9

Directions for Research
to Improve Care at
the End of Life

More light!

Goethe, last words, 1832

Goethe's dying words were not a plea for more research on the end of life, but they are fitting nonetheless. From the cellular to the social level, much remains to be learned about how people die and how reliably excellent and compassionate care can be achieved. Important, unanswered questions exist about the fundamental physiological mechanisms of the symptoms that cause so much suffering among dying patients and about the kinds of interventions that will relieve these symptoms. Basic epidemiological information on how people die is limited, and the influence of attitudes and beliefs on people's experience of dying and on caregiving practices is little charted. In addition, a better understanding of the reasons for the inadequate application of existing knowledge would help in identifying organizational, economic, and other incentives for the provision of accessible, effective, and affordable care at the end of life.

Deficiencies in knowledge and its use have important consequences. Hundreds of thousands of Americans face death each year experiencing severe symptoms and distress, including pain, fatigue, shortness of breath, nausea, wasting, delirium, anxiety, and depression. Although complete relief of such distress cannot be expected, relatively limited investments in cancer and pain research have already had important benefits, as described below. Furthermore, in this era of evidence-based medicine, the outcomes of clinical research are increasingly being used to determine what care is covered by public and private health insurance. In various ways, then, a lack of research on palliative therapies may translate into lack of resources to care for dying patients.

The shortage of scientific knowledge reflects a lack of clinical interest in high-quality palliative care, interest that would serve as a powerful stimulus to research. The committee argues that caring for dying patients can be personally rewarding and intellectually challenging. It demands that clinicians learn and apply a wide variety of skills and knowledge. For example, effectively balancing the benefits and adverse effects of multiple drug therapy in a metabolically compromised patient is clearly a complex and exacting task. Dying patients are not people for whom "nothing can be done" either now or in the future. Basic biomedical research into the pathophysiology of symptoms and mechanisms of symptom relief—following the model used in pain research—provides a promising foundation for improved patient care and a tighter focus for clinical research on care of patients likely to die.

This chapter focuses on directions for extending the knowledge base for effective, reliable care at the end of life. Unlike other chapters, this chapter uses certain technical terms that may not be understandable to some readers; it does so in an effort to underscore research gaps and opportunities. The following sections consider a symptom-based strategy for biomedical research, directions for behavioral and health services research, ethical issues in research involving dying patients, some important methods issues in end-of-life research, and barriers to research and consequent opportunities for leadership from the nation's research establishment. Although the committee recognized the methodological, ethical, and economic difficulties that can complicate end-of-life research, it did not believe that these problems are insurmountable if leaders recognize the relevance and promise of symptom and other research intended to help reduce suffering at the end of life. Such leadership would help attract more bright young trainees who could readily master the techniques required for the conduct of that research, both in the laboratory and at the bedside.




A SYMPTOM-BASED STRATEGY FOR
BIOMEDICAL RESEARCH

Earlier chapters of this report have described the symptoms that cause significant distress for many dying patients. Although deficits in the use of existing knowledge need to be corrected, it is also important to extend the base of knowledge about the pathophysiology of symptoms and the mechanisms for symptom relief and prevention. Pain is the most widely noted and feared symptom and has been the subject of an active and productive research strategy. This report particularly challenges the biomedical research community and its supporters to direct new intellectual energy to better understanding and countering the mechanisms of other physical and emotional consequences of advanced diseases (and treatment side effects), particularly fatigue, wasting, shortness of breath, nausea, confusion, anxiety, and depression.

Successful biomedical research depends on the interaction of basic scientists and clinical colleagues. Medical advances in managing pain and in understanding and treating chemotherapy-induced nausea reflect such interaction (MacDonald, 1995; Max, 1996). The committee concluded that progress in understanding and managing other symptoms would be advanced by applying and extending the research strategies that have been developed to investigate pain.

The committee also concluded that although palliative care research has received little emphasis in the past 20 years, modest investments appear to have paid substantial dividends in the prevention and relief of symptoms experienced by cancer patients. Advances resulting from cancer research—some of which benefit those who die of cancer as well as long-term survivors—have occurred in several areas (see, e.g., Holland and Rowland, 1990; Breitbart and Holland, 1993; Doyle et al., 1993; AHCPR, 1994a; Gavrin and Chapman, 1995; Portenoy and Kanner, 1996). These include

• managing cancer pain;

• managing nausea and vomiting associated with the use of chemotherapeutic drugs;

• using chemotherapy, radiation therapy, surgery, and other procedural interventions to palliate symptoms in selected cases;

• managing psychiatric symptoms by pharmacological and behavioral techniques;

• addressing special social and developmental issues (e.g., school, play) for children with cancer; and

• incorporating quality-of-life measures in cancer clinical trials.

Several of the above advances have improved symptom control and psychosocial well-being without adding substantial costs for care. For example, the conclusion that opioids and many other symptom-controlling drugs can be effectively administered subcutaneously when oral administration is not effective or feasible has simplified the care of patients and avoided the use of more expensive intravenous or intramuscular therapies (Bruera et al., 1988). Patients and families can readily learn to administer hypodermic injections, thus enabling them to spend longer periods at home with reduced costs and improved quality of life.




Directions for Pain Research

The discussion below begins by presenting pain as a model for symptom research. It then suggests how this model might be applied in three other areas: cachexia-anorexia syndrome, dyspnea, and cogntive/emotional symptoms. Although the physiological details will vary for different symptoms, the general strategy for directing research resources appears broadly applicable.




Pain as a Model for Symptom Research

Based on the committee's experience and investigations for this study, pain appears to be the only major symptom where a constellation of basic scientists and clinicians have worked together to understand the pathophysiology of the symptom and to develop logical therapeutic interventions on the basis of research observations. The most extensive body of work supported by NIH involves pain. In 1995, building on several years of work, 10 units of NIH announced an initiative "to stimulate and foster a wide range of basic and clinical studies on pain" (NIH, 1995, p. 1). Among the topics and areas mentioned for investigation, one singled out effectiveness of "biobehavioral pain management in terminally ill and dying patients" (NIH, 1995, p. 9), but many other topics (e.g., relationship to other symptoms, pain assessment strategies, comparison of different prevention and alleviation strategies) also should have some relevance for this patient population.

