|Home | About | Journals | Submit | Contact Us | Français|
Guidelines recommend advanced care planning for terminally-ill patients with less than one year to live. Few data are available about when physicians and their terminally-ill patients typically discuss end-of-life issues.
National survey of physicians caring for cancer patients about timing of discussions regarding prognosis, DNR status, hospice, and preferred site of death with their terminally-ill patients. We used logistic regression to identify physician and practice characteristics associated with earlier discussions.
Among 4,074 respondents, 65% would discuss prognosis “now” (patient has 4–6 months to live, asymptomatic). Fewer would discuss DNR status (44%), hospice (26%) or preferred site of death (21%) “now”, with most physicians waiting for patient symptoms or until there are no more treatments to offer. In multivariable analyses, younger physicians more often discussed prognosis, DNR status, hospice, and site of death “now” (all P<.05). Surgeons and oncologists were more likely than noncancer specialists to discuss prognosis now (P=.008), but noncancer specialists were more likely than cancer specialists to discuss DNR status, hospice, and preferred site of death “now” (all P<.001).
Most physicians report they would not discuss end-of-life options with terminally-ill patients who are feeling well, instead waiting for symptoms or until there are no more treatments to offer. More research is needed to understand physicians’ reasons for timing of discussions and how their propensity to aggressively treat metastatic disease influences timing, as well as how the timing of discussions influences patient and family experiences at the end of life.
Prior research has demonstrated large variations in the costs and aggressiveness of care delivered to patients with cancer and other chronic illnesses at the end of life.1–3 While such variation is not explained by variations in patients’ preferences for end-of-life care,4 it could be explained in part by physicians’ propensity to discuss end-of-life options with their terminally-ill patients. Patients whose physicians discuss end-of-life options sooner might be more knowledgeable about strategies to improve the quality of the end of-life experience and more interested in pursuing them based on this knowledge. Moreover, for patients who may prefer to avoid aggressive care at the end of life, physicians who discuss end-of-life options sooner might better understand these preferences and may be more likely to act on them.
National guidelines recommend advance care planning for patients with terminal illness and life expectancy of 1 year or less.5, 6 Important components of such advance care planning include clear and consistent discussions about prognosis, information about advanced directives and exploration of the “do not resuscitate” (DNR) option, information about palliative care options including hospice, and elicitation of patients’ preferences for site of death.5 California recently enacted legislation requiring doctors to counsel patients whose prognosis is less than a year about hospice and palliative care.7
Few data are available about when physicians and their terminally-ill patients typically discuss end-of-life issues. We surveyed physicians caring for cancer patients to describe their reports of the timing of discussions about prognosis, DNR status, hospice, and preferred site of death with their terminally-ill patients. We then assessed physician demographic, professional, and practice characteristics associated with their reports of the timing of these discussions. Finally, we assessed whether physicians who were more knowledgeable or comfortable addressing end-of-life issues were more likely than other physicians to discuss these topics while patients were still feeling well.
This study is part of a large multi-regional study examining processes and outcomes of care for a population-based cohort of cancer patients conducted by the Cancer Care Outcomes Research and Surveillance (CanCORS) Consortium.8 The cohort includes more than 10,000 patients diagnosed with lung or colorectal cancer during 2003–2005 living in Northern California, Los Angeles County, North Carolina, Iowa, or Alabama, or who received their care in one of 5 large health maintenance organizations (HMOs) or 10 Veterans Affairs Health Care System sites. Data were collected from patient (or surrogate) interviews, medical records, surveys of patients’ physicians, and surveys of informal caregivers. The study was approved by the human subjects committees at all participating institutions. This analysis uses data from the physician survey only.
As described previously,9 we surveyed physicians identified by patients participating in the CanCORS baseline interview as filling one or more key roles in their care, including providing surgery, chemotherapy, or radiation therapy (or discussing these treatments for patients who were not treated); referring patients to these key providers; helping to make treatment decisions; or knowing if the patient had symptoms related to their cancer or its treatment. Of 6,871 physicians whose contact information was verified, 4188 (61.0%) responded.
To assess timing of discussions about end-of-life issues, physicians were asked: “Assume you are caring for a patient who is newly diagnosed with metastatic cancer, but is currently feeling well. You estimate that the patient has 4–6 months to live. When, in the course of the typical patient’s illness, are you most likely, for the first time, to discuss the following with this patient or family? (1) prognosis, (2) DNR status, (3) hospice, (4) preferred site of death.” Response options were “now”, “when the patient first has symptoms”, “when there are no more non-palliative treatments”, “only if the patient is hospitalized”, and “only if the patient or family bring it up”. At the time of the survey administration, national guidelines5, 6 recommended discussion of these topics with terminally-ill patients having a life expectancy of 1 year or less.
