The health and long-term care costs in the last year of life in The Netherlands

Abstract

Knowing the determinants of rising health and long-term care costs is crucial to support cost containment policies and to predict future expenditures. According to the “red herring” debate, not ageing per se, but proximity to death is the most important determinant of future expenditures. This study aims to update and expand the existing Dutch literature after two major reforms in health and long-term care. Insurance claims data from 2018–2019 of 13,738,193 insured individuals were included. Using negative binomial regression analyses, the association between deceased individuals and survivors on total health and long-term care costs was investigated, as well as per health care sector. Costs rose sharply in the two months prior to death. Regression models showed an association with total health and long-term care costs of 10.8 for deceased individuals compared with survivors (crude model) and 3.3 (adjusted model). Especially including age and chronic diseases decreased the association. The largest differences in costs between deceased individuals and survivors in the adjusted model were found for geriatric rehabilitation care and primary care stays (16.7), home nursing (10,6), and long-term care (9.3). Not just the costs of deceased individuals are important for health care costs, but also age, as measured by being in the highest age category, and chronic diseases. The costs of deceased individuals were heterogeneous across health care sectors.

Supplementary Information

The online version contains supplementary material available at 10.1007/s10198-025-01763-w.

Introduction

Health and long-term care costs are rising in the vast majority of Organisation for Economic Co-operation and Development (OECD) countries. For example, in the Netherlands, the proportion of the gross domestic product spent on health care increased by almost one percentage point in twelve years: from 9.1% in 2007 to 10.1% in 2019 [1]. The Netherlands is a relatively high spender on long-term care. In 2005, this was four percent of GDP, and this rose to 4.9% of GDP in 2018 [2, 3]. Knowledge of the determinants of rising health and long-term care costs is crucial to support cost containment policies and to predict future health care expenditures.

There is a substantial body of health economics literature on the determinants of health and long-term care expenditures. Part of this literature is labelled the “red herring” debate, which was initiated by Zweifel et al. [4]. Breyer and Lorenz summarize the “red herring” debate in health economics as follows: “One of the most important controversies in the health economics discourse of the last twenty years concerns the question of whether the imminent ageing of the population in most OECD countries will place an additional burden on the taxpayers who finance public health care systems” [5]. Zweifel et al. argued that not ageing per se, but time to death, is the main determinant of health care expenditures [4, 6]. If Zweifel et al. were correct, and proximity to death instead of ageing was the main determinant of health care expenditures, “population ageing due to rising longevity would simply shift the high death-related costs to higher ages” [5].

Worldwide, several studies have investigated costs in the last year of life. A full discussion is beyond the scope of this paper. Zweifel et al. showed that “age has no effect, provided the individual was 65 + at the time of death,” using two relatively small datasets from Switzerland [4]. Breyer and Felder used claims data from a Swiss sickness fund from 1999 and combined this with data from the Statistical Office until 2050 to predict health care expenditures [7]. Their study showed that after correcting for the costs in the last year of life, a considerable part (60%) of the estimated demographic increase in health expenditures until 2050 remained. Yang et al. (2003) showed higher health care costs for deceased individuals in the year prior to death compared to survivors [8]. Shmueli et al. found that people in their last year of life spent twenty times more on health care than people who did not live in their last year of life, and that those differences might occur due to nursing costs [9]. Weaver et al. used survey data from the Health and Retirement Study (1993–2002) and concluded that “proximity to death significantly increased the probability of nursing home use by 50.0% and of formal home care use by 12.4%” [10].

It has been suggested that proximity to death might be a “red herring” itself for morbidity. Higher morbidity and disability in individuals in their last years of life may at least partly explain the increased health and long-term care costs in the last years of life. For example, Carreras et al. showed that including morbidity reduced the association between the last years of life and health care use considerably, for among other acute outpatient care, acute inpatient care, primary care, and pharmacy [11]. Howden and Rice showed that the impact of proximity to death on hospital care costs diminished when including morbidity in the model [12]. De Meijer et al. showed that after controlling for disability, time to death was no longer important in long-term expenditure [13]. These studies show the importance of taking morbidity and/or disability into account.

Increased health and long-term care costs for deceased individuals compared with survivors vary between health care expenditure components of the health care sector [14, 15]. Knowledge of the cost of the last years of life for various health care sectors is crucial for health care planning.

Dutch literature

As the current study is focused on the Netherlands, the Dutch literature is described in more detail. Stooker et al. and Polder et al. used claims data from the 1990s and showed, respectively, five and 13.5 times higher health and long-term care costs for deceased individuals compared with survivors [16, 17]. Using the same data as Polder et al., De Kok et al. showed high average health and long-term care costs for individuals dying from cancer compared with those dying from other diseases, and the mean health care costs by age group decreased [18]. Using 2009–2011 claims data, Bakx et al. showed that 9.4% of the total expenditures were for health and long-term care costs in the last year of life: 6.7% for medical care expenditures and 14.6% for long-term care expenditures [19]. Pot et al. used survey data and focused on utilization. Using regression analysis, they showed more health care utilization in the end-of-life group compared with survivors for contact with medical specialists, hospital care, informal personal care, professional home care, and institutional care [20]. After adjusting for, among others chronic diseases and functional limitations, the association between the last year of life and health care utilization diminished. Wong et al. focused on disease-specific hospital expenditures, using longitudinal data from the Dutch Hospital Discharge Register for the period 1995–2004 [21]. They found that proximity to death was important for the majority of diseases, and the influence of age seems, although statistically significant, modest compared with proximity to death. Using macro data, including mortality rates and age- and gender-specific per capita health expenditures for the years 1981–2007, Van Baal and Wong found that including time to death did not lower the forecasts of health care expenditures [22]. De Meijer et al. focused on long-term care expenditures, using a relatively small sample of survey data from 2004 [13]. Their main finding was that after controlling for disability, time to death was no longer important, i.e., it becomes a “red herring” in itself. In other words, time to death largely acted as a proxy for disability. Age and informal care availability remained important determinants of long-term care expenditures. Although costs in the end-of-life period have been studied extensively in the Netherlands, these studies were performed before two major reforms in health and long-term care, which may have affected the associations between the last years of life and health and long-term care costs. In addition, the potential heterogeneous effects of the last years of life have not been studied fully.

