to answer crazy questions about analysis. Thanks in particular to Chris McLeod, who shared his office with me, helped me learn Stata, and listened to my (many) frustrations throughout the process. Denise Morettin, Anna Low, Barbara Weinberg, Leanne Warren, Bev Dale and others helped shepherd my B C L H D data access request through the acquisition and data preparation processes. Stephen Lee at the B.C. Ministry of Health was happy to share their costing methodology for hospital stays, and to answer my questions about it. Mary Luebbe and staff at UBC Data Services were always willing to retrieve new data sets available through the Data Liberation Initiative. Statistics Canada Small Area and Administrative Data Division, especially Janie Saumure, provided enormous help in investigating and eventually requesting a special (custom) run of taxfiler data. Brian Murphy, also at Statistics Canada, helped with SPSD/M Russell Wilkins, from the Health Analysis and Measurement Group at Statistics Canada, was always happy to engage in discussion and debates about topics ranging from small area analysis to derivation of income per person equivalent. Geoff Ballinger and other staff from the Canadian Institute for Health Information helped clarify some of the trends discovered in national health expenditures data. Margaret McGregor, Marcy Cohen, Anne-Marie Broemeling, Bob Tate, Lisa Ronald and others from an ongoing research group, were supportive and always interested in my work as it progressed. Michael Mendelson provided huge assistance in identifying useful material on the historical development of Canadian federalism and the health care system Eddy van Doorslaer and Tom van Ourti were always happy to receive and respond to queries about the minutiae involved in equity analysis. And Eddy and Owen O'Donnell taught an excellent course on quantitative analyses in health equity that provided the backbone for the empirical part of this work. Alan Thomson, Jane Coutts, Heidi Matkovich and Nancy Meagher read and provided comments on parts of the thesis in its early stages and were always supportive throughout. Professor Michael McDonald read and provided thoughtful and very helpful comments and suggestions for earlier drafts of Chapters 1 and 2. Charles Normand and Allan Maslove read and critiqued early versions of my thesis proposal and kept me from being distracted from the main questions of interest. Cam Mustard and his co-authors provided the initial inspiration for the topic of the project, as well as enthusiasm to see further work in this area carried out. And, of course, Morris Barer, Bob Evans and Clyde Hertzman, my thesis committee. Thank you for everything. ix DEDICATION To the three boys. At last. Chapter 1: Introduction 1.1 Motivation The health care system in Canada is often described as a key national treasure, one that defines the country's identity. The health care system is regularly at or very near the top of public preoccupations, and so not surprisingly garners ample attention around federal and provincial elections. During the latter half of the 1990s there was a steady increase in the proportion of Canadians describing health care as the number one issue requiring the attention of political leaders (Mendelsohn 2004). Canadians in increasing number believe the system needs major reform, propelled by declining faith that the system will be there to support them if (or when) they need it (Blendon et al. 2003; Schoen et al. 2002). The ongoing presence of health care in discussions and debates reflects not only the importance of the system to the general public, but also the fact that health care in Canada is largely a public enterprise. The majority of health care services are publicly funded, the private insurance market is absent in the key areas of hospital and physician services, and the provincial government is the primary payer. There is not unanimous support for the public nature of health care financing and delivery in Canada. The clash of interests and political philosophies about the "correct" role for the state ensure there is always a crisis said to be brewing somewhere in the health care system. In many cases such apparent (or alleged) crises turn out to have no basis in evidence. Instead, myths are recycled, based either on unsubstantiated claims or pure fabrication (Barer et al. 1998). These "crises" are often statements about values disguised as irrefutable and undeniable facts. This is particularly true for claims about financing the health care system, where the "facts" put forward about financing inevitably reduce to differing views about who should pay for health care services, how much, and under what circumstances (Evans 2002; Stoddart et al. 1993). Though debates about access to and payment for health care services ultimately can be understood as differences in underlying values, those values are rarely explicitly exposed. When discussions about values do take place in Canada, there is remarkable consistency within the general public in support of a few basic underlying principles or values to guide the health care system. In general, the principles desired of the health care system are those embodied in the Canada Health Act: a system with universal coverage of the population that provides reasonable and fair access to a comprehensive set of services available to Canadians even as they move across the country, all administered by public (rather than for-profit) agencies. 1 The continual clash of values and interests - or the clash of values held by different interests -has influenced the development of health care systems in Canada. In one of the more famous examples, physicians staged a strike in 1962 in Saskatchewan to protest the implementation of that province's plan for universal insurance for physician services (Badgley and Wolfe 1967). The outcome of these clashes (past, present and future) depends on the relative influence of the contending forces, and their ability to sway the decisions of either influential actors or political institutions (Maioni 2004). In the case of the doctors' strike in Saskatchewan, the opinion of the majority of the public - inside and outside the province - and the support of the provincial government for universal health insurance eventually forced the physicians to back down. In Saskatchewan and elsewhere, the resulting health care systems reflect, albeit sometimes imperfectly, a nation's or community's relationship to and understanding of the role of government, and the rights and responsibilities resting with each citizen (Rawls 2003). Roy Romanow, in his 2002 report of the Commission on the Future of Health Care in Canada articulated the meaning of the values embodied in the Canada Health Act, saying that Canadians told the Commission that they remain committed to equity, fairness and solidarity (Romanow 2002). A system that values solidarity, for example, is one that strives to cover all residents of Canada, on roughly equal terms and conditions, rather than being limited to groups based on age, sex, income, geographic location or some other characteristic. Of equity, solidarity and fairness, it is equity that has been described as the "cornerstone" of the current Canadian health care system (Stoddart and Labelle 1985). The standard description of equity in relation to health care in Canada is the provision of health care services based on need rather than ability to pay. The Romanow report further clarified this, describing the structure of the Canadian health care system as one built on "double-solidarity". In this sense, contribution to the health care system through taxes ensures equity both in that contributions to financing are made based on ability to pay, and in that receipt of services is based on need. Despite general consensus among the public that equity is a fundamental value of the Canadian health care system, there are multiple definitions of equity at play. The Senate Standing Committee on Social Affairs, Science and Technology, chaired by Senator Michael Kirby (the Kirby Committee), in a national review of the health care system emphasized that individuals should not face undue financial hardship based on their use of the health care system (Canada.Parliament.Senate.Standing Senate Committee on Social Affairs 2002). This is a different approach from the equity model proposed by the Romanow report, where financing is separated from the use of health care services. 2 This project starts with the acceptance of equity as a central value of the health care system, a value held and expressed time and again by Canadians. The purpose of the present work is first to identify the various potential definitions of equity with respect to health care systems in Canada, and in British Columbia in particular, and second to investigate to what extent equity is embodied in the design and operation of the existing system in B.C. Health care systems evolve over long periods of time and are undeniably a product of experience. But that experience is replete with interest group pressures and countervailing forces. Interests, in whatever form, try to shape public opinion and the public's values. Both interest groups and public opinion influence (or at least try to influence) the formation of public policy. Values of the majority of the public may work their way into the organization of health care systems, but they may also be overcome by other considerations. The objectives of this thesis are: 1) To identify competing visions of equity and their implications for the design of health care financing and delivery; 2) To apply this framework of competing principles of equity to the legislative and policy history of health care systems in British Columbia, to identify the major equity principles at play; 3) To apply analytic methods for measuring equity in health, health care services utilization and health care financing originally developed for use with survey data, to administrative data in British Columbia; 4) To assess from an equity perspective how health, health care financing and health care services utilization have changed during a period (1992-2002) of restructuring in both health care financing and health care delivery; 5) To identify the equity principles implicit in the financing and organization of the health care system in British Columbia, as against those explicit in legislation and policy documents. 1.2 What is equity? Equity of what in British Columbia? An understanding of what principles of or objectives for equity are present in any health care system first requires a general framework for thinking about the notion of equity. A review of political philosophies sets out to develop this framework, with particular attention to the expression of equity within each. The political philosophies reviewed range from a libertarian pure market-based perspective as articulated by Robert Nozick, emphasizing protection of 3 individual freedom at all costs, to egalitarian approaches to sharing resources as articulated by Amartya Sen. A l l of these philosophies promote equity in some form. Isaiah Berlin argues that once rules of engagement in a society are set, equality demands that the rules be imposed fairly and impartially (Berlin 1978). This helps to distinguish equity and equality; the particular form of equity desired determines the rules a society establishes, and equality is then demanded in their application. Philosophies are delineated by the particular views each holds of the "correct" arrangement and articulation of rules that govern society, the set of rules that should be applied equally to all citizens (Sen 1992). Equity in the libertarian or entitlement philosophy developed by Robert Nozick is in the state imposing minimal restrictions on individual liberty (Nozick 1974). Equity is achieved when individuals are left alone, without regard for or concern with the distribution of goods, wealth or opportunities among them. At the other end of the spectrum, Amartya Sen describes a more egalitarian philosophy, where equity requires state intervention to ensure a (more) equal distribution of goods, wealth and opportunities. "Equal" has the same meaning in both philosophies, at least in the sense that once the rules are set, they must be equally applied. But the differences in the conception of equity will clearly lead to quite different societal outcomes. The general framework described by the review of philosophies is then extrapolated to apply to health and health care more specifically, offering an analysis of how each of the philosophies covered might play out in a modern-day organization of the financing of the health care system and the delivery of health care services. The society that follows from Nozick's philosophy, for example, would be based on health care being the responsibility of those who require services, meaning that services would flow to people who were able to pay for them. A society more closely aligned with Sen's philosophy, on the other hand, would be likely to provide health care services according to need rather than ability to pay, with a goal of minimizing inequalities in health status. The next step is a review of major health-related legislation and policy documents in Canada and in British Columbia in particular. This starts in the early 20 t h century with communities organizing to pay physicians in Saskatchewan in the 1910s, and proceeds to three federal reports on health care published in 1997 (one) and 2002 (two). The purpose of this review is to provide some perspective on how the health care system in British Columbia has evolved and what it is trying to achieve, at least as can be determined from official legal and policy positions. 4 Three related, but distinct, principles of equity emerge. The primary equity principle in health care in British Columbia is the provision of services (at least hospital and physician services) based on need rather than ability to pay. The second is the financing of those services from tax revenues, effectively separating financial contributions to the health care system from use of physician and hospital services. The views of equity expressed in the Romanow report reflect not only the present mood of the public, but also a long-standing Canadian commitment to equity in health care. These values may not be immutable, but they have been clear and consistent over several decades. More contentious is the third principle of equity, of financing hospital and physician services through progressive taxation. This principle emerges only in the Romanow Commission report in 2 0 0 2 , but seems to have longer-standing resonance with the public. The health care system in British Columbia is closer to the philosophy described by Sen than to the one described by Nozick, if one were to arrange the political philosophies along a spectrum. But that assessment is based on the stated intent of the health care system. The question that immediately follows is the extent to which the health care system in operation actually meets the two major equity objectives. 1.3 Measuring and monitoring equity in health, health care services utilization and health care financing The equity objectives identified for the health care system in British Columbia have a strong income interpretation. If services are provided based on need rather than ability to pay, then one would anticipate no differences in the use of services by income, after taking account of differences in need. Similarly, financing through taxes implies that contributions to the health care system will reflect "ability to pay", at least to the same extent that ability to pay is reflected in contributions to general government revenues that are used to finance the public portions of health care systems in Canada. The empirical work of this thesis (objectives 3 and 4 above) is thus focused on income-based equity in the payment for and use of physician and hospital services. These two sectors are currently (nearly) fully public, meaning that the majority of all physician and hospital services are financed from taxes. The expectation is both that services will be allocated based on need, and that contributions to financing will be related to ability to pay. There are other important health care sectors, such as pharmaceuticals and home care, that are financed from a mix of public and private sources, and therefore where payment and use will line up differently. Hospitals and physicians, however, represent the roots of health care systems in Canada. A n 5 understanding of equity in these two sectors is an important departure point for future investigation of equity in other areas. 1.3.1 Need for health care services An assessment of equity in the use of health care services depends on being able to identify or estimate what the expected use of services is. Measuring or estimating the need for health care services is a constant challenge in health services research. Many analyses of health care services utilization rely on income as a measure of need for health care services (see, for example, Roos and Mustard 1997). It is well understood that health is inversely related to income; wealthier people tend to be healthier, and poorer people tend to be sicker (Canadian Institute for Health Information.Canadian Population Health Institute 2004; Evans et al. 1994). This relationship means that it is legitimate, in some circumstances, to use income as a proxy for need for health care services. In the present case this is not possible, because the relationships among need, income, health care services use and payments to the health care system are of particular interest. The use of health care services will certainly be higher in lower income groups than in higher income groups because lower income groups have greater needs. But are the differences in use commensurate with the differences in need? A n ability to assess the effects of income - separated from need - on the use of or payment for health care services requires a separate measure of need for health care services. Many analyses use self-reported health status (Dunlop et al. 2000; Finkelstein 2001; Newbold et al. 1995), but such a measure is available only from survey data. There is no source of such information for the population-wide analyses undertaken here. This project develops a measure of need for health care services based on a commonly used and validated case-mix adjustment system (Reid et al. 2001). This system, developed at Johns Hopkins University, groups individuals based on their medical diagnoses. The groups were defined to reflect similarities in expected use of health care services. This measure of health status (or more accurately, i l l health) therefore reflects "need" for health care services, or perhaps capacity to benefit from the health care system, which is an appropriate construct for the questions at hand.1 For present purposes, the measurement of need for health care services is a means to an end rather than an end in itself. Achieving equity in "need" (that is, equal health status for each ' Culyer and Wagstaff (1993) provide an assessment of the implications of different connotations of "need" for health care services. They show that different definitions will not lead to the same conclusions about how health care services should be distributed. Need defined based on expected system response is most closely aligned with their description of "capacity to benefit". 6 individual in the population) does not appear as a stated policy objective of health care systems in Canada in general or in British Columbia in particular. Identifying inequities in need is of interest for some research questions, but it is not the purpose of this work to use inequalities or inequities in need as a tool to evaluate the health care system. The health care system will, to some extent, exert an influence on a population's health, but it is hardly the only, or perhaps even the most important, determinant of the health of a population. Quality of neighbourhoods, environmental factors, occupation mix and income levels, to name a few, are all important factors in the complex calculus of population health (Heymann et al. 2006). An understanding of the distribution of need nevertheless helps define the context within which the health care system operates. 1.3.2 Health care services use The primary equity principle of the provision of services based on need rather than ability to pay emerges because of the influence of health care on an individual's health status. Health care services and interventions have the intent of either maintaining health (in the case of preventive services) or returning people to prior states of health. This is likely why "equity" in health care services use is often seen as closely aligned with "access" in Canadian policy documents (Giacomini et al. 2004). The way a health care system is organized to provide services can have a direct impact on the distribution of the use of those services (Penchansky and Thomas 1981). Access, however, is of value only when the services made available are appropriate and effective. One of the struggles in health care delivery is creating the structure that achieves all of these objectives. Research suggests, for example, that patients requiring complex surgery will be most appropriately cared for by providers in "high-volume" settings (Tracey and Zelmer 2005; Urbach et al. 2004). But centralized location of services may impede the accessibility of those services unless other structures are put in place to support people who must travel to receive necessary services. There were many changes in the delivery of health care services in Canada in the 1990s. The role of acute care hospitals was de-emphasized (at least as measured in budgetary terms (Canadian Institute for Health Information 2005a)), there was a continued expansion of surgeries provided on a day (essentially outpatient) basis, there were changes in patterns of physician remuneration with more emphasis on salary and contractual arrangements in place of traditional fee-for-service payments, and there was increasing specialization among physicians. While these changes were not instituted with the direct intent of affecting who receives services, there clearly could be such 7 an impact. Assessment of equity in the use of acute and surgical day care services, and in the use of general practice and specialty services can provide information about achievement of the equity principle of service use based on need rather than ability to pay. Measuring equity in both 1992 and 2002 shows that while a significant degree of equity has been achieved in the public health care system, changes in delivery are tending to make the distribution of services less rather than more equitable. 1.3.3 Health care financing The form of financing of health care services has a tremendous effect on the distribution of the burden of payments. Generally speaking, there are five major potential sources of funds to support any health care system: direct taxation (e.g. income taxes), indirect taxation (e.g. consumption taxes), social insurance, private insurance and out of pocket payments (Wagstaff et al. 1999). The relative mix of these financing sources differentiates health care systems. In general, health care systems that rely more heavily on direct taxes as a source of finance tend to be more progressive, meaning that people with higher incomes pay a higher proportion of their income to support the health care system. This is true because income and other direct tax systems themselves are usually designed to be progressive, with tax rates being a direct (and non-linear) function of income levels. Indirect taxes, such as consumption taxes, tend to be regressive, with a greater (proportionate) burden of payment falling on lower income individuals because they tend to consume a greater portion of their income. This is a direct result of the fact that many of the consumption goods on which indirect taxes apply are non-discretionary in nature. They are purchased out of necessity by poor and rich alike, but obviously then account for a greater share of the disposable income of those less well off than of those better off. Out-of-pocket payments tend to be the most regressive form of health care finance because health care payments will be a much larger proportion of income of lower income individuals. A higher proportion of poor people have poor health which makes the impact even more pronounced (van Doorslaer et al. 1999; van Doorslaer et al. 2000; van Doorslaer and Wagstaff 1992; Wagstaff et al. 1999; Wagstaff and van Doorslaer 1992). The second equity principle that emerges from a review of policy and legislative history in Canada is the separation of financial contributions to the health care system from the use of health care services. This is a commitment to horizontal equity, meaning "income equals" would end up paying about the same amount for health care, regardless of health status. An additional principle 8 that emerges in the Romanow report is that this financing should be progressive. This is a commitment to vertical equity, meaning unequal treatment of income "unequals". The first part of the principle of "equity in finance" is achieved in Canada by relying on general tax revenues to finance the majority of services provided by physicians and hospitals. Restricting the analyses here to the hospital and physician sectors of the health care system implies that the overall progressivity of health care financing is equal to the progressivity of the overall tax system.2 Income taxes in Canada are progressive, but they account for less than one-quarter of government revenues in 2002, and were even less important a source of revenue in 2003 and 2004. Indirect taxes accounted for just under a quarter of government revenues, and the health insurance premium charged in B.C. gained importance over the study period, though it still accounted for less than 5% of the total in 2002. The overall effect is financing for the public portion of the health care system in British Columbia that is only very mildly progressive in both 1992 and 2002. Indications are that it will become less progressive over time, given the shifts in relative importance of direct and indirect taxes. If other sources of government revenue - which tend to be regressive - could be added to these analyses, financing of physicians and hospital services would be proportionate at best, and quite possibly regressive. If other health care sectors such as pharmaceuticals or home care services, which are based on more regressive forms of financing, were included, financing the system overall certainly would be regressive. Previous research in Manitoba indicated that the use of health care services has a redistributional effect on income (Mustard et al. 1998). This occurs not because of an explicit policy in the health care system to transfer money from higher income strata to lower income strata, but because the system is designed to transfer resources from the healthy (in a given year) to the less healthy. The resulting redistribution is precisely what one would expect in a system in which financing is based on ability to pay and use of services is based on need, and where there is a positive relationship between income and health status. Counting the use of health care services as a financial benefit in British Columbia shows that the health care system did redistribute income in both 1992 and 2002. The effect was about the same in both years, while income distribution itself became more unequal over time. The health care system had a redistributive effect, transferring money from the more to the less healthy, but this was not enough to offset increasing income inequality over the study period. 2 British Columbia charges premiums for medical services, but the revenues raised through this tax go into general revenues instead of being ear-marked for the health care system. 9 1.4 A c h i e v e m e n t o f o b j e c t i v e s ? Achievement of the major equity principles in British Columbia is mixed. Lower and higher income individuals clearly use the health care system differently after accounting for need. Lower income individuals tend to make more use of general practice and acute care services, while higher income individuals make more use of specialty and surgical day care services, and the income-related differences in use of specialty services tended to grow over time. The system does appear to allocate services largely on the basis of need, but this is truer of some services than of others, and there is clearly room for improvement. At the very least, there needs to be a better understanding of why the differences identified here are present, and whether they are consistent with the provision of equitable and high quality health care services. The public system covering physician and hospital services achieves one of the finance-related equity objectives simply by separating payments to the system from the use of services. Any proposal for changes in financing that would raise money from users of the health care system would have the effect of undermining this equity principle. User charges, co-payments and deductibles would all re-establish the carefully cut link between payment for and use of services. Continued changes in the relative importance of different sources of government revenues appear to be weakening the relationship between "ability to pay" and actual payment of taxes. This is a finding of general public policy rather than of health care policy specifically, but the effects can be felt directly in the sources of financing for health care, and could conceivably have an impact on the health of the population. The organization of the rest of this thesis follows as above. Chapter 2 provides an overview of philosophies and the historical review of policy and legislative documents. Chapter 3 then introduces the measures of income and need for health care services used throughout the empirical work, lays out the analytic approach, and assesses the distribution of need and income in British Columbia in 1992 and 2002. Chapter 4 analyzes equity in the use of physician and hospital services in those same years. Chapter 5 looks at equity in financing for hospital and physician services in 1992 and 2002. Chapter 6 concludes, providing a synthesis of findings, an analysis of prospects for the future and some thoughts on directions for equity research in Canada. 1.5 A b r i e f i n t r o d u c t i o n t o d a t a s o u r c e s a n d m e t h o d s The first part of the project (objectives 1 and 2, Chapter 2) relies on a literature review, first of philosophies and then of legislation and policy documents. The analyses described in Chapters 10 3 through 5 (objectives 3 and 4) required the construction of measures of income and taxes paid, a measure of need for health care services, and measures of health care services utilization and basic demographic and residence information, all at the individual and/or family level. Two primary sources of data provided these measures, as outlined in Table 1: Statistics Canada and the B.C. Ministry of Health through the BC Linked Health Database. The data used to identify income groups for families in British Columbia in 1992 and 2002 were available through a custom tabulation purchased from Statistics Canada. The data used to construct measures of need and health care services use were available from the BC Linked Health Database. Table 1: Data requirements and data sources for empirical analyses Data requirements Data sources Identification of the relevant populations of British Columbia, including demographics and residence and a "family grouper" B.C. Linked Health Database registry file Individual-level measure of need for health care services B.C. Linked Health Database, based on profile of diagnoses over the course of a year Individual-level measures of health care services utilization B.C. Linked Health Database utilization files, calculated in dollars of expenditure Income earned and taxes paid Statistics Canada custom tabulation of "tax filer data" The BC Linked Health Database is a resource maintained by the Centre for Health Services and Policy Research at U B C (Chamberlayne et al. 1998). Data from the BC Linked Health Database are available to researchers who meet access criteria, including peer and ethical review and review of data stewards, all according to an access policy that complies with B.C.'s Freedom of Information and Protection of Privacy Act (British Columbia Ministry of Health and Centre for Health Services and Policy Research (UBC) 1996; British Columbia.Legislative Assembly 1993). Peer review in this case was deemed met by committee approval of a thesis proposal3; ethics approval was obtained through the U B C Behavioural Research Ethics Board. Data were provided without personal identifiers but with study identifiers that allowed the connection of service use and registry information for (non-identified) individuals. More detail about the construction of variables is provided where each is first used. 3 The thesis proposal was also reviewed and approved by the UBC Department of Health Care and Epidemiology's Thesis Screening Panel, as required for PhD students in that department. 11 The empirical methods used to measure equity in the use of health care services and equity in financing follow from the work of the ECuity group in Europe.4 This group has conducted a number of international comparative studies over the past decade and more. Their methods have evolved over time, but they are based on Lorenz curves and Gini coefficients. Lorenz curves provide a graphical representation of the distribution of income within the population, and Gini coefficients are a way of summarizing those distributions into a single measure. These methods were developed for use with survey data, but are applied here to administrative data on the (actual) use of health care services, income and taxes paid. 1.6 Contributions of this work This work makes contributions both to methodological issues in health services and population health research, and to applied policy research with findings relevant to health and to general public policy. Finally, this work has the potential to contribute to broadening the type of health and health-care related information that can be communicated to the media and made available more generally to the public. One methodological contribution is the application of administrative data for purposes met largely in prior work by survey-based information. A significant advantage of administrative data in the analysis of equity in health care services use is that the measures of utilization represent actual use of services. What is recorded is what happened, subject to administrative errors. The data may be incomplete, because of the way they are collected, but what is there can be assumed to be relatively accurate (Roos et al. 1993; Warburton and Warburton 2004). Analyses based on survey data instead rely on self-reported measures of physician and hospital use, and are therefore subject to recall bias and response bias associated with surveys (Bancej et al. 2004; Evans and Crawford 1999; Rhodes and Fung 2004; Sackett 1979). Another significant methodological contribution is the development of need for health care services that can be used with administrative data, where those data include diagnosis codes. To this point, analyses of administrative data have relied on proxies for need, such as income, or measures of neighbourhood-based "health status" such as the premature mortality rate (McGrail et al. 2004; Roos and Mustard 1997). The present work also provides a comprehensive framework for thinking about equity in general and with specific application to health care. It offers an analysis of the equity principles or objectives that are central to the British Columbia health care system. The empirical part of 4 See the internet site for the ECuity project, accessed 25 January 2006. 12 the work assesses equity in a broader array of health care services than has previously been addressed. The most significant addition is that of the use of surgical day care services, which have not been included in previous survey-based analyses. There are significant inequities in the use of surgical day care, so this is an important expansion of the analytic base. Few studies in Canada have addressed equity in health care financing (Mustard et al. 1998; Smythe 2002) even though there is a clear equity principle in the form of financing for health care and public interest in this information. This work starts to fill this gap, at least for the medicare services that are fully financed from public sources. Finally, a fundamental change in the management of the health care system in Canada and elsewhere over the past few years has been the increasing prominence of efforts to report indicators of health care system performance and health status to the general public (2003a; Canadian Institute for Health Information and Canada.Statistics Canada 2003; Health Canada 2004; Statistics Canada 2004a). The provincial First Ministers made an explicit commitment in September 2000, reiterated in February 2003, to provide public reporting of "health performance indicators" as a condition of receiving additional federal funds for the health care system (Canadian Intergovernmental Conference Secretariat 2000; Health Canada 2003). Equity is clearly important to the public. The measures of equity in health care services use and health care financing developed here could (and arguably should) be added as indicators to the suite of information reported to the public. 13 Chapter 2 : Equity in health and health care 2.1 Introduction Many Canadians believe equity to be the great, unspoken principle of health care. The primary equity principle evident in Canadian health-care systems is the objective of providing health-care services based on need or capacity to benefit rather than ability to pay. This perspective on equity has a long history in Canadian health care, starting with the development of municipal doctor plans in Saskatchewan, through the Hall Commission and the development of the Canada Health Act in the 1980s (Begin 1988; Maioni 2004; Taylor 1978). It was also present in the report from the Commission on the Future of Health Care in Canada, in which Roy Romanow wrote: "Canadians consider equal and timely access to medically necessary health care services on the basis of need as a right of citizenship, not a privilege of status or wealth" (Romanow 2002, p. xvi). Equity may be important because of the value Canadians place on health. As Sen notes, "Health and survival are central to the understanding not only of the quality of one's life, but also for one's ability to do what one has reason to want to do. The relevance of health equity for social justice in general is hard to overstress" (Sen 2002, p. 663). At the very least, Canadians place value on equity in health care; this is a common theme of public opinion as reported by the media and in research (Giacomini et al. 2004; Kenny 1997; Maxwell et al. 2002). Despite its prominence in newspapers and journal articles about health care, it is not always clear to what people are referring when they use the word "equity". Without question, however, a concern with equity is fundamentally a concern about fairness and is related to but distinct from equality. Sporting events, for example, have winners and losers. The losing side is willing to accept its fate as long as a standard set of rules, fairly and impartially applied, governed the process. In this case, the outcome is unequal, but the contest is equitable. This concern for equity or fairness also embodies an interest in how rights, responsibilities and goods are distributed among individuals in a society. There is an interest in seeing some thing - an object, a service, a process or an outcome - distributed fairly. In the case of sports, this concern starts and stops with fairness in the process; it does not extend to a particular concern with the outcome. One might regard lopsided scores between mis-matched teams as 14 "unfair", but that is usually because one believes something unjust or inequitable led to the mis-match.5 In the case of health care, health is often depicted as the ultimate outcome of interest, and health care as one of the inputs that can contribute to that outcome. A n interest in the distribution of health care exists either in general, in as much as a society decides to take an interest in the processes that govern the distribution of any good or service, or in specific, with special attention to the distribution of health care not because of a particular interest in health care per se, but because of its potential to maintain or improve health status. The next section of this chapter sets out to address two questions - why equity, and why equity in health and health care? These are largely questions of political philosophy, addressing the structure and character of nations, and the rules that govern political and (formal) social interactions. The latter question is more or less prominent in these philosophies, to the extent that each gives special or exceptional status to equity as it applies to health and health care. Once the general frame of reference to equity is established, the third section moves to an examination of the presence and role of considerations of equity in legislation and in policy documents that provide structure for health care systems in Canada generally, and in British Columbia specifically. This history of health care services and health care financing is used to identify the principles of political philosophy that have influenced health care policy in Canada. The fourth section then attempts to identify the dominant equity-related motivations in British Columbia's health care system. This section also provides motivation for the empirical investigations in the following chapters, and their concern with measuring and monitoring equity in need for health care services, health care services use and health care financing. 2.1.1 Methods The section on political philosophy (section two) relied on a review of philosophies identified as having some influence on health-related research. These philosophies were chosen based on a review of academic and grey literature from a variety of disciplines that touch on health, including philosophy, population health, health services research and health economics. The literature search was conducted in three steps: a search through ISI Web of Sciences, a review of bibliographies of relevant papers, and an internet search for grey literature such as self-published reports and other sources outside academic journals. The 2004/2005 National Hockey League season was cancelled because owners were insistent on salary caps tied to revenues from the previous season. They believed this was the only equitable way to distribute access to high-quality players. 15 ISI Web of Sciences was chosen for the major search because it encompasses all of the disciplines of interest, and is relatively comprehensive in its coverage.6 The search used keywords to identify articles published in English from 1990 forward. The search identified 375 articles; a review determined relevance for the present project. Articles were not included in the final set if they were limited in scope (e.g. equity of a particular clinical intervention) or were clearly unrelated to health-care services (e.g. pay equity or equity investments). A review of bibliographies of review articles and those deemed most useful added articles not identified in the initial search. Authors of key papers were entered into the internet search engine GOOGLE in combination with the keyword "equity" to identify sources of grey literature, particularly reports funded by national and international agencies (e.g. the World Health Organization or the Nuffield Trust) Finally, a limited GOOGLE search of equity and health identified a small number of associations (International Society for Equity in Health) and individuals engaged in research in this area. The section on history (section three) identified major pieces of legislation and major policy documents based on a review of well-known accounts of the development of health care in Canada (Maioni 2004; Naylor 1986; Taylor 1978). The primary sources so identified were then reviewed for descriptions of coverage and financing arrangements, as well as for any specific reference to values or principles that specific pieces of legislation or specific policies were trying to achieve. 