Hanson LC, Earp JA, Garrett J, Menon M, Danis M. Community Physicians Who Provide Terminal Care. Arch Intern Med. 1999;159(10):1133-1138. doi:10.1001/archinte.159.10.1133
Most dying patients are treated by physicians in community practice, yet studies of terminal care rarely include these physicians.
To examine the frequency of life-sustaining treatment use and describe what factors influence physicians' treatment decisions in community-based practices.
Family members and treating physicians for decedents 65 years and older who died of cancer, congestive heart failure, chronic lung disease, cirrhosis, or stroke completed interviews about end-of-life care in community settings.
Eighty percent of eligible family and 68.8% of eligible physicians participated (N=165). Most physicians were trained in primary care and 85.4% were primary care physicians for the decedents. Physicians typically knew the decedent a year or more (68.9%), and 93.3% treated them for at least 1 month before death. In their last month of life, 2.4% of decedents received cardiopulmonary resuscitation, 5.5% received ventilatory support, and 34.1% received hospice care. Family recalled a discussion of treatment options in 78.2% of deaths. Most discussions (72.1%) took place a month or more before death. Place of death, cancer, and having a living will were independent predictors of less aggressive treatment before death. Physicians believed that advanced planning and good relationships were the major determinants of good decision making.
Community physicians use few life-sustaining treatments for dying patients. Treatment decisions are made in the context of long-term primary care relationships, and living wills influence treatment decisions. The choice to remain in community settings with a familiar physician may influence the dying experience.
FOR MANY older and seriously ill patients, medical care at the end of life is characterized by pain, suffering, and the use of invasive life-sustaining treatments.1,2 There is growing public concern about the quality of care for the dying, and patients and families desire greater control over treatment decisions.3 Physicians and nurses agree that they provide aggressive care to too many dying patients and want to provide treatment that is more consistent with patients' preferences.4,5 However, studies6- 10 from academic medical centers have shown little relationship between patients' preferences and the treatments they receive. In the Study to Understand Prognoses and Preferences for Outcomes and Risks of Treatments (SUPPORT), seriously ill and dying patients were randomized to a nursing intervention to increase communication about prognosis and preferences. This carefully designed intervention had no effect on life-sustaining treatment use, pain management, or cost.
Experts have offered several explanations for the use of aggressive life-sustaining treatment near death. If patient preferences do not affect treatment, then the answer may lie in physician behavior. Physicians initiate most discussions of treatment options and offer explicit or implicit advice. They help families interpret advance directives for incapacitated patients, and they give orders to initiate or withhold treatments. Physicians' decisions may be influenced by factors other than patient preferences, including the physician's medical knowledge and training, characteristics of their practice setting, or type of relationships they have with patients.11- 15
Although most patients die in community hospitals, in nursing homes, and at home,16 nearly all studies of care at the end of life use patients enrolled in academic hospitals. Such facilities have resources for aggressive life-sustaining treatment, and physicians' relationships with patients are often transient.17 Most seriously ill and dying patients are treated by physicians in full-time community practice, yet no major study has described end-of-life care in this setting. We studied terminal care provided by community physicians to describe their use of life-sustaining treatment and to examine predictors of life-sustaining treatment use and perceived influences on decision making in community settings.
We used consecutive death certificates to identify a sample of elderly decedents with common chronic illnesses. Between September 1994 and December 1995, research assistants screened death certificates in 12 North Carolina counties. They sampled each county for 3 months. Counties were selected to represent a range of rural and urban communities but to exclude teaching medical centers. Poverty rates for persons 65 years and older in these counties ranged from 13% to 36%, and physician supply ranged from 3 to 33 physicians per 1000 residents 65 years or older.18
Decedents were eligible if they were 65 years or older at the time of death and if their primary or secondary causes of death included chronic obstructive pulmonary disease, cancer, hepatic cirrhosis, congestive heart failure, or stroke. Since these last 2 diagnoses may be acute, such patients were excluded if the death certificate indicated onset at less than 1 month before death. These common diseases were chosen because they account for most deaths among elderly patients and their long-term course allows for the anticipation of treatment decisions.
Age, sex, race, and cause and location of death of decedents were abstracted from death certificates. The informant on each death certificate—usually a family member—was contacted for a telephone interview. Trained interviewers asked informants about their satisfaction with treatment decisions, treatment intensity, and symptom relief during the final month of life. Interviews with informants have been reported elsewhere.19 After reading definitions of cardiopulmonary resuscitation (CPR), ventilatory support, and intensive care, the interviewer asked "Did any of the doctors caring for [the decedent] discuss whether or not to use these treatments with him/her?" They asked a similar question about a discussion with family if the patient was not capable. We considered a discussion of life-sustaining treatment options to have occurred if the family reported a discussion with a patient or family surrogate. Family informants also indicated whether the decedent had a living will.
