Discussions With Physicians About Hospice Among Patients With Metastatic Lung Cancer FREE

Hospice includes a broad array of palliative and supportive services that can improve symptom management and quality of life for patients with a terminal illness.^1- 2 To be eligible for the Medicare hospice benefit, which covers 80% of hospice stays,³ a physician must certify that a patient has a prognosis of 6 months to live or less, and the patient must generally forego life-extending treatments for the illness. Despite the potential benefits of hospice, many terminally ill patients receive hospice services only in the final days before death or never enroll.^4- 5 Factors that may contribute to late or no enrollment include a lack of awareness of hospice, inaccurate understanding of one's prognosis, or a preference for aggressive treatment of the illness until all nonpalliative options are exhausted.^6- 11

Among patients with advanced cancer, discussing preferences about end-of-life care with their physicians is associated with less aggressive medical care near death, which in turn is associated with better quality of life.^12- 13 National guidelines recommend initial discussions about palliative care options including hospice for terminally ill patients with a life expectancy of less than 1 year, and the guidelines recommend the reintroduction of this topic as patients approach death.^14- 15 However, previous studies have documented shortcomings in communication between physicians and patients about end-of-life care.^16- 19 Physicians often delay discussions with patients about prognosis and end-of-life care or do not communicate their true survival estimates.^20- 25

Although prior studies have shown that fewer than half of patients who die of cancer are enrolled in hospice before their death,^26- 27 little is known about how often patients were aware of the option and declined hospice, perhaps to pursue treatments that could be life extending or because they were unaware of their prognosis. To our knowledge, there are no broad population-based data on discussions regarding hospice between patients with advanced cancer and their physicians. In this study, we used rich data on a large multiregional cohort of patients with metastatic lung cancer to identify factors associated with whether patients have discussed hospice with their physicians within approximately 7 months of their diagnosis.

Data were collected as part of a multiregional study of care for patients with lung or colorectal cancer conducted by the Cancer Care Outcomes Research and Surveillance (CanCORS) Consortium.²⁸ The cohort included over 10 000 patients 21 years and older, who received a diagnosis between 2003 and 2005 and were identified within weeks of their diagnosis through rapid case ascertainment. Subjects with lung cancer lived in 1 of 4 areas (Northern California; Los Angeles County, California; Iowa; or Alabama) or received care at 1 of 5 large health maintenance organizations or 1 of 10 sites in the Department of Veterans Affairs health system.

Patients (or surrogates of patients who were deceased or too ill) were interviewed using computer-assisted telephone interviewing approximately 4 to 7 months after diagnosis.²⁹ Survey instruments were translated into Spanish and Chinese and administered by bilingual interviewers for patients who preferred either of these languages. Following the standard definitions from the American Association for Public Opinion Research,³⁰ the response rate was 49.1% and the cooperation rate was 58.9% among patients with lung cancer. Relative to nonrespondents with lung cancer, respondents with lung cancer were younger (median age, 69 years for respondents vs 71 years for nonrespondents), more often female (43% of respondents vs 37% of nonrespondents), more often white (72% of respondents vs 60% of nonrespondents), more often Hispanic (6% of respondents vs 2% of nonrespondents), and less likely to be diagnosed at stage IV (40% of respondents vs 48% for nonrespondents) (all P ≤ .001). Medical records for up to 15 months after diagnosis were abstracted for those who provided consent. This study was approved by the human subjects committee of each participating institution.

We focused on the 1572 patients diagnosed as having stage IV lung cancer who were interviewed and provided consent for medical records abstraction (86% of patients with stage IV lung cancer). Median survival after diagnosis for patients with stage IV lung cancer is approximately 4 to 8 months,^31- 33 so a discussion with a provider about hospice within 4 to 7 months after diagnosis is appropriate for these patients.

Our primary outcome was whether patients had a discussion about hospice with a health care provider before the interview. If patients or their surrogates answered “yes” to the question, “After your cancer was diagnosed, did any doctor or other health care provider discuss hospice care with you?” or the medical record indicated that a hospice discussion had occurred before the interview date, we considered the patient to have had a discussion.

Covariates for the full sample included age (≤54, 55-59, 60-64, 65-69, 70-74, 75-79, or ≥80 years), sex, being married or living with a partner, education (<high school graduate, high school graduate, or some college or more), speaking English at home, income (<$20 000, $20 000-$39 999, $40 000-$59 999, or ≥$60 000), insurance (Medicare, Medicaid, private, or other), receipt of chemotherapy before the interview, comorbidity level measured by the Adult Comorbidity Evaluation 27 index (none, mild, moderate, or severe),³⁴ geographic region, treatment in a Department of Veterans Affairs facility, and health maintenance organization enrollment. We included self-reported race/ethnicity (white, black, Hispanic, Asian, or other) because prior studies have shown lower hospice enrollment among black and Hispanic patients with metastatic cancer.^26,35- 38 Given patients' limited life expectancy and variation in the time to interview completion, we included quartiles for the number of days from diagnosis to interview as covariates. Because expected survival may influence the likelihood of discussing hospice, we also included variables for the number of days from interview until death (died before interview or 0-59, 60-119, 120-179, 180-239, or ≥240 days after). In addition, we included a variable indicating whether the patient was alive at the time of the survey but had a surrogate complete the interview.

