AFRAMEWORK FOR LONG-TERM ACTUARIAL PROJECTIONS OF HEALTH CARE COSTS: THE IMPORTANCE OF POPULATION AGING AND OTHER FACTORS

Transcription

1 AFRAMEWORK FOR LONG-TERM ACTUARIAL PROJECTIONS OF HEALTH CARE COSTS: THE IMPORTANCE OF POPULATION AGING AND OTHER FACTORS Howard J. Bolnick* ABSTRACT Ever-expanding life expectancy is increasing the size of elderly populations with profound social and economic consequences for developed nations, including future cost of their health care systems. Most existing long-term health care cost projections are driven mainly by changing demographics (aging populations). This simplified approach fails to recognize the many variables, and complicated interactions among them, affecting the future of health, health care, and health care costs. This study presents a framework incorporating key health care cost drivers. Using the framework, the study then introduces three plausible futures for health care along with broad, nonmodeled estimates of their costs that point to a very wide range of potential future costs. By taking the next step and building actuarial models based on the framework presented in this study, actuaries and health economists can create a powerful tool for health policymakers and health officials to better understand the long-term consequences of decisions taken during their stewardship of health care systems. How admirable! To see lightning and not think Life is fleeting Basho 1. INTRODUCTION Life is fleeting, yet, as fleeting as life is today, it is far less so than only a few decades ago. Both life expectancy and health improved dramatically from beginning to end of the 20 th century. These results come from effective public health programs and, even more visible to us, almost miraculous improvements in medical care. We celebrate these marvels and, also, we pay dearly for them. Developed countries are aging. Their elderly populations are growing rapidly, both in numbers and *Howard J. Bolnick, FSA, MAAA, Hon FIA, is Chairman of InFocus Financial Group Inc., 20 N. Wacker Dr., Suite 4000, Chicago, IL 60606, hbolnick@kellogg.northwestern.edu. An earlier version of this paper was presented at the Institute of Actuaries of Australia Biennial Convention in May proportions of the total. Among the potentially serious consequences of demographic maturation is widespread concern about health care costs. Health care today is already enormously costly. The primary focus of many analyses of future health care costs is on the effect of one variable, population aging. Will rapidly increasing numbers of elderly, with their disproportionately high needs, simply overwhelm health care systems? Is demographics driving our destiny, or are there other more critical factors that will likely drive health care costs during the first half of the 21 st century? The possibility that rapidly aging populations will drive health care costs wildly higher seems at first glance to be a realistic concern, however, this particular concern becomes much less certain when the future of health and health care is carefully analyzed. Developing a reasoned framework for alternative future scenarios of health, health care, and health care costs is the goal of this paper. As we will see, careful analysis of the aging problem leads to very interesting considerations 1

2 2 NORTH AMERICAN ACTUARIAL JOURNAL, VOLUME 8, NUMBER 4 that, in turn, create a wide range of plausible outcomes by the middle of the 21 st century. The study is organized as follows. First, the paper analyzes the consequences of population aging by looking back at historical relationships between aging and health care costs. Next, it discusses a structure for uncovering the most important long-term variables that affect the future direction of health care. Identifying key variables then allows us to construct a framework for considering the future and three alternative scenarios that capture the wide range of plausible outcomes. Finally, I review current evidence to help understand the likelihood of each of these alternatives and end by discussing the implications of the analysis for those who are thinking about and planning for the future. 2. PERCEIVED PROBLEM Life expectancy increased significantly during the 20 th century, and, there is virtually unanimous agreement that this trend will continue into the 21 st century (Olshansky, Carnes, and Cassel 1993). For example, in the United States, a newborn was expected to live 47 years at the beginning of the 20 th century and over 75 years at its end; newborn life expectancy is now projected at more than 81 years by mid-21 st century. Equally dramatic increases were recorded over the 20 th century for the elderly; life expectancy at age 65 increased from 12 years to over 17 years, and at age 85 from 4 to 6 years. Similar trends in life expectancy are characteristic of national populations throughout the developed world. A consequence of increased life expectancy is large growth in elderly populations. Figure 1 summarizes projected percentage increases in population age 65 and over from 2000 to for the 1 This paper uses two sources for projections of demographics and public medical care and long-term care spending during the first half of the 21 st century: an Organisation for Economic Co-operation and Development (OECD) working paper (Dang, Antolin, and Oxley 2001; OECD Health Data 2001) and an EU Economic Policy Committee Report (EU Economic Policy Committee 2001). OECD and EU projections cover only public spending, and these projections are based on a limited range of assumptions. Public spending is used in this paper, since these are the only available projections and they provide a useful sense of the cost consequences of the analytical framework presented. Readers interested in details of the OECD and EU projections and methodologies can download the reports from the Web sites. Unless otherwise noted, figures and data in this paper come from these two sources. United States, Canada, Australia, Japan, and 15 European Union (EU) pre-expansion nations. Figure 2 presents this growth in elderly populations as a percentage of population totals. These data clearly show that developed countries populations will continue aging. We also know that health care 2 costs per capita increase rapidly with age. The EU Economic Policy Committee (2001) report presents data on age-related costs as percentages of GDP for medical care (Figure 3) and long-term care (Figure 4). Both components of health care costs show significant increases for people age 65 and older, long-term care costs being even more skewed towards older ages than medical care costs. A first estimate of the potential cost consequences of population aging can be made by simply combining demographic data with current age-related-cost data. This technique captures the consequences of population aging, assuming that the relative cost of health care by age remains unchanged. Projections in Figure 5 and Figure 6 3 use this method, and, even though only for public health care costs, they provide a baseline against which we can assess alternative cost projections. The cost effect of demographic aging on public health care spending (medical care plus longterm care) from 2000 and 2050 is a projected average increase in EU countries of 2.2% of GDP (country range 1.7% 3.0%) and 4% 5% of GDP increases in Australia, Canada, Japan, and the United States. The 2.2% of GDP projected average cost of aging in EU countries is split between 1.3% for medical care and 0.9% for long-term care. Of non-eu countries, only the United States is projected to have a large (2.1%) increase in public long-term care costs. Elderly populations in developed nations are rapidly increasing, and the cost of caring for their medical care and frailty needs is much more expensive than the cost for younger people. With no further analysis, analysts and public policymak- 2 Health care as used in this paper refers to medical care and frailty (long-term) care. 3 Luxembourg did not prepare projections. In addition, Germany, Greece, Spain, and Portugal did not prepare long-term care cost projections, and Canada did not break out costs between medical care and long-term care.

