Background Studying where people die across countries can serve as an evidence base for health policy on end-of-life care. This study describes the place of death of people who died from diseases indicative of palliative care need in 14 countries, the association of place of death with cause of death, sociodemographic and healthcare availability characteristics in each country and the extent to which these characteristics explain country differences in the place of death.
Methods Death certificate data for all deaths in 2008 (age ≥1 year) in Belgium, Canada, the Czech Republic, England, France, Hungary, Italy, Mexico, the Netherlands, New Zealand, South Korea, Spain (Andalusia), the USA and Wales caused by cancer, heart/renal/liver failure, chronic obstructive pulmonary disease, diseases of the nervous system or HIV/AIDS were linked with national or regional healthcare statistics (N=2 220 997).
Results 13% (Canada) to 53% (Mexico) of people died at home and 25% (the Netherlands) to 85% (South Korea) died in hospital. The strength and direction of associations between home death and cause of death, sociodemographic and healthcare availability factors differed between countries. Differences between countries in home versus hospital death were only partly explained by differences in these factors.
Conclusions The large differences between countries in and beyond Europe in the place of death of people in potential need of palliative care are not entirely attributable to sociodemographic characteristics, cause of death or availability of healthcare resources, which suggests that countries’ palliative and end-of-life care policies may influence where people die.
- PALLIATIVE CARE
- DEATH CERTIFICATION
- CHRONIC DI
- PUBLIC HEALTH
Statistics from Altmetric.com
The number of people who die as a result of prolonged chronic illnesses is increasing worldwide.1 Prior to dying, these people typically experience a wide range of complex needs and symptoms that require palliative care.2 In line with the view of the WHO, palliative care is an approach to care that goes beyond a medical specialty and that aims to control the various physical, psychological, social and other problems associated with a life-threatening illness.3 One important aspect of the palliative care approach is planning of care, including the location in which patients wish to receive care towards the end of life and the location in which they wish to die.
Twenty-six studies from 13 countries found that more than half of the patients preferred to die at home. Eight of these studies were conducted in the UK, four in the USA, three in Spain, two in Japan, two in South Korea and one each in Australia, Ethiopia, Rwanda, Sweden, Taiwan, Uganda and Zimbabwe, respectively.4 Furthermore, a population-based survey in seven European countries found that between 50% and 83% would prefer to die at home if faced with advanced cancer.5 However, it appears that this wish is often not met as in many countries people mainly die in hospitals,6–8 and in certain countries the percentage of hospital deaths is rapidly increasing.8 ,9 Research evidence has raised concerns that many hospital deaths are preceded by potentially burdensome and inappropriate hospital admissions and aggressive treatments shortly before death, which could be a threat to good end-of-life care, quality of life and ultimately a good death.10–13 Enabling people to die at home also has important cost implications as end-of-life care in hospital may incur higher costs than end-of-life care in community-care settings.14 ,15 Of course, a home death may be easier to achieve in countries that have developed adequate home care arrangements and may be different in certain middle-income or low-income countries where people may be more likely to receive appropriate care only in hospitals, for instance because formal home care is not available or not affordable (eg, in Mexico).16
Where people with palliative care needs die has been determined to be an issue for public health,17 and it should therefore be subject to the first function of public health: assessment and monitoring.18 ,19 Assessment and monitoring provide the scientific foundation that is needed to inform policies and interventions on a subject that has been determined to be relevant for public health. Continuous and systematic monitoring of the place of death in a population that died from causes indicative of palliative care need is therefore an essential cornerstone in the planning, implementation and evaluation of policies and programmes aimed at enabling people with chronic diseases to die in their place of choice, and it contributes to the scientific base that is necessary for the allocation of financial, material and human resources for palliative care. In this context, cross-national comparisons have the added value of providing points of comparison, and they can be a valuable source of hypotheses about alternative ways to provide end-of-life care by comparing where people die in countries with different strategies and traditions in end-of-life care.20 ,21 Yet relatively few cross-national comparisons of place of death have been conducted on the population level,22–25 and few population-level studies have investigated the place of death of people who died from diseases indicative of palliative care need.25 ,26
The aim of this study was to describe and compare the place of death of people who died from diseases indicative of palliative care need in 14 countries across four continents; to what extent the place of death is associated with sociodemographic characteristics, cause of death and availability of healthcare resources in each of the countries studied; and to what extent differences in these characteristics explain country-differences in the place of death.
