Can Clinical Interventions Change Care at the End of Life?

Physicians and nurses have long struggled with decisions about appropriate care for patients who are near death. During the past 20 years, patients and their families have become more likely to participate in these decisions. Many persons believe that Americans die in too much pain, with excessive technology and little human dignity, and at too great a cost. Most of the general public now supports physician-assisted suicide as a way to escape the failures of modern health care [1]. With increased public awareness has come a demand for new standards of end-of-life care that are better tailored to the needs of the dying patient. Despite intense attention to this problem, an effective solution remains elusive. Researchers have studied various interventions, including education of patients and physicians, written advance directives, and facilitated discussions with documentation of patient treatment choices in the medical record. These interventions have been designed to meet clinical goals, such as increasing the use of patient preferences for treatment, minimizing pain and suffering, reducing inappropriate use of life-sustaining treatments, and reducing costs. Some authors have assumed that these goals can be combined. They believe that if patients are able to express treatment preferences, physicians will provide terminal care that better meets the needs of patients and is less costly [2, 3]. Other experts have argued that care near the end of life will remain expensive [4], even if patients agree to forgo some high-cost treatments. Although some studies have claimed success, others have been disappointing. The most comprehensive intervention, SUPPORT (Study to Understand Prognoses and Preferences for Outcomes and Risks of Treatments) [5], was the largest and most carefully designed study in this field. Study nurses communicated detailed information on prognosis and preferences to physicians and their critically ill patients. This costly and time-intensive intervention yielded definitively negative results. After 2 years of study and random assignment of 4804 patients with an average life expectancy of 6 months, the investigators found no change in the frequency of discussions about treatment preferences, physicians' knowledge of patient preferences, or use of intensive care or ventilator support. Pain control did not improve, and cost of care was the same for the intervention and control groups. The failure of this intervention challenges many basic assumptions of clinical ethics. Changing medical care at the end of life is clearly difficult, and clinicians and investigators need to base future efforts on previously learned lessons. The primary objective of this review is to describe characteristics of successful clinical interventions. We examine the relation between the type of intervention and its success in achieving four clinical goals: increasing the use of patient preferences, minimizing pain and suffering, reducing the use of life-sustaining treatments, and reducing costs. Methods We conducted a systematic review of published interventions designed to change clinical practices and improve care at the end of life. We searched the MEDLINE database for English-language articles published from January 1990 to March 1996 using advance directive, terminal care, controlled trial, intervention, life-sustaining treatment, and prospective as key words or text words. To identify additional studies, we searched abstracts presented at recent national meetings and the reference lists of the selected articles. We included original research articles if data were presented on the effects of a clinical trial designed to change the experience of patients or the practices of physicians who provide care near the end of life. Because human behavior may be influenced by education, incentives, or organizational influences, we did not restrict our review to one type of intervention. We accepted studies that used historical controls or did not define a control group, because relatively few randomized, controlled trials have been done. We excluded any study that did not have the potential to affect clinical care, such as those limited to changes in attitudes or knowledge. We also excluded studies of patients or physicians outside the United States because of important variations in practice norms. Because of the heterogeneity of interventions, we grouped studies by descriptors of the interventions rather than by study design. Studies were first grouped by the target population: patients, physicians, or both. We then examined whether an intervention led to changes in four desirable clinical outcomes: use of patient preferences, pain control, use of life-sustaining treatments, or medical costs. Finally, we compared positive and negative trial results for differences in intervention, target group, or the strength of research design and methods. Funding sources for this review did not influence our analysis or interpretation of the data. Data Synthesis Sixteen studies met the inclusion criteria. Eight interventions targeted patients, five targeted only physicians, and three attempted to influence behavior in both groups. All studies attempted to increase use of patient preferences; some also examined changes in pain control, use of life-sustaining treatments, and cost. Interventions for Patients Eight studies used patient-education strategies to increase the recording of advance directives or proxy choices [6-13] (Table 1). Six of the eight trials were successful [6, 8-11, 13], and one achieved a particularly high rate of advance directive completion [10]. Five studies were randomized, controlled trials [7, 9, 11-13]; three used historical [10] or nonrandom [6, 8] control groups. Patients were enrolled in outpatient settings and were generally healthy. Five trials selected only elderly patients for study [7-10, 13]. Some sampling strategies also restricted the intervention to patients at higher-than-average risk for acute illness [13]. Baseline rates of documented preferences or proxies ranged from 0% to 31%. The rates of recorded proxies or advance directives after intervention ranged from 15% to 71%. The two studies in this group with negative results did not differ from the others in sampling strategies, but both were randomized, controlled trials. Table 1. Patient Education Interventions All interventions provided patients with education about life-sustaining treatments and their right to express preferences about use of these treatments. Most successful trials did not rely on written materials alone. In the two studies with negative results, written materials were supplemented only by interviews with researchers [7, 12]. The only successful study to use educational materials alone was a randomized, controlled trial in a large, well-established health maintenance organization [9]. In response, 19% of patients in the intervention group completed proxy forms. This success rate may have resulted, in part, from the investigators' choice of outcome measure. Unlike in other studies, advance directives were counted if they were mailed back to the research team; the directives did not have to be documented in the patients' medical records. These patients may also have been more inclined to complete written advance directives than those who were uninsured or otherwise disenfranchised [14]. All other successful studies in this group used patient-education materials and discussions with a physician or other clinician. In two of the earliest trials [6, 8], researchers reported success rates of 15% to 46% with the use of clinical discussions after they had found no change in behavior with the use of written materials alone. Richter and colleagues [11] designed a randomized, controlled trial that combined an initial discussion with a physician and mailed educational materials, resulting in increased completion of advance directives (3% before the intervention and 23% after the intervention). The highest rates of use of advance directives (71%) and proxy appointment (68%) were achieved when a social worker held repeated discussions with patients during follow-up clinic visits; this gave patients the time to ask questions and to reach thoughtful decisions [10]. Meier and colleagues [13] found similarly high rates of proxy appointment (48%) in a randomized trial of patient-advocate intervention in which repeated discussions were permitted. In summary, most educational interventions for patients increased documentation of patient treatment preferences. More patients recorded preferences when they had repeated opportunities to discuss their decisions with a physician or a social worker. Interventions for Physicians A second group of educational interventions was designed for physicians, who must elicit and respect patient preferences if they are to influence the treatments the physicians order [15-19] (Table 2). Investigators used sophisticated educational or motivational techniques in these trials, all of which achieved some positive results. However, these studies also required less rigorous standards of evidence. No trial in this group used randomly assigned controls, and some used patterns of practice for physician groups as a proxy for behavioral changes in individual physicians. All studies were conducted in a single institution or practice, where concerns about statistical power and cross-contamination make randomization less feasible. Table 2. Physician Education Interventions Two interventions were designed to increase use of patient preferences by increasing the number of written advance directives in outpatient records. Markson and colleagues [15] studied the impact of a 5-week didactic course for physicians. After the course, physicians still did not initiate discussions of advance directives; therefore, the trial was modified. In the final intervention, only patients receiving continued care were enrolled, extra clinic time was allow