Using claims data from a large representative sample of the Dutch population, the objective of this study was to analyze the association between the last year of life and health and long-term care costs in the Netherlands after two major reforms, for total health and long-term care costs and for several health care sectors, adjusting for age, gender, neighborhood socioeconomic status, and chronic diseases.

Institutional background

As our study is carried out in the Netherlands, we present some key features of the Dutch institutional context. The Dutch health care system is mainly financed by two schemes: one for curative care (Health Insurance Act) and one for long-term care (Long-Term Care Act), for which both major reforms were introduced. In 2006, the Dutch government introduced a system of managed competition in curative care, which aimed to promote efficiency, quality, acceptable societal costs (equity), reduce central governance, and improve access [23, 24]. In this system of managed competition, every citizen is compulsorily insured (social insurance) to avoid adverse selection. Health insurers receive compensation for predicted health care expenditures from a risk-equalization fund. The Dutch health insurers negotiate on behalf of their insured clients with health care providers about quality and costs. The basic benefit package includes, among others, care provided by general practitioners (GPs), hospital care, home nursing (since 2015), mental health care (since 2015, age ≥ 18 years), medical devices, maternity care, and pharmaceutical care [25]. A mandatory deductible is in place for the first €385 (in 2019), except for GP consultations, maternity care, home nursing care, and care for children under the age of 18.

In 2015, the Dutch government introduced a new Long-Term Care Act, which basically replaced the old one (the AWBZ, introduced in 1968) [2]. Cost containment was one of the aims of the reform. It is a social insurance scheme financed from income-dependent taxes. Only long-term care for persons who are permanently dependent on 24-h care or need permanent supervision is regulated by the Long-Term Care Act. From 2015 onward, other care previously covered by the AWBZ was transferred to the Social Support Act (Wmo; domestic care, social support), the Health Insurance Act (mainly home nursing), and the Youth Act (assistance and care for young people and their families coping with parenting and developmental issues, psychological problems, and disorders). The Long-Term Care Act reimburses stays in institutions like nursing homes, personal care (e.g., support with washing and dressing), daytime activities, additional care for people living in institutions (e.g., special clothes or medications), treatment, mobility devices, and transport [26]. Somewhat implicit in the reform is the assumption that people should first rely on informal care before they can access formal care. Both laws involved expectations with respect to cost containment, i.e., they would lower costs. In the case of managed competition, price competition between providers is encouraged due to the central role health insurers play in commissioning and contracting care. In the case of the Long-Term Care Act, the expectation is that people will rely on informal care and live at home as long as possible to avoid more expensive institutional care. Overall, we expected that the relative cost of the last year of life would decrease after these reforms and shift from long-term care to curative care.

Methods

Data and study population

This study used claims data from Dutch health insurers. The data were provided by the health insurers via the Centre for Information of Dutch Health Insurers, Vektis. The data were pseudonymized and contained all claims from all health insurers of all insured inhabitants in the Netherlands, i.e., the complete population, with the exception of 11,000 conscientious objectors to insurance [26]. For this study, data from 2018–2019 were used. Data on domestic care, social support (WMO), and the Youth Act were not available, and this accounted for about 9% of the governmental budget on health care expenditure [27].

The initial dataset included information on 17,602,681 insured Dutch inhabitants. Data from health care insurers with complete claims data for all sectors for 2018/2019 were included. People were excluded if gender, age, or neighborhood SES (socioeconomic status) data were missing. The final study population consisted of 13,738,193 individuals: 128,129 deceased and 13,610,064 survivors. This represented 77.3% of the Dutch insured population. Reasons for exclusion included incomplete claims data (21.2%), missing gender (< 0.01%), and/or missing neighborhood SES (3.7%). The final study population had a similar age-gender composition as the total Dutch population.

Variable of interest: deceased and survivors

Deceased individuals were those who died in 2019. We had information on the exact date of death. Survivors were individuals who were insured from July 2018 to July 2019 and did not die in 2020.

Dependent variable

Health and long-term care costs included the sum of yearly costs that were reimbursed for the following sectors:

1. General practitioner care: care provided in general practice and out-of-hours primary care.

2. Pharmaceutical care: extramural pharmaceutical care, excluding intramural pharmaceutical care.

3. Home nursing: nursing at home for people without an indication for long-term care.

4. Dental care: oral care for individuals under 18 years of age and dentures.

5. Obstetric care: obstetric care outside the hospital.

6. Medical specialist care: ambulatory and inpatient hospital care, including intensive care, emergency care, pharmaceutical care for inpatients, dental surgery, (primary care) diagnostics, and obstetric care in hospital.

