print logo

Forecasting the Cost of U.S. Healthcare

Thursday, September 3, 2009

There is no need to suppress the demand for healthcare. Expenditures on healthcare are driven by demand, which is spurred by income and by advances in biotechnology that make health interventions increasingly effective.

In attempting to forecast the cost of healthcare a generation in the future, several different issues must be considered when constructing estimates. The first of these is the likely downward trend in age-specific prevalence rates of chronic diseases and disabilities. Secondly, there is the rate of change in the cost of treating these conditions: will advances in biotechnology reduce or increase the cost of treatment? A third issue is the likely increase in the number and proportion of the population that is elderly, and a fourth issue is the rate at which the U.S. population will increase and the sources of that increase. The fifth issue is the rate of growth of per capita income and the impact of economic growth on the demand for the quantity and quality of healthcare.

Each of these issues is so large and complex that it would take a book to address them properly. My aim in this article is merely to outline the issues and to sketch some tentative answers.

Trends in Prevalence Rates and the Cost of Healthcare

One of the most important debates among health economists and biomedical specialists in the United States and other wealthy, Organisation for Economic Co-operation and Development (OECD) nations is whether rapid advances in biotechnology will spare their health systems from a financial crisis. This debate turns on four propositions.

The fact is that demographers’ past predictions of maximum life expectancy have been notoriously conservative when these forecasts were based on average experience.

First, there is now convincing evidence that prevalence rates of chronic diseases declined during the 20th century. Second, the rate of decline in these prevalence rates has accelerated. In the American case, prevalence rates declined at a rate of about 1.0 percent per annum between 1910 and 1980. Between the early 1980s and 1989, they declined at about 1.2 percent per annum. During the 1990s, the rate of decline further accelerated, reaching a level of about 2.0 percent per annum. Some investigators believe that a rate of decline in annual prevalence rates of 1.5 percent will be enough to offset the rising cost of healthcare, thus stabilizing the share of healthcare costs in GDP at its current level of about 15 percent.

Third, there is an unresolved issue regarding how much of the decline in the prevalence rates of OECD nations has been due to improvements in the environment and how much has been due to biomedical interventions. Partitioning the decline in prevalence rates into environmental effects and medical intervention effects is quite complex because of the long reach of nutritional and other biomedical insults at earlier ages on the odds of developing chronic diseases at middle and late ages. Although such life-cycle effects have long been suspected in particular diseases, only recently has a substantial body of evidence bearing on the interconnections been amassed. Longitudinal studies connecting chronic diseases at maturity, middle ages, and late ages to conditions in utero, infancy, and early childhood were reported with increasing frequency beginning in the 1980s and extending through the end of the 20th century. The exact mechanisms by which malnutrition and trauma at early ages affect waiting time to the onset of chronic diseases are still unclear, but it seems reasonable to infer that environmental insults during the period when cell growth is rapid could lead to long-lasting impairments of vital organs.

It is important to emphasize that medical interventions have not only contributed to the decline in prevalence rates of chronic conditions but also to the reduction in their severity. Advances in both surgical and drug therapies have significantly reduced the rate at which chronic conditions turn into disabilities that severely impair functioning. Such interventions have been especially effective in genitourinary, circulatory, digestive, and musculoskeletal conditions. However, many of the surgical procedures are quite expensive, and the cost of the new and more effective drugs is increasing sharply, mainly because of the large investments in developing these drugs.

The accelerating decline in the prevalence of chronic diseases during the course of the 20th century supports the proposition that increases in life expectancy during the 21st century will be fairly large.

Fourth, there is a debate over whether the mounting evidence of the long-term decline in the prevalence rate of chronic diseases means that the supply of treatable chronic diseases is declining. (The word “supply” here distinguishes the physiological burden of healthcare from the demand for healthcare services that may rise even if the physiological burden remains constant or declines.) Moreover, to address the question of whether declines in age-specific physiological prevalence rates will relieve current fiscal pressure on the healthcare systems of OECD nations, it is necessary to weight the existence of a particular chronic disease by the cost of treating that condition, which generally increases with age.

Figure 1

Such an index is shown in Figure 1. In this figure, the burden of per capita healthcare costs, which is based on U.S. data, is standardized at 100 for ages 50–54. Figure 1 shows that the financial burden of healthcare per capita rises slowly in the 50s, accelerates in the 60s, accelerates again in the 70s, and accelerates even more rapidly after the mid-80s. The financial per capita burden at age 85 and older is nearly six times as high as the burden at ages 50–54. Notice that the financial burden of healthcare for ages 85 and older is over 75 percent higher per capita than at ages 75–79. However, the physiological prevalence rates (number of conditions per person) is roughly constant at ages 80 and over.