3 A FRAMEWORK FOR LONG-TERM ACTUARIAL PROJECTIONS OF HEALTH CARE COSTS 3 Figure 1 65 Population Growth, ers are left with realistic worries about how their countries might plan and pay for the health care consequences of aging. Is an aging-driven health care cost crisis the whole story, or, are there other more important factors to consider in making long-term actuarial health care cost projections? It is to this question that we turn our attention. 3. HISTORICAL RELATIONSHIP BETWEEN AGING AND HEALTH CARE COSTS Increasing numbers of elderly citizens is a demographic characteristic not limited to the first half of the 21 st century: The last half of the 20 th century was also a time of large increases in many developed countries. This fact provides us with an Figure 2 65 as % of Total Population in 2000 and 2050

4 4 NORTH AMERICAN ACTUARIAL JOURNAL, VOLUME 8, NUMBER 4 Figure 3 Medical Care Costs Per Capita (% of GDP) Figure 4 Long-Term Care Costs Per Capita (% of GDP)

5 A FRAMEWORK FOR LONG-TERM ACTUARIAL PROJECTIONS OF HEALTH CARE COSTS 5 Figure 5 Public Medical Care Spending: 2000 and 2050 historical perspective to the aging population problem. Figure 7 displays historical growth of the elderly from 1960 to 2000 for 18 countries included in our study (1960 data is not available for Portugal). In the past 40 years, growth rates have been quite large, ranging from a low of 29.4% in Ireland to a high of 287.3% in Japan. This wide range provides a good basis for testing the proposition that increases in elderly populations drive increases in health care costs. Figure 8 compares the annual percentage increase in elderly population to the annual percentage growth in real health care costs for 21 OECD countries with complete data for 1970 and A least-squares regression of the data shows little discernible statistical relationship between Figure 6 Public Long-Term Care Spending: 2000 and 2050

6 6 NORTH AMERICAN ACTUARIAL JOURNAL, VOLUME 8, NUMBER 4 Figure 7 65 Population Growth, the two independent variables (r 2.074). Other investigators reach the same conclusion: There is no statistically significant historical relationship between health care cost increases and aging (Getzen 1992). Historical trend, then, does not support the strong concern that aging populations are the key to understanding the future of health care cost increases. We need to look elsewhere to better understand the future. A study of historical U.S. medical care costs (Peden and Freeland 1995) provides additional evidence to uncover long-term health care cost drivers. Their econometric analysis of medical care spending from 1960 through 1993 showed that U.S. cost increases were composed of the demographic, demand, and supply factors shown in Table 1. Despite a 102% increase in the elderly population from 1960 to 2000 (see Figure 7), and Figure 8 Relative Growth of Health Care Costs and Elderly Population:

7 A FRAMEWORK FOR LONG-TERM ACTUARIAL PROJECTIONS OF HEALTH CARE COSTS 7 Table 1 Causes of Growth in Real U.S. Medical Care Spending Demographic Changes in age/sex mix 7.2% Health Care Demand Increases in disposable income 17.6 Broader insurance coverage 5.3 Health Care Supply Technology induced increases 69.9 typically large differences in medical care costs between those under and over age 65, the contribution of aging to the overall increase in U.S. costs is negligible. Technology and increased disposable income account for virtually all of the observed increase. Other historical data reveal a factor that helps to explain why the relationship between aging and health care costs is weaker than expected. Figure 9 presents data demonstrating a very high relative cost of health care in the last year of life. The researchers (McGrail et al. 2000) used health care data from British Columbia to separate costs by age bracket for Canadians who died during and who lived through the study years. In their study, health care costs in the last year of life ranged from 16.7 times (at age 65) to 2.5 times (at ages 90 93) the cost for same age individuals who lived. McGrail also found that medical care costs in the last year of life are relatively much higher than social nursing (frailty care) costs. Very high relative costs of dying also have been reported in other countries, including the Netherlands (van Weel and Michels 1997) and the United States (Scitovsky 1984; Lubitz and Riley 1993). Health care costs are particularly steep in the last years of life. The longer people live, the higher the age that they enter their last years of life. This means that projections, like the ones described in Section 2, which do not reshape age-related costs to reflect a delay in the high cost of dying to later ages as life expectancy increases, overstate the effect of demographic aging on health care costs. This steepening of age-relatedcost increases has been reported in the United States for all health care expenses by Cutler and Meara (1997) and in Germany for medical care costs by Becher (2004). The EU Economic Policy Committee (2001, p. 60) report provides projections prepared by three EU member states that adjust for the cost of dying. Figure 10 compares public medical care spending projections for Sweden, Italy, and the Netherlands using a straight demographic (baseline) model and a death-cost model that adjusts age-related costs for expanding life Figure 9 Health Care Cost Ratios: Dying/Surviving

