1
Bachelor thesis; Economics
A long term projection of the
Dutch healthcare costs.
Date: January 30, 2018
Name: Sietske van Sloten
Student number: 10800581
2
Table of Contents
Introduction ___________________________________________________________________ 3 Literature review _______________________________________________________________ 5 Research design _______________________________________________________________ 9 Data analysis & Results ________________________________________________________ 11 Discussion ___________________________________________________________________ 20 Conclusion __________________________________________________________________ 22 Bibliography _________________________________________________________________ 23 Appendix ____________________________________________________________________ 25
Statement of Originality
This document is written by Student Sietske van Sloten who
declares to take full responsibility for the contents of this document. I declare that the text and the work presented in this document are original and that no sources other than those mentioned in the text and its references have been used in creating it.
The Faculty of Economics and Business is responsible solely for the supervision of completion of the work, not for the contents.
Abstract: Throughout Europa population is growing and aging, which leads to rising healthcare costs.
The healthcare costs in the Netherlands have been rising very quickly in recent years. The knowledge of healthcare costs in the future is important for the debate on required policy changes. In this paper the future Dutch healthcare costs are estimated for the long run, for three different scenarios.
The expected demographic changes, a growing and aging population, is examined in all three scenarios. The ‘last year of life’ healthcare costs are lot more expensive than healthcare costs for a normal year, both healthcare costs depend heavily on age. This difference in costs and trends for different age group are studied in the first scenario. The second scenario will include the effect of declining disability rate for elderly on the healthcare costs in the long run. Declining disability rate has a decreasing effect on long-term care volume, and therefore affects total healthcare costs. The effect of national income elasticity and the residual variable, representing mostly technological and policy change, will be studied in the last scenario.
The main conclusion is that national income elasticity and the residual variable have a major effect on healthcare costs, much more significant than the demographic changes.
3
Introduction
The British free healthcare, the National Health Service (NHS), finds itself in a big crisis this winter. Fifty thousand surgeries have been canceled last month (January 2018). Emergency rooms at various hospitals, are so overcrowded that patients have to wait in the ambulance for hours. The cause? A growing population and an aging population (De Wit, 2018).
In the Netherlands the problems have not yet been as pressing as they are in Great Britain, but above mentioned causing factors occur throughout Europe. The age structure of the EU
population is expected to change dramatically. The number of elderly people is projected to account for an increasing share of the population, due to the combination of the numerous cohorts born in the 1950’s and 1960’s and the continuing expected gains In life expectancy (European Commission, 2015). A paper commissioned by Netherlands Bureau for Economic Policy Analysis (CPB) states:
‘the aging population makes a claim on the public finances. The collective spending on aow (basic state pension) and healthcare will rise, income does not keep up. This deterioration of the budget balance is ultimately untenable, so that government must adjust its policy (Van der Horst et al., 2010).’
According to research done by the Central Bureau of Statistics (CBS) health costs will rise to 19-31% (Dutch definition) of GDP by 2040, because of more, better and more expensive healthcare (Van der Horst et al., 2011). A rise of the healthcare costs seems certain, but how fast and how high seems to be the question. Even the research by CBS ends up with a difference of 12 percentage points for the worst and best case scenario.
It is not surprising that healthcare costs is subject to a lot of discussion, on how to prevent a crisis in the healthcare system. But is crisis near, or are we jumping to conclusions too soon? To find the answer to this, a long term projection of the healthcare costs is required.
Many times the aging population is mentioned as one of the key factors behind the rising costs. Therefore it is important to take these demographic changes into account, for this long term projection. Demographic changes include the aging population, but also the increase of life
expectancy for the Dutch population. These demographic changes need to be put in perspective, is the future demographic situation really this important for the changes in healthcare costs? The effect of change in disability rates has on long term care is also examined in this paper. Finally economic growth, technology and policy, play a serious factor in the healthcare costs. Important is that the correct income elasticity on healthcare is selected when estimating the future healthcare costs.
This paper gives a few different future scenarios, and examines multiple trends that might affect healthcare costs. The knowledge of healthcare costs in the future is important for the debate on required policy changes. In this paper, policy makers can be find which fields require their attention. For the literature regarding healthcare, we want to make clear how important the income
4
elasticity is for future healthcare costs. This will be shown by estimating different scenarios for different income elasticities.5
Literature review
Since 2000 the healthcare costs are growing faster than GDP. In the time period 2000-2014 the costs grew with 7,5 percent per year, while GDP grew with 3.4 percent (Striem &
Bhageloe-Datadin, 2015). The aging population is often mentioned in the media as one of the main causes of the rising costs. The aging population is indeed a trend that will change a lot in Dutch society. At the peak of the aging population, in 2039, the Netherlands will have 4,6 million inhabitants of 65 and older (van Duin & Stoeldraijer, 2012, pp. 25). A stable decline in mortality rates can be observed in the Netherlands, according to van Duin & Stoeldraijer this trend is most likely to continue in the next decades. This means an increase in lifespan, people become older. Theories state that these extra years lead to the big increase in healthcare costs. More traditional projection methods overestimate the influence of health expenditure. Simply stating that healthcare costs will rise because life expectancy is increasing is too simple according to American research, they state that two factors work in the other direction (Cutler, Sheiner, 1998). Namely costs of the last year of life and disability rates. These factors will moderate the demographic effect on healthcare costs.