Family members identified the physician who was most responsible for the decedent's care during the final month of life. We invited this physician to participate in an in-person interview. If a physician cared for more than one decedent, he or she was interviewed about the first eligible death. To encourage participation, physicians were offered breakfast or lunch, a modest payment, and a summary of recent guidelines on end-of-life care.
Trained interviewers conducted the hour-long interview in physicians' offices, where physicians were asked to have the decedent's chart available for reference. The initial questions detailed treatment decisions made during the decedent's last month of life. Physicians indicated whether CPR, ventilatory support, intensive care, or tube feedings were used and whether active decisions were made to withhold other treatments or to give comfort care only. Although the family informant's answer was used to define whether a discussion occurred, physicians were also asked whether they discussed life-sustaining treatment decisions with the patient and family. They indicated when the first such discussion took place and evaluated the decedent's decisional capacity during the last month of life. Physicians also answered questions about the duration of their relationship with the decedent and whether they acted as his or her primary care physician.
Physicians answered general attitudinal questions about the influence of reimbursement, living wills, and duration of the patient-physician relationship on end-of-life decisions. Additional questions asked them to describe their personal and practice characteristics and education in ethics and care of dying patients.
The interview included 2 open-ended questions about influences on the decision-making process. Physicians were asked, "Think about the older patients in your practice who have died, for whom decisions about medical treatment in the last month of life went well, the right decisions were made, and the decision-making process went well. What made this happen?" They were then asked a parallel question about patients for whom decisions did not go well near the end of life.
All aspects of the study were reviewed and approved by the university Committee on the Protection of the Rights of Human Subjects.
Since any individual patient may have multiple treatment decisions, we created a summary score for the intensity of treatment to sustain life during the final month. Each decedent received a treatment score based on the most aggressive level of treatment given to them. A score of 7 corresponded to the use of CPR during the last month of life, 6 to ventilatory support without CPR, 5 to intensive care, 4 to a feeding tube, 3 to routine care without such treatments, 2 if other treatments were actively withheld, 1 if hospitalization was withheld, and 0 if comfort care was given during the entire last month of life.
Open-ended questions were coded using a consensus process. Investigators reviewed a 20% random sample of the responses to identify major themes and establish a coding scheme. Three investigators assigned codes to all responses and discussed disagreements until consensus was reached. If a physician answered with more than one concept, multiple codes were permitted. An experienced researcher who was not a member of the research team then reviewed coded data for completeness and consistency in the final codes.
Interview responses were analyzed using standard descriptive statistics. Physician and patient characteristics were tested for their relationship to the use of life-sustaining treatment measured by the treatment score. Physician characteristics were age, sex, religion, specialty, practice characteristics, ethics education, and duration of relationship with the decedent. Patient characteristics were age, sex, race, decisional capacity, living will, cause of death, and place of death. Correlation coefficients were used to test for associations between continuous variables and treatment score. One-way analysis of variance was used to test for associations of categorical variables with treatment score. A single multiple regression model tested physician and patient characteristics as independent predictors of treatment intensity in the last month of life.
Of 575 eligible family informants, 461 (80.2%) completed the telephone interview. Informants named 240 treating physicians, 165 (68.8%) of whom completed the in-person interview. Physicians who participated did not differ from those who refused when compared with decedent characteristics or family report of conflicts in decision making.
Physicians' responses were linked to 165 informant interviews and death certificates corresponding to the first eligible decedent they had treated. Decedents' mean (±SD) age was 79±8 years, 51.5% were men, and 88.4% were white (Table 1). The most common cause of death was malignant neoplasms (49.7%), followed by congestive heart failure (18.1%), stroke (17.0%), chronic obstructive pulmonary disease (14.0%), and cirrhosis (1.2%). Forty-one percent died in hospitals, 29.7% at home, and 27.9% in nursing homes. Forty-four percent of the decedents had living wills, and most were capable of a discussion of treatment options during their final month of life.
Physicians who treated these patients were usually white, male, and practicing for an average of 15 years (Table 2). About 28% could not recall any education in ethics or care of dying patients, 10.3% were self-educated, 45.4% had brief formal education, and 16.4% had an extensive educational experience such as a course or seminar. Slightly more than half had trained in the primary care fields of general medicine or family practice. Regardless of training, the physicians characterized 90% of patient visits as primary care. When asked about their relationship to the decedent, 85.4% viewed themselves as the primary care provider. Their relationships were also continuous; 68.9% had known the decedent more than 1 year before death.
