Abstract
Background: Place of death has for the past decade increasingly come to be regarded as a robust indicator of how palliative care is organized and provided, and is also recognized as an important factor for well being at the end of life. Variations in place of cancer deaths have previously been reported in the context of country-specific healthcare organization, but without differentiating between cancer types and national regional variations. Our aim was to examine, at a population level, where people with cancer diseases die in Sweden, and to investigate associations of place of death and cancer type with individual, socioeconomic and geographical characteristics of the deceased.
Material and methods: This population level study is based on death certificate data (sex; age; underlying cause of death and place of death) and population register data (educational attainment, marital status, living arrangements, area of residence, degree of urbanization, and healthcare region) of all 2012 cancer deaths in Sweden, with a registered place of death (hospital, nursing home, home, other places). Data were explored descriptively. To investigate associations between place of death and cancer types, and individual, socioeconomic and environmental characteristics, a series of multivariable logistic regression analyses were performed.
Results: The most frequent type of cancer death occurring at home was upper gastrointestinal cancer (25.6%) and the least frequent was hematological cancer (15.2%). Regional variations in cancer deaths occurring at home ranged from 17.1% to 28.4%. Factors associated with place of death by cancer type were age, educational attainment, marital status, healthcare regions and degree of urbanization.
Conclusion: Large healthcare regional variations in place of death among different cancer types were found. The socioeconomic inequality previously demonstrated for screening, diagnostic and treatment processes, rehabilitation and survival thus also seems to be reflected in the place of death.
For palliative cancer care policy, place of care is considered increasingly important [Citation1], for promoting equal and appropriate access to palliation regardless of care place. During the past decade, it has increasingly come to be regarded as a robust indicator of how palliative care is organized and provided in a country. In the end-of-life discussions, the desired general outcome is clearly death at the patient’s preferred place [Citation2], and worldwide the most common preference for people with cancer is to die at home [Citation3]. However, there is a large discrepancy between preferred and actual place of death. Hospital is a common or the unequivocal most common place of death across countries, although proportions of cancer deaths occurring in hospitals are reported to vary greatly, and both increasing and decreasing trends exist in different countries [Citation3].
Today, clinical practice models exist on how to integrate palliative care principles throughout cancer care [Citation4,Citation5], and awareness of the value placed on being at home by patients and families is apparent throughout the development of palliative care [Citation6]. However, despite increased attention being paid to palliative care, national guidelines for chronic conditions including cancer are lacking in terms of their end-of-life and palliative care guidance [Citation7,Citation8].
The determinants of place of death are multifactorial [Citation2]. Among factors influencing hospital deaths are underlying cause of death, for example hematological cancer, and the availability of acute hospitals and inpatient beds, whereas factors influencing home deaths include living with relatives, having a partner and socioeconomic status [Citation3]. Other factors that seem to influence hospital deaths are emergency department attendance in the last month of life for people with low socioeconomic status, being male and dying of lung cancer [Citation9]. Metropolitan populations have been shown to be less likely to die at home in some countries [Citation10], and impact of socioeconomic deprivation varies by geographical region rather than urban or rural residence [Citation11]. Equity analyses have previously been performed in cancer care, including screening [Citation12], diagnostic and treatment processes [Citation13], rehabilitation [Citation14,Citation15] and survival [Citation16]. In order to inform policy and to improve quality of care at the end of life, equity analyses related to end-of-life cancer care are equally important. Indeed, this will become even more critical with the expected increase of cancer deaths in the next decade [Citation2], especially given that socioeconomic inequalities in health are well known [Citation17].
The proportion of home deaths for people dying from cancer varies greatly in Europe and between studies; from 12.8% (Norway) [Citation18] to 46.3% (The Netherlands) and for hospital deaths from 44.3% (UK) to 72.8% (France), whereas nursing home deaths range from 3.3% (Spain; Andalusia) to 19.3% (The Netherlands).[Citation19] Furthermore, previously demonstrated international variations in place of cancer deaths have been associated with country-specific healthcare organization and resources for cancer care, but without differentiating between cancer types.[Citation19] A place-of-death study of the Swedish population showed that hospital is still the most common place of death for cancer deaths (50.9%), whereas only 21.1% of the cancer population die at home and 26.6% in nursing homes.[Citation20] However, there is a lack of studies focusing on geographical and other determinants for place of death for different cancer types.
