"Choice" in End-of-Life Decision Making: Researching Fact or Fiction?

Abstract

Purpose: The contribution of bioethics to clinical care at the end of life (EOL) deserves critical scrutiny. We argue that researchers have rarely questioned the normative power of autonomy-based bioethics practices. Research on the ethical dimensions of EOL decision making has focused on an idealized discourse of patient "choice" that requires patients to embrace their dying to receive excellent palliative care. Design and Methods:Our critique is based on a comprehensive review of empirical research exploring bioethics practices at the EOL. In addition we will provide a brief review of our own ethnographic, longitudinal study of the decision-making experience of dying patients, their families, and their health care providers. Results: There is little or no empirical evidence to support the autonomy paradigm of patient "choice" in EOL decision making. What we found is that (a) prognostication at the EOL is problematic and resisted; (b) shared decision making is illusory, patients often resist advance care planning and hold other values more important than autonomy, and system characteristics are more determinative of EOL care than patient preferences; and (c) the incommensurability of medical and lay knowledge and values and the multifaceted and processual nature of patient and family decision making are at odds with the current EOL approach toward advance care planning. Implications: It is exceedingly difficult to identify, study, and critique normative assumptions without creating them, reproducing them, or obliterating them in the process. However, a fuller account of the morally significant domains of end-of-life care is needed. Researchers and policy makers should heed what we have learned from empirical research on EOL care to develop more sensitive and supportive programs for care of the dying.

Book Reviews

Practice Concepts

The Forum

With the aging of the baby-boom generation and in the wake of SUPPORT (Study to Understand Prognosis and Preferences for Outcome and Risk of Treatment) has come a flurry of research and attention focused on end-of-life care. Major funding initiatives have been undertaken by the Robert Wood Johnson Foundation, Open Society Institute, Institute of Medicine, and even the National Institutes of Health. Much of this attention has been directed toward providing for a good death (Emanuel and Emanuel 1998). Bioethics—which has been concerned with issues of death and dying since its creation more than three decades ago—has made important contributions to clinical practice and is implicated in much of this social change. However, the contributions of bioethics regarding end-of-life practices deserve critical scrutiny, for aged persons and others confronting death. In this article, we will: (a) suggest that the bioethics practices governing end-of-life care (a focus on self-determination, advance care planning, and the permissibility of forgoing life-sustaining treatment) focus too much on problematic notions of patient "choice"; (b) review the empirical research examining bioethics practices at the end of life as well as our own ethnographic, longitudinal studies of the decision-making experience of dying patients, their families, and their health care providers; and (c) suggest directions for future research that will identify elements of care in which individual patient choices can be made meaningful.

It is exceedingly rare in our society to leave this life without encountering biomedicine. But those encounters have changed drastically over the years. Just 30 years ago, accepted biomedical practice recommended that all measures should be undertaken to resist death, no matter how inevitable the death, no matter how extreme the intervention. Before that, death was simply not spoken of—the physician's role was to provide comfort and reassurance, patients were not even told what was wrong with them (Novack et al. 1979; Oken 1961). Today, thanks in large part to the developments within bioethics, we require the individual to choose the approach to care at the end of life, rather than simply accept the interpretation of the situation and the ministrations offered by the clinician. The patient is expected to be an active and informed participant in every aspect of his or her care.

This approach is predicated on three unacknowledged assumptions: The first is that the timing of death can be predicted and that the concept of terminality can be objectively considered, measured, and shared with the patient—all necessary for patients to make realistic decisions about their care. The second assumption is that patients and providers recognize choice as a component of treatment decisions—that there are, in fact, meaningful options available for the decisions they confront that are amenable to individual choice rather than some more fundamental imperative, such as availability of resources or the limits of physiology. Finally, and most problematically, contemporary approaches to care of the dying assume that all individuals can routinely, comfortably, and meaningfully confront and consider not just their own mortality, but also the process of their physical decline and dying, in an engaged and rational manner.

The Contributions of Bioethics

Bioethics has developed a set of practices to provide guidance when confronting decisions about medical care before death; these practices include informed consent, advance directives, selecting a surrogate decision maker, and orders to forgo life-sustaining treatment or resuscitation (President Commission for the Study of Ethical Problems in Medicine and Biomedical and Behavioral Research 1982, President Commission for the Study of Ethical Problems in Medicine and Biomedical and Behavioral Research 1983). The current models of biomedical care, influenced by these bioethics practices, require the individual to confront numerous overt decisions about letting go of the physical life—often presenting the individual with mixed messages of hope, possibility, and speculation about what may be the right thing to do.

These practices are predicated on the widely accepted, yet unproven, notion that the individual can be an active, informed, and involved participant in medical care, even at a time of profound personal vulnerability—a vulnerability both existential and physical (Wolpe 1998). These bioethics practices were derived from complex ethical theories of moral integrity, individual rights, duties, virtues, and fairness. In turn, they have become accepted as the ethics of end-of-life care—instead of providing a framework of allowable actions, they have become prescriptive mandates in the clinical setting. These bioethics practices are based on what one of us has called the "autonomy paradigm," a theoretical construction founded on an idealized assumption of patient characteristics and equally idealized assumptions about the provision of clinical care (Koenig 1997).

The orientation of bioethics toward patient autonomy and individual choice was a natural development at the historical moment the field was conceived. Bioethics came into being during the 1960s and 1970s—a period of intense questioning of traditional authoritarian institutions and hierarchies. As many commentators on the origins of bioethics have noted, it was also a time when the rights of individuals belonging to particular social groups were established, including the idea of patient rights, along with a new emphasis on civil liberties (Flynn 1991; Jonsen 1998; Rothman 1991). As a result, the conventions for end-of-life care that emerged were oriented toward individualistic rather than social remedies. We will suggest that a bioethics emerging in a different sociohistorical context might have been very different; in fact, within academic bioethics discourse, profound change has occurred, and the reflexive focus on individual rights and choice has been reexamined and critically assessed. Unfortunately, the clinical practices guiding end-of-life care have been slow to change, because they are firmly entrenched in law and regulation.

The acceptability of the autonomy paradigm in the clinical arena is not because of its inherent or intuitive acceptance by clinicians as being the right thing to do; moreover, it was actively resisted when introduced in the 1970s (Rothman 1991). The seemingly simple principlist approach of much clinical bioethics—autonomy, beneficence, nonmaleficence, and justice—has been accepted in the clinical arena, in part because of the profound time compression clinicians face. There is little opportunity for time-consuming moral reflection in the everyday encounter; but, when a difficult situation cannot be ignored, clinicians prefer maxims to memorize, rules to apply, and if at all possible, an algorithm to follow to resolve the situation. Although the autonomy paradigm was initially resisted because of its inherent challenge to professional authority, its procedural approach has been easily adopted as the quick ethical fix to the challenges presented by the clinical demands of end-of-life care, symbolized most clearly by the availability of new technologies (Brody 1992). Despite recent strong critiques of principlism in general, and the overweighting of autonomy in particular, bioethics practices supporting the primacy of patient choice remain accepted dogma in the clinical setting.