Community physicians reported using few life-sustaining treatments for these decedents. During the last month of life, 2.4% of decedents underwent a resuscitation attempt and 5.5% received ventilatory support. In many cases, physicians reported a decision to withhold other therapies (21.3%), to withhold hospitalization (17.1%), or to provide comfort care only throughout the entire last month (15.2%). Thirty-four percent of decedents were enrolled in hospice services, although in some cases this decision was made less than a month before death. Since multiple treatment decisions could be made for any one patient, the most aggressive treatment decision in each case was used to assign a treatment score ranging from 0 to 7. Mean (±SD) treatment score was 2.7±1.7, which corresponds to no CPR, ventilatory support, intensive care, or feeding tubes and some decisions to withhold other treatments.
Family informants said that a physician discussed life-sustaining treatment options with the decedent or family 78.2% of the time; physicians reported a somewhat higher rate of discussions (84.1%). For 72.0% of decedents, the first such discussion took place more than 1 month before death; 39.0% of discussions were initiated more than 6 months before death. When asked about the usual timing of such discussions in their practice, 45.6% of community physicians reported that they discussed life-sustaining treatment when a serious illness was diagnosed, 26.9% had discussions when patients were hospitalized, and 17.5% regularly had discussions with healthy outpatients.
The intensity of life-sustaining treatment was associated with several patient characteristics. Patients received more aggressive treatment if they died in a hospital compared with a nursing home or home (average treatment scores, 3.3, 2.6, and 1.9; P<.001). Patients who were dying of cancer had lower treatment scores than those with other causes of death (2.2 vs 3.1; P<.001). White patients received fewer treatments to sustain life than did African Americans (2.6 vs 3.4; P=.04). Patients who had prepared living wills had less aggressive use of treatment during the final month of life (2.2 vs 3.1; P<.001). Patient age, sex, decisional capacity, and informant report of a discussion of treatment options were not associated with intensity of treatment use. No physician characteristics were associated with treatment use for decedents.
In multivariate analysis, treatment score was the dependent variable; place of death, cause of death, living will, and patient race were the independent variables. Such variables explained some of the variance in treatment (R2=0.24), and the overall model was statistically significant (P<.001). Race was not independently associated with treatment after adjusting for other variables. Place of death, cause of death, and having a living will remained in the model as independent predictors of treatment use during the final month of life (P<.01).
Physicians were asked general questions about the influence of additional reimbursement, living wills, and duration of the patient-physician relationship on decision making near the end of life (Table 3). Only 38.8% of physicians believed that reimbursement for an annual discussion of life-sustaining treatment would trigger more discussions. However, 88.5% agreed that an expanded use of living wills would make such discussions easier. Eighty-four percent agreed that withholding life-sustaining treatment was easier in the context of a long-term relationship with the patient.
In response to the open-ended question about positive influences on decision making, 37.7% of physicians named advance planning, 35.2% named a good patient-physician relationship, and 33.3% named family support for the decisions. Among physicians who named advance planning, 1 in 3 mentioned living wills. However, all physicians described advance planning as a process—not a document—and that decision making improved with psychological preparation and early discussions that included family members. Typical comments in this category included, "[I] encourage patients to discuss decisions regarding end-of-life care with family members well ahead of a crisis," and "[The] presence of a living will symbolizes prior thought on the issues." These community physicians said that when a patient thought about the decisions in advance of a crisis—and communicated those thoughts to family and physician—decision making went well.
One third of physicians discussed the importance of a good patient-physician relationship based on effective communication and mutual trust. For example, "The physician gains the trust of the patient and family by informing them of options, answering and eliciting questions, and implying that they have all the time they need," and "If the physician has a good relationship with the patient and family, they trust me and I feel compelled to make a good decision." Nearly all such answers linked trust to effective communication, and only one physician implied that a good relationship might lessen the need for talking: "Once you have an understanding with a patient, then decisions are easy to make and follow naturally from that understanding without much discussion." Although we considered the quality of the patient-physician relationship, duration of this relationship, and family-physician relationship as separate categories, in combination, they accounted for 65% of the physicians' responses to this question (Table 3).
Community physicians also thought that family support for decisions was essential. Decision making went well when family members acknowledged the patient as primary decision maker and supported decisions made by physician and patient. Physicians wanted to work with a "family [that] is in agreement with treatment" and has "no out-of-town kids, no emotional garbage, no legal conflicts."
In contrast, when physicians were asked about bad end-of-life decisions, 43.5% felt that lack of family support led to most problems in end-of-life care. Physicians acknowledged that family members might legitimately disagree with decisions. However, they also perceived that long-standing family conflicts intensify during the crisis of critical illness. Problems come from "family with guilt feelings, who never paid attention to her during good days but come around when decisions need to be made," and "internal family conflict, without a clear spokesperson or unified philosophy."
Twenty-six percent of physicians said unrealistic expectations of medical care caused poor decision making. Almost all comments attributed such expectations to families as well as patients, and several physicians characterized these individuals as "irrational," "unsophisticated," or "poorly educated." Other common barriers to good decision making were a lack of advance planning (16%), sudden onset of illness (14%), and physicians' lack of relationship with patient (13%) or family (12%).