7. Allied health care: physiotherapy, exercise, occupational and speech therapy, and dietary advice.

8. Medical aids and devices: including incontinence supplies, prostheses, diabetes aids, and auditory aids.

9. Mental health care: ambulatory and inpatient mental health care for inhabitants aged 18 years or older, long-term care up to 3 years.

10. Patient transport: transport by ambulance.

11. Geriatric rehabilitation care and primary care stay.

12. Postpartum care: postpartum care at home.

13. Other health care costs.

14. Long-term care: long-term (mental) care for people needing 24-h care or permanent observation.

Health care costs did not include patient transport by public transport, taxi, or private care, cross-border care, quality funds, and parts of other costs. Long-term care costs did not include medical aids and devices or dental care provided for patients in long-term care, nor were co-payments accounted for. 99.1% of the total health care provided under the Health Insurance Act and 96.3% of the total long-term care costs were included. Health and long-term care costs for deceased individuals were defined as the costs incurred in the year up to the date of death. For survivors, the costs were defined as those between July 1st, 2018 and July 1st, 2019. Costs for each sector were displayed for twelve time periods (t1–t12). The last five time periods consisted of 31 days, and the first seven time periods consisted of 30 days.

Potential confounders

Gender, age, neighborhood SES, and chronic diseases were included as potential confounders. Age was divided into ten groups, each covering a ten-year span. Neighborhood SES was measured using SES-WOA scores, which indicate the status of a neighborhood in comparison to other neighborhoods [28]. The score was derived from several characteristics of inhabitants: education, income, and position on the labour market, and was divided into quartiles for analysis (with a higher quartile indicating a higher SES). Chronic diseases were based on claims data from the year before death or between July 1st, 2018 and July 1st, 2019, and were included as binary variables (yes/no) for each condition (see Table 1). For people residing in institutional care (i.e., claiming long-term care) and receiving treatment from the same institution, general health care, including GP care and pharmaceuticals, is covered by the long-term care budget. Therefore, these individuals do not have claims for GP care and pharmaceutical care, which may result in a less accurate determination of chronic diseases.

Table 1.

Characteristics for all subjects

Total	n	%	Deceased	n	%
N	13,738,193	100
Survivors 2019	13,610,064	99.1	Deceased 2019	128,129	0.9
Age group
0–9 years	1,375,880	10.1		193	0.2
10–19 years	1,572,937	11.6		227	0.2
20–29 years	1,678,945	12.3		553	0.4
30–39 years	1,702,927	12.5		879	0.7
40–49 years	1,775,234	13.0		2270	1.8
50–59 years	2,009,293	14.8		7014	5.5
60–69 years	1,693,842	12.4		15,947	12.4
70–79 years	1,227,232	9.0		30,417	23.7
80–89 years	500,717	3.7		44,929	35.1
=  > 90 years	73,057	0.5		25,700	20.1
Gender
Males	6,716,977	49.4		62,265	48.6
Females	6,893,087	50.6		65,864	51.4
Neighbourhod SES
Quartile 1 (lowest)	3,273,349	24.1		35,332	27.6
Quartile 2	3,230,689	23.7		33,825	26.4
Quartile 3	3,463,810	25.5		32,365	25.3
Quartile 4 (highest)	3,642,216	26.8		26,607	20.8
Chronic diseases
Cancer	668,620	4.9		44,545	34.8
HIV/AIDS	20,085	0.1		252	0.2
Diabetis mellitis type 1	84,339	0.6		4285	3.3
Diabetis mellitis type 2	694,970	5.1		23,398	18.3
Cystic fibrosis	1326	< 0.1		12	< 0.0
Thyroid diseases	410,764	3.0		8707	6.8
Schizophrenic disorders	85,369	0.6		1446	1.1
Mood or anxiety disorders	470,504	3.5		6669	5.2
Personality disorders	51,742	0.4		248	0.2
ADHD	176,358	1.3		671	0.5
Parkinson’s disease	36,842	0.3		2999	2.3
Epilepsy	88,723	0.7		4718	3.7
Migraine	236,355	1.7		1129	0.9
Multiple scleroses	20,231	0.1		164	0.1
Chronic eye conditions	464,453	3.4		14,744	11.5
Hearing problems	542,904	4.0		22,734	17.7
Acute coronary syndrome	253,539	1.9		9237	7.2
Angina pectoris	144,612	1.1		10,603	8.3
Heart failure	115,807	0.9		16,691	13.0
Peripheral vasculair diseases	48,386	0.4		4289	3.3
Stroke	73,020	0.5		9793	7.6
Pulmonary hypertension	1995	< 0.0		254	0.2
Heart valve disorders	77,787	0.6		4272	3.3
Chronic venous insufficiency	67,552	0.5		1261	1.0
COPD or asthma	889,904	6.5		27,172	21.2
Crohn’s disease or ulcerative colitis	85,476	0.6		1136	0.9
Liver diseases	16,049	0.1		1363	1.1
Chronic skin disorders	931,167	6.8		15,094	11.8
Severe acne	132,213	1.0		306	0.2
Chronic inflammatory joint diseases	262,899	1.9		7156	5.6
Peripheral osteoarthritis	300,690	2.2		4645	3.6
Neck or back disorders	461,740	3.4		6313	4.9
Osteoporosis	229,991	1.7		10,223	8.0
Chronic schoulder problems	57,813	0.4		660	0.5
Kidney diseases	88,379	0.6		9388	7.3
Dementia	61,804	0.5		28,445	22.2
Intellectual Disablility	102,897	0.8		1577	1.2
	668,620	4.9		44,545	34.8