8 8 NORTH AMERICAN ACTUARIAL JOURNAL, VOLUME 8, NUMBER 4 Figure 10 Death Cost Projections expectancy. Differences in cost increases over a 50-year projection period ending in 2050 are quite significant. Longer life expectancy causes projected medical care costs to drop 0.3% 0.4% of GDP in 2050 versus the baseline demographiconly projections. This is large enough to offset 25% 33% of projected baseline medical care cost increases in these three countries, which average 1.2%, resulting in a death cost -adjusted increase of 0.8% 0.9% of GDP. The EU report does not provide any information on the effect of death cost adjustments to long-term care projections. Based on data from the studies described in this section, there is strong evidence to dispel concern that aging populations are the primary causative factor, or even a major factor, driving health care cost increases. There is, at best, only a weak historical relationship between aging and health care cost. While demographic health care cost projections indicate a more significant cost of aging from 2000 to 2050, adjusting medical care cost projections for increasing average age at death as life expectancy continues to expand, by itself, significantly reduces the effect of aging on demographic-driven cost projections. So, what is going on here? What can we expect to happen to health care costs as the 21 st century unfolds? 4. HEALTH A LONG-TERM HISTORICAL VIEW Taking a look at life expectancy, health and health care over centuries begins to bring the aging problem into a more understandable perspective. In a seminal work on demographic history, Omran (1971) argues that the modern history encompasses three distinct epidemiological eras: An Age of Pestilence and Famine predated This age was characterized by stagnation of death rates at extremely high levels. Life expectancy at birth fluctuated widely between 20 and 40 years. Major killers were infectious diseases: plague influenza, pneumonia, diarrhea, smallpox and tuberculosis. The greatest toll from infectious disease was among the young and childbearing women. Beginning about the civilized world began introducing public health measures like sanitation, and living habits changed due to

9 A FRAMEWORK FOR LONG-TERM ACTUARIAL PROJECTIONS OF HEALTH CARE COSTS 9 improved public health and higher incomes. These changes brought on an Age of Receding Pandemics, mainly by bringing the most deadly infectious diseases under control. Fewer deaths occurred at early ages; more people lived to older ages. This brought a redistribution of deaths from young ages to older ages. A receding risk of death from infectious diseases correspondingly increased the risk of death from chronic disease. During this period, life expectancy increased to roughly 50 years. The Age of Receding Pandemics continued through World War I. By then, medical knowledge and medical care became effective enough to begin affecting health and life expectancy. We entered an Age of Degenerative Diseases of Affluence. During the 20 th century, medicine increasingly made death from infectious diseases an historical curiosity in developed countries and increasingly focused its benefits on managing, and sometimes curing, chronic diseases. Major causes of death became today s familiar chronic conditions of heart disease, stroke and cancer. Life expectancy in developed countries broke through 50 years at birth and rose to above 70 years by end of the century. Omran (1971) argues that the Age of Degenerative Diseases of Affluence will continue for the foreseeable future: There will continue to be improvements in mortality and morbidity as medical care continues to improve and as people increasingly choose to live healthy lifestyles. Building on Omran s observations, Olshansky and Ault (1986) argue that the effect of changes in medicine and lifestyle during the 20 th century are so significant that they signal the beginning of a fourth epidemiological era, the Age of Delayed Degenerative Diseases. This new epidemiological era is characterized by: Rapidly declining death rates and improved survival concentrated at advanced ages. Major causes of death continuing to be chronic degenerative conditions: heart disease, stoke, and cancer. Age distribution of deaths from degenerative causes continuing to shift towards older ages. To the extent that this trend continues, an Age of Delayed Degenerative Diseases may herald an era of both longer life and healthier life. The 19 th and 20 th centuries were an era of rapid economic growth and rising incomes. Studies repeatedly have shown a strong positive correlation between income and health, both across countries and within national populations. Epidemiologists and economists are involved in a lively debate over the causes of these statistical relationships and their policy implications (Marmot 2002; Deaton 2002). Nonmedical lifestyle factors of socioeconomic status, risky behavior, education, health care access, and health-related behavior are all recognized as possible explanatory variables. What is important to our discussion is the strong support shown by these studies for the importance of lifestyle as a long-term driver of improved health. Epidemiological analyses point to a number of key long-term drivers of health and life expectancy: basic public health measures (e.g., good sanitation), healthier lifestyle provided for by rising incomes, significant advances in medical technology and medical care beginning in the 20 th century, and, during the last half of the last century, widespread reduction in population risk factors (high blood pressure and high cholesterol) leading to substantial reductions in the incidence of heart disease and stroke. In developed nations, these drivers have been complimented by public support for expanding public and private health care financing programs to improve population access to health care services. Rising incomes made it possible to expand the scope of national health care systems to all citizens, including the elderly and the poor. This combination of effective public health, medical advances, rising incomes, universal access to health care, and lifestyle improvements result in the significant increases in life expectancy and improved population health documented above. 5. LONG-TERM DRIVERS OF HEALTH CARE COSTS I have demonstrated that developed nations are undergoing a rapid upward shift in the age composition of their populations. I have examined historical evidence and feel comfortable concluding that aging populations are not a highly significant driver of increasing health care costs. We also know that there have been, and almost certainly will continue to be, advances in medicine

10 10 NORTH AMERICAN ACTUARIAL JOURNAL, VOLUME 8, NUMBER 4 and health care technology, and changes in peoples social and lifestyle habits that will significantly affect future mortality, future morbidity, the range of medical and social services available to support health and frailty, and, ultimately, the future cost of health care. We can use these observations to build a framework for projecting future health care cost increases. Making long-term health care cost projections poses a very different problem than making shortterm projections, such as those that actuaries use to manage private sector health plans and health insurance pools. In the short-term, actuaries can safely make an implicit assumption that mortality, morbidity, and the scope, intensity, and cost of health care services are closely related to experience from the recent past. It is highly unlikely that any of these variables will change so drastically that actuaries must take account of a wide potential range of short-term variations in their underlying assumptions. In the short-term, the past is a strong and valid predictor of the future. Long-term health care cost projections, though, must address potentially significant changes in life expectancy, health, and health care. In the short term: Life expectancy does not change enough to merit explicit consideration in health care projections. In the long-term, increases in life expectancy significantly affect health and steepen the age-related health care cost curve. Population health (burden of disease) can be safely assumed to be fairly stable. In the longterm, the array of diseases in a population changes (e.g., AIDS and SARS), and, of key importance to our inquiry, the elderly may live their additional years of life in generally good health or in relatively poor health. Lifestyle, social norms, and income do not change much. In the long-term, significant changes in these important external factors may greatly affect the burden of disease and the scope and intensity of health care. Medical care science and technology changes slowly and in relatively predictable ways. In the long term, medicine itself may change drastically. The scope and intensity of medical care services available to treat or prevent illness may well make medicine by mid-21 st century something unrecognizable by today s standards. In addition, the cost of health care relative to other goods and services also may differ significantly from what we experience today. The use of institutional services for the frail elderly is fairly stable. In the long-term, sites and relative costs of supportive services may change in unpredictable ways. It is virtually certain that over the next 50 years, there will be changes significant enough that it is impossible to accurately predict health, health care, and health care costs, particularly by assuming that the current burden of disease and medical care technology and techniques remain unchanged. Despite our inability to accurately predict longterm outcomes, we can understand possible futures based on three key interrelated drivers: Changes in life expectancy that capture a plausible range of added years of life. Biological morbidity scenarios that cover future changes in the burden of disease and a range of possibilities that added years of life expectancy will be lived as years of relative good health or relative ill health. Economic morbidity options that address external and internal changes affecting the future scope, intensity, and relative cost of health care. We can develop a framework using these three high-level factors to develop alternative scenarios that adequately capture the range of plausible health and health care futures, and their potential cost implications. 5.1 Life Expectancy Life expectancy is clearly increasing, but by how much? What, if any limits might there be to life expectancy and life span? There are two plausible theories that capture a range of potential limits. Natural aging theories build on the assumption that there is a fixed maximum life span, and an ultimate limit to life expectancy. Delayed death theories assume that life span and life expectancy can be increased significantly. Fries (1980), in an influential paper, analyzed historical mortality data to determine upper limits to natural life expectancy and life span. His analysis was based on an implicit assumption that