1. Last year of life
Improvements in life expectancy will postpone rather than raise health expenditure (Polder, Barendregt, Oers, 2006). This is also what Cutler & Sheiner state; an increase in life expectancy means that smaller share of elderly will be in the last year of life, when medical costs are very high. Polder et al. found that about 10 percent of total health expenditure in 1999 was associated with the healthcare use of people in their last year of life, and that healthcare costs in the last year of life of people who died in that year were on average 13,5 times higher than costs of people who did not die. Two patterns can be identified: one for decedents and one for survivors. The first one is that healthcare costs in the last year of life depend on age, costs were high for people dying at comparatively younger ages (<70 years), and turned out to decrease with increasing age of death, mainly due to a decrease in hospital care. The healthcare costs in the last year of life for elderly is mostly care, costs for nursing homes and home care. People dying at a younger age are mainly ill, their costs are not care costs but cure costs. Costs made through physicians, hospitals, drugs and related services. Therefore their last year of life healthcare costs are a lot higher, because they represent these more expensive curing costs. Especially the costs for the last year of life for people suffering from cancer, are a lot higher for younger people than older.
The second pattern is an increase of costs with age for survivors, in this case the normal annual health costs not the ‘last year of life’ costs. Given these patterns the costs-ratio between decedents and survivors depends heavily on age. So increasing longevity will still result in higher costs because people live longer, but the decline of costs in the last year of life with increasing age will have a moderate lowering effect.
6
2. Disability ratesThe second moderating factor are the declining disability rates among elderly, this trend leads to less need for long term care. Healthcare consists of curative care and long term care. The curative costs account for approximately 60 percent of total costs, the rest is long term costs. Elderly costs and costs for disabled together make up the main components for long term care costs, the elderly costs for about two thirds and disabled costs for one third (van Striem & Bhageloe-Datadin, 2015). Long-term care service is provided in institutions or at home, which one is more common is put forward as a crucial determinant of public expenditure on long-term care (Lipszyc, Sail, & Xavier, 2012). Because institutional care is seen as much more costly than home care, even though the difference is less clear for very severe cases. Lipszyc et al. found that in 2010 institutional care costs for one recipient equals 98 percent of GDP per capita for that year, while home care costs equals 14% of GDP per capita per recipient. So it is interesting to study trends for older long-term care recipients, and if they receive their care at home or in an institution. If elderly need less long term care because of declining disability rates, this could save a lot of money.
Declining disability rates, if sustained, will reduce medical costs on the elderly (Cutler, Sheiner, 1998). According to Dutch research disability rates in the Netherlands are also declining (de Hollander, et al., 2006). In this research they talk about how life expectancy in good health is increasing in the Netherlands. Especially the increasing trend in years without disabilities is notable. Since the eighties people have six to seven years more years without disabilities, these years are mostly due to a decrease in disabilities around mobility, like hearing and eyesight. According to de Hollander et al. there is no explanation for this trend. Other Dutch research found that disabilities in mobility among elderly decreased from 40 percent in 1987 to around 25 percent in 2001 (Puts et al., 2006). De Hollander et al. have the same conclusion as the American
researchers. Less disabilities means more self-reliance, more social participation, and in the end a decrease in demand for long term care.
3. Technology & Policy
The aging population only explains a small part of the rising costs. Due to the growing proportion of elderly people in the population, healthcare expenditure is one percent of GDP higher than in 1972 (Wouterse, ter Rele, Vuuren, 2016). According to a report on trends in healthcare costs from the RIVM, only 15 percent of the increase in healthcare costs for 1999-2010 was due to the aging population. Almost 50 percent of the increase can be attributed to a group of causes, such as broadened treatment indications, increase in number of patients, more intensive treatments, and new medical technology (Slobbe et al., 2011). The other 35 percent is caused by price
developments. In this research it is assumed that healthcare costs per capita is a fixed share of GDP, therefore it is not necessary to study the price developments. The focus will be on the volume development. According to Kommer et al. the most important factors in the 50 percent
7
group, are medical technology and effect of policy (2006). The government often decides toimplement a new policy, that affects healthcare volume and costs. For example, the government can decide that people have to pay a part of the treatments themselves, this might have the effect that people are less tended to go for a certain treatment. Or the government can implement a policy to eliniate the waiting lists, they could modernize the healthcare system, or decide to
implement a new law. This way the government can influence healthcare volume. Cutler & Sheiner (1998) but also Dutch research conclude that extra costs is mostly due to new more expensive technologies. Technological progress is expensive, because more and more diseases are treatable at a high price (van Striem & Bhageloe-Datadin, 2015). Because of the inelasticity of the demand for new treatments, the price does not really influence demand. So people will want the new treatments, despite the high price.
Researchers agree that a part of the increase in healthcare costs is due to medical technology, but they have not found a way to quantify it. We are constrained by a lack of data, but also the abstraction of the concept (Kommer et al., 2010). Medical technology is highly correlated with national income. High prosperity of a country determines the level of technology and the supply of healthcare in that country, therefore there is strong cross-action between these factors. This is the same for policy, it’s highly correlated to income. This is the reason that Jong (2011) decides not to include technology in his model, which he uses to make a decomposition of the healthcare costs. He states that a new treatment method for an earlier untreatable disease will lead to a volume increase. This is also the case for improvements of the quality of existing treatments. This means that the relation between income and healthcare volume is related to both number of treatments and operations as the quality of them. So in his decomposition he looks at the change in
healthcare price and healthcare volume.
