Why Don't Patients and Physicians Talk About End-of-Life Care?Barriers to Communication for Patients With Acquired Immunodeficiency Syndrome and Their Primary Care Clinicians

Background  Patients with chronic and terminal disease frequently do not talk to their physicians about end-of-life care. Interventions to improve this communication have generally been unsuccessful, suggesting that important barriers to this communication must exist.

Objectives  To determine the barriers to and facilitators of patient-clinician communication about end-of-life care and to identify barriers and facilitators that are more common among those patients who are least likely to discuss end-of-life care: minorities and injection drug users.

Methods  We conducted a prospective study of 57 patients with advanced acquired immunodeficiency syndrome and their primary care clinicians who were recruited from university and private clinics. Barriers to and facilitators of end-of-life communication were identified from a prior qualitative study and assessed for frequency and importance and for an association with the occurrence and quality of end-of-life communication.

Results  Clinicians identified more barriers than patients. Barriers identified by patients and clinicians fell into 3 categories of potential interventions: education about end-of-life care, counseling to help address end-of-life concerns, and health care system changes to facilitate patient-clinician communication. Although none of the patient-identified barriers was associated with the occurrence of communication, 2 clinician-identified barriers were associated with less communication: "the patient has not been very sick yet" and "the patient isn't ready to talk about end-of-life care." Nonwhite patients were more likely to identify the following 2 barriers than white patients: "I feel that if I talk about death, it could bring death closer" and "I don't like to talk about the care I want if I get very sick."

Conclusions  The diversity of barriers and facilitators relevant to patients with acquired immunodeficiency syndrome and their clinicians suggests that interventions to improve communication about end-of-life care must be focused on individual needs and must involve counseling interventions and health system changes in addition to education. Clinician barriers are more common and more strongly associated with the occurrence of end-of-life communication than patient barriers, suggesting that clinicians are an important target group for improving this communication.

PATIENTS WITH chronic and terminal diseases, such as acquired immunodeficiency syndrome (AIDS), cancer, and end-stage lung disease, frequently do not talk with their physicians about the kind of care they want at the end of their lives.^1-3 Randomized controlled trials of interventions to improve patient-physician communication about end-of-life care have included use of advance directives,⁴ educational interventions with patients and/or physicians,^5,6 and the provision of prognostic information to patients and physicians,⁵ but have not been successful in improving end-of-life care. Among severely ill, hospitalized patients who had not discussed end-of-life care with their physicians, 58% reported they do not want to have these discussions.⁷ What is not clear, however, is why these patients say they do not want to talk about end-of-life care. If barriers to and facilitators of communication about end-of-life care can be identified, interventions could be developed to improve communication about this issue and perhaps thereby improve end-of-life care.

Recent advances in the treatment of human immunodeficiency virus (HIV) infection have been dramatic and unprecedented, with marked improvements in survival and quality of life.^8,9 The new antiretroviral regimens, however, are not successful for all patients; many patients find themselves facing death from AIDS at the same time that many others are having favorable responses to antiretroviral therapy. Patients and physicians have described death from AIDS in the new protease inhibitor era as having transformed from fate to tragedy.¹⁰ In this way, caring for the patient who is dying of AIDS has become comparable to caring for patients who are dying of other treatable chronic diseases.

Like patients with other chronic diseases, most patients with AIDS have not discussed with their physicians the kind of care they want at the end of life. In a study of 1031 persons with AIDS, 68% knew of advance directives, but only 11% had talked with their provider about advance directives.¹¹ Furthermore, there are important differences in the proportion of patients who discuss end-of-life care with their physicians, according to HIV risk factor, race, and ethnicity. Gay men are more likely to have been counseled about and executed a prior directive than injection drug users or women,¹¹ and whites are more likely to have discussed their preferences for resuscitation with their physician than blacks and Hispanics.¹ Therefore, patients with AIDS, particularly women, minorities, and injection drug users, represent important groups of patients for whom the important barriers to and facilitators of communication about end-of-life care should be identified.