11 A FRAMEWORK FOR LONG-TERM ACTUARIAL PROJECTIONS OF HEALTH CARE COSTS 11 humans have genetic limits to life and on an explicit argument that advances in medicine along with social and lifestyle changes will increasingly reduce most early deaths. As we approach the natural genetic limits to natural life, Fries demonstrates a rectangularization of the survival curve (see Figure 11). The ultimate curve is characterized by very high survival rates until people reach their natural end of life (95% of all deaths in an age cohort will take place between ages 77 and 93), with a maximum life expectancy of 85 years and a maximum life span of 115 years. In the future, having reached the ultimate survival curve, causes of death usually will be ascribed to familiar chronic diseases (heart, stroke, and cancer); however, death is more correctly attributable to a general decline of the body s ability to withstand illness. These deaths are not curable; they are simply natural aging deaths that mark the end of life. A plausible alternative scenario is that death can be significantly delayed. In delayed death theories, both life span and life expectancy will be increased through advances in genetics and geriatric medicine. Research may uncover specific aging genes or the process leading to senescence. With new scientific knowledge, it may become possible to intervene or repair genetic damage, allowing for life span and life expectancy to increase well beyond the natural limits uncovered by Fries s research. The popular and scientific presses often carry articles and papers on scientific research being done in this area (e.g., see Lane, Ingram, and Roth 2002 or Hall 2003). A popular notion is that life span can be increased to 140 years with a correspondingly large increase in life expectancy. 5.2 Biological Morbidity The Burden of Disease War, hunger, a widespread and unchecked AIDS or SARS epidemic, or other currently unknown risks may counter the current trend; however, longer life expectancy in developed countries in the 21 st century is almost a universally agreed upon certainty. We will be living additional years of life, but in what state of health? Two theories described the range of plausible outcomes: a compression of Figure 11 Percent of Persons Surviving to Each Exact Age According to Life Tables: to 1991

12 12 NORTH AMERICAN ACTUARIAL JOURNAL, VOLUME 8, NUMBER 4 morbidity (Fries 1983, 1989) or an expansion of morbidity (Brody 1985). Fries s analysis of natural limits to life expectancy also looks at changes in the underlying morbidity patterns (burden of disease). He argues that morbidity due to chronic diseases of aging will be postponed as life expectancy continues to increase and approaches its ultimate natural limit. Fries sees evidence that social and lifestyle changes and early medical interventions can both postpone and actually prevent the onset of clinical morbidity. For example, even though lung cancers will continue to develop at the end of life, smoking cessation can clearly postpone or prevent the disease from breaching a symptom threshold that requires medical intervention. Fries sees application of this virtuous dynamic to most chronic diseases, which leads to a compression of morbidity where the elderly live both longer and healthier lives. Brody (1985) argues for a plausible opposite outcome, expansion of morbidity. In his future, longer life expectancy does not postpone the clinical onset of disease. Brody studied the relationship between increases in life expectancy and health using data from the 1950s to the early 1980s. He observed over this period that there was little evidence of any compression of morbidity. From these data, he concludes that, as life expectancy expands, it is not possible to argue from experience that additional years of life generally will be lived in good health Healthy Lifestyles A New Public Health Fries has been a strong advocate for the effectiveness of risk-factor reduction to delay clinical onset of chronic disability. As I will discuss in greater detail below, there are strong and significant relationships between known controllable risk factors and chronic disease. For example, nonmedical interventions such as smoking cessation, regular exercise, avoiding excessive use of alcohol, and weight control all have been shown to reduce the chances of chronic disease. In addition, early medical interventions such as prescription drugs to control high blood pressure and high blood cholesterol levels are also effective in delaying and reducing the burden of chronic disease (World Health Report 2002). One plausible future scenario is that effective new public health initiatives encourage healthy lifestyles that significantly reduce population risk factors and, in turn, reduce the burden of chronic disease. This future path is supportive of a compression of morbidity scenario. Adoption of healthy lifestyles can be looked at as a new public health effort, one that differs from the highly effective history of public health in that it focuses on individuals responsibility for their own good health and not on governmentmanaged public works (clean water, clean air, etc.). A new public health refers to government efforts to educate people about detrimental consequences of unhealthy lifestyles and to promote increasingly healthy lifestyles. At the other extreme, it is also plausible that new public health initiatives will not be tried, or that their efforts fail either because people will not adopt necessary lifestyle changes or because government will not develop effective programs to promote change. This unchanged lifestyle scenario will neither increase nor decrease the burden of disease. It simply results in a missed opportunity to manage future population health, and it increases the chances that will we follow the path of an expansion of morbidity. 5.3 Economic Morbidity What we learn from looking carefully at life expectancy and biological morbidity is that people, particularly the elderly, will live longer in the 21 st century, but that population burden of disease may either increase or decrease. Given any particular burden of disease that may exist in the future, we need to ask another independent question: What changes might we experience in the scope and intensity of medical care and frailty care, and in the resultant costs of treating diseases? The concept of scope captures the range of diseases subject to some type of medical intervention; intensity covers the preventive, diagnostic, palliative, and curative technology available to physicians to diagnose and treat diseases. The scope and intensity of health care, and its costs, are referred to in this paper as economic morbidity. Alternative plausible futures for economic morbidity are quite divergent. One key driver is an external factor that affects the demand for health care: social expectations and societal ethics. Two other key drivers, technology and health care sys-

