Although neurosurgeons and intensivists are confronted by similar clinical scenarios in many parts of the world, their approach to EOL decisions is highly variable. In European ICUs, decisions to WHLS vary from 16 to 70 % of deaths, depending on the country, while decisions to provide all therapies including CPR up to the point of death vary from 5 to 48 % (Wunsch et al. 2005; Sprung et al. 2003). In the United States, there is similar variability in EOL decisions among ICUs, with rates of CPR at the time of death ranging from 4 to 75 % and some ICUs never performing WDLS at all (Prendergast et al. 1998).

Clearly, this degree of variability in practice cannot be explained by local variability in case mix and illness severity. The fact is that patients, SDMs, and healthcare teams all interpret medical information differently and have different ethical norms that guide behavior. One study found that Canadian neurosurgeons, neurologists, and intensivists used highly variable approaches to prognosticate for patients with traumatic brain injury (Turgeon et al. 2013). In this study, when presented with a hypothetical case and asked if the 1-year prognosis was poor, one third agreed, one third disagreed, and one third were neutral; only 10 % were comfortable recommending WDLS. In another famous study, an international audience of healthcare practitioners was asked how they would manage a patient with a probable advanced cancer and an acute subarachnoid hemorrhage due to a ruptured aneurysm, whose family was divided over whether to continue aggressive life-sustaining measures or to WDLS (Kritek et al. 2009). Regional responses varied greatly; 24–68 % said they would WDLS, 15–30 % would continue aggressive care and consult an ethicist, and 16–49 % would write a Do Not Resuscitate order and transfer the patient to a skilled nursing facility.

Religion can also be an important consideration in EOL decision-making (Bulow et al. 2008). Jewish, Muslim, or Greek Orthodox patients who die in the ICU tend to have life support withheld (WHLS) rather than withdrawn (WDLS), whereas Catholic or Protestant patients have equal proportions of WHLS and WDLS (Sprung et al. 2007). Of note, the religion of the physician was as important as that of the patient for predicting whether life support would be withheld or withdrawn. There are many other demographic and system-related factors that are also associated with EOL decisions in the ICU, including race, socioeconomic status, insurance coverage, and bed availability (Muni et al. 2011; Barnato et al. 2007; Fowler et al. 2010; Stelfox et al. 2012).

The fact that physicians interpret clinical information inconsistently and that EOL decisions appear to be influenced by a host of nonmedical considerations should remind all clinicians of the need for humility and open-mindedness when discussing EOL decisions. This does not mean that physicians should refrain from providing an opinion or recommending an EOL decision when they feel it is appropriate, but physicians need to recognize the importance of an ethically sound approach to decision-making and conflict resolution.

Autonomy is a core bioethical principle that underpins most of modern western medicine. Our respect for autonomy is manifest in the importance we attach to informed consent, disclosure, and other everyday occurrences. Respecting autonomy is complicated when making EOL decisions in the ICU, because most ICU patients are incapable of participating in decisions about their care. We then rely on substitute decision-making to provide the perspective of the patient. Ideally, SDMs would make decisions on behalf of the patient using the patient’s values. In reality, SDMs find it difficult to make decisions without applying their own values and interests (Vig et al. 2006), and their predictions of patient preferences are often inaccurate (Shalowitz et al. 2006). Sometimes, patients have recorded their preferences in a document such as a living will or an advance directive. Unfortunately, such documents are often difficult to apply at the bedside because they do not contain instructions that can be applied in the present clinical situation (Teno et al. 1997).

Despite these limitations we should still strive to obtain a clear picture of a patient’s preferences when making decisions on their behalf. In neurosurgical patients, EOL decisions often hinge on a consideration of functional outcome and quality of life. These are highly subjective concepts, and a quality of life that would not appeal to a young, highly functional healthcare worker may be perfectly acceptable to other people, particularly when the alternative is death.

Non-maleficence is a principle rooted in the Hippocratic tradition and is a major driver behind EOL decision-making. Aggressive medical care, surgical procedures, life support, and cardiopulmonary resuscitation can all be associated with harms, such as pain, suffering, and loss of dignity for both the patient and the family. We accept the risk of harm when there is a reasonable prospect of benefit, but as the prospect of benefit decreases, the risk of harm is hard to justify.