Statistical analyses

Retrospective analyses

We estimated negative binomial fixed effects regressions with conditional mean health and long-term care costs:

$$C_i=\text{exp}(\alpha {·d}_i+x_i\prime \beta +{\epsilon }_i)$$

where $C_i$ is the mean conditional health and long-term care costs for individual $i$, $d_i$ is an indicator whether individual $i$ has deceased and $\alpha$ as its associated coefficient. $x_i$ represents potential confounders of individual $i$ which include gender, age categories, neighborhood SES and chronic diseases with associated coefficients $\beta$ for these confounders. When estimating the crude and intermediate models, the specific $\beta$ was assumed to equal 0. ${\epsilon }_i$ is a random unobserved cost shock where $\text{exp}({\epsilon }_i)$ is assumed to be distributed according to a gamma distribution. Both total health and long-term care costs, as well as costs for different sectors, were analyzed, excluding dental care, obstetric care, postpartum care, and other health care costs. We chose a general linear model with log-link and negative binomial distribution with conditional mean health care costs to account for the right skewness. No two-stage model was chosen, as almost all Dutch citizens have health and long-term care costs—most citizens are registered with a general practice for which a capitation fee is in operation or receive long-term care. For comparability, costs per sector were also analysed with negative binomial regression analyses. Six different models were conducted: (i) crude model with only$d_i$, (ii) crude model with only age (iii) model with $d_i$ adjusted for age, (iv) model with $d_i$ adjusted for age and gender, (v) model with $d_i$ adjusted for age, gender and neighborhood SES, (vi) model with $d_i$ adjusted for age, gender, neighborhood SES and chronic diseases. Using Bayesian information criterion (BIC), the different models were compared on their goodness of fit.

Results

Forty-nine percent (49.4%) of the total population were men. For the deceased, the percentage of men was 48.6% (Table 1). Almost a quarter of the deceased were aged 70–79 years, over a third were aged 80–89 years, and a fifth were 90 years or older. Cancer was the most common chronic condition among the deceased (34.8%), followed by dementia (22.2%) and COPD/asthma (21.2%). For survivors, the most common chronic diseases were chronic skin disorders (6.8%), COPD/asthma (6.5%), and diabetes mellitus type 2 (5.1%). Chronic diseases were more common in the deceased than in the survivors, with the exception of cystic fibrosis, personality disorders, ADHD, migraine, acne, and multiple sclerosis.

Health and long-term care costs

The average yearly costs for the survivor group were 3,317 Euro (SD = 12,200), while for the deceased, it was 50,423 Euro (SD = 39,051). The most expensive survivor had a health and long-term care expenditure of 2,324,284 Euro, and for the deceased, this was 708,485 Euro. Males had lower health care costs than females in both the survivor and deceased groups. For survivors, health and long-term care expenditure increased with age. For the deceased, health and long-term care costs were highest in the youngest age categories. The ratio between deceased and survivors decreased with age: for the 0–9 years group, the ratio was 88.5, compared to the ≥ 90 years group, where the ratio was 2.4 (Online Appendix I).

Of all health and long-term care sectors, nearly half (49.0%) of the total claims for deceased individuals were for long-term care, followed by medical specialist care (28.8%), home nursing (9.2%), and pharmaceutical care (3.6%) (Fig. 1). For survivors, the distribution was different: medical specialist care accounted for 36.9% of total claims, followed by long-term care (28.4%), pharmaceutical care (8.6%), and mental health care (6.8%).

Fig. 1.

Mean yearly health and long-term care costs per health care sector for survivors and deceased in 2019

Costs for the deceased increased in all sectors in the final months (Fig. 2), except for dental care, allied health care, and mental health care (not shown), which were smaller sectors in terms of increased costs in the last year of life. Costs rose significantly in month t-2 (two months prior to death) for general practitioner care, pharmaceutical care, medical aids, medical specialist care, and long-term care, with an even sharper increase in month t-1. Overall, month t-1 was the most expensive month for the deceased, with costs rising by 50% compared to t-2 (Fig. 2).

Fig. 2.

Mean health care costs per capita in the last 12 months of life versus regular 12 months of life. Costs for A general practitioner care, B pharmaceutical care, C home nursing, D medical specialist care, E medical aids and devices, F long-term care, and G all health and long-term care costs in the Netherlands per month. For deceased event at t0 equals time of death, for survivors t0 equals July 1st 2019

Negative binomial regression analysis

Without adjusting for potential confounders, deceased individuals had 10.8 (95% CI: 10.7–10.9) times higher health and long-term care costs compared with survivors (Table 2). After adjusting for gender, age, neighborhood SES, and chronic diseases, deceased individuals had 3.3 (95% CI: 3.3–3.3) times higher health and long-term care costs compared with survivors. Adjusting for age and chronic diseases greatly impacted the association between deceased and survivors, whereas gender and neighborhood SES had minimal effect on the association. After adjusting for chronic diseases, the association between age category and health and long-term care costs decreased. Adjusting for age, gender, neighborhood SES, and chronic diseases improved the fit of the model, as shown by the lower BIC.

Table 2.