This study is part of the International Place of Death study, which collected complete death certificate data over 1 year in 14 countries (table 1) that are situated across different levels of palliative care development.27 We obtained full country data for all countries, except for Spain, where the data are from the Andalusia province and Canada, where the data do not include the Quebec province. At the time of data collection (2011–2013), 2008 was the most recent year for which data were available in most targeted countries. In the USA, 2007 was the most recent available year, and in Andalusia (Spain) place of death was not recorded before 2010. In addition to place of death data, we obtained a limited number of clinical, sociodemographic and healthcare supply characteristics that correspond to factors identified as being associated with the place of death in people with cancer.28 The project lead coordinated all data requests to ensure similar variables and data and pooled all data into one common database. Ethics approval was not required as we studied anonymised death certificate data.
The population comprised all deaths of persons aged 1 year and over from 10 underlying conditions that previous research has found to be indicators for identifying a population potentially in need of palliative care, whether delivered through a specialist palliative care service or non-specialised care providers.29 The development of this criterion is described in online supplementary file 3. These 10 underlying conditions make up the criterion for the minimal (ie, most conservative) estimate of the population potentially in need of palliative care and include deaths caused by cancer (International Classification of Diseases 10 (ICD-10) codes C00-D48), heart failure (I500, I501, I509, I111, I130, I132), renal failure (N102, N112, N120, N131, N132, N180, N188, N189), liver failure (K704, K711, K721, K729), chronic obstructive pulmonary disease (J40, J42, J410, J411, J418, J430-J432, J438-J441, J448, J449), diseases of the nervous system (Alzheimer's disease (G300, G301, G308, G309), Parkinson's disease (G20), motor neurone disease (G122), Huntington's disease (G10)) or HIV/AIDS (B20-B24). We excluded neonatal deaths (deaths of persons younger than 1 year; 0.03% of the palliative care population subset) because the approach to and resources needed in palliative care in this population are very different from that of older children and adults.30
All included countries have a similar certification of deaths. The attending physician certifies the sex of the deceased, the cause of death and time and place of death. The authorities in each country check these certificates for inconsistencies and code the causes of death according to the ICD-10. In some countries, sociodemographic and environment-related variables are recorded on the death certificates; in other countries, they can be gained by linking the death certificates with other databases (eg, census data). Additionally, we linked available healthcare resource statistics per capita and data on the degree of urbanisation with the deceased's municipality/local authority of residence. Online supplementary table S1 provides additional information on these data.
The place of death was derived from the death certificate and comprised in all countries at least the categories ‘home’, ‘hospital’, ‘long-term care facility’ and ‘other’, except for Hungary, where only ‘hospital’ and ‘other’ were recorded and Mexico, where ‘long-term care facility’ was not recorded. Death certificates in England, Wales, New Zealand and the USA additionally recorded the category ‘palliative care institution’.
Independent variables included individual sociodemographic characteristics (age, sex, marital status), underlying cause of death, degree of urbanisation of the region of residence and national or regional measures of healthcare availability per region of residence (number of hospital beds/1000 inhabitants; number of long-term care beds/1000 inhabitants aged 65 years or over; number of general practitioners (GPs)/10 000 inhabitants). For Hungary, the Czech Republic and South Korea, we used nationally aggregated measures of healthcare availability because the individual's region of residence was not identified.
We calculated the percentage of deaths in the different locations in each country. As our data are population-level data, it was not necessary to calculate p values. For each country, a multivariable binary logistic regression analysis was conducted to determine the association between dying at home versus in a hospital and the cause of death (cancer vs non-cancer condition), age, sex, marital status, degree of urbanisation and per capita availability of hospital beds, long-term care beds and GPs per region of residence. For the countries in which palliative care institution was recorded as a place of death (England, Wales, New Zealand, the USA), we computed an additional multivariable binary logistic regression analysis per country to determine the association between death in a hospice versus in a hospital and the cause of death, age, sex, marital status and density of hospital beds and GPs per region of residence. The independent variables were entered simultaneously into the regression model. We included the independent variables that previous literature has shown to be associated with place of death28 and that could be obtained through death certificates.
To estimate the extent to which country differences in the place of death were explained by differences in healthcare availability, cause of death and sociodemographic characteristics, we conducted a hierarchical binary logistic regression analysis with the dependent variable death at home versus in hospital. Country was entered as the independent variable in the first step (model 1), cause of death, age, sex and marital status in the second step (model 2), and the density of hospital beds, long-term care beds and GPs per region of residence in the third step (model 3). Belgium was chosen as the reference country because it was in the middle in terms of the percentage of home and hospital deaths compared to the other countries. All models were checked for multicollinearity (tolerance values and variance inflation factors). All analyses were conducted in IBM SPSS Statistics V.22.