Costs rise, even though the number of conditions (comorbidities) per person remains constant, because the severity of the conditions increases or because the cost of preventing further deterioration (or even partially reversing deterioration) increases with age. It should be kept in mind that standard prevalence rates merely count the number of conditions, neglecting both the increasing physiological deterioration with age and the rising cost of treatment per condition. Figure 1 indicates that to forecast the future financial burden of healthcare, it is necessary to make use of a function of the age-specific cost of healthcare, such as that shown in Figure 1.

Figure 2

What, then, can be said about the likely movements in the curve of the relative burden of healthcare costs over the next generation? Figure 2 lays out three possibilities. The first possibility is that there will be a proportional downward shift in the curve (Case A). This is the curve obtained by using the change in the average prevalence rates, which implies a shift downward at a constant average rate at all ages. The example shown in Figure 2 implies a decline in average prevalence rates of 1.2 percent per annum, which locates all of the points in Case A at about two-thirds of the previous level. If 1.5 percent had been used, which is the high end of current forecasts of the decline in prevalence rates, the points on the Case A curve would all be located at about 60 percent of the original level.

The outlook for new and more effective technologies to deal with chronic disabilities through the marriage of biology and microchip technology is very promising.

A second alternative, shown as Case B in Figure 2, is that the curve of disease burden by age will shift to the right. The Case B curve was constructed on the assumption that over the course of a generation, the average age of onset of chronic conditions is delayed by about five years. This assumption is supported by a number of epidemiological studies in the Netherlands, Britain, the United States, and elsewhere. This forecast is based partly on the evidence that the average age of the onset of chronic disabilities has been declining since the start of the 20th century. It is also based on studies of the relative cost of healthcare by years before death. These studies have produced the curve shown in Figure 3, which is standardized on the average costs of healthcare for all persons age 65 and over in the U.S. Medicare program. Figure 3 shows that, five years before the year of death, annual health cost is virtually the same as all annual Medicare costs per capita. By the second year before death the cost has risen by about 60 percent, and in the year of death the annual cost exceeds the average by more than four times. Indeed, expenditure on persons during their last two years of life account for 40 percent of all Medicare expenditures.

Figure 3

The pattern portrayed in Figure 3 has not changed significantly over the past two decades. The relative constancy in healthcare costs by years before death supports Case B in Figure 2 because it implies that no matter how far to the right the healthcare curve shifts, age-specific costs will eventually rise sharply as the proportion of persons who die in any given age category increases. This line of reasoning implies that healthcare costs may continue to increase even if the age of onset of chronic diseases is delayed, because the proportion of a cohort living to late ages will increase. Moreover, the cost of keeping disabilities under control may rise because more effective drugs and procedures may be more expensive than the current set.

Trends in Life Expectancy

Will the 21st century witness as large an increase in the average life expectancy of the rich countries—30 to 40 years—as occurred during the 20th century? Most experts believe it will not. The middle estimate of the U.S. Census Bureau, for example, is that the increase in life expectancy between 2000 and 2050 will be only about seven years, and the estimated increase for the entire 21st century is just 13 years. This is less than half the increase that occurred during the 20th century. The same conservatism is evident in the projections of the United Nations, OECD, and other national and international agencies.1

These pessimistic projections rest on several propositions. Perhaps the most widely accepted is the proposition that opportunities for large reductions in mortality rates are possible only when death rates under age 5 are very high. Proponents of this view argue, for example, that the sharp decline in U.S. mortality rates during the 20th century was the result of a unique opportunity that cannot be replicated by those nations that have already experienced it: the opportunity to wipe out the majority of deaths due to acute infectious diseases, which were concentrated in infancy and early childhood. Whereas more than a third of all deaths at the turn of the 20th century were of children under 5, today infant and childhood deaths are less than 2 percent of the annual total. By contrast, deaths among persons age 65 and over, which accounted for just 18 percent of the total in 1900, have grown to three-quarters of all deaths today.2 Thus, at the start of the 21st century, the argument goes, the more than 90 percent of birth cohorts who live to age 50 begin to suffer from an increasing number of chronic diseases because their vital organ systems naturally lose their effectiveness with aging, and this deterioration eventually increases to a point where life can no longer be sustained. Empirical observations are buttressed by a variety of theories, some of them drawn from evolutionary biology, as to why the cells of vital organ systems decay. One prominent theory holds that because reproduction ceases at age 50, there is a sharp rise in deaths at post-reproductive ages because the forces of natural selection have not eliminated the genes that hasten rapid physiological decline past age 50.

Healthcare costs may continue to increase even if the age of onset of chronic diseases is delayed, because the proportion of a cohort living to late ages will increase.

