218 End-of-Life Issues in the Intensive Care Unit

Nicholas S. Ward, J. Randall Curtis, Mitchell M. Levy

In the last century, the process of dying changed dramatically. Previously, doctors simply did all they could for a patient, and when their treatments failed, the patient died, almost always at home. However, with the advent of more sophisticated medical technologies, even patients with severe organ failure can be kept alive. Unfortunately, with this progress has come a new set of complex medical, ethical, and societal issues.

In the United States, about 80% of people now die in healthcare facilities (60% in acute care facilities),¹ despite the fact that about 90% of Americans polled say they would prefer to die at home.² This disparity is due to two factors: (1) many people die while undergoing treatments meant to postpone death—treatments that are often futile; and (2) many families are unable to care for a dying person or are uncomfortable having a loved one die at home. The net result is that most people will die in a hospital or other healthcare facility, and many of them will undergo high levels of medical care before death. A recent study showed that about 20% of Americans will die in an intensive care unit (ICU) or be admitted to an ICU just before death.³

Two conclusions can be drawn from the preceding information. One is that a tremendous amount of health care is being delivered to dying patients. This has been reflected in several studies, such as that of Cher and Lenert, which showed that a relatively large percentage of Medicare expenditures goes to treat patients in the last weeks of their lives.⁴ The other conclusion is that doctors have to learn a new set of skills that were not necessary in the past. They need to be able to recognize patients who are going to die despite medical care and help decide which of the almost limitless supply of medical therapies available are appropriate and which are not. Physicians need to guide their patients through a maze of medical options in an attempt to balance preservation of life with quality of life—a daunting task, to say the least.

How Are Critically Ill Patients Dying?

In 1995, a landmark study in end-of-life issues was published. This was the first large-scale attempt to define how seriously ill people die in American hospitals. In a two-part study involving 4301 seriously ill, hospitalized patients, the investigators examined multiple aspects of end-of-life care and found major shortcomings in current practice. Only 47% of the time did the physician know when a patient wanted to avoid cardiopulmonary resuscitation (CPR), and the incidence of dying with moderate or severe pain was 50%.⁵

More insight into the dying experience came in subsequent studies that showed that the vast majority of ICU deaths occur only after a decision to limit life support has been made.^6,⁷ In two important studies, Prendergast and coworkers helped define how patients die in ICUs.^6,⁸ In their first study, they compared deaths in their ICU from two periods, 1987 to 1988 and 1992 to 1993, to determine how often CPR was performed before death and how often limits were placed on life support before death.⁹ Their data showed that the incidence of CPR before death had declined from 49% to 10% and that the incidence of withholding or withdrawing life support had increased from 51% to 90% of all ICU deaths.

In an effort to benchmark their data with the rest of the country, the same investigators then did a large follow-up study 1 year later. They collected data from more than 6000 patient deaths occurring in 131 ICUs in 38 states over a 6-month period and analyzed the data for the incidence of various limits on life support. They found that on average, only 25% of patients dying in ICUs were given CPR before death. About 70% of patients had some restriction on life support, and almost 50% of patients had some medical therapy withheld or withdrawn before death.⁶

The other striking piece of data to emerge was the degree of variability that existed among ICUs. The incidence of patients dying with full aggressive measures ranged from 4% in one ICU to 79% in another. Likewise, the incidence of withdrawing life support ranged from 0% to 79%, depending on the ICU. These data clearly show that although the overall practice of limiting life support in ICUs in the United States is common, there is tremendous variability from place to place in end-of-life care.

In 2003, a study examined deaths in 31 ICUs in 17 European countries. Overall, the percentage of patients dying with some limits on life support was 72.6%, which was very similar to studies done in the United States. As in the American studies, there was also tremendous variability in practices among the different ICUs, with rates of CPR before death ranging from 5% to almost 50%.⁹

What Accounts for Variability in Practice?

It should not be surprising that there is so much variability in a practice as multidimensional as end-of-life care. Even the standard practice of medicine varies from institution to institution. The decision to limit or not limit life support is generally a complex one that may reflect the personal biases of both physician and patient. Many attempts have been made to find patterns among different types of physicians and patients that can explain the variation. For example, one study showed that university-based physicians are more likely than community-based physicians to write do-not-resuscitate (DNR) orders and withhold or withdraw life support.¹⁰ A similar study showed that patients without private physicians in the ICU were more likely to undergo active withdrawal of life support.¹¹ Unfortunately, studies like this are hard to interpret unless one knows the contexts in which these decisions were made. An increased tendency to withhold life support from a terminally ill patient may reflect a weaker physician-patient relationship or a stronger one based on the patient’s preferences.

Other studies have sought to explain variation by culture, race, or religion.¹²^–¹⁶ Although such factors may play a role in these important decisions, and there may be some general trends in decision patterns, there is enough variation even within cultures, races, or religions to indicate that one cannot generalize this information to a given individual. Physicians need to be cognizant of the fact that their patients may have markedly different views of optimal end-of-life care, regardless of their culture, religion, or race.