Between 25% of people in Mexico and 49% in the Netherlands died from diseases indicative of palliative care need (table 1). For 52% the USA and 83% (the Czech Republic) of these deaths, cancer was the underlying cause. The density of hospital beds, beds in long-term care settings and GPs was highest in South Korea, the Netherlands and Belgium, respectively (see online supplementary table S2).
Place of death per country in the population potentially benefiting from palliative care
Between 13% (Canada) and 53% (Mexico) of deaths occurred at home (table 2). Between 25% (the Netherlands) and 85% (South Korea) of deaths took place in hospital. The Netherlands (35%) and South Korea (1%) had the highest and lowest proportions of deaths in long-term care institutions, respectively. In the countries where palliative care institution was registered as a place of death, between 4% (USA) and 13% (New Zealand) of deaths occurred there.
Characteristics associated with home death
In most countries, death from cancer was more likely to occur at home (as opposed to hospital) than death from a non-cancer disease (table 3). In France, Spain, Belgium, the Czech Republic and South Korea, home death was more likely for people with non-cancerous conditions. Women had higher chances of dying at home than men in France, Italy, Spain, England, New Zealand, the USA and Mexico, and lower chances in South Korea. In most countries, the probability of dying at home rather than in hospital increased with older age; but this was converse in England, Wales, New Zealand and Canada. Married people were more likely to die at home than people who were unmarried, widowed or divorced in all countries except Mexico where we found no difference. Residing in a rural rather than in a highly urban environment was associated with greater chances of dying at home rather than in hospital, except in France and Canada where the opposite was observed. The probability of dying at home generally decreased with a higher density of hospital beds, except in Spain and the Netherlands where the probability of home death increased. Associations between the probability of home death and the density of long-term care beds and GPs were not consistent across countries.
Characteristics associated with death in a palliative care institution
In the countries where palliative care institution was recorded as the place of death, patients with cancer and women (the latter not in New Zealand) had higher chances than patients without cancer and men to die in a palliative care institution rather than in hospital (table 4). Persons younger than 80 years and married persons in all countries except in the USA had higher chances of dying in a palliative care institution rather than in hospital. In the USA, persons younger than 80 and married persons were less likely to die in a palliative care institution. In England and Wales, there was a negative association between the probability of dying in a palliative care institution and the density of hospital beds.
Multivariable analysis aimed at explaining country-variation
Figure 1 shows to what extent sets of independent variables explained the variation between countries in dying at home rather than in hospital. The closer the OR is to 1, the smaller the difference in the probability of dying at home between the indicated country and Belgium as the reference country. The country differences in the probability of dying at home versus in a hospital (model 1) remained largely unchanged when cause of death and sociodemographic factors were added to the model (model 2). However, the adjustment for the density of hospital and long-term care beds and GPs per region of residence (model 3) considerably reduced or reversed the differences in the probability of dying at home between Belgium and the other countries. Only in the Netherlands did the probability of dying at home remain higher than in Belgium.
This study found considerable differences in the place of death of people who died from diseases indicative of palliative care need in 14 countries. The association between home death and cause of death (cancer vs non-cancer), sociodemographic factors, urbanisation and availability of hospital and long-term care beds and GPs differed in strength and direction in the countries studied. The differences between countries in the probability of dying at home were only partly explained by differences in the cause of death, sociodemographic characteristics and availability of hospital and long-term care beds and GPs, which suggests that factors related to end-of-life care policy may be important determinants of where people die.
There are at least two important roles of health and social care policy in general, and end-of-life care policy in particular, with regard to where people die. First, policy should be designed to enable people to die in their preferred location. Second, policy should be developed to ensure that high-quality care is available in the locations in which a large percentage of the population die. In the following sections, we elaborate how the findings of this study can inform these two tasks.
Place of death patterns help identify the care settings that have an important role in the delivery of end-of-life care in a country. They indicate which settings require particular attention from health services research and public health policy in terms of the availability, quality and cost of palliative care provided there and its integration with other health and social care services. We observed relatively high proportions (60% or more) of hospital deaths in France, the Czech Republic, Hungary, Canada and South Korea, which suggests that in these countries the hospital is where a major part of the palliative care population receive care in the final days or at least hours of life. This raises several concerns. Less than a third of those who die in hospital in France receive palliative care.31 Patients in the Czech Republic are often referred to hospitals at the end of life because they cannot receive adequate formal (palliative) care at home. At the same time, end-of-life care in Czech hospitals is described as often being inadequate.32 In Canada and South Korea, the percentage of hospital deaths is very high compared to the percentage of those who would prefer to be at home as they near death or when they die.33 ,34 The high percentage of hospital deaths in some countries should prompt considerations on the part of health policy whether patients who wish to die at home and their family carers receive appropriate palliative care and other forms of support that enable them to do so.