2.2 P o l i t i c a l p h i l o s o p h y a n d e q u i t y 2.2.1 Why equity? Most of the academic literature that deals with concepts or definitions of equity eventually refers to prominent political philosophers. Perhaps this is because it is of limited help to describe equity as concerned with fairness or justice, since that leaves open the question of how to judge what is fair or just. These are open questions precisely because there is no right answer. Instead, what is deemed just or fair is at least partially culturally constructed, based on history, legal norms and politics. The shape of these various components and their degree of influence shift over time, producing different conceptions of justice or equity across nations and across time within nations. Different philosophies or theories of justice are in perpetual competition for 6 The web site for this service says that more than 8,700 journals are included in its database. See accessed 23 October 2005. Access to this search engine was available through the University of British Columbia's library of electronic resources, available to all U B C faculty, staff and students. 16 preeminence in societies; a more structured understanding of and approach to philosophy provides a rational basis for choice of competing means and competing ends. These competing philosophies are not generally thought of through the perspective of equity, but are instead alternative depictions of how associations of individuals are constructed, how those individuals are intended to interact and what they are trying to achieve. The rules of engagement in a c iv i l society are established with or without explicit debate about or reference to a particular political philosophy. Whether they are explicit or not, the underlying political philosophies carry with them implications for, or different understandings of, the importance of equity (Rawls 2003; Redden 2002). Familiarity with the general principles of a few philosophical traditions provides the context in which the treatment of equity in health and health care develops. What follows is a (necessarily) brief sketch of a handful of political philosophies that have had, and continue to have, influence on political institutions in Canada. The discussion proceeds from those that emphasize liberty (or freedom) as the fundamental value that governs political and formal social relations, to those that temper liberty, to a lesser and greater degrees, with the values of equality and/or equity. Following this outline, the next section takes those general theories and extrapolates them to health and health care more specifically. The intent is to try and assess what each philosophy would say about society's role in influencing the distribution of health, the distribution of the use of health care services, and the distribution of payments to the health care system. Ultimately, one can identify varying influences of each of these philosophies in the policies and ongoing political debates about health and health care, because all have their adherents in any population. A brief groundwork in liberty andfreedoms - Isaiah Berlin Isaiah Berlin makes a distinction between positive and negative liberties or freedoms that can serve as the groundwork for talking about different philosophies. Negative liberty is the freedom to exercise free w i l l , within the constraints necessary to ensure that everyone else enjoys that same freedom. This concept of liberty outlines an expectation that limits on liberty must be minimized so that liberty itself can be maximized. But, Berlin says, "The bulk of humanity has certainly at most times been prepared to sacrifice this to other goals: security, status, prosperity, power, virtue, rewards in the next world; or justice, equality, fraternity, and many other values which appear wholly, or in part, incompatible with the attainment of the greatest degree of individual liberty..." (Berlin 1969, p 161). 17 The pursuit of these other goals, through exerting control over the process of governance, is what Berlin refers to as positive liberty. Rather than seeing a choice between negative and positive liberty, Berlin proposed that the two must operate in balance. Too much emphasis on negative liberty is a sacrifice of potentially competing values, such as equality or happiness, while too much emphasis on positive liberty will encroach directly on individual freedom. Berlin argues that equality is a competing value to liberty that has "universal and perennial appeal" (Berlin 1978, p. 82). This is not a statement that there is unanimity about what equality means or how it can and should be applied, but only that it is omnipresent. He then argues that equality ought to be thought of as the impartial application of rules, with deviation from those rules only in specified, agreed-upon or necessary circumstances.8 In this case equality and equity overlap; the rules must be applied equally, and if they are, one achieves equity, because all individuals have been dealt with in a fair manner. On the other hand, breaking the rules, or applying them unequally in preference of one individual or groups of individuals over another, is a matter of injustice and inequity. Objections to the rules that form the structure of society might come from three different directions. First is a charge that while the rules are fair, they are not impartially applied. This is quite different from a second charge, which is that the rules themselves are inadequate or otherwise inappropriate. The implication of this objection is that while some rules are acceptable, this one in particular either is not doing what it purports to do, or perhaps impinges too much on negative liberty. A third possible charge is simply that there are too many rules. Berlin concludes that: Plu ra l i sm, w i t h the measure of 'negat i ve ' l iber ty that i t enta i l s , seems to me a t ruer and more humane i d e a l than the goals of those who seek in the great , d i s c i p l i n e d , author i ta r ian st ructures the i d e a l of ' pos i t i ve ' se l f -maste ry by c lasses, or peoples , or the who le of m a n k i n d . It is t ruer , because i t does , at least , recognize the fac t that human goals are many , not a l l of t h e m c o m m e n s u r a b l e , and in perpetua l r ivalry w i t h one another . To assume that a l l values can be graded on one sca le , so that i t is a mere m a t t e r of inspect ion to d e t e r m i n e the highest , seems to me to fa ls i fy our knowledge that men are f ree agents , to represent mora l dec is ion as an operat ion wh ich a s l ide - ru le c o u l d , in p r inc ip le , p e r f o r m . (Berl in 1969, p. 171) 7 Berlin (1969) provides a lengthy critique of positive liberty, arguing that it could be used by rationalists to impose laws and governance structures that reflect what "rational individuals ought to want" rather than what the public at any given time actually desires. He appears to equate extreme versions of this with totalitarian regimes of the mid-20'h century. 8 Berlin recognized the fact that this sort of statement could be pushed to the point that it is completely meaningless. He nonetheless identifies rules and equal application of rules as one of the "deepest" needs of mankind; creation of and adherence to rules is what provides social order. 18 Berlin also believed that the weighting of values is likely to shift over time. This does not lessen the ability of individuals or societies to act based on values, but does caution against the assumption that once settled, structures built on those values will remain appropriate or acceptable. Equity is thus a moving target, not prescribed by any particular set of constraints, but negotiated while balancing the advantages and disadvantages of both negative and positive liberty. Entitlement theory / libertarianism - Robert Nozick In Anarchy, State and Utopia, Robert Nozick sets out the idea of a minimal state that identifies individual liberty as the central guiding value (Nozick 1974). Nozick begins with the idea that individuals are sovereign over themselves. What individuals owe each other is respect for this basic liberty. The minimal state is justified only because it is possible to construct a theoretical and "moral" progression from independent individuals to that state. Among a collection of individuals, living in geographic proximity, some conflict over scarce resources will inevitably arise, with charges of damage, theft or some other transgression. Individuals have a right to defend their liberty both individually and collectively. The collective expression of this is the formation of "protective associations", variations on insurance agencies that provide intermediary and collection services in exchange for a subscription fee, will develop in response. These protective agencies will evolve over time into larger associations, with one becoming dominant in each geographic area. This evolution is inevitable because one protective agency will emerge with more clients than the others, and will use that market power to assert control over the procedures used to sort out claims of wrongdoing and counter-claims of innocence. Once this dominance is achieved, that protective agency operates effectively as a "night watchman" in an ultra-minimal state. Nozick asserts that the process of moving from a collection of individuals to this minimal state violate no one's rights and therefore is legitimate. There is then a transition from this ultra-minimal state to a minimal state that is characterized by introducing an element of redistribution to the protective scheme. In this case it is recognized there will be individual independents in the dominant protective service's area of service who choose not to pay for protective services. The dominant agency is nevertheless forced to deal with the independents to the extent that the agency's clients make claims of loss as the result of action by the independents. The dominant agency will not allow the independents to apply their own procedural rules for determining guilt and relevant compensation, and must compensate the independents for this. Extending regular protective services to these independents would prove cheaper than providing this compensation, so there is redistribution in the sense the independents 19 are provided with de facto coverage at no cost, while regular clients of the protective agency continue to pay for those same services.9 Nozick goes on to argue that protecting individual rights means these individuals cannot be forced to bear costs aimed at contributing to a greater social good. Society is made up of individuals, so any "greater good" in society must have come at some cost to individuals, who themselves had no gain as a result. "He does not get some overbalancing good from his sacrifice, and no one is entitled to force this upon him - least of all a state or government that claims his allegiance (as other individuals do not) and that therefore scrupulously must be neutral between its citizens" (Nozick 1974, p. 33, emphasis in original). The key is neutrality, with Nozick's interpretation that "neutrality" requires the state not to intervene. The importance of neutrality extends also to the treatment of resources available to people. Individuals' wealth and other resources may not be distributed equally, but i f they were acquired on terms that did not harm others, then there is no injustice and the distribution must be considered equitable. This includes the transfer of wealth from one generation to another. Some people will have less and others will have more, but that is no business of the state, except to the extent that one claims wrongdoing. So the fact that we partially are "social products" in that we benefit from current patterns and forms created by the multitudinous actions of a long string of long-forgotten people, forms which include institutions, ways of doing things, and language ... does not create in us a general floating debt which the current society can collect and use as it will. (Nozick 1974 p. 95) In this social arrangement, the concept of equity is tied directly to liberty. A state that is equitable treats all individuals the same with respect to the protection of those liberties, regardless of their incomes or any other characteristic. Utilitarianism - John Stuart Mill The social construction that Nozick articulates features liberty as its centrepiece. Liberty trumps all other considerations; in fact, there are no other considerations that matter in any real way. In contrast to this, other political philosophies take for granted that there can be legitimate social arrangements that place some constraints on liberty beyond the minimal ones that Nozick allows. If the possibility of these additional constraints is taken as given, then the struggle of the 9 In economic terms, this is a free-rider problem, in which independents end up with equivalent coverage at lower prices than regular clients. Nozick argues that there will be a limited number of free riders, because the dominant protective agency will only provide protective services to its clients. In other words, as there are more independents, there is more possibility of attack by an individual not covered by the dominant protective agency. This is incentive for people to purchase protective services. 20 philosophy becomes identifying what those constraints are and what purpose(s) they serve. In essence, the question is how to define and then achieve an acceptable distribution of resources among individuals in a society. One of the traditional philosophies of evaluating how society distributes its resources is utilitarianism. The roots of utilitarianism, now associated with the discipline of economics, are in the mid-19 t h century writings of Jeremy Bentham and John Stuart M i l l . 1 0 The defining objective of utilitarianism is maximizing utility, which M i l l equated with happiness (Mill 1861). According to M i l l , things that are desired, and thus have utility, are those that bring happiness, or conversely, those things that help avoid pain or unhappiness. M i l l was careful, however, to qualify that his conception of "pleasure" was broad, not limited to physical sensation. Humans are, he believed, distinguished from animals precisely because of this more expansive ability to derive pleasure from experience, such as the experience of reading a good book or engaging in an invigorating conversation. His belief was that individuals have self-interest, but that self-interest should not triumph in cases where it causes harm to others. In fact, he believed that once people have the opportunity to develop their "higher level" intellectual and "noble" characteristics, they will develop a strong feeling of inter-connectedness such that an increase in another's happiness is one's goal as much as is increasing one's own happiness. That is, moral agents are such that they should show no special preference to themselves or to any particular set of persons, but should have an equal concern for all who can be affected. M i l l acknowledged differences among individuals, but believed that those differences, at least with respect to their ultimate goals in life, would narrow as societies continued to develop. This version of utilitarianism seems to allow some room for introducing distributions as a concern. The trouble with Mil l ' s version, however, is that it only provides an opening for this concern by making the assumption that //people were allowed or encouraged to develop intellectually then surely they would come out seeking more happiness for everyone. In other words, utilitarianism is what a rational agent will select on reflection or with the development of his or her more noble characteristics. In the absence of fully developed rational beings, society would depend on Berlin's notion of positive liberty to enforce this behaviour, and thus returns to Berlin's concerns about the careful balance between positive and negative liberty. 1 0 This section should not be interpreted as representing a history of utilitarian thought. Utilitarianism has a long and varied history, with several competing branches encompassing a correspondingly broad array of ethical principles. Presented here are Mill's views on utilitarianism, chosen because they were an early form that recognized some of the difficulties of trade-offs in utilitarian thinking, with specific attention to interpersonal comparisons and the role of justice in society. 21 Class conflict - Karl Marx The description of ongoing tension is a feature of all political philosophies, which after all are theories about how power and rights, including rights over decision-making, get allocated in society. The nature of political philosophy as a conflict of interests is the centerpiece of Marx's depiction of the relationship of capital and labour, bourgeoisie and proletariat." Marx is, of course, an important figure because of his development of a theoretical model that predicted the end of capitalism. He is included here because of his normative focus on what society would look like in its post-capitalistic form. Marx describes class conflict as the major and repeating theme in the evolution of social organization. In the capitalist and newly industrialized societies of the mid-19 t h century, this conflict was about control of the means of production, which were key to generating and then maintaining access to economic power. Marx believed that capitalist society required a fundamental reorganization that would give workers co-operative control over the means of production; according to Marx, this change was inevitable as workers gained a better understanding of their position in society and joined forces to change it. In the first instance Marx was not concerned with distributions of outcomes as much as he was concerned with workers controlling the capital or infrastructure that adds value to raw material. Gaining this control was the first and crucial step in transforming society, and it was necessary because the new society had to develop out of an entrenched capitalist arrangement. After the transition to a new society in which workers have this control, he recognized the inevitable challenge of balancing the equal call on the "products of labour" and the unequal ability to provide labour that is inherent in different men. A more fully realized version of communism - the next step - would follow. In a higher phase of communist society, after the enslaving subordination of the individual to the division of labor, and therewith also the antithesis between mental and physical labor, has vanished; after labor has become not only a means of life but life's prime want; after the productive forces have also increased with the all-around development of the individual, and all the springs of co-operative wealth flow more abundantly --only then can the narrow horizon of bourgeois right be crossed in its entirety and society inscribe on its banners: From each according to his ability, to each according to his needs!12 '' The material for this section was drawn from the writings of Marx, as archived on Accessed 25 October 2005. 1 2 From Part I of "Critique of the Gotha Programme", .htm. Accessed 25 October 2005 22 Like Nozick, Marx offers a fully-formed vision of society and relationships among men; like M i l l , Marx bases his vision of society on an optimistic view of human nature and how that nature is likely to develop given the opportunity. In his idealized version of communist society, Marx saw the production and distribution of goods as a societal function. Using Berlin's framework, Marx gave primary importance to positive liberty as achieved through co-operative action. In this case, equity is the contribution of labour according to ability, and receipt of the products of labour according to need, all based on the complete elimination of classes within society. Justice as Fairness - John Rawls While Marx's theory is based on the end of capitalism, John Rawls develops a theory of social interaction that is built solidly within western welfare-state capitalism. In addition to providing a theory of justice, Rawls also provides a rationale for arriving at the set of rules that underlie it (Rawls 2003). Rawls' main idea was that social policies or social rules should be set to be consistent with those that would be decided upon by individuals if they were in an "original position", behind a "veil of ignorance." In other words, the rules and structures of society should be those articulated by individuals who do not possess special knowledge about their position in society, or about their attributes, skills or goals in life. Without any special or specific knowledge, the structures or rules chosen would be the ones deemed most fair, regardless of an individual's eventual position in society. Rawls believed that what would emerge from such a hypothetical exercise would be a society that worked to achieve two fundamental principles13: 1. Each person is to have an equa l right to the most extens ive t o t a l system of equa l basic l iber t ies c o m p a t i b l e w i t h a s imi la r system of l iber ty for a l l . 2 . Socia l and e c o n o m i c inequal i t ies are to be arranged so that they are both : a . To the greatest benef i t of the least advantaged , consistent w i t h the just savings p r inc ip le , and b. A t t a c h e d to of f ices and posit ions open to a l l under condi t ions of fa i r equa l i t y of oppor tun i ty . (Rawls 2003, p. 266) The first principle is the familiar notion of maximized negative liberty. The first part of the second principle, the "difference principle", is based on the idea of Pareto efficiency in economics and is tied to changes in expectations of members of a society. The idea is that 1 3 The veil of ignorance is a neat rhetorical device, but Rawls is firmly in the camp of the rationalists, assuming that all rational individuals would come to precisely the same conclusion once behind the veil. In other words, the principles that come out are not those that are reached by a majority of individuals, but by one representative individual. 23 increasing expectations of the better off are only acceptable i f the rising tide that results will lift the boat of the most disadvantaged as well. In other words, someone aspiring to be an entrepreneur should have higher expectations of income than an unskilled labourer only if what the entrepreneur contributes to society will improve the prospects of the unskilled labourer as well. "Their better prospects act as incentives so that the economic process is more efficient, innovation proceeds at a faster pace, and so on" (Rawls 2003, p. 68). 1 4 Part b of the second principle, requiring equality of opportunity, is one of the possible justifications Rawls offers for instituting taxes and other means of constraining the transfer of wealth from one generation to another. Income inequality that is too great will lead to inequalities in political freedoms, such as unequal access to offices and positions, and therefore must be checked. Rawls also develops the idea of primary social goods, or basic things that people need in order to achieve their life goals. He defines the primary social goods as basic liberties, fair equality of opportunity, the rights and prerogatives of authority, income and wealth, and the social bases of self-respect. The appropriate distribution of these primary social goods is determined by the difference principle, subject to a social minimum that is in turn consistent with the principles appropriate savings (necessary for future generations) and maintenance of equal liberties. In this sense there is not an absolute social minimum that can be identified and provided across all societies. Instead, the social minimum will be determined by rules of society, including the application of the difference principle, which ensures that changes to distributions have to work to the advantage of the least well off. 2.2,1.6 Sovereign virtue - Ronald Dworkin Rawls describes what Ronald Dworkin calls an interest-based strategy for defining the role of liberty and equality in society. As an alternative, Dworkin develops what he calls a constitutive strategy, a strategy that identifies both liberty and equality as essential elements in the structure of society. "The constitutive strategy...builds liberty into the structure of its chosen conception of equality from the start. It insists that liberty must figure in the very definition of an ideal distribution, so that, for that reason, there can be no problem of reconciling liberty and equality" (Dworkin 2000, p. 135). Dworkin recognizes, like Berlin, that there are competing and conflicting values, of which liberty is but one. 1 4 Immediately following this, Rawls adds "I shall not consider how far these things are true. The point is that something of this kind must be argued if these inequalities are to satisfy the difference principle." 24 The quest ion raised by any such law is not whether i t at tacks l iber ty , wh ich i t does , but whether the a t tack is jus t i f ied by some c o m p e t i n g va lue , l ike equa l i t y or safety or publ ic amen i t y . If a soc ia l phi losopher places a very high va lue on l iber ty as l i cense , he may be understood as arguing for a lower re lat ive va lue for these c o m p e t i n g va lues . If he defends f r e e d o m of s p e e c h , for e x a m p l e , by some genera l argument in favor of l i cense , then his argument also supports , at least p r o tanto, f r e e d o m to f o r m monopol ies or smash storef ront w indows . (Dworkin 1974) While liberty is built in, it is equality that is the "sovereign virtue" of any society, in the sense that the one thing above all else that a government must provide is equal concern for all citizens. Dworkin, unlike Nozick, is concerned with equality of resources. Dworkin does not believe that personal resources like natural talents or abilities can be equalized, but that impersonal resources such as income and wealth can be made more equal. This is not, however, an argument for equality of income. Instead, the primary objective is having resources at an individual's disposal, or under an individual's control, to a sufficient degree that the person can set his or her life's goals and have some chance of achieving them. This does not require equality of income because individuals have different tastes and preferences, for work and leisure for example, that will make them willing to trade off income against other objectives. Dworkin is interested in equal resources as a means for individuals to exercise their liberty in pursuing these different preferences.15 If governments show equal concern for all individuals of society in this way, then it has achieved equity. Capabilities andfunctionings — Amartya Sen The view that the role of government is to ensure the possibility of equal participation by all members of society is consistent with Sen's theory of justice developed around his ideas of capabilities and functionings (Sen 1992).16 In Sen's view, "functionings" are the things that people can achieve in life, such as being well-nourished, avoiding illness, or having a sense of self-respect. Any individual will be able to achieve a series of functionings, all adding up to overall well-being. "Capability" is then each individual's ability to choose among different 1 5 Dworkin provides an elaborate thought experiment about an auction that starts with an equal distribution of clam shells to show that equality of resources is not the same as an equal distribution of income, or anything else for that matter. A further thought experiment about insurance is introduced to deal with the bad luck and other circumstances that are inevitable after this initial auction. 16Dworkin says that, in effect, his theory and his interpretation of Sen's say the same thing using slightly different language. 25 combinations of functionings, ".. .reflecting a person's freedom to lead one type of life or another" (Sen 1992, p. 40). Sen argues that health status in specific has to be a societal concern, because health is a fundamental necessity for exercising other freedoms and maximizing capabilities (Sen 2002). If people are in poor health, their choices are limited, their ability to fulfill their human potential is curtailed, and therefore they have been denied full participation in society. This is not an argument about "human rights" at a basic legalistic level, but is instead an argument about social citizenship, and what it means in the 21 s t century to be a citizen and have the capabilities to exercise all the rights and freedoms that such citizenship entails. If equity as a concept speaks to the fair or impartial application of a set of rules, then all political theories are concerned with equity, in the sense that each identifies the rules required to govern interactions among individuals. The theories are distinguished by what it is that the rules are trying to achieve, what they are trying to make equitable. Nozick believes social rules should be constructed to achieve maximal negative liberty, and M i l l a maximum of pleasure broadly defined. In contrast, Marx would use social rules to ensure contributions from citizens according to their ability, and the distribution of societal resources according to need. Rawls believes a society is just if it operates using the two principles of justice, while Dworkin would equalize resources and Sen would concentrate on equalizing capabilities. These different philosophies have different implications for health and health care policy. It is to these differences we now turn. 2.2.2 How does equity apply to health and health care? Sen is explicit in his treatment of health in his philosophy, because of the connection between health and the ability to achieve one's life goals. His approach focuses on health rather than giving a central position to health care (Table 2). Instead of arguing that access to health care is a fundamental right, the focus is on health care as a means by which individuals can achieve health, with the understanding that it is only in a state of health that individuals can fully exercise their (fundamental) rights of citizenship and participation in society (see also, for example, Culyer 2001). This is a view of health care as an instrumental or intermediate good, or as a means to an end, rather than as an end in itself (Culyer and Wagstaff 1993; Evans 1984). Given the emphasis on equalizing health, a society adhering to Sen's philosophy would be likely to provide universal and comprehensive insurance for health care services. The use of health care services would be tied to need (or some other metric that tries to minimize 26 differences in health status). Because of an emphasis on equality overall, it is also likely that such a society would finance its health care services from progressive taxes, taking an increasing income share with increasing income. Table 2: Theories of social organization and application to health and health care Equality of what? Distribution of health? Distribution of health care services? Distribution of health care financing? Entitlement / Libertarian (Nozick) Liberty. Cannot be constrained, except by natural endowments, which are not unjust. No state intervention; the bad luck of i l l health is borne by each individual. No state intervention. Either through direct purchase of services or through voluntary insurance schemes. The latter likely only where probabilities of illness are substantially <1 and premia are risk-rated. No state intervention. Payments to the health care system tied to the use of services and/or the payment of risk-rated insurance premia. Utilitarian (Mill) Rather than trying to equalize something, utilitarianism is concerned with maximizing overall utility, or pleasure and the absence of pain broadly defined. Maximize total utility. This will lead to a desire to improve population health status, to the extent that health is an important component of utility. Assuming that health care services do not carry "utility" except as a means to an end, services to be provided in a manner consistent with the goal of maximizing total utility. Likely to be compulsory insurance, at least catastrophic coverage. Class conflict (Marx) Control over the means of production. Equal. Receipt of services according to need. Contributions to health care financing according to ability to pay. Justice as fairness (Rawls / Daniels) Inequalities only where they work to the advantage of the most disadvantaged. Equal, as far as possible, in order to promote fair equality of opportunity. Receipt of services according to need. The burden of financing should not fall disproportionately on the i l l . Equality of resources (Dworkin) Impersonal resources Wil l work to equalize health, to the extent that it is a resource required to achieve one's goals Services covered in the prudent insurance package provided according to need. Community-rated premia to cover a prudent insurance package, based on what people would choose to insure, assuming equality of resources and perfect information. Functionings and capabilities (Sen) Capabilities Equal. Receipt of services according to need. Contributions to health care financing according to ability to pay. 27 A society adhering to Marx's philosophy would likely come out at the same place as Sen's, albeit through different reasoning. Rawls' theory of justice also leads in this direction, at least as it has been applied to health care by Norman Daniels (Caplan et al. 1999; Daniels 1985; Daniels 2001; Daniels 2002). Daniels takes Rawls' notion of primary social goods and adds health (not health care) as one of the things necessary to ensure fair equality of opportunity.17 Daniels focuses on health care as a means of either maintaining individuals at or returning them to "normal species functioning", or as close to that as one can get. Daniels further combines this commitment to health with an understanding that there are many things in addition to the health care system that can influence health status. So health care is "special" and must be distributed according to need because of its relationship to health, but the same principle of equality applies to the other determinants of health as well, for example education and income. Daniels is thus similar to Sen in his support for equality in a general sense, but is less direct on the financing side, stating only that "the burden of financing should not fall disproportionately on the i l l " (Daniels 1985, p.4). Dworkin takes a different approach in a short chapter of his book devoted to health care, wherein he suggests a prudent insurance scheme. This insurance scheme would provide a package of services the majority of people would choose to insure if they had the means available, if they had the same knowledge about costs and benefits of services as a "good doctor", and i f insurance premiums were community- rather than risk-rated.18 Dworkin's thought experiment - for he is explicit that it takes some imagination to make decisions as i f these three criteria are met - has some similarity to Rawls' veil of ignorance. A major difference, however, is that Dworkin does not assume that everyone would come to the same conclusion about services covered by the prudent insurance package. He acknowledges that there would be ongoing challenges around the fringes of medical care delivery, for example in the application of high technology care at the end of life. The insurance package might thus be adjusted over time. The overriding principle, however, remains the same; it is society's responsibility to equalize resources, and in health care this can be accomplished through the provision of insurance. The implications are that services would be provided based on need, with all individuals contributing 1 7 Daniels has expanded his views considerably since the publication of Just Health Care in 1985. In that volume, he was not prescriptive in terms of what a "just" health care system would provide or how it would be financed. He allowed, for example, that health care systems might be tiered, with a basic minimum for everyone, and an ability to purchase services beyond that based on ability and willingness to pay. This idea is absent in more recent writings, and in particular when he extends Rawls' theory in recognition of the range of determinants of health, where he begins to sound very much more like Sen. 1 8 Community-rated insurance premiums are applied equally to all insurees, while risk-rated premiums are adjusted based on past health care use and/or personal characteristics that, statistically speaking, increase the probability of future use of health care services. 28 equal payment to the health care system, in accordance with community rating of insurance premia. It is more difficult to apply Mil l ' s version of utilitarianism to health care, in no small measure because utilitarianism is focused on the end of maximized utility, so the importance of the distribution of health, for example, depends on the relative importance of health in overall measures of utility. If health is a major contributor to overall utility, then a more equal distribution of health might be consistent with greater utility when aggregated across all individuals in a population. M i l l did not advocate equalizing utility across individuals (though this could be one possible outcome) so it is not likely that his version of utilitarianism would advocate equalizing health status, in line with Sen, Marx and Daniels. It is possible, however, that Mil l ' s philosophy would result in the distribution of health care services based on need, as long as health care services did not themselves provide any direct utility. If health care services are viewed as a means to an end rather than an end in themselves, then the greatest utility of health care services would accrue to individuals with the greatest need (as long as need is measured in a way that implies need for health care services rather than a more general measure of i l l health). In Mil l ' s outward-looking and altruistic version of utilitarianism, this ranking of utility would determine the actual use of those services. It is even more difficult to say how financing would be arranged in a M i l l utilitarian society, though to the extent that he assumes decreasing marginal utility with increasing wealth, he would be in favour of some form of compulsory insurance, at least for catastrophic costs. He would, in other words, wish to protect individuals from large expenditures for health care, because they would have such a detrimental effect on utility. Finally, entitlement theory as articulated by Nozick would seem to leave little room for concern about the distribution of health per se. In this philosophy, an individual's i l l health is a matter of bad luck, and that bad luck does not imply any responsibility on the part of the minimal state. Nor would it be acceptable to set up a compulsory health insurance system to share the costs of providing health-care services. Any universal scheme for insurance or collective care, unless it was mutually agreed upon by every member of the state, would be coercion and as such an unacceptable attack on liberty.19 Instead, each individual would be left to decide how best to allocate whatever resources were available to him or her to health-care services as against all the other goods and services available for purchase or barter. This could potentially lead to the 1 9 Nozick does not rule out philanthropy, but any such giving has to be free from any real or perceived compulsion. 29 development of a market for health insurance as a way of pooling risks, as long as each individual was given full choice over whether to join the pool. Insurance premiums would be risk-rated, meaning adjusted based on expected need for services. Because of this, such insurance would likely only be available where the probability of illness was low. 2 0 In a society governed by this philosophy, health care services would be provided based on ability to pay, or ability to purchase insurance where it was available. The distribution of payments for health care services, even in the presence of insurance, would likely be strongly and positively correlated with ill health. These different philosophies place very different emphases on the distribution of health and health care. Some declare an interest in the distribution of health status as part of a general statement about how all resources in society should be distributed. Others do not necessarily call for equalizing health status, but still claim that health care is "special" and thus its distribution is a concern of the state, even where such concern is not shown to other goods and services.21 In general, it is easier to extrapolate the general philosophies to concerns about health care services utilization than to health care financing, except in those cases where general equality is given greater weight. 2.3 What is the place of equity in health care in Canada? The survey of political philosophy provides a brief introduction to the conditions or system of social "beliefs" under which health care may be given special status. The next logical question is, where does Canada (or British Columbia) fit along this spectrum? This question is addressed by taking an historical perspective. Major policy documents and legislation, for Canada in general and for British Columbia in specific, can provide evidence of the equity-related ambitions of the architects of health care systems. This review starts in the early part of the 20 t h century and proceeds to the reports of the Romanow Commission and Kirby Committee, as outlined in Figure 2.1. The long view helps to establish any patterns, i f they exist, in the commitment of governments in Canada to principles of equity in the stated purpose and legislative underpinnings of health care systems. It will also help identify any changes, especially " Community-rated systems that are not universal or that are not heavily regulated will tend toward risk-rating, as that is a way to split the market and manage risk (and make profits). (See, for example, Fein 1986) 2 1 This special status is not peculiar to health. Political liberty and access to justice, for example, are also governed by special provisions because they are not commodities per se, but instead are preconditions for full participation in society, including the enjoyment of other sorts of commodities (Evans 1984; Rawls 2003). 