We previously identified barriers to and facilitators of patient-physician communication about end-of-life care in a qualitative study using focus groups of patients with advanced AIDS and physicians with experience treating AIDS.¹² The goals of the current quantitative study are to identify the frequency at which these barriers and facilitators occur, to evaluate if these barriers and facilitators were associated with the occurrence and quality of this communication, and to identify barriers that are more common among those patients with AIDS who are least likely to discuss end-of-life care with their clinicians (ie, black and Hispanic patients, injection drug users, and women). The purpose of identifying these barriers and facilitators is to give clinicians insight to overcome these barriers and to provide researchers and policy makers direction for developing interventions that would be helpful in improving communication.

We recruited 57 patients with AIDS and their primary care clinicians into a prospective cohort study. We identified the primary care clinician as the clinician who, as reported by the patient, was primarily responsible for taking care of his or her AIDS-related problems. Enrollment criteria for patients included a prior AIDS-defining illness¹³ and a CD4 cell count of less than 100/µL at some time in the past. Patients were excluded if they could not speak English, if they had a diagnosis of dementia, or if they had been seeing their primary care clinician for less than 3 months. Patients would have been excluded if their primary care clinician was unwilling to participate, but no clinicians refused. None of the clinicians who participated were in residency training. Clinicians could have up to 3 patients enrolled in the study. All interviews were conducted with patients and clinicians between August 1996 and January 1997.

Patients were recruited in 3 ways. Advertisements were placed with community-based organizations, university and private clinics, and an AIDS Clinical Trials Unit. Academic and community clinicians were sent letters describing the study along with advertisements to be posted in their offices. Finally, a university-affiliated clinic in a large municipal hospital maintains a registry of clinic patients who are interested in hearing about studies for which they are eligible. Letters describing this study were sent to persons in this registry along with a letter of introduction from the primary care clinician. The University of Washington Human Subjects Committee approved all procedures, and all subjects signed approved informed consent forms.

One of 3 experienced research interviewers conducted each structured interview with patients at study enrollment. The interviewers completed interviewer training sessions together. All patient interviews took place in quiet settings with only the interviewer and patient present; patients read the questionnaire along with the interviewer, and the interviewer recorded responses. Patient interviews lasted from 60 to 90 minutes. Physician interviews were conducted by telephone using a structured script that took from 10 to 15 minutes to read.

Using definitions of the Centers for Disease Control and Prevention, we previously used focus groups of persons with stage C3 AIDS¹³ to develop a list of the barriers to and facilitators of patient-clinician communication about end-of-life care.¹² Patients in the focus groups were excluded from this study. These barriers and facilitators were transformed into questions by asking if the barrier and facilitator applied to each patient or their relationship with their primary care clinician; then, if the item applied, we asked if the item made it harder (for the barriers) or easier (for the facilitators) to talk with their primary care clinicians about the kind of care they would want if they became too sick to speak for themselves. Items were considered barriers or facilitators only if patients reported that the statement made it harder or easier, respectively, to discuss end-of-life care with their primary care clinicians.

To determine whether there had been patient-clinician communication about end-of-life care, we asked patients the following question: "Have you discussed with your doctor, in a face-to-face discussion, the kinds of treatments you would want if you became too sick to speak for yourself?"

To measure the quality of patient-clinician communication about end-of-life care, we used the focus groups to develop 4 generic items assessing the quality of communication about end-of-life care.¹² These items have been validated as a scale to measure the quality of communication about end-of-life care in a previous report.¹⁴

We assessed whether the occurrence of patient-clinician communication about end-of-life care was associated with the presence of each barrier and facilitator using the χ² test. There were several barriers and facilitators for which an association with the occurrence of end-of-life communication was tautological. For example, if a patient reported that a clinician was good at end-of-life communication, the patient and clinician must have had this communication to make this assessment. These barriers and facilitators were excluded from this analysis.

We divided the quality of communication into high and low as described previously¹⁴ and used the χ² statistic to test for an association between the quality of communication and the presence or absence of a barrier or facilitator. Similarly, we used the χ² statistic to determine which barriers and facilitators were associated with race/ethnicity or with an HIV risk factor. P<.05 was used to determine statistical significance.