What happens when this dogma is applied to clinical practice? Whereas common sense dictates that true choice only exists within the frame of actual possibilities, in practice the choice model paradoxically implies a broader range of options than are actually available. At the most extreme, dying patients seem to be required to make a choice between life and death. Choice is offered in a situation of statistical uncertainty, but it becomes transformed into a means of understanding and defining the situation itself. On the basis of our ethnographic research on end-of-life care (discussed later), the patient translates a choice about assumption of risk for the possibility of statistical benefit into the choice of doing something to live versus the option of giving up and dying. The irony of the choice paradigm is that it seems to present a choice that does not exist—the choice not to die of a terminal illness. The choice model further implies a distortion in the relationships among the patient, the clinician, and the health care system. Patient choice is presented in situations in which the possibility of choice is wholly dependent on the interpretation and presentation of that choice by the clinician. Common notions of shared decision making do not take into account the fact that the balance of power is weighted toward the clinician in these situations. In other words, the clinician, who further controls those options through their presentation and framing, wholly determines the range and possibility of patient choice by the treatment options that are offered. The clinician's role, however, is constrained by the systems within which he or she practices. The autonomy paradigm does not take into account the imbalance of power in these individual and organizational relationships (Brody 1992; Drought 2000; Lupton 1994).

Choice as Problematic

We propose that the choice model manifested in the ethics of end-of-life care is fundamentally flawed. The problems we now recognize in the care of people at the end of life—poor symptom management and inadequate palliative care, provision of expensive but ultimately ineffective interventions, social isolation of the dying and their caregivers, a lack of resources for basic custodial care—can, in most cases, be traced back directly to this illusion of patient choice. What do we mean by that? The concept of individual choice is an integral part of entrenched medical systems and bioethics practices like informed consent, do-not-resuscitate (DNR) orders, hospice eligibility criteria, compassionate use protocols, and Phase I clinical trials. But there are at least three fundamental problems in the choice model. In many, if not most of these situations, the choices patients confront are essentially meaningless. Resuscitation decisions are a telling example, because research reveals the dismal survival rate following resuscitation attempts (Blackhall 1987; Blackhall, Ziogas, and Azen 1992; Murphy, Murray, Robinson, and Campion 1989). Second, offering choice implies that one can decide for or against the suggested treatment, but the current bias in end-of-life care supports choices to limit treatment for the dying rather than fulfill individual patient choice per se (Field and Cassel 1997). Finally, aspects of the health care system in which the patient receives his or her care, insurance provides a good example, will control and constrain what choices can even be considered by the patient (Pritchard et al. 1998).

The ascendancy of a rational choice model is apparent in many areas of health care today, and bioethics is not alone in embracing it. Outcomes research and the creation of prescriptive practice guidelines are heralded as the solution to almost every clinical problem, but their success is dependent on the patient choosing the proffered approach to care. Computer programs help guide patients through complex decisions on topics ranging from which drug to take for migraine headache, to genetic testing, to choices among therapeutic alternatives for prostate cancer. Patients sit in front of computer terminals completing tasks based on game theory, making quantitative trade-offs between quality and quantity of life. However, anthropological research on medical computing has eloquently revealed the manner in which these practices disguise moral and political positions built into the systems by the designers. The choices offered mirror the assumptions of the clinicians consulted during development (Forsythe 1996). The art of healing and both the experience and practice of caring are delegitimized in the face of corrupted systems of autonomous patient choice. Social and contextual features of medical care disappear.

The appeal of the choice model is not surprising. As Kaplan states, "We live in a culture obsessed with the power of individuals to make decisions that control their lives. This obsession is, in part, wrapped up in the economic/cultural ideology that looks upon individuals as more or less solely responsible for their positions in life with respect to economic and political power" (Kaplan, n.d.). In essence, if the individual can control key arenas of life, including their social position, why not death as well? This is an old ethic firmly embedded in the American psyche and is a result of, and a defining feature of, our political philosophy. The history of this belief in control belies its illusory nature. Bioethics and its practices have pushed these ideals even further.

This critique is not new. Daniel Callahan 1995 has eloquently captured the essence of the issue:

By turning death into one more choice issue, we often do little more than deploy a clever way of avoiding the topic of death itself. An emphasis on choice allows us radically to relativize death, treating it not as our common problem but as just one more occasion for the expression of individual preference and multiculturalism: death as smorgasbord. If that bias is combined with the notion that most people have tidy, nicely rounded, and well-settled views on the way they want to die, then we end by depriving ourselves of any effort to achieve some common wisdom on the subject. (p. S34)

Perhaps the greatest problem with the choice model is that it assumes that we can accept and choose our death long before the inevitability and reality of our dying becomes apparent. In other words, it proposes that rational and cognitive ways of knowing can subordinate our emotional, psychological, and embodied experiences.

A Review of the Literature on Choice at End of Life

We would like to begin by pointing out that once one moves beyond the concerns of the psychological literature with themes of grief, mourning, acceptance, and coping styles, much of the empirical research on death and dying has implicitly accepted the choice paradigm as a normative ideal, evaluating clinical practices by whether they support patient choice. What can we learn about choice and how well it is working by looking at empirical end-of-life research? We begin with a brief look at informed consent. As currently construed, end-of-life decision making requires informed consent—a rational, deliberative process of balancing risks, burdens, and personal values. Informed consent comprises a set of legal rules, an ethical doctrine, and an interpersonal process.

A full review of the problems with informed consent is beyond our task here. However, it is noteworthy to point out that, as medical technology offers an ever greater array of possibilities (both for benefit and for simply doing something), the notion of consenting to a recommended treatment has been replaced by the opportunity to choose among a variety of treatments. In response to fears associated with the financial pressures of managed care, what we now see is a further evolution toward patient-generated requests for treatments not necessarily offered or recommended by clinicians or health care systems. In addition, direct-to-consumer advertising now exhorts the patient to "tell your doctor about" the newest drug for a condition that your physician may not realize you have. Such practices challenge the very basis of the physician–patient relationship and imply unlimited patient choice, not just consent, and eliminate opportunities for discussion of the real benefits and consequences of these choices.