Aim
The aim of the present study was to examine, on a population level, where adult people with cancer diseases die in Sweden, and to investigate associations with place of death and cancer type, and individual, socioeconomic and geographical characteristics of the deceased.
Material and methods
Design and population
This study is part of a larger population register-based place-of-death project.[Citation20,Citation21] The study is based on death certificate data of all 20 710 individuals who died of cancer in 2012 in Sweden and for which the place of death was registered (6.3% missing). The missing deceased individuals were (when statistically exploring the variables included in the study) randomly distributed within the cancer death population.
Death certificate data, processed by the Swedish National Board of Health and Welfare, were retrieved and included the following variables: sex; age; underlying cause of death and place of death. These data were linked, by Statistics Sweden, with data from their total population register, the multigenerational register, the historical population register and the register for data on properties, educational attainment and geographical location. Ethical clearance was obtained from the Regional Ethical Review Board in Stockholm (no. 2013/1745-31/5).
Variables
Place of death was used as the outcome variable for all analyses and was categorized as hospital, home, nursing home or other place (e.g. public places, roads, workplace). Nursing home refers to long- or short-term care facilities in municipalities, primarily for frail older people. However, a negligible number of care facilities for severely disabled younger people are also included in this category, as the different types of nursing homes are not distinguished in the death certificate. Hospices and specialized palliative care services exist in the country, but are not a place-of-death category on the death certificates.
Underlying cause of death, registered as ICD-10 classifications in death certificates, were grouped into seven distinct categories: upper gastrointestinal cancer; lower gastrointestinal cancer; pulmonary cancer; breast and gynecological cancer; prostate cancer; hematological cancer and other cancers (for ICD codes see supplemental data).
Individual and socioeconomic characteristic variables known to affect place of death were included in the analyses: sex (male/female), age (six distinct age categories), marital status (unmarried/married/widowed/divorced), and educational attainment (operationalized according to the Swedish classification system SUN2000 into the following categories: no formal or elementary education; lower secondary education; higher secondary education and higher education).
Finally, two additional environmental variables were used, based on: 1) the six geographical health care regions (Region 1, 2, 3, 4, 5, 6) in the country that cover health service responsibilities for a regional population, including special organizations for coordination of cancer care; and 2) degree of urbanization of the area of residence (urban/rural). The urban/rural variable is operationalized by Statistics Sweden, defining urban as: ‘an area with continuous settlements/houses, with 200 meters or less between the houses and with at least 200 inhabitants, and with no consideration taken to municipality or provincial borders’.
Analysis
Percentages were calculated for distribution of cancer types by place of death and healthcare region. To investigate associations between place of death and cancer types, and individual, socioeconomic and environmental characteristics, multivariable logistic regression analyses were performed. Using place of death as the dependent variable, five complementary models were developed stepwise, controlling for covariates (Model 1 = sex, age; Model 2 = sex, age, cancer type; Model 3 = sex, age, cancer type, educational attainment, marital status; Model 4 = sex, age, cancer type, educational attainment, marital status, urban/rural; Model 5 = sex, age, cancer type, educational attainment, marital status, urban/rural, health care region) in order to understand the respective role of these characteristics. The analyses were stratified (and performed separately) according to the living situation (home or nursing home) in order to understand the patterns of place of death (living at home and dying in hospital vs. dying at home or in nursing home; and living in nursing home and dying in hospital vs. dying in nursing home). For calculations of the model with nursing home residents we only included the last four age categories (≥60 years) in the multivariable logistic regression model. In the regression analyses, covariates were considered to have a significant association with the outcome if p < .05. The goodness-of-fit was systematically assessed using the area under the ROC curve (C-statistic) as a proxy.
For all statistical computations, SPSS 22.0 for Mac (SPSS, IBM Corp.) was used.
Results
Place of death by cancer type and health care region
Of the 20,710 cancer deaths (52.0% men and 48.0% women), 50.9% died in hospital, 21.4% at home, 26.4% in nursing home, and 1.4% in other places. The most frequent type of cancer death occurring in hospital was hematological cancer (64.5%), whereas the least frequent was prostate cancer (38.3%). Only 19.9% of all individuals dying of hematological and pulmonary cancer did so in nursing homes, whereas 38.1% of deaths caused by prostate cancer occurred in nursing homes. The most frequent cancer type death to occur at home was upper gastrointestinal cancer (25.6%) and the least frequent was hematological cancer (15.2%) (). Of all individuals who died of cancer, the proportion of women who died in hospital was 47.8%, with 50.9% occurring in nursing homes and 45.1% at home.
