Life expectancy and health care expenditures: A new calculation for Germany using the costs of dying
Introduction
Financing the modern welfare state in the future is commonly believed to be one of the most urging problems in today's politics. Presently, German workers pay 42% of their payroll for social insurance, a high tax wedge that is detrimental for employment. Therefore, politicians are desperately searching for ways to prevent or at least alleviate a further dramatic increase in the costs of social insurance over the next 30–40 years, e.g. by introducing a “sustainability factor” in the calculation of retirement benefits.
A big uncertainty is connected with the development of the expenditures of Social Health Insurance (SHI) because its benefits are predominately in kind rather than cash, and the nature of the respective product “health care” will change through technical progress in medicine as the structure of demand will change through a rising life expectancy. There have been numerous attempts to forecast the development of the average payroll contribution rate of German sickness funds. The spread of rates is rather wide. In his survey of forecasts, which have been published since 1995, Postler ([1], p. 23) reports on contribution rates for the year 2050 between 16.5 and 39.5% where the lower values refer to simulations of the effect of demographic change alone, ignoring any expenditure growth due to progress in medical technology.2
The uncertainty expressed in this wide spread, critical for health and social policy plans, has led to a fierce controversy on the need of a fundamental reform of the system of health care financing in the long run. Even the rather timid attempts to introduce more self-responsibility of patients, e.g. through a physician service fee of €10 per quarter, are attacked by trade unions and social democratic leftists to be the beginning of the dismantling of the welfare state. All the more, members of these circles openly deny that it may become necessary within the next decades to limit the set of basic benefits in SHI and in turn to widen the scope of supplementary private insurance. Politicians, for the same reason, are reluctant to introduce further legislation that would restrict the benefit package of SHI in the long run as such programs, interpreted as welfare state retrenchment, may risk the loss of many votes.
This reluctance runs the danger that inevitable reforms will later hit the citizens unexpectedly and might place a much heavier burden on them than a timely long-run policy. In particular, the build-up of private, capital-funded supplementary insurance requires a time lead of several decades.3 For these reasons, it is important to reduce those controversies which are derived from uncertainty of the future development.
A closer inspection differentiates three factors affecting future contribution rates for SHI:
- 1.
the impact of population ageing on the income basis, on which SHI contributions are levied; the latter shrinks when the share of workers in total population falls;
- 2.
the impact of ageing, in particular of the growth in life expectancy on per-capita expenditures;
- 3.
the impact of medical progress on per-capita expenditures.
The size of the first effect clearly depends on the way in which health care for the elderly is financed. In the U.S. Medicare system, the effect is very strong because medicare expenditures are financed mainly from general taxation which most heavily falls on the non-elderly. Thus, with the baby-boom generation reaching age 65, the ratio of beneficiaries to contributors rises sharply. The effect is less pronounced in an SHI system with universal coverage such as Germany's where all the earnings and retirement benefits are levied with the same percentage rate and derives from the fact that the incomes of the aged are lower than those of the working population. It would be eliminated completely if income-related contributions were replaced by fixed per-capita premiums since pensioners would then have to pay the same rates as workers.
The sign of the third effect is hardly controversial since there is by now a wide-spread agreement that medical progress predominantly consists of new products and procedures with both higher quality and costs. Only the size of this effect is still somewhat unclear.4 For Germany, Breyer and Ulrich [2] estimated it from the coefficient of time in a regression of per-capita expenditures on age structure and income, finding a growth rate of 1% per annum for the period 1970–1995.
Hence, it is the second effect that is probably the most controversial with respect to the future development of health care expenditures. There are at least three hypotheses on the impact of an increase in life expectancy on health care costs with constant medical technology:
- (a)
The status-quo hypothesis assumes that age-specific per-capita expenditures depend only on the state of medical technology and remain stable when the latter is controlled for. The impact of life expectancy can be calculated by applying present age–expenditure profiles on the future age distribution of the population (see, e.g. PROGNOS [3]).
- (b)
The expansion-of-morbidity hypothesis [4], [5], [6] is based on the observed multimorbidity of many elderly patients and states that the new possibilities of treating a specific type of illness (e.g. heart disease) prolong the patient's life without perfectly restoring his/her health so that shortly after another disease sets in (e.g. cancer), which requires additional treatment. According to this hypothesis, the main effect of technical progress in medicine is to prolong the life of those patients who are so sick that they would otherwise die, implying that the average health status of the population deteriorates over time. This explains Krämer's dictum ([5], p. 31) that “we spend the largest part of the additional years in the sickbed”.