Predicting Outcomes

A central problem complicating end-of-life decisions is the difficulty of predicting outcomes in critically ill patients. The combination of multiple coinciding medical problems and rapidly changing clinical status can make this a very difficult task. Essentially, two objective tools are available to a physician when trying to determine the prognosis of a critically ill patient: published outcomes and severity scores. Both can be helpful, yet have limitations. In addition, the role of a physician’s personal experience is important but can also have important limitations.

Severity Scores

Severity scores have been available for almost 3 decades, and much has been learned in that time. In most severity score algorithms, data are collected during the first 24 hours of admission and are then used to compile a score that predicts the risk of death during hospitalization. Many of these scoring systems were developed by reviewing data from thousands of ICU patients and employing logistic regression models to choose some important input variables. Other variables were simply chosen based on presumed clinical value. These scores were then validated prospectively on patients.

Unfortunately, there are several problems with these severity score–based systems. First, they make predictions based on hospital outcomes at the time of their creation. Thus, as medical treatments improve, the scores must be updated. In the 1970s, for example, acute respiratory distress syndrome (ARDS) had a hospital mortality approaching 80%, so its presence might justifiably increase a patient’s severity score. Presently, ARDS has about 40% hospital mortality. Thus, a severity scoring system using the diagnosis of ARDS, or even components of the diagnosis such as hypoxemia, would have to be adjusted. Some commercially available proprietary severity scoring systems, such as the Acute Physiology and Chronic Health Evaluation (APACHE) methodology, are updated and revalidated on a regular basis to avoid this problem, but many that are widely used today, such as APACHE II, are based on patient data collected as many as 2 or 3 decades ago.

Another problem with using severity scores is that most models derive their predictions from factors present at or shortly after admission to the ICU and do not provide updated mortality estimates as the patient’s condition changes. Further, severity scores often give intermediate mortality estimates such as 60% instead of a clear yes-or-no answer. Even these numbers are subject to confidence intervals. Perhaps the most glaring problem of using severity scores is that they say nothing about morbidity, disability, or survival after hospitalization. These factors are often just as important as risk of death in making end-of-life decisions. A patient may accept a 30% chance of survival if it were followed by a high quality of life, but not accept a 70% chance of survival if it were likely to entail a poor quality of life.

Outcome Data

Many of the same problems encountered with severity scores characterize the use of outcome data. Although published outcome studies are an essential tool for clinicians in predicting a course of illness, they suffer from two major problems. One is that the population studied for a particular illness may not share the same characteristics as a particular patient. For example, in a large multicenter clinical trial of a new therapy for sepsis, the mortality rate in the control (untreated) arm was 31%.¹⁷ It is important to note, however, that this trial excluded patients with conditions such as renal failure, liver failure, pancreatitis, acquired immunodeficiency syndrome (AIDS), and a variety of other comorbid conditions, thus limiting the usefulness of these data for prognostic purposes.

Another problem with using outcome data, similar to the severity scores, is the rapidity with which therapies can change and improve. In four published studies by different authors between 1981 and 2000 examining the mortality of Pneumocystis carinii pneumonia in ICU patients, the mortality decreased from 86% to approximately 50%.¹⁸ Similar changes in outcome over time have been reported with a variety of other illnesses, such as ARDS, as treatments have improved.

Caring for Families in the Intensive Care Unit

Impact of Family Meeting

Relatives, partners, and friends often provide support and care for a patient, which for some will include the responsibility of surrogate decision making. Surrogate decision makers are often under an enormous amount of emotional stress, and decision making during these circumstances can be difficult. In addition, one study revealed that despite discussions with ICU physicians, only half of families of critically ill patients adequately understood their loved ones’ diagnoses, prognoses, or treatments.¹⁹ Despite this, clinicians and health systems often neglect the care of the family as part of the overall care of a patient. Therefore, clinician-family communication is an important component of good quality care. In addition, effective clinician-family communication reduces the stress on family members of critically ill patients and improves the family members’ level of understanding. This is of critical importance in the ICU because if the patient’s family is under significant distress, their ability to provide surrogate decision making may be impaired, and the medical decisions they make may not accurately reflect the wishes of the patient.

In addition to the multitude of data showing that the way we communicate with families has a significant impact, there are also data showing that interdisciplinary team communication has a significant impact on important patient and family outcomes. Observational studies show increased survival, shorter lengths of stay, and improved patient satisfaction when there is good nursing-physician communication.²⁰ In addition, patients and families have reported that interdisciplinary communication is an important part of good end-of-life care, and most studies of interventions that have improved end-of-life care include an interdisciplinary team in the intervention. Unfortunately, some ICU family meetings occur only between the physician and the family. Underutilization of the many professionals involved with a critically ill patient’s care is a common mistake. Care of ICU patients is provided by a large interdisciplinary team that includes consulting physicians, nurses, social workers, and members of the clergy. These healthcare providers often know the patient and family from different perspectives, and holding a meeting without attempting to have all relevant members present can result in miscommunication and may result in missed opportunities to provide families with the best possible resources.

< div class='tao-gold-member'>

Only gold members can continue reading. Log In or Register to continue