Table 1 shows the demographic and clinical characteristics of the patients and clinicians. The demographic characteristics of enrollees are similar to those of persons who are HIV positive in the Pacific Northwest. Clinicians identified a mean of 2.6 barriers (median, 2) to end-of-life communication, with a range of 0 to 10 barriers identified per clinician. Only 26% of clinicians (10/38) did not identify any barriers that interfered with talking with their patients about end-of-life care. By contrast, patients identified a mean of 1.3 barriers (median, 0), with a range of 0 to 10 barriers per patient. Over half of the patients (56% [32/57]) did not identify a single barrier to talking with their primary care clinician about end-of-life care. Clinicians identified a mean of 5.0 facilitators (median, 5), with a range of 1 to 9. Patients identified a mean of 5.1 facilitators (median, 6), with a range of 0 to 8; only 2% of patients (1/57) did not identify any facilitators that helped them to talk with their primary care clinicians about end-of-life care.

Table 2 shows the patient-identified barriers to communication about end-of-life care in 3 categories according to the intervention we considered most likely to be successful in overcoming the barrier or making use of the facilitator. The 3 interventions are education about advance care planning and end-of-life care, counseling to help address end-of-life concerns, and a health care system change to facilitate patient-clinician communication about end-of-life care. Table 2 also shows the number of times a patient identified each barrier or facilitator as the single most important barrier or facilitator. Although patients identified barriers in each of these categories, the most frequent and most important barriers for patients were in the categories of educational or counseling interventions. Fourteen barriers were identified by at least 1 of the patients, but no more than 19% of patients (11/57) endorsed any one barrier. In contrast, Table 3 shows the clinician-identified barriers according to the intervention necessary. Clinicians identified barriers in all 3 categories, with the single most frequent and most important barrier identified being a system issue of having too little time to discuss end-of-life care. This barrier was identified by almost half of the clinicians, and a quarter reported that it was the single most important barrier.

Patients identified 10 facilitators of patient-clinician communication about end-of-life care. As shown in Table 4, the majority of the facilitators identified could be used in an education and counseling intervention. System change was not a common or important feature of these facilitators. Clinicians also identified 10 facilitators of end-of-life communication, 5 of which could be used in an educational intervention. However, as shown in Table 5, there were also 4 clinician-identified facilitators that would require system changes to use the facilitator in an intervention to improve end-of-life communication.

Despite the fact that there were 8 different barriers each identified by at least 5 patients as interfering with communication with their primary care clinician, none of these barriers was associated with the occurrence of this communication. There was, however, a patient-identified facilitator of end-of-life communication that was significantly associated with the occurrence of communication. Patients who reported that they had family or friends die were more likely to discuss end-of-life care with their primary care clinician (P=.02).

There were 2 clinician-identified barriers associated with the occurrence of patient-clinician communication about end-of-life care. Clinicians who reported as a barrier that the patient had not been very sick were significantly less likely to have had communication about end-of-life care with the patient (P=.003). Those patients for whom the clinician identified as a barrier that the patient had not been very sick yet did not have a significantly lower current CD4 cell count (median, 154/µL vs 216/µL; P=.30), lower nadir CD4 cell count (median, 33/µL vs 44/µL; P=.70), or more AIDS-defining illnesses (median, 1.7 vs 1.3; P=.20) compared with patients for whom clinicians did not identify this barrier. Also, if clinicians reported as a barrier that the patient was not ready to talk about the care they want if they get sick, they were less likely to have had an end-of-life discussion with the patient (P=.03). Conversely, if the clinicians reported as a facilitator that the patient had been very sick in the past, they were more likely to have had an end-of-life discussion with the patient (P=.02).

We also examined the barriers and facilitators associated with the quality of communication about end-of-life care for those 31 patients who reported having had communication about end-of-life care. Patients were more likely to report a lower quality of communication if they reported as a barrier that they do not like talking about getting very sick (P=.04). Similarly, patients rated the quality of communication lower if they reported as a barrier that they would rather concentrate on staying alive than talk about death (P=.04). There were no patient-identified facilitators to communication about end-of-life care that were associated with the quality of communication.

There were 2 barriers to communication about end-of-life care that were more common among nonwhite than white individuals. The nonwhite patients were more likely to report as a barrier that they do not like to talk about the care they want if they get very sick (P=.03). Nonwhite patients were also more likely to report as a barrier that they felt that if they talk about death, it could bring death closer (P=.02). White patients were significantly more likely to identify 3 facilitators of end-of-life care: my doctor is very good at talking about end-of-life-care (P=.03), my doctor is an expert on HIV and AIDS (P=.05), and I have been very sick in the past (P=.03). There was no significant difference in experience treating AIDS between the clinicians for white patients vs those for nonwhite patients (P=.40).