4. Income elasticity
As the incomes of countries rise, their spending on healthcare and on other goods and services increase. Income is one of the most recurring explanatory factor for healthcare costs in the
academic literature (Ligthart, 2007). How big the effect of income on healthcare costs is, depends on the income elasticity. Incomes elasticity indicates by how many percent the volume of
healthcare increases when real income increases with one percent, this income elasticity is on national level. When this elasticity is above one, healthcare is perceived as a ‘luxury’ good and this will make healthcare spending grow faster than income. This means that the extra money
someone earns all goes to healthcare, including some extra. With the result that there is less money for other goods or services. When elasticity is between zero and one, healthcare is
perceived as a ‘normal’ good. In that case, when income increases healthcare costs will grow at a lower rate. Economic papers on the income elasticity of healthcare have very different conclusions.
Getzen states: ‘At the macro level, studies of national expenditures consistently show income elasticities greater than 1.0, with 90+ percent of cross-sectional and time series variation
8
explainable by differences in per capita income’ (2000, p. 294). He looked at ten different articles on national income elasticity on healthcare costs, and concludes that elasticity lies between 1.2 and 1.6.At the other end of the spectrum is, for example, Ligthart. He concludes that on the grounds of his research in which he used multiple models, the income elasticity for de EU-15 is estimated at a value between 0.5 and 1. Estimates based on solely Dutch data shows an income elasticity of 0.5. He states that the coefficient related to income per capita is sensitive for the estimation period and for the specification. Earlier studies, which found income elasticity above one, did not include a trend or another proxy for technological development. This would lead to a higher income elasticity according to Ligthart. Secondly earlier studies have an earlier estimation period, it seems that these earlier estimation periods lead to a higher income elasticity. These results appear to indicate that income elasticity declines in the course of time. Ligthart states when countries are more industrialized and reach a higher level of prosperity, the income elasticity is lower.
De Jong assumes an income elasticity of one in his decomposition of the healthcare costs. A safe and easy assumption, in the middle of the spectrum. According to De Jong, this definition is also used by the European Commission (2008, 2009). In this case healthcare volume per capita grows one-to-one with income.
5. International healthcare definition
The international definition of healthcare differs from the Dutch definition. Big parts of the Dutch healthcare system are not covered by the international definition (van Striem, & Bhageloe-Datadin, 2015). Especially big parts of the long term healthcare that does not take place in institutions, that is care of a more social nature. Also a part of the long term healthcare in institutions is not included in the international definition, when there is very little evidence of treatment or nursing. Costs for well-being and social services are also excluded. These costs are not registered well
internationally, and therefore not included internationally. In this paper the international definition is adopted, to make sure that this paper can be used for comparison.
9
Research design
A long term projection on healthcare costs in the Netherlands is done in this paper. A paper on forecasting health expenditures gives some insight in the best ways to do forecasts on this topic (Getzen, 2000). Getzen states that in the long-term, inflation is just an superficial illusion. To measure healthcare costs in the long run, focus must be on how much society has to give up to pay for medical care, not price in euros. Therefore the appropriate measure of costs in the long run is the percentage share of GDP spent on health, and affordability depends fully on the grow of GDP. While economic logic and historical trends establish some limits, uncertainty remains very wide for the long run. The further the projection is in time, the less reliable.
In this paper the healthcare costs for the period 1999-2059 will be estimated for three different scenarios. All scenarios will take the effect of demographic changes on healthcare costs into account. The first scenario is a simple prognosis, that will be the basis for the other scenarios. In this scenario the healthcare costs grow with the same rate as GDP, so the average costs per person will be the same share of GDP for each year. The difference in costs for a normal year and a ‘last year of life’ year is included in all scenarios.
The decreasing disability rates are included in the second scenario, by looking at the long term care trends. Data on long term care recipients of 65 years and older is provided by OECD. By examining the change in each year, a trend can be detected. An average decrease or increase needs to be connected to the costs of long term care. After all this, the findings can be
incorporated in scenario one. The result in scenario two will be compared to the results in scenario one.
Scenario three will focus on the factors apart from demographic factors that lead to a change in healthcare costs. According to the literature these factors could account for 50 percent of the total change of healthcare costs. The factors studied are income, technological change and policy. Changes in healthcare due to income can be measured, when the national income elasticity for healthcare is known. The other part is the residual variable, which consists for the largest part of technological and policy factors. Again these factors will be incorporated in scenario one, and will be compared.