In several other countries, for instance in Italy, Spain, the Netherlands, USA and Mexico, we observed that one-third or more of people with a potential palliative care need died at home. However, population-based survey data on preferences from Italy, Spain and the Netherlands (not available for the USA and Mexico) indicate that this is still substantially lower than the proportion that would prefer to die at home in those countries, if circumstances allowed it.5 A high percentage of deaths at home should lead to considerations regarding the availability of formal home care, particularly home palliative care, and the availability of support for family carers. The relatively low public expenditure for long-term care at home in countries where many die at home (eg, in the Mediterranean countries)35 raises concerns about the quality of dying at home and calls for a closer investigation of the quality of care people receive at home. This applies to an even greater extent to Mexico where for a large part of the population dying at home is the only option. Home death in Mexico is associated with lack of health insurance,36 hospitals mainly focus on acute care, and palliative care is not integrated into mainstream healthcare services.
The country differences in dying at home versus in a hospital were only partly explained by differences in the cause of death, sociodemographic characteristics and availability of hospital and long-term care beds and GPs per region of residence. We hypothesise that at least part of the remaining variation may be accounted for by differences in policies in health and social care, and specifically end-of-life care. We make this assumption for two reasons. First, a systematic review of factors influencing death at home in patients with cancer found that where people die is strongly associated with availability of formal healthcare services in their local area.28 Second, we found that the patterns of place of death in certain countries could potentially be explained by the choices these countries have made with regard to, for instance, prioritising community-based care or public funding for palliative care services. For instance, the focus on palliative care in the community in the Netherlands may have contributed to the relatively high percentage of deaths at home and in long-term care institutions there. In the Czech Republic, on the other hand, home palliative care services are not recognised in the health insurance law, and therefore many patients are admitted to hospital when their condition worsens.32 One important task for future research should be to study whether greater availability of formal and publicly funded home care enables more people to die at home if they wish to. However, we must also assume that, next to policy-related aspects, there may be other factors that influence where people die, such as availability of extended family support28 or values and attitudes within the larger societal context.28 ,37
The use of death certificate data has the major advantage of providing population-level data that are not subject to potential bias inherent to sampling and that are cross-nationally comparable, as the procedure for registering deaths is similar across countries. However, conducting research based on death certificate data also has limitations. The validity of the cause of death as recorded on the death certificate has been contested,38 but the use of aggregated cause of death categories has most likely mitigated the risk of misclassification. On the basis of the underlying cause of death, it is impossible to estimate the precise content of palliative care that a particular patient needed. However, since we employed a conservative method for selecting the population in need of palliative care, we can safely assume that these people would have benefited from a palliative care approach. For the USA and Spain, we were not able to obtain death certificate data from 2008 (USA: 2007, Spain: 2010). Furthermore, the data for Spain (only the Andalusia region) and Canada (excluding the Quebec province) may not be generalisable to the whole country. For instance, the Spanish autonomous community of Catalonia has been undertaking an extensive palliative care programme over the past years which could affect the place of death patterns.39 Although we tried to obtain healthcare supply data that were as similar between countries as possible, there is a certain level of heterogeneity that was unavoidable due to differences between countries in the ecological levels of the data linked to the individual death certificates and potential differences in the terminology regarding healthcare resources. Information on household characteristics would have been a more valid proxy for potential social support than marital status, but such data were not available across all countries. It is not possible to include data on people's preferred place of death in population-level cross-national studies as these preferences are not routinely recorded. Lastly, the dying experience in a particular location may differ depending on how much time a person spent there, but we cannot conclude from the place of death record on death certificates how long a patient had been in the location in which they died.
This study found considerable differences between 14 countries in the place of death of people who died from diseases indicative of palliative care need and in the factors associated with place of death. We also found that a considerable proportion of the country-variation in the place of death remained unexplained after controlling for clinical, sociodemographic and healthcare availability characteristics, which suggests that other factors, potentially related to end-of-life care policy, have an important influence on where people die. These results inform end-of-life care policymaking in at least three ways. First, they indicate for each country the important settings for the provision of end-of-life and palliative care. Second, they highlight the care settings in which evaluating the availability and quality of end-of-life and palliative care should be a priority. Third, the differences between countries can serve as a source of hypotheses about the public health interventions that could be taken to ensure that people receive end-of-life and palliative care in locations that correspond to their wishes.