There are, however, persuasive arguments that spell out a more optimistic view of the course of changes in health and longevity during the 21st century. One of these arguments is based on the projection not of past changes in average life expectancy but of record life expectancy since 1840. Record life expectancy is defined as the highest life expectancy experienced by any country at each point in time. For example, the record life expectancy at birth in 1840 was found among Swedish women, who lived on average a little more than 45 years. In the year 2000, Japanese women achieved a record life expectancy of nearly 85 years. Fitting a curve to such best practice observations over a period of 160 years yields a linear curve, which suggests that for the foreseeable future, female life expectancy will increase at 2.4 years per decade and male life expectancy will increase at 2.2 years per decade. These equations lead to the prediction that by 2070 female life expectancy in the United States will be between 92.5 and 101.5 years, which substantially exceeds the forecast of 83.9 years made by the Social Security Administration in 1999.3

The fact is that demographers’ past predictions of maximum life expectancy have been notoriously conservative when these forecasts were based on average experience. In the late 1920s, L. I. Dublin, the chief actuary of the Metropolitan Life Insurance Company, put a cap of 64.75 years on life expectancy for both men and women. In 1936 he collaborated with the leading mathematical demographer of the first half of the 20th century to publish a revised upper limit of 69.93 years.4  More recently, a leading gerontologist set an upper limit on life (excluding some major breakthrough in molecular biology) of 85 plus or minus 7 years.5  Generally speaking, these caps tend to be in the range of five to ten years beyond the observed life expectancy at the time the forecasts were published.6

The accelerating decline in the prevalence of chronic diseases during the course of the 20th century supports the proposition that increases in life expectancy during the 21st century will be fairly large. At the beginning of the 20th century the burden of chronic diseases among elderly Americans was not only of greater severity but began more than ten years earlier in the life cycle than it does today. Moreover, the number of comorbidities at each age between 50 and 70 is well below levels that prevailed a century ago. This is, according to one study, equivalent to pushing back old age, since an increase of one unit in a comorbidity index is the equivalent of being a decade older. Studies of changes in functional limitations among persons who have reached age 65 since the early 1980s indicate that such limitations declined at an accelerating rate during the balance of the 1980s and the 1990s.7 

Medical interventions have not only contributed to the decline in prevalence rates of chronic conditions but also to the reduction in their severity.

Dora Costa has found that favorable changes in body size, particularly the decline in the waist-to-hip ratio (a measure of abdominal fat), explained close to half of the decline in mortality rates above age 65 during the course of the 20th century.8  Taking account of the characteristics of men of military age in 1988, she predicts that the annual decline in male mortality rates after age 65 will be nearly twice as high between 1988 and 2022 as it was between 1914 and 1988. Overall, the work on trends in chronic diseases and on frame sizes tends to support forecasts of continued linear trends in the extension of longevity during the 21st century.

One factor arguing in support of the optimists' projections is the increasing span of years that individuals have free of chronic conditions. For those who reached age 65 during the first decade of the 20th century, the average age of onset of chronic disabilities was about 51. By the 1990s, however, the average age of onset of chronic conditions was more than ten years later. Moreover, these disabilities are now generally milder, and many effective interventions to reduce the impact of chronic conditions are available.9  The outlook for new and more effective technologies to deal with chronic disabilities through the marriage of biology and microchip technology is very promising. Indeed, some devices that combine living cells and electronics to replace failed organs are already at the stage of human trials.10 Somewhat further off, but even more promising, are advances in genetic engineering that will produce cures for what are now untreatable diseases.11

Explaining the Paradox of Rising Expenditures on Health

Why is it that although the average age of onset of disabilities has been delayed by ten years, and that these disabilities have become milder than they used to be, the share of GDP spent on health is rising? One factor is the increase in the proportion of the population that is elderly. However, such changes in age structure account for a minor part of rising expenditures, on the order of 10 percent.

The main factor is that the long-term income elasticity of the demand for healthcare is 1.6—for every 1 percent increase in a family’s income, the family wants to increase its expenditures on healthcare by 1.6 percent. This is not a new trend. Between 1875 and 1995, the share of family income spent on food, clothing, and shelter declined from 87 percent to just 30 percent, despite the fact that we eat more food, own more clothes, and have better and larger homes today than we had in 1875. All of this has been made possible by the growth in the productivity of traditional commodities. In the last quarter of the 19th century, it took 1,700 hours of labor to purchase the annual food supply for a family. Today it requires just 260 hours, and it is likely that by 2040, a family’s food supply will be purchased with about 160 hours of labor.12

Consequently, there is no need to suppress the demand for healthcare. Expenditures on healthcare are driven by demand, which is spurred by income and by advances in biotechnology that make health interventions increasingly effective. Just as electricity and manufacturing were the industries that stimulated the growth of the rest of the economy at the beginning of the 20th century, healthcare is the growth industry of the 21st century. It is a leading sector, which means that expenditures on healthcare will pull forward a wide array of other industries including manufacturing, education, financial services, communications, and construction.