Research on decision making in the outpatient clinic reveals that the basic and minimum elements of informed consent—discussion of the nature of the decision and request for patient involvement—are rarely encountered even in routine decision making within the outpatient setting (Braddock, Edwards, Hasenberg, Laidley, and Levinson 1999). The SUPPORT study, a concentrated intervention to improve shared decision making for seriously ill hospitalized patients, failed to make any improvements in patient involvement in decision making (Covinsky et al. 2000; Golin et al. 2000; Lynn et al., 2000; SUPPORT Principal Investigators [SUPPORT], 1995). There are clearly problems with the ideals of informed consent, and informed consent at the end of life is particularly complicated by the emotional overlay that necessarily accompanies any awareness of a threat to one's life.

The current end-of-life literature is massive, and this review cannot be exhaustive. Even if limited to decision making at the end of life, hundreds of studies and commentaries have been published in the biomedical literature alone. We will organize our review by focusing on empirical studies that address or support the three assumptions we have proposed as underlying the autonomy paradigm and the choice model. These assumptions are: (a) there are reasonably accurate prognostic indicators that can be used for most conditions—patients can be given the information they need to make these choices; (b) there is widespread public and professional acceptance of the concept of patient involvement in advance care planning and the rational choice model it implies; and (c) patients can meaningfully participate in discrete and identifiable decisions around end-of-life care. (See Appendix for an overview of key findings in each area.)

Prognosticating Death

Increasingly effective quantitative models for prognosticating the risk of mortality are being sought and developed, and we will not review them here (see Field and Cassel 1997, for a review). What is of interest, however, is how rarely they are used in clinical practice (Christakis and Iwashyna 1998). Many of these models are developed from a decisions theory standpoint (Knaus et al. 1995), and clinicians may not know how to use them in a meaningful and effective way (Lynn et al. 2000). Prognostic models based on multiple measures of physiological functions also challenge the "art" of medicine, seeming to subordinate clinical skill and knowledge to a computer program or a collection of lab values (Murphy and Knaus 1990).

Do the human practitioners of the "art" of medicine do better than their computer counterparts? Studies of the prognostic accuracy of physicians' estimations of mortality reveal a consistent optimism favoring predictions of survival, an optimism that is enhanced when the duration of the physician–patient relationship is long (Christakis and Lamont 2000). However, the prognostic accuracy can be improved by combining physician estimation with the prognostic model (Muers, Shevlin, and Brown 1996; Teno, Harrell, et al. 2000; Vigano, Bruera, et al. 2000; Vigano, Dorgan, Buckingham, Bruera, and Suarez-Almazor 2000). Yet physicians are often reluctant to make mortality predictions or to adjust their approach to treatment when clear indicators of a poor prognosis are observed (Poses, Bekes, Copare, and Scott 1990; Weggel 1999). Operating behind this resistance to formalized prognostic models is the entrenched view within biomedicine of the uniqueness of every patient. This view holds that, although we may know much about populations, the individual is unique and what we know about the population may not apply to the individual (Hadorn 1997). Furthermore, recent research by Christakis on provision of survival estimates in the context of hospice care suggests considerable physician resistance to providing prognostic estimates, despite the theoretical utility of such information to patients managing the end of their lives (Lamont and Christakis 2001). The strong intuitions of clinicians, possibly based on the more difficult to study goal of providing patients with hope in the face of certain death, signal an arena deserving more research attention.

It is in taking into account the uniqueness of the individual patient that we encounter the first problematic chink in the choice theory—the uniqueness of the individual drives the choice paradigm, yet undermines the science of prognostication. Prognostic models apply to populations and cannot account for individual particularities. More importantly, the prognostic model cannot accommodate the human relationship between the physician and patient, and the meanings that a terminal prognosis holds for that relationship. This may be the core of the resistance to prognostication—physicians really do not want to play God.

A final dimension of the problem is that prognostic estimates are of even less utility once one leaves the domain of cancer, which often has a very predictable end course. (Another exception may be multisystem failure in the intensive care unit, a discussion beyond the scope of this article.) As the work of Joanne Lynn has revealed, when one considers the standard trajectories for death from heart or chronic lung disease, the numbers, that is predictions about the likelihood of surviving for 6 months, never look particularly bad. This is because in a disease characterized by multiple acute episodes and often leading to a modified recovery or at least return to baseline, the trajectory is not straightforward, and any prediction of which episode will be a patient's last is arbitrary (Lunney, Lynn, and Hogan 2002; Lynn et al. 2000). Thus, it might be rational for patients and care providers to remain unconvinced of the inevitability of death and focus on each episode as a unique event.

Advance Care Planning

The passage of the Patient Self-Determination Act in 1991 was based on widespread acceptance—with no empirical research backing the underlying premise or hindering the resolve of policymakers—that everyone should have an advance directive. Squarely within the autonomy paradigm, advance directives are viewed as a reliable way to preserve patient autonomy when individual integrity is threatened. The notion that individuals have clear-cut preferences about a completely unusual and by definition once in a lifetime experience like dying is never questioned—the existence of preferences is assumed and never validated.

Many patients say they want to discuss end-of-life issues with their physicians early and while they are healthy (Pfeifer et al. 1994). However, patients not only fail to complete an advance directive, but also do not discuss their preferences about end-of-life care, even when decisions may soon be required (Hofmann et al. 1997). When explicit decisions are made to forgo treatment and choose death, patients are most often older and sicker and with more functional impairment—the decisions support what seem like an inevitable consequence of the patient's current condition (Bordenave et al. 1998; Goodlin et al. 1999; Phillips et al. 1996). Patients' preferences are affected by their perception of their condition; however, patients are overly optimistic about their own conditions as well as the benefit various life-sustaining treatments can offer them (Coppola, Ditto, Danks, and Smucker 2001). Having trust in the health care system is also implicated in these decisions, especially in African American families who are often less willing to forgo treatments than other ethnic groups (Bonner et al. 1999).

Advance directives have been highly touted as a means of improving physician understanding of patient wishes by providing unequivocal documentation of those wishes. Wide-ranging research has been done on the means of increasing the use and effectiveness of advance directives in the clinical setting, which we will mention only in passing (Brent 1994; Hague and Moody 1993; Haisfield et al. 1994; Hogan 1992; Kupecz 1991; Levenson and Pettrey 1994; Madson 1993; Mezey and Latimer 1993; Molloy et al. 1996; Murphy 1994; Oleson, Jones-Schenk, and Tuohig 1994; Palker and Nettles-Carlson 1995; Rushton 1993). The most promising programs focus on extensive patient and family education about medical choices and the effectiveness of treatments like resuscitation coupled with system changes that incorporate advance directives into practice patterns in a variety of ways (Ditto et al. 2001; Molloy et al. 2000; Murphy et al. 1994; Prendergast 2001; Schneiderman, Kronick, Kaplan, Anderson, and Langer 1992; Schonwetter, Walker, Kramer, and Robinson 1993; Yamada, Galecki, Goold, and Hogikyan 1999).