Negative binomial regression analyses between survivors or deceased in 2019 and total health and long-term care costs

	Model 1	Model 2	Model 3	Model 4	Model 5	Model 6
Intercept	8.1067	7.2341	7.2240	7.2084	7.2956	6.9380
	IRR (95% CI)	IRR (95% CI)	IRR (95% CI)	IRR (95% CI)	IRR (95% CI)	IRR (95% CI)
Deceased (reference = survivors)	10.83 (10.74–10.92)		6.15 (6.10–6.20)	6.23 (6.18–6.28)	6.21 (6.16–6.26)	3.28 (3.25–3.30)
Age group (reference = age 0–9 years)
10–19 years		1.13 (1.13–1.13)	0.82 (0.81–0.82)	0.81 (0.80–0.81)	0.81 (0.81–0.82)	0.72 (0.72–0.72)
20–29 years		1.75 (1.75–1.76)	1.12 (1.11–1.12)	1.08 (1.08–1.09)	1.07 (1.07–1.07)	0.83 (0.83–0.83)
30–39 years		1.86 (1.85–1.87)	1.39 (1.39–1.40)	1.34 (1.34–1.35)	1.33 (1.33–1.34)	0.99 (0.99–1.00)
40–49 years		1.80 (1.79–1.80)	1.38 (1.37–1.38)	1.35 (1.34–1.35)	1.34 (1.34–1.35)	0.78 (0.78–0.79)
50–59 years		2.34 (2.33–2.34)	1.82 (1.81–1.82)	1.79 (1.78–1.79)	1.78 (1.78–1.79)	0.87 (0.87–0.88)
60–69 years		3.28 (3.27–3.30)	2.58 (2.57–2.59)	2.55 (2.54–2.56)	2.54 (2.54–2.55)	1.02 (1.01–1.02)
70–79 years		5.00 (4.99–5.02)	3.67 (3.66–3.68)	3.62 (3.61–3.63)	3.62 (3.60–3.63)	1.18 (1.17–1.18)
80–89 years		10.53 (10.48–10.58)	5.00 (4.98–5.02)	4.86 (4.83–4.88)	4.84 (4.81–4.86)	1.53 (1.53–1.54)
=  > 90 years		24.32 (24.08–24.56)	5.55 (5.50–5.60)	5.26 (5.21–5.31)	5.23 (5.18–5.28)	2.28 (2.26–2.30)
Gender (reference = men)
Women				1.18 (1.17–1.18)	1.18 (1.17–1.18)	1.21 (1.20–1.21)
Neighborhood SES (reference = quartile 1—lowest)
Quartile 2					0.89 (0.89–0.89)	0.93 (0.92–0.93)
Quartile 3					0.84 (0.84–0.84)	0.90 (0.90–0.91)
Quartile 4 (highest)					0.79 (0.79–0.79)	0.88 (0.87–0.88)
Chronic diseases (reference = without the disease)
Cancer						3.64 (3.63–3.65)
HIV/AIDS						10.09 (9.92–10.27)
Diabetis mellitis type 1						3.70 (3.67–3.73)
Diabetis mellitis type 2						1.68 (1.67–1.68)
Cystic fibrosis						62.53 (58.50–66.85)
Thyroid diseases						1.34 (1.33–1.34)
Schizophrenic disorders						8.39 (8.32–8.46)
Mood or anxiety disorders						2.93 (2.92–2.94)
Personality disorders						6.06 (5.99–6.12)
ADHD						1.81 (1.80–1.82)
Parkinson’s disease						3.17 (3.13–3.21)
Epilepsy						3.32 (3.29–3.35)
Migraine						1.57 (1.56–1.57)
Multiple scleroses						9.33 (9.17–9.49)
Chronic eye conditions						1.70 (1.70–1.71)
Hearing problems						1.53 (1.53–1.54)
Acute coronary syndrome						2.74 (2.73–2.75)
Angina pectoris						1.80 (1.79–1.81)
Heart failure						2.20 (2.19–2.22)
Peripheral vasculair diseases						2.78 (2.75–2.81)
Stroke						4.74 (4.70–4.78)
Pulmonary hypertension						10.51 (9.98–11.07)
Heart valve disorders						2.45 (2.43–2.48)
Chronic venous insufficiency						1.81 (1.79–1.83)
COPD or asthma						1.88 (1.88–1.89)
Crohn’s disease or ulcerative colitis						3.54 (3.51–3.57)
Liver diseases						3.46 (3.39–3.52)
Chronic skin disorders						1.42 (1.41–1.42)
Severe acne						1.37 (1.36–1.38)
Chronic inflammatory joint diseases						2.52 (2.51–2.53)
Peripheral osteoarthritis						2.31 (2.30–2.32)
Neck or back disorders						2.05 (2.04–2.06)
Osteoporosis						1.63 (1.62–1.64)
Chronic schoulder problems						1.60 (1.58–1.61)
Kidney diseases						3.74 (3.71–3.77)
Dementia						1.43 (1.42–1.45)
Intellectual disability						2.16 (2.15–2.18)
BIC	233,829,995	237,606,972	232,046,006	232,002,025	231,950,467	227,104,857

When examining age specifically, the highest age group (90 years and above) had 2.3 (95% CI: 2.3–2.3) times higher health and long-term care costs compared to the lowest age category (0–9 years). This result suggests that both being in the last year of life and being in the highest age category were positively associated with costs. People with chronic diseases had higher health and long-term care costs, especially those with cystic fibrosis, pulmonary hypertension, HIV/AIDS, multiple sclerosis, and schizophrenic disorder. People living in neighborhoods with higher SES had lower health and long-term care costs.