The data on healthcare costs for survivors and decedents for the year 1999 is provided by Polder et al. With the data on the average costs per capita and total GDP in 1999, the average healthcare costs as part of GDP will be calculated for each age cohort. With the population data from CBS, total population decedents and survivors can be set out for each age cohort for each year in the time period 1999-2059. Total costs as share of GDP for each age cohort for each year can be calculated with this data. All the costs per age cohort summed up will give the total
healthcare costs of GDP for a certain year. This number covers 68.5 percent of the total healthcare costs according to international definition, so it needs to be converted to 100 percent. In the end the residual is added, in two different ways. Method one is a fixed residual added up to the results,
10
method two is a multiplication with a fixed factor. This final numbers can be plotted in a graph, so the trend can be examined. The assumption in scenario one and two, that healthcare costs keep a constant share of GDP, makes method one a more logical choice. This residual is a fixedpercentage share of GDP, the assumption is kept intact. For scenario three, the second method might be more in line. In this scenario healthcare costs per capita is not a constant share of GDP anymore, it changes every year. It would be incorrect to keep the residual a constant share of GDP. Therefore the second method might give a more realistic result.
11
Data analysis & Results
1. General trends
In this paper the same age cohorts are used as in the paper by Polder et al. This was necessary because the data on average healthcare costs was taken from their study. These age cohorts are from 0 to 45, 45 to 55, 55 to 65, 65 to 70, this continues in steps of 5 to 95+. A few interesting trends and deviations can be identified in the different data on population provided by CBS. The large group of people born after the world war two, better known as the baby boom, can be followed through the years. The largest level of people for each age cohort represents this group (table one). After every few years the next age cohort shows a peak. Unfortunately the end of this trend is not visible in this data, this trend continues after 2059.
Another trend is the lowest number of people in each age cohort. This trend seems to be more stable, especially from age cohort 55 and up. The lowest level seems to be every 5 years. This trend is most likely also attributed to the baby boomers. When the largest part of the baby boomers are in the next age cohort, they leave a gap in the previous one (clearly visible in the population pyramids).
The following population pyramid show the demographic changes in the Netherlands. The number of young people keeps quite constant over the years. The large group of baby boomers can be followed through the time periods. The gap they leave is also very visible in these pyramids. It is evident that policy makers should take into account the big changes that are coming up for the group of elderly. The younger age cohorts will not change much in the next 40 years, while small changes can be identified in the middle age cohorts (see population pyramid to the far right)
12
2. Scenario one: total healthcare costsThe first scenario for forecasting healthcare costs creates the foundation for the other scenarios. The focus in scenario one will be on the demographic changes and their effect on healthcare costs. The group of elderly is growing in the Netherlands, this will have an effect on the healthcare costs. Table two shows that the average costs for elderly are a lot higher than for the younger people. The average healthcare costs for someone 95 years or older is about 6.5 times higher than for people between 0 and 45. The opposite trend can be identified for the costs for the last year of life, these costs start very high for the younger people. The costs are quite stable for ages 0 till 70, they peak at 70-75 and then decrease quite rapidly with the lowest point at 95+. At this point the costs of the last year of life are almost half of what they are at 0-75. Still these last year of life costs are a lot higher than the average costs for a survivor. The last year of life is an expensive year for
everyone.
In this scenario average healthcare costs per person keeps a stable share of GDP for the forecasting period, assuming an income elasticity of one. Therefore all the healthcare costs per capita are converted to share of GDP for 1999, the base year. For every year the number of decedents for each age cohort will be multiplied with the costs as share of GDP of the last year of life for their age cohort, this will give the total costs for decedents for each year as share of GDP. The same thing is done for the costs of the survivors. Total decedents and survivors for all age cohorts together gives the total healthcare expenditure for each year as share of GDP. The
following graphs provide some insight in the movement of the share of costs attributed to survivors and deceased. Graph 1 makes very clear that the costs of survivors leads to the increase in healthcare costs, with a peak in 2052. The total costs for the decedents is almost entirely flat for the whole projection period.
Graph 1: Total costs survivors (blue) & deceased (red), Graph 2: Trend total costs 0-65 (blue) & 65+ (red),
as % of GDP as % of GDP 0 .0 2 .0 4 .0 6 .0 8 2000 2020 2040 2060 Year Survivors Deceased .02 .025 .03 .035 .04 .045 2000 2020 2040 2060 Year 0-65 65+
13
Graph 3: Share population deceased Graph 4: Share of costs deceased
Graph 2 shows that the group of elderly cause the rise of healthcare costs.The red line,
representing the elderly, increases very fast. While the blue, representing the younger people, is almost flat. It is clear that policy makers should focus on the elderly, when trying to implement policies that will decrease the healthcare costs.
Graph 3 shows the share of population that die in a year, this rate increases over time. So a larger share of people will be in their last year of life every year until 2054. Because the last year of life is more expensive than a normal survivor year, you would expect that share of costs for deceased would rise. But the contrary is true, as shown with graph 4. Graph 4 plots the share of costs attributed to the deceased. Because the last year of life costs decline when people die at a older age, total costs for deceased can stay stable while a growing share of the population is in their last year of life every year. At the same time the survivors costs increase for older people, that is why total survivor costs increase. Which leads to a declining share of costs attributed to deceased.
