Introduction

Very few studies have addressed the question of the quality of death, despite the fact that this is, and has always been, a major concern to humanity.^1-3 When an incurable disease is diagnosed, as well as treating the symptoms presented, plans should be made for providing health care in the terminal phase. There is evidence that factors such as early palliative care and the control of symptoms are strongly associated with a better quality of death.⁴

It is known that most patients with advanced cancer prefer to die at home.^5-17 In Europe, according to a random sampling study of the general population of seven countries,⁸ more than half would prefer to die at home (in Spain, 66%). There is no evidence that the educational level of patients with terminal cancer is associated with their preference for the place of death. In this respect, preferences are associated with other factors, such as religious beliefs, personal values or the duration of the illness.⁸

Nevertheless, the majority of deaths by cancer patients occur in hospitals,^9-10 where the quality of death may be worse.¹¹ Many patients whose death was foreseeable on admission die after invasive diagnostic and therapeutic actions, while psychological, social and family needs, which can be as important as physical ones, or even more so, are ignored.¹²

There are three main places of death: at home, in hospital and in a long-term care centre (LTCC). Among them, the distribution of cancer deaths presents large variations between and within countries. A study conducted in six European countries reported that better chances of dying at home are associated with solid cancers, being married, higher educational attainment and living in less urbanised areas.¹⁰ Many studies have identified predictors of death at home, of which the most influential is the presence of an informal caregiver. In addition, the accessibility of palliative care services in the community at affordable cost is a significant prognostic factor, which might account for the large differences found between countries.¹⁴ Access to community palliative services is not homogeneous by socioeconomic groups: thus, it has been reported that “patients in higher socioeconomic groups were both more likely to die at home and to access home care”.¹⁵ Since the 1980s, the highest frequencies of deaths at home have corresponded to patients with higher levels of education;¹⁶ similar findings have been reported in Spain.¹⁷ One explanation for this could be that areas with higher socioeconomic status have more palliative care services available.^18-20 In Italy, the territorial differences between 13 provinces in the proportions of deaths at home persist even after adjusting for individual characteristics and context, which suggests there are systematic differences in access to palliative services.^21-22

Spain provides an interesting case study. It has a universal, free-of-charge national health system and an expanding specialised network of palliative care, following a national palliative care strategy that has been updated for the period 2010-2014. Accordingly, it might be expected that there would be no major personal or territorial differences in the quality of death or systematic differences in the place of death due to cancer, among patients or municipalities. But as a matter of fact, the development of palliative care differs substantially between regions.^23-24 The Spanish Society for Palliative Care (SECPAL) produces a standardised, detailed record of available infrastructure, human and organisational resources, providing a basic source of information.²⁴ There are important differences between autonomous communities in Spain in the allocation, organisation and management of palliative care resources.²⁵ Some regions have special plans for socio-health care and/or palliative care, following the guidelines of the National Strategy for Palliative Care that was launched nationally in 2006. That is the case of Catalonia, where palliative care is integrated in the strategy of social and long term care²⁶ and Basque Country, which have defined a plan coordinating primary and specialized resources.²⁷ Palliative medicine is not a medical specialty and the variability among health areas and regions in the organization of end-of-life care is noticeable, particularly regarding the role of primary care versus hospital care.

The main study aim is to examine individual and territorial variations in relation to the place of death due to cancer in Spain. To our knowledge, no previous studies have been conducted to analyse territorial variations in the place of death (home, hospital or LTCC) due to cancer using population data, including education among individual determinants.

Methods

Results

Table 1 presents the descriptive statistics for the total number of deaths due to cancer, differentiating according to whether the municipality of residence is identified or not. It also describes the municipal variables finally included in the models.

In large and medium-sized municipalities, 25.6% of deaths caused by cancer die at home, compared to 32.7% in small municipalities. Dying in hospital is more frequent in municipalities of more than 10,000 inhabitants (66.4% vs. 58.6%), while dying in a LTCC is very uncommon in Spain, the incidence being slightly higher among residents of small municipalities. There are other differences by size of place of residence: decedents in small municipalities tend to be older, there is a higher proportion of men and widowers in the population and a lower proportion of residents have higher education qualifications.

There are also large variations among identifiable municipalities in the percentage of cancer deaths at home. Counting only the municipalities that recorded more than 100 deaths from cancer in 2015, the raw data show an enormous range of variation, from 6.7% to 57.0% in the proportion of home deaths. Deaths in LTCCs range from 0% to 33.1% of the total, and those in hospital, from 39.3% to 90.5%.

The multilevel models with individual data and municipality effects clearly show that education is a very significant prognostic factor of the place of cancer death (Table 2). Thus, when the patient has higher education qualifications, the probability of dying at home vs. in hospital is increased by 0.047 compared with individuals without studies. The probability of dying in a LTCC decreases monotonically by education level. The other explanatory variables are very significant, in particular haematological cancer, but also sex and marital status, with the expected signs: female have a higher probability to die at home than male as the mainstream of the literature finds; the same stands for married patients. It is more probable for a female to die at home than for a male (0.033). Figure 1 represents the estimated municipal effects by size of population (10,000-20,000, 20,000-50,000, 50,000-100,000, larger than 100,000 and capitals of province irrespectively their population size). The municipal effects are very significant and represent 8.6% of the residual variance for death in LTCCs, 6.23% for death at home and 5.97% for death in hospital. That means that some characteristics of the municipality are influencing systematically the place of death even though it may be not possible to identify those characteristics to include them explicitly in the model. The municipality effects account for the “unobserved heterogeneity” among municipalities.