Patients who completed the interview themselves were asked additional questions. To assess expected prognosis, patients were asked, “Based on your understanding about what your doctors have told you about your cancer, your health in general, and the treatments you are receiving, how long do you think you have to live?” Patients' open-ended responses were coded as less than 1 year, 1 to 2 years, 2 to 5 years, more than 5 years, “in God's hands,” and “do not know.” We included a variable indicating whether patients prefer “treatment that focuses on relieving pain and discomfort as much as possible, even if it means not living as long,” as opposed to “treatment that extends life as much as possible, even if it means having more pain and discomfort,” and an index of fatalistic thinking (a shortened version of the Powe Fatalism Inventory³⁹ derived from responses to 4 questions on patients' views about health and life in general). We included scores from the Brief Pain Inventory,⁴⁰ the Dyspnea Scale,⁴¹ and a proxy for poor performance status defined as an EuroQoL-5 Dimensions (EQ-5D) health status index score of 0.60 or lower.⁴²

Additional variables of interest were whether the patient reported a discussion of do-not-resuscitate (DNR) preferences, whether the patient used hospice, and whether a provider recommended hospice. Patients were asked, “There are a number of things doctors can do to try to revive someone whose heart has stopped beating, which usually includes a machine to help breathing. Has a doctor ever talked to you about whether you would want to be revived or use life-sustaining machines?” Patients who answered no were asked, “Is this something you would like to discuss with your doctor?” We categorized DNR discussions as discussed, did not discuss but would like to, and did not discuss and would not like to. If patients or their surrogates reported hospice use in either the baseline interview or a follow-up interview at 12 to 15 months after diagnosis or if medical records indicated use in this period, we considered these patients to have used hospice within a year of diagnosis. Patients who answered yes to the discussion question were also asked, “Did any doctor or other health care provider recommend hospice care for you?”

Item nonresponse was less than 3% for most variables. Multiple imputation was used for missing values, and the analyses were performed using the multiple-imputed data sets.⁴³ We excluded 55 patients who did not respond to the hospice discussion question. The final study cohort included 1517 patients.

We estimated logistic regression models for the probability of having discussed hospice for the full sample. We then calculated rates for hospice discussion for subgroups defined by each covariate, adjusted for all other aforementioned covariates by direct standardization using the regression model.⁴⁴ A supplemental analysis of patients who completed the interview themselves (n = 510) also included the variables for expected prognosis, end-of-life preferences, symptom severity, and the proxy for performance status.

We assessed the association of hospice discussion with discussion of DNR preferences using a χ² test. Since some patients may have discussed DNR and hospice in the same conversation, we did not include this variable in the multivariate models. We also assessed the association of hospice discussion with hospice use and examined responses to the question about whether hospice was recommended. All analyses were conducted using SAS statistical software version 9.2 (SAS Institute Inc, Cary, North Carolina).

In the full sample of 1517 patients, the median time from diagnosis to interview was 5.0 months (interquartile range, 3.9-6.7 months). Surrogates completed the interview for 750 deceased patients (49%) and 134 living patients (9%).

Of the 1517 patients, 807 (53%) had discussed hospice with a physician before the interview (Table 1). Of these, 42% had a discussion reported in both the interview and the medical record, 51% in the interview only, and 7% in the medical record only. Among those who had discussed hospice, 72% reported that a physician or other health care provider recommended hospice. Patients who discussed hospice with a provider were more likely to enroll in hospice. Almost three-quarters (70%) of patients who had a discussion before the interview used hospice within a year of diagnosis compared with just more than a quarter (26%) of those who did not (P < .001).

In adjusted analyses, black and Hispanic patients were less likely than white or Asian patients to have discussed hospice (49% and 43% vs 53% and 57%, respectively; P < .001) (Table 1). More than half (54%) of English-speaking patients discussed hospice compared with 41% of non–English-speaking patients (P = .04). Medicaid enrollees were less likely to have discussed hospice than those covered by Medicare or private insurance and those who had other insurance or were uninsured (50% vs 52%, 52%, and 69%; P = .03). Those who were married or living with a partner were less likely to have discussed hospice (51% vs 57%; P = .02). Individuals who had received chemotherapy before the interview were less likely to have discussed hospice (51% vs 57%; P = .004). While 79% of patients who died before the interview had discussed hospice with a provider, only half (53%) of those who died within 2 months after the interview had done so, and discussion rates were much lower for patients who lived longer (eg, 33% of patients who died between 2 and 4 months after the interview and 21% of patients who died more than 8 months after the interview; P < .001). Living patients whose surrogates completed the interview were more likely to have discussed hospice than other patients (62% vs 52%; P = .01). Those interviewed sooner after their diagnosis were less likely to have had documentation of a hospice discussion (P = .04).