13 A FRAMEWORK FOR LONG-TERM ACTUARIAL PROJECTIONS OF HEALTH CARE COSTS 13 tems, affect mainly the supply side of health care. A range of plausible futures might result in an expansion of care or a compression of care. The longterm drivers of these alternatives are described next Social Norm and Medical Care Ethic The way in which we think about health and health care is a significant factor affecting economic morbidity. The fundamental importance of health and health care to each of us is captured by medical ethicists Mordecci and Sobel (1999): Health can be seen as a means, a foundation for achievement, as a first achievement itself, and a necessary premise for further achievement.... The sick individual suffers isolation, loss of wholeness, loss of certainty, loss of freedom to act, loss of the familiar world; the future is in doubt and all attention is concentrated on the present.... When ill, we no longer trust our bodies and...we no longer trust life (p. 34). This very emotional and basic drive clearly affects how people think about their health and the health care system that provides for their needs. Individuals personal sensitivities contribute to development of a health social norm that, in turn, shapes and constrains government stewardship of national health care systems. The high relative cost of the U.S. health care system has been ascribed by ethicist Callahan (1998) to a norm of individual rights to medical care: We have a system that has believed it could pursue unlimited medical progress to meet all individual needs at an affordable price.... Callahan (1998) challenges us to understand that, as long as we in the United States (and increasingly in other developed countries) demand all the care we want when we want it, government stewardship of our health care system has severe limitations to its ability to cost control. Alternatively, an ethic of social solidarity, which builds a meaningful sense of personal willingness to forego medical care for the common good, provides the U.K. National Health Insurance with social legitimacy to severely constrain health care resources and spending. This same end, constraint of services and costs, has also been achieved in authoritarian and previous communist systems that do not allow citizens any option to demand more than is made available to them by their leaders. The need to examine medical care ethics flows directly from Fries s compression of morbidity theory. Natural aging is based on an immutable 115-year life span, which allows for a maximum 85 years life expectancy. Death, from this pointof-view, is not a consequence of curable disease. Death, and the almost inevitable illness that accompanies it, is the consequence of living to a natural end of life. Earlier in this paper, I presented studies that demonstrate developed countries spend enormous amounts of money, and a very significant proportion of their health care expenditures, during the last year of life. This fact is understandable when we consider how the dying elderly, their families, and society think of medical care and death. A high cost of death is consistent with a medical ethics that usually commits physicians and their patients to do everything medically possible to stave off serious illness and death and a real hope that death can be conquered. The consequence of a conquering death ethic is pressure on the health care system to continually expand the scope and intensity of medical care, to do everything possible to treat or cure disease, even for the oldest citizens at their end of life, despite the inevitable high cost. Conquering death is not the only plausible future. There is an alternative that could develop, one recognizing aging and death as inevitable and natural (Callahan 1996). When death is near, and it will certainly come for all of us, physicians and their patients might more commonly than today recognize that end of life is approaching and choose death with dignity: An end-of-life health care aimed, not at treatment or cure, but at making inevitable death as peaceful as possible. Death with dignity does not affect the burden of chronic disease. It might, though, significantly reduce the intensity of medical care, perhaps with a corresponding, yet less costly, increase in social and supportive care for the frail elderly. Far more relatively lowcost palliative and supportive medical care might be offered to people nearing the end of their natural lives, and much more costly curative medical care increasingly be foregone. Development of a death with dignity norm and ethic would have a significant impact on the shape of health care systems and on future health care costs.

14 14 NORTH AMERICAN ACTUARIAL JOURNAL, VOLUME 8, NUMBER Health Care Technology Most observers assume that health care technology can only increase the scope and intensity of medical care. As we have seen, this cost-increasing technology dynamic has been the main cause of medical care cost increases during the last half of the 20 th century. But will technology continue to drive costs well in to the 21 st century? A complex interrelationship between medical care technology and health care financing has been called a health care quadrilemma by medical sociologist Weisbrod (1991). Weisbrod describes a circular causative relationship between technology and financing: A new medical technology is introduced. There is demand from patients and interest from providers in the new medical technology, which increases the scope, intensity, and cost of medical care. Third-party financing for medical care (private insurance and social health insurance) lowers barriers to the cost consequences of higher medical care costs and utilization. Higher costs increase the scope and demand for public or private insurance, which, in turn, provides a strong incentive for developing more new medical technology. The inevitable result of this health care quadrilemma is a system creating new, expensive technology that is demanded by patients and providers with no limit or self-correcting economic mechanism. The Peden and Freeland (1995) study cited previously demonstrates the enormous force of this dynamic. They found that from 1960 to 1993, 69.9% of the increase in U.S. medical care costs resulted from technology-induced demand. To the extent that the cost-increasing technology dynamic remains intact, the scope, intensity, and cost of health care will almost certainly continue to increase into the future. Weisbrod s (1991) argument rests, though, on the assumption that new technology is expensive; it only increases costs by expanding the scope and intensity of medical care. In the long-term this dynamic may not continue. As I will argue in much more detail below, recent advances in scientific knowledge make it plausible that future changes in medical technology may actually reduce the intensity and cost of medical care. This cost-reducing technology scenario, if it were to occur, clearly would have a major impact on the nature of health care and health care costs Frailty Care Aging often brings with it moderate or severe frailty. Whatever the burden of chronic conditions, elderly frailty needs are met through a combination of home care by informal providers (friends and family), formal home care, and institutionalized care. The cost of care differs significantly across these three categories, with formal home care and institutional being the most expensive options. A great deal of attention is being given to the continuum of care: providing frailty care in the most personally supportive and costeffective settings (Jacobzone 2000). One future scenario, de-institutionalized care, would find an increased proportion of the elderly s frailty care needs met with informal or formal home care, avoiding costly institutionalization. Another, much more expensive scenario would be an increasing proportion of institutionalized care Health Care Systems Health care systems are the means by which citizens health care needs and expectations are fulfilled. Health care systems are by nature of the problem they address, and by design, enormously complicated organizations in need of public stewardship. As such, health care systems are deeply involved with national political processes. In the short-term, health care systems, their cost, effectiveness, and continuous refinements are so important to us that they demand and receive a great deal of attention. In the long term, health care systems are also important, but less so than in the short term. Figure 12 compares healthy life expectancy (HALE) to the cost of health care (per capita total health care spending measured in U.S. dollars adjusted for purchasing power parity) in 2001 for each of the 19 countries included in this study. All of these countries have exceptional population health outcomes, but at widely varying costs. There is a large cluster of countries with HALEs between 69 and 72 that spent between $1,500 and $3,000 per capita in The three outliers are the United States (HALE 67.6, health spending $4,887), Japan (HALE 73.6, health spending $2,131), and Portugal (HALE 66.8, health spending $1,618).

