2.2 The epidemiology of death and symptoms: planning for population-based palliative care

Epidemiology is defined as ‘the study of the distribution and determinants of disease frequency’ (Hennekins et al., 1987). Epidemiology is central to the development of strategies for the prevention and management of disease in populations and for the planning of health services. Epidemiological data can also provide information about the nature of the progression of specific diseases and treatment outcomes. For palliative care, epidemiology can provide important information about disease and symptom occurrence as well as health-care needs. In this chapter, we take a broad view of the epidemiology of ‘disease’ towards the end of life and discuss epidemiology as it relates to the ‘human experience’ with an emphasis on disease, symptoms, psychosocial experiences, and access to health services. Areas where information about a population is available and where it is needed, but lacking, are highlighted. Examples of the way in which epidemiological data have informed planning, policy, or patient care are highlighted throughout this chapter. Epidemiological studies on populations at the end of life are relatively few. Where data from large or whole population studies are unavailable, smaller studies such as those derived from service-based data, have been used to illustrate aspects of the human experience towards the end of life. The attainment of high-quality population-based data is essential to understanding the needs that inform service development and provision of care across settings.

The World Health Organization (WHO) has defined palliative care. A definition of the palliative care population is important as it helps to articulate what palliative care is, who needs it, who should provide it (Rosenwax et al., 2006), and how well it is provided across health systems. A definition of the palliative care population is therefore a vital part of planning for palliative care service delivery. Notwithstanding this, defining the ‘palliative care population’ has been problematic. To date, among the approaches to defining the population have been the use of specific conditions, patient needs, and all deaths (Rosenwax et al., 2006). Definitions of the palliative care population may vary but it is essential to identifying who would benefit from palliative care and therefore remains a core challenge. The focus of research in this area has more recently centred on identifying markers of ‘palliative care needs’ as a basis for defining ‘the palliative care population’ (Boyd and Murray, 2010; Waller et al., 2010). Defining appropriate quality indicators is another important area of research. The type of diseases encountered as well as the socioeconomic, cultural, home, and natural environments that patients inhabit are all important variables that influence the spectrum of a population’s palliative care needs.

Incidence and prevalence are two important epidemiological measures. ‘Incidence quantifies the number of new events or cases of disease that develop in a population of individuals at risk during a specified time interval’ (Hennekins et al., 1987) and can be summarized as:

‘Prevalence quantifies the proportion of individuals in a population who have the disease at a specific instant and provides an estimate of the probability (risk) that an individual will be ill at a point in time’ (Hennekins et al., 1987) and can be summarized as:

Mortality statistics provide information on death rates and causes of death in populations and therefore provide important epidemiological data for palliative care. In reviewing mortality data, it is important to have an understanding of the limitations of data which are derived from diverse sources, each with its own limitations. The Global Burden of Disease (GBD) study, which was initiated in 1993 to provide comprehensive mortality and morbidity data, has been a collaboration involving the WHO, the World Bank, and other organizations. This study is ongoing, and since its first publication in 1994 the GBD data has been updated (Lozano et al., 2012; Murray, 2012; WHO, 2013a). Estimates of mortality relating to deaths in 2011 were published in 2013 and were largely obtained using the same methods as those used to produce previously published mortality estimates but were derived from more recent information from vital registration data (WHO, 2013b) (see Box 2.2.1).

Mortality estimates from the GBD are obtained from four general sources (Lopez et al., 2006):

Only a third of the world’s population resides in regions where complete civil registration systems exist that provide adequate, cause-specific mortality data. In most of Africa, South East Asia, the Middle East, and parts of the Pacific, where over one-quarter of the world’s population resides, there has until recently been little or no mortality monitoring (Rao et al., 2005; Lopez, 2006). Notable increases in the collection of data have been reported in Thailand and South Africa in publications from the WHO (2011).

It can be seen therefore, that reporting errors and inaccuracies relating to cause of death are a worldwide problem. Even in countries where deaths are reported with reasonable consistency, significant proportions of reports of death contain reporting errors (Maudsley et al., 1996; Rao et al., 2005; Mathers et al., 2006a). There are many reasons why death data may be unrepresentative. For example, in an effort to reduce complexity, mortality data are usually reported by single cause of death despite the fact that several co-morbidities and health risks may significantly contribute to death. This may introduce biases. These include, but are not limited to, economic constraints on various capacities for data collection and reporting, as well as political and other factors; these factors must be considered when assessing the limitations of global health data (Murray et al., 2004; Lopez et al., 2006; Mathers et al., 2006b; Wang et al. 2012).

Coding and reporting systems also influence the data available but increased adoption of standardized reporting systems such as the International Classification of Diseases (ICD; WHO a), by the majority of countries (from four countries in 1994 to more than one hundred in 2014) has resulted in improvements in ‘real-time’ availability of data relating to cause of death (WHO, b). Cumulative developments in information systems have potential for fostering further improvements in the comprehensiveness and accuracy of mortality statistics. Despite this, significant reporting delays exist in many regions in cause of death, location of death, and co-morbidities. The absence of these can impact on the service planning that is critical for optimizing the delivery of population-based care at the end of life (Wang et al., 2012).