In FY 1996, according to one estimate, NIH spent about $70 million on pain research out of an approximate total budget of $12 billion (John Retzlaff, NIH, personal communication, May 27, 1997). One researcher has suggested that this relatively modest public investment "primed the pump" for a substantial research investment by industry, which for many years had confined its work to anti-inflammatory drugs and short-acting opioids (Mitchell Max, personal communication, September 17, 1996 and March 18, 1997). For example, NIH-funded research into the relationship between opioid blood levels and pain relief led industry to develop sustained-release opioid tablets, skin patches, and patient-controlled analgesic infusion pumps. Other research on the pain-inhibiting neurotransmitters has also led to new pain treatments (Max et al., 1992; Wall and Melzack, 1994; Yaksh and Malmberg, 1994). In another previously neglected clinical area, pain related to nerve injury, NIH-supported development of animal models and clinical trial methods in pain have recently stimulated many companies to synthesize candidate compounds and initiate clinical trials. A similar industrial "second wave" of research might be predicted if public-supported research into other symptoms can suggest a physiological rationale for treatment and show that clinical trials can separate active from inactive treatments.

The many successes of clinical pain research in the past two decades—ranging from spinal opioid therapy for surgical and cancer pain to sumitriptan for migraine—rest on a foundation of basic neuroscience that could be applied to many other symptoms. Thousands of researchers have been studying specialized nerve endings and connecting circuits in the spinal cord and brain that give rise to pain and other sensations. Dozens of neuro-scientific techniques that have been developed to study pain and opioid effects could be adapted to the study of other symptoms. These techniques include mapping of neurotransmitters, identifying the changes in cellular and gene activity triggered by these impulses, developing animal models, and three-dimensional imaging of the brain activity triggered by symptoms (Hunt et al., 1987; Draisci et al., 1991; IOM, 1991; Kandel et al., 1991; Capechhi, 1994; Takahashi et al., 1994).

In the field of pain, many additional promising investigative avenues can be identified. These include:

• improving techniques for drug delivery (Storey et al., 1990; Berti and Lipsky, 199; Cleary, 19975; Coluzzi, 1997).

• introducing nonsteroidal anti-inflammatory drugs, which selectively affect the inflammatory components of pain and do not interfere with other essential body activities mediated by prostaglandins (Akarasereenont et al., 1994; Hayllar and Bjarnason, 1995).

• introducing more refined, receptor-specific opioids. Pain inhibition may be mediated by an opioid receptor type that does not influence sedation, bowel function, or respiratory function (Bigelow and Preston, 1995; Breitbart, 1996; Portenoy, 1996).

• identifying novel analgesics that influence nonopioid systems in the central nervous system (Eide et al., 1995; Rang and Urban, 1995; Nelson et al., in press). Particularly noteworthy is the recent discovery of multiple excitatory systems that maintain and enhance particularly severe forms of chronic pain. Current laboratory studies on animal models and initial clinical trials suggest that new types of nonopioid analgesics can successfully influence these pain-stimulating systems.

• strengthening developmental pharmacology intended to maximize benefits and minimize harms by taking into account patient differences in age, race, gender, weight, metabolic status, and other relevant factors.

• identifying prognostic factors to help guide preventive strategies and linking symptoms to quality of life measures to help guide care priorities (Donnelly and Walsh, 1995b).

• improving knowledge of the clinical features and treatment of pain arising in the viscera. Recent work in animal models suggests that visceral pain has some mechanisms that are distinct from other types of pain and that could provide opportunities for therapies. These mechanisms include the sensitivity of primary visceral afferents to kappa opioids (Sengupta et al., 1996) and transmission in the posterior funiculus of the spinal cord, an area readily accessible to surgical or electrical interventions (Al-Chaer et al., 1996).





Promising Leads for Research on
Cachexia-Anorexia-Asthenia Syndrome

Palliative care experts have identified the symptom complex of cachexia-anorexia-asthenia as a major complaint of patients, who rank the symptoms first or second in prevalence in many cancer studies (Bruera, 1993; Portenoy et al., 1994; Seale and Cartwright, 1994; Donnelly and Walsh, 1995b). In addition to cancer, this syndrome is common in many other illnesses, including AIDS, advanced congestive heart failure, and chronic obstructive pulmonary disease. It is also common in children with chronic lethal conditions (e.g., congenital anomalies) who are not actively dying. It creates dependency on others, with consequent emotional and fiscal cost. Moreover, cachexia can actually result in the death of many patients when sufficient protein to maintain vital body functions is lost.

Those clinicians and scientists who are not familiar with the concepts of preventable suffering and palliative care may assume that wasting, loss of appetite, and weakness are inevitable in the dying process. Some studies, however, suggest that these symptoms are amenable to treatment that affects patient's quality of life and, perhaps, survival time. Observations indicate that cachexia-anorexia may involve a cytokine cascade induced by the disease process or occurring in reaction to it (Grunfeld and Feingold, 1996; Mitch and Goldberg, 1996; Jaskowiak and Alexander, 1997; Moldawer and Copeland, 1997).¹ The specific influence of cytokines may vary from disease to disease. For example, in cancer patients, tumor products may play an important causal role with cytokines acting in an indirect supportive mode (Todorov et al., 1996; Tisdale, 1997). As research findings on the relative influence of cytokines and chemical products induced by disease processes emerge, interference with this process should allow patients to have an improved quality of life as they die and may prolong life for some. As with pain, when the pathophysiology of the symptom complex is more clearly understood, specific interventions can be formulated and tested. Starting points for research should include

• more precise descriptive terminology for the cachexia-anorexia-asthenia complex and the creation of working groups to create a "common language" for research studies;

• development of a standard symptom assessment format for studies on cachexia-anorexia-asthenia, similar to what is available for pain;

• an assessment of the costs of the cachexia-anorexia-asthenia complex; and

• investigation of particular therapy options based on fuller understanding of pathophysiology and the experience of symptoms as reported by patients and families.