According to Polder et al. the average healthcare costs for 1999 they provided accounted for 68.5 percent of the total healthcare costs, the findings need to be converted to 100%. According to these numbers 6.8 percent of GDP would be spend on healthcare in 1999. The outcome for 1999 deviates from the data provided by CBS on healthcare costs by international definition, CBS states that eight percent of GDP is spend on healthcare . The cause of this difference might be found in the average healthcare costs provided by Polder et al., they used an sample representing 13.4 percent of the Dutch population, which comes down to 2.1 million Dutch inhabitants. This is a big sample, but still too small to generalize for the entire Dutch population. but possibly still too small to generalize it for the whole Dutch population. The actual average healthcare costs could therefore deviate from the data found by Polder. CBS publishes healthcare data for the Dutch and the international definition. Data from CBS seems to be the most reliable data on this topic,
therefore it’s better to adapt the findings to theirs. Two different methods are used to add a residual to the findings. The first method increases total healthcare costs with a fixed residual every year, in this case the findings increase with 1.1667 percentage point of GDP. The second method
.8 .9 1 1 .1 1 .2 S h a re d e c e a s e d 2000 2020 2040 2060 Year 8 .5 9 9 .5 1 0 1 0 .5 1 1 S h a re c o s ts d e c e a s e d 2000 2020 2040 2060 Year
14
multiplies the findings with a fixed factor, in this case with 1.17. The two different methods are shown in graph 5.Graph 5: Healthcare costs as % of GDP for the Netherlands.
With both methods the peak of the healthcare costs is in 2052, in the time following it decreases again. The increase in healthcare costs is caused by the increasing costs for survivors, the costs for survivors is most expensive for elderly. Most likely the largest part of the baby boomers has passed away in 2052 or is in their last year of life, resulting in lower costs for total healthcare and total healthcare attributed to survivors. The second method lets the healthcare cost increase faster, with a peak at 12.55 percent of GDP spend on healthcare in 2052. With the first method this peak is at 11.88 percent of GDP.
3. Scenario two: long term care.
Data on disability rates corresponding costs are hard to find. But the consequence of declining disability rate is less need for long term care, and data on long term care can be found in the OECD database. OECD provides data on the percentage of 65+ who used long term care for the period 2004-2014. A decreasing trend can be identified (graph 4). These recipients of long term care are divided in care in institutions and care at home. Table 3 shows that especially the percentage of people receiving long term care at home fluctuates a lot.
.0 8 .0 9 .1 .1 1 .1 2 .1 3 2000 2020 2040 2060 Year Scenario1 Scenario1 M2
15
Graph 6: Long term care recipients, share of total aged 65 years old and over. Source: OECD
2004 2005 2006 2007 2008 2009 2010 2011 2012 2013 2014 in institutions. 7.3 7.1 6.9 6.9 6.7 6.7 6.6 6.5 6.2 5.6 5.3 At home 15.4 13.4 13.2 13.1 13.1 13 12.9 13.9 13.7 13.4 13.1 Total 22.7 20.5 20.1 20 19.8 19.7 19.5 20.4 19.9 19 18.4
Table 3: Long term care recipients, share of total aged 65 years old and over. Source: OECD
The percentage of people receiving long term care in institutions declines with a steady rate. Especially the large decrease from 2004 to 2005 is notable. This decrease is clarified by the following change in Dutch policy (translated from Dutch):
From 1 January 2005 the indication definition for AWBZ-care is in the hands of the Centrum Indicatiestelling Zorg (CIZ). The responsibility for defining the indication has shifted from the over 80 Regionale Indicatie Organen (RIO’s; regional indication organizations) to one nationwide indication organization, the CIZ.’ (Peeters & Francke, (2007), p. 7)
This explains why 2004 deviates so much from the other years. For this scenario we need the average annual decline of total long term care. Because the data on long term care in 2004 is defined very differently from the years 2005 to 2014, 2004 will be dropped from our calculations. Like discussed in the literature review, long term care consists of elderly care and care for
disabled, in this scenario the focus is on care for elderly. The RIVM part of CBS provides data on the costs of elderly care, for the years 2003, 2005, 2007 and 2011. Because 2004 is dropped from our data set, 2005 is used as the base year. RIVM provides the long term elderly care costs per capita per year for the age group 65+. This is per capita for the people in the 65 and older age
1 8 1 9 2 0 2 1 2 2 2 3 % L T C r e c ip ie n ts 2004 2006 2008 2010 2012 2014 Year
16
group, not the whole population. According to the international definition (SHA) the long term elderly care is 2385 euros per capita (65+) per year. In 2005 20,5% of the 65+ were provided with long term care, this means that the average costs for every person that used this service was 11634,15 euros (2385/0,205). GDP for the Netherlands in 2005 was 545690 million euro’s. So in percentage of GDP, one elderly person who uses long term care costs about 0.00000213 percent. To make it clear the total costs of long term elderly care in 2015 equals 11634 * 0.205 * 65+ population.The percentage of people who receive long term care changes every year, on average it declines with 0.233 percentage point. So in the prognosis the long term care recipients in 2006 decreased with 0.233 percentage point, in 2007 with 0.466 percentage point compared to 2005, in 2008 with 0.699. So the long term care costs per recipient as share of GDP (0.00000213 percent) multiplied with the decline in percentage (for 2006 multiplied with -0,00233) gives the decline per capita in the 65+ population. This number multiplied with the population of 65+ gives the yearly decline of healthcare costs due to declining long term care rates. This decline joined with the healthcare costs found in scenario one, gives the healthcare costs for scenario two.