30 more recent changes, in those patterns, and in the way that ideas about equity are incorporated into the financing and delivery of health care services. Figure 1 : A policy and legislative time-line for Canada and British Columbia British Columbia Royal Commissions on State Health Insurance Medical Care Act Hospital Insurance and Diagnostic Services Act British Columbia Royal Commission on Health Care and Costs Canada Health Act Three federal reports on health care 1 9 1 0 1 9 2 0 1 9 3 0 1 9 4 0 Rural Municipality Act and development of municipal doctor and hospital plans 1 9 5 0 1 9 6 0 Federal government post-war planning documents 1 9 7 0 The Hall Commission 1 9 8 0 1 9 9 0 British Columbia Medicare Protection Act Established Programs Financing Act 2 0 0 0 Canada Health and Social Transfer Despite the notion of a "system" of health care in Canada, the reality is that there are 14 provincial, territorial and federal systems linked by a set of common principles or minimal standards established at the federal level. The empirical work of this thesis focuses solely on physician and hospital services, which are the only services currently covered under federal legislation. It is those services that are the focus here as well. There is certainly a legislative and policy history with respect to equity for other sectors of the health care system, and for provinces and territories other than British Columbia, but those are not covered in the present work. 2.3.1 The first decades of the 20th century The financing and delivery of health care services for the majority of the population in Canada is the direct responsibility of provincial and territorial governments. The federal government has a direct delivery role in the case of a few special population groups including the armed forces, the Royal Canadian Mounted Police and status Indians. Beyond these groups, the role of the federal government is one of transferring funds to the provinces and territories, 31 providing leadership (with the provinces and territories) in setting overall policy and enforcing the Canada Health Act. The latter includes the ability of the federal government to withhold funds if the provinces do not enforce the principles articulated in the Act. It was the hospital insurance plan established in Saskatchewan in 1947 that became the model for all other hospital plans and later for the development of physician insurance. These plans formally established the form of comprehensive and universal, tax-financed care that remains the basis for services covered under the Canada Health Act. The groundwork for what was to come was laid in the years leading up to and following the depression of the 1930s. The Rural Municipality Act, passed by the Saskatchewan Legislature in 1916, allowed rural municipalities to levy property taxes to collect funds that could be used for hiring a physician on a retainer basis (Gelber 1980). These Municipal Doctors Plans transferred aspects of cooperative action common in rural Saskatchewan to the health field, and made possible the presence of physicians in rural areas where it would have been difficult (if not impossible) to retain doctors reliant on fee for service payment, especially during the depression.22 The same principles were applied to arrangements for hospital services, where funds were raised through a combination of property taxes and premiums for non-property owners. By 1939 there were 100 such municipal hospital prepayment plans, and by 1948 there were "107 municipalities, 59 villages, and 14 towns [holding] contracts with 180 doctors" (Taylor 1980, p. 184). This basic system structure spread to some municipalities in Alberta and Manitoba through the 1920s and 1930s, but remained most prominent in Saskatchewan (Gelber 1980). These early initiatives provided universal eligibility for access to a physician as determined by residence in a particular region. But the rationale for this arrangement appears to have been pragmatic more than ideological. Rural municipalities felt threatened by the potential departure of physicians, and the only recourse that appeared to be available was collective action by which each resident contributed a small sum that, aggregated across residents, added up to an acceptable, stable income for the contracted physician.23 Everyone contributed, but there is no evidence of discussion about adjusting payments based on income. The equity concept, in this case only implied, was equal access to services based on equal pre-payment. There was little or Maioni (2004) notes that the depression in Canada was of a different character than that of the United States. There were no bank failures in Canada. The problem for Saskatchewan, however, was the coincidence of a drought with a general downturn in other employment opportunities. The drought was disastrous for this one-crop economy. It is also worth noting that more current concerns about rostering patients and ensuring that the physician with whom they have a contract is the one from whom they receive services were not an issue n rural Saskatchewan in the 1920s and 1930s. Rural communities were relatively sequestered, and travel times great enough that physicians really could serve geographic catchments. 32 no opposition to these early initiatives, perhaps because for this brief period the interests of all parties pointed in roughly the same direction (Maioni 2004). 2.3.2 British Columbia Royal Commissions on Health Insurance Industrial and social relations were strained in British Columbia following the end of the First World War. In addition to war casualties and the needs of returning servicemen, there was the flu pandemic in 1918, and then the Winnipeg general strike in 1919 and sympathy strikes in Vancouver, Victoria and Prince Rupert (Naylor 1986). Health was a particular concern, in part because the war provided hard evidence of the poor health status of a large proportion of the population. For example, "of the 18,169 men who were registered in British Columbia under the Military Service Act, only 9,609 were found to be in sufficiently good physical condition for service in the firing line, and 4,075 were found to be unfit for military services of any kind" (British Columbia.Health Insurance Commission 1921, p. 44). One response of the British Columbia government was to appoint a commission to look into public insurance schemes. Information provided in the report of that commission makes it easy to see why health care services were of particular concern. A worker in the wooden shipbuilding industry earned an average weekly wage of $5.52, while the average for workers in "saw milling yards and delivery" was $3.92. At the same time, a consultation with a physician was $2.50, with additional charges for visits between 8pm and 8am and a $l-$2 premium per mile traveled for calls to the "country". One visit for the treatment of highly infectious diseases such as smallpox ran $5-$ 10, and a fracture cost $50-$ 150 depending on the requirements for care. In other words, a single visit with a physician might be affordable for an average worker, costing (only) a half a week's wages, but a fracture might produce a bill for more than six months of total earnings (British Columbia.Health Insurance Commission 1921).24 The Commission report tabled ten recommendations in total, two of which were: That a p rov inc ia l -w ide compulsory Heal th Insurance measure be enac ted and made app l i cab le to a l l wage -earners under the age of 65 years who have been bona f i d e residents of Br i t ish C o l u m b i a for at least 18 months , and whose incomes are not in excess of $3000 per year , such insurance to inc lude the fami l ies and other dependents of the insured. . . . That vo luntary hea l th insurance be prov ided for a l l wage -earners in excess of $3000 and for a l l others i r respect i ve of the 2 4 The average weekly wage rates quoted in this report were derived from the Workers' Compensation Board of British Columbia. The physician fees were from the schedule of fees adopted by the Vancouver Medical Association. There is no way of knowing whether these fees were actually applied in practice, but the fact that they were adopted by the association suggests that they must have reflected at least what the profession believed the prevailing charges should be. 33 amount of the i r i ncomes . (Brit ish C o l u m b i a . H e a l t h Insurance Commiss ion 1921, pp . 107-108) A combination of compulsory and voluntary insurance was recommended because the Commission's review of insurance schemes in other jurisdictions showed them that " A purely voluntary system does not reach those who most need it; its overhead charges are necessarily larger; compulsory contributions from employers, (to a voluntary system), are impracticable; and it does not admit of free choice of doctors, nor of exemption from medical examination" (British Columbia.Health Insurance Commission 1921). The costs of providing the insurance were to be shared, 2/5 by employer, 2/5 by employee and 1/5 by the state. The rationale was that all three of these parties would benefit from the presence of insurance, so all should contribute to the financing. The commitment of the Commission to insurance is even clearer in the following: Inasmuch as heal th is a basic asset of the n a t i o n , i t should not be used as a c o m m o d i t y fo r the making of prof i ts by so - ca l l ed sickness insurance compan ies . No more should sickness be a p r o f i t - m a k i n g business than the po l ic ing of our Prov ince . The main tenance of heal th should be a funct ion of good G o v e r n m e n t . If i t is a func t ion of good G o v e r n m e n t , i t should not be a leg i t imate f i e ld for a p r o f i t - m a k i n g business. This v iewpo in t has now been a c c e p t e d by six out of the nine Canadian Prov inces , in the case of Industrial acc idents . It should even to a greater ex tent be a c c e p t e d in the case of s ickness. (Brit ish C o l u m b i a . H e a l t h Insurance Commiss ion 1921, p. 8 5 ) 2 5 The Commission believed it had the support of the public for its recommendations. "In the evidence submitted to the Commission the principle of Health Insurance was endorsed by fifty-two women's organizations, twenty-seven fraternal societies...forty labor organizations of various kinds, and thirty-nine individuals speaking on their behalf (British Columbia.Health Insurance Commission 1921, p. 71). The Commission also noted that medical associations that presented did not oppose the idea of health insurance. The only formal opposition came from the Insurance Federation of British Columbia. From an equity perspective, there was a clear commitment to the improvement of health of all members of society, with individual monetary contributions pegged at a proportion of total costs. Despite the public support, however, Naylor (1986) notes that the report ".. .was never officially acknowledged, never published, and in fact remained a rather shadowy document for This position is perhaps not surprising given that the lead commissioner, E.S.H. Winn, was at the time the head of the Workers' Compensation Board of British Columbia. The WCB was established in 1917 as an "historic compromise" through which employers traded their ability to sue employers for work-related injuries for a guarantee of compensation for their injury and for wages lost. 34 some years" (p. 44). No action was taken, but the issue did not disappear, and at the end of the decade a liberal backbencher in the British Columbia legislature was successful in pushing for government action. This time it was a Royal Commission on State Health Insurance and Maternity Benefits. The Royal Commission produced an interim report in 1930 that said: Our invest igat ions thus far conv ince us that there is jus t i f i ca t ion and a genera l d e m a n d for the in t roduct ion in Brit ish C o l u m b i a of an e c o n o m i c a l l y sound and e q u i t a b l e p u b l i c - h e a l t h insurance p l a n , in the interests of the major i t y of Prov inc ia l workers , of Prov inc ia l industr ies , and of the State - in the more e f f e c t u a l safeguarding and preservat ion of c o m m u n a l h e a l t h , the more rat ional d is t r ibut ion of sickness costs , and the sc ient i f i c reduct ion of such charges to the Government , to employers , and to ind iv idua l c i t i zens . (Province of Brit ish C o l u m b i a 1930, p. Z8) The final report of the commission, similar to its predecessor, made a recommendation for compulsory state insurance for wage-earners with up to $2,400 annual income. Their estimates were that under these conditions insurance would extend to about one-quarter of the provincial population. The report acknowledges the "mental worry" about the ability to pay medical or hospital bills and its potential influence on an individual's health status. The particular form of the insurance scheme appears to be related to efforts of physicians to split the medical market. With insurance coverage to a maximum income ceiling, physicians would be assured of payment from government for lower income patients, but would still be able to charge higher income patients at their discretion (Naylor 1986). The Commission's view on financing their recommended health insurance system was expressed as: We have now for cons iderat ion whether cont r ibut ions f r o m the insured person should be lev ied on a percentage of i n c o m e or at a f la t ra te . If the benef i ts were d is t r ibuted in propor t ion to the amount c o n t r i b u t e d , there wou ld be some reason for levy ing contr ibut ions on a s l id ing scale accord ing to i n c o m e . But inasmuch as the benef i ts under our proposals are the same for the low as for the high w a g e - e a r n e r , the f la t rate of cont r ibut ion wou ld seem to us to be the equ i tab le one . This is the system in fo rce under the Brit ish scheme and w e r e c o m m e n d its adopt ion for Br i t ish C o l u m b i a . (British C o l u m b i a Royal Commiss ion on State Heal th Insurance and Matern i ty Benef i ts 1932, p. 31) These two royal commissions in British Columbia offer a consistent view of equity. They clearly argue that some form of insurance was required in order to improve access to individuals and families with limited financial means. This can be seen as a drive to create (some) equity in the use of health care services, at least to the extent that the cost of services impeded their use. 35 Equity, in the view of these commissions, was best (or at least most pragmatically) achieved through insurance that was compulsory to some specified income level and voluntary thereafter. This recommendation was in part a reflection of the review of international experience, but also aligned with physician interests, which were coalescing around a desire to be able to "split" the medical market and through that to maximize their incomes (Naylor 1986). This early version of equity therefore did not have a universal component; the poor and lower income individuals and families were viewed as the responsibility of the state, and higher income individuals and families would be left with an option to purchase insurance or to forego it. Payments for insurance, and thereby contributions to health care financing, were proposed as flat-rate payments, with each individual contributing the same amount. There was no progressivity in financing built into the system, but this may be because the proposed insurance was designed to cover a limited spectrum of incomes. Within a limited income range, equal payments are essentially the same as proportionate payments. It appears that the over-riding concern in organizing financing was risk-sharing rather than income redistribution. The health care system, organized this way, would have a significant redistributive effect from more to less healthy individuals and families, but less redistribution from higher income to lower income groups. 2.3.3 A short digression to federal-provincial fiscal relations When it decided not to take action on the first Royal Commission report, the British Columbia government "called on the federal government to shoulder its responsibility by enacting a variety of social insurance schemes" (Naylor 1986, p. 44). In the wake of the depression with its falling prices, unemployment and stagnation of industry, the provinces were increasingly turning to the federal government for assistance. Prime Minister Bennett's government responded in 1935 with the Employment and Social Insurance Act, outlining direct federal administration of a program for health and welfare that would be financed by premium payments. The federal government recognized that this proposal would raise questions about constitutional jurisdiction, so included a preamble arguing that social programs such as unemployment insurance were "essential for the peace, order and good government of well as to maintain interprovincial and international trade on equitable terms."26 Quoted from HRSDC. History of Unemployment Insurance - Developments Between the Wars. http://www.hrsdc.t; insurance chapter3.shtml&hs=tvt. Accessed 4 December 2005. 36 The opposition Liberal party, led to election victory by William Lyon MacKenzie King later that same year, supported the ideas embodied in this legislation in principle, but opposed it on constitutional grounds. Once in power, King referred the legislation for review, first to the Supreme Court of Canada, and then on to the Judicial Committee of the Privy Council in England. Ultimately, the Act was declared ultra vires, meaning it was deemed to exceed the scope and authority of the federal government to enact. (CBC Radio 1957; Hall 1980; Leclair 1975)27 Canada is a federal state, a country with federal and provincial governments, neither of which is subordinate to the other. This is a form of government distinct from unitary systems (like England) where there is a single, powerful governmental organization, with all sub-national government authorities acting as creatures of that higher authority. Provinces in Canada are not subordinate in this way to Ottawa. Federal systems are also distinct from international alliances, in that there is a central, even though not all-powerful, government authority (Canada.Parliament.House of Commons. Special Committee on the Federal-Provincial Fiscal Arrangements 1981). Modern federal systems are described as having two levels of government that are both interdependent and autonomous (Simeon 1972). The challenge in any federal system is balancing the powers of legislation and the powers of 28 the purse. The 1940 report of the Royal Commission on Dominion-Provincial Relations , or the Rowell-Sirois Commission, identified this issue. As it described, the conflict was that the provinces had constitutional authority for all policies and programs not specifically designated 29 for the federal government in section 91 of The British North America Act, 1867 , but did not have the fiscal power to put such programs in place. The federal government had the fiscal resources, but no legal or constitutional authority to act (CBC Radio 1957). The Rowell-Sirois Commission report reiterated the court's decision from 1935 that health was a provincial jurisdiction, and did not find any compelling argument to suggest a change in that arrangement. It did, however, warn "against rigidity in the matter of jurisdiction, because future conditions might warrant a national plan" (Hoppenrath et al. 1980). The federal government set up this Royal Commission after the Employment and Social Insurance Act was struck down. This was renamed the Constitution Act in 1982.Section 91 lays out the powers of the federal government, and section 92 the areas of exclusive jurisdiction of the provincial government. "Unemployment Insurance" was added to section 91 in 1940. Hospitals are the only health-related area named explicitly in the BNA Act, with Marine Hospitals the responsibility of the federal government, and all other hospitals falling to the provinces. The BNA Act can be found in several locations on the internet, but see for example e.html. Accessed 5 December 2005. 37 This set of events in the 1930s is only the beginning of a long history of federal-provincial conflict in Canada over appropriate contributions to health care financing, determination of fair shares of financing, and jurisdiction for policy development and implementation. The importance for the discussion here is that the realm of possibilities for policy development relevant to both financing and delivery of health care services is determined by both federal and provincial "desires". Provinces would generally find it difficult to act unilaterally in health care because of a limitation of funds. The federal government would find it impossible to act unilaterally because it has only a limited role in health care delivery. Health care policy thus rests on a tenuous balance of federal and provincial interests, and is a constant point of discussion between these levels of government. Finally, the federal-provincial disputes have the potential to over-shadow, or perhaps pre-empt, meaningful debates about individual (as opposed to federal-provincial) equity in health care finance and delivery. 2.3.4 The war years World War II bolstered public support for social welfare programs such as unemployment insurance, old age income protection programs and health insurance (Maioni 2004). Naylor (1986) sums up public sentiment by saying: Canada 's military endeavours unquest ionably had a broad base of popular support , but resentment of the high pr ice paid to earn v ic tor ies abroad led Canadians to look for a be t te r dea l on the home f ront . Membersh ip in t rades unions and the CCF grew apace . And the L ibera ls , pushed le f tward by this po l i t i ca l th rea t , eventua l l y s teered a hea l th insurance program to the br ink of i m p l e m e n t a t i o n , (p. 99) While the war was still in progress, the federal government appointed a Dominion Committee on Reconstruction to look into post-war planning. It was recognized the returning servicemen would require (and in all likelihood expect) assistance in re-integrating into the domestic workforce. The 1943 Committee report, Report on Social Security for Canada (the Marsh report), is considered a landmark document in the history of Canadian social policy. The Committee was chaired by Leonard Marsh, recently returned to Canada after working as a researcher in England with Sir William Beveridge, author of the 1942 Beveridge Report. Beveridge, in explaining the three guiding principles of his Report, wrote: " A revolutionary 30 moment in the world's history is a time for revolutions, not for patching." The Marsh report Modern History Sourcebook. Sir William Beveridge. Social and Allied Services (The Beveridge Report), 1942. Accessed 2 August 2005. 38 carries on in a similar vein, outlining a comprehensive system of income and insurance supports that would create a coherent social safety net for all residents of Canada. Marsh identified several principles underlying the organization of a health insurance system. "Universal risks" refers to social insurance, as opposed to a system determined by voluntary insurance, either through employment or individual policies. The system was also meant to be comprehensive, administratively efficient, and "constitutionally free from dispute" (Marsh et al. 1943). The available alternatives to accomplish these objectives / principles were described as: a constitutional amendment to the British North America Act, to give full jurisdiction over health to the federal government; allowance for the provinces to initiate their own legislation, under agreement of federal priority where there were conflicts between provincial and federal acts; and third, reliance on provincial initiative with more loose federal "coordination" (Marsh et al. 1943). Meanwhile, at the start of the Second World War, Ian Mackenzie, Member of Parliament from Vancouver, became Minister of Pensions and National Health. Mackenzie almost immediately started to think about post-war planning. Before the end of the year he wrote to Prime Minister K ing , saying, "Unemployment insurance w i l l be indispensable in coping with the reestablishment problem.... A n d a demand for a national health system is inevitable" (quoted in Taylor 1978, p. 16). It seems likely that Mackenzie was familiar with B . C . ' s failed efforts to create health insurance legislation, and that province's suggestion that the federal government should take action. With Mackenzie's support, an Advisory Committee on Health Insurance was finally appointed in 1942, chaired by Dr. J. J. Heagerty, Director of Public Health Services in the 31 Department of Pensions and National Health (Taylor 1978). The Marsh report and draft legislation developed by this advisory committee were presented to a Commons Special Committee on Social Security in 1943. Taylor (1978) describes six basic principles that the advisory committee thought important: 1. That no scheme of health insurance can be successful without a comprehensive public health program of a preventive nature; 2. That a real health program as distinguished from a policy of cash benefits can be effective only if it embraces the entire population; 3. That the principle of compulsory contributions should be embodied in any plan of health insurance to the greatest possible extent; 4. That public opinion and efficiency demand to the greatest possible extent a national plan; 3 1 Taylor (1978) Maioni (2004) and Naylor (1986) all provide (complementary) details of the process and politics of this committee. 39 5. That the Const i tut ion , as at present understood and in te rp re ted , prevents the Dominion Par l iament f rom adopt ing a single comprehensive nat ional Health Insurance A c t ; 6. That , for p ract ica l reasons, a const i tut ional amendment is not des i rable . (Taylor 1978, p. 18) Financing for this system of health insurance would come from individuals and employers as well as provincial and federal governments. Employed persons would contribute a percentage of income adjusted for family size, to a maximum cap. For example, a single person would pay three percent of income to a maximum of $26, an individual with one dependent 3.7% of income to a maximum of $52. Employers would contribute to the extent that employee contributions fell short, for example if three percent of an individual's income was less than $26, the employer would top up the employee's contribution so that it totalled $26. The financing required by the program not raised through this scheme would come from provincial and federal contributions, likely from general revenues, though this was not stated explicitly. Unemployed persons would make the same contributions, i f they had income from sources other than wages, or would be covered by contributions from their home province (Canada.Advisory Committee on Health Insurance et al. 1943; Taylor 1978). The benefits covered under the proposed plan included services from physicians and hospitals as well as limited dental benefits and coverage for pharmaceuticals and nurses. Separate grants were proposed to cover care for patients with tuberculosis and with mental conditions, and general public health initiatives (Canada.Advisory Committee on Health Insurance et al. 1943). What was proposed, then, represented a significant change in formulation from what was proposed in British Columbia on several points. First, there was a greater breadth of services to be covered, including at least some services in addition to physicians and hospitals. Second, the federal plan was designed to provide universal coverage, instead of "splitting the market" as proposed by British Columbia's two royal commissions. Third, financing was proposed as a percentage of income, up to a maximum, instead of the equal per person charges proposed in B.C. This is a proposal for (limited) proportionate financing, drawing an equal percentage of income to the maximum, and then a decreasing percentage of income above that.32 Tax analysts would describe this as a system that is proportionate to a maximum contribution, and then regressive among higher incomes. 40 Despite all of this activity, neither the recommendations from the Marsh report nor the legislation proposed by the Advisory Committee on Health Insurance were implemented at that time. Federal-provincial fiscal relations continued to frustrate efforts, but the Liberal government in Ottawa did recognize the public's interest in social reform, and the political threat of that interest in the growing popularity of the Co-operative Commonwealth Federation, the forerunner to the New Democratic Party. As an alternative, the new Minister of National Health and Welfare, Paul Martin Sr., suggested taking a "preparatory step" to health insurance of providing cash grants to the provinces for the construction of hospitals (Maioni 2004). Politically, this was what Mr. Martin believed was possible at the time. The other recommendations were shelved, but not forgotten. 2.3.5 The Hospital Insurance and Diagnostic Services Act 33 In 1947 Saskatchewan implemented a hospital insurance program. British Columbia followed in 1949 (Hoppenrath et al. 1980). The success of the program in Saskatchewan encouraged the public to believe that a full system of health insurance was possible. The push for national insurance was strong from the Co-operative Commonwealth Federation (CCF), the left-of-centre political party aligned with both labour and agrarian interests, which had strong support on the prairies and in British Columbia (Maioni 2004). This vocal support prompted the federal government to act, at least in part because of fear that not acting would increase support for the CCF and erode the political base of the Liberal party.34 Unlike the 1930s and 1940s there was no longer consensus among interest groups about insurance for health care. Physicians, through the Canadian Medical Association, continued to solidify their opposition to compulsory health insurance. Their alternative proposal continued to be subsidies of compulsory coverage for lower-income individuals, with others covered on a voluntary basis through private markets, in an effort to maximize potential incomes in the profession (Naylor 1986). The idea of "coverage to maximum income", consistent with the British Columbia proposals in the 1920s and 1930s, was echoed in the positions of business interests such as the Canadian Manufacturers' Association and the Chamber of Commerce. Despite this resistance, the federal government felt compelled to act. As summarized by Leclair(1975): Taylor (1986) suggests the impetus in Saskatchewan came from the lack of a Hospital Association-Blue Cross initiative to provide health insurance there, unlike in other provinces following discussions (and public support) in the 1940s. Maioni (2004) argues that this is precisely what the Democratic party in the U.S. did as well, co-opting the policies and therefore the support of a progressive party there. The difference was that the Democrats were able, ultimately, to subsume those interests, while the parliamentary system in Canada helped the CCF (and eventually NDP) to maintain itself as a separate political party. 41 In the midd le f i f t i es , in terest in hea l th insurance had grown and many provinces w e r e pressing the f e d e r a l government for some act ion in this f i e l d . The prov inc ia l governments w e r e f ind ing i t very d i f f i cu l t to raise the necessary revenues to m e e t the esca lat ing costs of provid ing hospi ta l serv ices . Many Canadian hospitals w e r e in f i nanc ia l t r o u b l e , espec ia l l y the smal le r and rural ones. Technology was beginning to burst for th and also unionizat ion of hospi ta l workers had b e c o m e a f a c t . Wh i le hospi ta l workers had previously been the poor cousins of the labor f o r c e , w i t h increas ing mi l i tancy the unions w e r e insist ing that the wage gap be c losed and eventua l l y e l i m i n a t e d b e t w e e n the hospi ta l workers and the rest of the work ing f o r c e . The hospitals w e r e f ind ing i t e x t r e m e l y d i f f i cu l t to f inance the i r operat ions and keep up the i r standards because less than 40 percent of the populat ion at that t i m e had some hospi ta l insurance coverage and much of it was not adequate , (p. 14) The Hospital Insurance and Diagnostic Services Act (HIDSA) was proposed in the m i d -1950s and passed i n 1957. The Act took effect Ju ly 1, 1958, w i t h a l l prov inces on board by the start o f 1 9 6 1 . 3 5 Th is legis lat ion has been descr ibed as the largest s ingle program ever undertaken i n peace- t ime Canada. A c c o r d i n g to H a l l (1980) , " The development o f the ten P r o v i n c i a l and two Terr i tor ia l hospital services programs became, i n fact, a nat ional program o w i n g to the combinat ion o f federal contr ibutions and federal standards, especial ly those requi r ing u n i f o r m def in i t ions o f residency and portabi l i ty o f benef i ts . " One d ist inguish ing feature o f this legis lat ion was that the federal fund ing w o u l d neither require nor i m p l y any change in ownership o f hospitals ( H a l l 1980). Government invo lvement in the organizat ion o f a universal hospi ta l insurance scheme was not to be confused w i t h government take-over o f the faci l i t ies that p rov ided services. Another feature was that the legis lat ion was not sole ly about fund ing arrangements. The Act spec i f ied a series o f audit controls to rev iew requests for fund ing and po l ic ies on l icensure, inspect ion and superv is ion o f hospi ta l standards. "In fact, the Act is p r imar i l y a legis lat ive enactment to enable people to obtain the services they require, and secondar i ly , a f inanc ia l arrangement to assist i n payment for those serv ices" ( H a l l 1980). The Act established several pr inciples required for provinces to receive funding f rom the federal government, as: an undertaking by the provinces to make insured services avai lable to a l l residents upon uni form terms and condi t ions ; to make arrangements to ensure the maintenance in hospitals of adequate standards; to mainta in adequate records and accounts See, for example, Government of Canada, Health Canada, Health Care System, e.html. accessed 25 January 2006. 42 in connect ion w i th the provision of insured services and the costs; and to permit access to these records and accounts ; and to make provision for the recovery of costs in th i rd party l iabi l i ty cases. (Canada.Department of Nat ional Health and Wel fa re 1960, p. 2) Funds to the provinces were provided based on services rendered, with the calculation described as: The federa l law provides that the contr ibut ions to the provinces shal l be paid out of the consol idated revenue fund by the Minister of F inance upon cer t i f i cat ion by the Minister of Nat ional Health and Welfare. . . The federa l contr ibut ion as out l ined in the A c t , is the aggregate in the year of twenty - f i ve per cent of the per cap i ta costs of in -pat ient services in Canada , that is the nat ional per cap i ta cost , and twenty - f i ve per cent of the per cap i ta cost of i n -pat ient services in the province less the amount of author ized charges, mul t ip l ied by the average for the year of the number of insured persons in the province. The e f fec t of this fo rmula is that the high-cost provinces receive a lower percentage of the i r costs f rom the federa l government than do the low-cost provinces. The inclusion in the formula of the nat ional per cap i ta cost , however , acts as a deterrent to the high-cost provinces, s ince the more that provincial costs exceed the nat ional costs, the lower the percentage of the federa l contr ibut ion w i l l be . General ly speaking, the federa l contr ibut ion amounts to approx imate ly half of the provincial costs across the country , although in indiv idual provinces it var ies. (Canada. Department of Nat ional Health and Wel fare 1960, pp. 11-12) There were no restrictions on the means by which provinces raised their portion of financing. Some instituted premia, others depended on special levies or general revenue. British Columbia financed its contribution through general revenue (Canada.Department of National Health and Welfare 1960). HIDSA was a critical step because it was the first piece of federal legislation that provided for insurance coverage for health care services. In making this step, the federal government also made a strong statement about equity with the requirement for treatment of individuals on uniform terms and conditions. The whole population was covered by this legislation, and was covered without respect to age, sex or income. Financing from general revenues at both the federal level and in British Columbia meant that while insurance provided a subsidy of individuals who used hospital services, contributions to health care finance would mirror contributions to general revenue. This legislation is more closely in line with the Marsh report and the Advisory Committee on Health Insurance proposal from the 1940s than it is with the proposals of the 1920s and 1930s in British Columbia. "Equity" in HIDSA meant universal coverage on uniform terms and conditions, 43 implying a principle of providing access based on need without respect to ability to pay. The decision to finance the system through general revenues meant that there was no designated health care fund, an understandable decision since the federal contribution was to be determined by provincial patterns and levels of expenditure. Financing from general revenues also meant that there was no health care-specific policy about its distributional or redistributional effects. Clearly there would be a transfer of income / benefits from the healthy to the sick, consistent with any form of insurance, but the transfer from wealthier to poorer was dependent on the general form of contributions to general tax revenue. The policy, though implicit, was to say that hospital services were to be financed according to the same principles used to raise revenues for all other government spending. 2.3.6 The Hall Commission With 10 years of experience operating a hospital insurance system funded fully from provincial resources, Saskatchewan was in the first group of provinces on board with the HIDSA. The federal funding came in almost as a windfall, so the government, under the leadership of Tommy Douglas from the CCF, immediately announced its intent to proceed with a physician insurance scheme (Badgley and Wolfe 1967). Prime Minister Diefenbaker, perhaps in part as a move to avoid (or at least delay) pressure for a similar move at the federal level, appointed a federal Royal Commission to look at the broader issue of health insurance. Mr. Justice Emmett Hall, Chief Justice of the Saskatchewan Court of Queen's Bench, and former law school classmate of Mr. Diefenbaker, was appointed the Commission's chair (Taylor 1978).36 Hall's review was comprehensive and the report is another landmark document on Canadian social policy. The conclusions were built on the principles of the HIDSA; Hall recommended that insurance for physician services should be universal, comprehensive, portable and publicly administered. The requirement for universality has a pragmatic flavour. The Canadian Medical Association suggested, as before, that subsidized, comprehensive coverage should be provided for lower-income individuals and families, while coverage for higher income individuals should be on a voluntary basis. The analysis for the Hall Commission considered costs and the requirement for The Canadian Medical Association was in support of this Royal Commission. Their experience with the Hospital Insurance and Diagnostic Services Act further cemented their position on health insurance, and the organization wanted to ensure a fair and impartial review of health insurance options. They believed this fair review was certain to conclude, as they had, that universal, government-mandated and operated insurance was not in the interest of the doctor-patient relationship (Taylor, 1978). 44 these subsidies based on adding the hospital and physician components together. In treating these two pieces as the beginning of a more comprehensive system of health care insurance, Hall came to the conclusion that perhaps 40% of the population would require a subsidy. In other words, a universal and comprehensive health care insurance program would require premiums high enough that a large proportion of the population would qualify for assistance. The Commission felt that the means tests necessary to run this sort of program were not in and of themselves desirable, and introduced unnecessary administrative costs and complexity. Since such a large proportion of the population would meet the means test, the Commission concluded that providing universal coverage was both fairer and more cost-effective (Canada.Royal Commission on Health Services 1964; Hall 1980). The Hall Commission report recommended that the federal government continue its (approximate) fifty percent funding as started through HIDSA. It also recognized that the provinces had the right to raise their contributions as they saw fit, but suggested the federal government raise its portion of the contribution through general revenue. The Commission gave careful thought to financing issues, and in particular to the fact that the proposed changes represented a significant expansion of a public program. With the i m p l e m e n t a t i o n of the recommendat ions of the Commiss ion there wou ld be an expansion of the proport ion of t o t a l spending f inanced in the publ ic sector . This wou ld be due part ly to new spending in i t ia ted under the programmes but pr imar i l y because pr ivate spending now wou ld b e c o m e publ ic spending . Much of the increased expendi tures associated w i t h our recommendat ions is the consequence of a shi f t in the method of paying for heal th services that wou ld be purchased whether or not any expansion of publ ic programmes takes p lace . (Canada.Royal Commiss ion on Heal th Services 1964, p. 865) In other words, the Commission very consciously chose a model that they projected would mean that public financing would account for 85% of total health care expenditures in 1971 (Canada.Royal Commission on Health Services 1964). Given the existence of public financing, the Commission was not strict about just how those finances should be raised, for example through income or other taxes, or how they should be shared by levels of government over the longer term. "Provided that the needs of Canadians are met; provided that the quality and scope of health services provided is equivalent to what we have recommended; provided such services are operated in the most economic manner and properly planned and integrated; the techniques adopted for their financing are of secondary importance" (Canada.Royal Commission on Health Services 1964, p. 873). 45 The Hall Commission justified its recommendations in part by arguing that the public would welcome the changes. How do Canadians in genera l f e e l about the possible higher cost of improved heal th care and the prospect of higher taxes or other forms of cont r ibut ion for this purpose? To answer this quest ion w e have throughout our hearings inqu i red f rom spokesmen of representat ive populat ion groups w i t h vary ing in terests , whether those they represented wou ld a c c e p t higher taxat ion to pay for the cost of increased and improved heal th serv ices . The answer has been yes, and there has been no dissent ing vo ice . (Canada.Royal Commiss ion on Heal th Services 1964, p. 878) The Hall Commission report further entrenched the views of equity that were established with HIDSA. A comprehensive set of services (albeit within specific sectors of the health care system) was to be provided on universal terms and conditions. Payment for those services was to 37 be from taxes, with no health-care specific policy attached to financing. Equity in access to services based on need rather than ability to pay begins to emerge as the primary equity principle in the organization of health care services in Canada. Equity in financing is more slippery, with the main principle seeming to be the separation of contributions to health care financing from the use of health care services, without any specific recommendations for raising the financing beyond that. 2.3.7 The Medical Care Act In 1961, Lester Pearson was leader of the (Liberal) opposition in Ottawa when that party " a federally subsidized, provincially administered national Medicare program" (Taylor 1986). The Liberals formed a minority government in 1963, and Walter Gordon was appointed as Minister of Finance. "He was the only strong supporter of health insurance to occupy that post in the fifty years that health insurance was a public issue. And it made the difference." (ibid., p. 6) The Medical Care Act, providing insurance for physician services, was passed in 1966, was effective July 1, 1968, and by 1971 all provinces had signed on. The principles embodied in the Medical Care Act were set "in keeping with a recommendation from the Royal Commission on Health Services in 1964" (Canada.Department of National Health and Welfare 1969, p. 64). These principles expanded those in the Hospital Insurance and Diagnostic Services Act and The Commission report notes that part of the reason the Commission did not wish to make specific recommendations in this area was that similar work was ".. .already being carried out by other Royal Commissions in the area of taxation and economic policy" (Canada.Royal Commission on Health Services 1964, p. 858). 