Among patients who completed the interview themselves, only 21% had discussed hospice before the interview. In contrast, 36% reported that they had discussed DNR preferences with their physician (Table 2). Among patients who had discussed DNR, only a third (35%) had also discussed hospice. More than a quarter (28%) of patients reported that they had not discussed DNR with their physician but would like to discuss this topic. Only 20% expected to survive less than 2 years; 51% reported a preference for relieving pain over extending life.

In adjusted analyses, those who expected to live less than 2 years were much more likely to have discussed hospice than those who expected to live longer or who responded their prognosis was “in God's hands” or “do not know” (eg, 45% of those who expected to live <1 year had discussed hospice compared with 12% of those who expected to live ≥5 years [Table 2; P < .001]). Those who were married or living with a partner were less likely to have discussed hospice (18% vs 26%; P = .04). Of the patients who died within 2 months of the interview, 40% had discussed hospice, and discussion rates were lower for patients who lived longer (eg, 21% of patients who died within 2-4 months of the interview and 19% of patients who died >8 months after the interview [P = .04]). Patients reporting more severe pain or dyspnea were no more likely to have discussed hospice than patients reporting less severe symptoms (P = .28 and P = .84, respectively). Patients reporting a preference for relieving pain over extending life were no more likely to have discussed hospice than patients reporting the opposite preference (23% vs 19%; P = .31).

Patients diagnosed as having metastatic lung cancer have incurable cancers with very poor prognosis and substantial palliative care needs at the end of life; thus, they are particularly likely to benefit from hospice. Nonetheless, using patient or surrogate reports and medical record documentation, we found that only half of patients diagnosed as having stage IV lung cancer, including patients who died before the interview, had discussed hospice with a physician or other health care provider within approximately 4 to 7 months of diagnosis. These discussions may have occurred very near death for a considerable number of patients who had a discussion. In fact, we found that only half of patients who died within 2 months after the interview had discussed hospice before the interview. Patients experiencing more severe symptoms were no more likely to have had a discussion than those experiencing less severe symptoms. More than three-quarters of patients who reported a preference for relieving pain over extending life had not discussed hospice with a physician. There was also evidence of “unmet demand” for end-of-life discussions—more than a quarter indicated that they had not discussed DNR preferences with their physician but would like to do so.

Among all patients with metastatic lung cancer, black and Hispanic patients were less likely to have discussed hospice than white or Asian patients. Medicaid enrollees and non–English-speaking patients were also less likely. These findings could explain lower rates of hospice enrollment observed for these groups in other studies.^{10,26,35- 38,45- 46}

Our large multiregional study builds on findings of prior smaller studies. We found that while patients who expected to survive less than 2 years were substantially more likely to have discussed hospice than those who expected to live longer, reported survival expectations were considerably more optimistic than suggested by mortality statistics for patients with metastatic lung cancer. One-third of patients expected to live more than 2 years after their interview, whereas only approximately 6% of patients diagnosed as having stage IV lung cancer are still alive 2 years after diagnosis.⁴⁷ These results are consistent with a study by Weeks et al⁸ that found that patients with metastatic lung and colon cancer overestimated their survival probability and their estimates affected their preferences for end-of-life care.⁸ Our finding suggests that physicians are not communicating effectively with patients about prognosis or perhaps that some patients may not be fully comprehending discussions of their prognosis with health care providers, choosing instead to maintain an optimistic attitude about their prognosis.

We found that patients who had a shorter survival time were more likely to have discussed hospice. This finding is similar to a recent study by Wright et al¹² that focused on general discussions about end-of-life care rather than discussions of hospice specifically.¹² However, in contrast to this prior study, we found no significant differences in the likelihood of hospice discussions based on symptom severity or reported preferences for relieving pain over extending life in our multivariate analysis.

We also found that hospice discussions were less common among patients who had received chemotherapy. This may reflect patient preferences, a more aggressive approach to treatment of metastatic cancer by their health care providers, or reluctance to discuss hospice while patients are receiving therapy that could be life extending because of the restrictions on such treatment imposed by the Medicare hospice benefit.