What is already known on this subject
Most people prefer to die at home, and end-of-life care in community care settings may be related to lower costs than end-of-life care in hospitals.
Cross-national population-level studies of place of death provide scientific data for the planning, implementation and evaluation of policies that are designed to ensure high-quality care in the locations where a large percentage of the population die and policies that enable terminally ill people to die in their place of choice.
What this study adds
By showing where the majority of the population die, the findings of this study highlight the care settings per country (eg, home, hospital) where evaluating and guaranteeing the availability and quality of palliative care should be a priority.
The large differences between countries in the place of death are only partly explained by differences in the cause of death, sociodemographic factors and healthcare supply, which suggests that country-specific palliative and end-of-life care policies may be among the factors that play an important role in where people die.
For the IPoD study: the authors would like to acknowledge the following conegotiators of national data: Katherine Hunt (University of Southampton), Régis Aubry (Observatoire National de la Fin de Vie), Luisa Frova and Stefano Marchetti (Italian National Institute of Statistics), Rod McLeod (University of Sydney). For the EURO IMPACT consortium: Zeger De Groote (Vrije Universiteit Brussel (VUB) & Ghent University), Sheila Payne and Sarah Brearley (Lancaster University), Augusto Caraceni (EAPC Research Network; Fondazione IRCCS Istituto Nazionale dei Tumori), Bregje Onwuteaka-Philipsen, Roeline Pasman and Anneke Francke (VU University Medical Center), Richard Harding and Irene J Higginson (King's College London; Cicely Saunders International), Stein Kaasa (Norwegian University of Science and Technology), Karin Linden (Springer Science and Business Media), Sophie Pautex (EUGMS European Union Geriatric Medicine Society). The authors thank the following agencies for the delivery of the death certificate data: Belgium: Flemish Agency for Care and Health, Brussels Health and Social Observatory, French Community of Belgium; France: Inserm-CépiDc (Centre d’épidémiologie sur les causes médicales de décès, Institut national de la santé et de la recherche médicale); Italy: Italian National Institute of Statistics (Unit for Cause of Death Statistics); the Netherlands: Statistics Netherlands (CBS); Spain (Andalusia): Instituto de Estadística y Cartografía de Andalucía; the Czech Republic: Institute of Health Information and Statistics of the Czech Republic (IHIS); Hungary: Central Statistical Office Hungary; USA: Center for Disease Control and Prevention (CDC) based on Data Use Agreement; Canada: Statistics Canada; New Zealand: New Zealand Ministry of Health; England and Wales: Office for National Statistics (ONS); Mexico: Secretaria de Salud and former Sistema Nacional de Informacion en Salud (now named Direccion General de Informacion en Salud); Korea: Statistics Korea.
DH and JC contributed equally.
Twitter Follow Lara Pivodic at @LaraPivodic
Contributors JC, DH and LD conceived of and designed the International Place of Death study. JC, DH, LM, JA-H, GM, MC-T, BO-P, WN, MRR, DW, ML, AC, YJR and JT were responsible for the acquisition of the data. JC and DH pooled the data from the different countries and did the data cleaning. LP, JC, LVdB, KP and LD conceived of the topic of this article. LP and JC performed the statistical analysis and interpreted the data, and KP, LVdB and LD made substantial contributions to the data analysis and interpretation. LP wrote the manuscript. All authors revised the manuscript critically for important intellectual content. All authors read and approved of the final version to be submitted for publication.
Funding This research is supported by the Research Foundation Flanders and the Willy Gepts Scientific Fund (Wetenschappelijk Fonds Willy Gepts). JC is a postdoctoral fellow of the Research Foundation Flanders. LP undertook this work as part of the European Intersectorial and Multidisciplinary Palliative Care Research Training (EURO IMPACT) project. EURO IMPACT is supported by the European Union Seventh Framework Programme (FP7/2007–2013 (264 697)) and aims to develop a multidisciplinary, multiprofessional and intersectoral educational and research training framework for palliative care research in Europe. The funders had no role in the study design; in the collection, analysis and interpretation of the data; in the writing of the report, and in the decision to submit the paper for publication.
Competing interests None declared.
Ethics approval Ethics approval was not required for this work as anonymised death certificate data were studied.
Provenance and peer review Not commissioned; externally peer reviewed.
Data sharing statement No additional data are available.
If you wish to reuse any or all of this article please use the link below which will take you to the Copyright Clearance Center’s RightsLink service. You will be able to get a quick price and instant permission to reuse the content in many different ways.