In graph 7 it is very clear that the peak of healthcare costs of scenario two is in a different year than in scenario one. The peak is in 2047, five year earlier. The expensive survivor years for elderly will decline over time because of less long term care, this explains why the peak is earlier. The decline after this peak is a lot steeper than in scenario one. With method one (shown in graph) the peak in scenario one is at 10.8 percent of GDP, method two shows a peak at 11.4 percent. A difference of 1.05 percentage point of GDP with scenario one, a significant difference. These findings are expected, because we calculated with a steady average decline of 0.233 percentage point.
Graph 7: Healthcare costs as % of GDP for the Netherlands.
.0 8 .0 9 .1 .1 1 .1 2 2000 2020 2040 2060 Year Scenario1 Scenario 2
17
But how likely is it that the long term care will keep declining for this long-term period. The dataset where the average decline is based on, is only nine years long. In the prognosis the percentage of long term care recipients declines from 20.5 percent in 2005 to around 8 percent in 2059. This is quite a big decline, and most likely not achievable. The longer the projection, the less reliable it will become. This is definitely true in this scenario, but it is still important to incorporate the trends in long term care in healthcare forecasts. To get a more reliable estimate, more data on long term care is needed. Therefore it is important to keep collecting data on long term care in the future, and investigate if this trend can continue. Policy makers and researchers need to follow this trend. Research is needed on the causing factors of this trend, and how to ensure this trend can continue.4. Scenario three: income and the residual variable
In this scenario the other factors for volume growth that make healthcare costs increase are taken into account. Scenario one will again be the base, scenario two will not be included in this scenario because this will make the result more uncertain. These remaining factors are economic growth and a the residual variable. The residual variable most likely represents technological growth and policy change, as discussed in the theoretical framework. Technological change and policy are intertwined with income change. If we want to see the effect of technological change and policy, the income effect needs to be taken into account. De Jong found that over the period 1973-2010 increasing income caused healthcare costs to increase with 1.8 percent per year (2012). The residual variable accounted for an increase of 1.1 percent per year. Summed up income and the residual variable caused for an annual healthcare costs growth of 2.9 percent. De Jong assumes an income elasticity of 1 in his decomposition.
The first analysis in scenario three will be done with an income elasticity of one. Healthcare grows at the same rate as income. The world bank provided data on the annual GDP growth per capita in percentages for the period 1961-2016. On average GDP per capita grows with 2.2 percent per year. This number subtracted from the 2.9 percent annual rise in healthcare costs related to income and the residual variable, leaves us with a healthcare growth rate of 0.78 percent per year. In this scenario all the healthcare costs per capita for survivors and decedents (table 2) increase with 0.78 percent per year. This data is added to the data of scenario one, and plotted.
18
Graph 8: Healthcare costs as % of GDP for the Netherlands (Method one).
The trends are very different for both scenarios. The demographic changes seems to have almost no effect on the rise of healthcare costs in scenario three. These findings are consistent with the majority of papers on this topic. Demographic changes do have an effect on healthcare costs, as can be seen in the blue line in the graph. But more importantly is the effect of the residual variable, technological change and policy. With an income elasticity of one, healthcare costs is estimated to be 18.5 percent (method one) of GDP in 2059 and keeps on rising. The second method estimates that 19.9 percent of GDP goes to healthcare, that is one fifth of all spending on national level. Scenario three differs a great deal from scenario one. Demographic changes and changes in long term care recipients can have a smoothing or steepening effect, but only up to a certain point.
But what if income elasticity at national level is not 1, but higher or lower? As earlier discussed, the range for income elasticities is broad. Researchers like Getzen, describe national health expenditure elasticities as typically greater than one with a maximum of 1.6 (2000). Ligthart states that income elasticity is definitely below one, and estimates Dutch national income elasticity at 0.5. The effects of these extremes, an income elasticity of 0.5 or one of 1.6, is also interesting to further investigate.
The average income growth in the Netherlands is 2.2 percent. The change in healthcare costs due to income changes with a 0.5 income elasticity would be 0.5 multiplied with 2.2 percent equals 1.06 percent. This would be 3.39 percent with a national income elasticity of 1.6. The true volume increase not caused by demographic changes is 2.9 percent per year. The growth that we are left with these different income elasticity is respectively 1.84 percent per year and -0.4917 percent per year. In this case we assume that the historical increase attributed to income and the residual variable remains the same in the future. The different income elasticities in this case mean that the ratio between the income effect and residual variable effect changes. When income
.0 8 .1 .1 2 .1 4 .1 6 .1 8 2000 2020 2040 2060 Year Scenario1 Scenario 3
19
elasticity is below 1, this means that the effect of the residual variable should be larger to keep the historic 2.9 percent intact. When income elasticity is above 1, the residual variable is smaller because a larger part of the 2.9 percent is dedicated to the income effect. The result for these three different national income elasticity is plotted in the graph below, with the residual added according to method two.Graph 8: Healthcare costs as % of GDP for the Netherlands (method two)
The effects of different national income elasticities are evident. An income elasticity of 0.5 gives healthcare costs exponential growth, the elasticity of 1.6 will decrease the healthcare costs, while an income elasticity of one show more of steady growth. The effect of the residual variable plays an important role in healthcare costs, as can be seen in the graph. In the 1.6 income elasticity scenario, policy and technological change have a lowering effect on the costs. Healthcare costs peak in 2040 and decline after this. A possible explanation is that technological change makes healthcare more efficient, maybe through the use of robots for example. The opposite is the case in the situation with 0.5 income elasticity, technological change and policy makes healthcare more expensive. This scenario seems to be the most supported by researchers. Because these
technological changes are more expensive, lead to more treatments and more pharmaceutical drugs. Only for this prognosis is the difference in using method one or method two has big effect. With an income elasticity of 0.5 the estimated costs of healthcare in 2059 according to method one is 32.98 percent of GDP and for method two even 37.28 percent of GDP, a difference of 4.29 percentage point.
