The results show that the total expenses associated with hospital treatment of persons who died in 2010 amount to 10.6 % of total hospital expenditure, but that the average expenditure per person decreased with age (Figure 2). Hospital expenses for women in their eighties who died in 2010 were on average more than NOK 50 000 (33 %) lower than for seventy-year-olds who died in the same year. Corresponding figures for men were a decrease of NOK 40 000 (24 %). The figure also shows that more resources were spent on women than men who passed away at a relatively young age (under 60 years of age), while more was spent on men who passed away at an older age (over 80 years of age).
There are several reasons why expenses vary with gender and age. First, studies show that those who die at an early age often succumb to diseases or injuries that require more hospital resources than those who die in old age (15). For example, many die from cancer at a relatively young age, especially women, whereas people in older age groups more frequently die of a myocardial infarction (16).
International research in this area indicates that one cannot explain the entire difference between the age groups in terms of the prevalence of diseases or events; some of it may also be due to the fact that identical diseases are not treated as aggressively in older age groups (17) – (19). There may be sound medical reasons for this, since elderly patients are unable to sustain the same tough treatment as younger patients, although the literature indicates that elements of age discrimination may also be involved (16, 18). Further studies are required to clarify this.
The variations between consecutive age cohorts illustrate the impact of historical and institutional issues. For example, health and expenses may be influenced by retirement or by having been born into an age cohort that has experienced particular events or epidemics. In addition to random variations, such issues illustrate why the expenses do not follow a straight decline, but vary around a decreasing trend at different age levels.
A potential problem associated with investigating hospital expenses for those who died in a particular year as opposed to those who lived through the entire year is that one fails to capture all death-related hospital expenses: If the costs associated with the cause of death are spread over several years, the investigation will be biased if only hospital expenses in the current year are included. In the data set, this applied in particular to those who died early in 2010, where the hospital expenses most likely had increased from late 2009. A key issue is the extent to which the costs were incurred at the very end of life, or whether they were more evenly distributed over a short or long period prior to death. This is illustrated in Figure 3, which shows two clear trends: First, expenses increased slowly towards the time of death. In analyses of data from the UK, a slight increase could be found as far back as 15 years before death (3). This could indicate that the definition of death-related expenses ought to be expanded. Second, the analysis of our material shows that even though the costs increased long before the time of death, this increase was minor compared to the increase in expenditure during the last few months. Hospital expenses accrued during the final month accounted for 30 % of all costs over the year for persons who died in December. Among patients in their eighties, the final three months prior to death accounted for 50 % of the total cost. We found a similar pattern for the other high-age groups. We may therefore assert that even though hospital expenses start increasing for some time prior to death, the bulk of the cost is incurred during a short period at the very end of life.
As a whole, this means that the figures probably capture a considerable proportion of the hospital expenses associated with the disease that caused death, but that the sum would have been greater if we could have obtained an overview of all death-related hospital expenses over several years. At the same time, by extending the time frame we would also run the risk of including treatment expenses that were unrelated to the end of life. For example, not all costs of treatment in January will necessarily be related to the cause of death if the patient dies in December. The correct interpretation of the figures is therefore «hospital expenses associated with persons who died in 2010», and that this represents a cautious, but not unrealistic, estimate of «hospital expenses towards the end of life».
The observation that large hospital costs are incurred by patients towards the end of their lives has major implications for the extrapolation of health expenditure. Assuming an increase in life expectancy, a traditional extrapolation of hospital expenses involves multiplication of the current average cost in various age groups with the number of people expected in these age groups in the future (20, 21). A problem associated with this method is that it fails to distinguish between expenses associated with the period prior to death, irrespective of age, and the expenses that follow from age alone. Today’s average expenses for people in their seventies include the costs of the many who die at age 70. If life expectancy rises, the proportion of people who die at age 70 will decrease, and the age-related average expenses for people in their seventies will thus also decrease. A more correct extrapolation of expenditure levels should take this into account.
A possible limitation of the analyses can be found in the use of DRG weights as a measure of expenses. In reality, some patients will cost more and others less than the DRG rate. If it costs more to treat a patient in her final year of life than another patient with an identical diagnosis and DRG category, the use of DRG rates will underestimate the costs associated with the end of life. There is also the possibility of an opposite bias: That the treatment of patients with the same diagnosis costs less when the patient is in her final year of life and unable to sustain similarly aggressive treatment. In this case, the DRG rates will overestimate the costs.
The finding that approximately 10 % of all hospital expenses were devoted to patients who died in the course of the year should be treated with some caution in the context of prioritisation. This requires another approach, investigating the extent to which the quality and length of life increase for those patients who receive hospital treatment in what could potentially be their last year of life. While some of these patients will die, many others will have a longer and better life as a result of the treatment. Moreover, many of those who die will have an enhanced quality of life as a result of hospital treatment in the period preceding their death. We have only investigated one half of this assessment, the costs, and can conclude that the costs associated with patients who die in the course of a year account for approximately 10 % of all hospital costs and vary somewhat according to gender and a lot according to age.
We would like to express our gratitude for the useful comments provided by the peer reviewers and the editors of the Journal of the Norwegian Medical Association, as well as colleagues such as Oddvar M. Kaarbøe, who pointed out the potential problems associated with using the DRG rates as a measure of costs. We would also like to thank the Ministry of Finance, the Research Council of Norway and The Commonwealth Fund for their financial support.