15 A FRAMEWORK FOR LONG-TERM ACTUARIAL PROJECTIONS OF HEALTH CARE COSTS 15 Figure 12 Health vs. Health Care Spending, 2001 It is clear from these data that greater health spending does not lead to better population health. In fact, the correlation among these 19 pairs of data is actually very low (r ) and the slope of a linear trend line is actually negative. The United States spends much more on health care than other nations, yet has one of the worst population health outcomes, while Japan has the world s best population health outcome with relatively low spending. It is also clear that developed nations do not deny their citizens access to effective health care interventions and technology. No matter the amount spent, each of the 19 countries populations has a HALE that could only result from widespread access and appropriate use of the best available medical care; effective medical technology must be available when needed to produce these excellent population health outcomes. It appears from these data that something other than the structure of developed nations health care systems accounts for differences in health care costs and health. Exploring this very interesting problem, though, is beyond the scope of this study. In the long-term, if health care system structure does affect health care costs, the dominant influence may well be the existence of and effectiveness of constraints: Private systems may constrain costs through a variety of supply-side and demand-side tools. Supply-side tools include negotiations among institutions that provide financing and those that provide care through managed care arrangements that attempt to control provider costs and the use of expensive medical care. Demand-side constraints include controlling moral hazard through provider copayments and other financial and contractual limitations included in most medical expense in insurance contracts. Public systems have additional, more effective tools at their disposal. Governments can constrain costs (e.g., provider salaries, cost of disposable materials, and/or availability of personnel, technology, and capital) through a global budgeting process or direct control of the supply of health care. Weak constraints may simply reduce use of services that are not necessary for good medical care, for example, private rooms in hospital versus wards or the convenience of accessing widely available diagnostic equipment. Stronger constraints may limit access, for example, by creating waiting lists for needed medical care or delaying the introduction of new technologies. The strongest constraints may simply make some types of expensive medical care unavailable to some patients to whom it might provide benefit.

16 16 NORTH AMERICAN ACTUARIAL JOURNAL, VOLUME 8, NUMBER 4 Table 2 Summary of Long-Term Drivers Cost Decreasing Cost Increasing Life Expectancy Natural Aging Delayed Death Biological Morbidity Compression of Morbidity Expansion of Morbidity Lifestyle Healthy Lifestyle Unchanged Lifestyle Economic Morbidity Compression of Care Expansion of Care Social Norm Social Solidarity Individual Rights Medical Care Ethic Death with Dignity Conquering Death Medical Technology Cost-Reducing Cost-Increasing Frailty Care De-Institutionalized Care Institutionalized Care Health Care Systems Increasingly Constrained Increasingly Unconstrained Weak constraints may be only marginally effective in reducing health care costs; strong constraints may effectively impact the growth of health care services and its costs over the short term and intermediate term. However, in developed countries, even the strongest constraints can only delay the consequences of Weisbrod s health care quadrillema; they cannot deny effective care nor cut costs too much without creating strong political counterpressures. The most important national health and health care policies will be those that directly or indirectly affect the burden of disease ( new public health), the scope and intensity of medical care interventions that are available to treat diseases (social norms and ethics and medical technology), and universal access to high-quality health care services. Stewardship of health care systems, then, is likely to have more effect on the pace of change than on its long-term direction. This long-term view of health care systems leaves us with two plausible futures. A system may become increasingly constrained as governments attempt to control costs, or the system may become increasingly unconstrained as citizens demand more access to health care resources Summary of Key Drivers The key drivers of the long-term future of health, health care, and health care costs are summarized in Table 2. The 21 st century may bring with it changes in life expectancy, biological morbidity, and economic morbidity that lead to less health care and lower costs or to more health care and ever higher costs. These key drivers are used in the next section of this paper to create alternative plausible futures. 6. ALTERNATIVE FUTURE SCENARIOS Life expectancy, biological morbidity, and economic morbidity all affect the long-term course of health, health care, and health care costs. Longterm health care cost projections can be built on alternative futures for these key drivers. Figure 13 provides an overview of how they interact 4. Life expectancy will almost certainly continue to improve; only its pace and ultimate limit is uncertain. Both natural aging and delayed death scenarios result in added years of life expectancy. Delayed death, though, will expand life expectancy further and, at some point, faster than natural aging. As life expectancy improves, biological morbidity may expand, stay the same or compress: Expansion of morbidity depicts longer life, but with average age of onset of chronic diseases rising slower than any increase in life expectancy. The elderly live longer, but their added years often will be unhealthy ones, adding significantly to the burden of disease. This adverse outcome is due, in part, to unchanged lifestyles that contribute to the lack of improvement in health. Compression of morbidity is the best plausible outcome. Longer life is accompanied by the average age of onset of disease increasing even faster than life expectancy. The elderly live longer and have fewer years of disease than at present, reducing the burden of disease. Healthy lifestyle and development of effective 4 The structure of Figure 13 is derived, in part, from schemas used in two research papers (Howe 1999 and Barer, Evans, Hertzman, and Lomas 1987).