In recent years, it has been demonstrated, for the first time, that the trajectory of anorexia and weight loss can be altered using hormonal agents (Bruera et al., 1990; Loprinzi et al., 1990). More specific interfering agents, acting directly on cytokines, are now available in the laboratory. The promise is that such agents may improve appetite and add weight in cachectic patients. The cytokine cascade, which contributes to the cachexia-anorexia syndrome, is becoming better understood, leading researchers to investigate inhibitors of cytokines or of agents that interfere with cytokine receptor expression and function (Plata-Salaman, 1995; Erickson and Hubbard, 1996; Laviano, 1996; Meydani, 1996). Cytokines are induced by cancer tissue and, under certain circumstances, may act as tumor inhibitors. The complex interaction of cytokine production <—> cytokine inhibition can be studied in animal tumor models. In addition, cytokines may act indirectly by inducing the production of other chemicals that directly influence nutrient metabolism. Identification of these substances may produce additional therapeutic leads. Another promising lead relates to the study of specific nutrients, for example, omega-3 fatty acids and branched-chain amino acids (Endres et al., 1989; Downer et al., 1993; Karmali, 1996). Finally, growth hormone or IGF-1 therapy may enhance well-being in some patients with chronic wasting (LeRoith, 1997).

Relative to other pharmaceutical interventions, the current drugs available for the control of cachexia-anorexia range from modest in cost (short-term use of corticosteroids) to moderately expensive (progestational agents). The cost of more specific interventions remains to be determined. Overall, given the number of patients affected, the costs of their dependency, the distress caused by the syndrome, and the intriguing findings of recent research, this is a priority area for collaborative investigation.




Directions for Research on Dyspnea

Dyspnea can be defined as "subjective difficulty or distress in breathing," often experienced as shortness of breath (Stedman's Medical Dictionary, 1995). Frequently occurring in combination with other symptoms, dyspnea is a common source of suffering for patients in the last months of life. Dyspnea can result from malignancy metastatic to the lungs or pleura, from heart failure, from chest wall weakness due to neuromuscular disease, or from primary pulmonary disease.

Until recently, if the cause of dyspnea could not be reversed, help for dyspneic patients was limited (Davis, 1994). Several lines of research now suggest that resigned acceptance should not govern the thinking on dyspnea. Promising areas for further investigation include (1) opioids; (2) respiratory stimulants; (3) oxygen therapy; and (4) support for respiratory muscles.

Opioids already have a role in controlling dyspnea similar to their use in the management of cancer pain. They moderate the reflexive drive to breathe and thereby decrease patient awareness of dyspnea; they may also improve breathing efficiency and exercise endurance (Doyle et al., 1993). Questions that warrant investigation include:

• How does morphine influence the pathophysiology of dyspnea?

• Are some opioids more effective than others for relieving dyspnea? Which have the optimal therapeutic margin? Do nebulized opioids have a role? Some pilot studies suggest that inhaled morphine relieves the dyspnea of metastatic disease in the lung (Davis, 1994). Would it show similar effects for congestive heart failure? Is there evidence for a specific effect on pulmonary receptor sites?

• If, as it appears to be, allowing a degree of opioid-induced hypercapnia (high levels of arterial carbon dioxide pressure) may be effective in relieving dyspnea, what levels of arterial carbon dioxide are safest and as free as possible of undesirable effects, such as clouding of consciousness?

• Is processing of the sensation of dyspnea in the brain organized in a way that would potentially allow one to relieve the perception of shortness of breath without suppressing the automatic drive to breathe?

Respiratory stimulants also warrant investigation in dyspnea therapy. Several agents, including phosphodiesterase inhibitors, progestational agents, and methylphenidate have stimulatory effects on respiratory centers in the medulla (Aubier et al., 1983; Mikami et al., 1989). Such effects prompt the question: Will these agents help to improve respiratory activities in chronically ill patients, or will the stimulatory effect on the respiratory center undesirably increase the perception of dyspnea?

Oxygen is clearly indicated in patients who are hypoxic (oxygen-deficient). Investigators might explore whether there is a role for oxygen therapy in nonhypoxic patients under any circumstances.

Support for respiratory muscles is another area for therapeutic investigation. Patients with advanced chronic illness often have the cachexia-anorexia syndrome and have lost a great deal of somatic protein. Their diaphragm and the external accessory respiratory muscles may be particularly vulnerable to wasting and fatigue. One pertinent research question, then, is whether improvement in patient nutrition, with consequent improvement in respiratory muscle content and strength, might limit dyspnea. Another question is whether newer techniques of ventilatory support might assist failing respiratory muscles and relieve the dyspnea.




Cognitive and Emotional Symptoms

In the arena of cognitive and emotional symptoms, a concerted effort to develop a multidisciplinary biomedical and psychological research strategy that focuses on the dying patient could bring progress in a number of important areas, including delirium and milder confusional states, depression, anxiety, sleep disorders, hallucinations, and suicidal ideation (see, e.g., Holland and Rowland, 1990; Breitbart and Holland, 1993; Hyman and Nestler, 1993; Breitbart et al., 1995). The biology of these symptoms as they affect patients with advanced life-threatening illnesses has been little examined; even their prevalence is not well charted.

For example, few studies of the prevalence of depression have investigated patients with far advanced disease (see, e.g., Hinton, 1972; Bukberg et al., 1984; Chochinov et al., 1994). The same holds true for delirium, which is common in advanced cancer patients and is important as a source of distress to families and health care personnel as well as patients (see, e.g., Lipowski, 1987; Fainsinger and Young, 1991; Bruera et al., 1992; Smith et al., 1995). Similarly, scientific study of the biology and importance of cognitive and emotional symptoms is limited for patients with severe congestive heart failure, chronic obstructive pulmonary disease, and diabetes (see, e.g., Littlefield et al., 1990; Leedom et al., 1991; Cowley and Skene, 1994; Hawthorne and Hixon, 1994).