Injection drug users and women with AIDS, when compared with patients who identified themselves as gay or bisexual men, were more likely to report 2 specific barriers: "I don't know what kind of care I want if I get very sick" (P=.03) and "I feel that if I talk about death, it could bring death closer" (P=.02). Gay or bisexual men were more likely to report as a facilitator the fact that their doctor was very good at talking about end-of-life care (P=.03). Despite the similarity between race/ethnicity and HIV risk factor in the barriers and facilitators identified, there was no association in our sample between race/ethnicity and HIV risk factor.

We identified a large number of barriers to and facilitators of patient-clinician communication about end-of-life care for patients with advanced AIDS and their primary care clinicians. While many of these barriers were interrelated, our findings suggest that addressing both the patient and the clinician barriers would require a multifaceted intervention, including educational programs, one-on-one or small-group counseling approaches (for patients and clinicians), and changes in the health care system that facilitate this communication. Each individual barrier was endorsed by no more than 19% of patients (11/20). With the exception of the system barrier of too little time available for these discussions, no more than 34% of clinicians (13/38) endorsed each individual barrier. Therefore, the diverse list of barriers and the multifaceted approaches that would be required to overcome them suggest that prior efforts to improve patient-physician communication about end-of-life care may have failed because they did not address enough of these barriers with individually targeted interventions.^4-6 We have also identified patient and clinician facilitators of communication about end-of-life care that may be incorporated into interventions designed to improve the quality of this communication.

Clinicians identified far more barriers to patient-clinician communication about end-of-life care than did patients. In addition, the only 2 barriers that were associated with the occurrence of end-of-life communication were clinician barriers. These findings imply that clinician barriers may be more important to the occurrence of end-of-life communication than patient barriers. One clinician barrier that was associated with the occurrence of end-of-life communication was the physicians' report that patients had not been very sick yet. This remains a barrier to communication about end-of-life care even among a group of patients with an AIDS-defining clinical illness and a history of a CD4 cell count below 100/µL. The other clinician barrier associated with the occurrence of end-of-life communication was the clinicians' impression that the patient was not ready to talk about end-of-life care. Patients, however, were asked if they felt ready to discuss end-of-life care, and this barrier was not associated with the occurrence of communication. While it is possible that clinicians have a better sense of when patients are ready, it seems more likely that clinicians make incorrect assumptions about whether their patients are ready to discuss end-of-life care. Clinician educational efforts should address these specific perceived barriers.

There was one patient and one clinician facilitator that was associated with the occurrence of communication about end-of-life care. Patients who reported having had friends or family members die were more likely to have had communication about end-of-life care. This information may be helpful in counseling interventions if it is used to teach physicians to help patients to think about their own treatment preferences by referring to their past experiences. For example, physicians could ask patients what was bad and what was good about the death of a friend or family member to help elicit values and preferences about end-of-life care. The clinician-identified facilitator associated with the occurrence of communication was the clinician's report that the patient had been very sick in the past. Similar to the barriers, this finding suggests that clinicians wait for patients to be seriously ill before they feel comfortable bringing up end-of-life care. An intervention teaching clinicians to view advance-care planning as a process that should begin early rather than being an outcome, such as the decision to write a do-not-resuscitate order, may help clinicians and patients to initiate discussion about end-of-life care earlier in the disease process.

The assessment of the occurrence of patient-physician communication about end-of-life care without an assessment of the quality of this communication is unlikely to improve the quality of care at the end of life.¹⁵ We have previously described a survey instrument to measure the quality of patient-clinician communication about end-of-life care.¹⁴ In the current report, we identified the barriers to and facilitators of end-of-life communication that were associated with the quality of this communication. Patients who report that they do not like to talk about getting very sick or that they would rather concentrate on staying alive than talk about death were more likely to report a lower quality of communication. Given the observational study design, we cannot determine whether patients who do not like to talk about end-of-life care are more likely to have poor communication or whether poor communication makes patients feel that they do not like to talk about end-of-life care. Regardless of the explanation, clinicians should be aware that some patients express these feelings, and clinicians should be sensitive to these concerns when initiating communication about end-of-life care.