However, to date, no studies have shown that advance directives significantly facilitate decision making, truly direct care, or save resources at the end of life (Emanuel 1996; Emanuel and Emanuel 1994; Emanuel, Weinberg, Gonin, Hummel, and Emanuel 1993; Ewer and Taubert 1995; Teno, Lynn, Connors, et al. 1997; Teno, Lynn, Wenger, et al. 1997; Teno, Licks, et al. 1997; Teno et al. 1994). In the SUPPORT study, the primary benefit of advance directives was the naming of a surrogate rather than provision of a clear direction for treatment. Further, because the situation was rarely a clearly hopeless one, physicians and surrogates were reluctant to act on the patient's stated wishes, even when they were available. It is not clear that the identification of a surrogate changed any treatment decisions (Teno, Lynn, Wenger, et al. 1997; Teno, Licks, et al. 1997). The utility of identifying a surrogate has not been proven, even though we may assume it increases clinician comfort level to have an identified person with whom to share the moral burden of decision making.

The extent to which surrogates actually discuss these issues with the patient and the concordance of the surrogates' views to patient-expressed preferences, remain problematic (Aikman, Thiel, Martin, and Singer 1999; Hopp 2000; Layde et al. 1995; Marbella, Desbiens, Mueller-Rizner, and Layde 1998; Mattimore et al. 1997; Seckler, Meier, Mulvihill, and Cammer 1991; Terry et al. 1999; Uhlmann, Pearlman, and Cain 1988). Surrogates' predictions of patient preferences are usually no better than chance. For many patients, this lack of agreement is unproblematic, because many voice the preference that the surrogates' wishes be followed and their own preferences overruled in situations of conflict (Hare, Pratt, and Nelson 1992; Sehgal et al. 1992; Terry et al. 1999; Zweibel and Cassell 1989). It is the relationship that matters, yet the relational values patients hold and want preserved through their advance directive (and symbolically charged appointments of surrogates) cannot be accounted for by the autonomy paradigm (Hofmann et al. 1997; Puchalski et al. 2000). There is some evidence that patients are also willing to defer to physician judgment over their own previously expressed preferences in a similar manner (Sehgal et al. 1992).

Clinicians often feel that they know the patient well, that this intimate knowledge is more important or valid in determining treatment choices for the patient than are statistical models. Furthermore, this perceived knowledge of the patient also inclines the physician to spare the patient a difficult conversation. The result is that these unverified assumptions of patient preferences impede clear communication and decision making (Wenger et al. 2000). Not surprisingly, if we look at the studies of how well physicians and others are able to identify patients' wishes for life-sustaining treatment, they are only moderately better than chance even for physicians with a long-standing relation with the patient and who have engaged in some discussion of preferences (Ewer and Taubert 1995; Mattimore et al. 1997; Seckler et al. 1991; Uhlmann et al. 1988; Weeks et al. 1998; Wilson et al. 1997).

How Choices Are Made

Recent studies on end-of-life care have described the factors important to patients, family members, physicians and other care providers at the end of life (Steinhauser, Christakis, et al. 2000; Steinhauser, Clipp, et al. 2000). There is agreement across all groups on the need for symptom management, preparation for death, completion about one's life, being treated as a whole person, developing a relationship with medical care providers, and the usefulness of advance directives (Steinhauser, Christakis, et al. 2000). Additional concerns center around avoidance of prolongation of the dying process, decreasing the burden on loved ones, and strengthening relationships with others (Singer, Martin, and Kelner 1999). Survivors of patients who have died express the need for more preparation around the details of dying and resources for handling the necessary tasks after a death. Relationships with the health care team are often problematic, and many cite the need for staff training in communication and more development of compassion (Jacobson et al. 1997). Although these findings reveal what might underlie the concept of a "good death," they do not say much about respecting choice for its own sake—instead they outline other values patients hold.

In studies examining what is important to patients who are dying, inconsistency is evident in many of the responses. Whereas patients and family members were more likely than physicians and other health care providers to feel that being mentally aware was important, they were less likely than their care providers to require a sense that the patient's life was still meaningful. Other evidence of ambivalence about what constitutes a good death is revealed by the responses of 48% of patients who said they wanted to use all available treatments no matter what the chance of recovery. This is of particular interest in that 64% did not want to be connected to machines and the general agreement described previously about wanting to prepare for death (Steinhauser, Christakis, et al. 2000). However, patients also felt that clear decision making and communication with physicians could relieve their fears and make the decisions they faced easier (Steinhauser, Clipp, et al. 2000). If this level of communication could be achieved in practice, it might diminish the apparent contradiction between wanting everything that can be done and, at the same time, wanting an idealized good death (Field and Cassel 1997; Steinhauser, Clipp, et al. 2000).

The difficulty of achieving clear communication and decision making may be explained by the indecisive and uncertain attitudes of providers about their care of dying patients, especially toward those not clearly dying and for whom extensive technological interventions may be undertaken. Nurses and physicians report extreme ambivalence about their experience in intensive care (Laubach, Brown, and Lenard 1996). This ambivalence stems from the repeated negative experiences they recount from their practice. Many of these complaints are directly related to questions about the appropriateness of care, and about utilizing their life-saving expertise on seriously ill and dying patients. Physicians and nurses often complain of disagreeing with, or feeling personally compromised by, the level of care provided to dying patients (Puntillo et al. 2001; Society of Critical Care Medicine Ethics Task Force 1992; Solomon et al. 1993). Yet, they are hesitant to accept the theoretical distinctions and guidance provided by bioethics and professional bodies (Dickenson 2001; Society of Critical Care Medicine Ethics Task Force 1992; Solomon et al. 1993).

The SUPPORT study illuminates the difficulty of achieving clear decision making even with a coordinated intervention—including professional staff dedicated to this purpose—for its promotion (Baker et al. 2000; SUPPORT, 1995; Teno, Fisher, et al. 2000). In-depth studies of the process of decision making shed some light on these difficulties (Hiltunen, Medich, Chase, Peterson, and Forrow 1999; Jacob 1998; Kaufman 1998, Kaufman 2000; Slomka 1992; Swigart, Lidz, Butterworth, and Arnold 1996). Research consistently demonstrates that the process of decision making requires significant cognitive, emotional, and moral work. These empirical studies using qualitative techniques of participant observation and in-depth interviewing allow access to the process that cannot be achieved in other ways.