46 included "comprehensiveness of insured services, universality of coverage, portability of benefits and public administration" (ibid., p. 1). This means that all Canadians were covered by universal and comprehensive hospital and physician insurance by 1971. Malcolm Taylor has described Canadians feeling "...pride that the society had taken Herculean measures to substitute medical need for ability to pay as the main determinant in the queuing process and to distribute the costs of so doing with reasonable equity" (Taylor 1980). The Medical Care Act provided for federal reimbursement of provincial costs, calculated based on 50% of average per capita costs of participating provinces, a different formula than used for hospital services. The means by which the provincial share was raised were left to the discretion of participating provinces. British Columbia and Saskatchewan were the first two provinces to join this program. In British Columbia, the Medical Services Act (effective March 1967) governed the operation of the program. The British Columbia Medical Services Plan was a governmental agency set up to provide what was described as a "voluntary premium plan", in which "coverage is contingent upon the required premiums having been paid with coverage ceasing when the premium is 15 days in default" (Canada.Department of National Health and Welfare 1969, p. 7). Other insurers were allowed to provide this same coverage on a group basis (i.e. the Medical Services Plan was not established as the sole insurer), and were able to sell policies for benefits that exceeded the provincial plan. The premia were determined by family size, with some adjustment for very low incomes. Single individuals, for example, paid $60 per year, $30 per year if taxable income was less than $1,000, $5.00 per year if taxable income was nil, and $0 if they qualified for social assistance. As of the early 1990s, the intent of the premia was to raise about half of the funds needed to reimburse physicians and alternative providers (such as chiropractors or physiotherapists) for services rendered (British Columbia Royal Commission on Health Care and Costs 1991). These were not earmarked taxes, however, so in actual fact they were pooled into general revenue, from which funds for health care were drawn. As of the end of the first year of operation, estimates were that 92% of the population of British Columbia had insurance coverage for physician services. The Medical Care Act thus enshrined in legislation the principles of equity suggested by the Hall Commission, although at this point they applied (in law) only to physician services. The choice of British Columbia to enact premia to raise revenues to meet the (approximate) provincial costs of the program shows there was no intent to create progressivity in the 47 payments. It is possible that the rationale for this choice reflects similar thoughts to those expressed in the earlier B.C. Royal Commission, suggesting that equal payments were appropriate since the benefits for lower-income and higher-income individuals would be the same. It is also possible that premia were instituted to allow the implementation of the program without the need to raise income taxes. Whatever the reason, the impact was a form of financing for the provincial portion that was regressive, taking a larger percentage of income from lower-income groups than from higher. 2.3.8 The Established Programs Financing Act A special committee on federal-provincial fiscal arrangements observed that once the goal of establishing a universal, comprehensive system of health insurance in Canada had been achieved . . .much of the reason for the par t i cu la r design of the f i sca l ar rangements was removed . In par t i cu la r , the costs of the programs w e r e now re lat i ve ly w e l l unders tood , and organized opposi t ion to the programs had d i sappeared . It no longer seemed necessary for the f e d e r a l government to play an advocacy ro le , through shared -cost programs, to deve lop a nat iona l system of med ica re and hea l th insurance . (Canada .Par l iament .House of C o m m o n s . S p e c i a l C o m m i t t e e on the F e d e r a l - P r o v i n c i a l F iscal Ar rangements 1981, p. 54) At the same time, there were concerns from the provinces about the rigidity of the funding from the federal government. They did not like the auditing function of the federal government to determine costs that were allowable for cost sharing. Perhaps more importantly, the provinces claimed that the limitation of health insurance to the two major areas of physicians and hospitals created "distortions in provincial priorities in health services" as these two programs were only pieces "of an increasing range of health services programs" (Taylor 1986). For its part, the federal government was concerned that with (essentially) open-ended 50-50 cost sharing it could exert no control over the increases in health sector spending, which were 22% in 1974 and 26% in 1975. Finally, there was concern that perhaps Canada was one of a growing number of countries that were no longer getting large improvements in health for increased health care spending (Taylor 1986). The federal response was the introduction of the Federal-Provincial Fiscal Arrangements and Established Programs Financing Act of 1977 (EPF). This Act created a formula and criteria for a block-fund federal contribution to hospital services (as defined by the Hospital Insurance and Diagnostic Services Act), physician services (as defined by the Medical Care Act), extended health care and post-secondary education. This major change introduced prospective rather than 48 historical-cost-based funding, and in general fundamentally altered the fiscal relationship between the provincial and federal governments with respect to health care. The federal contribution under EPF consisted of two components: a reduction in federal income tax rates and a set of cash contributions. The reduction in federal tax rates was a transfer of 13.5 personal tax points and 1 corporate tax point (1 point=l%) from the federal to the provincial governments, in essence a creation of "tax room" for the provinces (Broyles R.W. 1981; Canada.Department of National Health and Welfare 1978). The cash contributions were based on the total amount paid by the federal government for hospital insurance, medical insurance and post-secondary education programs in fiscal 1975-76 with adjustments to try to equalize payments among the provinces (on a per capita basis). The payments included an escalator adjustment equal to a three-year moving average of the rate of growth in per capita Gross National Product (Canada.Department of National Health and Welfare 1978). The effect of block funding was to shift the economic risk of health insurance to the provinces. The provinces were now given a budget tied to past spending and economic growth and were on the hook for expenditures exceeding that budget. The federal government was no longer tied to paying 50% of the cost of insured hospital and medical services. This change was consistent with the conclusion of the committee on federal-provincial fiscal arrangements, which wrote in 1981 that ".. .the proper role for the federal government is the formulation, monitoring and enforcement of conditions on its financial support of provincial programs" (Canada.Parliament.House of Commons.Special Committee on the Federal-Provincial Fiscal Arrangements 1981). 2.3.9 The Canada Health Act The connections between the introduction of EPF and equity for the Canadian public are perhaps not direct but are important. First, as it turned out, this new arrangement left the provinces at the federal government's mercy with respect to the funding formula and their desire to tinker with it. Second, payments were no longer tied to services rendered, which left the federal government without any stick to enforce the principles in the hospital and medical care legislation. It is perhaps no surprise then that almost as soon as EPF was in place there were "large increases in the number of doctors extra-billing38, an increased militancy among nurses Extra-billing is charging prices for services that are higher than the fees paid to physicians by the public health plan, and expecting patients to pay the difference. 49 and other hospital unions, and the beginning of charges that, as a result of EPF, provinces were diverting federal "contributions" to non-health purposes" (Taylor 1986).39 One question was then whether extra-billing by physicians and user charges imposed by some provinces for hospital services put the basic principles of hospital and physician insurance in jeopardy. In a commissioned review of the health care system released in 1980, Justice Emmett Hall concluded that " i f extra-billing is permitted as a right and practised by physicians in their sole discretion, it will, over the years, destroy the program, creating in that downward path a two-tier system incompatible with the societal level which Canadians have attained" (as quoted in Canada.Parliament.House of Commons. Special Committee on the Federal-Provincial Fiscal Arrangements 1981, p. 109). A 1981 federal parliamentary committee charged with reviewing EPF (the Breau Committee) quoted this statement by Judge Hall in its report, as well as another: Canadians understand the fu l l meaning of the Hospital Insurance and Medica l Care Acts . They sa id , through these two acts , that w e , as a society , are aware that the t rauma of i l lness, the pain of surgery, the slow dec l ine to d e a t h , are burdens enough for the human being to bear wi thout the added burden of med ica l or hospital bil ls penal iz ing the pat ient at the moment of vu lnerabi l i ty . The Canadian people determined that they should band together to pay med ica l bil ls and hospital bil ls when they were w e l l and income earning. Health services were no longer i tems to be bought off the shelf and paid for at the checkout s tand . Nor was thei r pr ice to be bargained for at the t i m e they are sought. They were a fundamenta l need , l ike educat ion , wh ich Canadians could meet co l lec t i ve ly and pay for through taxes, (p. 109) The Breau Committee did not explicitly endorse this view of medicare, but it did write that "the legitimate interests of doctors must give way to the broad public perception that uncontrolled billing of patients beyond the levels of provincial medical insurance plan schedules will ultimately destroy medicare" (Canada.Parliament.House of Commons.Special Committee on the Federal-Provincial Fiscal Arrangements 1981, p. 110). This committee recommended that physicians not be allowed to extra-bill. The committee's review of EPF included an assessment of its potential impact on the five principles embodied in HIDSA and the Medical Care Act. The issue of extra-billing fit under its assessment of accessibility. The report says that universality, another one of the five principles, can refer either to the number of individuals in a population covered by health insurance, or the As Begin (1988) points out, this last charge is silly. In the absence of cost-sharing (and in the case of block grants, as was EPF), the only way that the federal government could charge provinces of diversion is if their total health spending was less than the amount of the federal transfer. In any case, it was discovered that the shares of provincial government budgets spent on health care were the same before and after the switch to Established Program Financing. (Taylor, 1986) 50 range of services covered under the plan. Concerns about universality were tied to the fact that three provinces (Ontario, Alberta and British Columbia) plus the Yukon were charging health insurance premiums to raise general revenues for government. The report says that the premiums were "charged in a manner that might compromise satisfaction of the condition of universality. But the evidence is not conclusive" (p. 105). The report also provides a glimpse of provincial rationale in charging premia, given to the committee through its consultation process. A l l th ree provinces that now apply p remiums have s ta ted that they do not in tend to abandon t h e m for some other fo rm of f inanc ing . Premiums o f fe r one way , they say, to make the c i t i zen aware of the cost of heal th care serv ices . Fu r thermore , some provinces contend that residents should pa r t i c ipa te more d i rec t l y in the f inanc ing of these serv ices , as a m a t t e r of p r inc ip le . (Canada. P a r l i a m e n t . House of C o m m o n s . S p e c i a l C o m m i t t e e on the Federa l -P rov inc ia l F iscal Ar rangements 1981, p. 106) The Breau Committee concluded that: and medical care premiums constitute a regressive form of taxation and ...their use for financing a service as basic as health care is regrettable. The Task Force recommends that a clearer definition and measurement of universality of coverage be developed to ensure that the principle is respected. (Canada .Par l iament .House of C o m m o n s . S p e c i a l C o m m i t t e e on the Federa l -P rov inc ia l F iscal Ar rangements 1981, p. 106, bold and i ta l i cs in or iginal) This seems to takes the committee further than an analysis of EPF's potential impact on universality, albeit in a form that is a comment rather than a recommendation. They recommended that the Minister of National Health and Welfare work toward consolidating existing health care legislation "in order to establish clear program conditions supported by explicit criteria against which satisfaction of those program conditions can be monitored" (Canada.Parliament.House of Commons.Special Committee on the Federal-Provincial Fiscal Arrangements 1981, p. 115). The federal government's response was to draft the Canada Health Act. Monique Begin, the federal Minister of Health during this period, and the major sponsor of the Canada Health Act in the legislature, believed that passage of the Canada Health Act was necessary specifically because of EPF. Her rationale was that the transfer of funding of health care services through a pre-determined block grant instead of the review and payment of "bills" for services provided effectively removed the principles and requirements of the previous legislation. Her argument was that the only stick she had available to enforce the principles of HIDSA and the Medical 51 Care Act was withholding the entire monthly block grant. This would be out of scale in relation to the (alleged) infractions (Begin 1988). The Canada Health Act was a straightforward piece of legislation that laid out the principles underlying the transfer of funds from the federal to provincial and territorial governments. As well, it articulated the penalties applicable in the case of departure from this principle. There was widespread support for this legislation, excepting only (some) provincial governments and the medical profession, for familiar reasons. The Canada Health Act was passed unanimously and was proclaimed in 1984 (Canada.Parliament. 1984). Section 3 of the Act states "It is hereby declared that the primary objective of Canadian health care policy is to protect, promote and restore the physical and mental well-being of residents of Canada and to facilitate reasonable access to health services without financial or other barriers" (Canada.Parliament. 1984). The Act specifically identifies five principles of engagement: public administration, comprehensiveness, universality, portability and accessibility. As recommended, this legislation combined the hospital and medical acts, maintaining all of the principles embodied in each (including the non-profit administration principle of the Medical Care Act). This legislation came about at least in part as a response to concerns about direct charges to patients, mainly in the form of physicians charging fees that were higher than those reimbursed by provinces. The principles of the Canada Health Act thus further entrenched the notions of equity first articulated as dealing with individuals on equal terms and conditions. Despite the Breau Committee making its comment about the use of health care premia, the Canada Health Act says nothing directly about equity in financing, except in continuing to separate, at least for hospital and physician services, payment for and use of services. 2.3.10 The B.C. Royal Commission on Health Care and Costs and the Medicare Protection Act The British Columbia government established a Royal Commission on Health Care and Costs that tabled its final report in 1992. The Seaton Commission, as it was commonly known, was charged with a broad-based review of health and health care in British Columbia. B.C. was by no means alone in this move, as many similar provincial reviews were undertaken during the 1990s. The bulk of the final report was devoted to comments and recommendations on the way care is delivered, and things that could be done to improve the quality of services without necessarily increasing overall costs. In fact, a major theme of the report was that more could have and should 52 have been possible with already-committed funds before resorting to providing additional funding for the system. The report "fully endorses" the principles of the Canada Health Act. More than this, the report opens by outlining the principles of the Canada Health Act, and then recommending (as the first recommendation in the report) that "the government of British Columbia confirm the five principles of medicare, as described by the Canada Health Act, 1984, by enacting them in legislation" (British Columbia Royal Commission on Health Care and Costs 1991, p. A-5). In terms of financing, the commission report recommends (although not unanimously) that the premia charged in British Columbia be abolished. A New Democratic Party government was elected soon after the tabling of the commission report in 1991. This government introduced legislation called An Act to Protect Medicare in 1995, in part in response to the Seaton Commission (it appears), but also because of increasing concerns about physicians in B.C. charging patients for services in addition to the fees they received through the province's public Medical Services Plan. The Act passed second reading with 49 votes in favour and 1 against, and received royal assent on 13 July 1995.40 This was, in fact, not an entirely new act, but instead an amendment of an existing piece of legislation that governed the operation of the health insurance function in British Columbia. The amendment added details about restrictions on physician "extra-billing" and a preamble that stated the purpose of the legislation. There was no preamble in previous legislation. It now read: WHEREAS the people and government of Brit ish Co lumbia bel ieve that medicare is one of the def ining features of Canadian nat ionhood and are c o m m i t t e d to its preservation for future generat ions; WHEREAS the people and government of Brit ish Co lumbia wish to conf i rm and entrench universal i ty , comprehensiveness, accessib i l i ty , portabi l i ty and publ ic administ rat ion as the guiding pr inciples of the health care system of Brit ish Co lumbia and are c o m m i t t e d to the preservation of these pr inciples in perpetu i ty ; WHEREAS the people and government of Brit ish Co lumbia recognize a responsibi l i ty for the judic ious use of m e d i c a l services in order to mainta in a f iscal ly sustainable heal th care system for future generat ions; AND WHEREAS the people and government of Brit ish Co lumbia bel ieve i t to be fundamenta l that an individual's access to necessary med ica l care be solely based on need and not on the individual's abi l i ty to p a y . " (British Co lumbia Legislature 1996) The Minister of Health introduced the legislation by saying: The timeline and information about the legislative debate of this act are derived from Hansard, which is available at The relevant dates were June 29 (first reading), July 10 (second reading), July 12 (committee of the full house) and July 13 (royal assent), all 4 , h session, 35lh parliament (1995). 53 Medicare is a unique Canadian success. It grew out of the exper ience of mi l l ions of working Canadians who lost their life's savings to i l lness in their fami l ies or whose loved ones d ied for lack of af fordable t rea tment . Access to med ica l care , they sa id , must be based on a person's need for ca re , not on thei r abi l i ty to pay... Wi th this legis lat ion, Brit ish Co lumbia becomes the f irst province in Canada to entrench the founding pr inciples of medicare in law: universal i ty , comprehensiveness, accessib i l i ty , portabi l i ty and publ ic a d m i n i s t r a t i o n . " 4 1 The provincial government was now committed to the principles of the Canada Health Act through its own legislation rather than only through the federal legislation and associated threat of withholding of funds. The principle of equity explicitly identified in this legislation is the provision of services based on need rather than ability to pay. By now it is clear that this was (and is) the primary equity principle in the health care system in British Columbia. This Act does not address financing at all, except again in maintaining the separation of use of services and payments to the system, and this only implicitly. The use of the second part of the phrase "rather than ability to pay" reflects the alternative to providing services based on need rather than a commitment to progressive financing of the health care system. In other words, this legislation signals a strong commitment to the health care system redistributing income from the healthy to the less healthy, and nothing more. 2.3.11 The Canada Health and Social Transfer Another major change in the transfer of funds from the federal to provincial governments was introduced in 1996. The Canada Health and Social Transfer (CHST) combined funding for social services and social assistance with the health and post-secondary education funds formerly transferred through EPF. The transferred tax points were left untouched, but the cash component of the transfer was no longer tied to specific areas of spending, nor were increases in the transfer tied to economic performance. The CHST was unpopular from the start, in part because it merged spending for several areas of social investment, which made it difficult to identify the federal government's contribution to each, but perhaps more importantly because the cash contribution dropped significantly between 1995 and 1996 (Mcintosh 2004; Romanow 2002). Because the CHST combined funding for several programs without identifying the relative proportion to be dedicated to each, it is difficult to estimate how much the federal government contributed to each program. Estimates, however, are that the health portion of the cash transfer (for all provinces combined) decreased from $15.7 million in 1995 to $14.6 million in 1996 4 1 See Accessed 31 October 2005. 54 (Romanow 2002, Appendix E.l) . The transfers did increase in the following years, but the CHST continued to be a source of conflict in federal-provincial fiscal relations. For the provinces, the issue was the size of the transfer. For supporters of medicare, the issue was that the declining importance of the cash transfers meant the erosion of Ottawa's ability to enforce the terms and conditions of the Canada Health Act. If or when the federal government ceased to transfer cash to the provinces to fund health care systems, it would also lose any ability to ensure cross-country standards or principles in the operation of those systems. Upholding the now firmly established principles of equity would fall to each individual province and territory. The "health" and "social" components of the transfer were split starting in 2004, and the transfers were given a set schedule to increase the predictability of the federal cash contributions.42 This change diminished the immediate threat of the loss of cash transfers, but did little to improve federal-provincial fiscal relations (see, for example, Mcintosh 2004). 2.3.12 Three recent federal reports The federal government has been involved in three major reviews of the health care system over the last decade. The first of these, the National Forum on Health, was appointed in 1994 and reported to Prime Minister Chretien in 1997. This report endorsed a single-payer, publicly-funded health care system and the five principles of the Canada Health Act. It also recommended extending the current system (covered by the Canada Health Act) beyond hospital and physician services to include coverage for pharmaceuticals and for home care (National Forum on Health 1997). This report also explicitly addressed the issue of values and the role of values in public policy. Exploring va lues . . . is an exp l ic i t acknowledgement of the need to be fa i th fu l to what people real ly want and bel ieve in and of the obl igat ion for governments to set pol ic ies and a l locate resources accordingly. . . In recommending a course for Canada's heal th system, w e considered it essent ial that its foundations be consistent w i th the values of the major i ty of the publ ic . (National Forum on Health 1997) The forum went on to identify the health care related values important to the Canadian public, based on consultations and discussion groups with members of the public. These are: "equity, compassion, collective responsibility, individual responsibility, respect for others, efficiency and effectiveness" (National Forum on Health 1997). Canada Health Transfer. Finance Canada. Government of Canada. Accessed 31 October 2005. 55 Though the Forum made its recommendations in 1997, little action followed, and so the general concerns that it addressed did not disappear. The Commission on the Future of Health Care in Canada was established in 2001, headed by Roy Romanow, a former Premier of Saskatchewan, with a mandate to make recommendations to "ensure over the long term the sustainability of a universally accessible, publicly funded health system..." (Romanow 2002, p. xi). The mandate for the Romanow Commission also stated that these recommendations must respect "the jurisdictions and powers in Canada.. ."((Romanow 2002, p. xi), a direct reference to federal - provincial relations in health care. The Romanow Commission was set up following a September 2000 meeting of Canadian premiers (the "first ministers"). A communique following this meeting established an agreement that the provinces (and Health Canada) would engage in an effort to produce regular reports to the public on the performance of the health care system (Canadian Intergovernmental Conference Secretariat 2000). At the same time, the federal government agreed to increase its contributions to the provinces for health care.43 The Romanow Commission report, like the National Forum on Health, dealt explicitly with views on values and equity. For example: Based on ev idence both in Canada and internat ional ly , progressive taxat ion continues to be the most e f fec t i ve way to fund heal th care in Canada. From what the Commission heard f rom Canadians through the C i t i zens ' Dialogue and other consultat ions, the large major i ty of Canadians do not want to see any change in the s ingle -payer insurance pr inc ip le for core hospital and physician services. There also cont inues to be a strong consensus among Canadians that "abi l i ty to pay" should not be the predominant factor in how we fund key aspects of our heal th care system. Canadians want necessary hospital and physician services to be ful ly funded through our taxes. This may be because our tax -funded , universal health care system provides a kind of "double so l idar i t y . " It provides equi ty of funding between the "haves" and the "have nots" in our society and i t also provides equi ty between the healthy and the sick. (Romanow 2002), p.31) At about the same time, the Senate Standing Committee on Science and Technology, chaired by Senator Michael Kirby, began an investigation of the health care system. One specific objective of this committee was to identify the federal role in Canadian health care (Canada.Parliament.Senate.Standing Senate Committee on Social Affairs 2002, p. vii). The Romanow and Kirby commission reports were released within a month of each other in 2002, and both emphasized the need for a continued leadership role on the part of the federal Government of Canada. Department of Finance. A brief history of the health and social transfers. Accessed 3 August 2004. 56 government, and a need to increase the accountabi l i ty o f the health care system. Just what exact ly that accountabi l i ty should look l i ke , and what it covers (new money? a l l money?) have been points o f debate since then. The K i r b y Commit tee also addressed issues o f equity and f inanc ing direct ly . F o r example : On ba lance , the ev idence ava i lab le on how d i f fe ren t revenue sources a f fec t equity shows that equity is best served w h e n health care is funded through personal income taxation or consumpt ion taxes, rather than through pay ro l l taxes or f i xed p r e m i u m s . . . Furthermore, the Commit tee bel ieves that the increased federal revenue should be raised based on abi l i ty to pay; that is , to ensure equity, ind iv iduals w i t h higher incomes should pay more than ind iv iduals w i th lower incomes. (Canada.Par l iament .Senate.Standing Senate Commit tee on S o c i a l A f f a i r s 2 0 0 2 , p. 287) The K i r b y Commit tee u l t imately conc luded that the health care system required $5 b i l l i o n i n addit ional fund ing i n each year. They r e c o m m e n d ra is ing that money through a graduated health care p r e m i u m t ied to income taxes . 4 4 They also stated, however , that i f the federal government was u n w i l l i n g to implement such a p r e m i u m , or i f the pub l i c was u n w i l l i n g to accept it, then private health insurance m a y become a necessity. A l l three o f these reports support the pr imary equity pr inc ip le o f medicare , emphas iz ing that Canad ian health care systems should continue p rov id ing services based on need rather than abi l i ty to pay. F o r the first t ime, the latter two reports also make expl ic i t statements about f inanc ing medicare . The R o m a n o w report ident i f ied "progressive taxat ion" as the preferred opt ion (as in the quote above), and the K i r b y report recommended a dedicated health care p r e m i u m . The two reports c lear ly come out i n favour o f contr ibutions to f inanc ing that are separated f r o m the use o f services; they di f fer in their interpretation o f "ab i l i t y to pay" , w i t h R o m a n o w favour ing progressive payments and K i r b y proport ionate, to a m a x i m u m contr ibut ion. These reports prov ide a s ignif icant new art iculat ion o f the pr inc ip le o f equity in f inanc ing health care systems in Canada and h ighl ight the major contested issue i n health care f inanc ing . The premium outlined in this report was not, however, progressive. It was designed to take, on average, about 1% of income from each federal tax bracket, with contributions reaching their maximum at about $100,000 of income. Furthermore, the premium was designed in such a way that high and low ends of each of the federal tax brackets would pay substantially different proportions of their income through this premium. This is a substantial (and curious) violation of the principle of "equal treatment of income equals". 57 2.4 Identifying and monitoring equity There are several themes that emerge from this review of Canadian and British Columbia legislative and policy history. One important theme is the tortured and fractious history of federal-provincial fiscal relations. The Rowell-Sirois Commission of the 1930s identified the federal-provincial division of powers and responsibilities as an issue for health-care systems. This division of powers retains ongoing importance as a source of conflict, easily seen in recent tensions between provinces and the federal government over the amount and use of health care funding in the Canada Health and Social Transfer. While this theme may not always have direct consequences for equity, the jurisdictional debates can distract attention from other important issues (such as equity) in health care. The amount of money transferred from the federal government to the provinces also influences the range of policies open to the provinces. An expansion of the range of services covered under the Canada Health Act, for example, would be easier to implement in the presence of new federal funds. Perhaps more importantly, however, the federal government has to offer a significant enough "carrot" in the form of cash transfers to induce the provinces to continue their commitment to the principles of the Canada Health Act and to principles of equity. Without cash transfers from the federal government it would be easier for provinces to claim that the publicly funded health care system in their province is unsustainable, with the potential for these continued claims to alter public support for the program. Abandoning the principles of medicare clearly would mean abandonment of the equity principles as well. The carrot and stick approach has been in place since the passing of HIDSA, ensuring accountability of the provinces in the form of upholding a specified set of principles in exchange for the receipt of federal funding. The main principle of engagement required by the federal government in HIDSA was operating hospital insurance on "universal terms and conditions". This was expanded under the Medical Care Act to include comprehensiveness, portability and public administration in additional to universality. The Canada Health Act then formalized the five principles of universality, accessibility, comprehensiveness, portability and public administration. The issue of accountability, which the Romanow report suggested should be added as a "sixth principle" for health care, has become very contentious in recent federal-provincial negotiations. Provincial governments are reluctant to see conditional transfers, while the federal government is increasingly reluctant to increase transfers without more input into how and where the money will be spent (Canadian Intergovernmental Conference Secretariat 2000; Health 58 Canada 2003). This may be a hangover from the move to block funding with the introduction of EPF, which removed the federal government's direct oversight of compliance with legislation. Whatever the cause, the consequence is an environment that is more hostile to federal initiatives to mandate or enforce principles. A third theme, related to but distinct from accountability, is the commitment to access to health care services on "equal terms and conditions", to use the words of the 1950s. In the middle decades of the 20 t h century, this idea was presented as a justification for the creation of a universal health insurance system, covering first hospital and then hospital and physician services. In later decades, the expression of this intent became the provision of health care services based on need rather than ability to pay; different expression, same intent. This principle of equity has emerged as one of the Canadian public's fundamental understandings of how they wish their health care system to operate. There is not, of course, universal acceptance of this principle, or of the way in which it is expressed in current Canadian health care systems. Competing interests, or in the taxonomy of the first part of this chapter, competing political philosophies, are always present. Individuals with different philosophies are in competition to try to exert influence over public policy decisions. This creates a natural tension in health care; a largely public enterprise is subject to ongoing debate because of contested philosophical views over system objectives and ways of meeting them. These differences, in turn, are linked to different interests, and what they hope a health care system will achieve, for them. This perhaps explains why the public is consistent in ranking health and health care among its top priorities for politicians to address (Mendelsohn 2004). Regardless of the competing philosophies, however, consultations with and surveys of the Canadian public suggest an ongoing commitment to access to health-care services based on need rather than ability to pay (Maxwell et al. 2002; Mendelsohn 2004; Romanow 2002). Finally, a fourth theme is a more slippery view of the equity objective related to financing of health care systems. The impetus for development of health insurance was to spread the financial risk, which would both ensure that people could access the services they needed, and that those needed services would not cause financial ruin. This can be seen in the cooperative actions of the 1910-1930 period, where municipalities hired physicians and prepaid hospital services. It is equally apparent at the end of the century in two federal reports released in 2002, which included recommendations for financing through taxes or a dedicated health premium. The implications of the insurance perspective, or the pooling of risk in health care, is that the primary objective of the financial structure is to transfer income from the more healthy to the less 59 healthy. The primary equity principle of provision of services based on "need rather than ability to pay" only requires the separation of financing and delivery; it does not require (even though it sounds like it might) that financing be based on ability to pay. In fact, early proposals for and implementation of hospital and physician insurance schemes based on flat-rate charges mean that lower income individuals paid a larger proportion of their total income to finance health care services. The switch to universal programs started by the HIDSA changed this, introducing at least a component of financing from government general revenues. This change does not, however, necessarily signal a shift to progressive financing overall, where the proportion of income paid increases as income increases (the kind of financing most often thought of as reflecting "ability to pay"). Government revenues used to finance the health care system derive from a variety of sources such as income and consumption taxes and health care premia, each of which may be more and less progressive. The separation of financing from use in health care, and the generation of financing from general revenues, makes health care financing subject to general public policy rather than health care-specific policy. The commitment to equity in Canada is thus firm, if not unanimous. It is based on a strong sense of what equity means in the use of health care services, and a weaker and potentially more malleable sense of what equity means in raising the money to pay for those services. Some might describe Canadian health care systems in Marxist terms, operating on the principle of "from each according to his ability, and to each according to his need", but this is really only half right. From each according to his ability emerges only in the Romanow Commission report. Legislation and policy documents prior to this were more likely to suggest equal charges per person, or to suggest that federal contributions be derived from general revenue, with the provincial contributions left to each province's discretion. To each according to his need, on the other hand, is the primary equity principle of systems across the country. In terms of political philosophies, the Canadian medicare system is closer to Dworkin's notions of the role of a "prudent insurance package" to spread risk than they are to Nozick's pure market ideal. The emergent principles of equity, however, only describe the stated intention of policies governing the health care system and not their actual impact. Moreover, the legislation and major policy documents reviewed here are quite separate from the many governing policies that are implemented in the health care system on an ongoing basis. Acute care beds are closed, methods of physician remuneration are modified, changes to the tax system are, and all of these policy changes have the potential to affect the equity of the health care system. 60 The review of political philosophies made clear that equity is in the eye of the beholder, tied to what the "beholder" believes should be made equal to achieve his or her version of fairness. The review of legislation and policy documents shows that equity is also multi-dimensional. It would not be possible to summarize equity in a single measure. It is quite possible, even probable, that equity might be achieved in some aspects of health or health care and completely missed in others. The review also suggests that there are three major areas within which equity is of interest: health status of individuals and populations, both before and after the influence of the health care system; the financing structure of the health care system; and health care services utilization. Every policy decision has the potential for distributional consequences (Evans 1997). The purpose of the analyses in the next three chapters is to provide an empirical base for these constructs. More precisely, these analyses will look at what impacts a set of health care-related policies in British Columbia have had on equity in health care services use and equity in health care financing, in light of the historical and philosophical underpinnings of the system in that jurisdiction. 61 Introduction to Chapters 3, 4 and 5 The previous chapter showed that equity in some form is present in all health care systems. Systems can be distinguished by the particular vision of equity to which each is aspiring, and then whether or to what extent the system actually meets its objectives. Equity can be considered from several different angles. The focus here is on equity in need for health care services, in the use of health care services, and in financial contributions to the health care system, although there are several other angles, such as the availability of health care services, which could have been considered. The empirical work begins with an introduction to the measurement of (in)equality and (in)equity using analyses derived from the health economics and econometrics literature. The analyses follow the tradition established by a group of researchers based in Europe, and in particular methods developed by Eddy van Doorslaer and Adam Wagstaff, among others through the ECuity project.45 These methods are then used to quantify inequalities in the distribution of income, the inequalities in the distribution of overall need for health care services, and income-related inequalities in need, health care services use and health care financing in British Columbia, in 1992 and 2002. The approach taken here differs from the analyses of the ECuity group in two major respects; the present analyses use administrative data instead of survey data, and the British Columbia data used for this project are population-based rather than sampled. While these differences both represent potential improvements to the ECuity project approach, the use of administrative data is not without its downsides. The administrative data must also be used to develop a measure of need for health care services - there is no source for self-reported health status for every member of the population. In addition, and perhaps more importantly, the income variable available for this project was developed from neighbourhood-level rather than individual-level data. Each of these measures is described at the beginning of the chapter in which it is first introduced; the implications of each variable's use is discussed following the analytic results in each chapter. Information about this project and the research that has come out of it can be found at Accessed 11 October 2005. 62 Chapter 3: The distribution of income and need for health care services 3.1 I n t r o d u c t i o n This chapter is devoted to measuring inequalities in income and the need for health care services, and income-related inequalities in need in British Columbia, in 1992 and 2002. The major purposes of the chapter are to introduce the methods that will be used in the next three chapters, and to describe the measures of income and need for health care services that are central to this work. It is not the intent of this chapter to identify inequalities in need that the health care system is solely responsible for addressing. In that sense, this chapter will not conclude with an assessment of equity in health as it relates to the health care system. Instead, what is presented here will provide a picture of the context within which the health care system operates, both in the health status of the people the system in British Columbia is designed to serve, and in the economic conditions of the population from which funds for the system are derived. The specific hypotheses that are addressed are: • The distribution of individual-level need is very unequal across individuals in British Columbia in 1992 and 2002; • The distribution of need did not change significantly between 1992 and 2002; • The distribution of income became more unequal in British Columbia between 1992 and 2002; • There is an income gradient in the distribution of need across individuals in British Columbia; • The relationship between income and need did not change significantly between 1992 and 2002. 