Life expectancies vary greatly worldwide (see Fig. 2.2.1 ) (Institut national d’études démographiques (INED), 2012). These variations are associated with demographic characteristics such as occupational, political, cultural, and lifestyle risks as well as ethnicity, gender, and genetics (Commission on Social Determinants of Health, 2008; WHO, 2009, Institute for Health Metrics and Evaluation, 2010; Lim et al., 2013). Populations from low-income countries have not experienced the increase in life expectancy observed in the rest of the world, and communicable diseases and conditions of the newborn continue to be a significant cause of death. For several countries of sub-Saharan Africa, life expectancy in the 1990s had declined to 40 years or below (WHO, 2006). There have been some areas of notable improvement since that time; however, data demonstrate that countries in some regions have life expectancies less than or close to 50 years (INED, 2012).

Globally, life expectancy for 2012 has been estimated as 66 years for men and 71 years for women. In more developed regions the life expectancy at birth for both sexes is currently estimated to be 76.9 years and, in the least developed regions, 58.4 years (United Nations, 2013).

In addition to life expectancy, it is important to consider infant mortality. Within countries and across regions of the world, infant mortality varies. Across the world, the infant mortality rate has been projected to be 41.5 deaths per 1000 live births (INED, 2012). Two types of figures are reported for this very young age group—infant mortality per 1000 live births and infant deaths under the age of 1 year—and great variability exists for both. Afghanistan has, for example, had a very high infant mortality with projections of 123 deaths per 1000 live births in 2013 whereas projections for Iceland for the same year were for two deaths per 1000 live births (INED, 2012). In the contemporary era many countries have seen a decrease in child mortality. Others have been slower to achieve a decrease (Gapminder, 2010; INED, 2012).

In 2011, there were approximately 54.5 million deaths estimated throughout the world (WHO, 2013b). Fig. 2.2.2 presents the ten leading causes of death worldwide for 2011 (WHO, 2013c). As reports reflect individual cancers the general diagnosis of ‘cancer’ itself does not rank among the leading ten causes although lung cancers rank 6th with 1.5 million deaths (WHO, 2013d). The leading reported causes of death vary among regions at different levels of economic development and illustrate some of the health disparities associated with economic issues. The World Bank now classifies countries into four income groups: low, lower-middle, upper-middle, and high. The most recent data on causes of death by income categories are for 2011 (see Box 2.2.1). Globally, adult mortality rate declined from 204 per 1000 population in 1990 to 160 per 1000 population in 2011 (WHO, 2013a).

Disparities in causes of death between countries reflect divergent levels of economic development. More specifically, communicable diseases and conditions of the newborn are the predominant causes of death in low-income countries. These causes of death also vary between demographic groups within countries defined, for example, by socioeconomic status, gender, age, and ethnicity. The disparity in global mortality is greatest for low-income countries which constitute 11.4% of the world’s population but account for 16.3% of the world’s deaths (WHO, 2008).

Another example of where differences in health status, life expectancy, and causes of death exist is between ethnic groups within some high-income countries in which some indigenous populations have comparatively higher rates of death and morbidity from diseases including cancer, respiratory disease, stroke, injury, and diabetes when compared to the non-indigenous population of the same country (see Box 2.2.1) (Stevenson et al., 1998; Horton, 2006; Australian Bureau of Statistics, 2011a, 2011b; WHO, 2011).

Differences in the ranking of causes of death between time periods have been demonstrated, most notably by the data presented in the GBD study which presented comparative data from 1990 and 2010. For instance, while many communicable diseases have decreased in frequency, death from HIV/AIDS has moved from 33rd ranking to 6th, and road injury and self-harm have remained significant causes of death at 8th and 13th rank in 2010 (cf. 10th and 14th rank in 1990). Over those two decades, lung cancer has increased in frequency changing from 8th rank to 5th and diabetes has changed from 15th rank to 9th (Institute for Health Metrics and Evaluation, n.d.). WHO estimates comparing cause of death between the years 2000 and 2011 are available online (WHO, 2013d).

Mortality projections can assist in the planning of services that will be required to meet the needs of populations in the future and provide insight into requirements for the spectrum of knowledge and skills needed by palliative care clinicians (K. Strong et al., 2008).

