Special Feature

End-of-Life Care in the ICU: Perspectives on Surrogate Decision-Making and Managing Conflicts

By Leslie A. Hoffman, RN, PhD, Department of Acute/Tertiary Care, School of Nursing, University of Pittsburgh, is Associate Editor for Critical Care Alert.

Dr. Hoffman reports no financial relationship to this field of study.

One in 5 Americans dies in or shortly after discharge from an intensive care unit (ICU), an outcome that exposes critical care clinicians to frequent discussions that concern end-of-life care.1 This essay focuses on 4 topics of importance to end-of-life care in the ICU: 1) accuracy of surrogate decision-making; 2) dealing with conflicts; 3) surrogate willingness to endorse genetic research; and 4) opportunities for improving end-of-life care.

Accuracy of Surrogate Decision-Making

Prior to the 1980s, medical decision-making was generally paternalistic. Physicians, guided by ethical principles, decided what treatment was the most appropriate, with little input from the patient or family.1,2 Since 1980, there has been an increasing emphasis on allowing patients to make their own decisions, including the right to choose what medical care they do and do not want to receive.2 In 2004, strong support emerged for a shared decision-making approach.3 In this model, the physician works with the patient and family to determine health care values, beliefs, and preferences. The physician not only provides information, but also recommends the treatment that is viewed to be the most consistent with the patient's values and goals. This model blends the opposing forces of patient self-determination and the patient's need for guidance regarding complex management decisions.1 Because most patients who are near death in the ICU are unable to voice their preferences, a surrogate is typically asked to assume the role that would normally be assumed by the patient and make decisions that best fit with the patient's wishes or best interests.1

Surrogate decision-making is viewed as extending patient autonomy by allowing the preferences of patients to guide their care, even if they cannot make these decisions themselves. However authors of a recent meta-analysis concluded that surrogates incorrectly predict end-of-life treatment preferences in approximately one-third of cases.4 From a systematic literature search, 16 studies were identified that involved 151 hypothetical scenarios, 2,595 surrogate-patient pairs, and 19,526 patient-surrogate paired responses. Twelve studies assessed whether surrogates were more likely to favor or reject treatment the patient might want. Three studies found that surrogates tend to err in favor of approving interventions that the patient did not want and one found the opposite. Eight studies reported mixed results with no consistent trend in surrogates' mistakes. In 5 studies, investigators assigned the patient's surrogate based on the relevant state's relationship hierarchy and in 6 studies patients designated who they wished to be their surrogate. Patient designated surrogates were correct 69% of the time and legally assigned surrogates 68% of the time. There are also indications that surrogates and patients view consequences of treatment differently, which may preclude greater uniformity in opinion. In a study in which agreement occurred 70% of the time, patients selected "burden on the family" and "time left to live" as the most important factor to guide decision-making, whereas surrogates identified "pain" as the most important factor.5

Several studies have evaluated the potential of using a community standard for reference. In this model, the "community" consists of patients possessing experience with specific health states, diseases or treatments and the "standard" represents the preferences of these patients.6 The argument for using a community standard is based on the belief that groups of people share core values that can used analyzed to identify end-of-life treatment choices in various situations. In a study comparing the accuracy of such a model and decisions of surrogates, surrogates who had known the patient 47 ± 6 years were unable to predict preferences more accurately than the model. Both surrogates and the model had an average accuracy rate of 74-75%.6 This form of decision making has several advantages. If preferences are unknown, modal preferences can be used to guide the discussion.6 Preferences derived in this manner can be particularly helpful because they describe the choices of individuals who had actual experience with a specific disease or treatment.

Of interest, the rate of surrogate-patient agreement appears to be slightly higher when the choice involves whether to participate in a clinical research trial. Patients scheduled for elective cardiac surgery were asked to evaluate two study protocols (one low risk and one high risk). The patients and surrogates were then asked to indicate if they would consent to study participation. The surrogate's prediction was incorrect in 15% of cases for the low-risk study and 20% of cases for the high-risk study.7

Despite these findings, four studies have indicated that surrogates are more accurate in predicting treatment preferences than physicians.4 Therefore, in the absence of alternate methods, current reliance on surrogates or a community standard can be defended as the best available method for implementing the substituted judgment standard.1-7

Dealing with Conflicts

Interpersonal conflicts regarding complex decisional issues are inevitable in the high-pressured work of health care.8 Poorly managed, conflict saps productivity and spawns additional disputes. Well-managed conflict can enhance the self-esteem of all involved parties and engender creative solutions beyond expectations. Just as thoughtful differential diagnosis precedes optimum treatment, management of conflict involving different preferences is greatly enhanced when preceded by careful assessment.8

Conflicts commonly arise in the context of end-of-life treatment decisions. A common cause of such conflicts arises from family demands for care that the health care team believes is futile.9 When conflicts occur, the option exists to deal with the issue promptly and directly or delay action. Both choices can be appropriate.9 The most important pitfall in dealing with conflicts is to begin negotiations when you are angry, defensive, or preoccupied with emergent issues that make it impossible to devote sufficient time to resolving the conflict.9 Additional pitfalls and useful communications strategies are summarized in Tables 1 and 2.