Narratives by the nurses involved in the intervention phase of SUPPORT provide insight into the processes they observed in their work (Hiltunen et al. 1999). Narrative descriptions of individual cases were coded and analyzed for recurring themes around decisional conflict. Features of family decision making included recognition of a dilemma, a period of vacillation, moving to a turning point, and letting go. Hearing information, learning from others, and having someone facilitate the process were identified as enhancing the chance of reaching a turning point in which a decision could be made.

In observations of family decisions to forgo life support in the intensive care unit, similar processes were identified by Jacob 1998 and Swigart and colleagues 1996. Primary among these were the need for the family to arrive at a judgment about the patient's condition that was in line with the medical assessment presented to them and true to their perceptions of the patient. Families looked for cues from the patient's life and reviewed the patient's life story to validate decisions and provide assurance that they were doing the right thing. It was also important for the families to feel that they could work with the staff, that their concerns were heard, that they had easy access to information, and that their loved one was being well cared for. Connection and open communication were integral components of this process. The final phase encompassed work to allow the family to move on, and to make sense of the experience in light of ongoing family roles, relationships, and values.

Slomka 1992 describes a process of negotiated decision making. Decisions do not happen as discrete and isolated events but as a cascade of decisions, each decision and action affecting later possible actions. This type of negotiation model allows for a critical sharing of the moral burden of decisions to forgo treatment between the medical staff and the patient or family, in sharp contrast to the ideal of pure patient choice. Themes emphasizing the importance of communication and the need for placing empirical evidence of the patient's medical situation within a framework of shared moral concern resonate throughout all of these studies.

Kaufman 1998 places these findings within the social and political context of the modern hospital—identifying the inherent tensions and barriers to connection, communication, and empathy in the clinical setting. Medicine is the dominant framework for understanding illness and old age, a framework readily accepted by the public. To do what can be done until the outcome is repeatedly negative drives the default approach to care of every patient. There is no clear conception of how to balance goals of treatment and goals of comfort or the demarcation of where the limits of doing everything meet those of doing what can be done. A clear articulation of goals is almost impossible.

Further confounding this murky situation is the ambiguous role of the patient or family in controlling the treatment decisions. Although they often share the dominant view of medicine about its value, the lay understanding of what that entails is incommensurable with the professional understanding. End-of-life decisions require detailed knowledge of physiology and disease processes, as well as an understanding of the culture of medical knowledge—of the intricate rules, practices, and assumptions that give form to the medical knowledge and the choices presented to patients and families. Kaufman 1998 implicates the entire industrial and public enterprise of medicine in the problematics of end-of-life care. These are issues that will prove intransigent despite improved communication and increased availability of palliative care services.

End-of-Life Decision Making in a Multicultural Population of Cancer and AIDS Patients

Our own research provides further support for these findings. In a multiyear, longitudinal study of terminally ill patients with solid tumor cancer or AIDS, we conducted semistructured, in-depth interviews with 88 index patients, plus 1–3 family members or friends who supported the patient at home, and two of their outpatient clinic providers—in most cases a physician and a social worker or nurse. A total of 462 interviews were conducted and analyzed. The study was conducted in a county clinic for low-income, Medicaid, or indigent patients. Patients were referred to the study by their primary clinic physician if he or she felt the patient had 6 months or less to live. (The AIDS study was conducted just before the availability of life-extending drugs such as protease inhibitors.) All patients referred by their physician were approached for inclusion in the study and enrolled if they agreed to participate after obtaining informed consent as approved by the institutional review board. Patients ranged in age between 27 and 84. There were almost equal numbers of men and women in the cancer group; the AIDS group was entirely male because there were few women with AIDS seen in this clinic. Interviews were conducted in English, Cantonese, or Spanish, as the participant chose, and were then translated and transcribed (Drought 2000; Koenig and Gates-Williams 1995).

A similar interview schedule was used for the patients, family members, and health care providers. We asked what types of difficult medical "decisions" each patient had faced in the course of the illness and treatment. We also queried how specific practices—such as shared decision making, limiting treatment, formulating advance directives, and deciding about DNR orders—played out for each participant in the individual's care. Follow-up interviews and chart reviews were conducted to get a sense of the patient's course of treatment over time. In most of the 88 cases, all of the interviews and follow-ups were conducted within a 3- to 6-month period. The transcribed interview texts were coded and analyzed for recurring themes. Techniques of discourse analysis were used to compare the way respondents spoke about advance care planning with the way they described the process of decision making (Drought 2000).

None of the clinicians questioned the role of advance care planning in their clinical practice or expressed any doubts about its efficacy. They spoke passionately about how important it was for the patient to make an informed consent to treatment and that nontreatment always remained an option. Even though many of the clinicians identified themselves as being comfortable with and fairly good at talking to patients and addressing end-of-life issues, we saw little evidence of early, systematic attention to any of the end-of-life bioethics practices. In the majority of cases, they had not discussed limiting treatment, advance directives, or DNR orders at the time of our interview, even though they themselves had identified the patients as having a prognosis of less than 6 months by referral to the study (see Table 1 ). The providers offered detailed and well-thought-out reasons why each patient was an exception to the general rule—it was not time yet, the patient was not ready yet, the particular clinician interviewed was not the right person to initiate the discussion, a certain sequence of events (e.g., hospitalization or failure of response to chemotherapy) had not yet transpired. The uniqueness of the patient or the situation seemed to grant exemption from the normative discourse. Instead, they described the decisions made as reactive rather than proactive, highly contingent, and made late rather than early.

Patients seemed uninvolved or poorly informed about the details of their disease, their prognoses, and, often, their care. As in other studies, providers were generally ignorant of patient understandings and concerns as they were expressed in our interviews. Even so, and, despite the seeming chaos of the formal decision-making process as reflected in the interviews, patients were seen, evaluated, and transferred to hospice care, and few underwent aggressive interventions near the time of death. Indeed, very little dissatisfaction with the process of decision making was volunteered by the patients or their families. In fact, few patients, family members, or health care providers believed that many decisions were actually made. Options for palliative end-of-life care simply fell into place and ceased to be choices when the patients reached a physical state in which the fact of dying could no longer be ignored or denied. Patients who had demanded aggressive treatment for months expressed appreciation for the support of hospice. Patients who had actively resisted discussions of resuscitation yielded to a DNR order. Family members who had expressed frustration with the physician's unwillingness or inability to "do something" described their gratitude for a peaceful death at home when interviewed weeks later.