Recently updated projections of mortality and cause of death have been released by the WHO and are summarized in Table 2.2.1. The death rates from various causes were estimated using data from 2011 (WHO, 2013d). An in-depth discussion of age-standardized mortality projections of cause of deaths for 2030 using 2002 data has been published by Mathers and colleagues. Among other estimates, they present data demonstrating expected increases by 2030 in deaths from HIV/AIDS, lung cancer, diabetes, chronic respiratory diseases, road traffic accidents, violence, and war (Mathers et al., 2006a). These 2002 projections for 2030 provide three kinds of estimates: baseline, pessimistic, and optimistic (Mathers et al., 2006a). Under the optimistic scenario, 64.9 million deaths worldwide were projected and under the pessimistic scenario 80.7 million. With respect to HIV/AIDS, a baseline projection estimates deaths from this cause will increase from 2.8 million in 2002 to 6.5 million in 2030. Deaths from other infective conditions or perinatal conditions are projected to fall. Over the same time period baseline projections of deaths from cancer suggest an increase from 7.1 million to 11.5 million, and cardiovascular deaths from 16.7 to 23.3 million. Combined deaths from cancer and chronic non-infective and non-cancer illnesses are expected to account for 70% of deaths in 2030. The third major mortality grouping is related to deaths from injury and accidents and notable in this group was a 40% expected increase predominantly accounted for by road traffic accidents which were projected to increase from 1.2 in 2002 to 2.1 million in 2030 (Mathers et al., 2006a).

Projections are developed to reflect various contributing factors including the success of preventive action. Projections may vary if behaviours related to disease incidence or disease related-mortality change. Examples of important factors that may impact projections include tobacco use, the incidence of obesity, the incidence of infective causes of death, and whether prevention efforts have a greater or lesser impact than expected (Olshansky et al., 2005; Strong et al., 2008; Wang et al., 2012; WHO, 2012). For example, deaths attributed to tobacco use are estimated to be likely to account for 8 million deaths worldwide annually (WHO, 2013f); if preventive efforts are less or more successful than predicted, the death rates for tobacco-related diseases will change. Given the uncertainties that exist concerning influential variables including economic, environmental, social and, technological factors variations can be expected between actual and projected mortality (Mathers et al., 2006). Deviations from the projected mortality will affect palliative care services planning.

One aim of health-care service providers is to offer care, where possible, in a location that matches the patient’s preference. Data reporting concordance between preferred place of death and actual place of death has been much less common than data that simply reports place of death (Pritchard et al., 1998; Gomes et al., 2012a).

Data reporting ‘place of death’ are for the most part collected only in high-income nations. In general, it is clear that for many individuals in low-income countries hospital care is not available, which would suggest that the vast majority of people in these regions die outside the hospital setting (English et al., 2006). The opposite is true in many high-income countries where death in hospital is common. The available data suggest that more than 50% of deaths in England, the United States, Germany, Switzerland, and France take place in the hospital (Gomes, 2013). Significant variations in place of death nonetheless exist among high-income countries, with lower rates of hospital death reported in the Netherlands (35%), Ireland (30%), and Italy (35%) (Klinkenberg et al., 2005; Beccaro et al., 2007).

Some studies have provided projections of where people in high-income populations are likely to die in the future, and suggest that the rate of death at home will decline over time (Gomes et al., 2008). Despite this, other data indicate that this trend may not necessarily continue and reports from several countries suggest there will be an increase in the proportion of patients who die at home, at least with respect to cancer deaths in populations which are served by specialist palliative care (Gomes et al., 2012a).

When asked about preferred place of death, the overwhelming majority of well people indicate a preference for care at home up until, and including, the time of death (Foreman et al., 2006; Beccaro et al., 2007; Gomes et al., 2012b; Gomes, 2013). This contrasts with actual place of death for the majority (McNamara et al., 2007; Gomes et al., 2012a; Gomes, 2013). The reasons for this disparity have been investigated, revealing that many factors, including access to care, influence this observation. For example, where inpatient beds are available, more inpatient care occurs (Pritchard et al., 1998; Higginson and Costantini, 2008). Recent research that has focused on ‘preferred place of care’ rather than only on ‘place of death’ takes into account the observation that the preferences of ill patients may change along the illness trajectory, and may be different to the preferences elicited from well people with or without life-threatening illness (Storey et al., 2003; Munday et al., 2007; Agar et al., 2008; Gomes et al., 2008; Gomes, 2013). These types of data illustrate some of the subtle, less acknowledged, reasons behind patients’ changes in their preferred ‘place of care’ (see Fig. 2.2.3). When collected longitudinally, rather than at a single point in time, data that compares actual place of care to preferred place can also help to inform planning and support flexibility in service provision up until and including the time of death (Agar et al., 2008). Further research is required to explore factors associated with the disparities between desired place of death and actual place of death that have been identified (Gomes, 2013).

Epidemiological data related to the provision of palliative care are important for understanding whether there is population-wide access to appropriate care. It also helps to identify populations in need of palliative care and, in relation to care towards the end of life, whether care is available in the patient’s preferred place of care. These types of data allow for designing and refining models of palliative care delivery that aim to provide appropriate care and, where possible to provide it in patients’ locations of choice. The WHO has endorsed palliative care as an essential component of health care (see Box 2.2.1) (WHO, 2014). To achieve this, it is vital that generalist and most specialist health practitioners have fundamental competencies in symptom management towards the end of life (Dudgeon et al., 2008; Shipman et al., 2008). It is generally accepted that there is also a need for patients to have access to specialist palliative care services when needed (Field and Cassel, 1997; National Institute for Clinical Excellence, 2004; WHO, 2007; Martin-Moreno et al., 2008; Temel et al., 2010; National Gold Standards Framework Centre, 2012).