17 A FRAMEWORK FOR LONG-TERM ACTUARIAL PROJECTIONS OF HEALTH CARE COSTS 17 Figure 13 Changes in Life Expectancy, Biological Morbidity and Economic Morbidity early medical interventions contribute significantly to this positive outcome. Equilibrium of morbidity depicts longer life with average age of disease onset and life expectancy increasing by roughly the same number of years. The elderly live longer and the average number of end-of-life years with chronic disease is roughly the same as today, leaving the burden of disease relatively unchanged. Age of onset of chronic disease is delayed by healthy lifestyles aided by early medical interventions that delay the onset of clinical symptoms. Following the path of equilibrium of morbidity within biological morbidity in Figure 13, there are three plausible futures for economic morbidity. The equilibrium of care scenario depicts a course and treatment of disease that is generally the same as it is today. Social norms and medical care ethic remain unchanged, continuing a conquering death norm and individual rights ethic without significant changes in the scope or intensity of medical care. Medical technology continues to be cost-increasing. The needs of the frail elderly are met as they are today; the future brings no material changes in institutionalized care or de-institutionalized care. Health care systems are neither increasingly constrained nor increasingly unconstrained. Expansion of care depicts a future in which the time from biological onset of disease to onset of medical care is reduced. This happens due to changes in medical care technology that provide means to identify and treat illness closer to its onset than at present. In this scenario, new technology continues to be cost-increasing. Care for the frail elderly is increasingly institutionalized care, which also adds to future cost increases. Social norms and medical care ethic remain unchanged, continuing a conquering death norm and individual rights ethic that fuel the expansion of scope and intensity of medical care. Health care systems are neither increasingly constrained nor increasingly unconstrained. The compression of care scenario recognizes there may be changes in medical technology and/or in the medicalization of disease that will actually reduce the scope and intensity of medical care, so medical technology actually becomes cost-reducing. Care for the frail elderly is increasingly de-institutionalized care which

18 18 NORTH AMERICAN ACTUARIAL JOURNAL, VOLUME 8, NUMBER 4 helps to control future cost increases. Death with dignity becomes a widespread medical care ethic and the social norm of individual rights is understood as consistent with this new ethic. Health care systems are neither increasingly constrained nor increasingly unconstrained. Combining different outcomes for life expectancy, biological morbidity, and economic morbidity, we can create a large number of plausible future scenarios. Three scenarios that encompass a reasonable range of health, health care, and cost outcomes are described below. 6.1 Scenario I Continuing Today s Health Care Environment into the Future This first scenario describes a future not much different from today. Life expectancy follows the path of natural aging and continues increasing toward an ultimate 85 years expected life at birth. There is only a modest improvement in health caused by wider adherence to healthy lifestyles and discovery and use of effective preclinicalmorbidity interventions. Biological morbidity continues to improve so that the elderly both live longer and live about the same number of years in ill health from chronic diseases as they do today resulting in equilibrium of morbidity. Economic morbidity continues on its current path of expansion of care. Medical technology continues to increase the scope and intensity of health care interventions as it has in the past few decades. End-of-life debility and illness is increasingly recognized as it occurs, but patients and physicians continue to pursue aggressive medical care intervention. The frail elderly increasingly make use of expensive institutionalized care. Using OECD health data and projections prepared for the EU Economic Policy Committee (2001) report, we can make estimates of the cost consequences of this and our other scenarios for public spending in the 15 EU nations covered by this report. EU public health care spending in its 15 nations averaged 6.6% of GDP in Over a 25-year period from 1973 through 1998, these nations experienced health care cost increases equal to their GDP plus 1.4% per year (OECD Health Data 2001). The excess 1.4% growth incorporates the cost consequences of historical rates of population aging, changes in biological morbidity, and changes in economic morbidity including, most notably, increases in scope and intensity of medical care resulting from the introduction of new technologies. If this historical trend continues into the future, then public health care costs will grow from 6.6% of GDP in 2000 to 13.2% of GDP in Using EU Policy Committee alternative projections, a reasonable range of components of this increase can be estimated (see Table 3) Scenario II Adverse Future Health Care Environment A second plausible scenario can be constructed that creates very serious concern for developed countries ability to continue funding their public health care programs. Delayed death is realized; science discovers effective means to extend life span beyond its natural limits causing life expectancy to increase. Biological morbidity follows an expansion of morbidity. Longer life is not accompanied by better health so that added years are lived in general ill-health. People continue to pursue unchanged lifestyles. Economic morbidity continues to contribute to the problem through expansion of care. Medical technology expands the scope and intensity of costly, aggressive medical care interventions at an even more rapid pace than in the past; end of life is not recognized and disease at any age is treated aggressively with an aim to cure 5 The EU Economic Policy Committee (2001) report contains basic projections of health care costs calculated solely on changes in the demographic composition of 15 EU member nations. There are also alternative projections demonstrating cost sensitivity to various assumptions, including income elasticity of demand, cost of end-of-life medical care, improvement in health of the elderly, and different demographic assumptions. I use these various projections to estimate the cost consequences of the three scenarios presented in this paper. Cost estimates in this paper have not been modeled. They are based heavily on judgmental adjustments to the EU projections and, therefore, are only general representations of the cost consequences of the three scenarios.