3 . 2 M e t h o d s I - m e a s u r i n g i n e q u a l i t y There are three main sources of data used for analyses in the present chapter: the Canada Customs and Revenue Agency annual series on income statistics; a custom tabulation of data produced by Statistics Canada based on tax returns; and a measure of need for health care services derived from health care utilization files in British Columbia. Before providing the 63 details of these data sources, however, the measure of inequality that will be used throughout Chapters 3 through 5 is introduced. 3.2.1 A summary measure for inequality There are several standard approaches to measuring income inequality. The most commonly used of these measures is the Gini coefficient, which is derived from the Lorenz curve. A Lorenz curve is a line that represents the cumulative percentage of the population and the corresponding cumulative percentage of income (measured along the x- and y-axes respectively), where the population is ranked from lowest to highest income (Figure 2). The Lorenz curve representing perfect equality - the case in which every person has the same income - is shown by the diagonal line. In this case, 10% of the population has 10% of total income, 20% of the population has 20% of total income, and so on. In reality, the Lorenz curve for a given area or for a given population will lie below the line of equality, since the distribution of income in any population is unequal, and people are ranked from lowest to highest income (the first 10% of population along the x-axis is that 10% of the population with the lowest incomes, and so on). The extent of deviation of the actual curve from the diagonal line represents the deviation of the distribution of income from that of perfect equality. Figure 2: Hypothetical Lorenz curve of income inequality 1.00 -, 0.90 • 0.80 • £ o 0.70 • u C «-o 0.60 • c O l u aj 0.50 -o. ai > 1 0.40 -E 3 O 0.30 • 0.20 -0.10 -0.00'-0.00 0.10 0.20 0.30 0.40 0.50 0.60 0.70 0.80 0.90 1.00 Cumulative percent of population 64 The Gini coefficient is then an index that quantifies this deviation (the shaded area between the diagonal and the Lorenz curve). Formally, the Gini coefficient is defined as the ratio of the area between the Lorenz curve and the line of equality to the total area under the line of equality. The computation of the Gini coefficient can be written mathematically as a function of an integral, as: 3.1 G = l-2j[°Fi(jc)/(jc)fl6c, where Fi(x)f(x) defines the area under the Lorenz curve. In other words, the Gini is 1 - twice the integral under the Lorenz curve, when the Lorenz curve is defined within the unit square. In the case where all individuals have the same income, the Lorenz curve coincides precisely with the line of equality, and the Gini coefficient is 0. At the other extreme, where one individual controls all the income, the Lorenz coincides with the x-axis (increasing proportions of the population all have 0% of total income), until the last point, which is on the line of equality, meaning the Gini coefficient is 1. The Gini can thus range from 0 to 1, though in most cases, income distributions in developed countries are well under 0.5. Anand (1983), measuring inequality and poverty in Malaysia, provides in an appendix proofs of the equivalence of several standard approaches to calculating the Gini coefficient. The first of these is: 3.2 Gi = l-X(F/ + . - F / ) ( ^ + i + ^ ) , /=o where F refers to the x variable (population) and (|) refers to the y variable (income). The summation part of this equation calculates the area of a series of rectangles based on adjacent x and y values on the Lorenz curve. This is shown in Figure 3, below. In this case, hj+i — Fj is the horizontal distance of 0.1 (0.6 - 0.5), and <j)j+i + (j); is the height of each leg of the rectangle, as drawn. Assuming that the x variables are sufficiently close to each other that the Lorenz curve between the two is effectively a straight line, the Lorenz curve will cut the rectangle exactly in half; the Lorenz curve is the diagonal of the centre square (un-shaded) that sits between two rectangles of height §\ and width Fj+i - Fj (the shaded areas). With this neat construction, the summation of the areas captured by successive rectangles will equal exactly two times the area under the Lorenz curve. Hence, the definition above, that the Gini coefficient is equal to 1 minus this summation. 65 Figure 3: A method for calculating the Gini coefficient 1 q \ 0.5 R 0.6 F M Cumulative peicent of population While this approach is clear, and is built solidly on basic procedures in calculus, it is but one of several different ways that the Gini coefficient can be calculated. Another approach is: 2 3.3 G = l £ ( 1 - * / ) ; > » M" ,=i where ju is equal to the (population) mean of y (see, for example, Wagstaff 2002). This particular form of the Gini coefficient is worth consideration because it moves us along to one that is simpler from a computational standpoint, especially for large data sets. Equation 3.3 can be re-written as: 3.4 G = -—Y&-Ri)yi +—xy> jun (=| ju where x is the mean of the ranking term (1-Ri), which is one-half, and y is equal to //. The mean of the ranking term must be equal to one-half because by construction the ranking variable takes values from 0 to 1. The equation can then be simplified to: 3.5 ( " A - X V Daugherty46 shows that the equation inside the large square brackets is a variant expression of the covariance between x and y. This means that the Gini coefficient can be calculated as: 4 6 See Chapter 1, "Alternative expressions for the sample covariance". Accessed 4 July 2005. 66 2 3.6 G = cov(x,^) M as several authors have suggested (Anand 1983; Jenkins S 1988; Lerman and Yitzhaki 1984).47 This is referred to as the "convenient covariance" approach to calculating the Gini coefficient, because as noted above it is computationally much more straightforward, especially for large data sets, than are equations 3.2 or 3.3. The one remaining consideration in proposing an approach to computing Gini coefficients is that of standard errors. It will be useful, for example, to be able to ascertain that changes in Gini coefficients over time are more than might be expected based on random fluctuations between years. Building on the convenient covariance, Kakwani et al. (1997) propose a "convenient regression" approach, based on ordinary least squares regression, as: 3.7 2cr2R — =a + pRi + w, where &'R is the variation of the rank variable. The estimator of p equals: B = — 2>-//)(*/-i) np ,=i which can be seen to be a variant of the formal statement for the calculation of the Gini coefficient based on the covariance of x and R, since the mean of R equals xh. In other words, the beta coefficient from a regression of a scaled variance on the ranking variable provides an estimate of the concentration index. If the regression estimates are corrected for auto-correlation in the ranking variable and heteroskedasticity, then the standard errors from the regression are the standard errors for the concentration index. This sort of correction is provided by Newey-West estimators, in this case using a lag of 1 (Wagstaff et al. 2005b). This is the approach that will be used throughout Chapters 3 through 5 to calculate Gini coefficients and their standard errors. 3.2.2 Why Gini coefficients? Before proceeding, it will be useful to provide more justification for the choice of Gini coefficients as a basis for measure of inequality. There are several characteristics that, generally speaking, are desirable in any measures of inequality. First, a measure should be proportionate. If a measure is proportionate, an equal proportional addition of income to each person will not The steps here are shown not to repeat proofs already done, but to provide what I believe is a more straightforward proof of the covariance formula than has been offered elsewhere. 67 affect the calculation of the Gini coefficient (Chakravarty 1988). Similarly, an equal proportional addition of individuals to each income group will not affect the calculations. Second, measures of income inequality are more robust i f they can be decomposed, allowing the identification of sources of change in the measure over time. The Gini coefficient meets this test, though these decompositions are not pursued in the present analyses. The use of Gini coefficients in the present analyses does, however, set up the possibility of such future investigation. Measures of inequality should satisfy the Pigou-Dalton principle, that a transfer of income from a poorer person to a richer person will necessarily increase the inequality measure, and a transfer from a richer person to a poorer person will decrease that measure (Chakravarty 1988). In the case of the Gini coefficient, this will be true as long as the income-rank of individuals is unaffected by the transfers.48 Gini coefficients meet these tests, and furthermore have some history of being used to develop measures of (in)equality and (in)equity in health status, health care services use and health care financing. The present work thus builds on an already established field of inquiry that has seen a great deal of interest and development of methods over the last decade or so. Finally, it is worth emphasizing that Gini coefficients are relatively easily explained, and at least somewhat intuitive, given their relationship to Lorenz curves. This fact will be quite important given the ultimate (though not immediate) goal of producing measures of equity that can be used and understood by health care managers and the general public. 3.2.3 Preference weightings in the Gini coefficient The issue of weighting comes up in the literature on Gini coefficients and measurement of inequality more generally. There are three different statements made about the Gini coefficient. The first is that the Gini is more sensitive to changes in income in the middle of the income distribution (Anand 1983; Giorgi 1999). The second is that the Gini has a built-in social preference that gives more weight to lower income households than it does to higher income households (Wagstaff 2002). The third is that the Gini coefficient embodies no particular preference or set of preferences (Duclos 2000). Regardless of whether the "standard" Gini coefficient has built-in social preferences, the work of several authors (Kakwani 1980; Lerman and Yitzhaki 1984; Yitzhaki 1983) shows that the calculation of the Gini can be generalized so If income ranks change it is still possible to calculate a Gini coefficient, but the change in rank has to be taken into account. 68 that such weightings can be adjusted to reflect different preferences for or importance of inequality at different points in the income distribution. The generalized Gin i as proposed by Yitzhaki (1983) and used by Wagstaff (2002) is: where v is the parameter that can be modified to reflect different preferences with regard to inequality, p is the ranking (x) variable, and L(p) is the Lorenz curve. When v=2, this equation reduces to the standard Gin i coefficient. Wagstaff (2002) then presents this after integrating as: where w;(Rj,v) = v ( l - R j ) v l , which is the weight attached to the z'th person's share of total income (yi/nu). Based on this, and considering v parameters of 1.5, 2 and 4, the implicit weights of the (standard and extended) Gin i coefficient are shown in the graph below for a fictitious population of ten individuals. v>l G(v) = 1 - — X (1 - pT'yi= 1 - X -^-Wi(Ri, v) 69 Figure 4: "Weights" when the parameter v is varied from 1.5 to 2 to 4 3.5 3 The solid line is that for v=2, the standard Gini, and shows weights decreasing from (nearly) 2 to (nearly) 0 as one moves from the lowest to the highest incomes in the population (in this case assuming a population of 10 individuals). Wagstaff then interprets (his version of) this graph as implying that the standard Gini coefficient, represented by the v=2 line, is more "concerned" with deviations from an equal distribution of income among the poor, and less so with deviations among higher incomes. This interpretation is not, however, consistent with that of other authors, who characterize the v=2 situation - the standard Gini - as one where an equal weight is given to each individual. In other words, the standard Gini carries no implicit social preferences (Chakravarty 1988; Yitzhaki 1983). Wagstaff s interpretation of the "weight" part of the equation as inferring some sort of social preference also does not make intuitive sense i f one considers that the Gini is nothing more than an integral and therefore cannot embody social preferences, unless the Lorenz curve does as well. The disagreement in the literature revolves around differences in opinion about how one defines the "weight" given to each individual's place in the income distribution. This is an important question because the choice of a measure of inequality should be made with an understanding of any implicit social preferences embodied in that measure. If the standard Gini coefficient does, indeed, weight changes in the lower end of the income distribution more heavily than in the upper end of the income distribution, we need either to acknowledge that, or to use a modified Gini that will reflect our social preferences. In any case, the choice should be made explicit. The effects of changes in income distributions on resulting Gini coefficients can be tested quite easily with a fictitious population of 10 individuals. In our fictitious population, the lowest income individual receives 1% of total all income, the next individual receives 3%, the next 5%, all the way up to the highest income individual receiving 19% of total income available to the 10 individuals (Table 3). We then make six different modifications to the income distribution, all based on moving 2% of total income from one person to another. The first two modifications are between the tails of the distribution, transferring income from the 2 n d to the 9 t h individual, and then from the 9 t h to the 2 n d individual. The next two modifications transfer from the low end of the distribution (the 2 n d person) to the middle (the 6 t h person) and from the high end of the income distribution (the 9 t h person) to the middle (the 5 t h person). The third set of transfers is around the middle of the distribution, up from the 4 t h to the 6 t h individual, and down from the 7 t h to the 5 t h individual. If the Gini coefficient is more sensitive to changes in income of lower income individuals, the movement of income from the lower income individuals should increase the Gini (meaning a more unequal distribution) more than the movement of 2% of total income from the higher income individuals decreases it. In other words, for each pair of transfer, the Gini coefficient would be different. If the Gini coefficient is more sensitive to transfers around the middle of the distribution, then the third set of transfers should produce a larger change in the Gini coefficient compared to the two other sets of transfers. 71 Table 3: Initial and modified income distributions in a 10 person society Transfer between tails Transfers from a tail to the middle Transfer from and to the middle Initial income Individuals distribution 2nd to 9th9th to 2nd 2nd to 6th 9th to 5th 4th to 6th 7th to 5th 1 0.01 0.01 0.01 0.01 0.01 0.01 0.01 2 0.03 0.01 0.05 0.01 0.03 0.03 0.03 3 0.05 0.05 0.05 0.05 0.05 0.05 0.05 4 0.07 0.07 0.07 0.07 0.07 0.05 0.07 5 0.09 0.09 0.09 0.09 0.11 0.09 0.11 6 0.11 0.11 0.11 0.13 0.11 0.13 0.11 7 0.13 0.13 0.13 0.13 0.13 0.13 0.11 8 0.15 0.15 0.15 0.15 0.15 0.15 0.15 9 0.17 0.19 0.15 0.17 0.15 0.17 0.17 10 0.19 0.19 0.19 0.19 0.19 0.19 0.19 Gini coefficients 0.331 0.359 0.303 0.347 0.315 0.339 0.323 Change in Gini 0.028 -0.028 0.016 -0.016 0.008 -0.008 Table 3 shows that, in fact, each of the three pairs of transfers cause precisely the same magnitude of change in the Gini coefficient. The signs of the change are as expected given that Gini coefficients are consistent with the Pigou-Dalton transfer principle. The same table also shows that the Gini coefficient is not, in fact, more sensitive to changes in the middle of the income distribution. The size of the change in the Gini is greatest for transfers between tails, decreases for the transfers from the tails to the middle, and is smallest for transfers around the middle of the income distribution. A l l of these scenarios involved moving 2% of total income, which suggests that the Gini coefficient is, indeed, less sensitive to changes around the middle of the income distribution. Why this is so is more obvious when one examines the related Lorenz curves for the initial income distribution and the distributions that result from transferring between the tails (Figure 5) and around the middle (Figure 6). In Figure 6 the three Lorenz curves overlap for the first 30% and last 30% of the income distribution, while the overlap in Figure 5 is only the first and last 10% (these overlaps are also evident in Table 3). The total movement of income in both sets of transfers is identical, but because in the latter transfers income is only moved in the middle of the distribution, the Lorenz curve is only modified in that part of the distribution. The result is a smaller change in Gini coefficients. 72 Figure 5: Lorenz curves for the baseline distribution and transfers between the tails Initial income distribution 2nd to 9th 9th to 2nd 1 2 3 4 5 6 7 8 9 10 Cumulative percent of population Figure 6: Lorenz curves for the baseline distribution and transfers around the middle — Initial income distribution — 4th to 6th — 7 t h to 5th Cumulative percent of population Love and Wolfson (1975) suggest that the impact of a. transfer of income on the Gini coefficient depends on two factors: the amount of income transferred and the number of individuals between the two transfer points. The latter suggests that the Gini coefficient will be more sensitive to changes around the mode (rather than the middle) of the income distribution, because it is there where a greater number of individuals will be between transfer points. In any case, there neither Figure 5 nor Figure 6 appears to weight changes in the lower end of the income distribution more heavily than changes in the upper end of the distribution. The "standard Gini" is the one used in all subsequent analyses. 3.2.4 Concentration indices In addition to the Gini, this work will also use a generalization of it known as a concentration index. The Lorenz curve shows what Wolfson and Rowe (2001) refer to as a univariate relationship - the distribution of income within the population. Often one is interested "bivariate relationships", for example, the distribution of health status among the population, where the population is ranked by income. The two measures - the distribution of income and the distribution of health status by income - can be calculated the same way. For the former there is a Lorenz curve, or the cumulative distribution of income against the cumulative distribution of the population, with the population ranked from lowest to highest income. For the latter there is the concentration curve, where individuals are also ranked on income, and the x-axis represents the cumulative distribution of the population, but it is the cumulative distribution of health status rather than income that is on the y-axis. The relationship between the concentration index and concentration curve is exactly the same as the relationship between the Gini coefficient and the Lorenz curve. The concentration index is calculated in the same way as the Gini coefficient. The difference is in what is being measured rather than how it is measured. This is a useful generalization, as it means that all of the formulae and discussion about the Gini coefficient apply equally to the concentration index. The one difference between Gini coefficients and concentration indices is that while the Gini coefficient varies between 0 and 1, the concentration index varies between -1 and 1. The concentration index can be negative because it is possible to measure the distribution of something that is more heavily concentrated among the lower income groups. A good example of this situation, and one that will be used here, is measuring health as a negative, meaning lack of health, or i l l health, rather than as a positive, meaning good health. Ill health is, in fact, more 74 heavily concentrated in lower incomes groups, which places the concentration curve above rather than below the line of equality. 3.3 Methods II - quantifying income distribution in British Columbia 3.3.1 Income statistics from the Canada Revenue Agency Each year, the Canada Revenue Agency produces a series of publicly available data based on tax returns.49 The final version of each yearly installment to this series, which is intended to create a profile of Canadian taxpayers, is based on a stratified random sample of all individual tax returns filed.50 One of the tables in this series provides information on most major fields in the standard income tax form, including sources of income (e.g. employment, government transfers), deductions and taxes paid. The information is organized into 17 income bands, with the width of income bands increasing from $5,000 at the lower end of the income spectrum to $10,000 in the middle section and a maximum income group of $250,000 and over.51 These data are thus quite amenable to the construction of Gini coefficients for income distribution. The measurement of health and income-related health inequalities requires an ability to define both income status and health status for individuals or groups of individuals. This requirement to connect measures of income status and health status eliminates the potential to use the Canada Revenue Agency data directly in these analyses. Because of concerns about the protection of privacy, only geography-based rather than individual- or family-based measures of income were potentially available either through census data or a special tabulation of tax-filer data held by Statistics Canada. The Canada Revenue Agency data are useful, then, in that they represent a gold standard for the income distribution, and will provide an outside estimate of the magnitude of effect on Gini coefficients of moving from an individual to an ecological measure of income. Income statistics for the years 2000 and later are available on the Canada Revenue Agency website, (accessed 8 July 2005). Data for years prior to 2000 are available from the Agency on request. 5 0 This information is from the Canada Revenue Agency website quoted above. 5 1 The 1992 data actually were organized into 20 income groups. These data were aggregated so that the income groupings were consistent in both the 1992 and 2002 data. 3.3.2 Choosing an ecological measure of income Previous international comparative research in this area generally has used surveys to identify self-reported health status, health care services use and income. The exception to surveys was the use of tax information linked to surveys in some countries in these multi-country comparisons (Wagstaff et al. 1999). One purpose of the present research was to test the feasibility of using non-survey data sources to conduct analyses of (in)equality and (in)equity. The fact that individual-level data on income were not available for this project means that the approach here differs in both the source of data - administrative data vs. survey data - and in the level of measurement of the income variable - individual vs. ecological. Measurement of income at an ecological level has been shown to be a robust approach for analyzing socioeconomic patterns in health status and health care services use (Mustard et al. 1998), but there are limitations to this approach. The most important of these limitations is that the same income band must be assigned to all individuals and households resident in a particular geographic area, implying income misclassification of some (unknown) proportion of the population. Higher-income residents of poorer neighbourhoods, for example, w i l l be assigned to the same income group as their lower-income neighbours, and vice versa. It might be argued, then, that building the measure from units of geography that are as small as possible makes the most sense, as smaller geographies w i l l tend to be more homogeneous in income distribution, 52 which therefore w i l l minimize the proportion of individuals with misclassified income. Another possibility, however, is that smaller geographic areas may tend to represent people who are more alike in other ways, for example, populations that are more similar genetically because of clustering of ethnic groups by geography. Comparisons of these small areas then run the risk of real genetic differences confounding an apparent difference in health status by income group. This risk increases as the number of income groups increases. The use of deciles, in other words, would minimize the possibility that any differences by income were actually due to factors other than income. On the other hand, the use of deciles in the calculation of Gin i coefficients and concentration indices would likely diminish the Gin i coefficients and concentration indices; aggregation to 10 rather than (say) 1,000 income bands represents a loss of information on income distribution. Smaller geographies will be more homogeneous, but will still represent different individuals. If a neighbourhood gentrifies, for example, even a single block might contain lower income individuals who moved in prior to gentrification, along with higher income individuals who moved in later. In addition, the increasing prevalence of owner-occupied dwellings with income suites also produces a mix of incomes even within the same dwelling. 76 This means a classification decision must be made both with respect to the geographic unit used for analysis (postal code or dissemination area) and the number of bands ultimately created and used in analysis. The decision must strike a balance among the competing threats to the validity of the research results. Dealing first with geography, the choice of an income measure was between geography-based data from the census, and a geography-based custom tabulation of tax-filer data held by Statistics Canada. The smallest unit of geography available for the census was the dissemination area53 while the Statistics Canada tax-filer data could be analyzed (by Statistics Canada analysts) at the 6-digit postal code level. There are over 100,000 postal codes in British Columbia and about 7,000 dissemination areas (Canada.Statistics Canada 2004b), implying that the two sources will produce different results. Given the discussion above, it appears to make the most sense to use the small-area (i.e. postal code) approach available with the tax-filer data rather than the census data.5 4 Any threat to analyses based on these small areas sorting people into similar genetic pools can then be assessed by varying the number of income bands used to calculate the Gini coefficients and concentration indices.55 3.3.3 Income statistics from Statistics Canada tax-filer data A request for a custom tabulation of data from Statistics Canada required identifying the variables of interest related to income, and government transfers and taxes, as well as the process by which household-level data were to be aggregated to 1,000 income bands.56 The final list of income-related variables is provided on Table 4 and includes market income, income from government transfers, total income, federal and provincial income taxes, and disposable income. These concepts are linked in the order in which they are listed. Market income is the money accruing to people though employment and investments. Government transfers, such as public pensions and employment insurance, are added to market income to calculate total income. The Starting with the 2001 census, Statistics Canada reported census data using dissemination areas, while still collecting data using enumeration areas. There are small differences between the two, so for ease of reading, we refer to dissemination areas. Another drawback to census data is that they are collected only every five years, in years ending in "1" and "6" (e.g. 1996, 2001). Data collected in 2001 reflect income earned in 2000. Years of analysis for this project are 1992 and 2002, which means that using the census data would require an assumption that income distributions in 1992 and 2002 were similar to those in 1990 and 2000, as collected by the census. Ultimately, this decision has little impact on the Gini coefficient for income-related health status, as will be shown below. The request included clearly specified assumptions about what was available from the data and details about why each step and each variable were required. The details were intended both to ensure that we made no major mistakes in the initial submission and to encourage analysts at Statistics Canada to make suggestions for change or clarification if necessary. An iterative process over several months between the researcher and Statistics Canada ensured that this was, indeed, the case. A single person at Statistics Canada who liaised with the researcher and the Statistics Canada analyst coordinated the request. Statistics Canada provided a mock table outlining the steps to be taken for the production of data prior to finalizing the request and writing the contract with the researcher. This process proved invaluable as it raised some issues in the request that we then had the opportunity to address. 77 difference between total income and disposable income is determined by the amount of taxes paid to provincial and federal governments. Disposable income, the income individuals have available to spend, is the income concept used here to quantify and rank incomes so that people can be arranged along the x-axis of the Lorenz or concentration curve. Table 4: Income, government transfers and tax information requested Variable Definition Market income Derived: Total income that comes from employment, investments, business, farm, fishing and so on. It is all income accruing from private (as opposed to government) sources.. Income from government transfers Public pensions, child tax credits, employment insurance and so on Total income Market income plus government transfers Provincial taxes Income taxes paid by individuals / households to the provincial government Federal taxes Income taxes paid by individuals / households to the federal government Disposable income Total income less provincial taxes and federal taxes Families to postal codes Statistics Canada holds tax-filer data organized at the family level; individuals are merged into families using the fields on tax forms that identify a filer's spouse and dependents.57 This is a useful arrangement of the data, because income earned and taxes paid are most usefully thought of as household concepts, with households ranging in size from one member on up. There is, however, an important conceptual difference between families and households. Families, or more specifically in this case census families, are groups of individuals living together and related by blood, marriage, common law or adoption (see, for example, Statistics Canada 2003). In some cases, households and census families are the same thing, but in other cases a household may include multiple census families, as in the case of unrelated people sharing a house or an apartment. The implications of using families instead of households are explored below. The first step in the data request was then to aggregate or process the data from this family level to a postal code level. 5 8 This step had to consider two different types of aggregation. One is Statistics Canada used the census family definition. This definition of the family classifies people in the following manner: 1 ) couples (married or common-law) including same-sex couple living in the same dwelling, with or without children; and 2) lone-parents (male or female) with one or more children. The residual population is called "non-family persons" and is made up of persons living alone and of persons living in a households who are not part of a couple family or lone-parent family. This is, as mentioned, the smallest geographic unit that is maintained by Statistics Canada in these data. It is also the smallest geographic unit available for the health status and health care services use variables, to which this income variable must be connected. Postal codes will thus be the "bridge" among the different sources of data. 78 the equivalization of individuals within families and the other is the weighting of families within postal codes. The former is always a concern when trying to make comparisons among different families; ranking people on income generally means trying to find some way of making income comparable among "units". The weighting of families is relevant in general as it indicates how the population is to be cumulated along the x-axis of the Lorenz curve. In the case of ecological measures of income, family weighting also must be considered when calculating a comparable income for each geographic unit - in this case, 6-digit postal codes. A l l calculations that will be described were conducted using the disposable income variable. Adjusting for family size - "equivalizing income" - is meant to make incomes across families more comparable, taking into consideration that there are fixed costs of running a household, so that additional family members do not represent linear additions to cost.59 A couple, for example, does not require twice the income of a single person to achieve the same living standard (Ebert 1999). In all cases, the object is to create an "income per equivalent adult" for each family, which then allows a ranking of families from low to high income. There are a wide variety of equivalence scales available. In their early work, the ECuity group compared the progressivity of health care financing among countries while equivalizing income in those countries using locally defined methods (Wagstaff and van Doorslaer 1992). This meant that the equivalization methods were different across countries. In more recent work the group reconsidered the benefit of this approach, and chose instead to use a common equivalization metric. In a series of papers on financing (van Doorslaer et al. 1999; Wagstaff et al. 1999), this same group chose an equivalization metric recommended by Aronson, Johnson and Lambert based on an analysis of U K tax data. The generalized form of this metric is: E = (A + OKf, where A is the number of adults and K the number of children in the household. The specific form used by the ECuity group in 1999 was to set both rj) and 0 to 0.5. This group modified its approach again for a later series of papers on health status and health care services use (van Doorslaer et al. 2004a; van Doorslaer et al. 2004b; van Doorslaer and Jones 2004; van Doorslaer and Koolman 2004). In these papers, the metric was the "modified OECD equivalence scale", which set to the value of the first adult (defined as all individuals 14 years of age and over) as 1, each subsequent adult as 0.5 and all children (aged under 14) as 0.3. 5 9 It could be argued that it would be better to aggregate individuals to households rather than census families, because unrelated people sharing living space can also benefit from economies of scale. Using families may thus understate income in younger age groups, where there is more likelihood of living in non-family situations. It was not possible, however, to request the Statistics Canada by analyzed at the household level. 79 In Canada, Wilkins uses an equivalence scale that considers the size of the household without respect to the distribution of adults and children (Wilkins 2001). These scaling factors are based on Statistics Canada Low-Income Cutoffs, which in turn are constructed based on routine family expenditure surveys.60 The family expenditures surveys provide an average percentage of household income that is spent on food, clothing and shelter. This average, based on the 1992 survey, was 34.7%. The low-income cut-off is then set (somewhat arbitrarily) at 20 percentage points above this average. In other words, the average income of families that spend more than 54.7% of income on food, clothing and shelter is the low-income cut-off. There are adjustments made to this for family size and for size of community of residence, reflecting economies of scale within households, and differences in cost of living based on area of residence. The 2001 equivalence values, based on the low income cut-offs for different family sizes, were: 1.25 for 2 persons, 1.55 for 3 persons, 1.95 for 4 or 5 persons, and 2.44 for 6 or more persons sharing the same household (Wilkins 2001).61 Ebert provides a worked example of how the choice of equivalization approach influences the calculation and interpretation of household-level incomes. Consider a community that consists of two households, one a single person and the other a couple without children. In all cases, the household is the unit of analysis. The importance of this example is that the households have different compositions; the result does not depend on the particular household comparison chosen. The income distributions for this two-person community in Year 1 and Year 2 are shown in Table 5. The table also shows equivalized income calculated based on three different computation scenarios. In all cases, the ratio is calculated with total income in the numerator and the "equivalized" number of household members in the denominator. The first scenario is the unity or no equivalization approach. In this case, each household is given a weight of one, so total household income is divided by one. In other words, the total income of households is not adjusted based on the size of the household. The second scenario is the per capita equivalization approach, where total household income is divided by the number of (real) people in each household. In this case, there is no assumption of economies of scale; a household with four people would need four times the income of a household with one person in order to have "equal" income. povbk.htm. Accessed 3 July 205. 6 1 None of these equivalence scales vary with family income. In other words, there is no adjustment for the fact that patterns of expenditure change with increasing income. While it is theoretically possible to build such a scale, they are uncommon in analyses. 80 The third scenario is the equivalent adult approach, where total household income is divided by an adjusted number of household members, using an equivalization metric as described above. For this example, we use a weight of 1.0 for the first member of each household and 0.5 for the second adult. Table 5: Total income and adjusted income by household, year and equivalization method Single Couple Yearl 5 10 Year2 7 8 Household weighting Yearl 5 10 Year2 7 8 Per capita weighting Yearl 5 5 Year2 7 4 Equivalized adult weighting Yearl 5 6.67 Year2 7 5.33 Comparing the Yearl values in Table 5 shows that the different equivalization approaches will lead to different conclusions about the equality of the distribution of incomes across the two households. The "no equivalization" approach makes it appear that the couple is far better off than the single. With the "equivalent adult" approach, the couple is still apparently better off, but the degree of inequality between the households is much smaller. Where "per capita equivalization" is used, the single and couple appear to have equal income. As expected, the equivalization result lies between the household income and per capita income results. Between Year 1 and Year 2 there was a shift in income (of 2) from the couple to the single. Table 5 makes clear that conclusions about the impact of this income shift will be affected by the equivalization approach through which one chooses to compare the households. In the case of household weights (no equivalization), the inequality is diminished, but still favours the couple. In the per capita equivalization approach, what was formerly a state of equality is shifted to inequality favouring the single. And where equivalent adults are considered, the inequality is shifted from a situation favouring the couple to one favouring the single. Ebert and others (Ebert 1997; Ebert 1999; Van Ourti 2004) make the point that when constructing Lorenz curves it is also important to consider how much weight is given to each household. There are three choices available: each household is given equal weight (which is effectively the same as ignoring the issue entirely), households are weighted according to the number of individuals in each (n), or the weighting is the number of "equivalent adults" in each 81 household (m). In general, \<m<n, with m moving closer to either unity or n depending on the particular equivalization approach chosen. The difference between this issue and that of equivalization is that equivalization determines how one can make incomes of different units comparable, while household weighting determines the proportion of the total population accounted for by each household. The two issues become difficult to distinguish because, as Ebert (1999) points out, it would be quite inconsistent to use an adult equivalent approach to compare incomes and then to weight each household as unity for purposes of constructing a Lorenz curve. The Lorenz curves for each household income equivalization and weighting approach can be drawn based on the information in Table 5, by transforming the data as shown in Table 6. The lines of zeroes are added only to force the Lorenz curves to start at the origin. The population column presents the proportion of the total population represented by each of the (two) households. In the case of household weighting, there are two households as the unit of analysis, compared to three individuals with per capita weighting and 2.5 equivalent individuals with equivalized weighting. The proportion of income held by each household does not change. What does change in Year 2 is that with the latter two measures (per capita weighting and equivalized weighting) the rankings of the two households reverse. Because the rankings reverse, the order they are represented on the Lorenz curve must also be reversed from Year 1 to Year 2, since the Lorenz requires ranking of individuals by income. These reversals are reflected in the cumulative population and cumulative income columns of Table 6. Table 6: Calculations required for the construction of Lorenz curves Yearl Year2 Income/ Income/ Total Cumulative Cumulative Total Cumulative Cumulative Population income Population income Population income Population income Household weighting 0.00 0.00 0.00 0.00 0.00 0.00 0.00 0.00 single 0.50 0.33 0.50 0.33 0.50 0.47 0.50 0.47 couple 0.50 0.67 1.00 1.00 0.50 0.53 1.00 1.00 Per capita weighting 0.00 0.00 0.00 0.00 0.00 0.00 0.00 0.00 single 0.33 0.33 0.33 0.33 0.33 0.47 0.67 0.53 couple 0.67 0.67 1.00 1.00 0.67 0.53 1.00 1.00 Equivalized adult weighting 0.00 0.00 0.00 0.00 0.00 0.00 0.00 0.00 single 0.40 0.33 0.40 0.33 0.40 0.47 0.60 0.53 couple 0.60 0.67 1.00 1.00 0.60 0.53 1.00 1.00 82 The Lorenz curves drawn from Table 6 are shown in Figure 7. In all cases the darker line is the line for Year 1 and the lighter line represents Year 2. As was seen in Table 5, the choice of equivalization method will have an important influence on the interpretation of income distribution and changes in income distribution over time. Figure 7: Lorenz curves based on different household weighting concepts Household weighting Per capita weighting Equivalized adults The choice of equivalization and weighting method depends on the particular interest that drives a comparison of households. If the perspective is an interest in understanding differences in control over income by households regardless of household size or composition (e.g. how many children there are) then the first approach of "no equivalization" makes the most sense. If instead the goal is to see the per capita distribution of income, again independent of the composition of households, then the per capita equivalization approach is the most appropriate. If the perspective is the distribution of living standards, then the equivalent adult approach makes the most sense. This latter approach acknowledges the economies of scale associated with larger households through normalizing income across households by constructing an income per equivalent adult value for each household (Ebert 1997; Ebert 1999). The present work uses the modified OECD equivalence scale to calculate comparable household-level incomes. This scale weights the first adult in a household (aged over 14) as 1, subsequent adults as 0.5 and each child (aged 14 and under) as 0.3, and was chosen for consistency with previous work of the ECuity group The ecological income variable was constructed by weighting each household by the number of its equivalent adults, as the interest was in making comparisons among households based on living standards. In other words, the income measure for a postal code would be equal to a weighted average of each household's income per adult equivalent, with the weight equal to the number of equivalent adults in each household, as: 83 where h refers to each household, I is income and E is equivalent adults. Writing this formula out should make clear that when using the equivalent adult approach, the computation reduces to total income in an area (in this case postal code) divided by total number of equivalent adults. In this sense, each postal code is treated conceptually as a single, large household. Van Ourti (2004) conducted a range of sensitivity analyses, comparing multiple equivalence metrics and the three household weighting approaches, using data from Belgium. These analyses showed that the choice of equivalence scale and weighting had no bearing on results, except in the case of social insurance. A similar set of sensitivity analyses for British Columbia would have been quite costly, because of additional processing requirements by Statistics Canada analysts, and in any case was beyond the scope of this study. The data request did include, however, income measures using the two other weighting schemes to allow for a sensitivity analysis of the weighting approaches. The three income measures were: jee I equivalized adult weighting — Aggregation from postal codes to income bands A postal code level measure of income was a necessary first step, but further aggregation was required. This is because privacy concerns prevent release of data at the postal code level. From here, however, the processing steps are much clearer, including: 1) sort postal codes by average weighted disposable income measure, as calculated above. 