It is evident that specialist palliative care services are becoming available for cancer and non-cancer diagnoses in most high-income countries through mainstream health services in community, inpatient, and acute hospital settings but availability is much more limited in middle-income and low-income countries (Fig. 2.2.4) (Morris, 2011). In high-income countries, many tertiary-referral centres accommodate integrated consultative, specialist palliative care services within acute and subacute settings (Glare et al., 2003; Mercadante et al., 2008). Studies have highlighted that palliative care specialist services in most middle-income countries are available but only for some of the patients in need, and are frequently unavailable for the poor and those living in rural and remote regions (Morris, 2011). Specifically, in low-income countries, the majority of those dying do not have any access to specialist services and those services that do exist reach only a very small proportion of the people in need (Kikule, 2003; Morris, 2011). The majority of dying people in these areas are cared for at home and in communities by family and/or neighbours. Regarding needs in these countries, one cross-sectional study identified three main areas of palliative care need in low-income areas —symptom management, counselling, and financial assistance (Kikule, 2003).

Despite the availability of services, for a significant proportion of patients with far advanced disease, even in high-income countries there are limits on access to symptom management and end-of-life care (Higginson, 1997; Pritchard et al., 1998; Rosenwax et al., 2006; Beccaro et al., 2007; Goldsmith et al., 2008). As an example of the impact of epidemiological data supporting this claim, a West Australian study showed that patients dying from illnesses other than cancer are less likely (8%) than those with cancer (68%) to receive specialist palliative care (Rosenwax et al., 2006). Studies elsewhere have raised similar access issues when considering palliative care (Solano et al., 2006; Murray et al., 2008).

Examples of the wealth of information that can be provided by generating epidemiological reports from carefully collected clinical data are reports such as those from Seow et al. (2012) and Laugsand et al. (2011) that describe components of multidisciplinary service provision, symptom management, prognosis, diagnoses, and co-morbidity. While not necessarily generalizable to other regions, these data provide information that could be highly useful in planning and tailoring regional services for the areas in which data was collected. Of note is that local needs assessments, using defined criteria, such as performance status, symptom prevalence, and prognosis, may also allow further comparison and benchmarking in relation to palliative care access—an approach such as that recommended by Kaasa and colleagues and others (Kaasa et al., 2008; Boyd and Murray, 2010; Currow et al., 2012).

Studies such as those just mentioned can assist identifying the extent to which people and populations with palliative care needs are accessing services. Methodical enquiry into this matter is important especially where disparities exist and/or access is not clearly defined as needs based. An emphasis on needs-based care, rather than on diagnosis- or prognosis-based provision of care, has been advocated by many (Davies and Higginson, 2004; Rosenwax et al., 2006; Young et al., 2008; Boyd and Murray, 2010; Waller et al., 2012); however, presently there are limited epidemiological data to inform practice in regard to this area. The routine collection and systematized reporting of data relating to care needs is crucial to this and the review, identification, and refinement of tools to help in identifying those who would benefit from palliative care will assist this process (Boyd and Murray, 2010; Waller et al., 2010; Weissman et al., 2011).

In the previous section, access to health services in relation to care needs was discussed. Importantly, specific research questions in relation to how well those specific needs are being addressed within populations generally warrant epidemiological investigation. Health-care needs at the end of life have a direct relationship with symptoms experienced during that time; in the context of palliative care (Franks et al., 2000; Mirando, 2004; Higginson et al., 2007), and in relation to many specific diseases at the end of life (Edmonds et al., 2001; Elkington et al., 2005), symptom management is also generally acknowledged as an important need, among many palliative care needs. Some data which quantify and describe various aspects of the symptoms experienced towards the end of life in particular populations and to a certain extent are available (Solano et al., 2006; Teunissen et al., 2007). In relation to health-care needs, data relevant to palliative care have been reported but prevalence has not necessarily been well quantified.

Currently in palliative care there is a focus on developing and describing appropriate processes that can be used to identify, compare, and contrast the palliative care needs of different people and populations (Higginson et al., 2007; Currow et al., 2008a; Waller et al., 2010; Currow et al., 2012). Illness trajectories, as discussed below, are one useful way of mapping population-based functional support needs over time. Aside from the importance of addressing symptom burden and care needs from a health-service perspective, epidemiological data are also needed to provide adequate answers to questions that come from individuals in clinical settings such as ‘Am I likely to have pain that cannot be managed?’ (Franks et al., 2000; Edmonds et al., 2001; Currow et al., 2004; Davies and Higginson, 2004; Mirando, 2004; Elkington et al., 2005; Higginson et al., 2007) and ‘Who will care for me if I can’t care for myself?’. In essence these questions that may be asked by patients and carers are the corollary of questions asked by palliative care providers, that is, ‘What needs are best met by specialist palliative care providers?’ and ‘How can we ensure that people who may not require specialist palliative care have their needs at the end of life met?’