In addition, we asked physicians how strongly they agreed or disagreed with the following statement:10, 11 “I feel knowledgeable enough to discuss options for end-of-life care with my cancer patients and their families.” Response options were strongly agree, somewhat agree, neither agree nor disagree, somewhat disagree, strongly disagree. Physicians also provided information about the number of terminally-ill patients they cared for in the past year and personal and practice characteristics.
We excluded 114 physicians who did not respond to all 4 questions regarding discussions about end-of-life issues. We used multiple imputation to impute missing data for remaining items.12, 13 Item nonresponse was <2–3% for most variables, but was 7% for the proportion of patients enrolled in managed care.
We described physicians’ reports of discussing prognosis, DNR status, hospice enrollment, and preferred site of death. We used Chi-square tests to compare reports of discussing each option “now” by physician characteristics. We used logistic regression to identify physician factors associated with having discussions “now”. Independent variables that we believed may be related to timing of discussions included self-reported physician specialty (surgeon, medical oncologist, radiation oncologist, non-cancer specialist [predominantly primary care physicians, gastroenterologists, and pulmonologists, <1% were geriatricians and <0.1% were palliative care specialists]), age (<40, 40–49, 50–59, ≥60 years), sex, race/ethnicity (white, Asian, black, Hispanic, other), board certification, U.S./Canadian medical graduate, practice site (office, hospital), teaching involvement (none, 1–5 days/month, >5 days/month), proportion of patients in managed care (quartiles), number of terminally-ill patients cared for in the past year (quartiles), and study site. We calculated adjusted rates of discussing options “now” for each physician characteristic using a standardized regression approach.14
In a second set of models, we included variables reflecting whether physicians felt knowledgeable enough to discuss end-of-life options with their patients. Responses were categorized as strongly agree, somewhat agree, or other responses.
We conducted two sensitivity analyses. First, we changed the dependent variables in the models to “when the patient first develops symptoms” or “now” versus other options. Second, we repeated all analyses after restricting the cohort to physicians caring for at least 2 terminally-ill patients in the past year (95% of physicians), since some physicians (particularly non-cancer specialists) may have cared for cancer patients who did not die.
The mean age of the 4,074 physicians was 50.1 (SD 9.8), 82% were male, 22% were surgeons, 13% medical oncologists, 6% radiation oncologists, and 59% non-cancer specialists. Characteristics of the cohort are included in Table 1.
The Figure depicts physicians’ reports of when they would discuss prognosis, DNR status, hospice, and preferred site of death with their terminally ill patients, by specialty. Overall, 65% of physicians would discuss prognosis “now”, while 15% would have this discussion only if the patient/family brings it up. Forty-four percent would discuss DNR status “now” with 18% waiting for the patient to have symptoms and another 26% waiting until there were no more nonpalliative treatments. Approximately one-quarter (26%) of physicians would discuss hospice “now” with 16% waiting for symptoms and 49% waiting until there are no more nonpalliative treatments. Finally, 21% would discuss preferred site of death now, with 24% having this discussion only if the patient/family brought it up. Surgeons and medical oncologists were most likely to discuss prognosis “now”, whereas noncancer specialists were most likely to discuss the other topics now.
In multivariable analyses, surgeons and medical oncologists were more likely than noncancer specialists to report discussing prognosis “now” (P=0.008), but noncancer specialists were more likely than cancer specialists to discuss DNR status, hospice, and preferred site of death “now” (all P<0.001) (Table 2). Younger physicians were more likely to report discussing prognosis, DNR status, hospice, and preferred site of death “now” (all P<0.05). Female physicians were less likely than male physicians to report discussing prognosis “now” (P=0.05) and physicians with more terminally-ill patients were more likely than others to report discussing prognosis (P=0.003) and DNR (P=.04) “now”. Asian physicians were less likely than other physicians to report discussing prognosis “now”, and Hispanic physicians were most likely to discuss DNR status “now”. Physicians in office-based (vs. hospital) practices were more likely to report discussing hospice “now” (P=.007). Physicians with the most patients in managed care were most likely to discuss DNR status and hospice “now” (both P=.02). Discussion of all treatment options were highest in the Veterans Administration and the HMO sites and were generally lowest in Los Angeles County and Alabama.
In a second set of multivariable models that included physicians’ reports of whether they felt knowledgeable enough to discuss end-of-life options with their terminally ill patients, physicians who strongly agreed (55% of physicians, OR=1.89, 95% CI 1.55 to 2.30) or somewhat agreed (30% of physicians, OR 1.44, 95% CI 1.18 to 1.76) were more likely than other physicians to discuss prognosis “now”. Physicians who strongly agreed they were knowledgeable to discuss end-of-life options were also more likely to discuss DNR status (OR 1.94, 95% CI 1.58 to 2.37), hospice (OR 1.59, 95% CI 1.27 to 2.00), and preferred site of death (OR 1.60, 95% CI 1.27 to 2.05) “now”. In all cases, addition of these variables to the base models did not influence the associations in Table 2.