Certainly, some anxiety and depression are understandable responses to serious illness and loss. This contributes to disagreement about the nature and prevalence of clinical depression in the population of patients who are dying (Bukberg et al., 1984; Kathol et al., 1990; Chochinov et al., 1994). Others disagreements involve diagnostic criteria, treatment strategies, and assessments of the benefits and burdens of particular therapeutic agents.

General directions for biomedical research in cognitive and emotional symptoms in patients approaching death include some elements, primarily methodological, common to those proposed above for other symptoms. Directions for research include

• use of uniform descriptive terminology and classification schemes where they now exist and development of such terminology and taxonomies of disorders where they do not now exist;

• development, use, and refinement of reliable, valid, and practical symptom assessment tools and measures for studying the prevalence and severity of psychological symptoms in patients with advanced disease and for evaluating interventions appropriate for such patients;

• studies of the prevalence of neuropsychiatric symptoms and syndromes in patients with different diseases and circumstances (e.g., age, sex, care setting) and the distress created by these symptoms as reported by patients and families;

• examination of the relative contributions to diminished mental functioning and patient well-being of specific disease processes and the side-effects of therapeutic interventions to treat a disease or relieve some of its symptoms;

• investigation of the biochemical, immunological, neurobiological, and other physiological effects of the stresses imposed on patients and those close to them by life-threatening illnesses;

• assessment of the costs of different neuropsychiatric syndromes and symptoms as they affect patients with advanced disease; and

• comparison of alternative pharmacological and nonpharmacological therapy options (singly and in combination) including novel uses or combinations of existing therapies.

Social science and health services research, some directions for which are outlined below, offer additional opportunities for understanding and reducing the cognitive and emotional toll of advanced disease. In all areas, researchers need to consider (1) the extent to which different psychological symptoms coexist and interact with each other and with physical symptoms and (2) the patient and the family as a unit for purposes of assessing and managing symptoms, especially psychological symptoms (Breitbart et al., 1995).




SOCIAL, BEHAVIORAL, AND HEALTH SERVICES RESEARCH

A stronger foundation of scientific and clinical knowledge about symptom mechanisms and effective therapies for the dying patient is important but insufficient to guide improvements in care at the end of life. More needs to be understood about the extent and severity of physical, emotional, financial, and other adversities experienced by people dying from common diseases; about the possibility of spiritual growth and other valued experiences at the end of life; and about the attitudinal, organizational, legal, cultural, and other factors that impede the application of existing knowledge and principles of care. This kind of understanding will help in developing strategies that will make it easier for individuals and organizations to provide good care than to fail to do so.

Some research needs are so basic that it is surprising they still exist. For example, as Chapter 2 made clear, even elementary descriptive and epidemiological knowledge about how and where people die is very limited. Likewise, more is assumed than is known about what people want as they approach death; what is meaningful to their families and others close to them; and how the care and support that is valued can be effectively and efficiently organized. For example, although interdisciplinary care allows people with special training and experience to help patients with the many different burdens created by life-threatening illness, the benefits and costs of particular configurations of interdisciplinary expertise must still be evaluated. Similarly, continuity of care is widely viewed as desirable, but firm knowledge of how and to what extent it makes a difference is surprisingly limited (Donaldson, 1997). Evidence is beginning to accumulate about the contributions of continuous quality improvement strategies, but their value in end-of-life care is essentially anticipated rather than demonstrated. Chapter 5 identified the need for better conceptualization and measurement of outcomes meaningful to dying people and those close to them and described a number of the challenges of developing reliable, valid, and practical information to guide patient care and provide accountability.

Chapter 6 noted only limited information is available on the cost of dying in different ways, at different ages, in different places, and with different care. Research on the impact of different kinds of financial incentives is virtually nonexistent as it involves dying patients. Many welcome, while others worry about, the growth of managed care and for-profit health care institutions, but the consequences for dying patients are almost entirely unexamined. Demonstration projects testing the effects on end-of-life care of alternative practitioner and provider payment mechanisms are virtually unknown.

The next section identifies some methods issues that particularly complicate social, behavioral, and health services research. Following that is a discussion that suggests directions for future research. Although this discussion is not comprehensive or specific enough to direct particular projects, it provides a framework to be debated and refined.




Methods Issues in End-of-Life Research

No methods issues may be unique to research on care of the dying, but the mix of problems and the pivotal nature of certain problems create particular difficulties concerning the amount and quality of research in this field. Further, as pointed out in Chapter 5, many of the questions that face researchers also arise for administrators, clinicians, policymakers, and others attempting to measure, monitor, and improve the quality of end-of-life care and manage resources. Although some research questions may be addressed using the gold standard of clinical research, the randomized controlled trial, improvements in care may well require careful but innovative use of nonrandomized and sometimes uncontrolled settings. Methods to interpret and apply these research strategies, which may overlap with the methods of continuous quality improvement, merit further development.

For purposes of identifying patients, organizing services, and evaluating outcomes, one fundamental step would be the development of clearer concepts, definitions, and markers of the "last phase of life" as it has meaning in physiological, clinical, psychological, social, and cultural terms. Markers of the last phase of life may variously include (1) formal diagnoses; (2) easily observed physical signs and symptoms; (3) more subtle indicators from patient history, physical examinations, and tests; (4) scores generated by prognostic models; (5) acknowledgement by patients, families, or clinicians; or, (6) for some purposes, retrospective identification following death of a time or clinical juncture before death (e.g., one week before death, last hospitalization).