Black and Hispanic persons, women, and injection drug users represent groups that are disproportionately affected by HIV infection¹⁶ and are at increased risk for receiving lower-quality medical care.^17,18 These same patients are less likely to communicate with their physicians about end-of-life care than white homosexual or bisexual men with AIDS.^1,19 Consequently, these groups represent important populations toward which efforts to improve the quality of care should be directed. If efforts to improve the patient-physician communication about end-of-life care are to target these groups, it will be important to identify the barriers to communication that are unique to these patient groups and individuals. Our findings show that the nonwhite patients in this study are more likely to report that they do not like to talk about end-of-life care and are more likely to worry that talking about death could bring death closer. Similarly, injection drug users and women with high-risk sexual partners in this study were also more likely to worry that if they talked about death it could bring death closer. Previous authors have identified this barrier to discussing end-of-life care in case reports²⁰ and in Navajo persons.²¹ An effective approach to overcome this barrier needs to be further investigated and may vary for different groups.

There were also several facilitators of communication about end-of-life care that were more frequently reported among the white patients and homosexual or bisexual male patients. These include reports that the clinicians were very good at talking about end-of-life care and that the clinicians were HIV or AIDS experts. While there was no evidence that the clinicians caring for the homosexual or bisexual male patients or the white patients in this study had more experience treating AIDS, it is possible that these patients' assessment of their clinicians' skills are accurate. It is also possible that this represents higher satisfaction with care among white and homosexual or bisexual male patients. Future studies are needed to determine which of these possibilities account for the difference and the best intervention to minimize this difference.

There are several important limitations to this study. The sample represents a relatively small number of patients and clinicians in a single geographic area, and it is possible that a larger sample size would have identified small but significant associations with other barriers or facilitators. Since we interviewed only patients with advanced AIDS, further research will be necessary to determine if the findings in this study are generalizable to individuals with other chronic diseases. In addition, given the number of statistical comparisons, it is important to acknowledge that one or more of the associations seen may have occurred by chance. However, all the statistical associations occurred in the direction expected: each barrier associated with communication was associated with a decrease in the occurrence or quality of communication, and each facilitator was associated with an increase in the occurrence or quality of communication. Similarly, all of the barriers associated with race or with HIV risk behavior were more common in minority patients, injection drug users, or women, and all facilitators were more common among white patients and homosexual or bisexual men. This decreases the likelihood that these findings occurred by chance. Finally, each clinician could have more than one patient enrolled in this study, which could threaten the statistical assumption of independence of observations. However, since each clinician was limited to 3 patients, it is unlikely that an individual clinician unduly influenced the results.

Previous research has demonstrated the difficulty involved in improving patient-clinician communication about end-of-life care and the quality of care at the end of life. This study identifies a diverse group of barriers that are relevant to patients with AIDS and their clinicians and suggests that multiple interventions to overcome these barriers should involve not just patient and clinician education, but also counseling and health care system changes. In addition, these findings suggest that while interventions must target both patients and clinicians, clinician barriers are more common and more strongly associated with the occurrence of end-of-life communication. For interventions to be effective among nonwhite patients, injection drug users, and women with AIDS, these interventions must be culturally relevant and target the barriers that are more common in these groups. Whether these barriers can be overcome and, if so, how are important challenges facing medical researchers and educators, but we believe that with appropriate research, education, and system changes, many of these barriers can and will be overcome.

Accepted for publication November 17, 1999.

This study was supported in part by the Open Society Institute Project on Death in America Faculty Scholars Program, New York, NY (Dr Curtis); by the Parker B. Francis Fellowship Program, Boston, Mass (Dr Curtis); and by the University of Washington Royalty Research Fund, Seattle.

The authors thank the persons with AIDS and the clinicians who participated in this study.

The opinions expressed here are those of the authors and not necessarily those of the Open Society Institute.

Corresponding author: J. Randall Curtis, MD, MPH, Division of Pulmonary and Critical Care Medicine, Harborview Medical Center, Box 359762, 325 Ninth Ave, Seattle, WA 98104-2499 (e-mail: jrc@u.washington.edu).