Physicians face a number of barriers to the discussion of end-of-life options and preferences with patients and their families. These conversations are often difficult and time consuming, and time spent discussing these issues may not be adequately compensated. Also, although many patients and family members report a preference for earlier discussions of prognosis and end-of-life care,^22,48- 50 some prefer to delay the timing of these discussions or not discuss these topics at all—a finding confirmed in our results.^51- 54 However, among patients who discussed end-of-life care with a physician, many did not have a comprehensive discussion of end-of-life issues. For example, only a third of patients who reported discussing DNR preferences had also discussed hospice, representing a missed opportunity for health care providers and patients.

Important strengths of our study are the use of detailed information collected from patients, surrogates, and medical records for a multiregional cohort with metastatic lung cancer; the large sample size that allowed us to examine clinically important subgroups; and the sampling of new-onset cases from defined populations. Nevertheless, our findings should be interpreted in the context of several limitations. First, we relied on a combination of patient and/or surrogate reports and medical record documentation of hospice discussions, and we may not have captured all such discussions. Some patients who had a discussion may not have remembered it or may have chosen not to report it; some surrogates may have been unaware of discussions that took place; and health care providers may not have routinely documented all discussions. Alternative phrasings of questions about expected survival, hospice discussions, and discussions of DNR preferences may have produced different responses from patients. Also, patients who were more open to hospice may have been more likely to report a conversation about it. However, results were similar when we used only medical record documentation in identifying discussions. Second, because surrogates were not considered reliable informants about prognosis and end-of-life preferences, questions about these issues were asked only of patients, so analyses of these variables excluded surrogate interviews. Third, the interview asked about whether a discussion occurred but little about its duration or content. We found that almost three-quarters of those who had a discussion reported that the health care provider had recommended hospice. Fourth, our survey is subject to nonresponse bias, so the patients in our sample were not representative of all patients with metastatic lung cancer. However, demographic and clinical characteristics of the CanCORS population were similar to national estimates from the Surveillance, Epidemiology, and End Results (SEER) Program.⁵⁵

Despite evidence that hospice can improve quality-of-life outcomes and satisfaction with end-of-life care,^12- 13 many patients diagnosed as having metastatic lung cancer did not discuss hospice with a health care provider within approximately 7 months of diagnosis. Among those who did have a discussion, a sizeable number may have had the discussion very near death, at which point the patient may be unable to derive the full benefits of hospice. Our results suggest that many patients most in need of hospice and their physicians are not communicating effectively about end-of-life care options like hospice, and physicians may need to focus communication efforts on certain subgroups of patients, including racial/ethnic minorities, non-English speakers, and Medicaid enrollees. Incorporating a systematic assessment of patients' need for palliative care and hospice services into routine practice could help to foster earlier communication between physicians and patients regarding hospice.⁵⁶ Interventions to increase communication regarding patients' prognoses and goals of care may promote greater use of hospice for appropriate patients.

Correspondence: Haiden A. Huskamp, PhD, Department of Health Care Policy, Harvard Medical School, 180 Longwood Ave, Boston, MA 02115 (huskamp@hcp.med.harvard.edu).

Accepted for Publication: March 20, 2009.

Author Contributions: Dr Huskamp had full access to all the data in the study and takes responsibility for the integrity of the data and the accuracy of the data analysis. Study concept and design: Huskamp, Keating, Malin, Weeks, Teno, and Ayanian. Acquisition of data: Huskamp, Keating, Malin, Weeks, Kahn, and Ayanian. Analysis and interpretation of data: Huskamp, Keating, Malin, Zaslavsky, Weeks, Earle, Teno, Virnig, Kahn, He, and Ayanian. Drafting of the manuscript: Huskamp. Critical revision of the manuscript for important intellectual content: Huskamp, Keating, Malin, Zaslavsky, Weeks, Earle, Teno, Virnig, Kahn, He, and Ayanian. Statistical analysis: Huskamp, Zaslavsky, and He. Obtained funding: Kahn and Ayanian. Study supervision: Huskamp and Ayanian.

Financial Disclosure: None reported.

Funding/Support: 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 (grant U01 CA093344) and the NCI-supported Primary Data Collection and Research Centers (Dana Farber Cancer Institute/Cancer Research Network grant U01 CA093332, Harvard Medical School/Northern California Cancer Center grant U01 CA093324, RAND/UCLA grant U01 CA093348, University of Alabama at Birmingham grant U01 CA093329, University of Iowa grant U01 CA01013, and University of North Carolina grant U01 CA093326) and by a Department of Veterans Affairs Health Services Research and Development Service grant (CRS 02-164) to the Durham VA Medical Center.

Role of the Sponsors: The funding agencies had no role in the design and conduct of this study; collection, management, analysis, and interpretation of the data; or preparation, review, or approval of the manuscript.

Additional Contributions: Hocine Azeni, MA, provided expert statistical programming, and Paul Catalano, ScD, provided guidance on dataset construction.