Although the research, policy, ethical, and cultural merits, limitations, and implications of different markers of the dying phase need attention, an inescapable difficulty is that the anchor of end-of-life studies requires decisions about a prospectively uncertain end point. Will research questions be tied to timing relative to expected death (e.g., when a patient is predicted to have a 10 percent chance of living six months) or to timing relative to a particular event (e.g., when a patient receives a specific diagnosis or enrolls in hospice care)? Another complication is that the ultimate end point—death—varies somewhat with treatment itself as illustrated in some of the illustrative cases presented in Chapters 3 and 4.

For economic analyses, most cost accounting involves a fixed period of time or an episode of hospitalization. Here, at least for some purposes, studies count costs from a certain stage of illness through to death or for a particular period of time. How an analyst is to account for variation related to the starting point (e.g., when a diagnosis is made, which may be early or late in the course of some diseases) or the effect of treatment decisions (e.g., when the patient receives therapies that both relieve symptoms and prolong life) has not been worked out. Strategies that involve multiple data collection points add flexibility and information but are burdensome and expensive. For example, much end-of-life research requires interviews with patients or surrogates, as well as, perhaps, interviews with caregivers and reference to patient records. Such research is expensive and complicated. Thus, alternatives and trade-offs need to be explored.

Researchers also face complexities in studying patients with multiple symptoms that may interact with each other. Such interactions are not well described or understood. Similarly, researchers may find it difficult to distinguish symptoms caused by the patient's illness from those resulting from treatments.

Yet another problem with studies of people near the end of life involves missing or potentially distorted data. Many patients in these studies die without being questioned; others cannot be questioned directly about their symptoms, concerns, or attitudes because their illness or treatment or both have left them confused, weak, or unconscious. Their condition may also change frequently, which limits continued informed consent and participation in some studies. Missing data requires examination of possible imputation or substitution strategies, for example, using surrogates, but little research or analysis is available to guide choices.

In addition to the problem of whether a patient is functioning cognitively, issues related to diversity in language, religion, ethnicity, and other cultural attributes arise. How, for example, do and should measurement instruments and strategies take into account such diversity among patients and families?

Another challenge is that patient care occurs under many different financing arrangements and in myriad locations, many of which involve fragmentary and poorly developed record systems. Data systems for the nation's thousands of health insurance plans ordinarily do not distinguish between individuals who died and those no longer in the plan for reasons such as retirement or switch to another health plan so tracking people across time to determine outcomes can be difficult. In addition, current systems for measuring and monitoring the quality of care include very little that is directly related to care at the end of life. Many physical, emotional, and social issues of central concern (e.g., symptoms, counseling, satisfaction, impoverishment, meaningfulness) are not ordinarily part of patient data systems. Priorities for the development, testing, and refinement of reliable, valid, and practical quality measurement tools include not just the experience of symptoms but their burden as felt by patients. The point of citing such difficulties is not to discourage research but to encourage a realistic understanding of its challenges.




Directions for Investigation

As a starting point for establishing priorities and formulating a research agenda, the committee suggests a number of important areas for investigation. These investigations would draw on the expertise of epidemiologists, psychologists, economists, sociologists, anthropologists, health services researchers, and others. The list is only suggestive of the array of knowledge necessary to improve care at the end of life and the rich arenas for research by those in the behavioral and social sciences, humanities, epidemiology, and related fields.




Epidemiology of Dying and Death

Although refinements in the reliability, validity, completeness, and timeliness of "vital statistics" on the cause of death are a continuing objective, those interested in improving care at the end of life have a particular interest in better information on where people die and what paths they follow through different settings. Data on numbers, causes, and places of deaths and basic demographic characteristics (e.g., age, race) are routinely gathered by government, but information on dying pathways must come from special studies. In the absence of integrated, comprehensive, computer-based patient records, such studies will generally be small-scale and limited to a few institutions or communities and will thus present problems of generalizability. Nonetheless, a better descriptive understanding of common pathways during dying is possible for different settings of care.

An additional set of questions focuses on variations in the clinical pathways of dying—including the duration and pattern of the dying process (e.g., steady decline or relative stability marked by crises); symptom prevalence and level of distress; functional and cognitive impairments; and variation within and across populations of patients by diagnosis, demographic, and other characteristics. Because most people who die are older adults, much research should focus on this large population. Other populations are important but may be difficult to study. Little is known about the experiences of special populations such as younger people with preexisting physical and mental disabilities; children at various ages; people without insurance; those in the inner city and very rural areas; and religious, ethnic, and other minorities. Given the special sorrows and stresses created by life-threatening illness in children, the developmental aspects of illness and patient care involving neonates, infants, children, and adolescents need attention.




Delivery, Financing, and Improvement of Health and Other Services

As in much health services research, a useful descriptive step is mapping the nature and extent of variation in end-of-life care across clinicians and care settings. The more ambitious task is to investigate whether and how differences in outcomes and costs relate to differences in the clinical problems, characteristics, and preferences of dying patients; the settings and processes of care; and the mechanisms of structuring and financing care. Such research will help clarify the barriers to organizational excellence (or adequacy) in care at the end of life. To the extent that clinician attitudes, skills, and knowledge—for example, misplaced concern about addiction, ignorance of pharmacological and nonpharmacological strategies for symptom relief, disinterest in dying patients—contribute to inadequate care, then a broad range of techniques for changing attitudes, increasing knowledge, and modifying behavior may be employed. Modest amounts of research from other areas provide some hints about the effectiveness of different strategies (e.g., distributing information is usually ineffective as a sole strategy).

A series of questions can be raised about strategies for identifying problems, structuring accountability, and improving care. For example, are patient or family reports on care more reliable and useful than satisfaction measures in identifying problems and improving care? How well-suited are different quality measures for internal and external quality assessment? Can guidelines for clinical practice, decision-support tools, performance feedback, and similar strategies improve clinical practice and care outcomes? If certain structures and processes can be linked to better outcomes, can they—at least for some purposes—be used as reliable and valid but less costly proxies for outcomes?