2) create 1,000 income bands: Most income analyses tend to use deciles or quintiles of income. A concentration curve-based analysis, where the information in the curve can be collapsed to a single index measure, allows the use of much finer-grained income bands. The number of income bands requested, in this case 1,000, was chosen to maximize the information available without creating any problems of potential for individual household weighting r * = | y A W per capita weighting 84 identifiability. Bands were created so that (approximately) the same number of (real, not equivalized) people were captured in each. The income bands do not contain precisely the same number of people because of variations in the number of individuals assigned to each of the postal codes in British Columbia. The B.C. population estimate from the 1996 census (about the midpoint of the years of interest) was about 3.7 million people and approximately 1.4 million private households (Canada.Statistics Canada 2004c). This means that with 1,000 income bands, each band will represent (on average) 1,400 households, and close to 3,700 individuals. Once the bands were defined based on the ranking of postal codes and identification of cut-points based on each including about 1/1,000th of the population, the income measure was re-calculated. That is, all of the income was summed across all of the households in each of the postal codes included in a band, and then divided by the total number of equivalent adults in that band. 3) produce a data table with the following variables (columns) for each income band (rows): • equivalized disposable income for each band, as described above • 5 t h percentile, 25 t h percentile, median, 75 t h percentile and 95 t h percentile of equivalized disposable income for each band. The incomes reported were for the households that fell exactly at the percentiles specified (e.g. 25.00th percentile). These fields provided information on the distribution of family-level incomes within each of the 1,000 income bands. • number of adults, children and households in each band, with children defined as anyone under age 14. • average equivalized federal and provincial income tax payments, average equivalized government transfer payments, and average equivalized market income. These equivalizations were calculated using the same method as described for disposable income, and were calculated only for the equivalent adult weighting method. 4) provide in a separate table a crosswalk between postal codes and each income band This research requires that the ecological income measure be linked to individual-level information on health status and health care services utilization. These individual-level data include a postal code, which is the key to linking the different data sources together. The key to connecting the data sources is a crosswalk file prepared by Statistics Canada that lists all of the postal codes associated with each of the 1,000 income bands. 85 3.4 M e t h o d s III - M e a s u r i n g t h e n e e d f o r h e a l t h c a r e s e r v i c e s Understanding the relationship between income and need for health care services (and in later chapters, between health care services use and need) depends on identifying a reliable measure for need in the population. In this case, it is preferable to develop a measure of need at the individual rather than the ecological level. One possible approach to the measurement of need is the use of total expenditures or total number of visits as derived from health care services utilization data. In other words, actual use of services is used to proxy the level of need, on the premise that the use of health care services reflects need. For present purposes, there are at least two drawbacks to this approach. First, interest here includes an examination of the relationship between need and health care services use. If we were to measure need based on patterns of health care service use, we could not then examine the relationship between need and service use. They would have been defined as identical. Second, and more generally, there are factors other than need that influence the use of health care services. It is well recognized, for example, that the availability of hospital beds or physicians influences the rate at which those services are used (Wennberg et al. 2002). Taking current patterns of use as reflective of need is at best imperfect, and there is no easy way to predict the direction of bias. Another approach to the measurement of need is the use of survey information. Self-reported health status is well known to be related to mortality (Franks et al. 2003), and has also been shown to be predictive of health care services use (Bierman et al. 1999). Survey data are the main source of (self-reported) health status used by the ECuity project (van Doorslaer and Jones 2004; van Doorslaer and Koolman 2004). If the present project were to use survey-based data, the obvious source would be the Canadian Community Health Survey, a national survey conducted by Statistics Canada that included nearly 18,000 respondents in B.C. in 2001. 6 2 The intent of the present project, however, is to conduct a population-based analysis. Limiting analysis to 18,000 individuals would not allow as detailed a division of individuals into income groups. Perhaps more importantly, a population-based analysis is a prelude to analyses that will assess the differences in estimates of equity produced by different types of data sources. Furthermore, it is not necessarily clear that a measure of health status collected through self-6 2 See Statistics Canada, Canadian Community Health Survey (CCHS) - Cycle 1.1, accessed 28 January 2006. 86 report on surveys is the best measure of need for health care services. The two concepts certainly may overlap, but they are not precisely the same thing. A person might rate his or health poorly, for example, because of limitations in dexterity or mobility. This will represent a "need" for health care only to the extent that there are health care services that can address those particular complaints (see Culyer and Wagstaff 1993 for an interesting discussion of different approaches to measuring need). An individual-level alternative for a measure of need for health care services is offered by the Johns Hopkins University Adjusted Clinical Groups case-mix system.63 As described by the developers of this system: The Johns Hopkins A C G Case -Mix System is a s tat is t ica l l y v a l i d , d iagnos is -based , risk ad jus tment methodology wh ich a l lows heal th care providers , insurers and HMOs to descr ibe or p red ic t a population's past or fu ture heal th care u t i l i za t ion and costs . ACGs are also w ide l y used by researchers to c o m p a r e various pat ient populat ions' pr ior heal th resource usage w h i l e tak ing in to account the morb id i ty or "i l lness burden" of each popu la t ion . (2003b) This system accumulates diagnoses individuals receive from encounters with physicians and hospitals over a defined time period, ln this case, as in most applications, the time period is a year. Each diagnosis is assigned to one of 32 Aggregated Diagnosis Groups defined based on several criteria including clinical similarity, and expectation of health care service utilization such as follow-up visits or the likelihood of referral to a specialist. So, for example, a diagnosis of "dermatitis" is considered a "time limited: minor" condition (2003b). The Aggregated Diagnosis Groups are an intermediate step to assigning a single Adjusted Clinical Group to each individual. There are about 100 Adjusted Clinical Groups that depict the various constellations of Aggregated Diagnosis Groups assigned to individuals along with consideration of age and sex. These groups are mutually exclusive and exhaustive. As an example, one of the Adjusted Clinical Group categories is "4-5 other A D G combinations, age>=44, 2+ major ADGs" . This category includes people who are 44 and older, and have received diagnoses assigned to 4 or 5 different Aggregated Diagnosis Groups, with two or more of those Aggregated Diagnosis Groups considered "major" conditions. Non-users of health services are assigned to A C G 5200. Individuals who are recorded in the population file but are given only diagnosis codes that do not map to one of the 32 Aggregated Diagnosis Groups, are assigned to A C G 5110. ACGs are available through the BC Linked Health Database, which is described in more detail in the next chapter. 87 The advantage of this approach to measuring need for health care services for the project at hand is that the estimate of need is explicitly based on expected use of services given an identified constellation of medical diagnoses. Two individuals with equal "need" according to this definition will be expected to use (about) the same amount of health care, measured in terms of expenditures. There is the problem, of course, that assignment to an A C G other than the two indicating no use or limited use (i.e. no diagnostic codes) depends on accessing the health care system and being assigned a diagnostic code. Individuals who are sick but, for whatever reason, do not access health care services will be classified into A C G 5200, non-users. There is nothing that can be done to rectify any misclassification based on lack of use of health care services. This case-mix system has already been validated for use in British Columbia using administrative data from the BC Linked Health Database (Reid et al. 2001). The A C G system was shown to have high explanatory power for health care services use comparing A C G assignment with same-year use of health care services in British Columbia. In other words, the A C G categories were predictive of health care services utilization. Thinking about this in terms of need for health care services, the A C G assignment for an individual will predict his or her use of health care services; average need scores within an A C G indicate average need for health care services within that group. A programmer at the Centre for Health Services and Policy Research assigned Adjusted Clinical Group measures to B.C. residents in 1992 and 2002 based on a full year of both physician and hospital separations data. Along with A C G , a single record for each individual included age, sex and region of residence, dividing British Columbia into five large, geography-based health planning areas.64 For analytic purposes, age was aggregated to six age groups: 0-14, 15-44, 45-64, 65-74, 75-84 and 85+. Because the A C G is constructed based on an accumulation of diagnosis codes found in physician payment and hospital separations data across an entire year, analyses were limited to individuals registered for at least 275 days (9 full months) during the year. Registration for less than a full year reflects less opportunity (time) to use health care services in B.C. during the course of the year. This means there is less time to accumulate the diagnosis codes that determine classification into the Aggregated Diagnosis Groups and ultimately to the Adjusted Clinical Groups. Time of enrollment exerts a potential effect on A C G assignment. For example, Information on the five health authorities, based on the second edition of the B .C. Health Atlas, can be found here: 88 imagine two individuals who make the same use of the health care system over the course of a year. One of these individuals is registered for the whole year, but the other is registered for only six months because she moved to British Columbia in the middle of the year. If all visits for each individual have the same diagnosis code (say the individuals have chronic conditions) there will be no difference in the A C G assigned to each. If, however, they have different diagnosis codes in the spring, summer, fall and winter, they are likely to end up in different ACGs because the 6-month resident will be missing diagnosis codes associated with two of those seasons. The different assignment does not reflect an underlying difference in need, but rather the fact that they had different enrollment periods in British Columbia. The restriction to 9 months / 275 days is consistent with prior analyses of B.C. data (Reid et al. 2001). The measure of need chosen for analysis must be either continuous or binary in order to use linear regression approaches to calculate the Gini coefficients and concentration indices. The need variable available for this project (ACGs) is categorical, with nearly 100 different categories. It can be converted to a continuous variable using total health care expenditures to determine the average (expected) level of expenditure for each A C G category, where total health care expenditures are the sum of expenditures for all fee-for-service physician contacts and all inpatient or surgical day care hospital use over the course of the year across all individuals in the category.65 In order to deal with extraordinary expenditures, expenditure in each A C G category was truncated at the 99 t h percentile prior to calculation of the average. This average expenditure can be thought of as a (pseudo) continuous variable that identifies appropriate scaling of differences in need between individuals. Individuals in A C G 5200, indicating no use of health care services, are expected to have need scores of 0, consistent with good health status. The lower the index, the lower the need. In this sense, one individual with a need score of 200 as determined by the ACGs has twice as much ill-health, or need for health care services, as another individual with a score of 100. While the measure of need might sound a bit circular, the A C G assignment and associated measure of average il l health should be thought of more as a measure of anticipated need than of health status, per se. What is being measured is the requirement for health care services, based on both an individual's set of diagnoses, as provided by health care practitioners, and the average system response to all individuals with a similar profile of morbidity. The A C G groups were constructed to meet the specific goal of estimating need for health care services, so this use of the groups is in line with their construction. 6 5 The next chapter will provide more detail about the utilization data. 89 This average need score based on A C G assignment thus represents the y value for each individual when thinking about constructing a Lorenz curve for the distribution of need or a concentration curve of the distribution of need by income. Equation 3.7 provided a regression approach for calculating the Gini coefficient and concentration index. That was: 2a', = a + f3Rt + u, As an example of how this works, to calculate the concentration index for the bivariate relationship between need for health care services and income, the left hand side of this regression would be calculated as: f „ \ 2CJ\ y. \ y ) where G\ is the variance of the rank, which in this case is based on income, y; is the i l l health score for an individual, and y is the population average i l l health score.66 This is regressed on the rank by income, and the coefficient on that variable is the concentration index. This concentration index, however, does not take into account any differences across income bands in the proportion of the population from each age, sex and region of residence group. It is quite possible, for example, that the same A C G category may have different average expenditures by sex and by region of the province (using five large health authorities). The supply of specialists varies by geography, as is well documented in the research literature.67 Some regions may thus have higher average expenditures within A C G categories simply because they have a more readily available supply of providers of health care services. Standardizing for these influences will help isolate the variation in need that is related to income. The y variable can be indirectly standardized through linear regression to account for these differences (van Doorslaer et al. 2004c). This standardization requires two steps. The first step is to estimate the parameters of an explanatory model for i l l health, as: J> = « + X/?/*;/+ET'*Z*+£' where y is the modeled i l l health, Ex,, is a vector of age, sex and region of residence dummies used to standardize the outcome, and Zzki is a vector of control variables, in this case limited to 6 6 Note that the variance here is not the variance of income but the variance of ranks that are based on income. For an example of how this rank variable is calculated, please refer to the Stata code provided in Appendix 1. For information specific to British Columbia see, for example, R O L L C A L L , accessed 28 January 2006. 90 the natural log of equivalized disposable income. At first it may seem that it is inappropriate to include income as a control variable, since one of the objectives of these analyses is to identify income-related differences in need for health careservices. It is included, however, in order to control for the partial effects of income that operate through other (measured or unmeasured) variables. It would not make sense, for example, to predict an individual's need for health care services using his/her income as a predictor variable, and then assess the distribution of need among all individuals to see if there is variation by income. It would be equally inaccurate, however, to omit the income variable i f it does influence need, because leaving it out means the error of the regression is subject to omitted variable bias (van Doorslaer and Koolman 2004; Wagstaff et al. 2005a). The x (standardizing) and z (control) variables are separated because they are treated differently in the second step of the standardization process. In this step, a predicted il l health score is calculated for each individual, based on the results of the regression analysis above, but replacing each individual's income status with population average income. In other words, a standardized il l health score is calculated for each individual based on his or her actual values of age, sex and region of residence68 and population average value for (natural log of) income. The predicted values of i l l health can then be scaled to create a standardized ill health score for each individual, as: y*t = y-y + yt where y is the predicted value of i l l health for each individual and pis the overall mean of the ill health variable. Subtracting the predicted value from the actual value of i l l health provides a value for the degree to which i l l health is different from what was expected based on age, sex, region of residence and mean income. Adding the average value of i l l health then scales this so that the values for the population vary around the mean rather than varying around zero. y Calculating a Gini coefficient for the distribution of il l health (using on the left hand side of ys< the equation) will tell us how it is distributed across the population after standardizing for other factors. Calculating a concentration index for income-related i l l health will tell us whether there is a relationship between income and i l l health after standardizing for other factors. This method is used to construct concentration curves and concentration indices for 1992 and 2002. The first set is an analysis of the distribution of need for health care services across the In the model, age and sex are combined in a series of dummy variables, for example males aged 0-14, females aged 65-74. 91 population. In this case, a univariate analysis (Wolfson and Rowe 2001), the population will be ranked on need and a concentration curve / index showing the distribution of need in the population will be drawn / calculated. The second set of bivariate analyses will rank the population by income, and then calculate shares of need for health care services across the population. Analyses are conducted using Stata software. Sample code is provided as Appendix 1. 3.5 Results I - income distribution Data on individual tax returns from the Canada Revenue Agency indicate that the distribution of income in British Columbia became more unequal between 1992 and 2002 (Figure 8). The Lorenz curves for the two years are close, but they do not cross.69 As expected, the Gini coefficient for disposable income in British Columbia is higher in 2002 than it was in 1992. In fact, the Gini coefficients for market income, total income and disposable income all increased over that decade (Table 7). Figure 8: Lorenz curves based on Canada Revenue Agency data, 1992 and 2002 1992 - - - .2002 Line of equality 0 0.1 0.2 0.3 0.4 0.5 0.6 0.7 0.8 0.9 1 Cumulative percent of population Where Lorenz curves cross, there is no clear "Lorenz dominance". The implication is that the change in Gini coefficient cannot be said to be "better" or "worse" on its own; other factors such as desire for income in different parts of the population would need to be considered as well. 92 Table 7 : Gini coefficients for market income, total income and disposable income, British Columbia, 1992 and 2002 1992 2002 Statistics Statistics Canada, Canada, CRA unequivalized CRA unequivalized Market income 0.4096 0.2343 0.4211 0.2609 Total income 0.3692 0.1926 0.3835 0.2219 Disosable income 0.3270 0.1662 0.3432 0.1974 The income distribution tends to be most unequal for market income, slightly less unequal for total income, and slightly less unequal again for disposable income. The difference between market and total income is government transfers, which means that government transfers are disproportionately given to lower income groups. The difference between total and disposable income is income taxes paid to the provincial and federal governments; a decrease in Gini coefficient between these measures indicates that income taxes in British Columbia were progressive in both 1992 and 2002, taking proportionately more income from higher compared to lower income groups. The data from the Canada Revenue Agency and Statistics Canada are quite different in the magnitude of the Gini coefficient each produces, but are consistent in the size (real, not proportionate) and direction of change over time (Table 7). The Statistics Canada data, built from family-level tax information, but aggregated to 1,000 income bands, show much lower Gini coefficients than the individual-level Canada Revenue Agency data. The rest of the analyses in this project use the Statistics Canada data rather than the Canada Revenue Agency data, because it is the former that can be "linked" to measures of need and health care services utilization. Given this, it is important to consider the reasons for the differences. First, the two data sets are organized differently, with the Canada Revenue Agency data organized by individuals who file returns, and the Statistics Canada data organized by families, based on relationships defined on the tax form. The 1992 and 2002 Canada Revenue Agency data represent 1.72 million and 1.94 million tax filers in British Columbia, respectively. The Statistics Canada tax filer data are associated with 3,341,460 individuals in 1992 and 3,862,630 in 2002. The way that individuals form families will have an unknown effect on the level of Gini coefficient. If, for example, families tend to include one lower-wage earner and one higher-wage earner, the Gini for families would be lower than that for individuals. If, on the other hand, individuals with similar income tend to form families, the Gini coefficient would increase as we move from individuals to families. 93 Another difference between the Canada Revenue Agency and Statistics Canada data is that the latter have been equivalized to reflect (to the extent possible) differences in living standards associated with economies of scale of family composition. As above, it is not possible to predict what effect this adjustment will have on the Gini coefficient, because that depends on differences in family size across the different income groups. It was not possible to quantify the effects of either of these differences with the data available for this project. Finally, the Statistics Canada data represent weighted average incomes for families in 1,000 income bands that are in turn aggregated from postal codes. Since there will be a range of incomes in each postal code, this means that even though the Statistics Canada data have 1,000 income groups while the Canada Revenue Agency has only 17, there is more heterogeneity of income in the former. This heterogeneity will tend to lower the Gini coefficients for the ecological data, because the true range of incomes in the population is masked by the fact that higher income and lower income families often live in the same postal codes. This heterogeneity can be seen in Figure 9, which shows the range of family-level incomes, calculated as the 95 t h percentile of each income band less the 5 t h percentile of the same band, against income band, where the bands are ranked from 1 (lowest average disposable income) to 1,000 (highest average disposable income). Most income bands have a range of income of $20,000 to $80,000 or so, with a few income bands at the upper end of the distribution where the range is much greater. Even a $50,000 range is significant, however, compared to the $5,000 and $10,000 wide categories used for most income categories in the Canada Revenue Agency data. In that data set, only incomes above $150,000 are grouped into larger bands, and then because the number of individuals in those bands declines enough to warrant making them broader. 94 F i g u r e 9 In the case of the Statistics Canada data, the range increases with income, because the lower end changes far less than the higher end. This can be seen in Figure 10, which shows an X Y scatter plot of the income of the household at the 5 t h percentile of the distribution within each income band, against the income of the household at the 95 t h percentile of income in the same band. The household at the 5 t h income percentile in each band ranges from close to $0 of income to nearly $10,000, with the majority in the range of $4,000 to $8,000. The range for the 95 t h percentile is much wider, with a significant number of points above $50,000. This suggests that while some geographic areas are composed entirely of families at the lower end of the income distribution, lower income families are resident in every postal code-based income band in British Columbia. The lower incomes may be driven at least in part by the presence of lower income (but higher wealth) elderly in some areas, or by the increasing prevalence (especially in urban areas) of supplemental suites in houses. Again, the data available do not lend themselves to separating these effects. 95 Figure 10: XY scatter plot of 5 percentile against 95 percentile of income 0 2000 4000 6000 8000 10000 perc5 Note: This graph does not include one extreme value, the range for the 1 0 0 0 * income band The fact that the income inequality tends to be lower if based on the Statistics Canada data than on the Canada Revenue Agency data suggests that any findings here are likely to be conservative. The misclassification of incomes inherent in ecological measures limits the potential to find a relationship between need for health care services (or health care services use and health care financing in subsequent chapters) and income. 3.6 Results II - need for health care services The data from the BC Linked Health Database, from which the measure of need for health care services is derived, included 3,515,458 individuals in 1992 and 4,206,429 in 2002 (Table 8). In 1992, 473,043 of these observations, representing 13.5% of the total population, were dropped because their postal code could not be converted to one of the 1,000 income bands (137,652), because the postal code could not be mapped to a region in British Columbia (15,711), because the registration period was less than nine months (308,885), or because there was missing or invalid information for sex (10,688) or age (107). The percentage loss was lower in 2002, at 8.3%), both because of fewer records with postal code problems (105,497) and fewer individuals with less than nine full months of registration (237,728). The loss of observations due to 96 problems with postal codes was expected given what we know about the postal code recorded in the registry files in the B C L H D . 7 0 Table 8 : Population characteristics for records that were dropped from analysis 1992 2002 N % N % Total population 3,515,458 4,206,429 Dropped 473,043 13.5% 349,470 8.3% Study population 3,042,415 86.5% 3,856,959 91.7% For dropped records, where age and sex are known Males 259,658 54.9% 185,090 53.0% Females 202,590 42.8% 158,135 45.2% Age group 0-14 102,159 21.6% 70,923 20.3% 15-44 257,579 54.5% 197,383 56.5% 45-64 70,535 14.9% 46,265 13.2% 65-74 15,276 3.2% 10,483 3.0% 75-84 10,589 2.2% 9,922 2.8% 85+ 6,110 1.3% 8,249 2.4% Table 8 provides the age and sex distribution for observations that were dropped, excluding the minority of cases for which this information was missing or invalid. Compared to the population statistics for the study population (Table 9), the dropped records were more likely to be young and male. In some cases, it appears that a timing difference in the collection of postal code information was the cause of the mis-match. The B C L H D postal code is based on a cross-sectional snapshot of residence as of (approximately) July 1st of each year. The Statistics Canada postal code would reflect what was provided by individuals when they file their taxes in April of the following year. This mis-match presented a particular problem in 1992, when a large number of postal codes, concentrated in the quickly growing area of Surrey, were retired and replaced with new postal codes. The implication is that the B C L H D postal code likely represents the retired postal code and the Statistics Canada postal code the new postal code. There is no crosswalk between old and new, thus an income band could not be identified for individuals so affected. 97 Table 9: Characteristics of the study population, 1992 and 2002 1 9 9 2 2 0 0 2 B C L H D BCLHD population % of total BC Stats ( Vo of total population' Vo of total BC Stats c Vo of total Total population 3,042,415 3,470,307 3,856,959 4,141,272 Males 1,467,647 4 8 . 2 % 1,730,113 4 9 . 9 % 1,892,380 4 9 . 1 % 2,053,091 49 .6% Females 1,574,768 51 .8% 1,740,194 5 0 . 1 % 1,964,579 50.9% 2,088,181 50.4% Age group 0-14 608,593 2 0 . 0 % 696,872 2 0 . 1 % 656,032 17.0% 718,126 17.3% 15-44 1,368,785 4 5 . 0 % 1,634,490 4 7 . 1 % 1,620,110 4 2 . 0 % 1,827,269 4 4 . 1 % 45-64 642,313 2 1 . 1 % 698,566 2 0 . 1 % 1,029,512 26 .7% 1,044,057 25 .2% 65-74 247,103 8 . 1 % 259,669 7.5% 286,938 7.4% 290,384 7.0% 75-84 136,523 4 . 5 % 141,330 4 . 1 % 194,468 5.0% 194,643 4 .7% 85+ 39,098 1.3% 39,380 1.1% 69,899 1.8% 66,793 1.6% Income decile 1 291,761 9.6% 376,599 9.8% 2 299,006 9.8% 384,169 10.0% 3 302,059 9.9% 385,162 10.0% 4 301,397 9.9% 385,458 10.0% 5 301,645 9.9% 386,682 10.0% 6 308,661 1 0 . 1 % 389,315 1 0 . 1 % 7 310,134 10.2% 388,350 1 0 . 1 % 8 306,135 1 0 . 1 % 387,659 1 0 . 1 % 9 311,702 10.2% 387,835 1 0 . 1 % 10 309,915 10.2% 385,730 10.0% Region Interior 533,606 17.5% 594,915 1 7 . 1 % 644,837 16.7% 688,246 16.6% Fraser 936,669 30 .8% 1,097,458 31.6% 1,304,046 3 3 . 8 % 1,421,216 3 4 . 3 % Vancouver Coastal 773,633 25.4% 870,252 2 5 . 1 % 980,391 25.4% 1,035,101 25 .0% Vancouver Island 539,525 17.7% 619,788 17.9% 645,171 16.7% 695,942 16.8% Northern 258,982 8 .5% 287,894 8 . 3 % 282,514 7 .3% 300,767 7 .3% There were slightly more females than males in the final study population, consistent with the distribution reported in official population statistics from BC Stats (Table 9)(BC Stats 2003). Population estimates from BC Stats are provided in this table to see whether the study population is consistent in its composition to official estimates. The distribution by age group was also similar; in 1992, 45% of the population was in the 15-44 age range, one-fifth aged 14 and under, and about 14% aged 65 and over. The biggest demographic change between 1992 and 2002 was the movement of the "bulge" of the population into the 45-64 age group, which represented a little more than one-fifth of the total population in 1992 and more than one-quarter in 2002. 7 1 BC Stats produces yearly populaton estimates at provincial and sub-provinical levels, including regions used for health planning and administration. These estimates start with population figures produced by Statistics Canada based on the population census, with adjustments for known census under-counting and interpolation between census years. 98 About 10% of the population is assigned to each of the income deciles. The population is not precisely divided into deciles because of the use of the ecological measure of income. Each income band must be assigned to a decile in its entirety, which means that it is impossible to equalize the population of each decile. Over half the population of British Columbia resides in two of five health authorities, with that proportion increasing slightly over time. Distributions by region are roughly similar for the BCLHD-derived and BC Stats data. About 13%o of the population in 1992 and 17% in 2002 is classified into A C G 5200, indicating no use of physician and hospital services (Table 10) While this A C G no doubt includes some individuals who do have health problems, for purposes here this is the "good health" group, as indicated by the average need score of 0. The rest of Table 10 provides a summary of the average need scores for the most needy and least needy A C G groups in 1992 and 2002. Full tables are provided in Appendix 2. The "sum of need scores" column in Table 10 is calculated by adding the need scores for the individuals that fall into each A C G . The average is then calculated by dividing that sum by the 79 total N in assigned to each group. The five A C G groups with the lowest need scores (other than the non-users) represented about 9% of the total study population and less than 1% of total need in both 1992 and 2002. A l l five groups have need scores less than 100. The five lowest need ACGs change over time. The number of people in the A C G for "ophthalmological / dental" diagnoses decreased by half between 1992 and 2002, and the average need increased from 72 to 241. This implies that the average person in this A C G had more than three times the need for health care services in 2002 compared to 1992. In both years the list of five lowest need ACGs includes "preventive / administration" and "acute minor, age 5+", groups that would be expected to have low needs. The ill health score is an index meant to reflect the average need for health care services for each A C G group. The sum of ill health scores across the population represents the total ill health of the population, or the total need for health care services. Table 10: Need characteristics of selected Adjusted Clinical Groups (ACGs) in British Columbia, 1992 and 2002 1992 2002 Sum o f n e e d A v g n e e d Sum o f n e e d A v g n e e d ACG Description T o t a l N s c o r e s ( t runcated) s c o r e s ( trunc.) T o t a l N s c o r e s ( t runca ted) s c o r e s ( trunc.) Total - All ACGs 3 , 0 4 2 , 4 1 5 2 , 4 8 9 , 6 5 2 , 8 8 4 8 1 8 3 , 8 5 6 , 9 5 9 3 , 8 5 1 , 1 6 6 , 8 2 5 998 5200Non-Users (2 input files) 4 0 4 , 4 0 3 0 0 6 3 8 , 9 5 9 0 0 Percentage of total 13.3% 0.0% 76.6% 0.0% "Least needy" ACGs based on 1992 sorting 1600Prevent ive / Administrative 8 , 0 3 0 3 8 0 , 6 8 3 4 7 1 7 , 1 2 2 9 7 8 , 4 7 1 5 7 5 1 1 0 N o or Only Unclassif ied Diagnoses (2 input files) 2 1 , 1 3 4 1 , 5 3 9 , 2 1 2 6 6 2 8 , 0 6 2 2 , 7 5 4 , 4 3 0 94 300Acu te Minor, Age 6+ 2 4 3 , 2 7 8 1 7 , 1 3 7 , 8 8 8 7 0 2 9 0 , 7 1 1 2 3 , 8 3 8 , 6 6 1 82 1000Chron i c Specialty 2 , 9 2 5 2 0 6 , 6 7 0 71 2 ,551 2 6 0 , 5 3 2 102 110OOphthalmological /Dental 1 5 , 5 4 6 1 ,104 ,977 71 6 ,761 1 , 6 4 6 , 7 5 6 244 Percentage of total 9.6% 0.8% 9.0% 0.8% "Least needy" ACGs based on 2002 sorting 1600Preventive / Administrative 8 , 0 3 0 3 8 0 , 6 8 3 4 7 1 7 , 1 2 2 9 7 8 , 4 7 1 57 7 0 0 A s t h m a 3 , 3 7 3 3 7 1 , 0 8 0 1 1 0 7,211 5 2 5 , 3 5 4 73 300Acu te Minor, Age 6+ 2 4 3 , 2 7 8 1 7 , 1 3 7 , 8 8 8 7 0 2 9 0 , 7 1 1 2 3 , 8 3 8 , 6 6 1 82 200Acu te Minor, Age 2-5 2 5 , 7 4 9 2 , 0 2 2 , 8 6 7 7 9 2 5 , 9 8 4 2 , 3 0 1 , 9 7 9 89 600l_ikely To Recur, with Allergies 7 , 9 7 3 7 0 8 , 3 1 4 8 9 10 ,361 9 5 2 , 0 0 7 92 Percentage of total 9.5% 0.8% 9.1% 0 . 7 % "Most needy" ACGs (list consistent for both years) 503010+ Other A D G Combinat ions, Age 1-17, 2+ Major A D G s 8 3 5 2 0 , 4 0 6 , 4 1 8 2 4 , 4 3 9 8 4 7 4 3 , 3 5 9 , 3 3 3 5 1 , 1 9 2 507010+ Other A D G Combinat ions, Age 18+, 4+ Major A D G s 1 3 , 4 2 4 2 7 6 , 8 9 0 , 6 3 4 2 0 , 6 2 7 1 7 , 3 3 5 4 9 4 , 6 1 0 , 2 4 4 28,532 4 9 4 0 6 - 9 Other A D G Combinat ions, Age >34, 4+ Major A D G s 5 , 3 7 8 6 7 , 8 3 6 , 1 4 4 1 2 , 6 1 4 6 , 3 9 8 1 0 2 , 9 5 1 , 6 1 1 16 ,091 5340lnfants : 6 + A D G s , 1+ Major A D G 1,606 1 5 , 7 1 9 , 6 4 1 9 , 7 8 8 1 ,126 1 9 , 1 5 2 , 2 4 0 17,009 506010+ Other A D G Combinations, Age 18+, 3 Major A D G s 1 5 , 7 4 8 150,096,310 9,531 1 8 , 8 9 5 2 2 7 , 4 7 8 , 4 9 2 1 2 , 0 3 9 Percentage of total 1.2% 21.3% Most common ACGs based on population (list of top six consistent for both years) 1 2 % 2 3 . 0 % 1800Acu te Minor and Acute Major 2 6 2 , 9 7 8 7 4 , 8 7 9 , 6 3 9 2 8 5 3 1 2 , 2 4 2 1 0 5 , 1 1 6 , 9 2 5 337 300Acu te Minor, Age 6+ 2 4 3 , 2 7 8 1 7 , 1 3 7 , 8 8 8 7 0 2 9 0 , 7 1 1 2 3 , 8 3 8 , 6 6 1 82 4 1 0 0 2 - 3 Other A D G Combinat ions, Age >34 2 0 6 , 8 9 3 8 2 , 4 2 2 , 0 3 4 3 9 8 3 0 7 , 0 3 1 1 5 0 , 9 4 0 , 5 8 3 4 9 2 400Acu te : Major 1 6 1 , 5 8 7 1 9 , 9 9 5 , 6 3 4 1 2 4 2 1 5 , 4 8 8 3 0 , 3 9 4 , 5 5 0 141 Acute Minor and Major/Likely to Recur, age > 12,w/out 3200Al le rg ies 1 4 4 , 8 5 9 8 6 , 4 8 7 , 3 7 5 5 9 7 1 6 2 , 1 8 4 1 1 0 , 2 1 2 , 2 2 4 6 8 0 49106-9 Other A D G Combinat ions, age >34, 0-1 Major A D G s 1 3 7 , 5 9 9 2 0 8 , 3 9 8 , 2 5 4 1 ,515 189 ,551 3 4 6 , 7 0 3 , 3 3 3 1 ,829 Percentage of total ACGs accounting for most need (list consistent for both years) 38.0% 19.7% 38.3% 19.9% 507010+ Other A D G Combinations, Age 18+, 4+ Major A D G s 1 3 , 4 2 4 2 7 6 , 8 9 0 , 6 3 4 2 0 , 6 2 7 1 7 , 3 3 5 4 9 4 , 6 1 0 , 2 4 4 28,532 4 9 1 0 6 - 9 Other A D G Combinat ions, Age >34, 0-1 Major A D G s 1 3 7 , 5 9 9 2 0 8 , 3 9 8 , 2 5 4 1 ,515 189 ,551 3 4 6 , 7 0 3 , 3 3 3 1 ,829 4 9 2 0 6 - 9 Other A D G Combinat ions, Age >34, 2 Major A D G s 6 0 , 6 1 9 2 0 7 , 0 8 3 , 1 4 7 3 , 4 1 6 8 4 , 0 4 9 3 6 4 , 2 6 1 , 5 5 3 4 , 3 3 4 4 9 3 0 6 - 9 Other A D G Combinat ions, Age >34, 3 Major A D G s 2 2 , 3 9 6 1 5 3 , 7 9 4 , 2 8 3 6 , 8 6 7 2 9 , 3 4 8 2 6 0 , 3 4 4 , 3 0 2 8 ,871 506010+ Other A D G Combinations, Age 18+, 3 Major A D G s 1 5 , 7 4 8 1 5 0 , 0 9 6 , 3 1 0 9 ,531 1 8 , 8 9 5 227,478,492 1 2 , 0 3 9 100 The five A C G groups with the highest average need scores include 1.2% of the population in both 1992 and 2002 and account for 21%> and 23% of total need respectively. This list of five is consistent in 1992 and 2002, and there is a significant increase over time in the average need in each of the categories. The group with the highest average need doubles its average score in 2002 compared to 1992, from a score of about 24,000 to over 50,000. This A C G group is, by definition, limited to children aged 1-17, and includes less than 1,000 children in each year, all who had diagnoses indicating assignment to ten or more Aggregated Diagnosis Groups, two or more of which were considered major. Multiple morbidity is a universal feature of all five of these need groups. 73 Just under 40% of the population in both years is found in the six most common ACGs. This jumps to more than half the population if the list is expanded to seven to include the A C G for non-users. These groups account for only about 20% of total need. Ranking instead by the accumulation of need, the five top A C G groups account for about 40%> of the total need but only 8% of the population in 1992 and just under 9% of the population and 44% of total need in 2002. There is only one A C G that makes it to both the top six ranked by population and the top five ranked by total need. This A C G includes adults aged 35 and over who were assigned to between 8 and 9 Aggregated Diagnosis Groups, either none or only one of which was considered major. The average need scores in this A C G are between 1,000 and 2,000 in both years, far lower than the scores for the most needy groups, but there are 137,599 individuals in this A C G in 1992 and 189,551 in 2002. The need for health care services was clearly very unevenly distributed in British Columbia in both 1992 and 2002. As might be expected, the ACGs are also not evenly distributed across age, sex and income groups, nor are they evenly distributed across the five regions of the province. Table 11 includes the two most common ACGs and the two ACGs with the highest average need scores.74 Six are included in this list because the top six are common to both years while the top five are not. For display purposes, the 1,000 income groups have been aggregated to 10 groups (deciles). The division of deciles was based on each representing as close to 10% of the population as possible. All calculations of Gini coefficients and concentration indices used the 1,000 income bands and not deciles. 