Again these findings need to be interpreted with a cautionary note. The value of the
national income elasticities on healthcare vary a great deal in empirical literature. A lot can change in the time frame of this forecast, technological progress could stagger and policy can change at short notice. The main conclusion is that national income elasticity and the residual variable have a major effect on healthcare costs, much more significant than the demographic changes.
.1 .2 .3 .4 2000 2020 2040 2060 Year Scenario 3 M2 Scenario 3 IE 0.5 M2 Scenario 3 IE 1.6 M2
20
Discussion
The increasing healthcare costs is a problem of modern times in all developed countries and much has been written about it. Dutch researchers are also interested in this topic, but most research is done for the medium run. This paper attempts to give some insights in the long run trend for healthcare costs in the Netherlands. Many papers focus on the demographic side of this problem, but not all papers distinguish the difference between healthcare costs of survivors and decedents.
Academic literature mentions that declining disability rates among elderly are declining, and that this has an effect on the long term care volume. While this is trend is recognized, the possible impact it could have on the healthcare costs has not really been researched. In this paper
calculations are done with available data to convert this trend to real numbers, to get an impression of the impact of the declining disability rate on the healthcare costs.
Furthermore, the research agrees with the academic literature on the relevance of the residual variable, technological change and policy. This factor has a very large effect on the
healthcare costs, more than the future demographic changes. This is very clear in scenario three of this research.
The most obvious limitations of this research is that it is a long term forecast. Long term forecasting is very difficult, because a lot can change in this large period of time. Getzen (2000) says on long run forecasting ‘economic logic and historical trends establish some limits but the range of uncertainty is still wide’. Historical trends can be observed, the population can be
estimated quite accurately by CBS but a lot of data have a high level of uncertainty when using for long term forecasting. In scenario one and two of this paper we assume that healthcare costs will remain the same part of GDP for every year to come, this is most likely not the case. But this assumption needs to be made to give some insights in the future trends.
For scenario two the trend of long term care recipients is based on nine years of data, and then projected for the next 45 years. Long term care is assumed to decrease with 12.5 percentage point in this time period, to only 8 percent of elderly receiving long term care in 2059. How safe is this assumption, is 8 percent possible in the real world? Still this assumption is necessary to study the interesting trend that is provided by historical data.
The last limitations is the uncertainty of the residual variable and the national income elasticity. Academic literature have very different views on the value of the national income elasticity on healthcare, these different values result in estimations that sometimes can be in contradiction with each other. More research is needed on income elasticity for health care. How it changes over time, and what future values might be. Just like this paper, academic literature has not a method for measuring technological change and policy change. Nonetheless technological change and policy is included in this research through the residual variable, which represents these factors among other things. Therefore there are no hard numbers on the effect of
21
increasing effect in the past. This paper assumes that this trend continues in the future, especially technological change caused this increase in the past. But it might be possible that technological change will have a lowering effect on the healthcare costs in the future, while the income elasticity stays one. Policy makers could make a positive impact by promoting research into cheaper alternatives for existing treatments and medicines.The costs in this research are based on the research by Polder et al., which could have faults in their data due to their sample group. Further research might want to take a larger sample, to get data that represents the Dutch population more accurately.
22
Conclusion
In this paper long term forecasts on the Dutch healthcare costs are done for the time period 2018 till 2059, with the help of historic trends for the time period 1999-2017. A lot of different factors are taken into account, to get the most realistic view on this topic. In the first scenario demographic changes were the main factor, and the difference in costs for normal average costs and costs for the last year of life. Scenario two tried to incorporate the effect of decreasing disability rates among elderly, through looking at the trends in long term care. The last scenario was very broad, and mostly based on loose assumptions. This scenario tried to give an insight in the effects of changes in income, technology and policy.
The most optimistic findings could be found in scenario three, the situation where national income elasticity on healthcare equaled 1.6. It showed a peak of the healthcare costs in 2040 at 9.67 percent of GDP (method two: 9.97 percent). In scenario three were also the most pessimistic findings, an exponential growth of the healthcare costs when national income elasticity would equal 0.5. Were in the worst case scenario costs would rise up to 37.28 percent of GDP in 2059, A difference of 27.61 percentage point with the most optimistic peak. With both extremes in the same scenario we can conclude that national income elasticity on healthcare, technological change and policy have the biggest effect on healthcare costs. And what is the case, precisely these factors are hardest to predict and therefore also to predict right projection of the healthcare costs.