- (c)
The time-to-death hypothesis is based on the conjecture that the observed difference in the health care expenditures between young and old persons in cross-sectional data are not primarily due to the calendar age, but are caused by the differences of time-to-death (e.g. [7]); in higher age groups, a larger share of persons is in their last years of life in which in a futile attempt to prevent death, a high amount of money is spent for the medical treatment. In such a situation, an increase in life expectancy, caused by medical progress or simply by a healthier life style, lowers age-specific death rates and, thus, fewer persons are in their last years of life in each age group. A stronger version of this hypothesis relies on the “compression-of-morbidity” effect first postulated by Fries [8] which states that as more people reach the natural limits of longevity, sickness gets compressed in a shorter and shorter period before death.
There is hardly any empirical evidence in favor of the expansion-of-morbidity hypothesis. On the contrary, Dinkel [9], exploring data from the German microcensus, finds that younger cohorts (birth years 1919 and 1913) not only experienced more life years beyond age 60 than the older ones (birth year 1907) but also an even larger increase of healthy life years. On the other hand, the time-to-death hypothesis stands on a firm empirical basis: the increase of treatment costs in the years before death has been convincingly documented in numerous studies with data from various countries (e.g. [10], [11], [12], [13], [14], [15]).
An additional effect, supporting even the compression-of-morbidity hypothesis, is the reluctance of physicians to treat very old terminally ill patients as aggressively as they treat younger patients with similar symptoms, a behavior commonly interpreted as “age-based rationing”. To the extent that this lets costs of dying decrease with the age beyond a certain threshold, it accentuates the overestimation of future expenditures in status-quo predictions. The empirical evidence on the decline of the cost of dying at a very high age is unambiguous. Although Zweifel et al. [11], [12] found no significant age impact on the health care expenditures among deceased persons in Switzerland, they report a decline of the costs of dying among over 65-years-old decedents, a finding that was confirmed by Felder et al. [16] and Schellhorn et al. [17]. Lubitz et al. [18] showed that medicare expenditures in the last 2 years of life for 70-years-old decedents were around 50% higher than for persons who died at age 90. Similarly, Busse et al. [19] found in a German sample that the number of hospital days in the last year of life peaked at the age group 55–64 and declined steadily with rising age of dying thereafter.
Studies of the overall effect of longevity on health care spending for the elderly such as Miller [20] show that an increase in longevity as such dampens the growth of expenditures.
Even if it is no longer questionable that the status-quo hypothesis overestimates the impact of demographic ageing on per-capita health care expenditures, it is still interesting to assess the extent of the error. A first, rather crude attempt by Breyer [21], using hospital days of surviving and deceased patients, estimated that the true ageing-related increase in total expenditures will amount to only 40% of the one calculated with a status-quo projection.
Now, a new data set from a Swiss sickness fund allows a much more precise assessment of the error in a naïve forecast. This data set contains the annual health care expenditures of over 91,000 persons of whom approximately 4% died within a time span of 3.5 years. The claims data includes expenses for in- and outpatient care as well as the costs of medication. SHI in Switzerland currently covers 60% of the overall long-term care, of which four-fifth is paid for the inpatient long-term care and one-fifth for the care at home.
Using a regression analysis, we estimate age–expenditure profiles for men and women, each separated by survival status (survivors versus decedents). Applying these expenditure profiles to the age structure of the German population in the coming decades as predicted by the Statistische Bundesamt,5 taking the estimated age-specific death rates into account allows an assessment of the purely ageing-related increase in the health care expenditures. By contrasting this “sophisticated” projection with a “naïve” status-quo prediction, it will be possible to assess the extent of the error more precisely than it was possible with previous data sets.
In the following we shall compare three different scenarios. In the first scenario, age-specific average health expenditures of the year 2002 will be directly applied to the age composition expected for future decades. In the second scenario, we distinguish explicitly between persons in their last 4 years of life and those who survive longer than that. In the third scenario, we take the compression hypothesis literally by adjusting the age-specific expenditures rightward by the difference in age-specific remaining life expectancies. For example, if the remaining life expectancy of a 65-year old will increase by 4 years until 2050, we shall assume that a 65-year old in 2050 will only spend as much as a 61-year old today.