Looking into the specific health care sectors, deceased individuals had especially higher costs compared with survivors for geriatric rehabilitation care and primary care stays, home nursing, long-term care, and patient transport (Table 3; full models in Online Appendix II; Table 4: model 1 and 6 for the health care sectors contributing to the total health and long-term care costs mostly). Although deceased individuals also had higher costs for general practitioner care, pharmaceutical care, allied health care, and mental health care compared to survivors, the association between deceased versus survivors and costs was smaller in these sectors. After adjusting for gender, neighborhood SES, and chronic diseases, higher age was particularly associated with high costs for long-term care, geriatric rehabilitation care and primary care stays, home nursing, medical aids and devices, pharmaceutical care and patient transport. Allied health care was an exception.

Table 3.

Negative binomial regression analyses between survivors or deceased in 2019 and total health and long-term care costs, by health care sector^$

	Model 1	Model 3	Model 4	Model 5	Model 6
Deceased (reference = survivors)	IRR (95% CI)	IRR (95% CI)	IRR (95% CI)	IRR (95% CI)	IRR (95% CI)
Total health and long-term care	10.83 (10.74–10.92)	6.15 (6.10–6.20)	6.23 (6.18–6.28)	6.21 (6.16–6.26)	3.28 (3.25–3.30)
General practitioner care	4.46 (4.45–4.48)	2.66 (2.66–2.67)	2.69 (2.68–2.70)	2.68 (2.67–2.69)	2.35 (2.34–2.35)
Pharmaceutical care	6.29 (6.21–6.38)	3.24 (3.19–3.28)	3.24 (3.20–3.28)	3.23 (3.19–3.27)	1.75 (1.73–1.77)
Home nursing	30.69 (27.75–33.94)	13.59 (12.36–14.95)	13.81 (12.56–15.19)	14.8 (13.46–16.27)	10.62 (9.68–11.67)
Medical specialist care	11.86 (11.66–12.06)	7.36 (7.24–7.48)	7.47 (7.34–7.59)	7.44 (7.32–7.56)	3.64 (3.58–3.70)
Allied health care	4.23 (4.02–4.45)	2.88 (2.74–3.03)	2.98 (2.84–3.14)	2.98 (2.83–3.13)	1.88 (1.79–1.98)
Medical aids and devices	10.76 (10.33–11.21)	4.79 (4.60–4.98)	4.8 (4.61–5.00)	4.79 (4.60–4.99)	3.22 (3.10–3.35)
Patient transport	27.21 (24.87–29.75)	12.46 (11.40–13.61)	12.40 (11.35–13.55)	12.30 (11.26–13.44)	8.75 (8.01–9.55)
Mental health care	1.44 (1.34–1.56)	2.63 (2.43–2.84)	2.70 (2.49–2.91)	2.71 (2.51–2.93)	2.49 (2.31–2.68)
Geriatric rehabilitation care and primary care stay	39.47 (29.97–51.97)	27.52 (21.53–35.18)	27.63 (21.60–35.34)	28.21 (22.08–36.05)	16.69 (13.18–21.14)
Long-term care	26.24 (22.52–30.57)	8.47 (7.27–9.87)	8.37 (7.19–9.76)	8.37 (7.19–9.76)	9,26 (8.08–10.61)

^$Model 2 is not included in the table since this model does not include the variable deceased

Table 4.

Negative binomial regression analyses between survivors or deceased in 2019 and long-term care costs, costs for medical specialist care and cost for home nursing