With respect to symptom management, biomedical ethics and contemporary liberal philosophy suggest that there is not only justification for symptom management at the end of life, but an ethical (Sen, 1993; Freeman, 2007) and legal (Brennan, 2007) imperative for it. Within health services and/or specific populations, it’s important to develop benchmarks to assess outcomes and quality of care and to carefully describe study populations so as to facilitate comparisons across settings.

When reviewing and interpreting symptom-related epidemiological data there are a number of important aspects that must be considered. A summary of the key points to take into consideration in the interpretation of data reporting the epidemiology of symptoms follows:

The aim of this section has been to illustrate the relationship between empirical data and the symptom experience, and the uses, limitations, and challenges inherent in the interpretation of symptom-based epidemiological data. The following sections review existing epidemiological data related to symptoms and health-care needs at the end of life in the light of their incidence and prevalence, severity, frequency, associated distress, and in relation to impact on function and global burden for patients as well as caregivers.

Until recently, symptom prevalence studies in the palliative care setting have focused predominantly on patients with cancer diagnoses. There are now a number of good quality studies that have explored the prevalence of symptoms in patients with life-threatening and far-advanced chronic lung disease (Elkington et al., 2005; Walke et al., 2007) and cardiovascular disease (Addington-Hall et al., 1998a; Solano et al., 2006; Young et al., 2008). Large population-based data sets describing symptom prevalence are lacking; however, two systematic reviews of studies reporting point estimates of the prevalence of different symptoms have provided excellent overviews of symptoms experienced near the end of life (Solano et al., 2006; Teunissen et al., 2007).

Although generalizing is problematic, the available evidence suggests that for people with advanced, progressive life-limiting illness, there is a core group of symptoms, including pain, depression, dyspnoea, and fatigue, experienced across disease states in the last days, and probably the last year of life. Table 2.2.2 presents the data from a meta-analysis which included 64 studies across progressive cancer and non-cancer illnesses (Solano et al., 2006). The authors defined the study time-frame in the included patients’ end of life care and targeted 11 predicated symptoms. Studies restricted to the last hours of life were excluded. Notably this meta-analysis did not include cerebrovascular disease (stroke), now ranked the second most common cause of death worldwide, or Alzheimer’s dementia which is now reported as the fourth most common cause of death in high-income countries according to 2011 data (WHO, 2013e).

In a meta-analysis of 26 223 patients (Teunissen et al., 2007), studies relating to two time periods were analysed independently and data were presented in two groups relating to the time periods. The two groups were studies prior to the last 2 weeks of life (identified as ‘group 1’ in the original article) and studies conducted ‘in the last 1–2 weeks of life’ (identified as ‘group 2’ in the original article) (see Table 2.2.3 and Table 2.2.4). Despite the limitations discussed above, studies such as these help to define symptom experience over time and can be useful in service planning. As already discussed, point prevalence data and severity data for symptoms are influenced by, but do not provide detail of, access to quality care.

Data on the prevalence of specific symptoms, even in the setting of malignant disease, is more scarce than expected. With regard to the prevalence of pain, Bonica’s landmark review (Bonica, 1985) reported a prevalence of 71% in patients with advanced/metastatic/terminal cancer (Van den Beuken-van Everdingen et al, 2007). In addition to providing prevalence data, Bonica’s study highlighted methodological considerations in studies of symptom occurrence, including the presence of serious pain at all stages of cancer, its amenability to effective management, and the variation in pain experience over a day and over longer periods (Greenwald et al., 1987).

Other important studies have presented data relating to the experience of pain in cancer (Hearn et al., 2003; Holtan et al., 2007) and non-cancer (Solano et al., 2006; Murray et al., 2012) settings, or both. Most studies have also noted that effective management of pain can be achieved with adherence to the WHO ladder recommendations for management. An unexpected consequence of understanding available pain prevalence data is that it may also provide reassurance to some individuals as these statistics suggest that there is a proportion of patients with advanced cancer who do not report pain and that although pain is a significant problem for many, it is not an inevitable consequence of a cancer diagnosis. Communication of information about pain to patients and carers benefits from attention to these factors.

Van den Beuken-van Everdingen and co-authors (2007) provided pooled estimates of prevalence for four subgroups of patients based on different stages of the cancer care pathway. They report pooled prevalences of pain of (a) 33% (95% confidence interval (CI) 21–46% (from studies including patients after curative treatment), (b) 59% (95% CI 44–73%) from studies including patients under anti-cancer treatment, (c) 64% (95% CI 58–69%) from studies including patients with disease characterised as advanced/metastatic/terminal, and (d) 53% (95% CI 43–63%) from studies including patients at all disease stages. They also report that for one-third of patients pain was graded as moderate or severe. The pooled estimates of prevalence of pain were greater than 50% for the six cancer groups they examined. Age, continent of origin, and date of study did not contribute to statistically significant variation.