Table 1. Distribution of place of cancer deaths by sex, age and cancer type.
The cancer deaths occurring in hospitals in the six different health care regions ranged from 40.4% (Region 4) to 63.9% (Region 3). The number of nursing home deaths ranged from 15.7% (Region 3) to 33.0% (Region 5) and home deaths ranged from 17.1% in (Region 3) to 28.4% (Region 4).
By cancer type, the proportion of home deaths in the different healthcare regions ranged from 21.2% (Region 3) to 34.0% (Region 4) for upper gastrointestinal cancer, 17.5% (Region 3) to 33.6% (Region 4) for lower gastrointestinal cancer, 16.3% (Region 5) to 26.7% (Region 4) for pulmonary cancer, 12.7% (Region 3) to 27.9% (Region 4) for breast and gynecological cancer, 14.9% (Region 3) to 33.2% (Region 6) for prostate cancer, 12.5% (Region 5) to 22.5% (Region 4) for hematological cancer and 17.8% (Region 3) to 26.0% (Region 4) for other cancer types ().
Table 2. Distribution of place of cancer deaths by cancer type and healthcare region.
Factors associated with place of death by cancer type
Controlling for age and sex, people of all ages residing at home were associated with a significantly higher likelihood of dying in hospital than dying at home (Model 1), and this remained unchanged when also controlling for: cancer type (Model 2), marital status and educational attainment (Model 3), degree of urbanization (Model 4), and healthcare region (Model 5). Models 1–4 are provided as supplementary material (available online at http://www.informahealthcare.com).
The results from the final model () show that people residing at home were associated with a higher likelihood of dying in hospital if between 60 and 89 years old, compared to the oldest individuals (≥90) [odds ratios (ORs) ranging from 1.28 to 1.31]. The likelihood of dying in hospital was also greater when death was caused by pulmonary, breast and gynecological, hematological and other cancer types, compared to dying of lower gastrointestinal cancer (ORs ranging from 1.34 to 2.39). Moreover, living in urban (OR 1.20) rather than rural areas, and in the Stockholm healthcare region (OR 1.55), rather than the northern region was also associated with a higher likelihood of dying in hospital. Those living at home were also less likely to die in hospitals if married (OR 0.65) compared to being unmarried, and if living in the southeast (OR 0.60), compared to the northern region. People residing in nursing homes were associated with a higher likelihood of dying in hospitals if death was caused by hematological cancer (OR 2.95) compared to dying of lower gastrointestinal cancer, and if living in the Stockholm region (OR 2.32), rather than the northern region. Nursing home residents with higher secondary, or higher education (ORs ranging from 0.37 to 0.54), who were married (OR 0.24) were less likely to die in hospital than in the nursing home.
Table 3. Associations between place of death and individual, socioeconomic and geographic characteristics among individuals dying from cancer, final model.
Discussion
Our study confirms that hospitals are the most common places for cancer deaths, with variations among different cancer types. Similar variations in the place of death according to cancer type have also been shown in other countries.[Citation18,Citation19] The results also demonstrate large variations in the place of death between the different healthcare regions in the country.
The highest proportions of home deaths were found in gastrointestinal cancer (upper and lower) and the lowest proportions in hematological cancer. These variations could be regarded as clinically expected due to the varying end-of-life trajectories. For example, hematological malignancy is known to be associated with a high level of intensive medical care, and prospects of death are especially hard to predict.[Citation22] Several palliative care needs could be assumed to differ depending on the type of cancer involved. However, dying in hospital has previously been shown to be associated with challenges regarding the shift from curative to palliative care, resulting in futile treatment and poor symptom relief at the end of life.[Citation23] For example, across various cancer types, associations between palliative chemotherapy treatments in the last month of life and unfavorable patient outcomes have been demonstrated.[Citation24] This is indicative of the importance of developing strategies and skills for identifying and communicating time points when patients are approaching the end of life, regardless of cancer type. The category other cancer was associated with a higher likelihood of dying in hospital and among the various cancer types included in this category are those with unknown primary tumor. For this group, various examinations might be required even close to the end of life, which could be a reasonable explanation for hospital deaths. However, for all malignancies except for prostate cancer, roughly every second patient dies in hospital, as do two out of three with hematological cancer. This raises questions regarding which terminal hospitalizations could be avoidable and how to prevent unfavorable hospital deaths for those at risk. The inclusion of palliative care measures should be undertaken at the end of life regardless of the occurrence of oncological examinations or treatments.