19 A FRAMEWORK FOR LONG-TERM ACTUARIAL PROJECTIONS OF HEALTH CARE COSTS 19 death ; institutionalized care of the frail elderly becomes ever more common. Future health care costs in this scenario clearly exceed those in Scenario I. Longer life expectancy increases the average age of populations beyond those implicit in Scenario I, an expansion of morbidity adds to burden of disease, and greater increases in scope and intensity of care add even more to health care costs. A reasonable range of cost estimates (in relation to Scenario I) is shown in Table Scenario III Favorable Future Health Care Environment A plausible and relatively low cost long-term health care future might plausibly evolve as follows: Table 3 Potential Cost of Continuing Today s Health Care Environment into the Future %ofgdp EU Public Health Care Costs % Incremental Costs due to: Aging Populations and Increasing Life Expectancy Changes in Biological Morbidity (0.5) (0.5) Changes in Economic Morbidity EU Public Health Care Costs Life expectancy follows the path of natural aging and continues increasing toward an ultimate 85 years expected life at birth. People increasingly adopt healthy lifestyles. Interest in a healthy lifestyle is accompanied by low-cost, preclinical health care interventions that effectively postpone the onset of disease or delay the onset of clinical care. Added years of life are lived in relatively good health so that biological morbidity experiences compression of morbidity. These positive trends are complemented by significant improvements in economic morbidity. Health care systems experience an actual compression of care. Breakthroughs in scientific knowledge of the human body and chronic disease allow development of medical technologies that create low-cost preclinical interventions and curative treatments. Ethics and sensitivities change so that inevitable end-of-life debility and illness are clearly recognized and treated with palliative care and support, but not with aggressive medical interventions. The frail elderly are supported in their homes and lowercost institutional settings (e.g., day care). This scenario results in costs much lower that than in Scenario I due to favorable to very favorable changes in biological morbidity and a major change in the trend of economic morbidity caused by technology developing the means to reduce the intensity of medical care while improving health care outcomes (see Table 5). 6.4 Range of Plausible Future Scenarios The range of plausible futures for health, health care, and health care costs is quite wide. Costly outcomes include longer and relatively unhealthy lives, new and more expensive medical technology, and demands for aggressive pursuit of good health and cures for death. Less costly or even Table 4 Potential Cost of an Adverse Health Care Cost Environment Table 5 Potential Cost of a Favorable Future Health Care Cost Environment %ofgdp EU Public Health Care Costs % Incremental Costs due to: Aging Populations and Increasing Life Expectancy Adverse Changes in Biological Morbidity Changes in Economic Morbidity EU Public Health Care Costs %ofgdp EU Public Health Care Costs % Incremental Costs due to: Aging Populations and Increasing Life Expectancy Favorable Changes in Biological Morbidity (2.0) (0.5) Changes in Economic Morbidity (1.0) 3.5 EU Public Health Care Costs

20 20 NORTH AMERICAN ACTUARIAL JOURNAL, VOLUME 8, NUMBER 4 cost-reducing outcomes include longer and relatively healthy lives driven in large part by healthy lifestyles, favorable changes in medical technology making less intense medical care possible, development of a less aggressive medical care ethic particularly at the end of life, and de-institutionalization of long-term care for the frail elderly. I have incorporated combinations of these cost-increasing and cost-decreasing drivers into three plausible future scenarios that demonstrate the extremely wide range of resultant health care costs in The three scenarios are summarized in Table 6. Compare this to EU public health care cost in 2000 of 6.6% of GDP. Results are stated as percentages of GDP (i.e., health care costs divided by GDP) in current (2000) value (no general inflation). The three scenarios focus solely on the numerator of this ratio. By approaching the problem in this way, we do not explicitly take account of changes in the growth of non-health care sectors of the economy. To the extent that nonhealth care sectors of developing countries economies grow faster or slower than EU projections, results would change, perhaps dramatically. Exploring possible futures for GDP growth is beyond the scope of this paper. 7. EVIDENCE FOR A MORE PLAUSIBLE FUTURE The wide scope of plausible long-term scenarios and their public cost implications is almost overwhelming, ranging from a 50-year decrease of 1.4% of GDP to an increase of 11.2% of GDP. To help narrow this range, let us turn to evidence that might suggest a more likely path for the future. Table 6 Summary of Costs of Plausible Future Scenarios 2050 Average EU Public Cost Scenario (% of GDP) I. Continuing Today s Health Care Environment into the Future % II. Adverse Future Health Care Environment III. Favorable Future Health Care Environment Life Expectancy Projecting life expectancy is a familiar exercise to many actuaries. The major unresolved question for our investigation is whether or not science will discover effective means of extending natural genetic limits to life. If science accomplishes this goal, we can look forward to even larger populations and higher percentages of elderly than are currently incorporated into various long-term health care cost projections. Science has long been interested in understanding the reasons for aging and death. Readers interested in an overview of the current science of aging can refer to numerous books and articles; for example, see Olshansky, Carnes, and Cassel (1993); Held (2002); Lane, Ingram, and Roth (2002); and Hall (2003). They discuss how evolutionary and molecular biologists work on understanding the genetic reasons for aging and the mechanisms by which the process proceeds. Evolutionary biologists have developed a variety of theories that ascribe a genetic limit to life. Early researchers such as Williams and Hamilton proposed antagonistic pleiotropy theories that argue that reproduction is the genetic reason for existence, so genes having damaging effects later in life will not be as effectively eliminated by natural selection. Kirkwood later proposed a variation on these theories. His disposable soma theory, argues that humans divide their lifetime energy between sexual reproduction and maintenance of the body (soma). Energy is needed to repair genetic damage. Senescence is the genetic result of an energy trade-off favoring sexual reproduction at the expense of long-term perfect genetic repair. Recently, Lee (2003) has proposed another variation on antagonistic pleiotropy theories, arguing that children s need for parental care provides an evolutionary basis for long-term, postreproductive age survival. A genetic limit to life implies an upper limit to life span and life expectancy. Absent discovery of human means to change the natural genetic senescence of our bodies, medicine cannot cure death. Biologists and physicians focus their attention on the genetic aging mechanism. This line of inquiry has resulted in dysdifferentiative hypotheses of aging. These theories argue that, over time, there are gradual accumulations of random mo-