The differences in disability rate, long term care, declining costs for the last year of life for older people only have a minimum effect on healthcare costs compared to the factors in scenario three. Still the difference in the peak in scenario one and scenario two is very interesting. The peak in scenario two is five year earlier and around one percentage point lower. If the assumed trend in long term care can withstand in the real world, it is definitely a factor worth mentioning.
In conclusion, the scenarios in this paper give very different results for healthcare costs in the future on the speed of increase and the year when the costs peak. But we can be sure that healthcare costs will increase in the coming years. The most obvious question after this paper is; can we effort all these scenarios? Or does the government need to implement new policies, to obstruct these trends. More papers need to be conducted to see what options there are.
23
Bibliography
Cutler, D. M., & Sheiner, L. (1998). Demographics and medical care spending: standard
and non-standard effects (No. w6866). National bureau of economic research.
De Hollander, A. E. M., Hoeymans, N., Melse, J. M., Van Oers, J. A. M., & Polder, J. J. (2006). Zorg voor gezondheid-Volksgezondheid Toekomst Verkenning 2006.
De Jong, J. (2011). Decompositie van de zorguitgaven, 1972–2010. Centraal Planbureau
Achtergronddocument bij CPB Policy Brief, 11, 2012.
De Wit, T. (2018, January 21). Crisis Britse gezondheidzorg: ‘gaat tot doden leiden, hopelijk ben ik het niet’. Consulted at January 26, from
https://nos.nl/artikel/2213098-crisis-britse-gezondheidszorg-gaat-tot-doden-leiden-hopelijk-ben-ik-het-niet.html
Europea, C. (2015). The 2015 Ageing Report: Economic and budgetary projections for the 28 EU Member States (2013-2060). Directorate-General for Economic and Financial Affairs,
Economic Policy Committee (AWG),“European Economy, 3, 2015.
Getzen, T. (2000). Forecasting health expenditures: short, medium and long (long) term.
Journal of Health Care Finance, Vol. 26, No. 3, pp.56-72, Spring 2000
Getzen, T.E., 2000, Health care is an individual necessity and a national luxury: applying multilevel decision models to the analysis of health care expenditures, Journal of Health Economics, Vol. 19, pag. 259-270.
Kommer, G. J., Wong, A., & Slobbe, L. C. J. (2010). Determinanten van de volumegroei in de zorg. RIVM.
Ligthart, 2007, Determinanten van de gezondheidszorguitgaven, CPB Memorandum 186.
Lipszyc, B., Sail, E., & Xavier, A. (2012). Long-term care: need, use and expenditure in the
EU-27 (No. 469). Directorate General Economic and Financial Affairs (DG ECFIN), European
Commission.
Peeters, J. M., & Francke, A. L. (2007). Indicatiestelling voor AWBZ-zorg, sector Verpleging, Verzorging en Thuiszorg. Ontwikkelingen, knelpunten en oplossingsrichtingen. Utrecht:
24
Polder, J. J., Barendregt, J. J., & van Oers, H. (2006). Health care costs in the last year of life—the Dutch experience. Social science & medicine, 63(7), 1720-1731.
Puts, M. T. E., Deeg, D. J. H., Hoeymans, N., Nusselder, W. J., & Schellevis, F. G. (2008).
Changes in the prevalence of chronic disease and the association with disability in the older Dutch population between 1987 and 2001. Age and ageing, 37(2), 187-193.
Slobbe, L. C. J., Smit, J. M., Groen, J., Poos, M. J. J. C., & Kommer, G. J. (2011). Kosten van
ziekten in Nederland 2007: Trends in de Nederlandse zorguitgaven 1999-2010. RIVM
rapport 270751023.
Van der Horst, A., Bettendorf, L., Draper, N., van Ewijk, C., De Mooij, R., & Ter Rele, H. (2010). Vergrijzing verdeeld. CPB special publication, 86.
Van der Horst, A., van Erp, F., & De Jong, J. (2011). Zorg blijft groeien: financiering onder druk. The Hague: Centraal Planbureau.
van Duin, C., & Stoeldraijer, L. (2012). Bevolkingsprognose 2012-2060: langer leven, langer werken. Den Haag/Heerlen: CBS.
van Striem, F., & Bhageloe-Datadin, R. (2015). Ontwikkeling en de financiering van de zorglosten sinds 2006. Den Haag/Heerlen: CBS
Wouterse,B., ter Rele, H. & van Vuuren, D. 2016, Financiering van de zorg op de lange termijn, CPB Achtergronddocument, CPB, Den Haag.
25
Appendix
Table 1 Year peak population Year peak decedents Lowest level populationYear peak healthcare costs
0-45 2002 and 2035 unknown 2017 and 2051 2002
45-55 2016 2000 2030 2016 55-65 2026 2008 2040 2026 65-70 2032 2015 2045 2031/2032 70-75 2041 2020 2050 2041 75-80 2046 2025 2055 2046 80-85 2052 2044 Unknown 2051 85-90 2057 2050 Unknown 2057 Table 2
Age Healthcare costs survivors Healthcare costs decedents 0-45 615 16268 45-55 1050 16856 55-65 1521 16378 65-70 2208 16800 70-75 2816 17615 75-80 3705 16647 80-85 4456 14981 85-90 4841 13614 90-95 4956 12020 95+ 4590 9447