Pain is not only among the most common symptoms in terms of incidence, but it ranks highly with regard to intensity (Tishelman et al., 2007) and distress (Bruera et al., 1991; Portenoy et al., 1994; Hwang et al., 2003). Other common and distressing symptoms for which specific point prevalence data are available include fatigue (Solano et al., 2006; Tishelman et al., 2007), depression (Hotopf et al., 2002; Mitchell et al., 2011), delirium (Leonard et al., 2008), breathlessness (Elkington et al., 2005; Solano et al., 2006; Teunissen et al., 2007; Walke et al., 2007; Currow et al., 2010), disturbed bowel function (Clark et al., 2012), and psychosocial distress (Addington-Hall et al., 1998a; Hynninen et al., 2005; Jacobsen et al., 2005; Averill et al., 2007; Holland et al., 2007; Blinderman et al., 2008; Hill et al., 2008).

The evidence suggests that neuropsychiatric symptoms and syndromes are also particularly common toward the end of life (occurring in up to one in two patients) (Derogatis et al., 1983) and that under-recognition, misdiagnosis (Fallowfield et al., 2001), and under-treatment persist (Lloyd Williams et al., 2003). In the setting of far-advanced cancer anxiety (Roth et al., 2007), sleep disorders (Mercadante et al., 2004), post-traumatic stress disorder (Breitbart, 1995; Leonard et al., 2009), demoralization (Kissane et al., 2001), and suicidal ideation have all been identified as prevalent and distressing neuropsychiatric syndromes.

Much has been written in recent decades about the mandate for optimum care for all dying patients at the very end of life (Field and Cassel, 1997; National Institute for Clinical Excellence, 2004; WHO, 2007; Martin-Moreno et al., 2008) (see also Chapter 1.1). Recently, health-funding bodies, locally and nationally, have invested in comprehensive programmes to assist generalist and specialist clinicians to improve care for patients in the last days of life, regardless of setting or diagnosis. Epidemiological data about symptoms and health-care needs during this period of life can inform care provision and assist in defining priorities for the education of clinicians.

There are several problems inherent in reviewing data regarding symptoms at the very end of life. Fatigue, weakness, and lack of energy along with dyspnoea and pain are reported as highly prevalent in most studies examining symptoms at this time of life (Teunissen et al., 2007) Many of these symptoms and/or the distress associated with them are amenable to treatment and as a result, symptom prevalence data must be interpreted in the context of an understanding that some or many symptoms are amenable to treatment. It is important to note that point prevalence data alone do not give insight into the level of treatment provided. It must be noted that in some studies, data on symptoms such as fatigue and those symptoms related to delirium are conspicuous by their absence. This may result from bias of a particular tool (some tools have excluded fatigue) or, for example, the absence of delirium in an inventory of symptoms at the end of life. The omission of a symptom from a specific tool will unavoidably bias outcomes, so it is important to understand how a symptom assessment tool has been developed.

Reports from hospice programmes and pain studies suggest that despite the prevalence of symptoms, most deaths can be peaceful (Saunders, 1948; Lichter et al., 1990; Seeman, 1992; Hinkka et al., 2001). Although an early but informative study of this time of life was published in 1904 by Osler (Osler, 1904; Hinohara, 1993), studies on symptoms towards the end of life, especially in the very final days of life, have been rare until recently and often limited to small case series that relied on proxy reports. Table 2.2.4 provides pooled prevalence estimates from a comprehensive meta-analysis of cancer-related symptoms in the last 1–2 weeks of life (Teunissen et al., 2007). Symptoms during the last days or weeks of life have also been captured by other studies (Foley et al., 1995; Pritchard et al., 1998).

Longitudinal, population-based data relating to the experience of specific symptoms occurring at the very end of life are essential for planning service provision and setting training and research agendas but have only rarely been published. A study from Western Australia documented patient dyspnoea in 5862 patients who had rated their dyspnoea over time in a routine collection of symptom data in a clinical palliative care setting. Data were reported from patients over time, with a median of 48 days of data collection, up until the day of death. In the last days of life the proportion of patients with ‘no dyspnoea’ fell to 35% but those who rated dyspnoea as greater than 7 out of 10 rose to 26%. Moderate to severe dyspnoea in patients with respiratory failure was sustained over many months before death, and mild to moderate dyspnoea was also reported in other patients in the months preceding death (Currow et al., 2010). This study is a good example of how data collected in a standardized manner in a clinical setting can be analysed to provide important information about some groups and inform the development of treatment and support interventions, as well as patient, caregiver, and provider educational strategies.

Despite significant advances in palliative care treatments and interventions, there are data that suggest that, at the very end of life, unmet symptom-related health-care needs amenable to palliative care interventions persist (Okuyama et al., 2004; Goodridge et al., 2008; Rustoen et al., 2008; Laugsand et al., 2011). For example, the findings of the Study to Understand Prognosis and Preferences for the Outcomes and Risks of Treatment (SUPPORT) were published in a number of papers in the mid 1990s. This study followed 9105 adults hospitalized in the United States with at least one of nine life-threatening diagnoses. One finding was that proxies reported that 50% of conscious patients were in moderate to severe pain for more than half the time during the last 3 days of life (SUPPORT Principal Investigators, 1995). Epidemiological studies of cancer patients from other countries reveal similar rates of inadequate pain control (Laugsand et al., 2011) and unmet symptom management needs are also evident for patients with non-cancer, life-threatening conditions (Covinsky et al., 1996; Goodridge et al., 2008; Rustoen et al., 2008). As discussed above, unmet symptom management needs are important research findings; however, as previously emphasized, data must be interpreted in the context of access to skilled care as well as the amenability of symptoms to treatment and the ability of multidisciplinary services to meet complex needs (Meuser et al., 2001).