Several authors have developed a step-wise approach to resolving conflicts which involve issues related to end-of-life decision-making that involves surrogates.9,10 The first step involves acknowledging that a conflict exists. The second step involves preparing for discussion by reviewing known issues, what has been done, and selecting a time to address the issue when there are a minimum of distractions. Stone10 recommends approaching such situations with curiosity directed toward eliciting the "real" issues which are prompting the conflict. In such situations, affective issues are likely more important than the facts. Mediation experts recommend that, rather than ignoring emotions, one should openly acknowledge these on the part of both parties in the dispute.

Opportunities for Improving End-of-Life Care

Evidence from a number of studies suggests that families of critically ill patients are not satisfied with the amount and clarity of information they receive from health care providers.1 From focus group discussions, surrogates identified six domains of information that they believed would assist in patient-centered decision-making in the setting of critical illness.11 These 6 domains were:

  1. the nature of the patient's illness and treatment;
  2. prognosis for outcomes, including ventilator independence, survival, functional recovery and an acceptable quality of life;
  3. potential for pain and other distressing symptoms;
  4. potential complications of treatment;
  5. expected care needs after hospitalization; and
  6. alternatives to continuation of treatment into the chronic phase of critical illness.

Participants observed that the transition from acute to chronic critical illness passed without focused discussion of the potential for future recovery (functional capacity, quality of life) or alternative plans of care (withholding or withdrawing treatment). When discussions took place, participants observed that they typically occurred when treatment was prolonged and death was imminent.11 The participants all shared that they felt it was relevant and important to discuss possible options, including termination of mechanical ventilation and other life-sustaining treatments earlier in the course of illness.

Surrogate Consent for Genetic Research

Several clinical trials have examined or are actively recruiting patients for studies involving gene-based therapeutics.12 Studies that involve genetic research pose unique challenges in the critical care environment because patients are unlikely to have discussed their wishes with their surrogate decision-maker, the surrogate may have no personal knowledge about this treatment option and the surrogate may need to make a quick decision because the therapy must be administered promptly after diagnosis of a complication, eg, sepsis.12 Requests to participate in clinical trials that involve genetic testing introduce complex issues into the patient-provider relationship, due to concerns about potential loss of confidentiality, economic discrimination, and stigmatization.13

To explore how surrogates might react to a request for genetic samples, Freeman and colleagues12 interviewed surrogate decision-makers for 117 critically ill patients admitted to medical and surgical ICUs. The surrogates were 48.4 ± 1.2 years of age, 33% male, and 21% African-American. One-third (37.6%) had 12 or fewer years of education. Surrogate decision-makers were willing to permit genetic testing to aid diagnosis (95%), guide drug prescription (97%), or explain familial or ethnic disease predisposition (90%), without apparent concern that findings would be stigmatizing. Genetic testing was not perceived to be a barrier to participation in clinical trials, with 91.4% indicating they would be "very willing or somewhat willing" to give consent. Few (7.7%) indicated that availability of a small honorarium would make them "much more likely" to give consent. However, a majority were reluctant to permit testing if employers (79.5%), health insurers (76.9%), or life insurers (78.6%) could access results.

The subject of safeguarding clinical trial participants has received extensive coverage in the literature and media. In this discussion, limited attention has been given to exploring preferences toward enrolling in such studies in the setting of critical illness. Genetic research presents unique ethical challenges not encountered in other studies enrolling critically ill patients. This study, one of the few to address this issue, suggests that surrogate decision-makers are willing to support such studies providing that data are collected anonymously. Future studies are needed to more fully understand how patients and surrogate decision-makers view this treatment option.


Dealing with end-of-life decision making is a critical, but difficult, skill to master. Recognizing and actively dealing with such issues can improve dialogue, and help guide surrogates and other health care providers through difficult decisions in a manner that results in a solution acceptable to all parties. The effort is time-consuming and challenging but the rewards justify the effort. As new insights evolve, it is likely that ability to provide appropriate support will continue to improve.

Table 1 and Table 2


  1. White DB, Curtis JR. Curr Opin Crit Care. 2005;11:610-615.
  2. Emanuel EJ, Emanuel LL. JAMA. 1992;267:2067-2071.
  3. Thompson BT, et al. Crit Care Med. 2004;32:1781-1784.
  4. Shalowitz DI, et al. Arch Intern Med. 2006;166:493-497.
  5. Hare J, et al. Arch Intern Med. 1992;152:1049-1054.
  6. Smucker WD, et al. Med Decis Making. 2000;20:271-280.
  7. Coppolino M, Ackerson L. Chest. 2001;119:603-612.
  8. Siders CT, Aschenbrener CA. Physician Exec. 1999;25:32-37.
  9. Back AL, Arnold RM. JAMA. 2005;293:1374-1381.
  10. Stone D, et al. Difficult conversations: How to discuss what matters most. New York, NY: Viking Penguin Books; 1999.
  11. Nelson JE, et al. J Crit Care. 2005;20:79-89.
  12. Freeman BD, et al. Crit Care Med. 2006;34:986-994.
  13. Lapham EV, et al. Science. 1996;274:621-624.