We were surprised that neither the clinicians nor the patients perceived much choice when they were confronted with what are often described in palliative care as "decision points," in which one selects an option, seemingly from a full menu. The patients consented to treatment—by showing up, if by no other means. But, in most cases, there was not an experience or a perception of choice on anyone's part. Some patients denied outright that they had made any medical decisions, even though multiple decisions to pursue or forgo treatment (such as surgery) had, in fact, been made, or at least had occurred. This experience of not choosing may have been aggravated by the severity of illness, when the facts of the situation determined a fairly narrow realm of possibility. Theoretically, whereas one could choose among treatment options, the constraining or determinative particulars of physiology often limited the meaningfulness of choice. When questioned, some patients did speak passionately, and proudly, of making all their treatment decisions. The question remains as to whether these were active choices or situations in which the patient simply acquiesced in recognition of the biomedical reality and limited possibility posed by the situation. As one provider said of her patient for whom curative treatments had failed, "I mean, we discussed it, but honestly they weren't really decisions. They were ‘treatment’s not working, this is what we have to do now.' It wasn't a matter of ‘we can do this, or we can do that.’ It wasn't working." This theme of decisions evolving without any element of choice was recurrent in the provider interviews.

This lack of perceived choice is of particular interest because, in most cases, clinicians and patients expressed strong feelings that the patient is, or should be, in control of deciding the approach to treatment. Patients definitely wanted to feel that they had some say in what was to happen to them, to feel that they possessed some level of control over a situation that often felt well beyond their control. Most patients expressed an imperative to do whatever could be done. In the majority of cases, that meant choosing the recommendation of their physician. Even when the patient chose among an array of treatment options, it may not have been perceived as choice, because only one option was likely to meet the patient's definition of doing "whatever could be done." Additional research is needed to determine whether physicians provide options so that one option is logically the "correct" choice, an important clinical hypothesis to test. Paradoxically, many patients stated in interviews that they would never consider limiting treatment, even though they had actively made decisions to stop chemotherapy, enroll in hospice, or in some cases, opt against aggressive or curative treatment and palliation of symptoms. In the patient's mind, these decisions were not conceived as decisions about "limiting" treatment; rather, they were perceived as reasonable actions given the problems they were facing.

The difficult decisions patients did identify were more often related to managing life—broadly defined, even though our study was introduced as one focusing on medical decision making, and the questions that we asked were focused on the bioethics interventions designed to uphold choice. Patients described their most difficult decisions as being related to quitting a job, deciding where to live (often related to imminent care needs), disclosing their illness to loved ones, managing decreased finances, interacting with the medical community (frequent clinic visits, transportation problems, missed work for relatives, etc.), or planning for family members' needs after their death.

Patients were uninterested in discussing their imminent dying, struggling instead with the challenge of continuing to live—to have a life and remain engaged in the everydayness that gives life meaning—in the midst of their illness (Liaschenko 1993). A curious silence was "expressed" by most of the respondents when asked to consider the period of their actual dying. Individuals were willing to discuss their concerns in the present, and the difficulties they experienced with their treatment, symptom management, and the effort of staying connected with life. For the most part, they were willing to discuss the fact of their death and their concerns for their survivors, and they frequently volunteered information about their funeral plans and wills. But there was very little willingness to engage in discussions about the transition between life and death, about the practicalities of managing that transition—about how, where, and with whom the transition was to be made and what it might be like. This is the stuff of end-of-life treatment decisions. This is the presumed purpose behind the discourse of advance care planning. But, for the most part, respondents engaged directly with end-of-life treatment decisions only when it became a matter of practical necessity, when such decisions could no longer be ignored, and when the events requiring decisions were not just imminent but were actually evolving.

Even though the providers actively voiced support for patient involvement in treatment decisions, they offered true choice only when highly limited options shaped by dire biologic imperatives left little at stake. The providers we interviewed felt a greater responsibility for caring for the patient than for the mere presentation of treatment options and acquiescence to claims of patients' rights. An unarticulated value—one that superseded the value that these providers placed on patient autonomy—seemed to be in play. This value, closely related to providers' perceptions of responsibility, determined what choices they allowed patients to make. Intentionally or not, the providers categorized patients as being either appropriate or inappropriate candidates for whom aggressive treatment should be limited. Conflict occurred when the patient's desires were at odds with the physician's categorizations. Strong language of coercion was often used when patients refused treatments the physician felt would provide some benefit; in such cases, clinicians described their approach as "strong-arming," "terrorizing," or "hammering home" the need for treatment. Similarly, physicians would obstinately refuse to continue to provide chemotherapy or to refer patients to surgery if they felt that there was little or no benefit to be obtained at the expense of considerable suffering.

The physicians in our study professed a profound sense of obligation and responsibility to their patients. In some ways, this sense of responsibility fueled their commitment to the autonomy paradigm, because they believed that this provided a means of preserving the integrity of the patient. But that paradigm was not strong enough to overcome their abiding faith in their professional and technical ability to use their skills to help their patients within the defining conceptual framework of a biomedical model. If patient choice did not allow them to apply those skills in situations in which there existed a good probability of physiologic benefit, or if patient choice required physicians to apply those skills when there existed little or no probability of benefit, then there occasioned a deeply felt, ethical need to resist that choice. Fundamentally, these situations seemed to challenge their perception of the caregiving role. Patients and family members also expressed concern that requiring patient choice compromised the caregiver in some way. Some patients experienced a sense of abandonment at the moment of their greatest need when the physician appeared to abdicate the caregiver role by requiring the patient to make painful choices. A few patients expressed the feeling that offering choice reflected incompetence or indifference on the provider's part.

Patient and clinician responses about particular practices, such as the use of DNR orders and advance directives, reinforce Kaufman's findings about the incommensurability of lay and medical knowledge (Kaufman 2000). This incommensurability goes largely unexplored and unrecognized—the same words are used, but the meanings and understandings are never articulated or validated. Instead, using the same words passes for agreement. In the following description provided by one physician, we are left to wonder just what each party thought they had agreed on.

Well, I guess they hinge [the decisions] on understanding that this is a rapidly fatal illness and the goals from the onset are just comfort. I guess our discussions about DNR and extraordinary measures were some of the most important… The house staff wanted me to go over DNR because I had become [the patient's] primary care provider… I prefer to be the one to do this if I am the primary care provider. I have no problems doing it. I started to approach this situation with the social worker [present]. I generally say something like, "one of the things I would like to talk to you about and make clear is that if in the unforeseen future you have difficulty breathing or your heart stops working, would you like to have care?" I try to keep it medical. In this case, he had difficulty understanding that. So I said, "Would you like to be hooked up to a breathing machine or have your heart restarted if it stopped?" And he said that he would have to think about it. Then, thank God, the social worker stepped in and asked, "Do you have a living will?" He said "Oh, yes, and I don't want any heroic efforts to keep me alive." She really saved me and taught me that I can use other terminology when discussing these issues with patients. It was very fortunate that she was with me. He told us that he had thought about this and that his parents were aware of what he wanted. So, it was already resolved. He had done it on his own.