Another aspect of the physical experience that has been rated as highly important by cancer patients is the area of communication, consciousness, and mental acuity towards the end of life (Steinhauser et al., 2000). Few epidemiological studies have reported on the longitudinal trends in the level of consciousness towards the end of life. The large National Mortality Followback Study in the United States addressed many aspects of health care including the end-of-life experience (Seeman, 1992). This study sought the perceptions of family carers in regard to decedents and, with respect to cognitive function at the end of life, reported that 68.9% of patients ‘never or hardly ever’ had trouble in recognizing family members or friends during the last year of life (Seeman, 1992). Delirium is reported to frequently accompany the last hours of life for patients with cancer and non-cancer-related illnesses (Conill et al., 1997; Teunissen et al., 2007), but, as with other symptoms, such data must be considered in the context of the availability of effective management (Seeman, 1992; Lawlor et al., 2000).

Functional decline in the months before death has been described in several studies, including one by Glaser and Strauss (1968) that described the trajectories of dying. More recently, Lunney et al. (2003) described four general patterns of functional decline (see Fig. 2.2.5). Of note, these ‘general patterns’ of care needs are supported by detailed epidemiological data, in this case from US Medicare data sets (Lunney et al., 2003). It is important to acknowledge that while general trends exist, these trajectories are not necessarily applicable to individual patients. Functional decline can be viewed to some degree as a ‘proxy’ for health-care needs in that it has implications for personal care, physical support in the home, and caregiver supports. Clearly, on a national, regional, or institutional level it is important for health-care planning to accommodate the care needs of populations implied by trajectories of functional decline.

On an individual level, this type of epidemiological data about performance status and function can assist in facilitating discussion about an individual’s projected symptom experience and can help in answering such questions as: ‘Is it likely I will have months lying in bed unable to speak or get up?’, ‘Is it likely I will need someone to look after me?’, or ‘Is it likely I will be able to stay at home?’. As an example, it may appear almost redundant to many clinicians to state the general differences between the functional decline and care needs of a patient with Alzheimer’s disease and the needs of a patient with acute myeloid leukaemia.

For patients and carers this is most often far from obvious, and rather than being left to draw their own conclusions based on what they have observed in others, in the literature, and/or the media, patients and carers may benefit from timely and skilled communication with a health professional who has a good understanding of common functional trajectories.

Some information about population health-care needs at the end of life has been provided by a number of large, well-designed epidemiological studies (Seeman, 1992; Addington-Hall et al., 1995; McNamara et al., 2007; Coupland et al., 2010; Seow et al., 2012). Further information about the longitudinal trajectory of functional and other needs would be helpful for service planning.

With an increase in the potential to analyse large health-care administrative data sets through record linkage, there is increased discussion about the role that information extracted from these datasets can play in assessing the quality of care. Many studies of this nature have been carried out in recent years looking at the period towards the end of life including large epidemiological studies that have provided information about health-care utilization over the period near to the end of life. Although a full review of these studies is beyond the scope of this section, these include studies looking at costs of care (Fassbender et al., 2005; Yabroff et al., 2007), palliative care service use (Fassbender et al., 2005; Rosenwax et al., 2006), place of death (McNamara et al., 2007), and hospital and community health experiences including physician visits, procedures, intensive care admissions, emergency department presentations, and length of stay in hospital (Kaul et al., 2011; Unroe et al., 2011).

A recent Canadian study illustrates the types of findings possible in these studies. This study examined the public provider costs associated with the last 6 months of life for all cancer deaths in Ontario. Investigators found that 75% of costs associated with end of life and palliative care were incurred in the acute hospital setting (Walker et al., 2011). Other investigators found that 26.6% of costs were borne by family/carers (Dumont et al., 2010).

Increasingly, investigators have sought to explore the question of how to use data points within health administrative data sets to develop indicators of quality of care at the end of life and, to achieve this, have used methods including seeking expert opinion and patient and carer feedback about such indicators through focus groups (Earle et al., 2005; Grunfeld et al., 2008). It is becoming clearer with time that indicators could, with widespread use and further work on appropriate benchmarks for various regions, be used within and among health services to inform planning and predict service needs (e.g. home care, emergency room services, etc.) (The Dartmouth Atlas of Health Care, 2007). The ability of indicators in these data sets to truly capture the appropriate endpoints relating to quality of care remains complicated. Limitations include the challenge inherent in non-subjective indicators to provide a clear representation of many aspects of the patient’s experience. These data sets cannot generally answer questions such as whether patients’ preferences were honoured, and whether patients and carers were satisfied with care (Grunfeld et al., 2008). In summary, while there is a wealth of data related to health-care utilization there is a need to both refine validated quality indicators, and to develop systems to measure and record patient-reported outcomes to ensure information is meaningfully recorded to inform patient-centred care (Grunfeld et al., 2008; Higginson, 2013).