Figure 1 Municipality effects estimated with multilevel model M3 dying at home vs. in hospital by size of municipality. The average of all municipalities in sample is 0. In municipalities with positive sign dying at home is more frequent than for average, in municipalities with negative sign dying in hospital is more frequent than for average. Province capital includes all capitals irrespective of their population size. 

The second-stage models (Table 3) find that dying in hospital is more likely in provincial or island capital cities. Significant effects, which provide the highest standardised coefficients, were recorded for some of the 17 autonomous communities; especially noteworthy were the facts that death in a LTCC is associated with Catalonia, death in a hospital, with Madrid and the Canary Islands, and death at home, with Andalusia and Valencia.

According to the results of the models, the availability of palliative care services is only marginally significant, but always with the expected signs: the more availability of home services, the higher the probability to die at home. Death in a LTCC is more likely if palliative care services are provided at LTCCs in the municipality, while death at home is less likely if the only palliative services available are hospital-based. The lack of access to palliative services at home is not significant in the model.

The robustness tests with data for 2014 and identical models show similar results. The municipal effects obtained in the first stage in both years independently show a linear correlation of over 0.7 in all cases (0.81, 0.73 and 0.72 for models 1, 2 and 3 respectively).

Discussion

According to our results the place of death for patients suffering cancer is related to the municipality of residence, education and gender. Dying of cancer in hospital is an urban phenomenon while home deaths are more frequent in rural municipalities.

Male die more frequently in hospitals, women do it at home. This result is in accordance with the mainstream of the literature, which systematically finds a higher diagnostic and therapeutic intensity for male.²⁹ Our results coincide with those obtained for other countries, such as France,³⁰ on the influence of education and marital status on the place of death. Some studies conducted in countries where there is universal free-of-charge access to the public health service suggest that more highly educated patients may obtain a better quality of death, perhaps due to their greater competence in navigating the system.^31-32 If this were so, there could be a need for measures to facilitate access to the system and to eliminate non-monetary barriers. These barriers might be due to the fact that the homes of people with low educational level do not meet the minimum conditions to receive palliative care in them. If that is the case, the barriers might be counteracted by the intervention of social-health liaison officers. More highly educated people might obtain greater access to home-based palliative care at the end of life either because they reside in areas with better provision, or because they are better equipped to navigate through the administrative system or because the conditions of their homes facilitate it. Such inequalities would occur even within public, universal, free of charge health systems. Thus, the barriers are not institutional.

The finding that dying of cancer in hospital is an urban phenomenon while home deaths are more frequent in rural municipalities is robust. Both the provincial capital (and island capital) dummy variable and the dummy for large municipalities (over 100,000 inhabitants), which are included in some tests, are highly significant. Our model finds that the lack of access to palliative services at home is not significant in the model.

This unexpected result may be due to the high number of municipalities without such access in a specific region (Canary Islands), which is also included as a dummy in the model. When this dummy is excluded from the model, the lack of access is very significant at 1%, with a coefficient of −0.045 (result not shown).

Some autonomous communities, particularly Madrid and the Canary Islands, have extremely low percentages of home deaths compared with deaths in hospitals. This finding should be interpreted with caution. While in the Canary Islands it could be due to lack of access to palliative services at home, in Madrid it could be a consequence of hospital-centeredness derived from a hospital supply that is substantially greater than in others autonomous communities. At the other extreme, Valencia has a higher prevalence of home death, which may be associated with cultural and supply factors: it is the second region in the ranking of palliative care staff endowment (2.9 doctors per 100,000 inhabitants vs. 1.7 for Spain as a whole).

In economically depressed areas of the United Kingdom, terminal cancer patients have a lower probability of dying at home than in other locations.²⁰ However, our study found no evidence that the economic level of the municipality influences the probability of dying in one place or another. Although we analysed a comprehensive economic database of municipal information, including tax bases, social security affiliation data and detailed municipal budgets, the only clear conclusion to be drawn is that we cannot affirm the existence of such an association. As we used data from external sources and they showed no significant results, they are not shown in paper.

This study, which uses cross-information from various sources, such as the invaluable Directory of Palliative Care Resources, is the first to conduct an in-depth examination with population data on differences in the place of death in Spain, and it provides valuable information for cancer treatment plans and for palliative care.

However, it has limitations. The first is that small municipalities (fewer than 10,000 inhabitants) had to be excluded, for reasons of anonymity. Secondly, the fact of dying at home does not necessarily imply a good death, if the decedent failed to request palliative care services from the primary care physician or from specialised palliative care services. The lack of real measures of the quality of death is a limitation of the study. Another limitation is the fact that we are not able to know which resources have public or private ownership or which resources are accessible to publicly insured patients with no charge.

Taking into account the greater frequency of deaths in hospitals than in patients’ homes, and the ease of access to hospitals, at least in urban areas, it is to be assumed that a large proportion of home deaths are accompanied by the provision of palliative services. On the other hand, some patients and family members prefer death to take place in the hospital, because this environment offers the reassurance that unnecessary pain and suffering will be avoided.³³ In the autonomous communities in Spain there are different levels of implementation of the advance directives, which report the preferred place of death. Unfortunately, individual data are not available in the registry of death certificates. An acute care hospital, in many cases, is not the most appropriate place in which to die. In addition, a significant number of patients die in the emergency department, where staff focus more on trying to prolong life than in the palliative approach (approximately 20% of in-hospital deaths occur in this situation).^34-35

Developing specific plans for palliative care, with an active role being played by primary care teams, may help improve the situation. Therefore, health care policies should take account of the need to ensure access to palliative services in the home. On the other hand, there is an active debate on the complementary versus substitutive role of primary care and hospital care in the optimum palliative care provision. Professional specialization in palliative care lacks regulation. Specific training in the skills for palliative care must be a part of the core curriculum of all accredited training programs in different professions (physicians, nurses and others).