When we restricted the sample to physicians who reported that they would discuss prognosis “now” to assess how often these physicians would use the discussion or prognosis as an opportunity to discuss other topics “now”, 57% also reported that they would discuss DNR status “now”, 36% that they would discuss hospice “now”, and 30% that they would discuss preferred site of death “now”.
In sensitivity analyses, we repeated analyses based on whether physicians discussed end-of-life issues “now” or “when the patient develops symptoms” vs. other responses, and associations were generally similar (data not shown). Likewise, when we restricted the cohort to physicians caring for at least 2 terminally ill patients in the past year our results were similar.
In this large survey of physicians caring for cancer patients who practiced in various locations across the United States, we found that many physicians report not discussing end-of-life options with terminally-ill patients while they are still feeling well, instead waiting for onset of symptoms or until there are no more non-palliative treatments to offer. A significant minority of physicians reported they would discuss prognosis or preferred site of death only if the patient or family brought it up. Younger physicians and physicians who were not cancer specialists reported discussing end-of-life options sooner than older physicians and cancer specialists, as did physicians caring for patients in HMO or VA settings, compared with physicians whose patients were identified from population-based settings.
Although guidelines recommend advance care planning for terminally-ill patients with a life expectancy of 1 year or less, a minority of physicians in our study reported that they would discuss DNR status, hospice enrollment, or preferred site of death with patients who had 4–6 months to live and were feeling well. There are several potential explanations for this finding. Physicians may disagree with the guidelines (which are based primarily on expert consensus due to limited data), may be unaware of the guidelines, or may agree with the guidelines but still delay discussions that may be difficult.
Few data are available about the benefits of advance care planning on patient- and family-centered measures.15 Although seriously ill patients, families, and physicians may agree about the importance of preparation for the end of life16 and value good communication about end-of-life care,17 little is known about patients’ preferences regarding timing of discussions about end-of-life topics. Some reports suggest that patients prefer early discussions. For example, outpatients have suggested that advance care planning should occur when they are well, although ideally after the physician knows the patient.18
Nevertheless, not all patients want to discuss end-of-life issues early. An Australian study of metastatic cancer patients found that only 59% wanted to be told how long they had to live when first diagnosed.19 Indeed, some older individuals of certain ethnic groups do not believe patients should be informed that they have incurable cancer.20 Uncertainty also exists about preferences for timing of discussions about DNR status. Some evidence suggests that most patients want to discuss DNR status with their physicians.21–23 However, in the SUPPORT study, only 23% of patients had discussed DNR status with their physicians, and of those who had not, only 42% wished to discuss their preferences.24 A small study of hospitalized cancer patients found that only 9% had discussed advance care preferences with their outpatient oncologist and only 23% of the remaining patients reported wanting to do so, although 58% of patients supported policies requiring housestaff to discuss advance care preferences at hospital admission.25
Some physicians who are aware of the guidelines for advance care planning for terminally-ill patients may delay end-of-life discussions because they fear that discussing death early in a patient’s terminal course may be associated with less hope and poorer outcomes.26, 27 However, studies examining outcomes following end-of-life discussions have demonstrated no harm or positive outcomes.28, 29 Other physicians may have difficulty with end-of-life discussions because they perceive discussing end-of-life issues as an admission of failure to cure a patient’s cancer.30, 31 Additional training may benefit physicians who felt less knowledgeable discussing end-of-life options, since they were less likely to report having these discussions early.
Our finding that many physicians reported waiting until there are no more nonpalliative treatments available before discussing end-of-life issues suggests that some physicians favor providing treatment for terminally-ill patients before discussing end-of-life issues. This is consistent with the willingness of many terminally-ill patients to undergo therapy despite a high treatment burden or a high likelihood of an undesirable outcome.32, 33 However, physicians likely vary a great deal in their assessments of the availability of nonpalliative treatments and the number of regimens indicated,34 and waiting until all possible treatments are exhausted may delay discussions until it is too late for patients’ preferences and values to be addressed. In one study, only 41% of seriously-ill patients who expressed a preference for comfort care believed that their medical care was consistent with that preference.35 Given that use of chemotherapy at the end of life is increasing,3, 36 even for cancers generally considered unresponsive to chemotherapy,36 excessive treatment might delay discussions of end-of-life issues unnecessarily. Nevertheless, some physicians may defer hospice discussions because hospice care typically limits access to medical treatments.37
Some physicians may delay discussions of end-of-life issues because they are uncertain about the accuracy of their estimates for life expectancy. Physicians consistently overestimate survival for terminally-ill cancer patients, although their predictions are highly correlated with actual survival and have discriminatory ability even when inaccurate.38 Still, even after formulating survival estimates (accurate or not), physicians often have difficulty communicating them, frequently providing intentionally inaccurate estimates or no estimates at all.39 Yet, discussions of prognosis may be important, since terminally-ill cancer patients’ own estimates of prognosis are typically more optimistic than physicians’ optimistic predictions.40, 41 Patients’ beliefs about prognosis influence their treatment preferences,40 and may influence patients’ likelihood of initiating discussions about goals when physicians do not. Such delays could contribute to unrecognized wishes. For example, patients who express a preference to die at home are more likely to die at home than similar patients with wishes that are not expressed.42, 43
Although nearly two-thirds of physicians reported they would discuss prognosis “now”, among this group, only half reported they would also discuss DNR status “now”, and only one third would discuss hospice or preferred site of death “now”. This finding suggests that discussions of prognosis are not synonymous with advance care planning. Physicians may instead be using discussions of prognosis to initiate discussions about potential treatments.