Chapters 3, 4, and 5, which outlined the dimensions of care at the end of life and the arrangements for providing and evaluating care, suggested many questions. For example, will new criteria intended to guide hospice referrals for noncancer patients be practical to apply and will they affect levels of hospice use? What happens during the actual practice of home care for dying patients, and what are the effects of different care processes and structures (e.g., home care within or outside a hospice program)? The same general questions can be asked about care at the end of life in nursing homes. Hospital care is becoming somewhat better understood, but documented interventions to improve care remain elusive. Beyond the health care system, do the administrative boundaries separating social, medical, and other services create difficulties, and if so, how might these boundaries be redesigned?

With cost pressures encouraging health care managers to reduce staffing levels and reassign responsibilities from more highly to less highly trained personnel (e.g., from RNs to LPNs or aides), researchers need to investigate the impact of shifts in caregiving personnel on very sick and dying patients. Are patients suffering pain and other symptoms waiting longer for attention? What is happening to the interpersonal aspects of care and to the extent and quality of symptom assessment, management, and charting in patient records?

Similarly, the extent and impact of cost-driven shifts in the site of care for dying patients, which has been scarcely considered, warrants investigation to identify the benefits and burdens for patients and families as well as the impact on organizations delivering and financing care. Other important questions involve the characteristics and effects of different ways of managing transitions from one care setting to another and alternative methods for coordinating care for people with serious, progressive disease. In addition, it would be useful to learn about how strategies to prevent and relieve symptoms affect costs of care over the trajectory of serious illness.

As noted in Chapter 6, only the most global estimates of the cost of care at the end of life are available, and these estimates are largely limited to costs for older individuals covered by Medicare. Future research may be limited by losses in Medicare's administrative database as certain information on interventions and diagnoses is lost in capitated care. Such information is important for analyzing the impact of different financing mechanisms on the behavior and choices of practitioners and patients, the availability and quality of services, the clinical outcomes for patients, the perceptions of patients and those close to them, and the level and distribution of costs across different parties. One critical question is whether some financing arrangements pose particular threats to patient well-being, trust, and access to care. In this regard, it would be useful to learn about the patterns of enrollment, disenrollment, and continued enrollment by seriously ill patients in different types of managed care and in traditional health care plans.




Caregiver Experience and Bereavement

Palliative care is noteworthy in its concern for the patient in the context of the family (including nontraditional families). Many factors affect the physical, emotional, and practical experiences of those close to the patient and the types of interventions that will be practical and effective. Some issues—anticipatory grief, bereavement, anxiety—warrant research along the lines described in the earlier discussion of biomedical research on cognitive and emotional symptoms in patients. The efficacy and appropriateness of pharmacological therapies for grieving survivors remain important issues, but the role of different forms of social support and the relationship between caregiver experience and social and demographic characteristics also warrant further exploration.




Culture, Communication, Perceptions, and Decisionmaking

Changes in how people die have challenged clinical, cultural, and ethical norms and understandings about care at the end of life, as have changes in the ethnic, religious, linguistic, and social composition of the American population. The increasing racial and ethnic diversity in the United States has led those committed to improved end-of-life care to try to learn more about how perceptions, preferences, and behaviors vary and how communication and care strategies can respect differences in the context of organizational systems designed to provide reliable, excellent, and efficient care. This interest also reflects a recognition that—in addition to a strong basic science and clinical knowledge base—better support for patients and families will depend on a better understanding of the cultural, emotional, affective, and nonrational aspects of perception of symptoms, attitudes, behavior, and decisionmaking.

A modest body of research has accumulated in certain areas such as clinician, family, or surrogate awareness and acceptance of patient preferences; variation and stability in patient preferences about end-of-life care; communication of bad prognoses and treatment uncertainties; attitudes about assisted suicide; and measurement of attitudes and anxieties about dying and death. This research has laid out a range of questions for further investigation including: legal and other problems facing members of nontraditional families when their loved ones are not competent to make decisions; links between desire for assisted suicide and difficulties experienced or expected in receiving appropriate palliative care; and patient and family willingness or desire to rely more on clinicians to make certain important decisions about end-of-life care. Puzzled by patients reporting both high levels of pain and moderate or high levels of satisfaction with care, pain researchers have identified the determinants of patient and family satisfaction with pain treatment as an important research issue (APS, 1995).

As indicated in Chapter 8, the effectiveness of educational strategies in changing health professionals' knowledge, attitudes, and behavior is an important area for investigation. Research on the value of case-based learning, role playing, and similar strategies is generally promising, but overoptimistic claims should be avoided. Also relevant is research on strategies for using guidelines for clinical practice to educate clinicians and change their practices. For example, can academic detailing, computer-based decision support tools, and similar approaches improve the extent to which clinicians apply existing guidelines on pain management to the care of dying patients?

The role of the mass media in influencing knowledge, attitudes, and decisions about end-of-life care is just beginning to be explored. To the extent that the media can be interested in presenting a more balanced and realistic view of dying, research would be useful in designing programs that will attract public attention, improve understanding, and encourage people to behave in ways that allow a more comfortable dying.




ETHICAL ISSUES IN RESEARCH ON DYING PATIENTS

Ethical issues in research on care at the end of life require special consideration because patients with advanced chronic illness are a vulnerable population that tends to be particularly dependent on health workers. Variable degrees of cognitive failure such as dementia or fluctuating delirium are common and may limit a patient's ability to offer a truly informed consent. Moreover, patients may exhibit changing patterns of illness that will call upon investigators to constantly renegotiate informed consent and ensure that strict adherence to the protocol does not interfere with the patient's overall care. Patients with life-threatening illness and their families may be desperate as they sense that options have collapsed around them.

The ethical issues involved in enrolling and maintaining vulnerable subjects in clinical and other research have been the subject of much discussion. A variety of institutional mechanisms have been devised to provide for consideration of such ethical questions and oversight of research involving human subjects (President's Commission, 1982; Cassel, 1988; Annas, 1991; Kass et al., 1996; Vanderpool, 1996). More generally, the principles of research ethics must be applied within a compassionate environment by wise, charitable, and virtuous investigators.