A single death affects many others in terms of informal caregiving and grief (see Chapter 17.6). Documentation of the caregiver’s experience towards the end of life is therefore an important aspect of the epidemiology of the end-of-life experience.

Many caregivers willingly provide care, and indeed report that they find the role of carer a rewarding and an important part of family experience (Andrén et al., 2008). Notwithstanding this, research conducted across diagnoses and in various countries has tended to focus on the significant demands and burdens that arise from caring for patients with life-threatening illnesses and despite the positive aspects of the role, one in 13 carers in one study (7.4%) indicated that they would not take on the caring role again (Currow et al., 2011). Caregiving has been shown to affect both the physical (Christakis et al., 2006) and psychological (Zivin et al., 2007) health and the social and financial (Carmichael et al., 1998; Emanuel et al., 2000; Berecki-Gisolf et al., 2008) situation of caregivers (Aoun et al., 2005; Burton et al., 2012). Population-based sampling such as that undertaken by Addington Hall et al. (1998b) can provide important information about carers of patients with a particular diagnosis. For instance, the study by Addington-Hall et al. presents data from a representative sample of carers for a person who died from stroke and state that 43% of carers reported to have needed more assistance with personal care and 31% reported unmet needs regarding financial issues. For further information about caregivers, readers are referred to Chapters 6.1, 6.2, and 16.1 and to studies such as the 1999 National Long Term Care Study (NLTCS) (Wolff et al., 2007) and the SUPPORT study (Covinsky et al., 1996).

In addition to the impact of caregiving on physical and psychological health, from a societal and epidemiological perspective, the financial and social impact of caregiving is also significant. As an example, in the SUPPORT study, in which care in the United States was investigated, it was reported that 31% of families caring for patients near the end of life lost most or all of the family’s savings and, in 20% of cases, the caregiver had to resign from work or make another major life change to continue to provide care (Covinsky et al., 1996). Informal carers and their households are certainly at risk of suffering loss of income, and indeed epidemiological studies have demonstrated the significant impact that caregiving has on workforce participation (Carmichael et al., 1998; Berecki-Gisolf et al., 2008). Such data raise questions about availability of resources and policy to address the needs of carers. The importance of assessing the impact of caregiver interventions on carer burden beyond financial and health concerns is supported by existing literature. There is evidence suggesting that carers regard information, emotional support, practical care, and patient comfort as most important (Addington-Hall et al., 1998b; Burton et al., 2012).

The array of studies reporting on the needs and experiences of caregivers from high-income countries contrasts with the paucity of data relating to this from low-income countries. While less research has been done in low-income areas, an interesting small study by Grant et al. reported that for advanced cancer ‘the emotional pain of facing death was the prime concern of Scottish patients and their carers, while physical pain and financial worries dominated the lives of Kenyan patients and their carers’ (Grant et al., 2003). While more data are needed worldwide on the caregiver’s experience, the overlapping needs of caregivers and the contrasting needs reported in this study provide some insight into the spectrum of needs and the disparities that exist among caregivers.

Death is laden with emotional, social, and cultural significance (see Chapter 2.5). A diverse spectrum of beliefs exists about the spiritual/existential and cultural aspects of death and the period prior to, and after, death. Despite much being discussed in the popular media and the fact that many hold strong beliefs relating to this time of life, and the time before and after death, there is a paucity of epidemiological data concerning perceptions about the ‘metaphysical’ domains of experience near to the end of life with few publications in the peer-reviewed literature. Cultural factors are important in relation to symptom experience, distress, and communication at the end of life and addressed elsewhere in this textbook. Well-designed epidemiological studies investigating cultural aspects of the end-of-life experience within and across different social, geographical, and cultural groups would serve to illuminate this aspect of the end-of-life experience.

The study of the epidemiology of the end-of-life experience is an evolving and important field with an increasing number of studies being published that shed light on the experiences of those within populations who are nearing the end of life, and the experiences of caregivers. The use of validated tools, carefully designed studies, and record linkage will, it is hoped, shed more light on this important area over time. It is important for epidemiological enquiry that the symptoms and health-care needs at the end of life be a subject of focused study throughout the world if health policy is to truly reflect the spectrum of needs of individuals who are near to the end of life.

Professor Ingham’s research for this publication was undertaken, in part, with funding support from the Cancer Institute New South Wales Academic Chairs Program. The views expressed herein are those of the authors and are not necessarily those of the Cancer Institute of New South Wales. The authors wish to acknowledge the work of Dr Paula Mohacsi, PhD, MBA, MSc (Ed Studies) RN in the preparation of the manuscript.