To avoid possible misinterpretations of our study, we emphasize that we do not attempt to forecast future health care expenditures as a whole or to estimate the combined effect of demographic change on total health expenditures in Germany, which is at the center of the study by Schulz et al. [22]. Instead, we try to calculate a counterfactual, namely what the expenditures in 2002 would have been if the demographic composition corresponded to the predictions for certain dates in the first half of the 21st century.
The remainder of this paper is organized as follows. In Section 2, we describe the data and explain our methodology and in Section 3, we present the results, while Section 4 concludes.
Section snippets
Data and methods
A Swiss sickness fund made 1999 claims data of 91,327 persons available to us. Of these persons, 4% died between January 1, 2000 and June 30, 2003, i.e. 96% survived the year 1999 by at least 42 months. The data set allowed us to estimate the impact of both age and time-to-death on health care expenditures for both survivors and decedents. To account for the fact that in any given year there are persons with zero expenditures, we performed a two-stage estimation of individual health care
Results
We combined the two data sets described above, age–expenditure profiles for survivors and decedents from Switzerland and the population forecast for Germany, to estimate the purely demographic impact on per-capita health expenditures until 2050 when everything else, in particular technology and prices, is held constant. We distinguished the “q-model”, which takes into account the costs of dying and the expenditures in the last 4 years of life, from the “n-model”, which contains a naïve
Concluding remarks
In Section 1, we presented three alternative hypotheses referring to the impact of the increase in life expectancy on per-capita expenditures in SHI in Germany and countries with similar demographic challenges. Taking Swiss expenditure data as a basis, the (weak) time-to-death hypothesis gets the strongest confirmation of these hypotheses. Explicitly accounting for costs in the last years of life leads to a downward correction of the demographic impact on per-capita expenditures as compared to
Acknowledgements
We are grateful to Andreas Werblow for technical assistance and to the CSS Sickness Fund in Lucerne for the provision of the data set. This paper was presented at the annual meeting of the Health Economics Section of the Verein für Socialpolitik in Fulda on October 6, 2004. Valuable comments by the participants and by an anonymous referee of this journal are gratefully acknowledged.
References (30)
- et al.
A longitudinal study of the effects of age and time to death on hospital costs
Journal of Health Economics
(2004) - et al.
Health care expenditure in the last months of life
Journal of Health Economics
(2000) - et al.
The impact of ageing on hospital care and long-term care: the example of Germany
Health Policy
(2004) - Postler A. Modellrechnungen zur Beitragssatzentwicklung in der Gesetzlichen Krankenversicherung. Diskussionspapier,...
- et al.
Gesundheitsausgaben, Alter und medizinischer Fortschritt: eine Regressionsanalyse
Jahrbücher für Nationalökonomie und Statistik
(2000) - PROGNOS. Auswirkungen veränderter ökonomischer und rechtlicher Rahmenbedingungen auf die gesetzliche Rentenversicherung...
- et al.
Trading off longer life for worsening health: the expansion of Morbidity Hypothesis
Journal of Aging and Health
(1991) - Krämer W. Wir kurieren uns zu Tode. Die Zukunft der modernen Medizin. Frankfurt/New York;...
- Krämer W. Hippokrates und Sisyphus. Die moderne Medizin als Opfer ihres eigenen Erfolgs, in: Kirch W, Kliemt H,...
Though much is taken: reflections on aging, health and medical care
Milbank Memorial Fund Quarterly/Health and Society
(1984)
Aging, natural death and the compression of morbidity
New England Journal of Medicine
Trends in medicare payments in the last year of life
New England Journal of Medicine
Ageing of population and health care expenditure: a red herring?
Health Economics
Cited by (99)
An analysis of the thematic evolution of ageing and healthcare expenditure using word embedding: A scoping review of policy implications
2023, Socio-Economic Planning SciencesThe effects of ageing population on health expenditure and economic growth in China: A Bayesian-VAR approach
2020, Social Science and MedicineAge-specific patterns of health care expenditure in dying people
2017, Public HealthShould we adjust health expenditure for age structure on health systems efficiency? A worldwide analysis
2023, Health Economics ReviewHow Superintelligence Affects Human Health: A Scenario Analysis
2023, Journal of Futures StudiesMedical Progress, Ageing, and Sustainability of Healthcare Finance
2023, The Routledge Handbook of the Economics of Ageing
- 1
Tel.: +49 391 53 28050; fax: +49 391 67 190250.