On the other hand, some would argue that we have a duty to prolong life in any condition, regardless of the potential for harm or the prospects for recovery. According to this principle (“vitalism”), all life is of infinite value, and death is the ultimate harm that should be avoided. Although vitalism is not endorsed by any professional society, it is a principle that is sometimes invoked by SDMs to justify ongoing aggressive care for patients with profound neurological impairment (e.g., persistent vegetative state).

EOL decisions are typically motivated by a desire to avoid harm, but our illustrative case also emphasizes the importance of trying to do good. Family members will sometimes advocate for unproven or experimental therapies in the hope that these will restore health and function to patients with severe incurable illness. At times, healthcare workers may agree, simply because they feel that the patient has nothing to lose. But the condition of the patient does not override the principle of ensuring that all therapies offered to patients have a reasonable likelihood of being beneficial. This does not mean that a physician cannot offer off-label, unproven, or experimental therapies; he/she must be convinced that there is a realistic prospect that these therapies will be beneficial, either because of published reports, previous experience, or biological plausibility. But the principle of beneficence is as valid for incurable patients as it is for healthy ones.

Justice is arguably the most challenging principle to apply in EOL decisions, simply because there are so many concepts of justice. From a simplistic perspective, a physician should always try to make EOL decisions and recommendations based on purely clinical considerations, viewed in the context of the patient’s values. In other words, treat like cases alike. However, this is a naïve exercise. Physicians cannot avoid considering nonclinical factors and applying their own culture and values to a given decision. So to ensure that EOL decisions are made consistently, it may be necessary to create explicit standards of care (Umansky et al. 2011) or even use interdisciplinary committees to review challenging cases (Ford and Kubu 2006).

At a higher level, physicians must also consider the importance of “fairness” in making EOL decisions, and how “fairness” can be defined when allocating a scare resource such as a neurosurgical ICU bed to a patient with little or no hope of recovery. On one hand, a physician might consider Rawls’ argument that the greatest proportion of resources should be allocated to those who are most disadvantaged (Rawls 1999). Given our patient’s condition, this would suggest that she ought to receive a commensurately high level of resources. On the other hand, a physician could take the position that if every person in a persistent vegetative state were kept alive in a neurosurgical ICU bed, these resources would not be available to other patients seeking to have the same surgery as our patient. In other words, future patients would be denied the opportunity that was given to this patient, which would be manifestly unfair.

The approach described above does not always yield agreement between the patient, SDM, and healthcare team. Conflict is common in the ICU setting and often involves disagreements about the intensity of care provided for patients with a poor prognosis (Azoulay et al. 2009). In situations like our case, a common response to conflict is disengagement on the part of the healthcare team and continuation of aggressive care. This approach may avoid angry confrontations and medicolegal consequences, but it may ultimately be harmful for all concerned. The patient will continue to receive aggressive care without a realistic prospect of benefit. The medical team will lose the ability to provide support to the husband at a time when he clearly needs it. The ongoing care will consume a large quantity of healthcare resources, and it may contribute to burnout and distress among the ICU staff (Poncet et al. 2007).

The optimal approach would be to use an ethically sound decision-making process to develop a plan of care that is acceptable to all parties, demonstrated in a published checklist for meeting the ethical and legal obligations (Sibbald et al. 2011). This checklist is shown in Table 8.2 and can be applied in many other jurisdictions as well. If conflict escalates, physicians should attempt to use conflict resolution techniques shown in Table 8.3 (Knickle et al. 2012) or mediation by a third party. Mediated solutions are usually preferable to arbitrated solutions, because in mediated solutions the two parties to the conflict are able to craft the resolution themselves. In arbitrated solutions, the resolution is created by a third party and may be unacceptable to both the healthcare team and the SDM. It is often useful to consult one’s hospital Ethics Committee and/or the bioethics consult service which is usually staffed by appropriately ethics-trained physicians, social workers, PhDs, theologians, or others (Bernstein and Bowman 2003).