101 Table 1 1 : Study population characteristics for individuals in high frequency and high need ACGs 1992 2002 High frequency ACGs High need ACGs High frequency ACGs High need ACGs 10+ Other 10+ Other 10+ Other 10+ Other ADG ADG ADG ADG Combination Combination Acute Combination Combination Acute Minor s, Age 18+, s, Ages 1-17 Minor and s, Age 18+, s, Ages 1-17 and Acute 4+ Major 2+ Major Acute Acute Minor, 4+ Major 2+ Major Major Acute Minor, ADGs ADGs Major Age 6+ ADGs ADGs (1800) c /o Age 6+(300) % (5070) /o (5030) % (1800) < /o (300) % (5070) ( /o (5030) % Total population 262,978 8.6% 243,278 8.0% 13,424 0.4% 835 0.03% 312,242 8.1% 290,711 7.5% 17,335 0.4% 847 0.02% Males 128,835 8.8% 142,629 9.7% 7,145 0.5% 446 0.03% 151,351 8.0% 166,160 8.8% 9,131 0.5% 464 0.02% Females 134,143 8.5% 100,649 6.4% 6,279 0.4% 389 0.02% 160,891 8.2% 124,551 6.3% 8,204 0.4% 383 0.02% Age group 0-14 74,72912.3% 69,560 11.4% 0 0.0% 673 0.11% 76,771 11.7% 80,157 12.2% 0 0.0% 686 0.10% 15-44 132,516 9.7% 126,662 9.3% 1,627 0.1% 162 0.01% 151,818 9.4% 148,499 9.2% 1,790 0.1% 161 0.01 % 45-64 40,552 6.3% 37,029 5.8% 3,362 0.5% 0 0.00% 67,242 6.5% 53,135 5.2% 4,581 0.4% 0 0.00% 65-74 9,722 3.9% 7,018 2.8% 3,798 1.5% 0 0.00% 9,949 3.5% 5,814 2.0% 3,931 1.4% 0 0.00% 75-84 4,021 2.9% 2,384 1.7% 3,497 2.6% 0 0.00% 4,474 2.3% 2,225 1.1% 4,988 2.6% 0 0.00% 85+ 1,438 3.7% 625 1.6% 1,140 2.9% 0 0.00% 1,988 2.8% 881 1.3% 2,045 2.9% 0 0.00% Income decile 1 24,200 8.3% 21,902 7.5% 1,906 0.7% 130 0.04% 28,768 7.6% 27,981 7.4% 2,284 0.6% 134 0.04% 2 25,582 8.6% 24,143 8.1% 1,471 0.5% 104 0.03% 30,784 8.0% 29,614 7.7% 1,698 0.4% 92 0.02% 3 25,888 8.6% 24,476 8.1% 1,339 0.4% 93 0.03% 29,770 7.7% 29,173 7.6% 1,822 0.5% 90 0.02% 4 25,985 8.6% 24,830 8.2% 1,353 0.4% 67 0.02% 29,576 7.7% 28,829 7.5% 1,737 0.5% 91 0.02% 5 24,883 8.2% 24,486 8.1% 1,380 0.5% 80 0.03% 30,245 7.8% 29,397 7.6% 1,749 0.5% 80 0.02% 6 28,188 9.1% 25,856 8.4% 1,129 0.4% 78 0.03% 31,816 8.2% 29,829 7.7% 1,667 0.4% 64 0.02% 7 27,343 8.8% 25,054 8.1% 1,203 0.4% 75 0.02% 32,383 8.3% 29,576 7.6% 1,577 0.4% 87 0.02% 8 26,854 8.8% 24,925 8.1% 1,208 0.4% 59 0.02% 32,562 8.4% 29,244 7.5% 1,591 0.4% 73 0.02% 9 27,387 8.8% 24,697 7.9% 1,173 0.4% 87 0.03% 33,047 8.5% 29,534 7.6% 1,546 0.4% 74 0.02% 10 26,668 8.6% 22,909 7.4% 1,262 0.4% 62 0.02% 33,291 8.6% 27,534 7.1% 1,664 0.4% 62 0.02% Region Interior 52,092 9.8% 48,884 9.2% 2,135 0.4% 116 0.02% 51,591 8.0% 49,393 7.7% 2,784 0.4% 132 0.02% Fraser 78,926 8.4% 72,430 7.7% 4,635 0.5% 336 0.04% 106,509 8.2% 100,184 7.7% 5,567 0.4% 317 0.02% Vancouver Coastal 64,979 8.4% 58,549 7.6% 3,719 0.5% 177 0.02% 85,709 8.7% 73,378 7.5% 4,654 0.5% 191 0.02% Vancouver Island 44,091 8.2% 39,730 7.4% 2,208 0.4% 132 0.02% 48,846 7.6% 44,598 6.9% 3,291 0.5% 137 0.02% Northern 22,890 8.8% 23,685 9.1% 727 0.3% 74 0.03% 19,587 6.9% 23,158 8.2% 1,039 0.4% 70 0.02% 102 In both 1992 and 2002, A C G 1800, for acute minor and acute major conditions, is the single largest A C G category (other than 5200, which identifies non-users), with 262,978 and 312,242 individuals, respectively. The percentages reported in this table are the proportions of relevant populations represented. For example, A C G 1800 includes 8.6% of the total population in 1992 and 8.1% of the total in 2002. Males and females are equally likely to be in this category, but younger age groups are over-represented (for example, 12.3% of children aged 0-14 in 1992 are in this group) and older age groups are under-represented (e.g. 2.9% of 75-84 year olds in 1992 were assigned to this ACG). These patterns are pretty stable across time. The presence of this A C G across income deciles varies from 7.6% of the lowest income decile to 8.6% of the highest income decile in 2002; this increase is not monotonic across the income deciles. Nearly 10%> of all residents of the Interior Health Authority were assigned to this A C G in 1992, but this decreased to 8% in 2002. The Northern Health Authority saw a similar decrease, from 8.8% of its population assigned to this A C G in 1992 to 6.9% in 2002. The pattern for acute minor conditions (ACG 300) is much the same, with the exception that males (9.7% in 1992) are far more likely to be in this category than females (6.4%). This difference by sex was slightly less but still present in 2002. The high need ACGs each account for less than one percent of the population; a small fraction of a percent in the case of the multiple morbidity A C G with the highest average il l health (ACG 5030). There is no difference by sex in either of these groups, but older age groups are over-represented in A C G 5070 (ACG 5030 is, by definition, limited to the two youngest age groups). There is a slight declining likelihood of assignment to the high il l health A C G groups by income decile, but no pattern by region of residence. There are no significant changes over time. The average need score assigned to each A C G is based on total physician and hospital services use of all the individuals assigned to that A C G . The average thus masks a distribution of need within each A C G . Figure 11 shows the range within each A C G based on 2002 data, calculated by subtracting the value of the 5 t h percentile from the value of the 95 t h percentile of the distribution. Each bar along the x-axis represents one A C G , with the A C G groups sorted from lowest to highest average need.75 The range within each A C G group increases as the average need increases, meaning that this measure is heteroskedastic. There is good reason to use the Newey-West correction for standard errors, as described in the methods section. The furthermost bar on the right is truncated. The actual value for the range of this bar is 224,103. It would have been impossible to differentiate the majority of the bars (which are under 10,000) without this truncation. 103 Figure 11: Range (95 percentile - 5 percentile) of health care services use within each A C G 90.000 80,000 a • 70,000 c a> = 60,000 | 50,000 a 40,000 " a - r ^ o r o c o c n r g i n c D ^ ' n - r ^ o r o u D C D r v i u n c D ^ T r ^ c D c n c D o r N L D Q O ACG The distribution of need for health care services in the population can be summarized in Lorenz curves in which individuals are ranked by highest to lowest need (Figure 12, using data form 1992). The unadjusted Lorenz curve is furthest to the left, well above the line of equality. The curve is above the line of equality because the measure of need is a "bad" rather than a "good". This means that the Gini coefficients will be negative. Standardizing for age, sex and region of residence, and controlling for the effects of income, moves the adjusted Lorenz curve in a bit, somewhat closer to the line of equality. The Gini coefficient for the adjusted distribution of need will be closer to zero than the Gini coefficient for the unadjusted distribution. 104 Figure 12: Unadjusted and adjusted Lorenz curves for need for health care services, 1992 Cumulative piopiotion of population Both the unadjusted and adjusted curves for 1992 suggest that about half of all the need for health care services in the population is concentrated in 10% of the population. About the same is true for 2002 (Figure 13). 105 Figure 13: Adjusted Lorenz curves for need for health care services, 1992 and 2002 0 0.1 0.2 0.3 0.4 0.5 0.6 0.7 0.8 0.9 1 C u m u l a t i v e p r o p o r t i o n of p o p u l a t i o n The unadjusted and adjusted concentration curves for need in 2002 with individuals ranked by income show quite a different picture (Figure 14). As with the Lorenz curves for need, the curves lie above the line of equality, but they are much closer to that line; income-related inequality in need for health care services is much less than inequality in need overall. In this case, standardization moves the concentration curve very slightly away from the line of equality, suggesting that the distribution of age, sex and region of residence across the income groups is not equal. In this case, controlling for these differences increases the inequality in need. This will happen to the extent that, for example, age is associated both with lower income and higher need. The concentration curves 2002 (not shown) look much the same. 106 Figure 14: Concentration curves (unadjusted and adjusted) for need according to ranking by income, 1992 Unadjusted — -Adjusted Line of equality Cumulative proportion of population tanked by income The Gini coefficients and concentration indices for need and income-related need quantify what is shown by the Lorenz and concentration curves (Table 12). The adjusted Gini coefficient for need for health care services is -0.59 in 1992 (-0.60 in 2002), compared to a standardized concentration index for need by income rank of -0.04 (-0.03). Both are significantly different from equality, as indicated by the small standard errors. The difference between the Gini coefficients and concentration indices suggests that income is not the primary explanation of the variation in need in British Columbia. Table 12 also includes adjusted and unadjusted concentration indices calculated based on deciles rather than the 1,000 income bands. This was done to check the influence of small-area groupings on the results. If the results were quite different, one might question whether any differences found in need by income were actually due to other factors, such as genetics. As it turns out, the differences are small; for example a change in standardized concentration index in 2002 from -0.0313 to -0.0315. Subsequent analyses use 1,000 income bands. Finally, the table includes information on the sensitivity analyses run to test the effects of different household weighting approaches. Again, the alternative measures had a minimal effect on the concentration indices, in all cases changing them by 0.001 or less. Household weighting increased the unstandardized measure of inequality, meaning that weighting approach (i.e. no weighting) made need according to income rank appear (very slightly) more unequal. Per capita 107 weighting had the opposite effect, decreasing the concentration index. In both cases and in both years of analysis, the choice of weighting had essentially no impact on the standardized measure of inequality. The differences are small enough that subsequent analyses will be performed using only the equivalized weighting, which weights each household according to the number of equivalized individuals in it. Table 12: Unstandardized and standardized Gini coefficients for need for health care services and concentration indices for income-related need for health care services 1992 2002 Gini Coefficient / Concentration index Standard error Gini Coefficient / Concentration index Standard error Need for health care services Unstandardized -.70163465 .00152476 Standardized -.58889233 .00148948 Need for health care services by income rank Unstandardized -.03257056 .00079772 Standardized -.04381254 .00076852 -.72921308 -.59866423 -.01671108 .03126455 .00155953 .00153535 .00078550 .00076084 Need by income rank, calculated based on deciles rather than 1,000 income bands Unstandardized Standardized -.03313653 -.04409105 .00080331 .00077333 -.01588204 -.03148524 .00078601 .00076098 Sensitivity analyses Household weighting Unstandardized -.03221879 Standardized -.04390720 Individual / per capita weighting Unstandardized -.03286419 Standardized -.04374240 .00079833 .00076888 .00079793 .00076872 .01648765 .03243497 .01681539 .03239281 .00078441 .00076010 .00078416 .00075979 3.7 Discussion and conclusion Income data from the Canada Revenue Agency suggest greater income inequality than do data from Statistics Canada. Both data sets originate with income tax forms, but the former have been aggregated, at the level of tax filer, to groups based on total income. The latter are analyzed at the census family level within postal codes and then aggregated based on grouping postal codes with similar average equivalized family income levels. The difference in approach results in Gini coefficients based on Canada Revenue Agency data that are nearly twice the Gini coefficients derived from the Statistics Canada data. These differences are due to some combination of the effects of methods of aggregation and the impact of equivalization, but it is not possible with the data available to identify the independent effect of each. In all likelihood, 108 however, it is the heterogeneity in incomes introduced by using an ecological income measure that is producing the majority of this effect. Despite the differences in the magnitude of inequality, the two data sources show the same trends over time in market income, total income and disposable income. In general, income inequality in British Columbia increased over this decade, irrespective of income concept. This trend is new; throughout the 1980s, the impact of increases in inequality in market income was offset by transfer payments and the progressivity of income taxes (Frenette et al. 2004). Government policy in the 1990s has either been ineffective in offsetting these effects, or has been designed to allow the changes in market income inequality to flow through to total and disposable income (more on this in Chapter 5). Inequality in need for health care services as measured here is -0.60 in 2002 after standardizing for age, sex and region of residence and controlling for income. This is a relatively high (in absolute value) concentration index, compared to other research using similar analytic methods (more on this below). The concentration index drops significantly, to -0.04 in 1992 and -0.03 in 2002, when the ranking of individuals is based on income rather than need. Most of the inequality in need is associated with factors other than income. This conclusion must be tempered, however, with a reminder that the income measure used here is ecological rather than individual. Using the differences in Gini coefficient for groups based on individual income and groups based on neighbourhood income as a rough guide, one might expect the concentration index to double with a family- or household-level measure of income. In any case, the association between income and need for health care services is still significant, even after standardizing for age and sex, and favours those with more income. A concentration index of-.03 or -.04 should not be dismissed as insignificant from a policy standpoint. Consider, for example, life expectancy for Canadian males in 1996 according to neighbourhood income quintile as presented by Wilkins, Berthelot and Ng (2002) (Table 13). In 1996 a male living in a neighbourhood in the highest income quintile had a life expectancy that was five years longer than a male living in a neighbourhood in the lowest income quintile. It would be difficult to argue that the difference was unsubstantial, but as Table 3.12 shows, the proportion of total life years enjoyed by each income quintile did not differ by a great deal, simply because all income groups lived an average of at least 73.1 years. The difference between income groups is a small proportion of the total expected length of life. These differences translate to a concentration index of 0.01. 109 Table 13: Life expectancy for Canadian males, 1996, by neighbourhood income quintile Life Proportion Cumulative expectancy of total life proportion years of life years Quintile 1 (poorest) 73.1 19.2% 19.2% Quintile 2 75.9 19.9% 39.1% Quintile 3 76.7 20.1% 59.2% Quintile 4 77.2 20.3% 79.5% Quintile 5 (richest) 78.1 20.5% 100% The only other research using Canadian data that looked at the concentration index of income-related health status reported indexes of -0.099 and -0.12, depending on the measure of health status used (Humphries and van Doorslaer 2000). There are several important differences between that research and what is reported here. First, Humphries and van Doorslaer used data on self-reported health status (measured with the Health Utility Index) and self-reported gross income from the 1994/95 National Population Health Survey, whereas the present work uses an administrative data-based measure of need for heatlh care services and an ecological measure of income. Second, their data were for Canada as a whole, while in the present case the data are restricted to British Columbia. Finally, the survey-based measure of income was for total income rather than disposable income as used here. This may make a difference because gross (or total) income is more unequally distributed than is disposable income. Using the same logic as above, it seems most likely that the difference is due to the differences in income measures used. It is quite possible, however, that the health status measure exerts a strong influence as well. Results from previous international work vary greatly when different approaches are used to measure health status, in some cases producing up to a ten-fold variation in result for the same country (van Doorslaer et al. 1997; van Doorslaer and Koolman 2004). These varying results make it clear that comparisons across studies where different analytic methods were used can be a risky and misleading business. It would be difficult to make any comment about Canada's place in an international perspective based on the present analyses. One possible explanation for differences in results is that in all previous research the outcome variable is bounded, and there is a limit to how much worse " i l l health" is compared to "good health". For the Health Utility Index, for example, the bounds are 0 to 1, though the effective 110 bounds are narrower than this, since virtually everyone will have an index value that is greater than 0. 7 6 In contrast, the measurement of need for health care services based on ACGs, where average expected health care expenditure, adjusted for age and sex, is the outcome variable, effectively removes these boundaries on the upper end. In the present case, the least needy A C G category (other than the non-users) has a need score of about 50, while the most needy A C G category has an average need score of over 20,000 in 1992 and over 50,000 in 2002. In this approach, the most needy people are seen to bear thousands of times more of the need burden than the healthiest, while in the survey-based approached, the effective maximum is more like 10 times.77 Which approach is the more appropriate depends on the question one is trying to address. Where interest is in estimating the distribution of need for health care services, and in looking at the use of health care services after adjusting for morbidity, a case can be made that the A C G approach is more sensible. This variable is designed to capture "need" as it relates to the consumption of health care services. At the same time, the variable is not completely hostage to how people use health care services; while A C G assignment is dependent on diagnoses provided by physician or hospital encounters, the same A C G is applied to cases where there are 20 visits and where there is one visit, i f the diagnosis in each case is the same. One limitation to the use of ACGs for measuring need is that they do require use of health care services to get into a category beyond the non-users (i.e. healthy) group. Another limitation is that the range of expenditures within each A C G group suggests that this is an incomplete measure of need. They capture the need for health care services, but underlying that need in each A C G category is a latent severity, with no obvious means of assessing its extent or distribution within the A C G or of adjusting for it. The approach taken here was to apply an average need score to each A C G , and then assess to what extent variations in standardized (or adjusted) need was determined by income. This does not demand that the average use of health care services that determines the need score be correct. It is enough to have an average and then test whether deviations from the average are systematically related to income. A simple tabulation of the 2001 CCHS data from British Columbia respondents (using the Public Use Microdata File) shows that the health utility index at the 1st percentile is 0.134 and at the 5 th percentile is 0.368. The median health utility index is 0.931. Somewhat more than 10% of the population has a health utility index of 1, indicating perfect health status. Comparisons of the ill health scores across years must be made with caution. The ill helath score is based on average system response, which will by affected by inflation. This effect does not, however, appear to flow through to the calculation of Gini coefficients or concentration indexes. Re-running the 2002 data using a deflator of 17% (based on CPI growth between 1992 and 2002) has no impact on the results. I l l There were large inequalities in need for health care services in British Columbia in both 1992 and 2002. Income played a significant role in those inequalities, though there are clearly other influential factors as well. It was not the intent of the present work to estimate what role the health care system played in changes in need or income-related need over time. It is within this context of unequally distributed need, however, that the health care system operates. 112 Chapter 4: Equity in health care services use 4.1 Introduction (In)equality and (in)equity in the use of health care services have also been the subject of study by the ECuity group (van Doorslaer et al. 2000; van Doorslaer et al. 2004a; van Doorslaer et al. 2004b; van Doorslaer and Jones 2004; van Doorslaer and Wagstaff 1992; Wagstaff and van Doorslaer 2000). In the case of health care services, inequalities are differences in the use of health care services, and inequities are differences in the use of health care services that are systematically related to some other variable that a majority of a population would look dimly upon (in this case income). In general, income-related variations in the use of health care services differ by type of health care service but are generally consistent across countries. Most countries, for example, have a great deal of inequality in the use of GP (and other) services. There is, however, little inequity in the use of GP services, but some (and sometimes substantial) inequity favouring higher income groups in the use of specialty physician services. The opposite trend is generally seen for inpatient hospital services, the use of which is also inequitable, but in this case "favors" lower income groups. This previous research has all been based on survey data, which generally included information on all variables necessary to the analyses. That is, the surveys included self-reported income, self-reported health status and self-reported health care services use. There are well-known limitations to survey data. For example, both health (objectively measured) and income influence the way individuals answer survey questions about health status, such that subjective health measures tend to over-estimate the relationship between income and health status (Humphries and van Doorslaer 2000). Individuals also tend to under-report their use of services or supports (Evans and Crawford 1999; Warburton and Warburton 2004). Finally, there is some evidence that surveys, despite the best efforts of those designing them, may not accurately represent the population, for example in missing respondents in the tails of the income distribution (Frenette et al. 2004). A key objective of this thesis was to exploit the availability of person-specific, (reasonably) comprehensive administrative data to overcome these limitations in the analysis of equity of health care services use. Administrative data can be expected to provide a more precise measure of health care services utilization. What is recorded is what happened, subject to administrative 113 errors. The data may be incomplete, because of the way they are collected, but what is there can be assumed to be relatively accurate (Roos et al. 1993; Warburton and Warburton 2004). The objectives of this chapter are to: • assess the likelihood of use of health care services by income status, after adjusting for age, sex and morbidity • assess whether the amount of health care services use, given that any use occurred, differs by income status, after adjusting for age, sex and morbidity • assess the extent of change in these relationships between 1992 and 2002 The hypotheses are: • the use of health care services overall shows no significant relationship to income, after adjusting for demographics and need for health care services. • After adjusting for demographics and need, the likelihood of use of general or family practitioner services does not vary by income status, but the use of services by those who use any is inversely related to income. • After adjusting for demographics and need both the likelihood of use and amount of use of specialist physician services are directly related to income. • After adjusting for demographics and need the likelihood of any inpatient hospital care is inversely related to income, but there is no variation in use of services by income status amongst those who use any. • After adjusting for demographics and need the likelihood of surgical day care services is directly related to income, but there is no variation in conditional use of services by income status. • After adjusting for demographics and need the likelihood of use of hospital services overall will not vary by income status, nor is there variation in the use of services by income status amongst those who use any. 4.2 Methods and data sources 4.2.1 Statistics Canada The measure of income based on tax-filer data described in the previous chapter is used here to rank individuals by income status. 114 4.2.2 The BC Linked Health Database The central file of the BC Linked Health Database is a registry of all residents who meet 78 eligibility criteria for B.C. health care insurance, register and pay the required premiums. This file includes demographic and location information for each individual regardless of whether they actually use health care services. The information used here is age (birth year and month only), sex, and postal code of residence as recorded in the middle of each calendar year. Data in the registry have been cleaned and validated, so are assumed to be accurate. For analytic purposes, individuals are aggregated into six age groups: 0-14, 15-44, 45-64, 65-74, 75-84 and 85+. Postal codes are aggregated to the five geography-based health authorities responsible for 79 health care planning and administration in British Columbia. The health care services utilization files of the BC Linked Health Data base include information on: 1) all separations from hospitals in British Columbia, as well as all out of province (and some out of country) hospitalizations for residents of B.C.; 2) services provided by physicians to B.C. residents, and reimbursed on a fee-for-service basis, as well as out-of-province services provided to B.C. residents. There are some limitations to these data. First, the physician data do not include any detailed information on the use of services provided by physicians paid by non-fee-for-service methods through an "alternative payments" arrangement (contracts, salary or through sessions — payment for services rendered in a block of time). These alternative payment arrangements have traditionally represented less than 10% of total payments to physicians, but the proportion has been rising in recent years (Canadian Institute for Health Information 2005b). Furthermore, the relative importance of alternative payments varies by region (with greater use of alternative payments in rural and remote areas of the province, which also tend to be poorer) and specialty of physician. This information gap means under-representing use of services by (some) lower income individuals, as well as loss of information on some tertiary care services (such as cancer and neonatal care) that cut across all income strata. It is in part because of this limitation of the data that a region of residence indicator is included in the analyses, though it is still possible that the analytic outcomes will underestimate income-based gradients in use purely because of missing data. 7 8 Eligibility criteria are set in part by the Canada Health Act, e.g. coverage begins in the third month of residency in the province for Canadian citizens and landed immigrants. Presence in the registry also depends on individuals signing up for the Medical Services Plan, the B.C. version of health insurance. Most people do register, but comparisons of registry and population data suggest that young males are among the population groups where there is under-counting in the registry. 7 9 Information about these five health authorities, from the second edition of the B.C. Health Atlas, can be found here: 115 The analysis requires the conversion of all health care services utilization into a common metric. In this case, the choice is expenditures, in part because this is a standard approach, and also because the physician payment files include information on actual dollars paid, and there are methods available for converting hospital separations to dollars. The alternative would be to use counts of visits or service rather than expenditures, but this approach has limitations. That approach assumes that all visits are equivalent, regardless of whether provided by a general practitioner or a specialist, or whether a simple office visit or a more involved consultation. Furthermore, it is not clear how visits could be equated across hospital and physician services. Since one of the objectives here is to look at equity in overall service utilization, this is a major consideration. Finally, the ability to make cross-sectional comparisons using number of visits would rest on an assumption that the mix of things called "visits" was relatively static across time. There is no attempt to adjust to expenditures to constant dollars, because the primary questions of interest do not require this. Instead, the primary interest is in changes in use by income stratum. Each database, the associated approach to data preparation and conversion of service units to expenditures (where necessary) is described in turn. Physicians Fee-for-service records in the BC Linked Health Database include a study ID, a date of service, a fee item indicating the service provided, a specialty code indicating the most recently registered specialty of the physician providing the service and the total amount paid for the service. The 2002 data included payments to out-of-province physicians for services provided to B.C. residents. Similar data for 1992 were not available. The inclusion of these data in one year and not the other introduces the possibility that changes over time might reflect the additional data rather than actual changes in patterns of use. The decision was to include the data for completeness, given that these out-of-province services are an important source of care for residents of B.C., especially for those who live in areas that border Alberta. Analyses were, however, conducted excluding these payments to test for the potential influence on their addition for cross-year comparisons. In any case, the out-of-province payments are included only in "total physician" and "total expenditure" groupings, because it is not possible to identify specialist vs. generalist services in the data. 116 There is interest in looking at equity in the use of physician services overall, as well as equity in the use of general/family practitioner services and specialty services separately. With this in mind, processing of the data proceeded in the following steps: 1. A l l records of service provided by non-physicians (e.g. chiropractors, osteopaths, physical therapists) were removed. This eliminated about 11.5 million records of a total 52+ million record for services provided in 1992 and less than 3 million of a 81 total of more than 115 million records in 2002. 2. The "family practice" grouping was defined as records with a provider specialty code indicating "general practice" (there is no code for family practice). A l l other records were assigned to a "specialty group 1", indicating that those services were provided by a physician other than a general practitioner. 3. As an alternative, we created a "primary health care" grouping, which included internal medicine, paediatrics, emergency medicine, geriatric medicine and obstetrics / gynecology. In this case, all other records were assigned to a "specialty group 2". This grouping captures a larger group of practitioners who are thought to be providing a significant proportion of primary care services. 4. Interest is really in the type of service provided rather than in the characteristics of the provider.82 Because of this, and following the work of a large primary health care project using the same data source83 the decision was to assign groups based on their likelihood of billing fee items. The idea was to identify the functional (specialty) "ownership" of each of the several thousand fee items in the fee schedule that governs fee-for-service payments to physicians, and use this functional ownership to determine whether a particular fee item is part of the family practice, primary health care or specialty groupings. 5. To do this, the frequency of billing for each fee item in each year was calculated, as well as the proportion of those billings that were submitted by physicians in the Data preparation for this chapter was conducted using SAS software. Sample programs are provided as Appendix 3. 8 1 The increase in number of records does not necessarily reflect an equivalent increase in the number of services provided. There may be changes in the way certain services, such as laboratory services, are billed that could inflate the number of records. The non-physician records decreased in 2002 because of change in reimbursement policies that significantly curtailed their coverage under the public plan, effective January 2002. (The latter is detailed on page 14 of the 2001/2002 Annual Report of the British Columbia Ministry of Health Services, 8 2 This is based on the perspective of the service the patient receives rather than the (training) specialty of the practitioner providing the service. For example, a general service provided to a child by a paediatrician, where there was no referral, is "primary care" or "general practice", but not paediatrics, despite the specialty of the physician. 8 3 Publications are not yet available, but a description of the program of research is at accessed 2 February 2006. 117 family practice and primary health care groups. Fee items for which this percentage was equal to or greater than 70% were assigned to the family practice / primary health care group; those for which it was less than 30% were assigned to the specialty groups. The choice of the 70% cut-off is based on previous research on primary care using the BC Linked Health Database. 6. In a small number of cases (about 3 percent of records) the percentage fell between 30 and 70; these were assigned to the family practice, primary health care or specialty groups based on the (training) specialty of the providing physician. 7. Once each fee item was assigned to a particular group, total expenditures were accumulated for each individual over the course of the year84, creating a measure of 1) total expenditures 2) family practice expenditures 3) specialty group 1 expenditures 4) primary health care expenditures and 5) specialty group 2 expenditures. By definition: Total expenditures = family practice expenditures + specialty group 1 expenditures = primary health care expenditures + specialty group 2 expenditures Hospitals Records in the hospital data set (or "discharge abstract database") are generated when patients leave (separate from) the hospital through discharge or death (the latter is rare). This applies both to inpatient hospital stays (acute care) and admissions that occur for same day services (surgical day care). The hospital records generated for day services and inpatient acute care are identical and include: study ID; hospital number; level of care, which differentiates inpatient acute care from day surgery; up to 16 diagnoses; diagnosis types, indicating which was the most responsible for the length of the hospital stay, which contributed to the stay, and which arose after admission; and Resource Intensity Weight / Day Resource Intensity Weight, indicating the relative service intensity of acute and surgical day care stays. Hospital services were converted to expenditures using Resource Intensity Weights, Day Resource Intensity Weights and costing data from the B.C. Ministry of Health (Lee 2002). The Resource Intensity Weights and Day Resource Intensity Weights are intensity scores developed and implemented by the Canadian Institute for Health Information. Intensity is measured by 8 4 Some of the records on the physician payment file have a negative paid amount. In these cases, the record indicates a reversal of a previous payment, and are in most cases accompanied by another record indicating the appropriate (fixed) payment. No special care was given to these, or to 0 dollar records, since the negatives will be netted in the course of adding over the year and the 0s will have no effect. In a handful of cases, the totals for the year summed to a negative number, likely indicating reversals of payments at the beginning of the payment year. In these cases, the negatives were transformed to 0s. 118 estimating (relative) resource requirements for Case Mix Groups, in some cases stratified by age 85 group. Case Mix Groups are, in turn, groupings of diagnostic codes (ICD-9 ) based on clinical similarity. In brief, the principal diagnosis of a hospital stay, indicating the primary influence on the length of stay, is used to classify the stay into a Case Mix Group, which in turned is combined with age to estimate a Resource Intensity Weight. The implication is that a stay with a Resource Intensity Weight of 2 is, on average, twice as costly to provide as a stay with a Resource Intensity Weight of 1. As might be expected, the Day Resource Intensity Weights, for surgical day care cases, all tend to be far lower than Resource Intensity Weights for acute inpatient care, and are in most cases well below unity (Table 14). A l l acute and surgical day care separations in both the 1992 and 2002 hospital data had a Resource Intensity Weight or Day Resource Intensity Weight > 0 assigned to them; missing data does not present a problem in converting the hospital stay to a cost using the Resource Intensity Weights and the Day Resource Intensity Weights. Table 14: Resource intensity weight by level of care, 2002 <0.25 0.25- 0.5 0.5-1 >1 Total Acute care 27,639 72,644 149,126 127,086 376,495 Surgical day care 202,504 83,695 5,317 3,791 295,307 Total 230,143 156,339 154,443 130,877 671,802 The calculation of intensity weights has changed over time, based on better understanding of relative costs of services and changes in the composition of Case Mix Groups. When changes are made, older hospital data are updated with the newer versions of the methodology. This study used the 2001 methodology, which was applied to both the 2002 and 1992 data, meaning that the same Resource Intensity Weights and Day Resource Intensity Weights were applied in both years. This means that no "noise" was introduced to the analysis through using different weights in the two years. Estimated costs per hospital separation can be calculated by multiplying each record's intensity weight by a cost per weighted case. The calculation of the cost per weighted case requires identification of both total hospital costs associated with providing acute and day care services (the numerator) and the total number of resource intensity weighted cases for acute and 8 5 Hospitals in B.C. coded using ICD-9 (or ICD-9-CM) in 1992. In 2002 they coded using ICD-10, and these codes were back-coded to ICD-9 by CIHI. 8 6 The result is an estimated cost rather than a specific case cost. In other words, two individuals in the same age group assigned to the same Case Mix Group will automatically be assigned the same intensity score. 119 day surgery combined (the denominator). The denominator is the simple sum of Resource Intensity Weights and Day Resource Intensity Weights assigned to each hospital record. The numerator requires detailed information about hospital costs. The B.C. Ministry of Health provided total hospital-specific costs for acute and day care services (Lee 2002). The method used to calculate these costs was to take the overall hospital budget and remove from it cost components unrelated to providing either inpatient or day care services, such as extended care and outpatient services. The proportion of the overall hospital budget accounted for by these other areas was applied to support and administrative costs, to reduce those to the estimated acute and day care-related amount as well. The general method was the same in 1992 and 2002, but the earlier year required more estimating because of the financial reporting systems in place at the time. With the numerator and denominator of the weighted cost per case calculation in place, processing of data occurred in two ways. First, a cost per case was calculated for each hospital separation based on the provincial average cost per Resource Intensity Weight and Day Resource Intensity Weight. Second, a hospital-specific cost per case was calculated by dividing hospital-specific total costs, as calculated by the B.C. Ministry of Health, by the hospital-specific sum of Resource Intensity Weights and Day Resource Intensity Weights, as calculated using the 2002 CIHI methodology.87 Different equity results for hospital-specific vs. provincial average costing would indicate differences in admission to higher and lower cost facilities by income group. It was then necessary to move from costs per case to hospital costs per individual, through a simple summation of costs at the level of study id. In some cases, patients are in acute care beds across years, for example admitted in December 1991 and discharged January 1992.88 The data available for analyses were based on separation date. That is, the 1992 data include all records of hospitalization that ended in 1992, including the days of care provided in 1991. At the other end of the year, there are patients admitted in December 1992 (for example) who do not leave hospital until January 1993 (or later). This means that a certain portion of care provided to patients who entered the acute care system in 1992 but were discharged in 1993 (or later) is excluded from these analyses. This year There were a few cases, in both years of data, where a hospital number did not have a cost figured provided by the Ministry of Health. In these cases, the provincial average cost per weighted case was applied. By definition, this applies to acute care but not to day care cases. 120 crossover affects about 3% of the days of care. The assumption made here is that what is lost on 89 one end is made up on the other. There are no adjustments to deal with this issue. "Linking" data from Statistics Canada and the BC Linked Health Database Data from Statistics Canada are, as discussed in the previous chapter, based on neighbourhoods rather than individual persons or families. The connection of these two data sources is thus made at an ecological level. That is, anonymized, individual-level data on need for health care services and health care services use are assigned to one of the thousand income bands based on postal codes of record in 1992 and 2002. The BC Linked Health Database includes one postal code per year for each (registered) individual. One or more postal codes are present in the central registry file for each B.C. resident, depending on the number of years he or she has been registered. In addition, postal code information is associated with each record of health care services use in the physician payment and hospital files. A comparison in the mid-1990s of two B.C. Ministry of Health sources for postal code data showed little difference in assignment to local health area or health region (McGrail and Wong Fung 1999). But neither of these files was based on health care services use, and there has not been any analysis of agreement / disagreement across multiple files. Nor can it be assumed that good agreement at the local health area level necessarily implies good agreement for the small-area assignment necessary for this project. Given the importance of postal codes for connecting income with data from the BC Linked Health Database, a further investigation of postal codes was in order. What do we know about postal codes in the BCLHD? The B.C. Ministry of Health Services has two major sources of client-based information collected to the level of mailing address, including postal code. One is the Client Registry, established in the early 1990s as the system that was meant to provide and track Personal Health Numbers. The other is the Registration and Premium Billing file, in existence in one form or another since the start of publicly funded physician services in British Columbia. The purpose of the Registration and Premium Billing file, or R&PB, is to identify who, at any particular time, is registered with the medical services plan, who is responsible for paying the premiums associated with that coverage, and when that responsibility lies with the individual, how to contact him or her. A more exacting approach would be to accumulate several years of hospital data and process them to create a file of total days of care provided in 1992 and 2002, irrespective of dates of admission and discharge. That was not possible here because more recent years of hospital data were not available. 