Living in urban areas was associated with a higher likelihood of dying in hospital, thus confirming the findings of previous research.[Citation10] This could be considered paradoxical since, for example the Stockholm region (urban) has the smallest number of hospital beds and the northern region (rural) has the largest number.[Citation20] However, it might be that proximity to a bed – rather than availability of a bed – is the underlying determining factor. Furthermore, associations between place of death and socioeconomic factors (marital status and educational attainment) were found, which has previously also been confirmed in other countries.[Citation3] Taken together, these differences point to socioeconomic inequality complexities in care for the dying, which previously have been demonstrated in research on cancer screening, diagnostic/treatment processes and rehabilitation.[Citation12–15]
The large variations in place of death between cancer types in the different healthcare regions are striking, given that the country has national policy documents for palliative care, and a nationally coordinated cancer care structure and organization. However, the first national clinical practice guidance for palliative care was published in 2012, whereas the first national policy guidance (governmental initiative) for palliative care was published in 2013. Quality indicators pertaining to palliative care existed in 2012 in Swedish national policy guidelines only for breast, colorectal, prostate and lung cancer. However, none of these indicators referred to place of care or the preferences of the patient in this regard.[Citation8] Only the national clinical practice guidance for one disease (colorectal cancer) included recommendations related to palliative care. These recommendations, however, were not about place of care at the end of life (www.cancercentrum.se).
Another important aspect that may be considered in relation to large cross-regional variations is that the six healthcare regions in the country have legislative freedom to organize their healthcare services. Similarly, the municipalities are responsible for organizing the care in nursing homes, in collaboration with healthcare services. As a result, the organization and development of hospice and specialized palliative care services vary nationally. Despite an increase over the past decade, according to an update from 2016, these services are still only available for a maximum of one fifth of the population, including individuals dying of cancer, and there are large variations between urban and rural areas as well as between health care regions (www.nrpv.se/palliativguiden). It should also be noted that the regional cancer organizations primarily emphasize cancer prevention, treatment and care, which is monitored in national cancer registers, and these registers have no variables pertaining to end-of-life care or place of death. The cross-regional variations disclosed in this study thus point to potential inequities regarding end-of-life cancer care within the country. However, further investigation is required in this regard.
Future research might consider special initiatives to support people whose preference it is to die in places other than hospitals—initiatives such as palliative care consultation teams, multidisciplinary palliative home care teams and palliative care champions in non-palliative care specialized settings.[Citation25] Moreover, actual place of death should be considered in relation to preferred place of death.[Citation2]
The study has limitations. Death certificate data are always potentially prone to errors in the registering of the underlying causes of death.[Citation21] Furthermore, deaths occurring in specialized palliative care versus non-specialized settings were not included in the analysis. It was not possible do so because Swedish death certificates only include the categories used in the study. Hence, deaths occurring in hospice and specialized palliative care settings could potentially have been registered in anyone of the place-of-death categories. Although place of death does not disclose the quality of care provided per se, it is regarded as a robust indicator of how end-of-life care is organized and provided in a country.[Citation1–3]
In conclusion, hospitals are the most common places for cancer deaths in the country, with large variations between cancer types, determined by socioeconomic and geographical factors. These variations, potentially indicative of inequities regarding the end of life for patients with cancer, may be related to vague national cancer care policy pertaining to palliative care, and may therefore be linked to differing structure and organization of care within the different healthcare regions.
IONC_1250946_supplemental_tables.docx
Download MS Word (38.1 KB)Acknowledgments
We thank The National Board of Health and Welfare and Statistics Sweden for extracting and providing data used in this study.
Disclosure statement
The authors report no conflicts of interest. The authors alone are responsible for the content and writing of this article.
References
- Hales S, Chiu A, Husain A, et al. The quality of dying and death in cancer and its relationship to palliative care and place of death. J Pain Symptom Manage. 2014;48:839–51.
- Gomes B, Higginson IJ. Where people die (1974-2030): past trends, future projections and implications for care. Palliat Med. 2008;22:33–41.