We asked patients if they had been asked or had made any decisions about resuscitation status. The patient responses were fascinating in their description of the need for that choice and its implication. Many patients did not relate the use of resuscitation to their underlying disease state. Instead, they referred to its use in the event of a hypothetical heart attack or stroke and the possibility that its use would leave them in a limbo state of dependence on life-support, possibly for months and months. Most said they saw little use in that—it was not really living. There was almost no reference to its actual use in the face of their impending death from cancer or AIDS. One woman said that her grandson would definitely want everything done, but in the next breath said that when it was his time, though, they should just let him go, let him have some dignity.

It was not clear from our interviews if there was simply a complete disconnect between the use of cardiopulmonary resuscitation (CPR) and the patient's impending death from cancer or AIDS, or if patients simply made the assumption that CPR would not be offered or applied when a death was inevitable or expected, a not illogical view for patients to hold. In either situation, having to choose to forgo CPR was problematic for many patients. Its symbolic meaning, even when there was confusion as to its actual meaning, was quite strong. For the most part, these patients were not denying that they were going to die, but forgoing CPR was not seen to be about dying. Most patients said they would want it if it would leave them in a desirable state. Several said they would want it so they could say goodbye to loved ones, and then they would have the machines disconnected. There was no reference to the possibility that it might not work.

The resuscitation decisions were also quite symbolic for the providers, signifying more than the attempt to restore heartbeat or the support of breathing with a ventilator as they typically described it to patients. For providers, the resuscitation decision was symbolic of the general treatment orientation toward the patient and also functioned as a prognostic cue. Clinician discussions of code status included the following descriptors: The patient would or would not have been a candidate for surgery based on code status; the patient was inappropriate to have a DNR, because she was still undergoing chemotherapy; he must not have been DNR, because he was in the hospital last week. The DNR order signified more than applying a particular technology at the moment of death, a fact that may explain why discussion of the order is so often delayed.

Pellegrino and Thomasma 1988 have said that, for the dying person, "reasonable assurance that one's life plans can be fulfilled is no longer available. Though we still consider such persons free to make decisions and function as human beings, they are futureless. Since the future and hope are very much part of normal decision making, we must conclude that these persons are in a state of diminished humanity" (p. 142) and are subject to an increased vulnerability. This may explain why patients and providers cling to the possibility of a future and resist accepting the label of dying or making early preparation for it. This may also explain why clinicians are often more comfortable in answering the question (proposed by Joanne Lynn, among others), "Would you be surprised if your patient died in the next six months?" than in determining that a patient has a prognosis of less than 6 months to live. Many of the patients in our study were quite indignant that they were directly required to confront and admit to their dying through the decisions they were asked to make about end-of-life care. Although the clinicians sometimes labeled these patients as "in denial," most patients were fully aware that they were dying—that was not what was at issue. What they wanted was to focus on living while they could. A social worker described one patient's refusal of hospice home care because he did not want someone coming in every week reminding him that he was dying. Early hospice referral may provide needed support in the home, but at too great an expense to the patient's continued efforts to maintain an identity and the possibility of a future.

Recent work has posited that patient reluctance to engage in end-of-life decision making is often related to cultural norms, values, and expectations (Blackhall, Murphy, Frank, Michel, and Azen 1995; Blackhall et al. 1999; Ersek, Kagawa-Singer, Barnes, Blackhall, and Koenig 1998). In our study, which purposefully included a diverse population, we did find some differences among cultural groups (Barnes, Davis, Moran, Portillo, and Koenig 1998; Davis 1996; Hern, Koenig, Moore, and Marshall 1998; Koenig and Gates-Williams 1995); however, a full range of responses was noted across all groups. There were individuals who strongly adhered to the choice model and wanted involvement in and control over the medical decisions, as well as individuals who wished to defer decision making to their physicians or others, in each of the four ethnocultural groups we studied (European Americans, African Americans, Chinese immigrants, and Hispanic immigrants). The factors affecting patient decision making are complex and varied, and are problematic for many if not most patients.

The autonomy paradigm did work well for a few patients, however. Homosexual men with AIDS who were strongly connected with the local gay community used the model frequently and quite effectively—requesting DNR orders, appointing health care agents, and refusing aggressive treatments. Because our study was conducted before the introduction of protease inhibitors, the patients did not foresee any reprieve in their illness. The nontraditional kinship relationships frequently found in this group could also be better protected and legitimized in the health care arena by the use of advance directives. Other patients who used the model effectively tended to have one or more of the following characteristics:

1. An understanding of the health care system, the types and limits of treatments available to them, and the expectation of informed consent and patient involvement

2. Prior experience with the end of life of a loved one who had received aggressive intervention

3. Understanding and acceptance that death from their disease was inevitable, and treatment would not provide significant benefit.

Future Research Directions

One challenge of empirical research is to describe what is happening without structuring what one sees by the questions that are asked. Much end-of-life research to date has simply assumed the salience of the choice paradigm, partly because it provides seemingly clear-cut "decision points" to study using quantitative methods, avoiding the messy realities of dying and the need to create new research tools and methods focusing on process. How do you identify, study, and critique normative assumptions without creating them, reproducing them, or obliterating them in the process? This task is particularly challenging when those normative assumptions are based on positivist ideals of a priori, universal values (Pearlman, Miles, and Arnold 1993). We admit to complicity in this regard. When we first began our ethnographic project with cancer and AIDS patients, we "naturally" focused on the primary role of cultural difference among patients and on bioethics practices as key domains of end-of-life care. Fortunately, an advantage of qualitative methodologies, when employed flexibly, is that they allow for midcourse corrections. Thus, we were able to "hear" the data when patients did not fit neat cultural stereotypes and when our focus on decision making proved to be illusory from the point of view of subjects.

A full account of the morally significant domains of end-of-life care is needed. As Hoffmaster has proposed, "Morality only becomes intelligible when the background that makes it possible is considered" (Weisz 1991, p. 250). Research that examines and interprets bioethics practices within the context of clinical care can provide insight not only into what is going on, but what ought to be as well.