Notwithstanding legitimate concerns, ethical risks should not be overstated. By its nature, research involving symptom-relief strategies may be associated with fewer adverse effects than many trials of agents intended to cure or significantly retard the advance of life-threatening illnesses; these latter trials are commonly regarded as acceptable. Restricting patients suffering from advanced illness from clinical trials may deprive them of improved treatments for symptoms. In addition, to deprive palliative care patients of the opportunity to make their own decisions about participation deprives them of the special meaning of doing something to benefit future generations. Altruism is an important value for some people faced with irreversible illness.

Concern about research that potentially exploits specially vulnerable subjects led to strict limits on research in nursing homes (Cassel, 1988). In recent years, these concerns have been reevaluated, and more flexible guidelines are now commonly observed. Although it is true that seriously ill patients or nursing home residents may, by virtue of their illness or the fact of institutionalization, be uniquely dependent on caregivers and thus vulnerable to exploitation, the committee believes that is possible to create standards for consent and oversight that protect the subjects but still allow them the right to participate in studies. Research may be ethically conducted if certain principles are followed and certain questions appropriately answered. The broadest principles rely on the fundamental precepts of respecting persons, doing good, avoiding harm, and treating people justly (Beauchamp and Childress, 1994). More specific considerations include the following (MacDonald, 1993; Vanderpool, 1996):

• Purpose. Trials with the objective of improving patient quality of life using low-risk interventions usually will not present ethical dilemmas. In contrast, participation in Phase 1 drug trials, where the chance of benefit is small, can present serious ethical concerns. Nevertheless, in these situations, if the purposes of the patient, family, and physician are consonant and clearly understood, ethical dilemmas may be resolved.

• Patient competence. Informed consent can only be given by competent patients, and research on those who cannot give consent (e.g., unconscious emergency department patients) has raised serious concerns and prompted the development of guidelines on the conduct of research for such patients. Cognitive defects are common in patients with advanced chronic illness, many of whom are elderly. Screening tools for competence may be needed for research involving such patients.

• Double agency. Health professionals engaged in research are acting as double agents in that their research may help the research subject but also create benefits for society at large, advance basic knowledge, enhance careers, and improve the competitive position of sponsors. However, the current medical milieu commonly places health professionals in patient care settings in the role of double agents with responsibilities not only to patients, but to the patient's family, and to their own colleagues, departments, institutions, medical faculties, and, most problematic of all, the state or private health agencies. Although this double-agency status must be accepted, the resolution of the dilemmas it may create can generally be resolved by making clear that the patient's well-being is the top priority.

• Industry priorities and research directions. Given limited resources for research, the research activities of many health professionals are undoubtedly influenced by the support available from the pharmaceutical industry. This may limit or skew the direction of palliative care research because most large firms appear to be less interested in drugs that interfere with the pathophysiology of symptoms but do not affect the overall disease trajectory. Thus, some promising lines of scientific investigation may be neglected. Earlier in this chapter, the committee suggested a symptom-based research strategy following the pain model. Better understanding of the pathophysiology of other symptoms may particularly attract manufacturers interested in identifying new uses of existing drugs such as opioids.

In sum, that ethical questions deserve attention does not automatically preclude the conduct of research on end-of-life issues. For researchers, inducing harm out of proportion to potential benefit is a risk that must be carefully considered with the patient, primary caregivers, and appropriate family members and evaluated within the context of established institutional review processes and policy guidelines.




RESEARCH LEADERSHIP

Box 9.1 lists a number of factors that have contributed to the dearth of research on the experience of dying patients and on interventions to improve, prevent, or relieve symptoms and improve patients' quality of life while dying. In the committee's view, underlying many of these obstacles is the shared view of many researchers, educators, and practitioners that dying is a failure, that end-of-life care is unrewarding, and that scientific research in the area is uninteresting. Within academic health centers, the same structural and leadership factors that lead to inattention to end-of-life care in the curriculum contribute to the lack of research in the area.

The neglect of research also reflects economic forces that drive research funding priorities. One apparent impediment is that a number of effective approaches to symptom relief use morphine and other drugs that are available generically, which may make them less interesting to industry-supported researchers. However, as the introduction of new longer-acting oral morphine formulations demonstrates, it is possible to market profitable versions of generic drugs such as morphine. Such profitable drugs may, however, create affordability problems for patients or for hospice programs operating on a fixed per diem payment. Also, if health care organizations do not see palliation as important to their institutional well-being, they will be tempted to cut costs by skimping on palliative care including pharmaceuticals.

Unfortunately but unsurprisingly, dying patients do not form an energetic, ongoing constituency to advocate for more research. As observed earlier, their survivors tend to be exhausted and grieving and, after a time, they are expected to put their lives back together and move on. In addition, hope inclines people—especially patients, families, and clinicians—to focus on curing disease rather than on developing more effective interventions to comfort those approaching death. These dynamics can deflect advocacy groups that champion research to prevent or cure various diseases from vigorously and publicly promoting palliative research to help those who will, inevitably, die of these causes. By devoting at least a portion of their resources to promoting symptom-focused and other research to reduce suffering for those with advanced progressive illness, these advocacy groups could play an important role in creating public demand for improved care. It is also critical for those in leadership roles in research, clinical practice, education, and public policy to advocate for dying people who cannot advocate for themselves or those who will follow them.

The major public funder of biomedical research is the National Institutes of Health (NIH).² NIH is an enterprise that is structured around research on organ systems, driven by investigator interests, and focused on mechanisms of disease. This stance has attracted fairly generous funding, even in recent years when other government health programs have faced reduction or elimination of funding. In the future, however, the issue of research priorities will get more attention. Also, the controversy over assisted suicide may prompt policymakers to propose research mandates that do not have much scientific potential to produce results that help dying patients.