Our finding that younger physicians were more likely than older physicians to discuss end-of-life options with their patients may reflect greater awareness of guidelines44 or more emphasis on palliative care issues during their medical training. Alternatively, physicians who are older may have more experience with difficult end-of-life conversations and be more skeptical about the potential benefits. Our finding that physicians in HMO or VA settings reported having end-of-life discussions sooner than physicians of patients in population-based settings may reflect a less-intensive approach to end-of-life care in these managed care models of care. Alternately, it may reflect the ability of integrated organizations to influence behavioral norms. Other evidence suggests greater hospice use in managed care systems45 and less intensive end-of-life care for cancer patients in the VA versus fee-for-service Medicare settings.46 The geographic variations we observed for the population-based sites is consistent with other studies demonstrating large variations in care, particularly for situations where few data are available to guide treatment recommendations.1, 47
Our analysis is strengthened by the large, geographically diverse sample of physicians from multiple disciplines who care for cancer patients. Nevertheless, our findings should be interpreted in light of several limitations. First, our vignette had a limited list of response options to learn when physicians would likely discuss end-of-life issues with a terminally ill patient, and responses to this vignette may not reflect reality. Our response options, for example, did not include an option such as “at the next visit, but before symptoms develop”, which may have been a choice of physicians who may want to establish more of a relationship with their patients before discussing sensitive end-of-life issues.48 Second, as CanCORS participants are not a national sample and our survey is subject to nonresponse bias, we cannot be certain that the physicians in our sample were representative of physicians caring for cancer patients nationally or in the regions studied. In addition, physicians’ responses may have been influenced by social desirability bias; but the low prevalence of “now” responses suggests that social desirability was not the main determinant of responses. Finally, we surveyed only physicians and thus do not know the role of non-physician providers in initiating end-of-life discussions with terminally ill patients.
In conclusion, despite guidelines recommending discussion of prognosis, DNR status, hospice, and preferred site of death with terminally-ill patients while they are still feeling well, our findings suggest that different types of physicians have very different views regarding the appropriate timing of these discussions. More research, both quantitative and qualitative, is needed to understand physicians’ reasons for timing of their discussions, the role of physicians’ propensity to aggressively treat metastatic cancer on the timing of discussions, and patients’ preferences for timing and how timing influences patients’ and families’ experiences at the end of life. Education and patient, physician, and institutional interventions may be necessary to increase advance care planning for terminally-ill cancer patients. Programs such as the Education for Physicians on End-of-Life Care-Oncology (EPEC-O)49, 50 provide a promising model for influencing behavior change.
Funding: This work of the Cancer Care Outcomes Research and Surveillance (CanCORS) Consortium was supported by grants from the National Cancer Institute (NCI) to the Statistical Coordinating Center (U01 CA093344) and the NCI-supported Primary Data Collection and Research Centers (Dana Farber Cancer Institute/Cancer Research Network U01 CA093332, Harvard Medical School/Northern California Cancer Center U01 CA093324, RAND/UCLA U01 CA093348, University of Alabama at Birmingham U01 CA093329, University of Iowa U01 CA01013, University of North Carolina U01 CA093326) and by a Department of Veteran’s Affairs grant to the Durham VA Medical Center U01 CDA093344 (MOU) and HARQ 03-438MO-03).
Investigators from the National Cancer Institute and the Department of Veterans Affairs participated in this study’s design and data collection but not the analysis, interpretation of findings, reporting of this work or the decision to submit the article for publication.
Financial Disclosures: None of the authors have any potential conflicts of interest or financial disclosures to report.
The authors would like to thank Robert E. Wolf, M.S., and Yang Xu, M.S. for expert programming assistance and Yulei He, Ph.D. for assistance with the multiple imputation.