The ethics practices of end-of-life care are a response to technological developments within biomedicine, the public focus on individual rights, the societal disenchantment with authority in general and biomedicine in particular, the rise of consumerism especially with regard to health care, and the cost constraints currently faced in health care because of the failures of market reforms. These forces have coalesced into a discourse—made up of ethical principles, legal doctrine, regulatory mandates, clinical practices and expectations, and the ideology of the good death—that requires the patient to speak his or her death, to grieve in advance, literally to decide on the "terms of engagement" with death, and paradoxically, to choose to participate in it or not (Armstrong 1987). It is an attempt to normalize and tame death, to bring it into public consciousness. These practices require that the individual perceive him or herself as an object that (a) is dying and (b) must be medically managed into that death. This poses a profound challenge for the individual attempting to continue to live his or her known life through the dying process. A reformulated research agenda, one that does not reify choice above all other domains, can help reveal the full dimensions of the dying process so that meaningful interventions can be developed and evaluated.

Death is still, despite the ethics practices we employ at the end of life, difficult for mere mortals to accept, approach, and embrace. Talk to the contrary does not make death easy to accept; instead, it can do violence to the processes of denial, negotiation, acceptance, and living that engage the dying and those who care for them. We propose that these ethics practices protect us from truly acknowledging and coming to terms with death primarily by introducing the notion of choice into dying. We are distracted from the reality of death by our preoccupation with the choices presented by the discourse. We focus attention on the choices, rather than on the situation creating the need for the choices, and the reality of death becomes lost. Patients and providers resist the choices surrounding death, but the resistance is to the admission of death itself. It is only when death's proximity can no longer be denied, and choice has given way to fate, that the end-of-life practices are readily accepted and employed. At that point there is no more choice in the matter.

Paradoxically, the key for patients to choose and receive end-of-life care seems to be the ability to understand that there is no choice in the matter. Outside the narrow confines of the hastened death debate, dying is simply not a choice. Patients need to understand—in more than a cognitive sense—that they are at the end of life, then they readily accede to the end-of-life decision-making model. But we know little of how that actually transpires—particularly how physicians, families, and patients construct a mutually agreed-on narrative—because of the limited bioethics research focus on choice.

We were originally asked to comment on whether guidelines could be developed for limiting treatment in older patients who have seriously diminished quality of life. We would say no—not as long as we cling to the choice model. For guidelines to work, such individuals must be able to understand and accept that they have a seriously diminished quality of life, are at the end of life, and that there is no medical treatment to reverse or ameliorate the situation drastically. If they "choose" that position, however, it is probably unnecessary to impose a limitation of treatment on them. Empirical research shows that the controlling patient choice is about the perception of the person's stage in the illness trajectory, not about the treatments he or she wishes to receive.

For patients with a predictably deteriorating illness, such as cancer, there comes a point when the physical experience of symptoms allow no other meaning to be applied than that they are dying. At that point, a referral to hospice is usually readily accepted. In the critical care unit, families reach a point when they can accept that the failing physical body will not be able to respond to the technologies being offered, and they "choose" to withdraw treatment. This shift in subjective experience may account for the repeated data showing referral to hospice in the last 2 weeks of life and DNR orders written and treatment withdrawn only within days of death in the intensive care unit. But how does the patient with end-stage organ failure know which dip in function will not turn around? Do we expect him or her to choose which will be the final dip? How does the frail patient identify, in the gradual and imperceptible evanescence of function, that it is time to let go of life? Is he or she to choose to stop treatments that have been relied on for many years or not to treat a relatively minor new problem? Can these choices be made sooner than the last weeks or days of life? Should they be made sooner? Can we afford for them not to be?

Descriptive research on the experience of patients through the disease and dying trajectory can help us fashion an ethics of end-of-life care that reflects the values and concerns of patients and their families. Is there an identifiable moment or process for recognizing death's inevitability? Is there a point of transition toward disengagement with life? How do patients, families, and clinicians identify that a patient is dying? What are the sources of resistance to this identification and what values does that resistance signify? How are power relationships between clinicians and patients affected? What is the experience of accepting and preparing for death early in the process? What are the costs and the benefits?

Research on physician–patient communication should focus on ways to help the patient understand and interpret the physical symptoms in light of the medical possibilities and the likely disease trajectory. In other words, how can patients understand and identify when they are near death? Research is needed to develop communication techniques that focus less on death and more on the realities of the medical system, the types of decisions patients actually—rather than hypothetically—confront, and the finite possibilities of medicine. Both patients and providers could benefit from a clearer understanding of these issues, especially given the technological imperative and the media coverage of every new possibility for extending life. How can this discrepancy be decreased?

There are current efforts to increase physician knowledge and comfort with end-of-life care, establish systems for the provision of palliative care, and develop community networks to increase public awareness. Research on these programs should focus on their effects and outcomes in a descriptive manner, not just on their efficacy in perpetuating the autonomy paradigm. Innovative programs may generate organizational models and clinical practices that move beyond meaningless recitation of nonexistent choices—such as dying at home if one is elderly and alone—to fully integrated systems of care that meet most needs. Research might then identify situations in which real choice is possible and beneficial.

Overview of Empirical Research on Choice in End-of-Life Care

Assumption: Reasonably accurate prognostic indicators can be utilized for most conditions.

Physicians resist and may not know how to use them.

Physicians and patients are consistently overoptimistic about prognosis.

Assumption: There is widespread public and professional acceptance of the concept of advance care planning and the rational choice model it implies.

Basic informed consent remains problematic.

Measures to increase the effectiveness and incidence of advance directives have largely failed.

Many patients resist engaging in advance care planning and find other values than autonomy important.

System characteristics are more determinative of end-of-life treatment than are patient preferences.

Assumption: Patients or their surrogates can meaningfully participate with their physicians to make discreet and identifiable decisions around end-of-life care.

Patients express conflicting values—want to know everything has been done, do not want to die on machines.

Decisions to withdraw or limit life-sustaining treatments are a process rather than an event and require working through the cognitive, emotional, and moral components of the decision.

Patients and providers have differing conceptions and understanding of patient and family needs at the end of life.

Decisions to forgo life-sustaining treatments are more readily made by patients of increased age, with limitations in activities of daily living, whose self-estimation of prognosis is poor.

Incommensurability of medical and lay knowledge prohibits clear communication and may prohibit shared decision making.

The authors gratefully acknowledge the support of the following: American Foundation for AIDS Research (1772); National Institutes of Health (R01 NR029060); Open Society Institute Project on Death in America; The Robert Wood Johnson Foundation; State of California AIDS Research Program (R95-ST-188); The Greenwall Foundation; University of California, San Francisco, AIDS Clinical Research Center; and the University of California, San Francisco Regents Fellowship (to T.S.D.).

References