121 As noted in previous analyses of postal code agreement between these two data sources, the difference in the purposes of these two files dictates differences in how addresses are collected, maintained and updated (McGrail and Wong Fung 1999). Figure 15 provides an overview of the flow of information relating to postal codes in the structure of health insurance and health care services use in British Columbia. Figure 15: What do we know about postal codes in the BC Linked Health Database? Jane Public p.c. Jane Public p.c. Jane Public p.c. Employers PharmaNet Hospitals Vital Stats Registration and P remium B U l m s file B C L H D When individuals make contact with different service areas of the health care system, or with Vital Statistics for registration of a birth or death90, they are often asked for identification and to verify their current home address. In the case of PharmaNet, hospitals, and Vital Statistics (i.e. when a person receives a prescription drug, enters hospital, or dies (or in the event of a birth)), that information may be communicated to the Client Registry.91 The bi-directional arrows between some hospitals92 and Client Registry, and Vital Statistics and Client Registry, indicate a 9 0 In British Columbia, the Vital Statistics Agency is a special operating unit within the Ministry of Health. 9 1 There is no assumption or proof that this is always the case. 9 2 In fact, this applies only to a few hospitals, not all of them. 122 special relationship with these program areas because, in the case of newborns, there is an initial assignment of Personal Health Numbers. When individuals have contact with their physicians, office staff may also ask for confirmation of current address. This information may be passed on to the Ministry of Health as part of a request for payment for services rendered, but it does not appear that that postal code information is collected or used in any way by the physician payment or registration (R&PB) functions of the Ministry. Even if it is available it may not be used. The R&PB address information, on the other hand, can be updated either by patients directly through an on-line system or through their employers, in cases where employers pays for health insurance premiums.93 One of the long-standing concerns with postal codes in British Columbia, however, is that there is no incentive for employers to pass employee address changes along to the Ministry. In cases of employer premium payment, the Ministry has neither a need for nor a clear interest in updating address information for registrants. The incentive to have up-to-date addresses is more obvious where individuals and families are responsible for paying their premiums directly, but this is the case for only about one third of provincial residents.94 The BC Linked Health Database derives its address information from an annual snapshot of the R&PB file. The R&PB file also provides updated information to the Client Registry file, but the information does not flow in the other direction, from the Client Registry to the R&PB file, and thus to the BC Linked Health Database. The situation can thus be summarized as follows: • patient encounters with the health care system are not used to provide routine updating of the Registration and Premium Billing file; • the flow of information implies that the Client Registry (which is not available to the BC Linked Health Database) will have more accurate postal codes through updating of address information form other files, although they are not validated in any systematic way; The R&PB address is that of the coverage holder, which may not be the same as the dependants, especially in the case of university students. Eligibility rules for MSP also create confusion in addresses; individuals may have out-of-country addresses and be B .C. residents eligible for ongoing MSP coverage. The MSP service verification audit system had about a 95% response rate in the mid-1990s using R&PB addresses. That response rate is now (2005) around 70%. The return to sender percentages increased slightly, but it can be assumed that people who receive service verification mail for prior residents may just throw it away. In 1992 when the PHNs were first introduced the R&PB address quality was very good, but it has degraded considerably due to employers not updating their employee addresses with R&PB. Large employers with many job sites are more prone to this problem, and the B.C. government is one of the worst offenders. 123 • BC Linked Health Database postal codes are particularly suspect where employers provide premium coverage. In this case, postal codes may be out of date, reflecting previous rather than current place of residence; • While physician encounter data provide the appearance of an opportunity to either validate or update postal codes in the R&PB file, in fact, there is virtually complete agreement between the two sources, indicating that the physician data are drawing their postal code of record from the R&PB file.95 • Hospital data may offer an opportunity to update postal codes, but only a small percentage of individuals (10-12%) have an encounter with the hospital system in any given year. It would be more promising to use a source like PharmaNet to update files, but those data were not available to the present project, nor are they currently part of the BC Linked Health Database. • Therefore, given the files currently available through the B C Linked Health Database, there are certainly some errors in the recording of individual level postal codes, but there is little that can be done to improve that situation. The final analysis file For all utilization files, after the processing described above was complete, records were sorted by study ID, and the total amount paid was summed for each study ID. Sums were transferred to a central analytic file that included study ID, sex, age, regions of residence, income stratum and expenditure totals for each health care service sector. The resulting list of fields available for analysis was: Study ID Sex Age Region of residence (five health authorities) Income band Various income concepts associated with each band (as described in Chapter 3) Need fo health care services (ACG, as described in Chapter 3) General / family practice expenditures (two measures) Specialist expenditures (two measures) Total physician expenditures Acute hospital expenditures (two measures) Surgical day care expenditures (two measures) Total hospital expenditures (two measures) 'This does not imply that the postal codes physicians submit with their request for payment do not differ. But in the process of moving from a billing file to a payment file, the Ministry appears to substitute whatever postal code is submitted by the physician with the postal code that is current on the Registration and Premium Billing file for that individual, based on the Personal Health Number. 124 Total expenditures (two measures) 4.2.3 Analysis The methods used here were very similar to those used to analyze need for health care services (Chapter 3). The distribution of health care services use, measured in terms of expenditures, with respect to income status, is assessed using concentration curves and concentration indices, both prior to and after adjusting for need, as determined by ACGs. A major difference between the two sets of analyses is in the treatment and use of the A C G variable. The analyses in the previous chapter used the average need scores for each A C G as the outcome variable. In the case of health care services use, the outcome variable is actual expenditures, either in raw form for the unadjusted analyses, or indirectly standardized in the case of adjusted analyses. As before, the standardization is conducted using a regression analysis approach, as: Also as before, the x variables, the standardizing variables, include the demographics age group, sex and region of residence.96 In this case, however, the x variables also include the A C G , as an estimate of need for health care services, coded for the regression model as a series of dummy variables. The vector of z variables, the control variables, is the natural log of income. These variables are used to develop a model of health care services use. Again, as in the case of the analysis of need for health care services, a predicted is generated for each individual using the actual values for the values for the x variables, but replacing the z vector variables with their means. This predicted;; can then be used to produce a standardized y value for each individual, as: where yt is each individual's expenditures on health care services, y is predicted expenditures, and y is the mean of expenditures across all individuals. The interpretation of the standardized y is the same as before. The standardized outcome for each individual is the actual expenditure value, less the predicted value, plus the mean of expenditures across all individuals. The first two terms on the right hand side of the equation measure the deviation of the particular y value from its predicted value. The addition of the last term scales that difference so that all deviations 9 6 As before, age and sex are combined in the model in a series of dummy variables, for example males aged 0-14, females aged y* = y i - y + y , 65-74. 125 (positive or negative) vary around the mean. Variation in the standardized y across income groups therefore represents the variation by income groups in use of services that is not associated with differences in age, sex, region of residence or need for health care services (as measured by ACGs) (van Doorslaer et al. 2004b).97 If there is a relationship between income and health care services use after standardizing for age, sex, region and need, then the standardized y values across income groups will exhibit a pattern. That pattern can be captured either in picture form with the concentration curve or as a scaled index with the concentration index. One of the drawbacks to the use of ACGs is that the assignment to a category other than "non-user" is contingent on using health care services. One way to deal with this drawback is to divide the health care services use analyses into two parts — an analysis of the likelihood of using a service, and an analysis of the use of services, conditional on some use. An income effect in the likelihood of services use indicates income-related variations in (realized) access to health care services. Analysis of a binary outcome is still possible with the methods described, if a linear probability model rather than a logistic model is used. It is still possible to calculate the concentration index using the regression methods described in the previous chapter. In the analyses of service use conditional on some use, each type of health care service (e.g. acute inpatient hospital care, day care surgery) is analyzed separately. The conditional models are based on use of those particular services. That is, the first model for general practice services looks at whether the likelihood of use varies by income. The second conditional model then looks at whether total expenditures on general practice services vary by income, given that use of general practitioners is greater than zero. Previous work by the ECuity group has shown that the linear specification of the two models does not produce results that are significantly different from the more traditional (but less flexible) logit (part one) and negative binomial (part two) specifications (van Doorslaer et al. 2004b). The present work uses two separate models rather than a formal two-part model. The difference is that there will be no explicit computational connection between the two models; the The ECuity group generally include only demographic variables in the vector of x variables. The reasoning is that the x variables should include only those outside of policy control. Standardizing for}' based on the means of the z variables then shows how much inequality there is in the outcome variable that cannot be explained by whatever is in the vector of x variables. This is a useful approach when the intent is then to decompose or disaggregate that total (leftover) inequality into its constituent components, such as income, region of residence, and whatever other health-influencing variables might be available. Since the aim here is to identify the amount of income-related inequity, rather than to decompose total inequity, the choice was to standardize for all available variables, and to predict the outcome based on the average (across all observations) of income. 126 predictions of the first are not used in the analyses undertaken in the second. Instead, the purpose of the two models is to assess two different questions. In the case of health care services, the unstandardized concentration index is often referred to as depicting income-related health care inequality and the standardized concentration index as depicting income-related horizontal inequity - or treating people at the same illness level differently based on income (van Doorslaer et al. 2004b). While the term "horizontal inequity" is not used here, it is important to recognize that what is being measured is the extent to which income is associated with differences in expenditures for individuals, after removing the potential influence of age, sex, region of residence and need. This analytic approach assesses the extent to which income dictates differences in access to and use of health care services for individuals who are otherwise the same. 4 . 3 Results The overall study populations for 1992 and 2002 totaled just over 3 million and 3.8 million, as described in the previous chapter (see Table 9). There were slightly more females than males in the study population, with about half the population in the 15-44 age group, though this proportion declined over time. More than half the population lived in the two health regions that make up the Lower Mainland of British Columbia (Fraser and Vancouver Coastal), and this concentration increased between 1992 and 2002. Less than 10% of the population of British Columbia lived in the very large geographic area of the Northern Health Authority. There were 413,240 individuals in 1992 (681,325 in 2002) that did not use any health care services during the year (Table 15).98 In both 1992 and 2002 these non-users were more likely to be male and in the 15-44 age group, consistent with other research on use patterns. The non-users were also more likely to live in the Northern health authority and less likely to live in the Fraser health authority. The former finding may well relate to the higher prevalence of non-fee-for-service physician arrangements in the relatively sparsely populated northern half of the province (McGrail et al. 2004). This is why a regional adjuster is added to the regression analyses. Finally, the non-users were slightly more likely to be assigned to the lower income deciles, with this difference more pronounced in 2002. This suggests there may be differential access by income group. The data include some individuals who have minor use of health care services. For purposes here, "non-users" are all individuals who have total expenditures that are less than the cost of one physician visit ~ $23.50 in 1992 and $29.47 in 2002. 127 Table 15: Population characteristics for users and non-users of health care services 1 9 9 2 2 0 0 2 Non-users Users Non-users Users Total population 413,240 13.6% 2,629,175 86.4% 681,325 17.7% 3,175,634 82.3% Males 262,390 63.5% 1,205,257 45.8% 423,495 62.2% 1,468,885 46.3% Females 150,850 36.5% 1,423,918 54.2% 257,830 37.8% 1,706,749 53.7% Age group 0-14 83,930 20.3% 524,663 20.0% 136,000 20.0% 520,032 16.4% 15-44 204,470 49.5% 1,164,315 44.3% 339,720 49.9% 1,280,390 40.3% 45-64 91,430 22.1% 550,883 21.0% 160,866 23.6% 868,646 27.4% 65-74 21,938 5.3% 225,165 8.6% 23,091 3.4% 263,847 8.3% 75-84 8,556 2.1% 127,967 4.9% 12,620 1.9% 181,848 5.7% 85+ 2,916 0.7% 36,182 1.4% 9,028 1.3% 60,871 1.9% Income decile 1 42,259 10.2% 263,605 10.0% 77,826 11.4% 310,008 9.8% 2 41,578 10.1% 262,621 10.0% 70,917 10.4% 313,627 9.9% 3 41,989 10.2% 260,723 9.9% 72,327 10.6% 315,179 9.9% 4 43,344 10.5% 262,452 10.0% 70,568 10.4% 315,528 9.9% 5 41,570 10.1% 261,397 9.9% 68,308 10.0% 314,792 9.9% 6 42,484 10.3% 263,600 10.0% 67,154 9.9% 318,317 10.0% 7 41,155 10.0% 261,670 10.0% 67,842 10.0% 320,118 10.1% 8 40,025 9.7% 264,337 10.1% 64,508 9.5% 324,672 10.2% 9 40,023 9.7% 265,782 10.1% 62,081 9.1% 320,968 10.1% 10 38,813 9.4% 262,988 10.0% 59,794 8.8% 322,425 10.2% Regional health authority Interior HA 79,326 19.2% 454,280 17.3% 117,432 17.2% 527,405 16.6% Fraser HA 114,171 27.6% 822,498 31.3% 216,758 31.8% 1,087,288 34.2% Vancouver Coastal HA 102,366 24.8% 671,267 25.5% 182,218 26.7% 798,173 25.1% Vancouver Island HA 68,992 16.7% 470,533 17.9% 100,907 14.8% 544,264 17.1% Northern HA 48,385 11.7% 210,597 8.0% 64,010 9.4% 218,504 6.9% Nearly all individuals identified as "users" of health care services had some contact with family or general practitioners in both 1992 and 2002 (Tables 16 and 17). This reflects the strong gatekeeper function that general practitioners play in the health care system in British Columbia. About 59% of system users had at least one contact with a specialist (using the broader definition) in 1992. By 2002 two-thirds of service users consulted with a specialist physician. This trend is even more pronounced - an increase from 55% to 65% - for the specialist grouping that excludes "primary care specialties" (paediatrics, obstetrics / gynaecology and internal 128 medicine). There is an increase overall in the proportion of the population that saw a specialist, but the increase is greater for the non-primary care specialties. The likelihood of at least one referral to a specialist increases with age in both years, except for a slight drop off in the oldest age group, and the increase across the years is itself larger in the older age groups. By 2002 nine out of ten individuals age 75-84 who used any health care services received some services from a specialist during the year. Given that 94% of this age group are "users" (Table 15), an overwhelming majority of this age group receives specialist services. 129 Table 16: Number and percentage of population that use services by health care service type, 1992 (excluding non-users) Physicians Hospitals Total General Specialist Primary health care Specialist Total fee-for-service practitioners (definition 1) provid ers (definition 2) physcian Acute care Surgical day care Total hospital %o f % o f %o f %o f %o f % o f % of total total total total % o f total total total user user user user total user user user N pop N pop N pop N pop N user pop N pop N pop N pop N Total population 2,603,156 99.0% 1,557,799 59.3% 2,607,845 99.2% 1,440,469 54.8% 2,627,561 99.9% 266,945 10.2% 178,464 6.8% 408,901 15.6% 2,629,175 Males 1,189,240 98.7% 660,853 54.8% 1,191,554 98.9% 617,965 51.3% 1,204,348 99.9% 107,504 8.9% 75,965 6.3% 166,945 13.9% 1,205,257 Females 1,413,916 99.3% 896,946 63.0% 1,416,291 99.5% 822,504 57.8% 1,423,213 100.0% 159,441 11.2% 102,499 7.2% 241,956 17.0% 1,423,918 Age group 0-14 520,481 99.2% 197,305 37.6% 521,467 99.4% 166,083 31.7% 524,476 100.0% 33,043 6.3% 15,674 3.0% 47,005 9.0% 524,663 15-44 1,151,229 98.9% 673,024 57.8% 1,153,505 99.1% 616,190 52.9% 1,163,280 99.9% 110,546 9.5% 72,502 6.2% 171,700 14.7% 1,164,315 45-64 544,760 98.9% 381,866 69.3% 545,739 99.1% 365,193 66.3% 550,608 100.0% 49,783 9.0% 43,503 7.9% 83,919 15.2% 550,883 65-74 223,492 99.3% 174,996 77.7% 223,775 99.4% 168,142 74.7% 225,096 100.0% 35,304 15.7% 26,207 11.6% 53,966 24.0% 225,165 75-84 127,223 99.4% 103,153 80.6% 127,354 99.5% 99,101 77.4% 127,940 100.0% 28,546 22.3% 17,174 13.4% 40,350 31.5% 127,967 85+ 35,971 99.4% 27,455 75.9% 36,005 99.5% 25,760 71.2% 36,161 99.9% 9,723 26.9% 3,404 9.4% 11,961 33.1% 36,182 Income decile 1 260,607 98.9% 156,581 59.4% 261,216 99.1% 143,573 54.5% 263,323 99.9% 33,630 12.8% 18,260 6.9% 47,635 18.1% 263,605 2 260,219 99 .1% 152,894 58.2% 260,684 99.3% 140,922 53.7% 262,441 99.9% 28,888 11.0% 17,678 6.7% 42,843 16.3% 262,621 3 258,543 99.2% 152,060 58.3% 258,906 99.3% 140,515 53.9% 260,551 99.9% 28,939 11.1% 18,115 6.9% 43,140 16.5% 260,723 4 260,151 99 .1% 152,779 58.2% 260,550 99.3% 141,195 53.8% 262,273 99.9% 27,945 10.6% 17,743 6.8% 41,937 16.0% 262,452 5 259,149 99.1% 153,540 58.7% 259,505 99.3% 141,828 54.3% 261,207 99.9% 27,094 10.4% 17,832 6.8% 41,204 15.8% 261,397 6 261,180 99.1% 152,988 58.0% 261,567 99.2% 141,251 53.6% 263,452 99.9% 26,303 10.0% 17,496 6.6% 40,229 15.3% 263,600 7 258,951 99.0% 154,842 59.2% 259,435 99.1% 142,946 54.6% 261,518 99.9% 25,484 9.7% 17,837 6.8% 39,800 15.2% 261,670 8 261,776 99.0% 157,558 59.6% 262,228 99.2% 146,032 55.2% 264,218 100.0% 23,983 9.1% 17,790 6.7% 38,304 14.5% 264,337 9 263,083 99.0% 159,623 60.1% 263,546 99.2% 148,334 55.8% 265,677 100.0% 23,380 8.8% 17,860 6.7% 37,925 14.3% 265,782 10 259,497 98.7% 164,934 62.7% 260,208 98.9% 153,873 58.5% 262,901 100.0% 21,299 8.1% 17,853 6.8% 35,884 13.6% 262,988 Regional health authority Interior HA 450,359 99.1% 257,971 56.8% 451,062 99.3% 237,407 52.3% 453,957 99.9% 53,627 11.8% 31,375 6.9% 78,050 17.2% 454,280 Fraser HA 816,273 99.2% 488,432 59.4% 817,280 99.4% 451,356 54.9% 822,120 100.0% 78,267 9.5% 57,152 6.9% 123,511 15.0% 822,498 Vancouver Coastal HA 663,099 98.8% 413,013 61.5% 664,765 99.0% 383,474 57.1% 670,861 99.9% 59,252 8.8% 40,584 6.0% 92,146 13.7% 671,267 Vancouver Island HA 466,146 99 .1% 284,628 60.5% 466,810 99.2% 264,954 56.3% 470,303 100.0% 48,998 10.4% 36,726 7.8% 78,645 16.7% 470,533 Northern HA 207,279 98.4% 113,755 54.0% 207,928 98.7% 103,278 49.0% 210,320 99.9% 26,801 12.7% 12,627 6.0% 36,549 17.4% 210,597 130 Table 17: Number and percentage of population that use services by health care service type, 1992 (excluding non-users) Physicians Hospitals Total General Specialist Primary health care Specialist Total fee-for-service practitioners (definition 1) provid ers (definition 2) physcian Acute care Surgical day care Total hospital % of %of %of %of %of %of % of total total total total total total total user user user user % of total user user user N pop N pop N pop N pop N user pop N pop N pop N pop N Total population 3,129,090 98.5% 2,117,770 66.7% 3,136,974 98.8% 2,050,954 64.6% 3,175,269 100.0% 227,854 7.2% 237,447 7.5% 425,062 13.4% 3,175,634 Males 1,442,429 98.2% 921,131 62.7% 1,446,691 98.5% 889,462 60.6% 1,468,698 100.0% 90,678 6.2% 104,990 7.1% 176,600 12.0% 1,468,885 Females 1,686,661 98.8% 1,196,639 70.1% 1,690,283 99.0% 1,161,492 68.1% 1,706,571 100.0% 137,176 8.0% 132,457 7.8% 248,462 14.6% 1,706,749 Age group 0-14 511,339 98.3% 210,580 40.5% 513,725 98.8% 181,592 34.9% 519,930 100.0% 22,094 4.2% 14,766 2.8% 35,556 6.8% 520,032 15-44 1,260,434 98.4% 785,891 61.4% 1,262,993 98.6% 764,681 59.7% 1,280,212 100.0% 82,138 6.4% 70,855 5.5% 144,679 11.3% 1,280,390 45-64 856,073 98.6% 672,395 77.4% 858,194 98.8% 661,070 76.1% 868,586 100.0% 49,941 5.7% 76,532 8.8% 114,257 13.2% 868,646 65-74 260,916 98.9% 232,660 88.2% 261,388 99.1% 230,105 87.2% 263,836 100.0% 29,164 11.1% 38,285 14.5% 58,904 22.3% 263,847 75-84 180,145 99.1% 163,792 90.1% 180,394 99.2% 161,986 89.1% 181,840 100.0% 30,802 16.9% 30,228 16.6% 53,177 29.2% 181,848 85+ 60,183 98.9% 52,452 86.2% 60,280 99.0% 51,520 84.6% 60,865 100.0% 13,715 22.5% 6,781 11.1% 18,489 30.4% 60,871 Income decile 1 304,877 98.3% 199,973 64.5% 305,712 98.6% 192,878 62.2% 309,933 100.0% 26,793 8.6% 21,369 6.9% 44,252 14.3% 310,008 2 309,321 98.6% 203,347 64.8% 310,058 98.9% 196,263 62.6% 313,580 100.0% 23,235 7.4% 21,737 6.9% 41,168 13.1% 313,627 3 310,098 98.4% 207,430 65.8% 310,930 98.7% 200,654 63.7% 315,137 100.0% 23,755 7.5% 23,697 7.5% 43,289 13.7% 315,179 4 311,074 98.6% 208,716 66.1% 311,817 98.8% 202,124 64.1% 315,470 100.0% 23,849 7.6% 24,318 7.7% 43,802 13.9% 315,528 5 310,893 98.8% 207,115 65.8% 311,590 99.0% 200,655 63.7% 314,773 100.0% 23,624 7.5% 23,822 7.6% 43,145 13.7% 314,792 6 314,053 98.7% 211,839 66.5% 314,766 98.9% 205,461 64.5% 318,297 100.0% 22,820 7.2% 24,290 7.6% 42,916 13.5% 318,317 7 315,529 98.6% 213,763 66.8% 316,296 98.8% 207,304 64.8% 320,082 100.0% 22,707 7.1% 24,380 7.6% 42,976 13.4% 320,118 8 320,012 98.6% 218,559 67.3% 320,806 98.8% 211,830 65.2% 324,641 100.0% 21,457 6.6% 24,602 7.6% 42,075 13.0% 324,672 9 316,372 98.6% 218,656 68.1% 317,141 98.8% 212,113 66.1% 320,951 100.0% 20,188 6.3% 24,179 7.5% 40,711 12.7% 320,968 10 316,861 98.3% 228,372 70.8% 317,858 98.6% 221,672 68.8% 322,405 100.0% 19,426 6.0% 25,053 7.8% 40,728 12.6% 322,425 Regional health authority Interior HA 520,528 98.7% 346,984 65.8% 521,748 98.9% 335,983 63.7% 527,353 100.0% 45,122 8.6% 43,084 8.2% 80,075 15.2% 527,405 Fraser HA 1,073,640 98.7% 713,609 65.6% 1,075,991 99.0% 688,851 63.4% 1,087,224 100.0% 72,058 6.6% 81,708 7.5% 140,090 12.9% 1,087,288 Vancouver Coastal HA 784,216 98.3% 541,717 67.9% 786,652 98.6% 524,183 65.7% 798,094 100.0% 47,925 6.0% 52,876 6.6% 92,758 11.6% 798,173 Vancouver Island HA 537,816 98.8% 379,646 69.8% 538,893 99.0% 369,838 68.0% 544,235 100.0% 43,813 8.0% 42,000 7.7% 78,449 14.4% 544,264 Northern HA 212,890 97.4% 135,814 62.2% 213,690 97.8% 132,099 60.5% 218,363 99.9% 18,936 8.7% 17,779 8.1% 33,690 15.4% 218,504 131 There is some gradient in the likelihood of seeing a specialist by income decile considering either specialist definition, and the gradient became more pronounced over time. In 2002, nearly 71% of individuals in the highest income decile had contact with a specialist (by the first definition) compared to 64.5% of individuals in the lowest income decile. The likelihood of seeing a specialist grew five percentage points between 1992 and 2002 for the lowest income decile, but increased by eight percentage points for the highest income group. Only 10.2%) of the user population (less than 9% of the total study population) was admitted to an acute care hospital bed during 1992, and this decreased to 7.2% by 2002. There was less change in the likelihood of use of surgical day care services, which increased from 6.8% to 7.5% of the user population between 1992 and 2002. The users of acute and surgical day care services are mostly different populations; 15.6% of the users accessed one or the other of these hospital services in 1992, implying that only 1.4% of the user population use both. This overlap decreases only slightly to 1.3% in 2002. Fewer than 2,000 individuals in the user group in 1992 and less than four hundred in 2002 had recorded hospital expenses but no recorded (fee-for-service) physician encounters. The use of both acute and surgical day care generally increases with age, as does the use of hospital services overall. The major exception is a drop-off in the use of surgical day care in the 85+ age group. On the acute side, all age groups showed a decrease in overall use between 1992 and 2002 (no surprise, as the availability of acute care beds was declining (McGrail et al. 2001)), while for surgical day care there were actually slight decreases in the likelihood of use in the two younger age groups and increases for all others. In both years, the use of acute care was inversely related to income. There were no major differences in surgical day care use across income classes in 1992, but a positive relationship in 2002, mostly because of a large jump in the likelihood of use from the 2 n d to the 3 r d income decile. 132 T a b l e 1 8 : A v e r a g e u s e o f s e r v i c e s ( i n $ $ ) 1 9 9 2 Physicians Hosp-specific costing Hosp average costing Total fee Surg day Total Surg day Total Gen Spec Prim Hlth Spec for svc Acute carecare hosp Acute carecare hosp TOT exp TOT exp Prac (def. 1) Care (def. 2) phys (def. 1) (def. 1) (def. 1) (def. 2) (def. 2) (def. 2) (def. 1) (def. 2) Total population 231 191 283 140 422 527 47 573 528 47 574 996 996 Males 196 174 239 130 369 520 45 565 515 45 559 935 929 Females 261 206 320 148 467 533 48 580 539 48 586 1048 1054 Age group 0-14 151 76 178 49 227 213 24 237 182 20 203 464 430 15-44 220 158 262 116 377 343 41 384 345 41 387 762 764 45-64 238 243 298 183 481 523 52 575 520 53 573 1056 1054 65-74 327 362 426 263 689 1170 82 1252 1189 84 1273 1941 1962 75-84 406 409 515 300 815 1844 97 1941 1912 100 2011 2756 2826 85+ 456 314 537 233 770 2387 69 2456 2505 70 2575 3226 3345 Income decile 1 265 202 323 144 467 716 49 765 715 48 763 1232 1230 2 243 189 297 136 433 576 46 622 578 46 624 1055 1057 3 241 186 292 135 428 561 46 607 582 47 629 1035 1057 4 232 185 282 135 417 534 45 579 547 46 593 995 1010 5 232 187 283 136 419 515 45 560 537 47 584 979 1003 6 224 179 272 132 404 510 45 555 510 45 555 958 959 7 225 186 275 137 412 482 47 529 477 47 524 941 936 8 221 191 271 141 412 455 47 502 451 47 498 914 910 9 218 193 268 143 411 464 47 511 449 46 495 921 906 10 212 210 265 158 422 456 50 506 430 47 477 928 899 Regional health authority Interior HA 230 175 277 128 405 578 41 619 628 45 673 1024 1078 Fraser HA 236 193 289 140 429 480 48 528 487 48 535 957 964 Vancouver Coastal HA 226 213 283 157 440 531 48 579 476 42 518 1019 958 Vancouver Island HA 239 192 288 142 431 532 52 584 554 55 610 1015 1041 Northern HA 217 144 255 105 360 574 37 610 574 39 612 971 973 133 Table 19: Average use of services (in $$) 2002 P h y s i c i a n s Hosp-spec i f i c cost ing H o s p average cost ing Total fee Acute Surg day Total Acute Surg day Total Gen Prac Spec (def. 1) Prim Hlth Care Spec (def. 2) for svc phys care (def. D care (def. 1) hosp (def. 1) care (def. 2) care (def. 2) hosp (def. 2) T O T exp T O T exp (def. 1) (def. 2) Total population 285 280 345 220 569 636 94 730 638 95 733 1299 1302 Males 251 261 308 204 517 619 97 716 621 98 720 1233 1237 Females 314 295 376 233 614 651 91 742 652 92 744 1355 1358 A g e g r o u p 0-14 158 103 197 64 263 277 42 319 195 32 227 582 490 15-44 251 205 288 167 460 405 58 463 383 58 442 923 902 45-64 289 337 354 271 630 521 107 628 521 108 629 1258 1259 65-74 426 522 547 401 956 1237 205 1442 1296 214 1510 2398 2466 75-84 566 613 713 466 1187 2122 246 2368 2332 268 2600 3554 3787 85+ 585 493 698 380 1082 3148 167 3316 3527 184 3711 4398 4793 Income deci le 1 316 285 379 223 605 844 88 932 837 86 924 1537 1528 2 298 270 358 209 570 655 87 742 659 88 747 1312 1318 3 292 276 352 216 572 669 93 761 680 96 776 1333 1348 4 292 275 352 216 572 654 93 746 665 98 763 1318 1335 5 286 272 345 213 563 649 93 742 658 97 755 1305 1318 6 282 275 340 217 564 609 94 703 621 97 718 1267 1282 7 277 273 335 215 557 618 93 712 623 96 718 1269 1275 8 274 276 332 218 554 558 94 652 565 96 661 1206 1215 9 268 284 327 224 556 546 96 642 538 95 633 1198 1189 10 266 309 331 244 579 566 107 673 541 101 642 1251 1220 Reg iona l health authority Interior HA 292 264 348 208 569 683 85 768 733 98 831 1337 1400 Fraser HA 281 272 341 212 555 541 93 634 591 97 688 1188 1243 Vancouver Coastal HA 273 300 338 235 575 659 104 763 557 89 646 1338 1222 Vancouver Island HA 308 305 372 242 616 704 88 792 737 96 832 1408 1448 Northern HA 270 218 315 172 501 740 97 837 688 95 783 1338 1285 134 The population average expenditure (among users) in 1992 was $422 for physician services and $573 for hospital services, rising to $569 and $730 in 2002 (Tables 18 and 19)." In both years, more than half of physician expenditures were for family practice or primary care services and the majority of hospital expenditures (nearly 90%) were for inpatient acute care. Females had higher average health care expenditures than males for every service category in both years, except for surgical day care in 2002, where average costs were slightly higher for men. As expected, average expenditures generally increased with age; physician expenditures for the oldest age group were more than three times the level for the youngest age group, and hospital expenditures were ten times greater for the oldest compared to the youngest. Parallel to the likelihood of any service use, average expenditures for specialist physician services and surgical day care in both 1992 and 2002 were lower amongst the eldest age group (85+) than amongst the 75-84 year olds. Also similar to the likelihood of service use, average expenditures for general or primary care physician services and for acute inpatient hospital services tended to be inversely related to income, while expenditures on specialist physician services were inversely related to income from the lowest to the middle income groups and then directly related thereafter. Expenditures on surgical day care followed a similar income-related pattern in 1992; by 2002, there was a general positive relationship between surgical day care expenditures and income. Residents of the Northern Health Authority had the lowest per capita expenditures for physician services (both general and specialist) in both 1992 and 2002. This was expected, given the prevalence of alternative physician payment arrangements in the Northern Health Authority. Table 11 provided some descriptive information about the number of individuals assigned to a select set of four ACGs, the top two for frequency and the top two according to need for health care services. Tables 20 and 21 expand the information for these four ACGs a bit, showing average expenditures by income decile and health care sector. There is no discernible pattern of total physician and hospital expenditure by income decile in any of the four A C G groups. In fact, the only clear gradient in health care expenditure by income is for general practice and specialty services in the two high-frequency ACGs in 2002. In these two groups the average expenditure for general practice services is inversely related to income and the average expenditure for 9 9 These tables show average expenditures for the user group as a whole, rather than average expenditures among those who use particular services. The average for total physician expenditures in 2002 decreases to $564 excluding payments to out-of-province physicians for services provided to B .C. residents. Finally, the reader may notice that the overall average cost reported for hospital-specific and hospital-average costing are similar but not identical. This is likely due to rounding errors in the process of calculating episode-specific costing. 135 specialty services is directly related to income. The same general phenomenon is present, but is less clear-cut, in 1992. These expenditures are specific for need category but are not standardized for age, sex or region of residence, which may vary across the income groups. In fact, all of the preceding tables are descriptive, providing comparisons of unadjusted percentages and expenditures. The descriptive analyses will only take us so far because of the number of variables involved. We proceed now to an examination of trends in the use of health care services by income status, after adjusting for age, sex, region of residence and need for health care services, all of which clearly have an impact on utilization. 136 Table 20: Average use of services (in $$) by income decile for select high frequency and high cost ACGs, 1992 Physicians Hosp-specific costing Hosp average costing Tot Acute Surg Acute Surg TOT exp TOT exp G P S p e d P H C Spec2 phys care day care Tot hosp care day care Tot hosp 1 2 Deciles High frequency ACG - Acute Minor, Age 6+ (300) 1 56 13 58 11 69 6 2 8 6 2 9 78 78 2 55 13 58 11 69 4 3 7 4 3 7 76 76 3 56 14 58 12 70 4 2 7 5 3 7 77 77 4 56 15 58 13 71 4 2 6 4 2 6 76 77 5 55 15 57 12 70 4 3 7 4 3 8 77 78 6 55 14 57 12 69 3 2 6 3 3 6 75 75 7 55 15 57 13 70 3 3 6 3 3 6 76 76 8 54 15 57 13 70 2 3 5 2 3 5 74 74 9 54 16 57 13 70 3 3 6 2 3 5 76 75 10 54 17 57 15 71 2 3 5 2 3 4 76 76 High frequency ACG -- Acute Minor, Age 6+ (1800) 1 151 63 164 50 214 104 13 116 104 13 117 330 331 2 147 64 160 51 211 91 15 106 94 15 109 317 320 3 149 66 162 53 215 93 15 109 99 16 116 324 330 4 146 66 158 53 212 87 15 103 91 16 108 314 319 5 144 67 158 54 212 95 15 110 98 17 115 322 327 6 145 67 159 53 212 100 15 115 102 16 118 327 330 7 145 67 158 54 212 83 14 98 83 15 98 310 310 8 141 67 154 54 208 80 16 96 80 16 96 304 304 9 142 69 155 55 211 75 16 90 73 16 88 301 299 10 141 73 156 58 213 61 16 77 60 15 75 291 289 High cost ACG --10+ Other ADG Combinations, Age 18+, 4+ Major ADGs (5070) 1 1356 1967 2086 1237 3323 17547 199 17746 17423 195 17618 21070 20941 2 1341 2059 2131 1270 3401 16626 226 16852 16392 221 16613 20252 20014 3 1396 2089 2150 1334 3484 17352 235 17587 17518 236 17754 21071 21238 4 1358 2134 2124 1368 3492 16979 231 17210 16968 230 17198 20702 20690 5 1413 2214 2210 1418 3627 17151 241 17393 17776 243 18019 21020 21646 6 1413 2180 2184 1409 3593 18717 228 18945 18348 225 18573 22537 22166 7 1355 2221 2145 1431 3576 17018 261 17279 16608 250 16857 20855 20433 8 1293 2172 2077 1389 3465 16806 282 17087 16478 275 16752 20553 20218 9 1256 2224 2010 1470 3480 17864 287 18151 17282 274 17556 21631 21035 10 1241 2248 2036 1453 3489 18161 287 18448 16834 266 17100 21937 20589 High cost ACG --10+ Other ADG Combinations, Ages 1 -17 2+ Major ADGs (5030) 1 980 1778 1781 976 2757 21790 347 22137 15951 233 16185 24894 18942 2 1022 1944 1805 1161 2966 19353 317 19670 13882 216 14099 22636 17064 3 929 1932 1657 1205 2861 18342 421 18763 13024 286 13310 21624 16171 4 971 1834 1651 1153 2804 26184 341 26525 17150 222 17373 29330 20177 5 868 1605 1504 969 2473 19462 377 19839 13133 248 13381 22312 15854 6 971 2018 1799 1190 2989 29661 270 29932 18734 222 18956 32921 21945 7 913 1530 1548 895 2443 23197 280 23477 17126 204 17330 25920 19773 8 985 2100 1762 1322 3084 26758 369 27127 17351 264 17615 30211 20699 9 995 2272 1792 1476 3268 22879 374 23252 14387 237 14624 26520 17892 10 770 1637 1386 1022 2408 21516 307 21824 13975 202 14177 24231 16585 137 Table 21: Average use of services (in $$) for select high frequency and high cost ACGs , 2002 Physicians Hosp-specific costing Hosp average costing GP Sped PHC Spec2 Tot phys Acute care Surg day care Tot hosp Acute care Surg day care Tot hosp TOT exp TOT exp 1 2 Deciles High frequency A C G • •- Acute Minor, Age 6+ (300) 1 63 18 64 16 81 2 3 5 2 3 5 86 87 2 63 18 64 16 81 2 4 5 2 3 5 87 86 3 62 19 64 18 83 1 3 5 1 3 4 87 87 4 62 19 63 17 81 2 3 5 2 3 5 86 86 5 62 19 63 18 82 1 3 4 1 3 4 86 86 6 61 20 63 19 82 6 3 9 2 3 5 92 87 7 61 21 63 19 83 2 3 5 2 3 5 88 88 8 62 21 63 19 84 2 3 6 2 3 6 89 89 9 61 22 63 20 84 2 3 5 2 3 5 89 89 10 61 24 63 22 86 2 3 4 2 2 4 90 90 High frequency A C G • •- Acute Minor, Age 6+ (1800) 1 174 88 188 74 264 83 21 104 88 20 108 368 372 2 171 90 185 75 263 70 20 90 77 20 97 352 359 3 170 94 185 80 266 80 23 104 87 24 111 370 377 4 170 94 185 80 267 85 24 110 89 25 114 377 381 5 170 96 184 81 268 88 24 112 94 24 118 381 386 6 169 98 183 84 270 81 24 105 85 24 109 375 379 7 169 99 184 84 271 79 25 103 82 25 107 374 377 8 166 100 181 85 268 70 22 92 76 22 98 361 366 9 165 102 180 86 269 75 25 100 78 23 102 369 371 10 165 110 183 92 277 63 23 86 63 22 85 363 362 High cost A C G --10+ Other ADG Combinations, Age 18+, 4+ Major ADGs (5070) 1 1924 2874 2911 1887 4843 24617 547 25164 24625 525 25150 30008 29993 2 1982 3002 2979 2005 5060 24353 728 25081 24952 695 25647 30141 30707 3 2006 3012 2995 2023 5097 23232 734 23966 23749 739 24488 29063 29585 4 2070 2966 3054 1982 5125 23734 786 24520 24520 785 25305 29646 30430 5 1948 3034 2907 2075 5116 24453 744 25197 24084 737 24820 30313 29936 6 1979 3049 2967 2061 5190 24126 786 24912 24318 780 25098 30101 30288 7 2001 3139 2974 2166 5291 25382 795 26178 25746 752 26498 31469 31790 8 1911 3054 2888 2077 5027 21938 753 22691 23011 720 23731 27718 28758 9 1894 3115 2888 2121 5080 22576 796 23372 22933 778 23711 28451 28790 10 1756 3217 2790 2183 5019 21768 945 22714 21186 822 22009 27732 27028 High cost A C G - 10+ Other ADG Combinations, Ages 1 -17 2+ Major ADGs (5030) 1 1423 3012 2649 1786 4554 42730 934 43664 22902 481 23383 48217 27937 2 1187 2608 2292 1503 4133 46456 1018 47474 25945 502 26447 51607 30580 3 1216 3223 2665 1774 4810 60708 742 61450 31341 395 31736 66260 36546 4 1270 2889 2353 1806 4546 46115 895 47010 25191 498 25689 51556 30235 5 1299 2963 2553 1710 4852 44946 724 45670 26136 423 26560 50521 31412 6 1153 2888 2188 1854 4377 30228 848 31076 17103 432 17534 35453 21911 7 1265 3479 2473 2271 5466 52436 793 53229 28914 457 29371 58694 34837 8 1155 3386 2568 1973 4737 48850 1198 50048 26705 636 27341 54785 32079 9 973 3002 2059 1916 4003 41973 830 42803 20518 454 20972 46806 24975 10 1141 3292 2480 1953 4477 48004 1015 49019 25064 515 25579 53496 30056 138 Tables 22 and 23 show the unadjusted (as a measure of overall inequality) and adjusted (as a measure of income-related inequality) concentration indices for the probability of use and the amount of use conditional on any use, by type of service, and for all services. These tables also show the standardized probability and average hospital and physician expenditures by income decile. The concentration index and values for which the (two-tailed) test was significant at the 0.01 level are in bold font; they are italicized for significance at the 0.05 level; and they are in normal type for non-significant results. Table 22: Inequality and inequity in health care services utilization, 1992 Physicians Hosp-specific costing Hosp average costing Acute Surg day Acute Surg day G P Sped PHC Spec2 Tot phys care care Tot hosp care care Tot hosp Tot1 Tot 2 Probability of use Inequality 0.002 0.011 0.002 0.013 0.002 -0.069 -0.002 -0.042 -0.069 -0.002 -0.042 0.002 0.002 Inc-related inequity 0.000 0.009 0.000 0.010 0.000 -0.019 0.015 -0.006 -0.019 0.015 -0.006 0.000 0.000 Decile 1 0.855 0.502 0.857 0.461 0.863 0.094 0.055 0.138 0.094 0.055 0.138 0.864 0.864 2 0.856 0.501 0.857 0.463 0.864 0.089 0.057 0.134 0.089 0.057 0.134 0.864 0.864 3 0.857 0.505 0.858 0.467 0.864 0.089 0.058 0.136 0.089 0.058 0.136 0.864 0.864 4 0.856 0.508 0.858 0.470 0.864 0.088 0.058 0.134 0.088 0.058 0.134 0.864 0.864 5 0.856 0.511 0.858 0.472 0.864 0.089 0.058 0.135 0.089 0.058 0.135 0.864 0.864 6 0.856 0.511 0.858 0.472 0.864 0.089 0.060 0.136 0.089 0.060 0.136 0.864 0.864 7 0.855 0.516 0.857 0.476 0.864 0.087 0.060 0.134 0.087 0.060 0.134 0.864 0.864 8 0.855 0.517 0.857 0.479 0.864 0.086 0.060 0.133 0.086 0.060 0.133 0.864 0.864 9 0.855 0.519 0.857 0.481 0.864 0.084 0.060 0.132 0.084 0.060 0.132 0.864 0.864 10 0.854 0.531 0.856 0.493 0.864 0.082 0.061 0.132 0.082 0.061 0.132 0.864 0.864 Conditional use Inequality -0.034 -0.003 -0.031 0.003 -0.016 -0.005 0.010 -0.024 -0.014 0.002 -0.034 -0.046 -0.052 Inc-related inequity -0.014 0.016 -0.009 0.018 0.003 -0.003 0.005 -0.011 -0.009 0.001 -0.017 -0.006 -0.010 Decile 1 245 308 295 243 423 5,392 688 3,936 5,423 687 3,960 1,040 1,043 2 237 311 288 244 418 5,266 679 3,749 5,286 681 3,766 996 999 3 239 314 289 248 421 5,101 681 3,630 5,171 684 3,681 994 1,004 4 233 317 284 251 419 5,175 681 3,690 5,248 686 3,744 982 990 5 234 319 286 251 422 5,031 670 3,621 5,151 683 3,706 984 999 6 234 318 283 254 420 5,120 678 3,749 5,191 682 3,771 1,012 1,012 7 232 325 285 257 425 5,233 687 3,618 5,218 686 3,606 991 984 8 230 329 283 261 425 5,121 696 3,621 5,045 694 3,571 983 981 9 228 331 280 263 424 5,179 693 3,614 5,088 681 3,555 989 978 10 224 339 278 270 429 5,278 705 3,637 5,130 686 3,538 986 973 139 Nearly all of the concentration indices in both Tables 22 and 23 are significant. Despite this, in some cases the numbers are quite small; for example the concentration index for the probability of use of health care services overall (having total expenditures greater than 0) is 0 when displayed to three decimal places. The very small standard errors that result from the regressions (see Appendix 4) occur in part because of the large number of observations in the data set. Some researchers have identified changes in Gini coefficients of 0.01 or greater as significant, so that will be used as a rough guide for "how large" the concentrations have to be to be considered to have "clinical significance". Table 23: Inequality and inequity in health care services utilization, 2002 Physicians Hosp-specific costing Hosp average costing Acute Surg Acute Surg day GP Sped PHC Spec2 Tot phys care day care Tot hosp care care Tot hosp Tot 1 Tot 2 Probability of use Inequality 0.008 0.022 0.008 0.023 0.008 -0.045 0.025 -0.009 -0.045 0.025 -0.009 0.008 0.008 Inc-related inequity 0.000 0.011 0.000 0.011 0.000 -0.018 0.025 0.003 -0.018 0.025 0.003 0.000 0.000 Decile 1 0.810 0.533 0.812 0.516 0.825 0.06