Certainly, research on the pathophysiology of symptoms, mechanisms of symptom relief, and the dying pathways and experiences of patients with different diseases has not been a priority at NIH, and such research is even difficult to identify in its databases. For example, in material provided to the committee, the National Cancer Institute identified 15 active research protocols (as of summer 1996) potentially relevant to this committee's work, but most of them appeared to involve side effects of therapies intended primarily to prolong life (e.g., managing chemotherapy side effects, coping with radiation treatment for prostate cancer, alleviation of platinum-induced peripheral neuropathy) (Varricchio et al., 1996). In some cases, such research may be generalizable to relief of symptoms induced by disease or by treatments to palliate symptoms (e.g., bowel symptoms related to opioid therapy for pain), but the extent to which this is the case needs explicit investigation.

One promising development has been the work by patient groups, clinicians, health services researchers, and others to encourage biomedical researchers and public and private research sponsors to include quality-of-life measures in clinical trials (Johnson and Temple, 1985; Moinpour et al., 1989; Nayfield et al., 1991; Varricchio et al., 1996). Quality-of-life measures, however, often emphasize physical activities and return to work with less attention to "issues that are documented to be relevant to terminally ill persons such as hope, loneliness, depression, anger and anxiety about what is happening to the patient and to family and friends" (McMillan and Mahon, 1994, p. 438). Thus, although the committee believes that the introduction of quality-of-life measures in clinical trials is an important step, research specifically focused on the quality of life for dying patients is also needed.

The committee could not locate any specific data on the number of patient deaths occurring within NIH clinical trials and other research programs, but it expected that clinical trials on heart disease, cancer, stroke, and AIDS—major causes of death in the U.S. adult population—record thousands of deaths. Little or no systematic data are collected on how these patients die and what their symptoms and needs are. Although the number of deaths in trials involving neonates, infants, children, and adolescents is smaller, the same dearth of information applies.

Because resources are limited and priorities must be set, those advocating specific studies related to end-of-life care must be able to argue the case for these studies. They will generally need to address several core questions (IOM, 1992, 1995c) including

• How significant is the clinical problem—its prevalence (current, estimated future) burden of illness (e.g., morbidity, quality of life), economic burden, and variability in treatments or outcomes across regions or population subgroups?

• What is the likelihood that research results will affect decisions and outcomes?

• Will the research constructively supplement conclusions from other research?

To help in setting research priorities as well as increasing knowledge to guide both scientific and clinical research, the committee strongly encourages NIH and other biomedical research groups to consider the inclusion of data collection protocols that would gather more information on people who die from the diseases being investigated in clinical trials. This information would provide a clearer picture of patients' experience during the months before they died of conditions such as congestive heart failure, chronic obstructive pulmonary disease, and end-stage renal disease. Such protocols could be added to all clinical trials of potentially fatal illnesses, much as the National Cancer Institute added quality-of-life measures to its cooperative protocols some years ago. Such data would provide guidance for decisions about priorities for symptom research based on the burden of distress and suffering as well as on scientifically promising questions for investigation.

Public resources for health services research and other topics outside biomedicine are more limited. One major sponsor of such research, the Agency for Health Care Policy and Research (AHCPR), has faced funding threats and political controversy in its efforts to promote evidence-based medicine and guidelines for clinical practice.³ The committee saw the research supported by this agency as a valuable national resource (see, also, IOM, 1995c). Its limited resources notwithstanding, the committee also believed the agency can play an important leadership role in directing attention to end-of-life issues in health services research. The NIH should be encouraged to work with other federal agencies, such as the AHCPR, the Department of Veterans Affairs, and the Food and Drug Administration to consider developing cross-cutting initiatives or to nurture and coordinate pain and other symptom research and research on the effectiveness of other interventions frequently used at the end of life. The initial objectives would be (1) to promote more sharing of ideas and concepts among researchers from different disease fields and disciplines, (2) to provide a focal point for proposals by those researchers interested in symptoms, quality of life for dying patients, and related issues, and (3) to urge investigators and staff to consider creative, economical ways of incorporating questions on end-of-life issues in a wider range of studies. (One example is a study sponsored by the National Heart, Lung, and Blood Institute that looked at risk factors and prognoses for cardiovascular diseases for people over age 65 and included an ancillary study that looked at behaviors and burdens of caregivers for those with severe disease (Schulz et al., 1997). Other longer-term objectives would be to encourage interdisciplinary research and to develop career tracks for researchers interested in these topics and to attract researchers to areas not previously identified as being "interesting." The committee recognized that these kinds of initiatives are difficult to create and sustain but believed that policymakers and research leaders can devise creative ways to support focused biomedical research that will ultimately help people have a better quality of life while dying.




CONCLUSION

Progress in a clinical field is dependent on its research base. Deficiencies in basic scientific and clinical knowledge affect both the availability of effective palliative therapies and the reimbursement of palliative services. The recruitment of excellent young professionals to an area of medical care depends in some part on the opportunities in the field for posing and resolving stimulating questions. This chapter has highlighted a number of exciting areas of inquiry that could build on findings that are accumulating in fields such as pain and immunology research. A major challenge to the field of palliative medicine is to work with colleagues in other fields to generate and conduct research in these and related areas. A broader task is to cultivate understanding of the potential benefits of such research among government and industry sponsors of research, to develop stronger ties with patient advocacy groups, and to promote public demand for knowledge that will improve the quality of life for people who are dying.



---

NOTES

¹Cytokines are hormone-like proteins that are secreted by various kinds of cells and that are involved in immune responses, cell-to-cell communication, and metabolism. Interferons and interleukins are cytokines.

²In an effort to understand the scope of government biomedical research on end-of-life issues, the committee met with representatives of the National Institutes of Health (NIH). As this report was going to press, NIH staff were organizing a conference for September 1997 to consider directions for NIH-sponsored research on symptom management in palliative care settings.

³For example, NIH staff noted that although they sought advice from the Agency for Health Care Policy and Research (AHCPR) about alternative end points for clinical trials, AHCPR's resources were so limited that it could not support its own research agenda in the area.