- Tang ST, Chen C-H, Place of death and end-of-life care. In: Cohen J, Deliens L, editors. A public health perspective on end of life care. New York: Oxford University Press; 2012. p. 21–34.
- Partridge AH, Seah DS, King T, et al. Developing a service model that integrates palliative care throughout cancer care: the time is now. J Clin Oncol. 2014;32:3330–6.
- Zimmermann C, Seccareccia D, Clarke A, et al. Bringing palliative care to a Canadian cancer center: the palliative care program at Princess Margaret Hospital. Support Care Cancer. 2006;14:982–7.
- Higginson IJ, Sarmento VP, Calanzani N, et al. Dying at home-is it better: a narrative appraisal of the state of the science. Palliat Med. 2013;27:918–24.
- Mast KR, Salama M, Silverman GK, et al. End-of-life content in treatment guidelines for life-limiting diseases. J Palliat Med. 2004;7:754–73.
- Lind S, Adolfsson J, Axelsson B, et al. Quality indicators for palliative and end of life care: a review of Swedish policy documents. BMJ Support Palliat Care. 2013;3:174–80.
- Henson LA, Gao W, Higginson IJ, et al. Emergency department attendance by patients with cancer in their last month of life: a systematic review and meta-analysis. J Clin Oncol. 2015;33:370–6.
- Houttekier D, Cohen J, Bilsen J, et al. Place of death in metropolitan regions: metropolitan versus non-metropolitan variation in place of death in Belgium, The Netherlands and England. Health Place. 2010;16:132–9.
- Lavergne MR, Lethbridge L, Johnston G, et al. Examining palliative care program use and place of death in rural and urban contexts: a Canadian population-based study using linked data. Rural Remote Health. 2015;15:3134.
- Edgar L, Glackin M, Hughes C, et al. Factors influencing participation in breast cancer screening. Br J Nurs. 2013;22:1021–6.
- Cavalli-Björkman N, Lambe M, Eaker S, et al. Differences according to educational level in the management and survival of colorectal cancer in Sweden. Eur J Cancer. 2011;47:1398–406.
- Holm LV, Hansen DG, Larsen PV, et al. Social inequality in cancer rehabilitation: a population-based cohort study. Acta Oncol. 2013;52:410–22.
- Moustsen IR, Larsen SB, Vibe-Petersen J, et al. Social position and referral to rehabilitation among cancer patients. Acta Oncol. 2015;54:720–6.
- Quaglia A, Lillini R, Mamo C, et al. Socio-economic inequalities: a review of methodological issues and the relationships with cancer survival. Crit Rev Oncol Hematol. 2013;85:266–77.
- Mackenbach JP, Stirbu I, Roskam AJ, et al. Socioeconomic inequalities in health in 22 European countries. N Engl J Med. 2008;358:2468–81.
- Cohen J, Houttekier D, Onwuteaka-Philipsen B, et al. Which patients with cancer die at home? A study of six European countries using death certificate data. J Clin Oncol. 2010;28:2267–73.
- Cohen J, Pivodic L, Miccinesi G, et al. International study of the place of death of people with cancer: a population-level comparison of 14 countries across 4 continents using death certificate data. Br J Cancer. 2015;113:1397–404.
- Håkanson C, Öhlén J, Morin L, et al. A population-level study of place of death and associated factors in Sweden. Scand J Public Health. 2015;43:744–51.
- Cohen J, Bilsen J, Miccinesi G, et al. Using death certificate data to study place of death in 9 European countries: opportunities and weaknesses. BMC Public Health. 2007;7:283.
- Cheng BH, Sham MM, Chan KY, et al. Intensive palliative care for patients with hematological cancer dying in hospice: analysis of the level of medical care in the final week of life. Am J Hosp Palliat Care. 2015;32:221–5.
- De Korte-Verhoef MC, Pasman HRW, Schweitzer BP, et al. Reasons for hospitalisation at the end of life: differences between cancer and non-cancer patients. Support Care Cancer. 2014;22:645–52.
- Näppä U, Lindqvist O, Rasmussen BH, et al. Palliative chemotherapy during the last month of life. Ann Oncol. 2011;22:2375–80.
- Houttekier D, Cohen J, Van den Block L, et al. Involvement of palliative care services strongly predicts place of death in Belgium. J Palliat Med. 2010;13:1461–8.