	Long-term care		Medical specialist care		Home nursing
	Model 1	Model 6	Model 1	Model 6	Model 1	Model 6
Intercept	6.8474	− 0.0558	7.1097	6.2211	5.0232	3.3732
	IRR (95% CI)	IRR (95% CI)	IRR (95% CI)	IRR (95% CI)	IRR (95% CI)	IRR (95% CI)
Deceased (reference = survivors)	26.24 (22.52–30.57)	9.26 (8.08–10.61)	11.86 (11.66–12.06)	3.64 (3.58–3.70)	30.69 (27.75–33.94)	10.62 (9.68–11.67)
Age group (reference = age 0–9 years)
10–19 years		11.89 (11.21–12.60)		0.58 (0.58–0.58)		0.37 (0.35–0.38)
20–29 years		39.93 (37.66–42.34)		0.73 (0.73–0.74)		0.27 (0.26–0.28)
30–39 years		70.20 (66.20–74.44)		0.95 (0.94–0.96)		0.28 (0.27–0.29)
40–49 years		103.91 (98.02–110.15)		0.77 (0.77–0.78)		0.46 (0.44–0.48)
50–59 years		163.64 (154.56–173.27)		0.87 (0.86–0.87)		0.75 (0.73–0.78)
60–69 years		261.94 (246.84–277.97)		0.98 (0.97–0.98)		1.26 (1.21–1.31)
70–79 years		449.61 (421.54–479.55)		1.04 (1.03–1.05)		2.91 (2.79–3.03)
80–89 years		1706.08 (1570.45–1853.43)		0.98 (0.97–0.99)		12.66 (12.00–13.36)
=  > 90 years		6589.02 (5604.21–7746.89)		0.86 (0.84–0.87)		36.52 (32.79–40.68)
Gender (reference = men)
Women		0.63 (0.62–0.65)		1.24 (1.24–1.25)		1.09 (1.07–1.11)
Neighborhood SES (reference = quartile 1—lowest)
Quartile 2		0.92 (0.89–0.96)		0.95 (0.94–0.95)		0.66 (0.64–0.68)
Quartile 3		1.06 (1.03–1.10)		0.95 (0.95–0.95)		0.59 (0.58–0.61)
Quartile 4 (highest)		0.89 (0.86–0.92)		0.93 (0.93–0.93)		0.53 (0.51–0.54)
Chronic diseases (reference = without the disease)
Cancer		1.07 (1.01–1.14)		5.85 (5.81–5.89)		2.38 (2.28–2.48)
HIV/AIDS		1.78 (1.27–2.49)		5.03 (4.83–5.23)		3.20 (2.55–4.02)
Diabetis mellitis type 1		1.24 (1.05–1.46)		3.06 (3.00–3.12)		10.22 (9.16–11.4)
Diabetis mellitis type 2		0.78 (0.74–0.84)		1.32 (1.31–1.33)		3.28 (3.15–3.41)
Cystic fibrosis		29.31 (7.87–109.18)		13.34 (11.45–15.54)		99.72 (41.12–241.84)
Thyroid diseases		0.69 (0.64–0.74)		1.39 (1.38–1.40)		1.54 (1.46–1.62)
Schizophrenic disorders		4.21 (3.57–4.96)		1.32 (1.29–1.34)		6.88 (6.16–7.69)
Mood or anxiety disorders		1.18 (1.10–1.26)		1.42 (1.40–1.43)		2.43 (2.32–2.55)
Personality disorders		0.68 (0.55–0.85)		1.51 (1.47–1.55)		1.75 (1.52–2.02)
ADHD		0.84 (0.74–0.94)		1.16 (1.14–1.17)		1.06 (0.98–1.14)
Parkinson’s disease		6.96 (5.46–8.85)		2.41 (2.34–2.48)		9.50 (8.07–11.19)
Epilepsy		20.62 (17.60–24.16)		3.71 (3.65–3.78)		10.25 (9.22–11.40)
Migraine		0.76 (0.69–0.84)		1.73 (1.71–1.75)		1.91 (1.78–2.04)
Multiple scleroses		15.32 (10.94–21.45)		6.88 (6.62–7.16)		42.63 (34.05–53.36)
Chronic eye conditions		4.71 (4.38–5.06)		1.96 (1.94–1.97)		2.43 (2.32–2.55)
Hearing problems		3.64 (3.41–3.89)		1.60 (1.59–1.61)		2.07 (1.98–2.17)
Acute coronary syndrome		0.51 (0.46–0.56)		3.78 (3.74–3.82)		1.49 (1.39–1.59)
Angina pectoris		0.70 (0.61–0.79)		2.00 (1.98–2.03)		1.69 (1.55–1.83)
Heart failure		2.26 (1.97–2.58)		2.65 (2.61–2.69)		3.31 (3.02–3.62)
Peripheral vasculair diseases		1.31 (1.06–1.61)		3.83 (3.74–3.92)		3.11 (2.70–3.59)
Stroke		8.26 (6.99–9.77)		6.62 (6.49–6.75)		7.03 (6.28–7.87)
Pulmonary hypertension		2.44 (0.89–6.72)		4.68 (4.16–5.26)		17.88 (9.09–35.19)
Heart valve disorders		0.79 (0.67–0.94)		3.66 (3.59–3.73)		1.53 (1.37–1.72)
Chronic venous insufficiency		0.69 (0.57–0.83)		2.29 (2.24–2.34)		2.47 (2.18–2.80)
COPD or asthma		1.11 (1.05–1.17)		1.84 (1.83–1.85)		2.62 (2.53–2.72)
Crohn’s disease or ulcerative colitis		0.68 (0.58–0.81)		5.28 (5.18–5.38)		2.67 (2.39–2.99)
Liver diseases		2.74 (1.90–3.96)		4.09 (3.92–4.27)		4.17 (3.26–5.33)
Chronic skin disorders		1.00 (0.95–1.05)		1.49 (1.48–1.49)		2.25 (2.17–2.33)
Severe acne		1.20 (1.05–1.38)		1.43 (1.41–1.46)		1.78 (1.63–1.95)
Chronic inflammatory joint diseases		0.88 (0.80–0.97)		3.44 (3.40–3.47)		3.44 (3.23–3.67)
Peripheral osteoarthritis		0.51 (0.46–0.56)		3.25 (3.22–3.29)		1.96 (1.85–2.08)
Neck or back disorders		4.33 (4.02–4.67)		2.59 (2.57–2.61)		6.71 (6.39–7.04)
Osteoporosis		1.18 (1.07–1.31)		1.62 (1.60–1.64)		2.63 (2.46–2.82)
Chronic schoulder problems		0.43 (0.35–0.53)		2.00 (1.95–2.04)		1.65 (1.44–1.88)
Kidney diseases		1.97 (1.69–2.30)		5.03 (4.94–5.13)		6.31 (5.69–7.00)
Dementia		50.08 (42.58–58.90)		1.18 (1.16–1.20)		3.59 (3.22–4.00)
Intellectual disability		1904.25 (1638.81–2212.68)		2.24 (2.20–2.27)		2.23 (2.01–2.46)
BIC	8,807,269	8,699,488	155,824,797	154,173,034	12,673,741	12,518,225

Results for long-term care adjusted for chronic diseases should be interpreted with caution, as chronic diseases were less accurately recorded for people residing in institutional care and receiving treatment from the same institution. In general, people living in neighborhoods with lower SES had higher health and long-term care costs. As with the total health and long-term care costs, adjusting for age and chronic diseases had a significant impact on the association between deceased versus survivors and costs. Adjusting for age had the most noticeable impact on this association for home nursing, long-term care, and patient transport. Adjusting for chronic diseases most strongly impacted the association between deceased versus survivors and costs for medical specialist care, geriatric rehabilitation care and primary care stays, and pharmaceutical care.