Robert Fogel is the Charles R. Walgreen Distinguished Service Professor of American Institutions at the University of Chicago Booth School of Business. He won the Nobel Prize in Economics in 1993.


Image by Darren Wamboldt/Bergman Group.





1. US Census Bureau 2000b.
2. Change in the age structure accounts for some of the change in the distribution of deaths. According to this proposition, countries in which this decline has already occurred are unlikely to be able to produce another such reduction in mortality rates. U.S. National Center for Health Statistics 1997; Preston, Keyfitz, and Schoen 1972; Preston 1985; Linder and Grove 1947.
3. Oeppen and Vaupel 2002.
4. Dublin 1928; Dublin and Lotka 1936.
5. Fries 1980, 1990.
6. See Oeppen and Vaupel 2002 suppl.
7. Helmchen 2003; Charlson et al. 1994; Stuck et al. 1999; Manton and Gu 2001.
8. Costa 2004.
9. Helmchen 2003.
10. Arnst 2003.
11. Economist 2003.
12. Fogel 2008.


Arnst C. 2003. Off -the-shelf body parts. Business Week (18–25 Aug.): 106–107.

Charlson M, Szatrowski TP, Peterson J, Gold J. 1994. Validation of a combined comorbidity index. Journal of Clinical Epidemiology 47: 1245–1251.

Costa D. 2004. The measure of man and older age mortality: Evidence from the Gould Sample. Journal of Economic History 64: 1–23.

Dublin LI. 1928. Health and Wealth: A Survey of the Economics of World Health. New York and London: Harper and Brothers.

Dublin LI, Lotka AJ. 1936. Length of Life: A Study of the Life Table. New York: Ronald Press.

Economist. 2003. A voyage of discovery: Biotechnology may yet renew the pharmaceutical industry (in survey section: Climbing the helical staircase: A survey of biotechnology) 36: 7–9.

Fogel RW. 2008. Forecasting the cost of U.S. healthcare in 2040. NBER Working Paper 14361.

Fries JF. 1980. Ageing, natural death, and the compression of morbidity. New England Journal of Medicine 303: 130–136.

Fries JF. 1990. The sunny side of aging. Journal of the American Medical Association 263: 2354–2355.

Helmchen L. 2003. Changes in the age at onset of chronic disease among elderly Americans, 1870–2000. Typescript: Center for Population Economics, University of Chicago.

Linder FE, Grove RD. 1947. Vital Statistics Rates in the United States 1900–1940. U.S. Government Printing Office: Washington, D.C.

Manton KG, Gu X. 2001. Changes in the prevalence of chronic disability in the United States black and nonblack population above age 65 from 1982 to 1999. Proceedings of the National Academy of Sciences, USA 98: 6354–6359.

Oeppen J, Vaupel J. 2002. Broken limits to life expectancy. Science 296: 1029–1031.

Oeppen J, Vaupel J. 2002 suppl. Broken limits to life expectancy, supplementary material. Available on the Internet at (last accessed 6 August 2008).

Preston SH. 1985. Resources, knowledge, and child mortality: A comparison of the U.S. in the late nineteenth century and developing countries today. In International Property Conference, Florence, 5–12 June, vol. 2, 373–386. Li├Ęge, Belgium: International Union for the Scientific Study of Population.

Preston SH, Keyfitz N, Schoen R. 1972. Causes of Death: Life Tables for National Populations. New York: Seminar Press.

Stuck AE, Walthert JM, Nikolaus T, Bula CJ, Hohmann C, Beck JC. 1999. Risk factors for functional status decline in community-living elderly people: A systematic literature review. Social Science and Medicine 48: 445–469.

U.S. Census Bureau. 2000. Table C: Projected life expectancy at birth by race and Hispanic origin, 1999 to 2100. Available on the Internet at (last accessed 6 August 2008).

U.S. National Center for Health Statistics. 1997. Monthly Vital Statistics Report 46(1) suppl. (11 September).





Most Viewed Articles

3-D Printing: Challenges and Opportunities By Michael M. Rosen 10/19/2014
With physical copying now approaching digital copying in terms of ease, cost, and convenience, how ...
Why Privilege Nonprofits? By Arnold Kling 10/17/2014
People on the right view nonprofits as a civil-society bulwark against big government. People on ...
Chinese Check: Forging New Identities in Hong Kong and Taiwan By Michael Mazza 10/14/2014
In both Hong Kong and Taiwan, residents are identifying less and less as Chinese, a trend that ...
The Origins and Traditions of Columbus Day By Amy Kass and Leon Kass 10/10/2014
Columbus Day is a most unusual American holiday and has become a day 'to celebrate not only an ...
How Green Is Europe? By Vaclav Smil 09/30/2014
A superficial look might indicate great achievements. Yet a closer view reveals how far European ...