Physicians in community practice provide terminal care characterized by early discussion, respect for living wills, and few life-sustaining treatments for their elderly patients with chronic illness. In the last month of life, only 2% of these older dying patients had CPR, only 5% received ventilatory support, and most died after decisions were made to withhold hospitalization and other treatments or to provide only comfort care. Many patients died at home or in a nursing home treated by primary care physicians with whom they had long-standing relationships. Physicians discussed treatment options for 4 of 5 patients who died, and most initiated these discussions a month or more before death.
Decisions to forgo aggressive therapy were related to the patients' place of death. If a patient does not leave home or the nursing home, the resources to provide aggressive life support will be limited. Within this study population, the place of death was a predictor of the intensity of treatment use. No patient who died at home received CPR or ventilatory support, and 70% had hospice referrals. Even the decedents whose place of death was a community hospital received life-sustaining treatments infrequently. Six percent of patients who died in community hospitals received CPR and 12% received ventilatory support during their last month of life. Comparison with other research suggests that life-sustaining treatment use in community hospitals differs from academic medical centers. Lynn et al2 found more aggressive treatment of elderly or seriously ill decedents in SUPPORT hospitals. Three quarters died in the hospital, and during their last 3 days of life, 11% experienced CPR, and 25% received ventilatory support.2 Other studies in academic hospitals have shown that CPR is attempted for 20% to 30% of all patients who die. The CPR survival rates are actually higher in community hospitals than in referral centers, which may be explained by more selective use.20- 23 A national study of Medicare patients found that those who were admitted to smaller, nonteaching hospitals were less likely to receive high-cost treatments before death.24 The decision to enter a hospital—particularly an academic one—may be an important determinant of the type of terminal care a dying patient experiences.
The patient-physician relationship is also what distinguishes community practice from referral centers. All the treating physicians in this study were experienced practitioners, not physicians in training. Most had long-term primary care relationships with the decedents. Longer relationships facilitate advance planning, as evidenced by the high rate of early treatment discussions. More than 80% of these physicians believed it was easier to withhold life-sustaining treatments in the context of a long-term relationship. Furthermore, in their responses to open-ended questions, the physicians said that good relationships between physician, patient, and family led to good decision making. The duration and quality of their relationships with the decedents might have allowed for the communication and trust necessary to forgo treatments near the end of life.
Several limitations of the present study should be considered. First, because we used physician report of treatment use, results could be affected by physicians' desire to report fewer life-sustaining treatments. To guard against this social desirability bias, we asked physicians to review medical records on the decedents and have them available during the interview—and most did so. Second, our data show that community physicians frequently discuss treatment options near the end of patients' lives, but we cannot judge the quality of communication or ethical rationales for decisions. In a separate analysis, family members indicated whether they were aware of these discussions and whether they were satisfied with decisions made. Family members were generally satisfied with treatment decisions, but they wanted improved communication and more time and attention from treating physicians.19 A third limitation is our exclusion of patients who were referred away from home and died in more distant hospitals. The decision for a seriously ill patient to enter a tertiary hospital is related to the patient's age, functional status, and diagnosis. We cannot comment on the referral practices of these community physicians, and our findings may be due in part to patient characteristics associated with nonreferral. However, since most seriously ill older patients are not referred and die in community settings, these study results generalize to most similar patients.16 Finally, physicians in this study practiced in only one state. However, data on regional variation in terminal care show that patients in North Carolina have terminal care costs near the national average.25 To the degree that costs reflect intensity of treatment, terminal care practice patterns in North Carolina are representative.
Most older patients will experience terminal care in community settings and be treated by primary care physicians with whom they have established relationships. Less intensive treatment is not necessarily better treatment, but it may better meet the needs of patients who die of chronic and incurable diseases. Since the treatment practices of community physicians will affect most dying patients, it is important to ask them and their patients about end-of-life care. Community physicians and their patients made many decisions to limit treatment near the end of life, and the surviving family members had few regrets about these decisions.19 Additional interventions to enhance and preserve trust, relationship, and communication in varied practice settings may help all physicians improve the process of decision making. Decisions about where one dies and who the treating physician will be may determine many aspects of the experience of terminal care. These decisions should be considered as part of advance care planning.
Accepted for publication October 1, 1998.
This study was funded by a Robert Wood Johnson Generalist Faculty Scholars Award, Chapel Hill, NC (Dr Hanson).
Presented at the national meeting of the Society of General Internal Medicine, Chicago, Ill, April 25, 1998.
Reprints: Laura C. Hanson, MD, MPH, CB 7110, Division of General Medicine, University of North Carolina, Chapel Hill, NC 27599-7110 (e-mail: firstname.lastname@example.org).