Discussion and conclusion

This study examined health and long-term care costs in the last year of life following two major reforms in the Netherlands, using administrative claims data from Dutch health insurers. Our findings confirm that health and long-term care costs in the final year of life matter, but age and chronic diseases as well. The association between the last year of life and health and long-term care costs was heterogeneous across different health care sectors. Deceased individuals had notably higher costs than survivors in specific sectors, particularly in geriatric rehabilitation care and primary care stays, home nursing, long-term care, and patient transport. Furthermore, higher age was strongly associated with increased costs in long-term care, geriatric rehabilitation care and primary care stays, home nursing, medical aids and devices, pharmaceutical care, and patient transport.

Average costs of deceased individuals in the Netherlands were almost 38,000 Euro in the last year of life compared with around 2,700 Euro for survivors. Particularly, the final two months of life showed increases in costs. Medical specialist care, long-term care, and home nursing contributed mostly to the high health and long-term care costs in the last year of life. For long-term care this is largely due to a more expensive intensity care package that people receive when their health declines, which include housing costs [29]. Home nursing and medical specialist care has also previously been shown to be common at the end of life [10, 21].

Overall, we expected that the relative cost of the last year of life would decrease after the reforms, shifting from long-term care to curative care. Without adjusting for age, gender, neighborhood SES, and chronic diseases, we found that deceased individuals had 10.8 times higher health and long-term care costs compared to survivors. This ratio was lower than the 13.5 times ratio reported by Polder et al. [17], but higher than the 5 times ratio reported by Stooker et al. [16]. Stooker et al. included only individuals with compulsory public health insurance, which was in place until 2006 and covered residents earning below a certain income threshold [16]. This population is expected to have higher health and long-term care costs, which may explain the lower ratio in their study. The study population of Polder et al. is more comparable to ours. However, Polder et al.'s crude analyses did not account for changes in population composition. With an aging population, the ratio is expected to decrease. Therefore, it remains uncertain whether the observed lower ratio between deceased and survivors is due to the reforms or changes in population composition.

After controlling for age, gender, neighborhood SES, and chronic diseases, deceased individuals had 3.3 times higher health and long-term care costs compared with survivors. This highlights the importance of adjusting for these factors when calculating end-of-life costs. Specifically, adjusting for age and chronic diseases had a significant impact on the association between the last year of life and health and long-term care costs. While some studies have suggested that the time of death could be a "red herring" for morbidity [12, 13], our study indicates that the last year of life does indeed affect health and long-term care costs. This suggests that the increased costs in the last year of life are not merely a "red herring," but also reflect the effects of age and morbidity.

Strengths and limitations

The primary strength of our study lies in the large and comprehensive dataset of health and long-term care claims for the Dutch population. This provided us with the opportunity to examine the costs in the last year of life at a micro level, offering great precision across various health care sectors. Additionally, we were able to control for key confounders such as gender, age, neighborhood SES, and a wide range of chronic diseases. However, there are a few limitations to our study. First, for individuals residing in institutional care who receive treatment from the same institute, the identification of chronic diseases may have been less accurate. This could have led to an overestimation of the age-effect and a potential underestimation of the effects of chronic diseases. As Wong et al. found a decreasing disease-specific deceased/survivor ratio with age, this could have influenced the differences between deceased and survivors [21]. Since only 1.1% of the Dutch population falls into this group, the impact is likely small, with the exception of long-term care. Second, our study did not include data on the Social Support Act, Youth Act, unclaimed care, or additional insurance claims. Finally, we did not have information on informal care support, which is a limitation shared by most Dutch studies, with the notable exception of De Meijer et al., whose sample size was relatively small [13].

Conclusions

Our results suggest that health and long-term care costs are influenced not only by the last year of life, but also by age and chronic diseases. This indicates that the "red herring" debate in health economics is not solely about time-to-death costs, but also involves the costs associated with aging (specifically, being in the highest age category) and the costs of morbidity. The costs in the last year of life were found to be heterogeneous across different health care sectors. These varying effects of age and the last year of life should be considered when planning for future health and long-term care needs.

Supplementary Information

Below is the link to the electronic supplementary material.

Author contributions

CD, TL, JD and TH contributed to the study conception and design. Material preparation, data collection and analysis were performed by CD, TL and JD. The first draft of the manuscript was written by BB and all authors commented on previous versions of the manuscript. All authors read and approved the final manuscript.

Funding

This research received no specific grant from any funding agency, commercial or not-for-profit sectors.

Availability of data and materials

The datasets generated and/or analyzed during the current study are not publicly available due to the General Data Protection Regulation (GDPR). According to the GDPR, the National Health Care Institute has a legal basis to process this health data. Datasets are only available for organizations with a legal basis to process this health data which is in accordance with the legal basis for which the National Health Care Institute has collected the data. Data on aggregated level are available from the corresponding author.

Declarations

Conflict of interest

The authors declare no conflict of interest.

Ethics approval and consent to participate

Not applicable. Pseudonymized (claims) data is personal data according to the GDPR, but can’t be traced back to individual citizens. The National Health Care Institute uses the pseudonymized data for its legal tasks; no objection or individual complaint can be submitted.